Introduction

This study focuses on whether and to what extent the institutional complexity of a multinational corporation (MNC) shapes its accounting quality. In our setting, institutional complexity pertains to both the number of institutional contexts and the degree of discrepancies, or tensions, among these contexts (e.g., Arregle, Miller, Hitt, & Beamish, 2016; Greenwood, Raynard, Kodeih, Micelotta, & Lounsbury, 2011). MNCs represent a pivotal and still evolving organizational form in today’s business world (Aguilera, Marano, & Haxhi, 2019; Cuervo-Cazurra & Ramamurti, 2014). However, there is limited knowledge about the forces driving their accounting quality, particularly given that, in contrast to firms operating in a single country, MNCs face more severe agency costs and information asymmetry, arising from foreignness, cultural and language differences, geographic distances, and divergent operating and legal institutions (Bushman, Chen, Engel, & Smith, 2004; Kostova & Zaheer, 1999; Kostova, Roth, & Dacin, 2008, 2009; Shroff, Verdi, & Yu, 2014). Consequently, it is an open question as to how tensions arising from their heterogeneous institutional environments influence MNCs’ accounting quality.

Cumming, Filatotchev, Knill, Reeb, and Senbet (2017) develop the theory of international mobility of corporate governance, pertaining to how MNCs’ divergent institutional contexts at the headquarters level and the subsidiary level define their strategic choices. International mobility of corporate governance rests on two fundamental mechanisms, which are corporate governance bonding and corporate governance arbitrage. The corporate governance bonding hypothesis largely relates to the mobility of good governance, such as through cross-listing in a country with a stronger legal regime (Coffee, 1999; Cumming et al., 2017). In contrast, the corporate governance arbitrage hypothesis refers to the mobility of weaker governance, such as placing subsidiaries in weaker legal institutions to circumvent some corporate governance requirements (Aguilera et al., 2019; Allred, Findley, Nielsen, & Sharman, 2017). Hence, within a given firm, tensions may arise as to which hypothesis prevails, with implications for various corporate outcomes, including accounting quality. To investigate this issue, we employ a unique sample of non-US MNCs operating subsidiaries in offshore financial centers (OFCs)1 either with or without a cross-listing in the US. The OFC setting permits us to examine the underexplored corporate governance arbitrage hypothesis. In essence, we seek to address three questions.

First, the corporate governance bonding hypothesis predicts that cross-listing in the US can serve as an effective institution to reduce information asymmetry and agency cost for foreign firms. US investor protection laws and regulations are widely regarded as among the most effective in the world (Ball, Kothari, & Robin, 2000). However, findings from prior studies may not be applicable to cross-listing MNCs (Khanna & Palepu, 1997). In particular, different from other cross-listing firms, such as those operating in a single country, MNCs face more severe agency cost and information asymmetry (Bushman et al., 2004; Reeb, Kwok, & Baek, 1998; Shroff et al., 2014). Therefore, it is debatable how the Securities & Exchange Commission (SEC) can monitor the accounting quality of these MNCs that face tensions between complex, heterogeneous, and dynamic institutions. In this context, the first question we seek to answer is whether cross-listing in the US enhances the accounting quality of MNCs.

Second, it is important to understand how parent companies and foreign subsidiaries mutually shape MNCs’ performance (Aguilera et al., 2019). In this regard, the corporate governance arbitrage hypothesis posits that some MNCs can set up shell companies or operate subsidiaries in countries with less stringent legal institutions to bypass corporate governance requirements. Our sample of MNCs register subsidiaries in OFCs that exhibit attributes such as tax avoidance opportunities, secrecy, and weak legal rules and enforcement, thereby increasing the risk of insider expropriation and undermining the bonding effect (Desai, 2005; Durnev, Li, & Magnan, 2016). Consequently, the second question is whether an MNC’s choice of OFC subsidiaries negatively moderates the association between cross-listing and accounting quality.

Third, prior evidence indicates that the institutional void of emerging markets (Khanna & Palepu, 1997) and the institutional distances (or differences) between an MNC’s home country and subsidiary institutions influence its knowledge sharing (Meyer, Li, & Schotter, 2020), as well as its financial disclosure (Shi, Magnan, & Kim, 2012). Therefore, our third question follows: How does an MNC’s OFC subsidiaries choice moderate the association between its home-country governance and accounting quality?

We contend that an MNC’s accounting quality hinges on the tension among heterogeneous and conflicting external/internal institutions. Cross-listing is a visible strong external legal institution, whereas the use of OFC subsidiaries is a less visible internal institution. Our setting thus allows us to examine the interactions between external and internal governance mechanisms, as suggested by John and Senbet (1998). Some anecdotal evidence provides support to our conjecture. For example, on December 21, 2016, Braskem S.A., a Brazil-registered entity with American Depositary Receipts (ADRs) traded on the New York Stock Exchange (NYSE), agreed to pay US$325 million in disgorgement of profits under the terms of a resolution with the SEC. Over a decade until 2014, in conjunction with its parent company, Odebrecht S.A., a large Brazil-based engineering firm, Braskem engaged in a sophisticated bribery operation of government officials as well as of executives at Petrobras, Brazil’s national oil company, to obtain political, tax, and commercial advantages. OFC affiliates played a key role in the scheme as the funds for such bribery operations were “…funneled…to a series of off-shore entities.”2 This conduct resulted in corrupt payments and/or profits totaling approximately US$465 million for Braskem, leading the firm to restate its previously issued financial statements.3 Other noteworthy disclosure and reporting scandals involving US cross-listed firms with offshore affiliates include Parmalat4 and, more recently, Wirecard.5

Our study exploits a large international dataset provided by OSIRIS, which includes information on subsidiaries of MNCs. Our sample consists of 4,265 unique MNCs from 31 non-US countries operating subsidiaries in OFCs in the period of 2002–2016, yielding 16,993 firm-year observations. With respect to our first research question, we find that MNCs cross-listed in the US are associated with lower abnormal accruals, higher accruals quality, and a greater level of earnings persistence. In an economic sense, cross-listing in the US is associated with a 10.39% (27.54%) reduction in the absolute value of abnormal accruals (positive abnormal accruals), thus supporting the corporate governance bonding hypothesis that MNCs raise the quality of their accounting to attain legitimacy in the US stock markets.

Regarding our second research question, we document that the positive association between cross-listing and accounting quality is reduced if MNCs operate their business via subsidiaries in OFCs with high OFC attributes. Furthermore, among the three unique attributes of OFCs, i.e., regulation arbitrage, secrecy policy, and tax avoidance opportunity, the first two contribute the most to the lower level in accounting quality, thereby providing strong support to the corporate governance arbitrage hypothesis. Finally, for our third research question, our analyses reveal that OFC attributes also negatively moderate the positive association between an MNC’s home-country governance and accounting quality, consistent with the corporate governance arbitrage hypothesis. Our results are robust to a battery of sensitivity checks, including employing the impact of an exogenous shock related to Organisation for Economic Co-operation and Development’s (OECD’s) Common Reporting Standards (CRS) regulation for OFCs, which mandated the exchange of tax and financial information among OFCs for OECD countries.

Our research contributes to three streams of literature. First, our research adds to the literature on the impact of institutional complexity on the corporate governance of MNCs. In their review paper, Aguilera et al. (2019) call for more research to understand the financial perspectives of the corporate governance of MNCs’ parent–subsidiaries relationship. By employing a unique sample of MNCs registering subunits in OFCs, we heed the call of Aguilera et al. (2019: 473) by simultaneously testing the corporate governance bonding and the corporate governance arbitrage hypotheses. Our results underline how parent companies and foreign subsidiaries jointly determine the accounting quality of MNCs. It is therefore imperative to have effective parent–subsidiaries cooperation to enhance accounting quality. This may represent a challenge for corporate directors if they focus solely on parent company accounts.

Second, our work contributes to the literature on firms operating in OFCs. Over past decades, more and more companies have chosen to set up subsidiaries in OFCs. By 2015, OFC-based institutions managed wealth equivalent to 12% of global gross national product (GNP), or around $9 trillion US (Alstadsaeter, Johannesen, & Zucman, 2017). However, much of the OFC world remains little known and underexplored (Durnev, Li, & Magnan, 2016, 2017). Heeding the call by John and Senbet (1998) to expand the scope of corporate governance, our work explores how reliance on OFC subsidiaries, a unique internal governance mechanism, interacts with external governance mechanisms, including cross-listing and home-country governance, to impact accounting quality.

Finally, our findings also extend the institutional complexity theory by suggesting that legitimacy and reputation concerns outweigh agency cost for MNCs cross-listing in the US. From a practical perspective, our results suggest that regulators such as the SEC and the Public Company Accounting Oversight Board, as well as investors, should enhance enforcement and monitoring for firms with opaque or complicated governance structures.

The remainder of our paper is organized as follows. Section 2 presents the theory underpinning our analyses, while Section 3 develops hypotheses. Section 4 outlines data selection and research design, and Section 5 reports our results. In Section 6, we provide additional analyses and sensitivity checks, and, finally, Section 7 presents our conclusions.

Theoretical underpinnings

On the basis of both economic (North, 1991) and organizational institutional theory (Scott, 2001; Williamson, 1998), it is widely argued that regulative, normal, and cognitive institutions affect MNCs’ activities (Arregle et al., 2016; Cantwell, Dunning, & Lundan, 2010; Jackson & Deeg, 2008). Specifically, institutional theory suggests that institutional complexity, defined as the divergent tensions resulting from many and varied institutional logics, has an impact on MNCs (Greenwood et al., 2001; Kostova et al., 2008, 2009; Kostova & Zaheer, 1999; Regner & Edman, 2014). Further, Arregle et al. (2016) theorize that institutional complexity relates to both the number of institutional contexts (countries) and the degree of discrepancies among these contexts. Along this line of reasoning, the comparative corporate governance research stream focuses on how MNCs’ multiple institutional settings at the headquarters and at the subsidiary levels shape their strategic behaviors (Aguilera et al., 2019; Meyer et al., 2020).

For instance, prior research shows that different domains of institutions, country-level institutions, and institutional distances affect MNCs’ organizational legitimacy (Kostova & Zaheer, 1999), systematic risk (Reeb et al., 1998), firm performance (Chacar, Newburry, & Vissa, 2010), and internationalization (Arregle et al., 2016; Regner & Edman, 2014). In this regard, culture and cultural diversity, as potential manifestations of institutional domains, may also affect MNCs’ outcomes, such as capital structure (Zhang, 2022). However, the impact of institutional complexity on MNCs’ accounting has attracted scant attention from scholars. Meek, Roberts, and Gray (1995) explore the impact of institutions on voluntary annual report disclosures by MNCs. More recently, Huang (2018) illustrates that decision structures of US MNCs influence earnings management of subsidiaries. Beuselinck, Cascino, Deloof, and Vanstraelen (2019) take into account how MNCs’ complexity, as represented by their subsidiaries, affects their accounting quality. However, our research is distinct from Beauselinck et al. (2019) in two critical ways. First, Beauselinck et al. (2019) focus on earnings management at the subsidiary level of MNCs. In contrast, our research examines the overall accounting quality of MNCs. Second, we adopt a unique sample of MNCs operating subsidiaries in OFCs and cross-listing in the US, whereas Beauselinck et al. (2019) test earnings management choices for a generic sample of MNCs.

Cumming et al. (2017) advance the theory of international mobility of corporate governance. Based upon agency theory and institutional theory, they maintain that MNCs can import or export corporate governance practices to enhance efficiency and achieve legitimacy from foreign stakeholders. International mobility of corporate governance entails two distinct mechanisms that coexist and generate tensions as to decision-making and ultimate outcomes within MNCs: corporate governance bonding and corporate governance arbitrage. Corporate governance bonding refers to importing good governance practices, such as through cross-listing in countries with stronger legal regimes (Coffee, 1999; Cumming et al., 2017). Corporate governance arbitrage exploits the institutional differences between different countries, and largely pertains to the mobility of governance from stronger institutional environments to weaker ones, possibly through setting up shell companies (Allred et al., 2017; Cumming et al., 2017).

In order to simultaneously test a corporate governance arbitrage hypothesis and a corporate governance bonding hypothesis, we adopt a unique setting of non-US MNCs operating in OFCs and cross-listing in the US.

Hypotheses

Our first hypothesis relates to the effect of US cross-listing on the accounting quality of MNCs. It is grounded on the corporate governance bonding hypothesis (Coffee, 1999; Cumming et al., 2017; Stulz, 1999), which recognizes the legal consequences of a US cross-listing. The argument is that US disclosure requirements, exposure to SEC enforcement, and the threat of shareholder litigation make it harder, and costlier, for controlling owners and managers to extract private benefits from outside investors. For instance, Doidge, Karolyi, and Stulz (2004) show that firms with a US cross-listing exhibit a valuation premium relative to non-cross-listing firms.

Prior studies document that cross-listing in the US reduces information asymmetry, thereby improving the precision of both private and public information for non-US firms (Fernandes & Ferreira, 2008; Herrmann, Kang, & Yoo, 2015). However, theory underpinning and conclusions from the prior literature may not be applicable in the context of MNCs, particularly those registering subsidiaries in OFCs (Khanna & Palepu, 1997). In contrast to other cross-listing firms, MNCs face more severe agency cost and information asymmetry arising from foreignness, multiple currencies, cultural and language differences, geographic distances, and different operating and legal institutions (Bushman et al., 2004; Reeb et al., 1998; Shroff et al., 2014). Consequently, it is debatable how the SEC can monitor the accounting quality of these MNCs with complex, multifaceted, and dynamic institutions.

The institutional complexity theory posits that MNCs face multifold influences from both the host and home countries (Hillman & Wan, 2005). Therefore, there may be an institutional contagion or governance spillover effect on MNCs listed in the US as they attempt to obtain legitimacy in the host market (Bell, Filatotchev, & Aguilera, 2014; Cumming et al., 2017). The specific governance spillover mechanisms include board governance, incentive contract design, or attracting more foreign institutional investors and foreign directors (Areneke & Kimani, 2019; John, Saunders, & Senbet, 2000; John & Senbet, 1998; Lel, 2019). Along this vein, Gao, Zuzui, Jones, and Khanna (2017) advocate that reputation is an important strategic lever for MNCs to sustain legitimacy in the global market. Therefore, it is conceivable that cross-listed MNCs’ quest for legitimacy may incentivize them to bond to US practices and enhance their accounting quality.

Hence, given the legitimacy arguments for MNCs in prior literature, we put forward our first hypothesis (framed in null form):

Hypothesis 1:

Ceteris paribus, an MNC’s US cross-listing status is positively related with its accounting quality.

Building on the corporate governance arbitrage hypothesis, our second hypothesis focuses on MNCs’ subsidiaries’ structure. MNCs typically conduct their international activities through foreign subsidiaries operating in different institutional environments. However, the bonding literature generally overlooks MNCs’ internal legal institutions (Bushman et al., 2004; Shroff et al., 2014). Nevertheless, the corporate governance arbitrage hypothesis complements the corporate governance bonding hypothesis by suggesting that there is mobility of governance from a stronger institution to a weaker institution (Cumming et al., 2017).

It is imperative to understand how parent companies and foreign subsidiaries jointly contribute to MNC’s performance (Aguilera et al., 2019; Meyer et al., 2020). Most prior studies focus on parent–subsidiary relationships (Arregle et al., 2016; Regner & Edman, 2014), while Aguilera et al. (2019) reveal that there are limited studies on understanding the financial perspective of corporate governance of parent–subsidiary relationships for MNCs. In this regard, Akamah, Hope, and Thomas (2018) provide evidence that MNCs with tax-haven operations are more likely to aggregate their geographic disclosures. Furthermore, Huang (2018) finds that the external legal institutions of MNCs’ subsidiaries influence their earnings management. Adopting a similar perspective, we conjecture that the complex and opaque structure of MNCs operating subsidiaries in OFCs, along with OFCs’ secrecy policies and regulation arbitrage, make it easier for firms to engage in earnings manipulation (Durnev et al., 2017).

Using the subsidiary-weighted offshore attribute index developed by Masciandaro (2008), we explore how the OFC overall institutional setting, an internal governance mechanism, moderates the association between cross-listing and accounting quality. Higher offshore attribute indexes imply zero or low taxation, more flexible legal institutions, existing secrecy policies, and less scrutiny from capital market regulators and auditors. All these factors potentially affect their accounting quality, and can help MNCs obscure earnings manipulation and mitigate the bonding function arising from cross-listing. Additionally, prior studies show that the SEC is not effective in monitoring firms from less rigorous investor protection environments (Gong, Ke, & Yu, 2013; Lang, Ready, & Wilson, 2006; Shi et al., 2012). Therefore, we predict that an MNC’s choice of OFC subsidiaries will negatively moderate the association between cross-listing and accounting quality. Accordingly, we postulate the following hypothesis:

Hypothesis 2:

Ceteris paribus, an MNC’s subsidiary-weighted offshore attribute index negatively moderates the positive association between its cross-listing status and accounting quality.

In addition to the heterogeneous institutions of cross-listing and OFC subsidiaries, an MNC’s institutional complexity also encompasses another external governance mechanism, i.e., its home-country institutions. Prior research mostly supports the view that the strength of country institutions is positively associated with firms’ accounting quality and disclosure transparency (e.g., Francis, Michas, & Seavey, 2013; La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1998). However, differences between an MNC’s parent and its subsidiaries’ cultural and regulatory contexts create an institutional distance that can negatively impact the MNC’s knowledge-sharing process (e.g., Meyer et al., 2020). Further, Khanna et al. (1997) posit that, in emerging markets that lack transparent reporting standards and efficient judicial systems, there is an institutional void that allows the emergence of governance practices distinct from other institutional environments.

In alignment with institutional void and distance arguments, the corporate governance arbitrage hypothesis suggests that the mobility from a strong institutional setting to a weaker one will affect MNCs’ governance practices (Aguilera et al., 2019; Cumming et al., 2017). Within this perspective, Allred et al. (2017) document that many MNCs bypass international standards by registering shell companies in foreign jurisdictions. Similar to MNCs operating with shell companies, our sample of MNCs with OFC subsidiaries can take advantage of higher levels of regulatory arbitrage, secrecy policy, and tax avoidance, thus providing them more opportunities to manipulate earnings. Hence, it is plausible to conjecture that OFC attributes negatively moderate the positive association between home-country legal institutions and accounting quality. In light of these views, we offer the following hypothesis:

Hypothesis 3:

Ceteris paribus, an MNC’s subsidiary-weighted offshore attribute index negatively moderates the association between its home-country legal institutions and accounting quality.

Data selection and research design

Sample Selection and Data

To construct our sample of MNCs cross-listed in the US, we merge the OSIRIS international database with the data of firms cross-listing in the US. We obtained a list of ADRs that list on NYSE/AMEX/NASDAQ from the BNYMellon website.6 We exclude over-the-counter and Rule 144a private placements firms, given that they are not required to register with the SEC and incur legal bonding costs by following US disclosure practices (Coffee, 2002).

The OSIRIS international database (maintained by the Bureau Van Dijk Electronic Publishing) is a comprehensive dataset that contains over 60,000 public companies from more than 130 countries. It provides country-level and firm-level data, such as the economic development and the accounting standards relating to different countries. For most firms, OSIRIS also has information on subsidiaries, thereby allowing us to identify MNCs with subsidiaries conducting business in OFCs. Following IMF surveys (2000, 2006, 2008), Zoromé (2007), and Dharmapala and Hines (2009), we identify 40 available OFCs in OSIRIS.7 We then get MNCs that establish subsidiaries in OFCs while having their headquarters registered in non-OFC countries or jurisdictions. We limit every country to include both cross-listing and non-cross-listing firms, and firms with OFC subsidiaries. Meanwhile, all countries where MNCs have their headquarters registered are non-OFCs and must have at least 10 firm-year observations. Furthermore, we restrict our sample to all non-financial listed firms and require each firm-year observation to have information on all measures of accounting quality and control variables. After dropping observations with missing firm-level variables, there are 4,265 MNCs, all of which have OFC subsidiaries, and 16,993 firm-year observations from 31 developed and developing countries in the full sample, including 345 cross-listing firms with 1,669 firm-year observations and 3,920 non-cross-listing firms with 15,324 firm-year observations during the period of 2002–2016.8

Panels A, B, and C of Table 1 describe the country, industry, and year data distribution, respectively. Panel A reports that, among our 16,993 firm-year observations, 4,477 (26.35%) are from Japan, 1,902 (11.19%) from the UK, 1,616 (9.51%) from Taiwan, and 1,144 (6.73%) from China, while Columbia and Qatar have just 22 (0.13%) and 35 (0.21%) firm-year observations, respectively. For industry distribution, Panel B presents that industrial industry accounts for 8,847 (52.06%) observations, whereas information technology and energy industries only represent 281 (1.65%) and 539 (3.17%) observations, respectively. For sample year distribution in the period of 2002 to 2016, Panel C indicates that most observations are concentrated in the period after 2008, with 2013 representing the peak (1,883 observations or 11.15%); conversely, only 153 (0.91%) observations occur in 2002. To mitigate the impact of uneven country, industry, and year distributions and unobservable country or industry effects, we adopt industry-level and year-level random fixed-effect models in our empirical tests.

Table 1 Sample distribution
Full size table

Accounting Quality Proxies

We follow previous literature and adopt five proxies for accounting quality: (1) absolute value of abnormal accruals (ABS_ACCRUAL), (2) positive abnormal accruals (ACCRUAL+), (3) accruals quality (ACC_QUALITY), (4) modified accruals quality (MACC_QUALITY), and (5) earnings persistence (EARN_PERS). Based on prior studies (e.g., Dechow, Sloan, & Sweeney, 1995; Francis et al., 2013), we employ the modified cross-sectional Jones (1991) model to compute accruals-based earnings management proxies. Specifically, firm-year observations from different countries for each industry and year are pooled to estimate the coefficients in Eq. (1):

$$\frac{ACCRUALit}{ASSETit-1}=\alpha 1\frac{1}{ASSETit-1}+\alpha 2\frac{\Delta REVENUEit}{ASSETit-1}+\alpha 3\frac{PPEit}{ASSETit-1}+\varepsilon it$$
(1)

where ACCRUALit represents total accruals for firm I in year t, computed as the difference between earnings before extraordinary items and discontinued operations and cash flow from operations. ∆REVENUEit signifies changes in revenue between the current year and the previous year, whereas ASSETit−1 denotes total assets of the prior year, and PPEit means gross value of property, plant, and equipment of the current year.

Similarly, we use the estimated parameters from Eq. (2) to calculate nondiscretionary accruals (NDACCRUAL):

$$NDACCRUALit=\widehat{\alpha }1\frac{1}{ASSETit-1}+\widehat{\alpha }2\frac{\Delta REVENUEit-\Delta ARit}{ASSETit-1}+\widehat{\alpha }3\frac{PPEit}{ASSETit-1}$$
(2)

Here ∆ARit represents the change in net accounts receivable.

The difference between Eqs. (1) and (2) is that, in Eq. (2), change in revenues (∆REVENUEit) is adjusted by the change in accounts receivable (∆ARit), which is consistent with previous literature (e.g., Dechow et al., 1995). Further, abnormal accruals (AA) are essentially the discrepancy between total accruals (ACCRUAL) and nondiscretionary accruals (NDACCRUAL): AAit = [ACCRUALit /(ASSETit−1)] - NDACCRUALit.

We then derive two proxies for accounting quality based on abnormal accruals. The first proxy is the absolute value of abnormal accruals (ABS_ACCRUAL), which may relate to either income-increasing or income-decreasing accruals (Dechow et al., 1995). Conversely, our second proxy, ACCRUAL+, only focuses on income-increasing management in accruals.

While our first two proxies center on abnormal accruals, our third and fourth proxies for accounting quality are built on the estimation error in relating total current accruals to operating cash flows. The rationale is that accounting quality is higher when earnings can be more predictive of future operating cash flows, as advanced by Dechow and Dichev (2002). Therefore, we build on previous research (e.g., Francis, LaFond, Olsson, & Shipper, 2005) to estimate the following model for accruals error to determine our third proxy accrual quality (ACC_QUALITY):

$$CACCRUAL_{it} = \, \lambda_{0} + \, \lambda_{{1}} CFO_{it - 1} + \, \lambda_{{2}} CFO_{it} + \, \lambda_{{3}} CFO_{it + 1} + \, \lambda_{{4}} \Delta REVENUE_{it} + \, \lambda_{{5}} PPE_{it} + \varepsilon_{{{\text{it}},}}$$
(3)

where total current accruals (CACCRUAL) is the dependent variable and operating cash flow (CFO) is the test variable. CACCRUAL reflects the changes in current assets, changes in current liabilities, changes in cash, and changes in short-term debt.9 In addition, all variables are scaled by average total assets. Equation (3) is thus estimated for each industry according to the Industry Classification Benchmark (OSIRIS database), which needs a minimum of 20 observations in year t. Consequently, the accruals error in year t is computed as the standard deviation of the firm- and year-specific residuals of Eq. (3) from year t−3 to year t. If the standard deviation of residuals is larger, the accruals quality should be lower. However, to make our results easier to explain, we multiply the residual by − 1 to derive our third proxy accrual quality (ACC_QUALITY). Therefore, a higher ACC_QUALITY denotes enhanced accounting quality.

Our fourth proxy is modified accruals quality (MACC_QUALITY). Wysocki (2009) develops this proxy to mitigate the confounding effect of the accruals quality (ACC_QUALITY) used by Dechow and Dichev’s (2002), thus drawing simultaneous association between current accruals and operating cash flows. More specifically, we estimate Eqs. (4) and (5) during the year t−3 to t, and the ratio of the standard deviation of the residuals between Eq. (4) and Eq. (5) represents the modified accruals quality (MACC_QUALITY).

$$CACCRUAL_{it} = \lambda_{0} + \lambda_{{1}} CFO_{it} + \varepsilon_{{{\text{it}},}}$$
(4)
$$CACCRUAL_{it} = \lambda_{0} + \lambda_{{1}} CFO_{it - 1} + \lambda_{{2}} CFO_{it} + \lambda_{{3}} CFO_{it + 1} + \varepsilon_{{{\text{it}},}}$$
(5)

Finally, our fifth proxy for accounting quality is earnings persistence (EARN_PERS), which is defined as the negative standard deviation of a firm’s average earnings from year t−3 to year t (Dechow, Ge, & Schrand, 2010). Specifically, EARN_PERS = (− 1) × σ (Earningsit).

Empirical Models

To test Hypotheses 1, 2, and 3, we estimate the following regressions, which link cross-listing, OFC attributes, home-country governance, and their interactions with our five proxies for accounting quality:

$$\begin{aligned} Y_{it} = & {\text{b}}_{{1}} Cross\_Listing_{it} + {\text{b}}_{{2}} SUBindex \times Cross\_Listing_{it} + \, \beta_{{3}} RuleofLaw_{t}^{c} (or \, HRuleofLaw^{c} ) \\ & + \beta_{{4}} SUBindex \times RuleofLaw_{t}^{c} \left( {or \, SUBindex \times RuleofLaw^{c} } \right) + {\text{b}}_{{5}} SUBindex_{it} + {\text{b}}_{{6}} LGDP_{t}^{{\text{c}}} + {\text{b}}_{{7}} Tax_{t}^{{\text{c}}} \\ & + {\text{b}}_{{8}} Eng^{{\text{c}}} + {\text{b}}_{{9}} Mcap_{t}^{c} + {\text{b}}_{{{1}0}} Big5_{it} + {\text{b}}_{{{11}}} Litigate_{it} + {\text{b}}_{{{12}}} Inde_{it} + {\text{b}}_{{{13}}} Ifrs_{it} + {\text{b}}_{{{14}}} Aanalyst \, Following_{it} \\ & + {\text{b}}_{{{15}}} Size_{it} + {\text{b}}_{{{16}}} Mtb_{it} + {\text{b}}_{{{17}}} Lev_{{{\text{i}}t}} + {\text{b}}_{{{18}}} Neti_{it - 1} + {\text{b}}_{{{19}}} \sigma \_CFO_{it} + {\text{b}}_{{{2}0}} \sigma \_Rev_{it} + {\text{ orb}}_{{{21}}} CFO/Sales_{it} \\ & + industry \, fixed \, effects \, + \, year \, fixed \, effects \, + \varepsilon_{{{\text{it}}}} , \\ \end{aligned}$$
(6)

where Yit represents one proxy for accounting quality: the absolute value of abnormal accruals (ABS_ACCRUAL); positive abnormal accruals (ACCRUAL+); accruals quality (ACC_QUALITY); modified accruals quality (MACC_QUALITY); and earnings persistence (EARN_PERS). c denotes countries or jurisdictions where the headquarters of MNCs are registered, I refers to firms, and t to years.

Given the diversity of country and industry attributes, as well as time variation on measures of accounting quality, we incorporate industry and year fixed effects in every regression for our multi-level random effect models (Roychowdhury, 2006). We control for country-level institutional factors (i.e., countries or jurisdictions where the headquarters of firms are set up), which include the market capitalization of countries (Mcap), average corporate tax rate (Tax) of a particular country (Farah, Elias, Chakravarty, & Beamish, 2021), natural log of country gross domestic product (GDP) (LGDP), and whether English (Eng) is the official language.

Our primary testing variables in Eq. (6) are Cross_Listing, SUBindex × Cross_Listing, and SUBindex × RuleofLaw (or SUBindex × HRuleofLaw). Cross_Listing is an indicator variable that equals one if an MNC is cross-listed in the US and zero otherwise. It is adopted to test Hypothesis 1. Consistent with corporate bonding hypothesis, we predict that MNCs cross-listed in the US have higher accounting quality. Specifically, in the models for ABS_ACCRUAL and ACCRUAL+, the coefficients for Cross_Listing should be negative, while in the models for ACC_QUALITY, MACC_QUALITY, and EARN_PERS, the coefficients should be positive.

In addition, we adopt the subsidiary-weighted offshore attitude indexes (SUBindex) to measure the legal institutions among MNCs’ subsidiaries. This index is advanced by Masciandaro (2008), encompassing diverse aspects of MNCs subsidiaries. Specifically, to construct SUBindex, an OFC’s manifestation of economic crime, rigor of political systems, enforcement of regulations, possible national benefits, and whether it is included in one of the OFC blacklists10 are considered. To comprehensively account for MNCs’ subsidiary institutions, we incorporate all subsidiaries of MNCs, i.e., those registered in OFCs and those that are non-OFC registered. If a country does not demonstrate any OFC attribute, the offshore attribute index is defined as zero. If a country shows some OFC attributes but was not included in any blacklist, the offshore attribute index is defined as one. Similarly, if a country indicates some OFC attributes and was also listed in one, two, or three blacklists, the offshore attribute index is defined as two, three, and four, respectively. Furthermore, one is added to the offshore attribute index if a country or jurisdiction is on the market list of OFCs.11 For non-OFC subsidiaries, the offshore attribute index is equal to zero or one.

In sum, the offshore attribute index spans from zero to five, with zero representing the lowest level of OFC attribute and five signifying the highest level of OFC attribute. To capture the overall MNC subsidiary institutional complexity, we weight the offshore attribute index by the number of subsidiaries (divided by the total number of subsidiaries of an MNC). Specifically, the following formula depicts how SUBindex is calculated:

$$SUBinde{x}_{it}={\sum }_{c}\left(offshore\,attribute\,inde{x}^{c}\times subsidiar{y}_{it}^{c}\right)/{\sum }_{c}subsidiar{y}_{it}^{c}$$
(7)

The variable subsidiary refers to the number of subsidiaries that a firm I has in country c in a given year t.12 Essentially, if an MNC has more subsidiaries in countries with larger offshore attribute index scores, SUBindex will demonstrate larger values. The interaction between Cross_Listing and the SUBindex is to examine Hypothesis 2 (i.e., SUBindex × Cross_Listing), whereas the interaction between RuleofLaw and SUBindex is to test Hypothesis 3 (i.e., SUBindex × RuleofLaw).

RuleofLaw is employed to measure the home-country institutions of MNCs. This proxy reflects the strength of contract enforcement, property rights, and the police and the courts, as well as the likelihood of crime and violence for a particular MNC (Kaufmann, Kraay, & Mastruzzi, 2011; La Porta et al., 1998). In addition, we define a binary variable HRuleofLaw as another proxy for home-country institution. It is equal to one if an MNC’s headquarters are registered in a country with a value of RuleofLaw that is higher than the median of our sample and zero otherwise.

Consistent with Hypotheses 2 and 3, we predict the coefficient signs for SUBindex × Cross_Listing and SUBindex × RuleofLaw to be positive in the models for ABS_ACCURUAL and ACCRUAL+ and negative in the models for ACC_QUALITY, MACC_QUALITY and EARN_PERS.

Following previous literature (e.g., Roychowdhury, 2006; Zang, 2012), we also control for various firm-level variables that potentially influence a firm’s accounting quality: firm litigation risk (Litigate), firm size (Size), market to book ratio (Mtb), firm leverage (Lev), profit in the previous year (Neti), Big 5 auditor (Big5), the independence of controlling shareholders, the ratio of operating cash flow to net sales (CFO/Sales), and the adoption of International Financial Reporting Standard (Ifrs). Further, given that previous studies (e.g., Hribar & Nichols, 2007) posit that fundamentals volatility and unsigned accruals are positively related to each other, we therefore control for volatility of operating cash flows (σ_CFO) and volatility of sales (σ_Rev) using four years of historical data.13 “Appendix A” provides a summary of all variable definitions.

Empirical results

Descriptive Statistics

Table 2 reports descriptive statistics for country and firm-specific variables, with Panel A presenting the statistics for 1,669 cross-listing firm-year observations, Panel B summarizing the statistics for 15,324 non-cross-listing firm-year observations, and Panel C providing the results of univariate tests for the mean and median differences for the five accounting quality proxies between cross-listing and non-cross-listing samples.14 As indicated in Panel A, for cross-listed MNCs in the US, the mean and median of absolute discretionary accruals (ABS_ACCRUAL) are 10.5% and 5.2%, respectively, comparable to the mean (median) ABS_ACCRUAL for US domestic firms of 11% (6%) reported by Cohen, Dey, and Lys (2008). Relatedly, the mean and median of positive discretionary accruals (ACCRUAL+) are 6.1% and 5.1%, respectively. The mean (median) ACC_QUALITY for the cross-listed MNCs is − 1.674 (− 0.358), while the mean (median) MACC_QUALITY and EARN_PERS are 2.986 (1.041) and − 0.032 (− 0.020). Panel B reports that, for non-cross-listing OFC MNCs, the means and medians of ABS_ACCRUAL are 15.9% and 7%, while for ACCRUAL+ they are 7% and 6.2%, respectively. The mean (median) ACC_QUALITY for the non-cross-listing MNCs is − 2.171 (− 0.564), while the mean (median) MACC_QUALITY and EARN_PERS are 2.267 (0.895) and − 0.037 (− 0.020), respectively.

Table 2 Summary statistics and results of univariate tests
Full size table

Panel C compares the mean (median) differences for the five accounting quality measurements: ABS_ACCRUAL, ACCRUAL+, ACC_QUALITY, MACC_QUALITY, and EARN_PERS. Results of the t and Z tests show that the mean and median differences between cross-listing and non-cross-listing OFC MNC samples are significant for all five proxies. More explicitly, pertaining to Hypothesis 1 at the univariate level, we find that cross-listing OFC MNCs have lower absolute value of abnormal accruals (ABS_ACCURAL), lower positive abnormal accruals (ACCRUAL+), higher accruals quality (ACC_QUALITY) and modified accruals quality (MACC_QUALITY), and more persistent earnings patterns (EARN_PERS) than non-cross-listing OFC MNCs.15

Results of Primary Regressions

Table 3 reports the results of our baseline regressions in Eq. (6). For Tables 3, 4, 5, and 6, the t and Z values of all the models are built on multilevel year and industry random fixed effects clustering in estimated coefficients, as well as in their standard. Columns 1 and 2 present the results for the regressions, with the absolute value of abnormal accruals (ABS_ACCRUAL) as the dependent variable. The coefficients on Cross_Listing are significantly negative (− 0.016) in these two columns, suggesting that cross-listing in the US is related to a lower absolute value of abnormal accruals. The coefficients are also economically significant. Specifically, cross-listing in the US is associated with a level of abnormal accruals that is 10.39% lower than for non-cross-listing firms (− 10.39% = − 0.016/0.154, where − 0.016 is the coefficient on Cross_Listing and 0.154 is the sample mean of the entire sample for ABS_ACCRUAL).

Table 3 Baseline results
Full size table
Table 4 Heckman two-stage model
Full size table
Table 5 Difference-in-difference analyses
Full size table
Table 6 Three distinguishing aspects of offshore financial centers
Full size table

Similarly, columns 3 and 4 report the results on our second proxy ACCRUAL+. The coefficients on ACCRUAL+ are also negative and significant (coefficients = − 0.019 and − 0.018), signifying that MNCs cross-listing in the US exhibit 27.54% fewer positive abnormal accruals than non-cross-listing counterparts (− 0.019/0.069, where − 0.019 is the coefficient on Cross_Listing and 0.069 is the sample mean of the entire sample for PAAC).

Furthermore, columns 5 and 6 summarize the results for the models with accruals quality (ACC_QUALITY) as the dependent variable, while columns 7 and 8 are for the models with modified accruals quality (MACC_QUALITY) as the dependent variable. The coefficients on Cross_Listing are positive and significant across all four models, implying that cross-listed MNCs are associated with better accruals quality than MNCs not cross-listing in the US. Finally, columns 9 and 10 show the results for regressions with earnings persistence (EARN_PERS) as the dependent variable: the coefficients on Cross_Listing are again positively significant, showing that cross-listing MNCs exhibit more persistent earnings patterns compared to non-cross-listing MNCs. Taken together relative to Hypothesis 1, these findings suggest that by cross-listing in the US, foreign MNCs commit to higher accounting quality, which is consistent with the corporate governance bonding hypothesis, as it implies that investors are unlikely to be expropriated by insiders.

With respect to Hypothesis 2, the coefficients on SUBindex × Cross_Listing are positive and significant from columns 1 to 4 (coefficients = 0.013, 0.013, 0.012, and 0.010, respectively), but negative and significant across columns 5 to 10 (coefficients = − 0.011, − 0.033, − 0.038, − 0.016, − 0.007, and − 0.007, respectively). These results suggest that the accounting quality of cross-listed MNCs is reduced as an MNC’s OFC subsidiaries exhibit increasing offshore attribute indexes: higher SUBindex scores for cross-listed MNCs are associated with increased absolute values of abnormal accruals (ABS_ACCRUAL), increased positive abnormal accruals (ACCRUAL+), decreased accruals quality (ACC_QUALITY), decreased modified accruals quality (MACC_QUALITY), and reduced earnings persistence (EARN_PERS). For example, for cross-listed MNCs, if the SUBindex increases by one, the absolute value of abnormal accruals (ABS_ACCRUAL) rises by 0.028 (0.028 = 0.015 + 0.013, where 0.015 is the coefficient on SUBindex × Cross_Listing in column 1 and 0.013 is the coefficient on SUBindex in column 1).

In addition, a series of F tests on the coefficients of the sum of Cross_listing and SUBindex × Cross_listing show that we can reject the hypothesis that the sum of the coefficients equals zero. For instance, the sum of the two coefficients is − 0.007 (− 0.019 + 0.012) in column 3, and the F value equals 16.9.16 This evidence is consistent with Hypothesis 2 that the association between cross-listing and accounting quality for MNCs is weaker for MNCs operating in OFCs with higher SUBindex, which signifies that the corporate governance bonding benefits associated with accounting quality for MNCs are mitigated by the corporate governance arbitrage effect.

Regarding Hypothesis 3, the coefficients on SUBindex × RuleofLaw (HRuleofLaw) are positive and significant from columns 2 and 4 (coefficients = 0.006 in both cases), but negative and significant across columns 6–10 (coefficients = − 0.168, − 0.175, − 0.113, − 0.001, and − 0.003, respectively). These results suggest that the impact of an MNC’s home-country governance on its accounting quality is negatively moderated by SUBindex. A series of F tests on the coefficients of the sum of RuleofLaw (HRuleofLaw) and SUBindex × RuleofLaw (SUBindex × RuleofLaw) show that we can reject the hypothesis that the sum of the coefficients equals zero in most regressions. This evidence is consistent with Hypothesis 3 that the association between home-country governance and accounting quality for MNCs is weaker for MNCs operating in OFCs with high OFC attributes, thus lending further support to the corporate governance arbitrage hypothesis.

Robustness checks and additional analyses

Self-Selection Bias

An MNC’s decision to cross-list in the US is voluntary, because it is plausible that some firms are more likely to cross-list (Doidge, Karolyi, & Stulz, 2004). Therefore, we cannot rule out the possibility that MNCs with better accounting quality are more likely to cross-list in the US, and that a firm’s decision to cross-list is a function of unobservable omitted variables that are correlated with our test variables (an endogeneity issue). We address the above issues by performing the Heckman (1979) two-stage treatment effect model. In the first stage, following previous literature (e.g., Herrmann et al., 2015), we estimate a probit choice model in which the likelihood of an MNC cross-listing in the US is linked with firm-specific and country-wide factors. We employ ave_Disclosure (the average of the World Bank business disclosure index, which measures the country-level extent of business transparency) and ave_LnBM (the average of the log of book-to-market ratio, which is a firm-level measure of information transparency) as the exogenous variables in the first stage, as they impact the likelihood of cross-listing but it is unlikely that they will influence the accounting quality of MNCs.17

In the second stage, we estimate the regressions of Eq. (6) after including the inverse Mills ratio, obtained from the first-stage probit model. As indicated in Panel A of Table 4, at the country level, MNCs originating from countries with a civil law system (Legal Origin), higher regulation quality (RegulationQuality), better economic development (LnGNP), and higher country market capitalization (Mcap) are more likely to cross-list in the US. At the firm level, our results show that larger firms (Size) and those with better financial performance (Roe) are more likely to list their shares in the US. Panel B of Table 4 shows that the results of Heckman two-stage models are, overall, consistent with those of multi-level random fixed effect regressions in Table 3.

Endogeneity

In addition to the Heckman two-stage model, we also employ three difference-in-difference models to mitigate the endogeneity concerns for cross-listing, as well as for OFC choices. First, we adopt a difference-in-difference model for OFC MNCs changing from non-cross-listing to cross-listing. Second, we undertake another difference-in-difference model for cross-listing firms with strong home-country institutions (RuleofLaw) changing from non-OFC to OFC status. Third, we exploit a unique exogenous regulation shock, the OECD CRS, to examine the changes before and after the CRS adoption.

Table 5 reports the results for these difference-in-difference analyses. Panel A compares accounting quality of an MNC before and after cross-listing in the US during our sample period. The treatment group comprises OFC MNCs changing from non-cross-listing to cross-listing, whereas the control group incorporates non-cross-listing OFC MNCs. The coefficients for PostCross_Listing are significant in all models, lending further credence to Hypothesis 1. In addition, the results for SUBindex × PostCross_Listing and SUBindex × HRuleofLaw remain robust, consistent with Hypotheses 2 and 3.

Panel B examines the accounting quality of a cross-listing MNC changing its status from non-OFC to OFC, using a cross-listing subsample for firms with strong home-country legal institutions (HRuleofLaw). The treatment group includes cross-listing MNCs that alter their status from non-OFC to OFC, while the control group includes MNCs that retain their non-OFC status throughout the sample period. The coefficients for PostOFC × SUBindex are positive and significant for ABS_ACCRUAL and ACCRUAL+, whereas they are negative and significant for ACC_QUALITY and EARN_PERS, suggesting that the change of status from non-OFC to OFC further reduces the accounting quality of cross-listing MNCs.

Finally, in Panel C, we report the results of the exogenous impact of an OECD regulation. In 2014, OECD issued CRS, which mandated the exchange of tax and financial information among OFCs for OECD countries. The treatment group comprises cross-listing MNCs from OECD countries, while the control group encompasses non-cross-listing MNCs from OECD countries. Post is equal to one after 2014 and zero otherwise. As presented in Panel C, SUBindex × HRuleofLaw retains the expected sign in all five models. Nonetheless, the coefficient for Post × SUBindex × HRuleofLaw is negative and significant in model 2 for ACCRUAL+, and positive and significant for models 3 and 5 for ACC_QUALITY and EARN_PERS. This result implies that OFC MNCs enhance their accounting quality after the adoption of CRS. In sum, our difference-in-difference analyses provide strong support to our primary findings.

To further mitigate the concerns of endogeneity, we also perform the propensity score matching (PSM) methodology.18 For brevity, we do not report the results. However, results of the PSM estimation lend further credence to our main results reported in Table 3.19

The Impact of Offshore Characteristics on Accounting Quality

To further explore the influence of OFC characteristics on accounting quality, we adopt three proxies, i.e., secrecy policies, regulation arbitrage, and tax avoidance opportunities, to replace the overall measure of OFC characteristics, SUBindex. Secrecy policies are computed as a firm’s subsidiary-weighted secrecy indicator on the IMF group index. According to the IMF (2000), OFCs are categorized into three categories based on transparency for international cooperation and supervision. OFCs are listed in Group I if they have the highest level of transparency, and listed in Group III if they exhibit the lowest level of transparency.20 We code one, two, and three for OFCs in Group I, Group II, and Group III, respectively, and compute the subsidiary-weighted transparency index as the proxy of the secrecy policies (Secrecy). Regulation arbitrage represents the flexible regulation gap between an MNC’s subsidiaries and its home country; therefore, it is directly related to the corporate governance arbitrage hypothesis. The proxy of tax avoidance (TaxDiff) is defined as the difference between the average corporate tax rate of an MNC’s home country and the weighted corporate tax rate of its subsidiaries.21

Panels A, B, and C of Table 6 summarize the regression. For Hypothesis 1, the coefficients for Cross_Listing appear to be negative and significant in models of ABS_ACCRUAL and ACCRUAL+, while positive and significant in models of ACC_QUALITY, MACC_QUALITY, and EARN_PERS. Related to Hypothesis 2, we find that, in the models with secrecy policies and regulation arbitrage, the evidence is consistent with our primary results. However, in models with tax avoidance opportunities, the coefficients on TaxDiff × Cross_Listing are only significant in one model (EARN_PERS), implying that secrecy policies and regulation arbitrage related to MNCs’ OFC subsidiaries likely dominate tax avoidance in mitigating the cross-listing benefits.22

Additional Analyses

To further enhance the robustness of our results, we undertake two additional tests. For the first additional test, we add Canadian firms back into our sample. We excluded Canadian MNCs from our primary sample because Canadian firms, different from ADRs, list their shares directly on the US stock exchanges and are exempt from many disclosure requirements under the multijurisdictional disclosure system (Foerster & Karolyi, 1999). The second additional test is to employ country-weighted least squares to diminish the concerns of uneven sample size across countries. The results, untabulated for brevity, generally support our primary findings.

Concluding remarks

This study investigates how tensions related to an MNC’s institutional complexity affect its accounting quality. We employ MNCs registering subsidiaries in OFCs as our sample, given that these firms are characterized by complex and multi-level legal structures. Hence, these MNCs offer a unique opportunity to simultaneously test the corporate governance bonding hypothesis and the corporate governance arbitrage hypothesis (Aguilera et al., 2019; Cumming et al., 2017). In terms of the corporate governance bonding hypothesis, we hope to examine whether the enforcement from the SEC and other regulators is effective in reaching MNCs that may expect to enhance their legitimacy in the US markets (Bushman et al., 2004). In light of the corporate governance arbitrage hypothesis, we seek to examine whether setting up subsidiaries in OFCs with opaque corporate structures and less stringent legal institutions negatively moderates the effect of cross-listing or the effect of MNCs’ home-country institutions on accounting quality.

We find that MNCs cross-listing in the US exhibit higher accounting quality, which supports the corporate governance bonding hypothesis that cross-listing MNCs benefit by reducing information asymmetry, thus leading to higher accounting quality (Coffee, 2002). Further, we document that the positive association between cross-listing and accounting quality is negatively moderated by the internal institutions of OFC subsidiaries, thereby providing support to the corporate governance arbitrage hypothesis. We also explore the channels through which the subsidiaries’ legal framework plays a role, and find that the moderation effect is through regulation arbitrage and secrecy policies in OFCs, rather than through tax avoidance. Finally, we show that the internal OFC attributes also negatively moderate the positive association between MNCs’ home-country institution and accounting quality, which offers further evidence to verify the corporate governance arbitrage hypothesis. Overall, we conclude that tensions arising within MNCs because of their institutional complexity need to be taken into account when examining corporate outcomes, with the impact on accounting quality being one illustration of such forces.

As in other studies, our study is subject to caveats. For instance, our study only focuses on MNC firms. Do non-MNC firms exhibit different characteristics of accounting quality? In addition, how would firm-level corporate governance factors, such as the board of directors’ attributes, impact the accounting quality of MNCs? Finally, do broader constructs such as culture affect MNCs’ accounting quality and, if so, to what extent do they interact with corporate governance? Given the scant evidence on these questions, we leave them for further research.

Notes

1In this study, we follow Zoromé’s (2007: 7) definition of an OFC as “a country or jurisdiction that provides financial services to non-residents on a scale that is incommensurate with the size and the financing of its domestic economy.” Based on this definition, we identified the world’s 40 primary OFCs using the OSIRIS database (Appendix A). The Zoromé (2007) definition is also consistent with the International Monetary Fund’s (IMF) definition of OFCs (IMF, 2000). The IMF defines OFCs as “(I) jurisdictions that have relatively large numbers of financial institutions engaged primarily in business with non-residents; (ii) (jurisdictions with) financial systems with external assets and liabilities out of proportion to domestic financial intermediation designed to finance domestic economies; and (iii) more popularly, centers which provide some or all of the following services: low or zero taxation; moderate or light financial regulation; banking secrecy and anonymity” (Offshore financial centers. IMF background paper. Available at: http://www.imf.org/external/np/mae/oshore/2000/eng/back.htm).

2https://www.justice.gov/opa/pr/odebrecht-and-braskem-plead-guilty-and-agree-pay-least-35-billion-global-penalties-resolve. Retrieved January 18, 2021.

3https://www.sec.gov/Archives/edgar/data/1071438/000119312517291642/d446350d20f.htm#fin446350_5. Retrieved January 18, 2021.

4https://www.sec.gov/litigation/litreleases/lr18527.htm. Retrieved January 17, 2021.

5https://www.reuters.com/article/wirecard-accounts-ey/german-prosecutors-probe-ey-auditors-over-wirecard-collapse-idUSKBN28E29Z. Retrieved January 18, 2021.

6Please refer to https://www.adrbnymellon.com/directory/dr-directory for a complete list of ADR firms. ADR listing is different from direct stock listing.

7Please refer to online Appendix Table A3.

8We chose the sample period starting from 2002 to mitigate the possibility of changes among cross-listed firms before and after the adoption of the Sarbanes-Oxley Act in early 2002.

9CAACRUALit is defined as ΔCA-ΔCL-ΔCASH+ΔSTDEBT. Here, ΔCA denotes the change in current assets, ΔCL is the change in current liabilities, ΔCASH is the change in cash, and ΔSTDEBT represents the change in short-term debt.

10The blacklists include the Financial Stability Forum list, the Financial Action Task Force (FATF) list of Non-Cooperative Countries and Territories, and the OECD list of tax havens.

11The market list of OFCs is obtained from the International Financial Centers’ Yearbook (IFCY) dataset from 2006 to 2007, by which a country or jurisdiction is classified as an OFC if the authorities of a country or jurisdiction approved it. See Masciandaro (2008) and Rose and Spiegel (2007) for a description of the index.

12Almost all of the offshore attribute indexes (Masciandaro, 2008) for non-OFC countries or jurisdictions are zero or one.

13Following previous literature, we also control for foreign institutional investors (Lel, 2019) and foreign directors (Areneke & Kimani, 2019) in a reduced sample. Our results are not sensitive to these additional variables.

14We winsorize the primary firm variables at the 1% and 99% levels.

15We also perform other univariate tests, while, for brevity, we do not tabulate the Pearson cross-correlation matrix. We have performed variance inflation factors tests to ensure that multi-collinearity is not a concern for all the variables included in the model. The two accruals quality measures (ACC_QUALITY and MACC_QUALITY) display a positive correlation of 0.197, consistent with previous research (e.g. Biddle et al., 2009). However, these two accruals quality proxies are negatively related to the absolute value of abnormal accruals (ABS_ACCRUAL). Our primary testing variable,Cross_Listing, is negatively associated with ABS_ACCRUAL and ACCRUAL+, whereas it is positively associated with ACC_QUALITY, MAC_QUALITY, and EARN_PERS, signifying that MNCs cross-listing in the US enjoy higher accounting quality compared to their non-cross-listing counterparts.

16All the F tests for other columns are significant.

17We also use an alternative instrument variable, “country-level liquidity” (Doidge et al., 2004), to perform our Heckman two-stage model. Our results are not sensitive to this correction.

18For the advantages of using the PSM, refer to Larcker and Rusticus (2010).

19For PSM results, please refer to online Appendix Table A2.

20Data source: International Monetary Fund (2000). Offshore Financial Centers: IMF Background Paper. The IMF has not updated its categorization of OFC groups since 2000, so we hold the group numbers of OFCs constant throughout our sample period.

21OECD Tax Database, Table II.1 – Corporate income tax rates: basic/non-targeted (last updated December 2018), http://www.oecd.org/tax/tax-policy/tax-database.htm. Corporate income tax rates around world, 2020 (Tax Foundation, https://taxfoundation.org/publications/corporate-tax-rates-around-the-world).

22It is worth noting that we do not include the interaction term SUBindex × RuleofLaw (HRuleofLaw) in Table 6 because the three testing variables here (i.e., Secrecy, Regulation Arbitrage, and TaxDiff) are calculated as the differences between a firm’s home-country index and its subsidiary-weighted index, which are similar to SUBindex × RuleofLaw.