Subsidiary financial reporting quality, geographic diversification and analyst forecast accuracy: Evidence from European private subsidiaries

(1)

1

Subsidiary financial reporting quality, geographic

diversification and analyst forecast accuracy: Evidence

from European private subsidiaries

Master thesis Accountancy

University of Groningen, Faculty of Economics and Business

June 22, 2020

Charlotte van den Hurk

Student number: 2771187

Folkingestraat 7a

9711 JS Groningen

(06) 34499323

c.b.m.van.den.hurk@student.rug.nl

Supervisor: S. Rusanescu, PhD

Word count: 11.889

(2)

2

Abstract

This paper examines the relation between the financial reporting quality of foreign European subsidiaries and the forecast accuracy of the analysts who are following their U.S. parents. Prior literature suggests that multinationals manage their earnings through subsidiaries. Next, researchers found that low financial reporting quality is associated with less accurate forecasts. Therefore, this study predicts that the financial reporting quality at the level of the subsidiary is positively associated with the forecast accuracy of the analysts. This research further investigates the influence of geographic diversification on this relation. Geographic diversification leads to an increase in information asymmetry and therefore I predict that it strengthen this relation. Using data from I/B/E/S for the forecast accuracy, COMPUSTAT for data of the parent company and Orbis for data of the European subsidiaries, I find that (i) financial reporting quality of the subsidiaries does not affect the forecast accuracy and (ii) geographic diversification does not influence the relation between the financial reporting quality of the subsidiaries and the forecast accuracy of the analysts. This implies that the financial reporting quality of the subsidiary does not influence the reporting quality of the whole corporate group and therefore not affect the forecast accuracy.

Keywords: Forecast accuracy, Financial reporting quality, Earnings management,

(3)

3

In the last few years, research regarding accounting has increasingly paid more attention to the financial reporting quality within corporations (Beuselinck et al., 2019; Bonacchi et al., 2018; Dyreng et al., 2012). A report of the OECD in 2018 stated that multinationals and their foreign affiliates produce almost one-third of the global production. Due to the economic importance, manipulation in the financial statements of the MNCs in for instance the earnings can result in immense consequences for the economy and the society. Therefore, a better understanding of the whole corporate group is important since the financial reporting of the foreign subsidiaries is possibly influenced by the parent company of the multinational (Beuselinck et al., 2019). Different firms (e.g. Enron and Parmalat), have shown this in the past by publishing consolidated financial statements that did not reflect the real economic position of the firm. Enron used subsidiaries to hide debt off the balance sheet and disclose minimal details on these subsidiaries. Therefore, the liabilities were understated and the equity, as well as the earnings, were overstated. The earnings in the consolidated financial statement were manipulated and did not show the real economic performance. Financial analysts, who made forecasts about the value of Enron, used these earnings for their forecast. Therefore, analysts were not able to accurately forecast the future performance of the firm, which affected the investors’ investment decisions. Considering the MNCs’ economic impact, Beuselinck et al. (2019, p. 45) stated that ‘‘financial reporting practices of MNCs warrant

careful attention.’’

According to La porta et al. (1999), MNCs often have a very complex corporate structure, which consists of domestic and foreign subsidiaries. Over the last years, MNCs increasingly expanded their operations to foreign countries (Chin et al., 2009). According to a report of the United Nations Conference on Trade and Development in 2014, the top of the largest MNCs held 70 percent of their assets abroad. Expanding operations abroad leads to multiple advantages, for instance, cost savings, access to other resources, tax reduction, and growth opportunities (Bodnar & Weintrop, 1997; Daniels & Bracker, 1989). This increases the value of the firm and thus for the shareholders. However, there is also a downfall since an increase in foreign operations leads to an increase in firm complexity (Callen et al., 2005). A complex firm environment results in more information asymmetry between the managers and the investors. Managers could exploit this, to employ activities for their self-interest (Chin et al., 2009). An increase in information asymmetry creates flexibility for managers to engage in

(5)

5 for instance more earnings management. Managers make use of earnings management for different reasons, for instance, to meet certain earnings requirements, stated in their contracts, to achieve bonuses (Dye, 1988). According to Dye (1988), an increase in information asymmetry results in a higher degree of earnings management. According to Healy and Wahlen (1999), Earnings management is defined as ‘Earnings management occurs when

managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.’

According to Beuselinck et al. (2019), MNCs use their subsidiaries to conduct earnings management. MNCs use their subsidiaries to conduct these risky activities because when it is discovered, the reputational damage is lower at the level of the subsidiary than at the parent level (Dearborn, 2009). Furthermore, according to Fan (2012), MNCs manage their earnings through the subsidiaries, to avoid reporting a loss in the consolidated financial statement. Meeting or beating the benchmark is another incentive for managers to manage the earnings through the subsidiaries (Bonacchi et al., 2018). The costs of not meeting these benchmarks are high, it leads for instance to a large decline in the stock price (Brown, 1993; Brown, 2001; Skinner, 1994). The parent company can use their subsidiaries because an increase in the earnings at the level of the subsidiary directly translate to an increase in the earnings in the consolidated financial statement.

Prior research also examines drivers of earnings management through subsidiaries, for instance, MNCs manage more earnings if their subsidiaries are established in tax havens (Prencipe, 2012). Dyreng et al. (2012) concluded that this is not due to the reducing tax expense since they included the earnings management before taxes in their research, but rather the result of the little or no local tax costs which comes from the earnings that are managed before the tax is determined. Further, earnings management increases extensively if the quality of the shareholders’ rights in a country is low (Leuz et al., 2003). In line are the findings of Dyreng et al. (2012), which concluded that U.S. MNCs manage their consolidated earnings more if they have large foreign operations in weak rule of law countries. Foreign operations in these weak rule of law countries are less confronted by the authorities with earnings management, and if so, the consequences are less harsh than in a strong rule of law country (Dyreng et al., 2012). Finally, it is more difficult for the auditors of the parent

(6)

6 company to audit the financial statement of the subsidiary and thus to detect earnings management (Stewart & Kinney, 2013).

Earnings are considered as the most important item in the financial statements by analysts and investors (Degeorge et al., 1999). When the earnings have been managed, the financial statement does not give a true image of the value of the firm, which increases the information asymmetry (Dye, 1988). A lower financial reporting quality increases both over-and under-investment problems for the investors (Biddle et al., 2009). The analysts, therefore, have an important role as intermediaries between the firm and the investors (Clement, 1999). However, this information asymmetry also impacts the forecast of the analysts. Due to the manipulations in the earnings, analysts are unable to accurately predict future earnings (Burgstahler and Eames, 2003; Salerno, 2014).

Prior studies (e.g. Chin et al., 2009; Degeorge et al., 1999; Salerno, 2014) examine the relationship between earnings management and the forecast accuracy of the analysts. However, these studies investigate earnings management at the firm level. Bonacchi et al. (2018) stated that only assessing the consolidated financial statement is not sufficient since the financial reporting of the subsidiaries is possibly influenced by the parent. According to Dyreng et al. (2012), there is a need for further research relating to earnings management within the MNC. This research is necessary because the earnings of the subsidiaries are included in the consolidated financial statement and therefore influence the decisions of the investors, through the forecasts of the analysts. Accordingly, the research question of this paper is guided by the following research question:

RQ: How does the FRQ of the subsidiaries influence the forecast accuracy of the analysts

who are following the MNC?

Beuselinck et al. (2019), Bonacchi et al. (2018), and Dyreng et al. (2012) found that MNCs manage their earnings through their subsidiaries. It is therefore expected that the financial reporting quality of the subsidiaries influences the financial reporting quality of the corporate group as a whole and as a consequence influence the forecast accuracy of the analysts who are following the MNC-parent. Considering this, I predict a positive relationship between the FRQ of the subsidiaries and the forecast accuracy of the analysts who are following the MNC. Further, this research attempts to examine the role of geographic diversification on the relation between the financial reporting quality of the subsidiaries and the forecast accuracy of

(7)

7 the analysts. As stated by Chin et al. (2009), a more geographic diversified firm, leads to an increase in complexity of the corporation. This creates more opportunities for managers of MNCs to conduct earnings management because the corporate group consists of more subsidiaries, which are located in different countries (Chin et al., 2009). MNCs tend to manage their earnings more in foreign subsidiaries than in domestic subsidiaries, for instance, to avoid reporting a loss in the consolidated financial statement (Beuselinck et al., 2019; Fan, 2012). Furthermore, Duru and Reeb (2002) concluded that a more diversified firm increases the complexity of the firm, which makes it difficult for analysts to predict the value of the firm. This implies that geographic diversification leads to less accurate forecasts of the analysts (Duru and Reeb, 2002). This study would like to explore if there is also a moderating effect of geographic diversification on this relation between FRQ of the subsidiaries and the forecast accuracy of the analysts. Taking into account the findings of prior studies, I predict that geographic diversification strengthens the relationship between a low financial reporting quality of the subsidiary and a less accurate forecast of the analysts who are following the MNC.

This study is based on 219 U.S. multinationals including their 7,189 European subsidiaries, from 2011 until 2017. The data collection of the list of material subsidiaries including their jurisdiction is hand-collected out of the Edgar database. Furthermore, the financial data regarding the parent are obtained by Compustat and the data regarding the forecasts of the analysts are gathered from the I/B/E/S database. While the data of the subsidiaries are obtained through the Orbis database. I use the ordinary least square (OLS) to test my hypotheses. Additional tests show the robustness of my results. My findings suggest that the forecast accuracy of the analysts who are following the MNCs is not influenced by the FRQ of the subsidiaries. Contrary to my predictions, I also find that geographic diversification does not influence this relation.

My study adds to the growing literature on how the MNCs manage their earnings from within. This paper contributes to four strands of literature. First, different studies (Beuselinck et al., 2019; Durnev et al., 2017 Dyreng et al., 2012) researched earnings management within the multinationals. They documented that MNC-parents tend to manage earnings through their subsidiaries. However, I complement these findings by including the effects of earnings management at the subsidiary level on the forecast accuracy of the analysts who are following the MNC. This has economic importance, since the manipulated earnings at the subsidiary

(8)

8 level are included in the consolidated financial statement, and affect the investors’ decisions, through the forecasts of the analysts.

This paper contributes to the second strand of literature by adding how earnings management affects forecast accuracy. Previous studies that researched the relation between earnings management and forecast accuracy like Chin et al. (2009), Degeorge et al. (1999), and Salerno (2014), focussed on the financial statements of stand-alone firms. These studies found that if earnings management is higher, the forecast accuracy of the analysts is lower. However, this study focus on the FRQ of the subsidiary of the MNCs rather than the prior studies which focussed on stand-alone firms. This is important since the financial reporting quality of the subsidiary can influence the financial statement of the corporate group as a whole and therefore affect the forecast accuracy of the analysts. Contrary to the findings of prior studies, I find no evidence of this relation.

Furthermore, prior studies concluded that firms tend to manage their earnings away from the enforcers (Ayers et al., 2011; Choi et al., 2012). In line are the studies of Beuselinck et al. (2019) and Fan (2012), which stated that MNCs tend to manage earnings more in foreign subsidiaries than in domestic earnings, for instance, to avoid losses in the consolidated financial statement. Moreover, Duru and Reeb (2002), investigated corporations that are internationally diversified and concluded that a more geographically diversified firm results in forecasts that are less accurate made by analysts. This paper contributes by examining if there is also a moderating effect of geographic diversification on the relation between financial reporting quality and forecast accuracy, which is also studied in this research. However, this study did not find evidence that geographic diversification affects the relation between the FRQ of the subsidiary and the forecast accuracy of the analysts who are following the MNC. Finally, the findings of my study provide more insight for European regulators on accounting standards regarding how the FRQ at the subsidiary level influence the forecast accuracy of the analysts who are following the MNC. This is important since one of the goals of accounting standards is to provide insight for investors, which is obtained through the forecasts of the analysts. This paper finds that the FRQ of the subsidiaries does not affect the forecast accuracy and geographic diversification does not have a moderating effect.

The remainder of the paper is organized as follows. Section 2 revise the literature and develop the hypotheses. In Section 3, I describe the research design. The results are presented in

(9)

9 Section 4. And in section 5, I conclude, point out limitations and make suggestions for future research.

2 Literature review and hypotheses development

Literature review

Financial reporting quality of MNCs

According to Herath and Albarqi (2017), several regulators, such as the International Accounting Standards Board (IASB) and the Financial Accounting Standards Board (FASB), define financial reporting quality as ‘‘financial reporting quality represents financial

statements that provide accurate and fair information about the underlying financial position and economic performance of an entity’’. The objective of financial reporting is to attract and

to inform potential investors and other creditors (Chen et al., 2011). The higher the financial reporting quality of the firm, the more benefits are gained by users of the financial reports, such as investors, creditors, and analysts (Herath & Albarqi, 2017).

Prior studies concluded that a higher financial reporting quality leads to better investment decisions by investors, results in market efficiency, and reduce information asymmetry between the managers and the investors (Chen et al., 2011; Herath & Albarqi, 2017; Jung et al., 2014). In contradiction to a lower financial reporting quality which creates problems regarding both over- and under-investment by investors (Biddle et al., 2009).

Next to this, Burgstahler and Eames (2003) stated that analysts take into account the financial reporting quality of a firm when making forecasts about future performance. Salerno (2014) finds that higher financial reporting quality positively influences the forecast accuracy of the analysts. This implies that analysts are better able to forecast the future performance of the firm if the firm has a high financial reporting quality.

In the literature, earnings management is used as a proxy to measure financial reporting quality. Prior research established three general earnings management tools: accruals management, manipulation of real economic activities, and classification shifting (McVay, 2006). The difference between the net income and the cash flows are called the accruals (Li et al., 2009). When managers use accrual-based earnings management, they created accruals to manipulate the earnings, which can either increase or decrease these earnings. To increase the earnings, managers borrow earnings from future periods, through revenue or delay current

(10)

10 expenses (Healy, 1985; Jones, 1991). Managers can also increase the bad debt reserves or write down inventory, to decrease the earnings (Li et al., 2009). If managers use the tool of manipulating real economic activities, managers give discounts to increase the sales and cut expenses, for instance, R&D costs (Baber et al., 1991). These actions increase the current earnings but are also costly because cutting R&D expenses can jeopardize future earnings (McVay, 2006). The third earnings management tool, classification shifting, is introduced by McVay (2006). Classification shifting is the misclassification of expenses within the income statement. Core expenses (cost of goods sold and selling, general and administrative expenses) are shifted as special items expenses, to increase the core earnings. Managers use this classification shifting, to mislead outsiders, for instance, analysts who use earnings benchmarks to make earnings forecasts. In these benchmarks, the special items of the income statement are excluded (McVay, 2006). Therefore, the analysts overstate the earnings, and this results in analysts´ forecasts which are less accurate since they do not show the real economic performance of the firm.

Prior research established different incentives for managers to engage in earnings management (Healy & Wahlen, 1999). As stated by Healy and Wahlen (1999) this includes (1) The expectations and valuations of the capital market; (2) accounting numbers which are included in contracts; and (3) anti-trust or other government regulations.

First, investors and financial analysts use accounting information such as earnings to forecast the value of the firm. Managers can use earnings management to manipulate the earnings, in an attempt to influence these forecasts with the purpose of increasing the stock price (Healy & Wahlen, 1999). The second category of incentives for managers can be found in contracts, which are written in terms of accounting numbers. Contracts between stakeholders and managers consist of accounting data to be able to monitor and regulate the actions of the managers. For instance, contracts between the manager and the shareholders and lending contracts with creditors. Contracts between the managers and the shareholders consist of specific targets concerning the performance of the company. When the manager is able to meet those targets they will get a reward, for instance, a bonus. Managers conduct earnings management to meet those targets, so they will receive these compensations (Healy, 1985; Scott, 2014). Furthermore, lending contracts between creditors and managers often consist out of debt covenants. Debt covenants are restrictions for the managers to be able to monitor their actions. Research shows that these debt covenants influence the accounting choices of the managers. (DeFond & Jiambalvo, 1994; Scott, 2014). Managers engage in earnings

(11)

11 management, due to the fact that the costs of violating these debt covenants are high. It leads, for instance, to reputational damage and increases the cost of capital (Healy & Wahlen, 1999; Scott, 2014). The third incentive for managers to engage in earnings management relates to anti-trust and other governmental regulations. Healy and Wahlen (1999), discuss several industries that are regulated by the government with the use of accounting data, such as the banks and the pension funds. Governmental bodies have set certain capital requirements for firms that operate in for instance the banking industry. Managers have incentives to meet the capital requirements and as a consequence, managers engage in earnings management to be able to meet those demands (Healy & Wahlen, 1999). Furthermore, managers make more use of earnings management if the firm is approaching delisting, is exposed to an anti-trust investigation, receiving a subsidy or governmental protection (Cheng et al., 2010, Healy & Wahlen, 1999; Watts & Zimmerman, 1978).

MNCs consists of a parent company, and foreign and domestic subsidiaries (Beuselinck et al., 2019). The parent company has to prepare the consolidated financial statements, which consists of all the subsidiaries, including their foreign subsidiaries. The consolidated financial statement aims to show the MNC as if it is one single entity.

The parent company set the tone by issuing targets and monitor the performance of all its subsidiaries (Beuselinck et al., 2019). The subsidiaries have to report their financial statements in order for the parent company to be able to assess their performance (Beuselinck et al., 2019). In turn, the parent company suggests adjustments in for instance their targets, to ensure the reporting objectives of the MNCs as a whole (Beuselinck et al., 2019; Busco et al., 2008). The possibility to make adjustments in the financial statements of their subsidiaries creates opportunities to influence the financial reporting at the benefit of the parent (Dyreng et al., 2012; Robinson & Stocken, 2013).

The parent company has several incentives to manage the earnings at the subsidiary level. First, the parent company manages their earnings through the use of their subsidiaries in order to meet or beat the benchmark that has been set by the analysts (Bonacchi et al., 2018). The costs of not meeting these benchmarks which are set by analysts’ are high, for instance, there will be a large decline in the stock price and legal actions can be taken by shareholders against the managers of the firm (Brown, 1993; Brown, 2001; Skinner, 1994). The second incentive for the parent company to manage their earnings at the subsidiary level relates to reputational damage. The reputational damage when discovering earnings management is lower at the

(12)

12 subsidiary level than at the parent level (Dearborn, 2009). Finally, the parent company can use the subsidiaries to manage earnings, to avoid reporting losses in the consolidated financial statement (Fan, 2012), since an increase in earnings at the subsidiary directly translates in an increase in earnings in the consolidated financial statements.

Forecast accuracy

The analysts can be seen as the intermediaries between the company and the investors (Luo et al., 2010). The analysts gather and analyze industry, market, and company information in order for the capital market to be better able to make investment decisions (Payne & Robb, 2000). Yet, the forecast is only valuable for the investors if it is accurate (Francis et al., 2019). Different studies examine determinants that influence the forecast accuracy of the analysts. The accuracy is negatively influenced by tax planning (Francis et al., 2019), internationalization (Duru & Reeb, 2002), and earnings volatility (Byard et al., 2006). And positively influenced by more informative disclosure policies (Lang & Lundholm, 1996), insider cash flow rights (Forst et al., 2019), bank ownership (García-Meca & Sánchez-Ballesta, 2011), and analysts’ experience (Clement, 1999).

Furthermore, Salerno (2014) provide evidence that the earnings quality increase the forecast accuracy of the analysts. This is important, because analysts use the earnings, from the financial statements, to make forecasts about future performance. The forecast of the analysts reflects the financial reporting quality of the financial statements. According to Salerno (2014), higher financial reporting quality leads to a more accurate forecast.

Hypotheses development

Bonacchi et al. (2018) investigated a sample of Italian non-listed subsidiaries and concluded that subsidiaries are conducting earnings management for their listed parent to meet or beat the benchmark. The literature established these benchmarks of the analysts, as a proxy for the expectations of the investors. According to previous research, the costs of not meeting these earnings expectations are high. For instance, there will be a large decline in the stock price and legal actions can be taken by shareholders against the managers of the firm (Brown, 1993; Brown, 2001; Skinner, 1994). Thus, the managers have high incentives to meet these expectations, and therefore engage in earnings management (McVay, 2006). Furthermore, according to Fan (2012), MNCs avoid reporting losses in their consolidated financial statement and therefore manage their earnings through their subsidiaries. Finally, when

(13)

13 earnings management is discovered, the reputation damage at the subsidiary level is lower than at the parent level (Dearborn, 1999).

Prior literature established different drivers of earnings management within the MNCs. First, subsidiaries which are located in tax havens are often more used to manage their earnings, due to earnings that are managed before the tax rather than the tax rate itself (Dyreng et al., 2012). Further, auditors of the parent company find it more difficult to audit their subsidiaries, and therefore, will be less able to detect earnings management at the subsidiary level than at the parent level (Stewart & Kinney, 2013). Next, Beuselinck et al., (2019) found evidence that MNCs manage the earnings more if the subsidiaries are highly integrated into the corporate group. At last, MNCs that are operating in a country which has a high-quality institutional environment, which means that the costs of detecting earnings management are high, tend to manage their earnings through subsidiaries which are established in countries with a low- quality institutional environment (Prencipe, 2012).

Thus, prior studies (e.g. Dyreng et al., 2012; Beuselinck et al., 2019; Bonacchi et al., 2018) have shown that the MNC tend to have large incentives and different drivers to manage the earnings through the use of their subsidiaries instead of managing the earnings at the parent level. However, as discussed previously, earnings management result in a financial statement that does not show the true economic value of the firm, and it, therefore, decreases the financial reporting quality (Dye, 1988). Since the financial statements of the subsidiaries are also included in the consolidated financial statement, it can be expected that the reporting quality of the subsidiary influences the reporting quality of the group as a whole.

Considering prior studies, I expect that a lower financial reporting quality at the subsidiary level affects the financial reporting quality of the whole corporate group and leads to a less accurate forecast of the analysts’ who are following the MNC. This leads to my first hypothesis:

H1: A lower FRQ at the subsidiary level leads to less accurate forecasts of the analysts who

(14)

14 MNCs often have a very complex structure, which includes earnings from domestic and foreign affiliates (La porta et al., 1999). Expansion of MNCs by establishing or acquiring foreign subsidiaries increases the complexity of the firm (Chin et al., 2009; Duru & Reeb, 2002). This expansion to foreign markets may also increase the value of the firm due to exploiting this uncertain environment (Kogut, 1983). However, this uncertain environment leads to more information asymmetry and creates room for manager discretion (Chin et al., 2009; Duru & Reeb, 2002). It creates, for instance, more opportunities for the parent company to engage in earnings management (Chin et al., 2009). Chin et al (2009) have established a negative relation between the financial reporting quality and a geographically diversified firm. Despite that this study is conducted with stand-alone firms, literature relating to MNCs shows that MNCs also tend to manage their earnings more in foreign subsidiaries (Beuselinck et al., 2019; Fan, 2012). In line is the finding of Fan (2012), who stated that MNCs attempt to avoid reporting consolidated losses, and for this purpose use more foreign than domestic subsidiaries to manage their earnings.

Besides a lower financial reporting quality, prior studies also established multiple difficulties for analysts regarding a more geographic diversified firm. First, a more diversified firm increases the complexity of the firm, which results in information asymmetry and less transparency (Chin et al., 2009; Duru & Reeb, 2002), Further, analysts tend to have more knowledge about operations in their home countries than about operations in foreign countries. Differences in, for instance, institutional regulations, culture, and language increase the difficulty of forecasting these operations (Ashbaugh & Pincus, 2001; Duru & Reeb, 2002). Duru and Reeb (2002), argue that due to the larger size of MNCs, there is more information available about the firm. However, Duru and Reeb (2002) stated that the increase of available information of MNCs does not outweigh the unfamiliarity of analysts regarding foreign operations. According to the study of Duru and Reeb (2002), who investigate corporate international diversification of MNCs on the forecast accuracy, a firm that is geographically diversified leads to analysts’ forecasts which are less accurate.

In sum, prior studies concluded that firms manage their earnings away from regulators (Ayers et al., 2011; Choi et al., 2012). In line is the research of Fan (2012) who finds that MNCs manage more earnings in foreign subsidiaries to avoid monitoring actions from the parent’s country of residence. Furthermore, the research of Duru and Reeb (2002), find that geographic diversification has a negative relationship on the forecast accuracy. This study also examines whether geographic diversification has a moderating effect on the relation

(15)

15 between the financial reporting quality of the subsidiaries and the forecast accuracy of the analysts who are following the MNC. As discussed previously, I expect that lower financial reporting quality results in forecasts that are less accurate. Taking into account prior literature regarding geographic diversification, I expect that geographic diversification strengthens this relation. This leads to my second hypothesis:

H2: Geographic diversification strengthens the positive relationship between financial reporting quality of the subsidiary and the forecast accuracy of the analysts who are following the parent.

3 Sample selection and research design

Sample selection

This study is based on a large sample of U.S. Multinationals and their European subsidiaries for the period of 2011 till 2017. The data from the listed U.S. parent company are gathered from the database Compustat. I eliminate the financial firms because of the high leverage, and those firms are therefore not comparable to non-financial firms, where high leverage indicates financial distress (Fama & French, 1992). I also discard the firms that do not have the necessary financial information to carry out the research. The financial information of the parent company is obtained from Compustat, while the information about the forecast accuracy of the analysts following the parent company is gathered from the Institutional Brokers Estimate System (I/B/E/S) database. The information regarding the actual earnings and the forecasts are collected from the summary file of the I/B/E/S database (Das, 2002; Hope, 2003). The reason that both the actual and the forecast information are gathered from the I/B/E/S database is to ensure that the data is comparable. (Das, 2002). In some papers, researchers use the actual earnings from Compustat (O’Brien, 1990; Sinha et al., 1997). However, the use of the actual earnings per share from the I/B/E/S database has another major advantage since it also controls for the earnings metrics (Das, 2002). To be in line with prior studies and to ensure the consensus forecast can be made properly, I require a minimum of three analysts who made a forecast about a U.S. multinational in one specific year (Hope et al, 2006; Payne & Robb, 2000).

Furthermore, for the listed non-financial firms I hand-collect the names and jurisdiction of all material subsidiaries included in exhibit 21 of 10-k filings from Edgar. Unfortunately, not all

(16)

16 the multinationals could be found within the Edgar database or did not disclose their relevant subsidiaries. Those are excluded from the final sample. Given the coverage of Orbis, which I use to gather the financial data of the subsidiaries, I select only the subsidiaries of 30 European countries. Next, the subsidiaries that are operating in the financial industry are removed, because of the above-mentioned reason of the comparability issue. After discarding material subsidiaries with unavailable information in the Orbis database, the final sample of subsidiaries consists of 7,189. These material subsidiaries are owned by 219 parent companies about which 13,986 forecasts have been made by analysts during the period of 2011 till 2017. This leads to total observations of 1,115. The parent companies are classified by industry using the 2-digit SIC codes. The U.S. multinationals in my sample operate in 28 different industries including manufacturing instruments and related products (16.7%), business services (14.8%), and electronic and other electric equipment (14.6%).

Research Design

Accuracy of the analysts’ forecast

The forecast accuracy (ACCURACY) is measured for each firm-year, as the absolute value of the difference between the actual earnings and the forecast consensus, scaled by the absolute value of the actual earnings (Horton et al., 2013). Furthermore, the variable ACCURACY is multiplied by (-1) for the ease of interpretation (Duru & Reeb, 2002). Therefore,

ACCURACY does not explain the direction of the accuracy of the forecast, but rather capture

the magnitude of the forecast accuracy. This means that when the value of ACCURACYt increases, this indicates a higher forecast accuracy of the analysts (Duru & Reeb, 2002).

ACCURACYt = (-1) | AEjt-FCjt-1|

|AE_jt| (1)

AEjt is the actual earnings, measured as the earnings per share in period t. FCjt-1 is the consensus forecast of the analysts in period t, which is the forecast of the earnings that have been made in period t-1.

In line with the research of Forst et al. (2019), I used the consensus forecast that has been made 3 months prior to the end of the fiscal year. To test the robustness of the results, I perform the analysis with the use of alternative forecast horizons (Byard et al., 2006). This includes the consensus forecasts that have been made nine months, six months, and the most recent one before the end of the fiscal year (Forst et al., 2019). The result of these robustness

(17)

17 test can be found in table A.1 Further, I performed an additional test using the median instead of the consensus forecast. The results can be found in table B.1.

Measurement of Financial Reporting Quality of the subsidiaries

In this paper, earnings management is used as a proxy for financial reporting quality. To measure earnings management, I used accruals. This is a well-known way to measure earnings management that is extensively done in empirical research (Klein, 2002; Payne & Robb, 2000). Accruals are non-cash working capital, for instance, inventory and accounts receivable. However, accruals can be misused, this is called discretionary accruals. Those accruals are present and cannot be explained via normal operating activities (Payne & Robb, 2000).

In line with prior research (DeFond & Jiambalvo, 1994; Payne & Robb, 2000) I used the Modified Jones Model proposed by Dechow et al. (1995) to estimate the discretionary accruals of each subsidiary. The Modified Jones model is similar to the standard Jones Model (Jones, 1991), except for the adjustment of including sales-based manipulation, since changes in sales can imply a higher use of discretionary accruals (Dechow et al., 1995). The discretionary accruals are calculated as the difference between the total accruals and non-discretionary accruals.

The total accruals (TA) are calculated as the net working capital minus the depreciation expense (Peasnell et al., 2012).

TAit = ΔNon-cash current assetsi,t – ΔCurrent liabilitiesi,t – Depreciation expensei,t (2)

The modified Jones Model is presented in equation (3):

TAit

Assetsit-1 = α1t

1

Assetsit-1 + α2t

ΔRevit-ΔReceivablesit

Assetsit-1 +α3t

PPEit

Assetsit-1 +єit (3)

Where:

TAit = total accruals for sample firm i in year t

Assetsit-1 = total assets for sample firm i in year t-1

ΔRevit = change in revenues for sample firm i in year t

ΔReceivables = changes in accounts receivable for sample i in year t PPEit = gross property plant and equipment for sample firm i in year t

(18)

18 Each component is divided by the total assets (Assetsit-1), therefore, the Modified Jones

Model, controls for the firm size of the company. In addition, the model also controls for the firm-year. This is important since the firm size and the year from the firms which are included in my sample differ across companies and could influence the presence of discretionary accruals. Equation (3) is computed separately for each industry on cross-sectional data for the companies that are listed under the same 2-digit SIC codes (DeFond & Jiambalvo, 1994; Haw et al., 2004; Paynee & Robb, 2000). Further, representative estimates can only be made when there are at least six companies in a certain industry (Defond & Jiambalvo, 1994; Subramanyam, 1996). Therefore, I eliminated the subsidiaries, if they are operating in an industry that contains less than six firms in my sample.

The value of the discretionary accruals can either be positive or negative. In this study, I used the unsigned value, because I am interested in the magnitude of earnings management rather than the direction of the discretionary accruals. This means that when the unsigned value increases, there is also an increase in the discretionary accruals. The use of absolute value is in line with prior research that has been done regarding discretionary accruals (Haw et al., 2004; Klein, 2002).

In this research, the financial reporting quality of all the subsidiaries (FRQ_SUB), is measured as the weighted average of the absolute value of discretionary accruals, reported by all the material subsidiaries of the parent in a given year.

Geographic Diversification

The U.S. multinationals have established subsidiaries in many countries throughout Europe. Since I focus on European subsidiaries, I consider that multinationals have a higher degree of geographic diversification, if the corporate group owns more foreign subsidiaries established in Europe. Prior literature shows that the use of a so-called ‘country scope’ is a good indicator to measure the geographic diversification of a firm (Delios and Beamish, 1999; Tallmand & Li, 1996). Country scope is defined as the number of countries in which a firm has established subsidiaries. Similar to those papers, I use the number of European countries in which the U.S. multinationals have subsidiaries as an indicator of geographic diversification (GEO_DIV). In other words, I am interested in the number of countries where the listed U.S. parent has subsidiaries, rather than the number of foreign subsidiaries.

(19)

19

Empirical Model

Equation (4) examines the association between the financial reporting quality of the subsidiaries and the forecast accuracy of the analysts, who are following the parent after controlling for different factors that prior literature has shown to influence forecast accuracy.

ACCURACYt = β0 + β1FRQ_SUBit + β2EARN_SURit + β3EVOLATit (4)

+ β4LOSSit + β5FIRM_SIZEit + β6FOLLOWit-1

+ β7HORIZONit-1 + β8YEARi + β9INDUSTRYit +єit

I use equation (5) to test whether the geographic diversification influences the relation between the financial reporting quality of the subsidiaries and the forecast accuracy of the analysts. Equation (5) is obtained by adding the moderating effect of geographic diversification (FRQ_SUB * GEO_DIV) to the model presented in equation (4).

ACCURACYt = β0 + β1FRQ_SUBit + β2GEO_DIVit +β3FRQ_SUBit * GEO_DIVit (5)

+ β4EARN_SURit +β5EVOLATit + β6LOSSit + β7FIRM_SIZEit + β8FOLLOWit-1

+ β9HORIZONit-1 + β10YEARi + β11INDUSTRYit + єit

In line with Call et al. (2009), there is controlled for outliers by using the winsorizing technique. All the continuous variables are winsorized at the 1% and 99% level.

Control variables

When examining the relationship between the financial reporting quality of the subsidiaries and the forecast accuracy of the analysts and the potential effect of geographic diversification on this relation, I include two sets of control variables. The first set controls for the firm characteristics that influence the forecast accuracy of the analysts and the second set controls for the characteristics of the analysts who are making the forecasts.

Control variables for firm characteristics

Based on prior studies, I controlled for four different firm characteristics for the reason that these influence the forecast accuracy of the analysts. The first set of controls contains the earnings surprise (EARN_SUR), the earnings volatility (EVOLAT), the likelihood of reporting a loss (LOSS), and the firm size (FIRM_SIZE).

(20)

20 In the paper of Lang and Lundholm (1996), the authors discuss earnings surprise as an important determinant that influences the forecast accuracy of the analysts. When a company launches for instance a major new product, the earnings of a company can increase extensively. The realized earnings could deviate from the expected earnings, therefore the forecast consensus among analysts could be very low. Revisions of the forecasts that differ significantly from previous forecasts are expected to happen. To control for earnings surprise, I include the control variable EARN_SUR. Earnings surprise is measured as the absolute difference between the current year earnings per share and the last year’s earnings per share divided by the price at the beginning of the year (Lang & Lundholm, 1996).

Different academics suggest that long-term earnings volatility since it is more difficult to predict these earnings, results in forecasts that are less accurate (Francis et al., 2019; Lim, 2001; Byard et al., 2006). Therefore, I controlled for the earnings volatility (EVOLAT), measured as the standard deviation of the actual EPS five years prior to the end of the fiscal year divided by the stock price (Byard et al., 2006).

Analysts find it more difficult to predict a stock price for firms that are making a loss instead of a profit. The reason for this is that negative earnings are less persistent than positive earnings, and analysts find it more difficult to predict negative earnings which relate to temporary earnings (Hwang et al., 1996). This means that the analysts’ forecasts for firms that reported a loss are less accurate than if the firm reported a profit (Hwang et al., 1996). I include a dummy variable (LOSS), which takes the value equal to one if the firm has reported a loss (EPS<0), and zero if otherwise (Byard et al., 2006; Garía-Meca & Sánchez-Ballesta, 2011).

The firm size of the parent company could result in two contrary effects. On one side, the firm size could lead to more complexity and more dispersion among the analysts. (Duru & Reeb, 2002) On the other side, firms that are larger tend to be more transparent and disclose more information which could lead to a more accurate forecast (Duru & Reeb, 2002). To mitigate these potential associations, firm size (FIRM_SIZE) is controlled for as the natural log of the total assets of the parent (Dyreng et al., 2012; Garía-Meca & Sánchez-Ballesta, 2011).

(21)

21

Control variables for forecast characteristics

The second set of controls is related to the characteristics of the forecast that have been made by the analysts. This set control variables include the number of analysts who are making forecasts about the U.S. multinationals (FOLLOW) and the time period between the forecast date and the earnings announcement (HORIZON).

Duru and Reeb (2002) stated that the firm disclosed more corporate information when the number of analysts who are following the firm increases. This leads to the fact that firms with a higher degree of analysts’ following, tend to have a more accurate forecast (Lung & Lundholm, 1996). Further, the number of analysts (FOLLOW) is also an indirect control for the growth opportunities and the firm size of the multinationals (Byard et al., 2006). High growth firms and firms that are larger have on average a higher degree of analysts who are following the firm (Barth et al., 2001). FOLLOWis measured as the number of analysts who made a forecast and are included in the consensus forecast (Forst et al., 2019). In line with prior studies (Byard et al., 2006; Garía-Meca & Sánchez-Ballesta, 2011), I performed an additional test by adding the market-to-book ratio as a direct growth control. The results can be found in table B.1.

The empirical findings of different studies stated that the forecast horizon strongly affects the forecast accuracy (Brown & Mohd, 2003; Sinha et al., 1997). Forecasts that have been made with a longer horizon to the end of the fiscal year are less accurate (Brown et al., 1987). I include horizon (HORIZON) as control variable as the days between the date that the consensus forecast is recorded in I/B/E/S and the date of the earnings announcement (Duru & Reeb, 2002). I additionally test whether the results are robust to changes in the measurement of the control variable HORIZON. Using the days between the consensus forecast and the end of the fiscal year instead of the earnings announcement date (AHORIZON). The results can be found in table B.1.

Finally, analysts can undermine additional difficulties regarding forecasting, across different years and industries. Therefore, I include dummy variables for the industry (INDUSTRY) using the 2-digit SIC code and for the fiscal year (YEAR) (Bhat et al., 2006; Garía-Meca & Sánchez-Ballesta, 2011).

(22)

22

4

Empirical analysis

Descriptive statistics

The descriptive statistics of the sample are shown in Table 1. The sample consists of a total of 219 U.S. multinationals with 1,115 observations. Results suggest that the mean forecast accuracy (ACCURACY) of my sample is -0.12. The forecast accuracy is measured to capture the magnitude instead of the direction. It, therefore, implies that on average the forecast differs 0.12 USD from the actual EPS reported by the firm. In comparison to the study of Duru and Reeb (2002) and Byard et al. (2006), which report a forecast accuracy of 0.05 and 0.06, respectively, the analysts in my sample made less accurate forecasts.

The financial reporting quality of the subsidiaries (FRQ_SUB), measured as the average discretionary accruals, has an absolute value of 0.17. This is quite similar in comparison with the study of Beuselinck et al. (2019), who also researched the subsidiary discretionary accruals.

Furthermore, the U.S. multinationals have obtained subsidiaries which are located in, on average, 4 different countries throughout Europe. The earnings surprise (EARN_SUR), is measured as an absolute value to be able to capture the magnitude of the surprise. On average, the surprised earnings for the analysts have a value of 0.01 USD. And the mean of earnings volatility (EVOLAT) has a value of 0.02 USD. Only 4% of the U.S. multinationals reported a loss in the fiscal year. This amount is lower than in the studies of Byard et al (2006) and Duru and Reeb (2002), who reported 8% and 11%, respectively. The mean of the sample regarding the firm size (FIRM_SIZE) takes the value of 8.04. Table 1 also shows that 12 analysts have made a forecast about the earnings per share of the U.S. multinational. This is around the same number reported in the study of Duru and Reeb (2002). On average, the consensus forecast has been made 144 days before the earnings announcement of the firm.

(23)

23

Table 1

Descriptive statistics

Variable N Mean St. Dev. p25 Median p75 Min Max

ACCURACY 1115 -0.12 0.26 -0.10 -0.04 -0.01 -2 0 FRQ_SUB 1115 0.17 0.15 0.09 0.14 0.21 0.01 0.95 GEO_DIV 1115 4.37 3.73 1.5 3 6 1 15 EARN_SUR 1115 0.01 0.02 0.00 0.01 0.02 0 0.12 EVOLAT 1115 0.02 0.02 0.01 0.01 0.02 0.00 0.17 LOSS 1115 0.04 0.20 0 0 0 0 1 FIRM_SIZE 1115 8.04 1.55 6.90 7.85 8.98 4.66 11.87 FOLLOW 1115 12.49 8.42 6 10 18 3 37 HORIZON 1115 144.24 12.74 133 140 154 118 180 This table shows the descriptive statistics, whereby all the continuous variables are winsorized at the 1% and 99% level. The sample size, mean, standard deviation, percentile 25, median, percentile75, minimum, and maximum are shown. The forecast accuracy (ACCURACY) is a measure of analysts’ forecast accuracy and it is calculated as -1 * (|Actual EPS – Consensus Forecast|/Actual EPS). FRQ_SUB is a measure of the financial reporting quality of the subsidiaries, calculated as the weighted average of the absolute value of discretionary accruals, reported by all the material subsidiaries of the parent in a given year. GEO_DIV is the number of different countries in which the parent company established material subsidiaries. The earnings surprise (EARN_SUR), is calculated as the absolute value difference between the current year earnings per share and the prior year’s earnings per share scaled by the stock price at the beginning of the year. EVOLAT, is the earnings volatility measured as the standard deviation of the actual EPS five years prior to the end of the fiscal year scaled by the stock price. LOSS is a dummy variable which equals to one if a firm reported a loss in a specific year (actual EPS <0), and equals to zero otherwise. The FIRM_SIZE is calculated as the natural log of the total assets.

FOLLOW, is the number of analysts who are following the multinational. HORIZON is the number of days

between the date that the consensus forecast is recorded in I/B/E/S and the date of the earnings announcement.

Correlation

The results of the Pearson correlation between the variables are shown in table 2. Results show that the correlation between forecast accuracy and subsidiary FRQ is not statistically significant. Contrary to my predictions, this finding suggests that the analysts’ forecast accuracy is not influenced by the average level of subsidiary earnings management.

Table 2 reports a positive correlation between the analysts’ forecast accuracy and geographic diversification (GEO_DIV) that is significant at the 5% level. This suggests that the forecast accuracy of the analysts increases with the number of foreign countries in which the MNC has subsidiaries (0.09). This is not consistent with my expectations, which stated that more geographic diversification is associated with lower forecast accuracy. Expanding operations to foreign countries can lead to less forecast accuracy because analysts experience more difficulties related to institutional regulations, culture, and language when making forecasts. However, the analysts, who are making forecasts about the U.S. multinational, are based around the world. It is, therefore, possible that the U.S. Multinational has established their foreign subsidiary in the same country or in a country which for instance has the same

(24)

24 regulations and language. Therefore, it brings more understanding of the operations of the foreign subsidiary for the analysts.

The four control variables regarding the firm characteristics (EARN_SUR, EVOLAT, LOSS, and FIRM_SIZE) are statistically correlated at the 5% level with the forecast accuracy of the analysts. Earnings surprise, earnings volatility, and the likelihood of reporting a loss are negatively correlated and are therefore associated with less accurate forecasts. Thus, this finding suggests that if the EARN_SUR (-0.38), EVOLAT (-0.30), and LOSS (-0.45) 1 increases, the forecast accuracy will decrease. This result is in line with my predictions. As discussed previously, the forecast accuracy may change depending on the firm size. Similar to Byard et al. (2006), the FIRM_SIZE in table 2 shows a positive correlation (0.19) with the forecast accuracy. Forecast accuracy is also positively and significantly associated with the number of analysts following the firm (0.17). In contrast with the other control variable for forecast characteristics, HORIZON is negatively correlated (-0.28) with the forecast accuracy. This is in line with my predictions as discussed in the methodology section. The coefficients of the correlations between the other explanatory variables are relatively small, except for the positive and significant association between FOLLOW and FIRM_SIZE (0.65). This association is not surprising since prior literature established an association between the number of analysts who are following the firm and the firm size (Bushan, 1989). However, the coefficients of the correlations do not reach the cut-off level of 0.7. Thus, it indicates no multicollinearity issues in my empirical model.

1_{Loss firms have a large impact on the forecast accuracy. Duru and Reeb (2002) came to this conclusion after} they removed the loss-making firms and calculated the forecast accuracy of the profitable firms. I confirm that this finding also apply to my sample. Similar to Duru and Reeb (2002), I analyze if this also applies to my sample. I exclude the 59 firms who made a loss in a specific year and measure the forecast accuracy mean of the profitable forecast as -0.10. By excluding only 59 of the total sample of 1,115, the forecast accuracy is 0.02 USD closer to zero.

(25)

25 Table 2 Correlation matrix Variables 1 2 3 4 5 6 7 8 9 1 ACCURACY 1 2 FRQ_SUB -0.04 1 3 GEO_DIV 0.09 -0.04 1 4 EARN_SUR -0.38 0.01 -0.10 1 5 EVOLAT -0.30 0.01 -0.12 0.53 1 6 LOSS -0.45 0.08 -0.09 0.33 0.26 1 7 FIRM_SIZE 0.19 -0.10 0.49 -0.18 -0.22 -0.19 1 8 FOLLOW 0.17 -0.04 0.17 -0.15 -0.21 -0.10 0.65 1 9 HORIZON -0.28 0.03 -0.22 0.15 0.17 0.21 -0.47 -0.46 1

The correlation matrix shows the Pearson correlation coefficient of the variables in the study, whereby all the continuous variables are winsorized at the 1% and 99% level. The correlation coefficients that are statistically significant at least at the 5% level are shown in bold. The forecast accuracy (ACCURACY) is a measure of analysts’ forecast accuracy and it is calculated as -1 * (|Actual EPS – Consensus Forecast|/Actual EPS). FRQ_SUB is a measure of the financial reporting quality of the subsidiaries, calculated as the weighted average of the absolute value of discretionary accruals, reported by all the material subsidiaries of the parent in a given year. GEO_DIV is the number of different countries in which the parent company established material subsidiaries. The earnings surprise (EARN_SUR), is calculated as the absolute value difference between the current year earnings per share and the prior year’s earnings per share scaled by the stock price at the beginning of the year. EVOLAT, is the earnings volatility measured as the standard deviation of the actual EPS five years prior to the end of the fiscal year scaled by the stock price. LOSS is a dummy variable which equals to one if a firm reported a loss in a specific year (actual EPS <0), and equals to zero otherwise. The FIRM_SIZE is calculated as the natural log of the total assets. FOLLOW, is the number of analysts who are following the multinational. HORIZON is the number of days between the date that the consensus forecast is recorded in I/B/E/S and the date of the earnings announcement.

(26)

26 Regression analysis

To test my hypotheses, I conducted two ordinary least square regressions. The results of the regression for equation (4) and equation (5), are reported in table 3 and presented under column (1) and (2), respectively.

The first column tests the relation between the average discretionary accruals, as a proxy for the financial reporting quality of the subsidiaries (FRQ_SUB), and the forecast accuracy of the analysts who are following the multinational (ACCURACY). The relation between

FRQ_SUB and ACCURACY is not statistically significant. This finding suggests that the

analysts’ forecast accuracy is not influenced by the subsidiary FRQ. This finding does not support hypothesis 1, which stated that there would be a positive relation. A possible explanation for my results is that the earnings management conducted by the European subsidiaries might not be large enough to alter the financial reporting quality of the corporate group as a whole. The lower amount of earnings management could be due to the strict rules of the European Commission regarding the financial reporting standards. This leaves less space to conduct earnings management at the subsidiary level and thereby will not have an excessive impact on the FRQ of the consolidated financial statement of the MNC.

The coefficient of the earnings surprise, earnings volatility, the likelihood of reporting a loss, and the forecast horizon is negative and statistically significant. This suggests that if

EARN_SUR, EVOLAT, LOSS, and HORIZON increases, this results in a less accurate forecast

of the analysts who are following the MNC. This is in line with my predictions and prior research (Brown et al., 1987; Byard et al., 2006; Hwang et al., 1996; Lang & Lundholm, 1996).

Based on the results of table 3, the number of analysts who are following the U.S. multinational (FOLLOW) and the firm size (FIRM_SIZE) are not significantly related with the forecast accuracy of the analysts. These results are not in line with my predictions and prior research, which reports a positive statistically significant relation (Duru & Reeb, 2002; Francis et al., 2019). However, other studies who examined the forecast accuracy also reported an absence of a statistically significant relation related with the number of analysts following (FOLLOW) and the firm size (FIRM_SIZE) (Byard et al., 2006; Garía-Meca & Sánchez-Ballesta, 2011).

(27)

27 The second column in table 3 reports the findings of the OLS regression regarding hypothesis 2. This includes the moderating effect of geographic diversification on the relation between the FRQ of the subsidiaries and the forecast accuracy (FRQ_SUB * GEO_DIV). I find that geographic diversification (-0.025) does not significantly affect the relation between the financial reporting quality of the subsidiaries and the forecast accuracy. This is not in line with hypothesis 2, which stated that geographic diversification strengthens the relation between a low financial reporting quality of the subsidiary and less accurate forecasts of the analysts who are following the MNC.

As discussed previously, the European Commission set strict rules regarding financial reporting standards. In addition, the second reason for these strict rules is to ensure consistency and for outsiders to be better able to compare the firms which are established in different European countries. The absence of the moderating influence might be due to these strict rules of the European Commission to which the European subsidiaries of the U.S. multinational have to comply to. Therefore, the geographic diversification of the MNC in Europe does not influence the relation between the FRQ of the subsidiaries located in Europe and the forecast accuracy of the analysts.

Furthermore, table 3 also includes the direct effect of the geographic diversification (GEO_DIV) on forecast accuracy. Geographic diversification (0.005) is also not related with the forecast accuracy. This means that geographic diversification does not influence the forecast accuracy of the analysts. This is not in line with the findings of Duru and Reeb (2002), who report that a more geographic diversified firm leads to a less accurate forecast. However, in my study, I use a sample that consists of only European companies to measure geographic diversification, instead of Duru and Reeb (2002), who conducted their study with companies which are located all around the world. This sample difference explains the absence of this relation. As discussed previously, the European Commission has set rules to ensure the high financial reporting quality of the companies. However, these rules are also set to create consistency among European companies. Meaning, that an increase in geographic diversification in Europe, still results in subsidiaries that have to comply to the same rules. For that reason, an increase in geographic diversification will not influence the forecast accuracy of the analysts who are following the MNC.

The coefficients of the control variables reported under column (1) and column (2) are similar to each other.

(28)

28

Table 3

Ordinary least square regression analysis

This table reports the results of the OLS regression between the forecast accuracy and (1) the relation between financial reporting quality of the subsidiaries as the main independent variable, and (2) the relation between financial reporting quality of the subsidiaries as the main independent variable and the moderating effect of geographic diversification on this relation. The continuous variables have been winsorized at the 1% and 99% levels. The standard errors are reported in the parentheses. *, **, and, *** significance at the 10%, 5% and 1% levels, respectively. The forecast accuracy (ACCURACY) is a measure of analysts’ forecast accuracy and it is calculated as -1 * (|Actual EPS – Consensus Forecast|/Actual EPS). FRQ_SUB is a measure of the financial reporting quality of the subsidiaries, calculated as the weighted average of the absolute value of discretionary accruals, reported by all the material subsidiaries of the parent in a given year. GEO_DIV is the number of different countries in which the parent company established material subsidiaries. The earnings surprise (EARN_SUR), is calculated as the absolute value difference between the current year earnings per share and the prior year’s earnings per share scaled by the stock price at the beginning of the year. EVOLAT, is the earnings volatility measured as the standard deviation of the actual EPS five years prior to the end of the fiscal year scaled by the stock price. LOSS is a dummy variable which equals to one if a firm reported a loss in a specific year (actual EPS <0), and equals to zero otherwise. The FIRM_SIZE is calculated as the natural log of the total assets.

FOLLOW, is the number of analysts who are following the multinational. HORIZON is the number of days

between the date that the consensus forecast is recorded in I/B/E/S and the date of the earnings announcement. ACCURACY (1) (2) Hypothesized variables FRQ_SUB -0.0129 0.044 (0.047) (0.074) GEO_DIV 0.005 (0.004) FRQ_SUB * GEO_DIV -0.025 (0.025) Firm characteristics controls

EARN_SUR -2.631*** -2.618*** (0.444) (0.444) EVOLAT -0.876** -0.863** (0.363) (0.364) LOSS -0.340*** -0.341*** (0.038) (0.038) FIRM_SIZE 0.004 0.002 (0.007) (0.008)

Forecast characteristics controls

FOLLOW 0.001 0.001

(0.001) (0.001)

HORIZON -0.003*** -0.003***

(0.001) (0.001)

Fixed effects

YEAR yes yes

INDUSTRY yes yes

Observations 1,115 1,115

(29)

29 Additional Analyses

In order to test the robustness of the results, I performed four additional tests. The results are shown in tables A.1 and B.1, respectively. Similar, to table 3, column (1) refers to hypothesis 1 and column (2) refers to hypothesis 2.

Change of forecast horizons

To test the robustness of my results, I examine if the findings differ when there is a change in the forecast horizon (Byard et al., 2006). Similar to Forst et al. (2019), I use the most recent forecast, the forecast that has been made 6 months, and 9 months prior to the end of the fiscal year. The results regarding the change of the forecast horizons are reported in table A.1. The findings are similar to those reported in table 3.

Alternative forecast accuracy measurement

In line with the studies of Das (2002) and Forst et al (2019), I test the robustness of my results, by adjusting the mean consensus forecast to the median. The results are presented in table B.1 and are comparable to the results in table 3.

Including growth measurement

As stated by Byard et al. (2006), the measurement of FOLLOW is an indirect indicator for the growth opportunities. In line with the study of Byard et al. (2006), I performed an additional test in which I added the market-to-book ratio as a control variable (Garía-Meca & Sánchez-Ballesta, 2011). As shown in table B.1, there is no statistically significant relation between the market-to-book ratio and the forecast accuracy.

Change of measurement control variable horizon

Prior studies regarding forecast accuracy added the variable HORIZON in their empirical model. Similar to those studies, I included HORIZON as a control variable measured as the days between the consensus forecast and the earnings announcement of the firm (Duru & Reeb, 2002). However, the measurement of this indicator differs among studies. Forst et al. (2019) and Horton et al. (2013) measured the control variable HORIZON as the days between the consensus forecast and the end of the fiscal year. To test if my findings are robust, I change the measurement of HORIZON (Forst et al., 2019). I changed the control variable

HORIZON to the days between the consensus forecast and the end of the fiscal year, the

Subsidiary financial reporting quality, geographic diversification and analyst forecast accuracy: Evidence from European private subsidiaries