Audit industry specialization and
financial reporting quality of
multinational companies
Name: Leonie Russchen
Student number: S2686422
Supervisor: Simona Rusanescu
Address: Tweede Hunzestraat 37a
Phone number: 0615648916
Email: L.Russchen@student.rug.nl
2 Abstract
Based on the agency theory, this research examines the relation between industry specialization of the auditor of a multinational company and the financial reporting quality of the subsidiaries, measured by earnings management. Industry specialists possess industry specific knowledge which gives them a more comprehensive understanding of the client and therefore improves the financial reporting quality. Subsidiaries are monitored by separate auditors, which have to deliver a summary of the work they did to the auditor of the parent company. Therefore, the expectation is that industry specialization of the auditor of the parent company improves the evaluation of this summary and therefore enhances the financial reporting quality of the subsidiaries. Using a sample of private European subsidiaries of US multinationals, results indicate that there is no relation between audit industry specialization and subsidiary accrual based earnings management. Additionally, audit industry specialization does only partially relate to real earnings management. The absence of a relation might be explained by the influence of geographic proximity. The distance between the auditor of the parent and the subsidiary auditor might mitigate the effects of industry specialization. On the other hand, the subsidiary might not operate in the same industry as the parent.
3
TABLE OF CONTENT
INTRODUCTION ... 4
THEORY AND HYPOTHESES DEVELOPMENT ... 9
Reporting quality of MNCs ... 9
Audit industry specialization ... 13
METHODOLOGY ... 17
Sample selection ... 17
Main independent variable - Parent auditor industry specialization ... 17
Dependent variable – Subsidiary AEM ... 19
Dependent variable – Subsidiary REM ... 20
The empirical model ... 22
Control variables ... 22
RESULTS ... 24
Descriptive statistics ... 24
Correlations ... 25
Regression analysis ... 26
DISCUSSION AND CONCLUSION ... 29
Findings ... 29
Implications ... 31
Limitations ... 32
REFERENCES ... 33
APPENDIX A – SIC code definitions... 37
4
INTRODUCTION
Multinational companies (hereafter: MNCs) and their subsidiaries play a vital role in the modern economy (Corry & Cormican, 2019). According to De Backer et al. (2019) and the OECD (2018) MNCs are responsible for more than half of the world exports. Especially their subsidiaries abroad that are trading internationally, account for over 30% of the world exports, which is more than MNC headquarters and domestic plants together (24%). Furthermore, the contribution of MNCs to the world gross domestic product is estimated at nearly 30%, of which roughly one third is contributed through foreign subsidiaries. Additionally the overall share of MNCs in global employment is about 23%. Therefore, considering the vital role of MNCs in the current economy, the effect of their corporate structure deserves closer attention.
The corporate structure of MNCs is often very complex (Beuselinck et al., 2019). This complexity exists because they (usually) have a large number of subsidiaries, which in some cases exceeds 100 (Grewal et al., 2018). The conventional belief of viewing the foreign subsidiary as a subordinate entity within the MNC has been the subject of continuing debate in current literature (Corry & Cormican, 2019). On the one hand, parents prefer centralized decision rights in order to fully control foreign subsidiaries (Huang, 2018). On the other hand, parents may delegate their decision rights to subsidiary managers to adapt to the foreign environment (Huang, 2018). The role of a subsidiary is developing as described by Delany (2000) from doing only what is expected, to doing what also makes good business sense. Therefore researchers argue that managing a subsidiary effectively is not simply about following the mandate prescribed by the parent (Corry & Cormican, 2019; Delany, 2000). It is rather about fulfilling the current mandate in a superior way and pursuing strategic initiatives that add new value to the corporation (Corry & Cormican, 2019; Delany, 2000). However, eventually the MNC-parent still has a lot of power over their subsidiary, since the subsidiary is influenced by the mandate from the parent.
MNCs must provide a consolidated statement that contains the financial information of all their subsidiaries (Beuselinck et al., 2019). The purpose of the consolidation process is to present the results and the financial position of a parent company and its subsidiaries, as if the group were a single entity (Beuselinck et al., 2019). Therefore, not only the financial reporting of the MNC-parent is important, but also the financial reporting quality (hereafter: FRQ) of the subsidiaries.
5
Investors and users are interested in achieving a high quality of financial information, and this quality, among other things, depends on having a high quality of earnings (Herath & Albarqi, 2017). Therefore, the focus of this paper is on the earnings quality in financial reporting and follows the definition of Dechow et al. (2010, p. 344), which states: “Higher quality earnings provide more information about the features of a firm’s financial performance that is relevant to a specific decision made by a specific decision-maker”. Because earnings management (hereafter: EM) is an alteration of the reported accounting numbers (Mangala & Isha, 2019), it is used in this research as an inverse measure of FRQ. It occurs when managers use their own judgment to alter the numbers in the financial statements to mislead their stakeholders about the actual performance of the company (Healy & Wahlen, 1999). EM is further explained using the agency theory. Regarding MNCs, agency problems arise whenever the interests of the CEO or managers of the parent company are not aligned with those of the monitors or shareholders of the MNC (Beuselinck et al., 2019). In this case the CEO of the MNC-parent has the opportunity to manage earnings at subsidiary level, which affects the consolidated statements since these include the financial information of the subsidiaries. Thus, if there is more EM, the FRQ of the MNC decreases because the consolidated financial statements are less useful for shareholders or decision makers. There are two types of EM, namely accrual based earnings management (hereafter: AEM) and real earnings management (hereafter: REM). AEM is the managers’ manipulation of discretionary accruals to alter the numbers in the financial statements (Jones, 1991). REM refers to managerial intervention by manipulating earnings through actions that depart from normal business practices (Anissa et al., 2019).
Reporting practices of subsidiaries are likely to be influenced by many local as well as MNC headquarters-level factors (Beuselinck et al., 2019). For example, the opportunity for tax avoidance (Durnev et al., 2017; Dyreng et al., 2012; Leuz et al., 2003), the legal environment of the subsidiary (Durnev et al., 2017), the level of centralization at the MNC-parent (Huang, 2018) or corporate governance characteristics of the MNC-parent (Mangala & Isha, 2019). A better understanding of the factors influencing the FRQ of the MNC is essential, given the economic importance of MNCs nowadays. Therefore, this research focusses on a factor that is not yet researched in relation to subsidiary FRQ: audit industry specialization at the MNC-parent.
The issue of industry specialization has become increasingly relevant to the auditing profession as audit firms organize their practices along industry lines rather than traditional
6
service lines (Knechel et al., 2007). The research of Romanus et al. (2008) suggests that auditors with industry-specific knowledge are more likely to possess a comprehensive understanding of a company's characteristics. Understanding the client’s industry enhances the auditor’s ability to critically evaluate the recognition and valuation of transactions and events related to that industry (Scott & Gist, 2013). Therefore, research suggests that audit industry specialization is associated with lower levels of EM (Krishnan, 2003; Balsam et al., 2003; Anissa et al., 2019). However, there are also studies that contradict this relation (Chi et al., 2011).
The influence of audit industry specialization on FRQ within MNCs is very limited in the current literature. Even so, the available research shows that auditors with more experience possess expertise in performing global group audits in a way that positively affects the earnings quality of consolidated statements (Gunn & Michas, 2018). Additionally, Downey and Bedard (2018) find that experience and knowledge has a positive influence on resolving challenges within MNCs. Since there is no research done yet regarding the relation between audit industry specialization of the MNC-parent and subsidiary FRQ, measured by the level of EM, this research studies this relation by examining the following research question:
RQ: How does audit industry specialization at the MNC-parent influence the FRQ of the subsidiary?
According to the agency theory external audits work as a mechanism to reduce information asymmetry between the managers and monitors of a company and to mitigate the possibility of self-serving reporting by management (Greenwood & Zhan, 2019; Shahzad et al., 2019). Previous studies show that audit industry specialization is associated with lower levels of AEM, because industry specialists have the expertise, resources and the incentive to constrain opportunistic reporting of accruals (Krishnan, 2003; Balsam et al., 2003). Although there is no research yet regarding the relation between subsidiary AEM and audit industry specialization of the MNC-parent, previous studies do show a positive relation between audit industry specialization and the consolidated statements of MNCs (Downey & Bedard, 2018; Gunn and Michas, 2018). Additionally, geographic proximity causes auditors to more effectively monitor client reporting behavior, constrain EM and improve the FRQ, because they have more client specific knowledge (Choi et al., 2012). Since subsidiaries are audited by separate auditors than
7
the parent company, this auditor has more client-specific knowledge because of the proximity to the subsidiary. Therefore, subsidiary AEM decreases because audit industry specialization at the MNC-parent improves the evaluation of the work of the subsidiary auditor. Thus, it is expected that the association between subsidiary AEM and audit industry specialization of the MNC-parent is negative.
Regarding the relation between audit industry specialization and REM studies vary in their findings. On the one hand, the research of Anissa et al. (2019) shows that industry specialists have a deeper client knowledge than non-specialists and stronger incentives to provide higher quality audits to protect themselves against reputational damage. On the other hand, Chi et al. (2011) find evidence of a negative relation, suggesting that managers switch from using AEM to using REM, when higher audit quality constrains the use of AEM. Regarding MNC-parent audit industry specialization and subsidiary REM there is no previous research examining this specific relation. Even so, it is expected that subsidiaries might switch from using AEM to REM, when higher audit quality, caused by auditor industry specialization at the MNC-parent, constrains the use of AEM at both MNC-parent level as well as subsidiary level. Therefore, the expected association between audit industry specialization at the MNC-parent and subsidiary REM is positive.
A quantitative approach is used to examine the research question. The data for this research is collected from annual reports of (1) listed parent companies from the US and (2) private subsidiaries from European countries and covers the years 2012 to 2017. The size of the final sample is 3954 observations. The dependent variable is the FRQ of the subsidiaries. This is measured using both AEM and REM. For the measurement of subsidiary AEM the modified Jones model (Dechow et al., 1995) is used to compute subsidiary abnormal accruals. For the estimation of subsidiary REM the methodology of Roychowdhury (2006) is followed. Specifically, the abnormal operating cash flow (CFO) of subsidiaries is used to measure the level of manipulation by subsidiary management through boosting the sales volume. Additionally, the subsidiary abnormal production costs is used to measure the level of overproduction used to lower the costs of goods sold. The independent variable of this research is the auditor industry specialization of the MNC-parent. This is defined following Balsam et al. (2003), who describe specialists as the largest supplier in the industry when the difference between the first and second supplier is at least 10 percent. Additionally, if the largest suppliers in the industry differ less than
8
10 percent, and the different between them and the remaining auditors is more than 10 percent, these are all considered industry specialists (Balsam et al., 2003).
The results of this research show that there is no association between auditor industry specialization at the MNC-parent and subsidiary AEM or REM, as measured by abnormal CFO. However, industry specialization of the MNC-parent auditor does decrease the level of abnormal production costs of the subsidiary. Following the explanation of Anissa et al. (2019) auditor industry specialization decreases the level of REM, because the auditor has more experience and knowledge and is therefore able to detect REM better than non-specialists. Regarding the absent relation between audit industry specialization at the MNC-parent and subsidiary AEM and subsidiary abnormal CFO, there are multiple explanations. First of all, Choi et al. (2012) argue that geographic proximity mitigates information asymmetries and enhances monitoring effectiveness. Since the MNC-parent and the subsidiary are in different countries, the effects of not being in close proximity might mitigate the influence of audit industry specialization. Secondly, since subsidiaries are audited by separate auditors, asymmetric information between the auditor of the MNC-parent and the subsidiary auditor reduces monitoring effectiveness (Gul et al., 2018). Consequently, the effects of audit industry specialization are reduced by the information asymmetry. A third explanation might be that the subsidiary does not operate in the same industry as the MNC-parent, eliminating the effects of industry specific experience and knowledge.
This research contributes to the current literature in multiple ways. First of all, the current literature on the FRQ of MNCs are mostly only using the consolidated financial statements (Kim et al., 2001; Dyreng et al., 2012) despite the relevance of subsidiary reporting for the FRQ of MNCs. Therefore, by looking at not only the MNC-parent, but focusing on the financial reporting of the subsidiaries, this research expands the existing literature on the FRQ of MNCs. Secondly, the existing research on audit industry specialization and FRQ are mostly focused on stand-alone firms (Krishnan, 2003; Balsam et al., 2003; Knechel et al., 2007). Instead, this study focusses on the relation between audit industry specialization and FRQ in the context of MNCs. Therefore, this study contributes by bringing existing literature regarding the relation between audit industry specialization and FRQ into a new context. Thirdly, the results of this research add to the literature on REM, because the current research regarding REM in relation to audit industry specialization is not showing very consistent results (Anissa et al., 2019; Chi et al., 2011).
9
Therefore, even though this study does not find conclusive evidence, it does find some evidence of a negative relation between audit industry specialization and REM. Fourthly, previous studies have examined for example the location of EM within the MNC (Beuselinck et al., 2019), the influence of industrial and geographic diversification of subsidiaries on the earnings quality of the MNC-parent (Kim et al., 2001) or the difference between domestic and foreign subsidiaries (Durnev et al., 2017). By looking at audit industry specialization at the MNC-parent, this research explores a new determinant of EM in relation to MNCs.
The rest of the research is organized as follows: first the theoretical framework is explained, followed by the hypotheses development. Secondly the methodology is discussed. Thirdly the results of the research are presented. Lastly, these results are discussed and a conclusion is formed.
THEORY AND HYPOTHESES DEVELOPMENT
In this section the theory and the development of the hypotheses are discussed. First the reporting quality of MNCs is discussed using the agency theory to explain the motivations behind EM, including the discussion of the two types of EM: AEM and REM. This is followed by the relation between audit industry specialization and the FRQ of MNCs, ending this section with the development of the two hypotheses and a conceptual model showing the expected relations.
Reporting quality of MNCs
MNCs are defined as corporations that hold assets and conduct operations in more than one country (De Simone et al., 2020). They must provide a consolidated financial statement including the financial information of their subsidiaries (Beuselinck et al., 2019). A subsidiary is any operational unit controlled by the MNC and for the purpose of this research it is situated outside the home country (Corry & Cormican, 2019). The purpose of the consolidation process is to present the results and the financial position of a parent company and its subsidiaries, as if the group were a single entity (Beuselinck et al., 2019). Therefore, the MNC-parent must include the assets, liabilities, revenues and expenses of their subsidiaries into their own financial statements (Behn et al., 2020). This information is provided to the MNC-parent by each subsidiary, which has to set up their own financial statements.
10
To monitor the MNC as a whole, the external auditor of the MNC-parent looks at the consolidated financial statements (Docimo et al., 2020). According to the International Standard on Auditing (ISA) 600 the auditor is required to set individual materiality amounts for each subsidiary (IAASB, 2007). These amounts determine the effort that needs to be put into the audit for each subsidiary and therefore ultimately determines the quality of the consolidated financial statements (Stewart and Kinney, 2013). Since the subsidiaries often operate in different industries, cultures and jurisdictions, their audit is performed by different audit teams or firms than the parent company (Stewart and Kinney, 2013). The MNC-parent auditor evaluates a summary of the work the subsidiary auditor did, because it would not be efficient to examine all the actual work (Downey & Bedard, 2018). Based on this evaluation, it combines the subsidiary financial statements with the financial statements of the parent company.
The FRQ of MNCs and their subsidiaries is in this study measured by looking at EM. The agency theory can be used to explain the motives and incentives of managers to engage in EM. Agency theory addresses the problems that occur when one party (the principal or shareholders) delegates work to another party (the agent or managers) (Kaymak and Bektas, 2017). One of the agency problems is the information asymmetry that exists between these two parties, because the agents have more information regarding everyday operations than the principals do (Kaymak and Bektas, 2017). The second problem is that the principals and agents may have different goals or interests (Kaymak and Bektas, 2017). The combination of these two problems creates an opportunity for managers to follow their own interests above the interests of their shareholders (García‐Sánchez et al., 2020). Therefore, managers have the opportunity to manage their earnings to achieve their own goals above the goals of the shareholders (García‐Sánchez et al., 2020).
Regarding MNCs, the agency problem occurs between the monitors or shareholders of the parent company (principals) and the CEO or managers of the parent company (agents) (Beuselinck et al., 2019). Agency problems arise in this relationship whenever the CEOs’ interests are not aligned with those of the shareholders of the MNC. As mentioned before, the consolidated statement contains information about all the subsidiaries combined with the parent information. Therefore, the CEO or managers of a MNC-parent have the ability to not only manage the earnings of the parent company, but also those of the subsidiaries over which the MNC has sufficient power (Beuselinck et al., 2019). This results in an opportunity for the CEO of
11
the MNC-parent, because it could potentially use the foreign subsidiaries to avoid strong monitoring on the MNC-parent.
Thus, the agency theory explains why MNCs have an opportunity (i.e. information asymmetry) and incentive (i.e. different interests) to engage in EM. There are other explanations for MNCs to choose to manage earnings at subsidiary-level. First of all, monitoring of the consolidated statements is more difficult because of the complexity of the corporate structure. Greater organizational complexity limits the transparency and increases information asymmetry (Gul et al., 2018). The MNC-parent auditor only evaluates the work done by subsidiary auditors, which might make it more convenient for the MNC-parent to manage earnings at subsidiary level (Beuselinck et al., 2019). Secondly, when EM is detected at the subsidiary, the MNC-parent can (try to) deny any involvement in the matter (Dearborn, 2009). As explained by Dearborn (2009) companies might be able to separate themselves from harmful subsidiaries, even though they still gain profit from their activities. This means that the MNC-parent would not be accountable for the EM detected at the subsidiary. For the same reason, the consequences after getting caught may have less impact regarding financial or reputational damage when it occurs at subsidiary-level (Beuselinck et al., 2019; Dearborn, 2009).
There are two types of EM used to measure the FRQ of subsidiaries: AEM and REM. The first type, AEM, is the managers’ manipulation of discretionary accruals to alter the numbers in the financial statements (Jones, 1991). Accruals are the portion of the earnings that are not received yet (Dechow et al., 1995). However, not all accruals are manipulated because accruals also result from the normal course of business activities (Jones, 1991). Therefore, the abnormal component (i.e. discretionary accruals) is used as a measurement for FRQ. Previous research is mainly focused on AEM as a measure for FRQ, but research examining AEM in the context of MNCs is limited. One example of research on AEM is the study of Beuselinck et al. (2019). They examine the location of AEM within MNCs, using the financial statements of subsidiaries for their research. Their results show that the influence of parent companies over their subsidiaries is related to subsidiary integration and earnings management opportunities. Furthermore, they find that MNCs tend to use more AEM through subsidiaries located in low-quality institutional environments. A second research focused on EM within MNCs is the study of Dyreng et al. (2012). They examine whether the institutions and laws, including tax laws, in the locations of the subsidiaries, are associated with differences in EM in the consolidated statements of the entire
12
company. Their results show that (1) there is less EM when the firm has a high concentration of subsidiaries in foreign countries with a strong rule of law, (2) profitable firms with extensive tax haven subsidiaries engage in more EM and (3) EM appears to be more prevalent in domestic income than in foreign income. In their research Dyreng et al. (2012) mostly focus on AEM, but additionally use earnings restatements as an alternative measure of EM. A disadvantage of this measure is that for a restatement to occur the violation must be discovered (Dyreng et al., 2012). A third relevant research is the study of Durnev et al. (2017). They examine the FRQ of offshore firms by looking at AEM, accruals quality and earnings persistence. With offshore firms they mean subsidiaries (of MNCs) that are offshore financial centers (OFCs). For their research they used consolidated financial statements. The findings of Durnev et al. (2017) show that offshore firms use more AEM, have lower accruals quality and lower earnings persistence than non-offshore firms. Additionally, they find that MNCs with OFC subsidiaries have lower FRQ compared not only to US MNCs with non-OFC subsidiaries, but also to domestic firms.
The second type of EM, REM, is defined as managerial intervention by manipulating earnings through actions that depart from normal business practices (Anissa et al., 2019). It is motivated by managers’ desire to mislead stakeholders into believing certain financial reporting goals have been met (Roychowdhury, 2006). Certain real activity manipulations are possibly optimal actions in some circumstances. However, when managers engage in these activities more extensively than appropriate giving the circumstances, or even for the sole purpose of manipulating earnings, they are actually manipulating the real activities (i.e. REM) (Roychowdhury, 2006). Managerial interventions on earnings are hidden in normal business activities (Anissa et al., 2019). Therefore, it is more difficult for auditors to detect REM than it is to detect AEM. Additionally, REM does not refer to generally accepted accounting principles, and does not need to be disclosed in financial statements (Anissa et al., 2019). Examples of REM are overproduction to lower the cost of goods sold or price discounts to increase sales volumes. The research regarding REM in relation to MNCs is very limited. However, the previously mentioned study of Durnev et al. (2017) did not only examine AEM in offshore firms but also REM. Regarding REM they found that offshore firms are less likely to engage in REM than non-offshore firms, while they found the contrary for AEM. Additionally their results showed that offshore firms use REM to supplement AEM.
13
Durnev et al. (2017) are not the only study that finds that AEM and REM supplement each other. Previous research on both AEM and REM shows that the two types of EM work as substitutes (Cohen et al., 2008; Zang, 2012). Zang (2012) finds that when AEM is constrained, due to for example higher level of regulatory investigations or limited accounting flexibility, firms use more REM. Chi et al. (2011) explain that firms switch to REM when AEM is constrained, because REM does not involve direct violation of regulations. Additionally, Cohen et al. (2008) find that companies switch from AEM to REM because REM is harder to detect for an auditor.
Audit industry specialization
Besides explaining the motives and incentives for managers to engage in EM, the agency theory also explains that the external audit works as a mechanism for reducing agency costs and mitigating the possibility of self-serving reporting by management (Greenwood & Zhan, 2019). Therefore, audit improves the FRQ, because it mitigates the asymmetric information problem and acts as a monitoring mechanism for shareholders on managers (Shahzad et al., 2019). Since audit has a positive influence on FRQ, this research focusses on auditor industry specialization. Industry specialists are defined by Romanus et al., (2008) as auditors whose training and experience are largely concentrated in a particular industry. According to Owhoso et al. (2002), auditors with industry specialization are better able to detect errors in their industry of specialization than in other industries. The reason for this is that auditors with industry-specific knowledge possess a comprehensive understanding of a company’s characteristics, which improves their ability to detect errors (Romanus et al., 2008). Additionally, auditors who have a more comprehensive understanding of an industry's characteristics and trends will be more effective in auditing than auditors without such industry knowledge (Krishnan, 2003). According to Krishnan (2003) industry specialists also have the resources, industry specific-expertise and the incentives to detect and constrain EM and therefore enhance the earnings quality. To become industry specialist an investment in these resources and knowledge is necessary. Therefore, industry specialization is costly to develop (Balsam et al., 2003). However, once developed, the knowledge will increase the auditors’ ability to detect EM and minimize intentional errors (Balsam et al., 2003).
14
The relation between audit industry specialization and the FRQ of MNCs is very little researched in current literature. One example of a relevant research is the study of Gunn and Michas (2018), who argue that auditors with more experience possess expertise in performing global group audits in a way that positively affects audit quality. They find that auditor expertise is negatively associated with client restatements, indicating higher audit quality of MNCs. Since client restatements are also an indicator of earnings quality (Dechow et al., 2010), it can be expected that audit industry specialization not only has a positive influence on audit quality, but also on the earnings quality of MNCs. A second relevant study is the research of Downey and Bedard (2018), who examined the experiences of U.S. auditors in managing engagements of multinational entities. They use survey evidence to document the challenges associated with performing audits on MNCs and the strategies that experienced auditors use to overcome these challenges. They conclude that experience and knowledge (i.e. industry specialization) have a positive influence on resolving challenges within MNCs. Although this is not directly related to the FRQ of subsidiaries, it does again show a positive influence of audit industry specialization on the consolidated statements.
In their study, Choi et al. (2012) argue that geographic proximity of the auditor is associated with higher audit quality, because it enables informational advantages that help the auditor to develop client specific knowledge. This knowledge is important to auditors for planning the audits effectively, identifying audit risks and properly interpret audit evidence (Knechel et al. 2007). Therefore, geographic proximity leads to auditors being able to more effectively monitor client reporting behavior and constrain EM, improving the FRQ. When the MNC-parent CEO or managers are closely monitored, they have the opportunity to resort to their subsidiaries to engage in EM. Audit industry specialization at the MNC-parent increases the industry specific knowledge of auditors (Romanus et al., 2008), and therefore decreases the level of EM at the MNC-parent since it improves monitoring effectiveness. This results in a better FRQ of the parent financial statements, but the level of EM at the subsidiaries is likely to increase. The subsidiaries are monitored by the separate auditors, which are in close proximity to them. The subsidiary auditors are subject to the same laws, regulations and cultural norms as the subsidiary client and therefore have more client-specific knowledge (Downey & Bedard, 2018). When the auditor of the MNC-parent has the advantages of industry specialization, it is better
15
able to evaluate the work of the subsidiary auditor, which improves the FRQ of both the subsidiary and the MNC.
Previous studies show a very consistent relation between audit industry specialization and AEM. Both Krishnan (2003) and Balsam et al. (2003) find that audit industry specialization is associated with lower levels of AEM, because industry specialists have the expertise, resources and the incentive to constrain opportunistic reporting of accruals. This suggests that industry specialization might help mitigate the use of AEM tactics (Romanus et al., 2008).
Although there is no research available which specifically focusses on the relation between audit industry specialization of the MNC-parent and subsidiary AEM, it is expected they will be negatively related. First of all because the majority of previous studies show a negative relation between AEM and audit industry specialization, which is consistent with the argument that audit industry specialization enhances the overall reporting quality of the MNC. Secondly, audit industry specialization at the MNC-parent improves the monitoring of the MNC-parent. This is likely to cause the MNC-parent to resort to engaging in AEM at the subsidiary level. However the proximity of the subsidiary auditor to the subsidiary will improve the monitoring of the subsidiary and therefore decreases the level of subsidiary AEM. Therefore, audit industry specialization at the MNC-parent reduces AEM, leading to an improved FRQ, resulting in the following hypothesis:
H1: Auditor industry specialization at the MNC-parent increases the FRQ of subsidiaries, by decreasing their level of accrual based earnings management.
Since REM is more difficult for auditors to detect than AEM (Anissa et al., 2019), most studies focus on AEM. Still, there are several studies regarding the relation between REM and audit industry specialization, which have mixed results. For example, Anissa et al. (2019) find in their research that audit industry specialization has a negative relation with REM. They argue that specialists are expected to have deeper knowledge than non-specialists because of their expertise and experience in certain industries. Additionally, audit industry specialists have strong incentives to provide a higher quality audit since they want to protect themselves against reputational damage (Anissa et al., 2019). Contrary to Anissa et al. (2019), Chi et al. (2011) find in their research that managers switch from using AEM to using REM, when higher audit quality
16
constrains the use of AEM. Since audit industry specialization is a critical indicator of audit quality, Chi et al. (2011) find that audit industry specialization has a positive influence on the level of REM.
Regarding industry specialization of the MNC-parent auditor and subsidiary REM there is no previous research examining this specific relation. Additionally, previous studies are inconsistent regarding the sign of the relation between REM and audit industry specialization. As mentioned before, auditor industry specialization increases the audit quality and is therefore likely to reduce AEM at the MNC-parent. In this case the MNC-parent CEO is likely to resort to subsidiary AEM. Since several previous studies see AEM and REM as substitutes, the subsidiaries might switch to REM, when strong monitoring of the subsidiary also reduces subsidiary AEM. Therefore the expected relation between audit industry specialization of the MNC-parent and subsidiary REM is positive, leading to the following hypothesis:
H2: Auditor industry specialization at the MNC-parent decreases the FRQ of subsidiaries, by increasing their level of real earnings management.
17
METHODOLOGY
In this section the methodology to test the hypotheses is explained. First the sample selection is discussed. This is followed by the description of the independent variable (MNC-parent audit industry specialization) and the dependent variables (subsidiary AEM and REM). Hereafter follows the empirical model used for testing the hypotheses. This section ends with the discussion of the control variables.
Sample selection
First all listed firms from the US are identified using Compustat. From these firms, the financial firms are excluded from the sample because their financial environments differ from other industries and they have different accruals processes that are not likely to be captured by the models used in this research (Peasnell et al., 2000). Additionally, the firms with missing information regarding their external auditor are also excluded, because this information is necessary to determine audit industry specialization of the MNC-parent. For these listed non-financial firms the names and jurisdiction of all material subsidiaries included in exhibit 21 or 21.1 of 10-k filings are hand-collected from Edgar. This is followed by matching the names and countries of the material subsidiaries with those of the firms covered by the Orbis database. Given the coverage of Orbis, only privately held subsidiaries from 30 European countries are selected. Next, the subsidiaries operating in the financial industry are excluded for the same reason as the parent companies from the US. For the remaining subsidiaries, the financial information is collected from the Orbis database. Finally, after deleting firms with unavailable financial information, the final sample consists of 3954 subsidiaries-year observations.
Main independent variable - Parent auditor industry specialization
Because the specialist status of the auditor cannot directly be observed, prior research has used several proxies to estimate industry specialization (Balsam et al., 2003; Krishnan, 2003). Mostly, these measures are based on market share, because the assumption is that industry specialization is created through repetition and a large volume of business in an industry therefore indicates expertise (Balsam et al., 2003). Following Balsam et al. (2003) industry specialists are measured using auditor industry share based on the client sales in a specific industry. All the necessary information regarding MNC parents is collected from Compustat. The MNC-parents are classified into the industry in which they operate by using the standard industrial classification
18
codes (i.e. SIC codes). Appendix A shows a table explaining the SIC codes and the associated industries. Next, the MNC-parent sales and auditor code are collected for calculating the audit industry specialization. An overview of the auditor codes and their meaning is included in Appendix B. The unaudited companies are excluded from the sample, because this research looks at the difference between audit industry specialization and non-specialization, which means the companies included in the research must be audited. Additionally, firms with auditor code 9 (i.e. other) are directly coded as non-specialists because this code represents multiple smaller audit firms and therefore none of them is an industry specialist based on industry share.
Following Balsam et al. (2003), auditors are defined as industry specialists when they are: (1) the largest supplier in the industry and the difference between the first and second supplier in the industry is at least 10 percent, or (2) the largest suppliers in the industry between which the difference is less than 10 percent and the difference between them and the remaining auditors is more than 10 percent. Based on this distinction in sales, the auditor industry specialist is determined for the years 2012 to 2017, as shown in table 1.
With this information a dummy variable (Par_AIS) is created with a value of 1 if the MNC-parent of a subsidiary is audited by a specialist in that year and a value of 0 otherwise.
Table 1: Audit industry specialists per industry and year
Year / Industry 1 2 3 4 5 6 7 9 10 2012 Deloitte KPMG PWC KPMG PWC Deloitte EY Deloitte EY Deloitte PWC PWC 2013 Deloitte KPMG PWC KPMG PWC PWC Deloitte EY Deloitte EY Deloitte PWC PWC 2014 Deloitte KPMG PWC PWC PWC Deloitte EY Deloitte EY Deloitte PWC PWC 2015 Deloitte KPMG PWC PWC PWC Deloitte EY Deloitte EY Deloitte PWC Deloitte 2016 Deloitte KPMG PWC
PWC PWC Deloitte EY Deloitte EY Deloitte
2017 Deloitte KPMG PWC
PWC PWC Deloitte EY Deloitte EY Deloitte
This table shows the industry specialists according to their market share. The industry is classified using the 4 digit SIC codes as presented in appendix A.
19 Dependent variable – Subsidiary AEM
Subsidiary AEM is measured with the discretionary accruals estimated with the Modified Jones model. In their study, Dechow et al. (1995) compare five different models to measure AEM, namely the Healy model (1985), the DeAngelo model (1986), the Jones model (1991), the modified jones model and the industry model (Dechow and Sloan, 1991). Their results show that the modified version of the Jones model best measures AEM in comparison to the rest of the models. Based on these results and also prior research using this model (Balsam et al., 2003; Becker et al., 1998), the Modified Jones model is used to measure AEM in this research.
The Modified Jones model divides the total accruals into discretionary and non-discretionary accruals, where the former is a proxy for AEM. When a company has an excessive amount of discretional accruals, it can be a sign the accounting numbers are altered and therefore the reporting quality decreases. In order to estimate the discretionary accruals of the subsidiary, first the total accruals are computed using the following equation:
𝑇𝐴𝑐𝑐𝑡 = (∆𝐶𝐴𝑡− ∆𝐶𝐿𝑡− ∆𝐶𝑎𝑠ℎ𝑡+ ∆𝐷𝐶𝐿𝑡 − 𝐷𝑒𝑝𝑡) (1)
Where:
𝑇𝐴𝑐𝑐𝑡 = Total accruals in year t
∆𝐶𝐴𝑡 = Change in current assets between year t and t-1 ∆𝐶𝐿𝑡 = Change in current liabilities between year t and t-1
∆𝐶𝑎𝑠ℎ𝑡 = Change in cash and cash equivalents between year t and t-1
∆𝐷𝐶𝐿𝑡 = Change in debt included in current liabilities between year t and t-1 𝐷𝑒𝑝𝑡 = Depreciation and amortization expense in year t
Secondly, the modified jones model is estimated using the following equation which results in the estimated values of firm-specific parameters α1, α2 and α3:
𝑇𝐴𝑐𝑐𝑡 𝐴𝑡−1 = 𝛼1( 1 𝐴𝑡−1) + 𝛼2 ( ∆𝑅𝐸𝑉𝑡 𝐴𝑡−1) + 𝛼3 ( 𝑃𝑃𝐸𝑡 𝐴𝑡−1) + 𝜀𝑡 (2) Where:
20 𝐴𝑡−1 = Total assets at t-1
∆𝑅𝐸𝑉𝑡 = Change in revenues between year t and year t-1 𝑃𝑃𝐸𝑡 = Gross property, plant and equipment in year t
Finally, the non-discretional accruals are calculated using the estimated firm-specific parameters from model (2). Following the modified Jones model according to Dechow et al. (1995), this is done for each year and industry combination:
𝑁𝐷𝐴𝑡 𝐴𝑡−1 = 𝛼1 ( 1 𝐴𝑡−1) + 𝛼2 ( ∆𝑅𝐸𝑉𝑡− ∆𝑅𝐸𝐶𝑡 𝐴𝑡−1 ) + 𝛼3 ( 𝑃𝑃𝐸𝑡 𝐴𝑡−1) (3) Where:
𝑁𝐷𝐴𝑡 = Estimated non-discretionary accruals in year 1 𝐴𝑡−1 = Total assets in year t-1
∆𝑅𝐸𝑉𝑡 = Change in revenues between year t and year t-1 ∆𝑅𝐸𝐶𝑡 = Change in net receivables between year t and year t-1 𝑃𝑃𝐸𝑡 = Gross property, plant and equipment in year t
After calculating the non-discretionary part of the total accruals, the discretionary component (Sub_DACC) can be calculated, by subtracting the non-discretionary accruals from the total accruals. The first dependent variable is |Sub_DACC|, defined as the absolute value of the discretionary accruals, which is used to test for overall differences in EM rather than a specific sign.
Dependent variable – Subsidiary REM
Previous studies use different proxies for the intensity of subsidiary REM (Roychowdhury, 2006; Kim and Sohn, 2013). These proxies are developed by focusing on three methods of manipulating real operational activities:
1. Offering excessive sales discounts to temporarily boost sales revenues in the current period 2. Overproduction to report a lower cost of goods sold in the current period
21
Only the first two methods of manipulation are used in this research, because the financial information necessary to calculate the third method is not available for the entire sample. For the first method of manipulation the focus is on the subsidiary cash flow from operations (CFO). Following the methodology of Roychowdhury (2006), the actual CFO is divided into a normal (expected) part and an abnormal (unexpected) part. This division is calculated for each industry and year. After estimating the parameters using equation 4, the normal CFO is calculated using these estimations. Hereafter, the abnormal CFO of the subsidiary is calculated by subtracting the actual CFO of a subsidiary from the normal CFO. Finally, this is multiplied by negative one, so the higher the amount of abnormal CFO of the subsidiary (Sub_Abn_CFO), the more likely it is for that company to be engaged in REM.
𝐶𝐹𝑂𝑡 𝐴𝑡−1 = 𝛽1 1 𝐴𝑡−1+ 𝛽2 𝑆𝑡 𝐴𝑡−1+ 𝛽3 ∆𝑆𝑡 𝐴𝑡−1+ 𝜀𝑡 (4) Where:
𝐶𝐹𝑂𝑡 = Cash flow from operations in year t 𝐴𝑡−1 = Total assets in year t-1
𝑆𝑡 = Total sales in year t
∆𝑆𝑡 = Change in sales between year t and year t-1
For the second method of manipulation (i.e. overproduction) a smaller sample is used because the financial information regarding the production costs of the subsidiaries was not available for the entire sample. For the estimation with the second measure of REM (i.e. overproduction) the actual production costs have to be divided into normal and abnormal production costs. Similar to the abnormal CFO, first the normal production costs are estimated based on the parameters from equation (5). Hereafter, the subsidiary abnormal production costs (Sub_Abn_PC) are calculated by subtracting the actual production costs from the normal production costs. A higher amount of abnormal production costs of the subsidiary indicates that is is more likely that this subsidiary is engaged in REM. 𝑃𝐶𝑡 𝐴𝑡−1= 𝛽1 1 𝐴𝑡−1+ 𝛽2 𝑆𝑡 𝐴𝑡−1+ 𝛽3 ∆𝑆𝑡 𝐴𝑡−1+ ∆𝑆𝑡−1 𝐴𝑡−1 + 𝜀𝑡 (5)
22 Where:
𝑃𝐶𝑡 = Production costs in year t
∆𝑆𝑡−1 = Change in sales between year t-1 and year t-2 The rest of the variables have the same meaning as in equation (4).
The third form of manipulation (i.e. reducing discretionary expenses) is not included in this research, because the necessary financial information on subsidiary discretionary expenses, such as R&D expenses or other expenses, is not available from the Orbis database.
The empirical model
The following model is used to test the hypotheses:
𝑆𝑢𝑏_𝐸𝑀𝑖,𝑡 = 𝛽0+ 𝛽1 𝑃𝑎𝑟_𝐴𝐼𝑆𝑖,𝑡+ 𝛽2 𝑆𝑢𝑏_𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛽3 𝑆𝑢𝑏_𝑅𝑂𝐴𝑖,𝑡+ 𝛽4 𝑆𝑢𝑏_𝐿𝐸𝑉𝑖,𝑡 + 𝛽5 𝑆𝑢𝑏_𝐿𝑂𝑆𝑆𝑖,𝑡+ 𝛽6 𝑆𝑢𝑏_𝑌𝐸𝐴𝑅𝑖,𝑡+ 𝛽7 𝑆𝑢𝑏_𝐼𝑁𝐷𝑖,𝑡+ 𝛽8 𝑆𝑢𝑏_𝐶𝑂𝑈 𝑖,𝑡+ 𝜀𝑖,𝑡 (6)
The dependent variable Sub_EM stands for either the amount of |Sub_DACC|, Sub_Abn_CFO or
Sub_Abn_PC. The main independent variable is Par_AIS, an indicator variable for auditor
industry specialist at the MNC-parent. It is expected that the coefficient of Par_AIS is negative in relation to |Sub_DACC| and positive in relation to both Sub_Abn_CFO and Sub_Abn_PC.
Control variables
Based on prior studies several control variables are included in this research. First of all, according to Siregar and Utama (2008) larger firms engage in less earnings management than smaller firms because they are more easily scrutinized by investors or regulators than smaller firms. Since Beuselinck et al. (2019) find evidence that smaller subsidiaries have higher values of EM, the size of the subsidiary (Sub_SIZE) is included as control variable. It is calculated following Beuselinck et al. (2019) as the log of subsidiary total assets. Secondly, subsidiary performance is included as a control, since previous research suggests that better performing companies are less likely to engage in EM to reach their objectives (Khani et al., 2019). Additionally, Beuselinck et al. (2019) find evidence that subsidiary performance, measured by subsidiary return on assets (ROA), is negatively related to EM. Therefore, following Beuselinck et al. (2019), the control variable Sub_ROA is included to control for subsidiary. Furthermore,
23
according to Becker et al. (1998) managers of highly leveraged firms have more incentives to use EM to avoid debt covenant violation. Therefore the third control variable is subsidiary leverage (Sub_LEV), which is calculated by dividing long term debt by total assets. Fourthly, Beuselinck et al. (2019) find that subsidiaries that are facing losses are associated with more EM, because they want to avoid or reduce these losses. Therefore subsidiary loss (Sub_LOSS) is another control variable, which is defined as 1 if the subsidiary made a loss that year and 0 otherwise. Since the population of this study covers multiple years, industries and countries, the model includes dummy variables for year effects (Sub_YEAR), industry effects (Sub_IND) and country effects (Sub_COU). All information regarding the dependent, independent and control variables is shown in table 2. For the regression analyses, all continuous variables are winsorized at the 1st and 99th percentiles of their distribution.
Table 2: Variable definitions
Variables Definitions
Par_AIS = Audit Industry Specialization of the parent - dummy variable with value 1
for audit industry specialists and value 0 for non-specialists
|Sub_DACC| = absolute value of the subsidiary discretionary accruals, estimated with the modified Jones model.
Sub_Abn_CFO = subsidiary abnormal CFO, calculated as abnormal cash flow from
operations following the model of Roychowdhury (2006)
Sub_Abn_PC = subsidiary abnormal production costs, calculated following the model of Roychowdhury (2006)
Sub_SIZE = the logarithm of subsidiary total assets
Sub_ROA = subsidiary return on assets calculated as net income divided by lagged total
assets
Sub_LEV = subsidiary long term debt divided by total assets
Sub_LOSS = dummy variable with value 1 if there is a loss at the subsidiary that year and
0 otherwise
Sub_YEAR = year dummies for the period 2012 to 2017
Sub_IND = subsidiary industry dummies indicating SIC codes
24 RESULTS
This section starts with the descriptive statistics, followed by the correlations between all the variables. Hereafter the results of the regression will be presented.
Descriptive statistics
The descriptive statistics of the variables are shown in table 3. This table shows that, on average, subsidiary discretionary accruals represents 14 percent of its total assets and has a standard deviation of 0.172. Both values are similar to the study of Beuselinck et al. (2019) which are 13 percent and 0.166, respectively. The subsidiary abnormal CFO has an average of -0.004 and standard deviation of 0.210. The standard deviation of abnormal CFO is higher than the standard deviations in similar studies. For instance, Chi et al. (2011) and Kim and Sohn (2013), who look at standalone companies, have standard deviations of 0.116 and 0.069, respectively. This means that the amount of abnormal CFO of this research is more spread out across the sample than in the previously mentioned studies. The sample used to calculate subsidiary abnormal production costs is 2636, which is roughly two third of the final sample. The average is -0.002 and the standard deviation 0.602. Again, the standard deviation of Chi et al. (2011) is lower (0.210), which shows that for this measure of REM the amounts are also spread out more. Table 3 also shows that audit industry specialization at the MNC-parent has a mean of 0.350 and a standard deviation of 0.478. This means that approximately one third of the sample has a MNC-parent which is audited by an industry specialist. The descriptions of the control variables are similar to those reported in other relevant studies such as Chi et al. (2011) and Beuselinck et al. (2019).
Table 3: Descriptive statistics
Variable N Mean Std. dev. P25 Median P75
|Sub_DACC| 3954 0.144 0.172 0.038 0.087 0.181 Sub_Abn_CFO 3954 -0.004 0.210 -0.091 -0.005 0.080 Sub_Abn_PC 2636 -0.002 0.602 -0.227 0.079 0.279 Par_AIS 3954 0.350 0.478 0 0 1 Sub_SIZE 3954 9.683 `1.768 8.438 9.682 10.900 Sub_ROA 3954 0.052 0.117 0.015 0.050 0.098 Sub_LEV 3954 0.007 0.045 0.000 0.000 0.000 Sub_LOSS 3954 0.170 0.372 0 0 0
This table contains descriptive statistics for all the variables used in the analyses. All the variables are defined in table 1.
25 Correlations
In table 4 the correlations between the variables are presented. The table shows that audit industry specialization at the MNC-parent is not significantly correlated with subsidiary AEM or subsidiary abnormal CFO, which could indicate that there is no relation between them. However, audit industry specialization is significantly related to the abnormal production costs of the subsidiary. This negative correlation could mean that a subsidiary with a parent that is audited by an industry specialist has a lower level of abnormal production costs than a subsidiary with a parent company audited by a non-specialist. Furthermore, the absolute value of subsidiary discretionary accruals is not related to subsidiary abnormal CFO, while it is significantly related to subsidiary abnormal production costs. Therefore, not much can be said regarding the relation between subsidiary AEM and REM. The two measures of REM are positively correlated, which indicates that subsidiaries with a higher amount of abnormal CFO are likely to also have a higher amount of abnormal production costs. Furthermore, table 4 shows that the amount of subsidiary AEM is higher for subsidiaries that are smaller, have a lower ROA or are making a loss. The magnitude of abnormal CFO is higher for subsidiaries that have a lower ROA, are more leveraged or are making a loss. The amount of abnormal production costs is higher for subsidiaries that are bigger, have a lower ROA, are more leveraged and are making profit. After testing for it, using the VIF method, the results show that there is no multicollinearity between the independent variables.
Table 4: Correlation matrix
1 2 3 4 5 6 7 8 1 |Sub_DACC| 1** 2 Sub_Abn_CFO .007** 1** 3 Sub_Abn_PC -.081** .089** 1 4 Par_AIS .027** .004** -.063** 1** 5 Sub_SIZE -.116** -.005** .106** .003** 1** 6 Sub_ROA -.119** -.434** -.066** -.038** .035** 1** 7 Sub_LEV .025** .033** .048** .002** .100** -.071** 1** 8 Sub_LOSS .118** .256** -.256** .053** -.081** -.634** .036** 1** * , ** Denote statistical significance at the 5 percent and 1 percent levels, respectively. The sample consists of 3954 observations. All variables are defined in table 1.
26 Regression analysis
The results of the estimation where the dependent variable is the level of subsidiary discretionary accruals are reported in table 5. The results in table 5 show that MNC-parent audit industry specialization is not significantly related with subsidiary discretionary accruals. Therefore, these results do not support hypothesis 1, since there is no evidence of the expected relation between audit industry specialization of the MNC-parent and subsidiary AEM. The results are consistent when including subsidiary abnormal CFO and abnormal production costs as additional controls to account for the substitution effect of AEM and REM. Moreover, the coefficients of the control variables are similar in all estimations. Furthermore, subsidiary discretionary accruals decrease with its size and performance and increase with the likelihood of reporting a loss.The adjusted R square shows that in all estimations approximately 5 percent of the variability of the discretionary accruals of subsidiaries is explained by the independent variables.
Table 5: Regression analysis – the effect of parent audit industry specialization on subsidiary discretionary accruals
Dependent variable: |Sub_DACC|
Independent variables (1) (2) (3) Par_AIS 0.000 0.001 (0.039) (0.129) Sub_SIZE -0.008** -0.008** -0.006** (-4.559) (-4.555) (-2.859) Sub_ROA -0.085** -0.085** -0.133** (-2.852) (-2.852) (-3.291) Sub_LEV 0.118 0.118 0.087 (1.896) (1.896) (1.105) Sub_LOSS 0.033** 0.033** 0.033** (3.541) (3.534) (2.881) Sub_Abn_CFO -0.055** (-2.976) Sub_Abn_PC -0.008 (-1.205) Observations 3954 3954 3954 R2 0.061 0.061 0.072 Adjusted R2 0.052 0.052 0.058
This table presents the results of the estimation of model 6 when the dependent variable is the absolute value of subsidiary discretionary accruals. Column (1) shows the results of the model including only the control variables. Column (2) shows the results of the full model. In column (3), the main model includes subsidiary abnormal CFO and abnormal production costs as an additional control, since AEM and REM are likely to be substitutes. Year-, industry- and country fixed effects are included, but the coefficients are not reported. All variables are defined in table 1. * , ** shows statistical significance at the 5 percent and 1 percent levels, respectively.
27
The results of the estimation where the dependent variable is the level of subsidiary abnormal CFO are shown in table 6. This table shows that audit industry specialization of the MNC-parent has no significant relation with subsidiary abnormal CFO. Consequently, hypothesis 2 is not supported by these results, since there is no evidence of the expected association between MNC-parent audit industry specialization and abnormal CFO of the subsidiary. These results are not in line with the current literature, since they mostly find a significant relation, disregarding the sign, between audit industry specialization and REM (Anissa et al., 2019; Chi et al., 2011). Furthermore, table 6 shows that the magnitude of subsidiary abnormal CFO decreases when the subsidiary is performing better. The results are consistent when the absolute value of discretionary accruals is included as a control for substitution between subsidiary AEM and REM. Additionally, the coefficients of the control variables stay roughly the same in all estimations. Approximately 68 percent of variability of abnormal CFO of subsidiaries is explained by the independent variables. This is higher than similar studies, such as Anissa et al. (2019) which has an adjusted R square of 0.151.
Table 6: Regression analysis - the effect of parent audit industry specialization on subsidiary abnormal CFO
Dependent variable: Sub_Abn_CFO
(1) (2) (3) Par_AIS 0.001 0.001 (0.122) (0.123) Sub_SIZE -0.002 -0.002 -0.003 (-1.144) (-1.149) (-1.324) Sub_ROA -0.824** -0.824** -0.828** (-24.530) (-24.527) (-24.629) Sub_LEV 0.011 0.011 0.016 (0.152) (0.151) (0.225) Sub_LOSS -0.017 -0.017 -0.016 (-1.635) (-1.639) (-1.500) |Sub_DACC| -0.044** (-2.442) Observations 3954 3954 3954 R2 0.205 0.205 0.206 Adjusted R2 0.197 0.196 0.198
This table presents the results of the estimation of model 6 when the dependent variable is subsidiary abnormal CFO. Column (1) shows the results of the model including only the control variables. Column (2) shows the results of the full model. In column (3), the main model includes the absolute value of subsidiary discretionary accruals as an additional control, since AEM and REM are likely to be substitutes. Year-, industry- and country fixed effects are included, but the coefficients are not reported. All variables are defined in table 1. * , ** shows statistical significance at the 5 percent and 1 percent levels, respectively.
28
The results of the estimation where the dependent variable is the level of subsidiary abnormal production costs are shown in table 7. The results of table 7 show that the amount of subsidiary abnormal production costs is lower for subsidiaries of a MNC-parent that is audited by an industry specialist. Therefore this result contradicts hypothesis 2, because a negative relation is found instead of the expected positive relation. The level of subsidiary abnormal production costs increases with the size of the subsidiary and decreases with its performance. The results are consistent when controlled for substitution effects between AEM and REM, and the coefficients of the control variables are similar in all estimations. Approximately one third of the variability of abnormal production costs of subsidiaries is explained by the independent variables.
Table 7: Regression analysis - the effect of parent audit industry specialization on subsidiary abnormal production costs
Dependent variable: Sub_Abn_PC
(1) (2) (3) Par_AIS -0.048* -0.048* (-2.336) (-2.332) Sub_SIZE 0.031** 0.032** 0.031** (4.806) (4.937) (4.856) Sub_ROA -0.422** -0.421** -0.427** (-3.898) (-3.891) (-3.947) Sub_LEV 0.312 0.312 0.318 (1.371) (1.370) (1.396) Sub_LOSS 0.005 0.010 0.013 (0.165) (0.309) (0.386) |Sub_DACC| -0.076 (-1.333) Observations 2636 2636 2636 R2 0.333 0.335 0.335 Adjusted R2 0.324 0.325 0.325
This table presents the results of the estimation of model 6 when the dependent variable is subsidiary abnormal production costs. Column (1) shows the results of the model including only the control variables. Column (2) shows the results of the full model. In column (3), the main model includes the absolute value of subsidiary discretionary accruals as an additional control, since AEM and REM are likely to be substitutes. Year-, industry- and country fixed effects are included, but the coefficients are not reported. All variables are defined in table 1. * , ** shows statistical significance at the 5 percent and 1 percent levels, respectively.
29
DISCUSSION AND CONCLUSION
In this last section of the research, the findings will be discussed based on the results of the previous section. This is followed by the implications of the research and lastly the limitations will be discussed.
Findings
According to the agency theory, managers have the opportunity to manage earnings because there is asymmetric information between the managers and the shareholders of a company (Beuselinck et al., 2019; Cuervo‐Cazurra et al., 2019). For MNCs this asymmetric information exists between the CEO of the MNC-parent and the shareholders of the MNC (Beuselinck et al., 2019; Cuervo‐ Cazurra et al., 2019). Based on this theory, the CEO of MNC has incentives to manage their earnings when their interests are not aligned with those of the shareholders of the MNC (Beuselinck et al., 2019; Cuervo‐Cazurra et al., 2019). According to the agency theory external audit works as a mechanism to reduce information asymmetry and to mitigate the possibility of self-serving reporting by management (Greenwood & Zhan, 2019; Shahzad et al., 2019). Audit industry specialization reduces and constrains EM, since audit industry specialists have a more comprehensive understanding of an industry's characteristics and have the resources to be more effective in auditing than auditors without such industry knowledge (Krishnan, 2003; Romanus et al., 2008). Previous studies show that audit industry specialization is associated with lower levels of AEM, because industry specialists have the expertise, resources and the incentive to constrain opportunistic reporting of accruals (Krishnan, 2003; Balsam et al., 2003). There is no research yet regarding the relation between audit industry specialization at the MNC-parent and subsidiary AEM. However, Gunn and Michas (2018), suggest that auditors with more experience in auditing MNCs possess expertise in performing global group audits in a way that positively affects audit quality. Additionally, Downey and Bedard (2018) conclude that industry specialization has a positive influence on resolving challenges within MNCs. Furthermore, geographic proximity causes auditors to more effectively monitor client reporting behavior, constrain EM and improve the FRQ, because they have more client specific knowledge (Choi et al., 2012). Since subsidiaries are audited by separate auditors than the parent company, this auditor has more client-specific knowledge because of the proximity to the subsidiary. Therefore, subsidiary AEM decreases because audit industry specialization at the MNC-parent improves the evaluation of the
30
work of the subsidiary auditor. Thus, it is expected that the association between subsidiary AEM and audit industry specialization of the MNC-parent is negative.
Regarding the relation between audit industry specialization and REM studies vary in their findings. Anissa et al. (2019) find evidence that industry specialists have deeper client knowledge than non-specialists and stronger incentives to provide higher quality audits to protect themselves against reputational damage. On the contrary, Chi et al. (2011) suggest that managers switch from using AEM to using REM, when higher audit quality constrains the use of AEM and therefore find a negative relation. It is expected that subsidiaries might switch from using AEM to REM, when higher audit quality, caused by auditor industry specialization at the MNC-parent, constrains the use of AEM at both MNC-parent level as well as subsidiary level. Therefore, the expected association between audit industry specialization at the MNC-parent and subsidiary REM is positive.
The results suggest that industry specialization of the MNC-parent auditor is not related to the level of subsidiary AEM or subsidiary abnormal CFO. However, the results do show that the level of abnormal production costs is lower for subsidiaries that have a parent company that is audited by an industry specialist. This second finding is not as expected but is similar to the findings of Anissa et al. (2019). According to their study REM can be detected better by auditors who have special expertise in certain industries because REM is directly related to a company's business transactions and company cash flow. Therefore, an industry specialist at the MNC-parent might have enough experience and expertise to be able to reduce subsidiary REM. Additionally, the results show no evidence of audit industry specialization reducing subsidiary AEM. Therefore, the reasoning of Chi et al. (2011) regarding companies to switch from AEM to REM when monitored closely does not apply here. Therefore, the expected increase in subsidiary REM, because of audit industry specialization, is actually a decrease.
There could be several reasons the results did not show any relation between industry specialization of the MNC-parent auditor and the level of subsidiary AEM or subsidiary abnormal CFO. The first reason could be found in the research of Choi et al. (2012). Their results show that geographic proximity helps auditors to develop better knowledge about client-specific incentives and abilities, improving their effectiveness in monitoring client reporting behavior and constraining biased financial reporting (i.e. EM). Thus, Choi et al. (2012) argue that geographic proximity mitigates information asymmetries and enhances monitoring effectiveness. Since this
