Do managers assist investors, or mislead them? The impact of political risks on the readability of narrative disclosures

SJOERD HARMENS (S2600617)

Department of Accounting; Faculty of Economics and Business; University of Groningen; Groningen, Netherlands

Supervisor:

prof. dr. R.B.H. (Reggy) Hooghiemstra +31(0)50 36 33772

r.b.h.hooghiemstra@rug.nl

Second assessor: dr. C.P.A. (Coen) Heijes

+31(0)50 36 37784 c.p.a.heijes@rug.nl

[word count: 11,869] June 24, 2019

2

TABLE OF CONTENTS

ABSTRACT 3

INTRODUCTION 4

THEORY & HYPOTHESES 7

RESEARCH DESIGN & METHODS 16

EMPIRICAL RESULTS 21

ROBUSTNESS CHECKS 30

3

Do managers assist investors, or mislead them? The impact of

political risks on the readability of narrative disclosures

Department of Accounting; Faculty of Economics and Business; University of Groningen; Groningen, Netherlands

June 24, 2019

MSc Thesis Accountancy (EBM869B20) ABSTRACT

This paper examines whether the readability of narrative disclosures is associated with the degree of firm risks emanating from politics. Specifically, I examine which of the two perspectives in literature on managers’ disclosure practices (i.e. opportunistic or informational) prevails under these circumstances. Using a sample comprising of U.S. firms between 2002 and 2016, I find that, when political risks are high, the readability of narrative disclosures decreases. In other words, my findings corroborate the opportunistic perspective: managers tend to make it difficult for investors to extract and process information from narrative disclosures in political uncertain times. In particular, high-able managers have stronger incentive to act in this way. A possible explanation is that they attempt to hide overcompensation and over-engagement in uncertain activities. Contrary to my expectation, I find that managers’ possibility to act in this way is not constrained by independent directors. This may be due to a lack of knowledge about firm-specific political risks. The results are robust after controlling for several alternative explanations and measurements, as well as controls for the selection bias. My findings fuel the ongoing debate among economist, business leaders, and politicians about the consequences of political risks for businesses and society.

Keywords: political risk; readability; managerial ability; board independence

Acknowledgements: I appreciate the helpful comments and suggestions from Reggy Hooghiemstra (supervisor) from the University of Groningen. I am also grateful for the critical comments from Johan Karssenberg from the University of Groningen, Roy Hoekstra from the University of Groningen, and Jens Thomasson from Ernst & Young.

4

“

”

INTRODUCTION

More and more companies are suffering large losses due to continuing political tensions (e.g. the escalating trade war between the U.S. and China or the U.K. leaving the European Union). As a result of uncertainties about governments’ future actions, share prices fluctuate on a daily basis (Pastor & Veronesi, 2013). The ambiguity about firm value provides incentive for managers and investors to emphasize the exchange of information (Verrecchia, 2001). Investors, when faced with political uncertainty, attempt to identify and quantify how the prospects of companies are affected by these risks. In order to do so, they demand information from managers, which is supplied in the form of disclosures (Nagar, Schoenfeld, & Wellman, 2019). In particular, narrative disclosures are fruitful as they provide contextual information about the company (Merkley, 2014); that is, managers explain to investors how they deal with political risks (e.g. perform lobbying activities or dialogues with government agencies). However, the information shared by the self-interested manager may be strategically chosen (Abrahamson & Park, 1994; Bernard, 2016). Literature suggests that if the manager expects investors will react negatively to political risks, he will attempt to conceal information in narrative disclosures (i.e. to present themselves in the most favourable way). Despite its current importance and relevance, literature is ambiguous about which managerial disclosure practice prevails: opportunistic or informational (Merkl-Davies & Brennan, 2007; Riedl & Srinivasan, 2010). Hence, this study attempts to answer the empirical question whether managers use the information in narrative disclosures to assist investors, or to mislead them, when confronted with political risks.

Despite publicly traded firms being required to issue periodic disclosures that discuss the firm’s operations and financial condition, authorities struggle with the notion of readability (Loughran & McDonald, 2014). Preparing and presenting readable narrative disclosures is important since it determines investors’ ability to “assimilate information into asset prices” (Loughran & McDonald, 2016, p. 1198). Yet, due to lack of regulation, it depends on the manager’s motivation how readable he presents the narrative disclosures. To conceal the impact of political risks on the firm’s prospects, the opportunistic perspective suggests that managers are inclined to lower the readability. By doing so, it becomes more difficult for investors to

5 extract and process the relevant information from narrative disclosures – ultimately mitigating a lower valuation of their company (Bloomfield, 2002). Consistently, several studies demonstrate that managers tend to mask negative information in disclosures by lowering the readability (Li, 2008; Ajina, Laouiti, & Msolli, 2016; Lo, Ramos, & Rogo, 2017; Hooghiemstra, Kuang, & Qin, 2017; Asay, Libby, & Rennekamp, 2018; De Souza, Rissatti, Rover, & Borba, 2019). In contrast, the informational perspective implies that managers with an exceptional way of dealing with political risks want to signal this; communicate their excellence to investors. To succeed, narrative disclosures must be of high-quality because it enhances the credibility of the information (Eccles, 2001). Therefore, in line with the suggestion of Rennekamp (2012), managers may be motivated to present more readable narrative disclosures since this results in a higher valuation of their company. Several studies find corroborating results that managers, in order to communicate the high-quality of their firm more effectively to investors, enhance the quality of disclosures (Zhang & Wiersema, 2009; An, Davey, & Eggleton, 2011; Merkley, 2014; Guay, Samuels, & Taylor, 2016; Bonsall Miller, 2017; Nagar et al., 2019).

Additionally, managers’ disclosure practices are likely determined by their ability to manage political risks. That is, managers are less likely to cloud information by lowering the readability of narrative disclosures if their performance is superior (Merkl-Davies & Brennan, 2007). Prior studies provide evidence (Demerjian, Lev, Lewis, & McVay, 2013; Baik, Brockman, Farber & Lee, 2018; Hasan, 2018) that highly-able managers have a stronger tendency to communicate their superior understanding of economic environment changes to investors (Trueman, 1986). In addition, Andreou, Philip, & Robejsek (2015) suggest that more able managers are better prepared against risks – indicating superior performance in uncertain times. From the informational perspective it can be argued that high-able managers, when confronted with political risks, have a stronger incentive to enhance the readability of narrative disclosures. Ceteris paribus, the incentive to act in line with the opportunistic perspective is weaker. Additionally, effective monitoring by the board of directors may influence managers’ possibilities to act in accordance with either perspective. As proposed by the early research from Forker (1992), director’s independence is an important characteristic when pressuring managers to disclose their private information. Later studies confirm this idea (Chen & Jaggi, 2000; Gul & Leung, 2004; Abraham & Cox, 2007; Laksmana, 2008); that is, boards with less ties to management are more effective in disciplining management, thereby enhancing the quality of disclosures. In line with these findings, I predict that a more independent board of directors is associated with a weaker (stronger) incentive for the opportunistic (informational) perspective to materialize.

6

Answering calls from scholars to examine determinants of the readability of disclosures based on the circumstances in which it is produced (Rutherford, 2003; Clatworthy & Jones, 2006), as well as to expand our theoretical understanding of the impact of political risk (Pastor & Veronesi, 2013; Hassan, Hollander, Lent, & Tahoun, 2017), I aim to answer the following question: ‘How is the readability of narrative disclosures affected when managers are exposed to political risks, and what is the effect of managerial ability and directors’ independence on this association?’. In line with the study of Hassan et al. (2017), who provides the first firm-level measure of political risk, the sample of this study consists of quarterly observations from U.S. firms between 2002 and 2016. In contrast to earlier studies on this topic (Bird, Karolyi, & Ruchti, 2017; Boone, Kim, & White, 2017; Nagar et al., 2019), I find that, after controlling for alternative explanatory factors, there is a negative association between political risks and the readability of narrative disclosures; managers tend to act in accordance with the opportunistic perspective when confronted with political risks. Moreover, I find, in contrast to my prediction, that the high-able manager has a stronger incentive to act in accordance with this perspective. Lastly, the results indicate that this disclosure practice of managers is not effectively counteracted by independent directors.

This study makes several contributions to literature. First, I contribute to literature on textual analysis of narrative disclosures. Prior scholars (Rutherford, 2003; Clatworthy & Jones, 2006; Merkl-Davies & Brennan, 2007) underscored the importance of enhancing our understanding of the association between uncertainties and the readability of disclosures. Given the current relevance of political risks, I contribute to the ongoing debate about whether managers tend to assist investors in these circumstances or will try to mislead them. The results support literature on the opportunistic perspective (Li, 2008; Ajina et al., 2016; Lo et al., 2017).

Second, this study complements recent literature examining the impact of political risks. Drawing on the suggestion from Hassan et al. (2017), I attempt to identify and quantify the effects of political risks within the context of disclosures. Even though recent literature suggests that managers attempt to assist investors (Bird et al., 2017; Boone et al., 2017; Nagar et al., 2019), we know little about this motivation when it comes to firm-specific rather than aggregated (e.g. national level) political risks. My findings are relevant for economists, business leaders, and politicians, as they fuel the ongoing debate about the effects of risks originating from politics on businesses and society. Furthermore, I contribute to literature examining the important role of managerial ability in disclosure practices (Bamber, Jian, & Wang, 2010; Baik et al., 2018; Hasan, 2018). Despite investors seeing managerial ability as positive and resulting in superior performance (Bertrand & Schroar, 2003; Demerjian et al., 2013; Andreou et al.,

7 2015), high-able managers, when confronted with political risks, have a stronger motivation to act opportunistically. This emphasises that managerial ability can backfire at the expense of investors, thereby contributing to the strand in literature discussing the disadvantages of high-able managers (Habib & Hasan, 2017).

The remainder of the paper is organized as follows. Section 2 provides the theoretical background of political risks, readability of narrative disclosures, managerial ability and board independence, which results in the hypotheses. Next, section 3 discusses the research design and methods. In section 4, the empirical results of this study are presented. Subsequently, section 5 provides the robustness checks. Finally, section 6 comprises the conclusion and discussion.

THEORY & HYPOTHESES Political risk and the readability of narrative disclosures

Recent events have fuelled debates and negotiations about the effects of risks originating from the government on investments, employment, and other aspects of firm behaviour (Hassan et al., 2017). In a recent analysis of topics discussed most extensively in the board room of 3.000 public companies in the US, EY (2018) concludes that corporate boards consider political uncertainty to be ‘the greatest near-term risk’ their company faces. Broadly speaking, political risks are defined as the uncertainties regarding government’s future actions1 _{(Pastor &}

Veronesi, 2013). Prior studies have primarily applied two techniques of measuring aggregate political risk: one based on newspaper coverage of political topics (Pastor & Veronesi, 2013; Baker, Bloom, & Davis, 2016; Nagar et al., 2019) and the other on uncertainty shocks during events such as gubernatorial elections (Li & Born, 2006; Boutchkova, Doshi, Durnev, & Molchanov, 2012). Despite calls from prior studies to increase our ability to interpret the impact of political risk, few studies to date examine how these risks vary between companies. However, very recently Hassan et al. (2017) provided the first large scale measurement of firm-level political risk2 _{based on the time that managers devote to discussing political risks during}

quarterly conference calls. This enables researchers to expand current knowledge about the impact of political risks. Accordingly, Hassan et al. (2017) underscore the importance of future research to improve our ability to identify and quantify causal effects of political risk.

Political risks influence the decisions of managers because they cause changes in the operating environment of the company (e.g. changing monetary policies, trade guidelines, or

1 In this research, I assume that political risks potentially have negative impacts on the prospects of a company. 2 In the remainder of this paper, the term ‘political risk’ refers to firm-level political risks.

8

industry regulations) (Julio & Yook, 2012). In response, managers take measures to reduce their exposure to these political risks. For example, firms’ investment behaviour becomes more conservative (Bloom, Bond, & Van Reenen, 2007; Handley & Limao, 2015), and their hiring rates fall dramatically (Bloom, 2009; Stein & Stone, 2010). A potential reason for this more conservative decision-making is that firms likely apply a strategy of ‘wait and see’ in times of political uncertainty; postponing decisions until the prospects become more clear (Bloom, 2007; Gilchrist, Sim, & Zakrajsek, 2014). In addition to delaying decisions, managers tend to actively manage those political risks deemed relevant by performing lobbying activities, forging links to politicians, and donating more to political campaigns (Hassan et al., 2017). In other words, managers, when confronted with political risks, attempt to ensure the future success of their company by preparing themselves for the potential negative consequences.

Similarly, investors react to changes in a company’s exposure to political risks because of its association with the value of the firm (Pastor & Veronesi, 2012, 2013). Specifically, evidence suggests that risk premia are revised upwards for companies that are exposed to more risks in their operating environment (Hassett, 2011). This demonstrates that investors demand higher returns for investments that bear more uncertainties – for example, regarding the outcomes of political events (Pastor & Veronesi, 2013). Subsequently, when faced with increased uncertainty, investors will, to make well-informed decisions, demand more information about the prospects of the firm which ultimately reduce information asymmetries (Lev & Zarowin, 1999; Nagar et al., 2019). Put differently, the exchange of information among managers and investors is essential in ensuring the effective allocation of capital between companies and investors (Healy & Palepu, 2001).

Corporate disclosures are the most important communication-channel that managers use to provide information to investors. Specifically, narrative disclosures are important as they describe, discuss and evaluate critical contextual and non-financial information that provides a more meaningful understanding of the company (PwC, 2007; Jones & Smith, 2014). From the investor’s perspective, the narrative disclosures are crucial to understand how political risks affect the firm’s prospects. In order to be effective, users should be able to extract and process the relevant information from the narrative disclosures (Bonsall & Miller, 2017). In the accounting literature, this is referred to as the readability of a text, which is defined as the “measure of the ability of the reader to decipher the intended message of a text” (Loughran & McDonald, 2016, p. 1188). For example, using frequent and complex structures in a text hampers investor’s ability to extract the relevant information (De Souza et al., 2019).

9 This underscores the economic significance of readable narrative disclosures: “a message which is not understood is useless either for decision-making or for monitoring” (Smith & Taffler, 1984, p. 139). Readable narrative disclosures make it easier for investors to extract the relevant information. In turn, this improves their ability to make well-informed assessments about how political risks influence the prospects of firms and adjust their decisions accordingly. Yet, it depends on managers’ motivation whether the investor is being assisted in this manner. Literature on this topic shows two competition schools of thought: the assumption that managers disclosure motives are opportunistic and include impression management (opportunistic perspective), or the assumption that managers will try to provide value-relevant information to improve the decision-making process (informational perspective) (Merkl-Davies & Brennan, 2007; Riedle & Srinivasan, 2010). Consistent with this dichotomy, a priori it is ambiguous how political risks affect the readability of narrative disclosures. In other words, do narrative disclosures provide an opportunity to assist the investors, or are they used to mislead them?

Opportunistic perspective

Most of the studies related to the readability of narrative disclosures adopt the opportunistic perspective: managers exploit the information asymmetry between the manager and the investors to engage in biased, i.e., self-serving, reporting (Merkl-Davies & Brennan, 2007). The fundament for this perspective is drawn from agency theory of Jensen & Meckling (1976) which assumes that managers and investors act in an opportunistic manner to maximize their own utility (Deegan & Samkin, 2009). Because of a more direct involvement in the daily operations of the company, managers have an information advantage over the investors (An et al., 2011). Accordingly, this provides them with the opportunity to engage in self-serving behaviour – the choice not to share private information with investors. Resultingly, agency costs arise which are born by both parties. That is, investors suffer from higher estimation/information risks because they are less informed (i.e. bigger information asymmetry). Consequently, these investors expect to be compensated by demanding a higher return on their investment, which involves that managers suffer from higher costs of capital (Healy & Palepu, 2001).

Since this is not favourable for either party, narrative disclosures, as “a channel for managers to convey contextual information about their firms to market participants”, potentially provide a solution (Merkley, 2014, p. 725). In these sections, managers describe, discuss, and evaluate their company’s exposure to risks and the way in which they manage

10

these (Jones & Smith, 2014). Accordingly, the estimation/information risk is reduced as investors can evaluate whether the information aligns with their expectations (Healy & Palepu, 2001). Although one problem is solved, another is created. That is, the self-interest manager does not always profit to the maximum by bridging the information gap. Another, more profitable option, may be to conceal information about negative organizational outcomes (Abrahamson & Park, 1994). In a recent study, Bernard (2016) provides corroborating evidence; the benefits of reducing information asymmetry do not outweigh the costs for financially constrained firms. In a similar manner, managers of firms exposed to political risks may not profit from sharing all private information with investors. This suggestion lays the foundation for the opportunistic perspective, which assumes that the information shared by the self-interest manager is well thought out (i.e. present the company in the best possible light).

In the accounting literature, presenting or writing information in a certain way to be perceived more favourably by the reader is called ‘impression management’ (Hooghiemstra, 2000). This entails “managers opportunistically taking advantage of information asymmetries” (Merkl-Davies, Brennan, & McLeay, 2011, p. 318). In this case, narrative disclosures are vehicles by which managers present a self-interested view of corporate performance to investors. That is, “managers are assumed to manipulate the presentation and disclosure of information in corporate narrative documents resulting in reporting bias to mislead the investors about organisational outcomes” (Merkl-Davies, Brennan, & McLeay, 2011, p. 318). In general, the consensus in the impression management literature is that managers attempt to emphasize good news and conceal bad news (Clatworthy & Jones, 2006; Cho, Roberts, & Patten, 2010). For example, Fiol (1995) find that internal documents of companies tend to include more references to threats than the documents presented to external stakeholders. Moreover, Clatworthy & Jones (2006) find evidence suggesting that when corporate performance is poor, managers tend to use fewer quantitative results, fewer personal references, and more passive sentences in their disclosures to manage the impression of investors. Collectively, these results indicate that managers adjust their disclosure practices to give investors a more positive impression. Accounting scholars point specifically to narrative disclosures as being suitable for impression management since these disclosures are not subjected to strict regulations (see Merkl-Davies & Brennan (2013) for a complete review).

A specific form of impression management is obfuscation, which includes a writing-style that obscures the intended message by lowering the readability of narrative disclosures (Courtis, 2004). This entails using long-sentences, non-relevant information, or complex grammatical structures. Due to the cognitive limitations of attention and processing power, information

11 written in a less readable form is less easily extracted and processed by investors (Hirshleifer & Teoh, 2003). Drawing on agency theory, Bloomfield’s (2002) ‘management obfuscation hypothesis’ explains that these limitations are used by managers to maximize their own utility. His suggestion implies that managers are likely to lower the readability of information that threatens the value of the firm or their reputation. By doing so, the self-interested manager maximizes his utility since this is often linked to firm performance – among which the value of the firm (Hooghiemstra et al., 2017). The first large scale study providing support for this hypothesis comes from Li (2008), who finds that managers lower the readability of narrative disclosures when firm performance is bad. Subsequent research confirms Bloomfield’s (2002) hypothesis in several other circumstances (Ajina et al., 2016; Lo et al., 2017; Asay et al., 2018; De Souza et al., 2019). In these situations, managers feel threatened by the private information they need to share with investors. Hence, disclosures become less readable to make it more difficult for investors to extract and process this information. Building on these findings, the same may be the case in a situation characterized by political risks; managers feel that their reputation (i.e. ability to manage these risks) and the value of the firm is threatened, so he lowers the readability of narrative disclosures to prevent this from happening.

In sum, neither managers nor investors appreciate political risks. As such, investors assess the negative consequences of political risks on the company’s prospects based on the information in narrative disclosures. The self-interested manager knows that if this information is considered negative, his utility will be hampered. Therefore, he lowers the readability to hide this information (i.e. present the company in the most favourable way as possible). Accordingly, I expect that if a company is confronted with more political risks, managers will lower the readability of narrative disclosures and hypothesize:

𝐇𝟏𝐚: Companies that face more political risks will have less readable narrative disclosures.

Informational perspective

The informational perspective assumes that managers will “contribute to useful decision making by overcoming information asymmetries between managers and firm outsiders” (Merkl-Davies & Brennan, 2007, p. 119). The fundament for this perspective is drawn from signalling theory. In contrast to agency theory, this view assumes that managers bridge the information gap because they benefit from sending their private information to interested parties (Anctil, Dickhaut, Kanodia, & Shapiro, 2004; An et al., 2011; Jiang & Yang, 2017). According to prior literature (Whiting & Miller, 2008; Scott, 2015), high-quality companies are

12

motivated to credibly reveal their superiority compared to their low-quality counterparts. By doing so, they provide investors with considerably useful information which enables them to distinguish between high- and low-quality companies, ultimately improving their return on investment. This illustrates another important role of narrative disclosures: communicate the underlying quality differences to investors. Yet, literature accentuates that investors are sceptical in accepting alleged high qualities of companies (Watson, Shrives, & Marston, 2002). Therefore, high-quality firms should communicate information in an observable credible way. A crucial aspect is that this way should be (more) costly for low-quality firms to mimic (Scott, 2015). One possible way to achieve this is through high-quality disclosures which illustrate that managers “do not shrink back to tell their current and future performance” (Eccles, 2001, p. 192). By providing high-quality disclosures (i.e. reducing information asymmetry), managers illustrate that they attempt to assist investors in understanding the high-quality of the company (Nagar et al., 2019). In contrast, it is costly for low-quality firms to imitate this disclosure practice since it makes their low-quality more apparent. This captures the essence of the informational perspective: managers enhance the quality of disclosures to effectively communicate their high-quality to investors.

Prior research found evidence confirming this notion. For example, prior studies find that managers of high-quality firms provide voluntary disclosures to assist investors in processing and understanding the information (An et al., 2011; Guay et al., 2016). Another strand in literature examines why managers voluntary certify their financial statements (Zhang & Wiersema, 2009; Marinovic & Sridhar, 2015). They conclude that this behaviour is motivated by the need to add credibility to the information they disclose. Furthermore, Riedle & Srinivasan (2010) examine management’s disclosure practices when reporting special items in disclosures (i.e. unusual or infrequent items). They demonstrate that managers present information in footnotes to assist investors in understanding the economic implications of these items. In another study, Merkley (2014) examines how managers present the information in R&D disclosures. He demonstrates that managers, in order to maximize the firm value, tend to “modify R&D disclosures…to provide relevant information to the market participants” (Merkley, 2014, p. 727). More specifically in the context of readability, Rennekamp (2012) suggests that more readable disclosures result in stronger price reactions – indicating that investors accept information with more ease. Accordingly, enhancing the readability of narrative disclosures could be an effective mean for high-quality firms to credibly communicate their superiority. In line with this suggestion, Bonsall & Miller (2017) find that managers that present more readable narrative disclosures profit from a lower cost of capital. Unlike

13 Bloomfield’s (2002) ‘management obfuscation hypothesis’, these studies demonstrate that enhancing the readability is a beneficial disclosure practice for managers.

In times of political risks, investors, when becoming aware of political risks, generally revise the value of firms downwards (Pastor & Veronesi, 2012). As a reaction, in these uncertain times, when adopting the informational perspective, managers are inclined to reveal their high-quality to mitigate this loss in firm value. According to Merkley (2014), narrative disclosures are a useful tool to do so: managers communicate information about business fundamentals to investors, including political risks. Even though political risks themselves might be considered negative, managers can illustrate how their exceptional way of dealing with these risks ensures future success of the company. For example, by providing a description of specific dialogues with government agencies or lobbying activities. Nonetheless, literature points to the fact that in order to be effective, the quality of the narrative disclosures should be high. As such, prior studies examined several attributes of financial reporting quality and find corroborating results (Bird et al., 2017; Boone et al., 2017, Nagar et al., 2019); managers enhance the quality of financial reporting when confronted with political risks. However, no study to date examines an important attribute of financial reporting quality in this context: the readability of narrative disclosures. Drawing on previous findings, I expect managers to increase the readability of narrative disclosures to effectively communicate their superiority to investors. Therefore, I hypothesize that:

𝐇𝟏𝐛: Companies that face more political risks will have more readable narrative disclosures.

The moderating effects of high-able managers

Prior accounting research often ignored the impact of management heterogeneity on corporate outcomes. However, recent studies highlight the effects of various managerial characteristics on firm performance (Adams, Almeida, & Ferreira, 2005; Gabaix & Landier, 2005; Kaplan, Klebanov, & Sorensen, 2012). The starting point for this strand in literature is the study by Bertrand & Schroar (2003) who examine how the presence of managerial fixed effects explain corporate practices. Subsequently, studies focus on explaining these managerial fixed effects by considering different observable characteristics of a manager such as gender, education, and experience (Gul, Wu, & Yang, 2013). More recently, research focuses on the role of managerial ability, which is defined as the “relative efficiency and ability of managers in transforming resources in revenues” (Hasan, 2018, p.3). Prior studies found associations between managerial ability and various corporate practices: tax avoidance (Koester, Shevlin, &

14

Wangerin, 2017), mergers and acquisitions (Chen & Lin, 2018) and quality of disclosures (Cantrell, 2013).

Although the effects of managerial ability have been explored frequently in the past years, until recently, no study has examined its impact on the readability of narrative disclosures. However, Hasan (2018) filled this gap in literature and found that “firms with more able managers are less likely to cloud their superior performance” (Hasan, 2018, p. 4). Rather, he explains, they will be more forthcoming to signal their superior ability through more readable disclosures. In contrast, low-able managers tend to conceal this fact. The fundamental argument supporting this notion comes from the early work of Trueman (1986). He suggests that high-able managers have a superior understanding of the environment and want to communicate this to investors. Later studies (Baik, Farber, & Lee, 2011; Demerjian et al., 2013) provide evidence that managerial ability is positively associated with the quality of disclosures – confirming that high-able managers want to show their superiority to investors.

Another strand in literature focuses on the way high-able managers deal with risks. The consensus in literature is that these managers, due to their high ability, have a superior understanding of the operating environment of the company (Koester et al., 2017). This enables them to identify and respond to risks in a superior way. Consistently, Andreou et al. (2015) examine how differences in managerial ability explain how well companies are prepared against economic shocks. They find that during financial crises, high-able managers are better able to protect the prospects of their company. In a similar manner, managerial ability may explain how well companies are prepared against political risks. Ultimately, one would argue that high-able managers are more inclined to communicate this to investors instead of hiding it.

In line with the findings of Andreou et al. (2015) and Hasan (2018), I expect that managerial ability moderates managers’ tendency to act in accordance with either the opportunistic or informational perspective and hypothesize:

𝐇𝟐: The negative (positive) association between political risks and the readability of narrative

disclosures is weakened (strengthened) by managerial ability.

The moderating effects of independent directors

Corporate governance can be viewed as “a set of contracts that help align the actions of managers with the interest of shareholders” (Armstrong, Guay, & Weber, 2010, p. 181). In line with agency theory of Jensen & Meckling (1976), it protects investors against the self-serving behaviour of managers. The board of directors is considered as one of the fundamental

15 mechanisms as it solves problems that are inherent to managing organizations (Hermalin & Weisbach, 2003). Namely, as described in the early research of Fama (1980), the task of the board of directors is to be the ‘ultimate internal monitor’ of managers. As part of their duties, directors pressure managers to disclosure their private information to investors (Raheja, 2005). In practice, this entails that directors oversee all decisions of management and ensure they align with the interest of investors, including those related to disclosure practices (Beasley, 1996). Subsequently, literature has emphasized that appointing a board of directors improves the quality of disclosures (Cohen, Krishnamoorthy, & Wright, 2004).

However, the effectiveness of the board of directors in its role as internal monitor may be constrained because of their dependence on managers (Kim, Mauldin, & Sukesh, 2014). That is, in practice, their independence is jeopardized due to two reasons: (1) selection or appointment of directors by management, and (2) directors might be former employees or have business relationships with the company (Brennan, 2006). Ultimately, their dependence on management will result in costlier and less efficient monitoring (Adams, Hermalin, & Weisbach, 2010). In contrast, directors are considered independent if they are non-employees with no ties to the firm except in their roles as director (Cohen et al., 2004). The importance of this characteristic is frequently emphasized in early literature (Fama, 1980; Fama & Jensen, 1983). A director who is independent from management only values his reputation as monitoring expert, which entails being a strong internal decision manager (Fama & Jensen, 1983; Armstrong et al., 2010). Moreover, the economic significance is underscored by standard setters and regulators, who encourage or require companies to appoint directors that are independent from management (SEC, 2003).

One strand in accounting and governance literature examines how independent directors can pressure managers to disclose high-quality information. For example, the early research by Forker (1992) proposes that the inclusion of independent directors should result in more comprehensive and higher quality disclosures. Later studies (Beasley, 1996; Chen & Jaggi, 2000; Gul & Leung, 2004) find evidence that confirms this proposition. In particular, Abraham & Cox (2007) examine whether an independent board of directors is associated with the amount of information a company discloses about their corporate risks. In line with the rationale described in the study of Eng & Mak (2003), they find that independent directors are better able to pressure managers to reveal their private information. Similarly, Laksmana (2008) describes that boards with the power to act independently provide more information to investors. Collectively, these findings suggest that an independent director is an effective governance mechanism to enhance the quality of disclosures. Despite the apparent association, few studies

16

to date focus on the relationship between board independence and the readability of narrative disclosures. Filling this gap, I examine whether manager’s opportunity to act in accordance with either the opportunistic or informational perspective is moderated by an independent board of directors. Accordingly, I hypothesize:

𝐇𝟑: The negative (positive) association between political risks and the readability of

narrative disclosures is weakened (strengthened) by board independence.

RESEARCH DESIGN & METHODS Sample selection and data collection

The sampling procedure starts with the data about political risk of Hassan et al. (2017) who analyse the quarterly conference calls of firms listed in the United States. As a result, I use an initial sample consisting of 208,794 observations from 9,478 firms for the period 2002 to 2016. To retrieve data about the readability of narrative disclosures I made use of the measures provided by Loughran & McDonald (2016). Their dataset contains multiple measures of textual aspects from narrative disclosures in annual and quarterly reports. Data on managerial ability comes from Demerjian, Lev, & McVay (2012). The remaining data about governance-, firm-, and CEO-characteristics is collected from Compustat, MSCI GMI Ratings, ExecuComp and BoardEx.

In line with prior studies (Lo et al., 2017; De Souza et al., 2019), I exclude companies active in the financial and insurance sectors due to their different operating and financing structures (representing 16,787 observations). Furthermore, I match all observations with data about readability and as a consequence I have to delete 60,612 observations. The remaining observations are merged with data about managerial ability and board independence. Consequently, an additional 22,690 and 39,285 observations are omitted from the sample, respectively. By merging the data about the control variables, I exclude 18,002 observations from the sample due to missing data. Finally, I delete 993 observations from companies that have less than four consecutive observations in total. In summary, the final sample comprises of 50,425 observations representing 1,443 unique companies. Table 1 illustrates the construction of the final sample.

17 Table 1. Sample used in analyses.

Observations

Initial sample from Hassan et al. (2017) 208,794

Firms in financial sector (SIC code: 6000-6199) Firms in insurance sector (SIC code: 6300-6411) Missing data on readability

Missing data on managerial ability Missing data on board independence Missing data on other variables

Companies with less than four observations Final sample (10,378) (6,409) (60,612) (22,690) (39,285) (18,002) (993) 50,425 Measurement of dependent variable

As a measure of readability of narrative disclosures, many studies to date use the Fog-Index (Ajina et al., 2016; Guay et al., 2016; Hooghiemstra et al., 2017). However, Loughran & McDonald (2014) argue and demonstrate that the wide application of this technique is inappropriate since the Fog-index is likely to be inaccurate and poorly specified. They suggest that general measures of readability are more suitable; for example, the gross file size of disclosures. Nevertheless, variability in the gross file size may be explained by factors not related to the textual aspects (e.g. encoded images and tables). As such, I use the natural logarithm of the net file size (FSIZE), which is the plain text file, as a proxy for the readability of narrative disclosures. Another simple proxy for readability applied by prior studies (Li , 2008; Guay et al., 2016; De Souza et al., 2019) is the natural logarithm of the number of words (NWORDS). That is, longer texts may be used to influence the reader’s ability to extract and process the information. In sum, for both proxies of readability, a higher number implies that narrative disclosures are harder to read.

Measurement of independent variable

Building on research by Hassan et al. (2017), I measure political risk (PRISK) based on the amount of words managers use to discuss political risks in their quarterly conference calls with financial analysts. Conference calls are a well-suited source of data since risks are often the subject of the discussion (Hollander, Pronk, & Roelofsen, 2010). In order to distinguish between language associated with political risks versus non-political risks, Hassan et al. (2017)

18

create two sets of words related to politics (from political science textbook and newspapers) and non-politics (from accounting textbook). Based on the transcripts of the conference calls, a pattern-based sequence-classification determines the relative amount of times both sets are used in combination with synonyms for risk. Ultimately, having a higher share of the quarterly conference call related to political risks implies that political risks in that period are higher. Measurement of moderating variables

Drawing on the study by Demerjian et al. (2012), I measure managerial ability (MANAB) by focussing on the inputs and outputs of companies. Specifically, the technique divides firm efficiency drivers in two groups: those caused by firm-specific drivers, and those caused by managerial ability. Reducing the overall firm-efficiency by those efficiencies caused by non-personal attributes (e.g. firm size, cash available, and firm market share) isolates efficiencies caused solely by managerial ability. Since the technique is only applicable on an annual basis, I assume that the measure remains stable during all four quarters of a year.

Consistent with prior studies (Eng & Mak, 2003; Laksmana, 2008) I measure board independence (IND) by the percentage of outside directors on the board. In general, a higher percentage of outside directors indicate that the board is able to act more independently from management. To construct this measure, directors are divided into three groups: insiders, outsiders, and affiliated. Insiders are current employees of the company, while outsiders are only connected to the company through their directorship. The latter group, according to the NYSE regulations (2019), are directors that are directly or indirectly connected with the company through one or more intermediaries, making them not independent of management. Measurement of control variables

To avoid biased results, I include control variables that might affect the readability of disclosures. Firstly, in addition to board independence, other governance attributes are likely to influence how companies write their disclosures. For example, boards that consist of more directors are better able to reduce information asymmetries (Upadhyay & Siram, 2011) since they possess better monitoring techniques and more knowledge (Ji, 2016). In a similar manner, larger audit committees are more efficient in monitoring the financial reporting process of the company (Klein, 2002). Hence, I control for the impact of board size (BSIZE) and audit committee size (ACSIZE) on the readability of narrative disclosures. Furthermore, prior research found that a company’s information environment improves when boards are more gender

19 diverse (Gul, Srinidhi, & Ng, 2011). In line with this finding, I control for board diversity (DIV) as measured by the number of female directors.

Secondly, since the CEO has a dominant position within the firm they manage and can therefore influence many disclosure practices, I control for various CEO characteristics. In line with prior studies (Laksmana, Tietz, & Yang, 2012; Hooghiemstra et al., 2017), higher CEO compensation (COMP) may indicate a stronger tendency to lower the readability of narrative disclosures to hide the fact that they are overpaid. I use the total of remuneration as a measure of CEO compensation as reported in the ExecuComp database, which includes both cash and stock (option) payments. Moreover, research suggests that CEO tenure (TEN) and CEO age (AGE) will have an impact on disclosures (Zhang & Wiersema, 2009). Namely, when time passes, CEOs get more experience, skills, and improve their decision-making process which ultimately enhances their disclosure practices. Therefore, I control for CEO age as well as CEO tenure, as measured by the number of years after appointment.

Thirdly, firm characteristics can influence the readability of narrative disclosures. For example, companies with more debt have a stronger tendency to attract and retain capital providers by disclosing information that aligns with their expectations. Consistent with prior research (Ajina et al., 2016; De Souza et al., 2019), I control for firm leverage (LEV), as measured by the sum of current and non-current liabilities scaled by total assets. Another aspect related to the financial statements which may influence the readability of narrative disclosures is firm size (SIZE). That is, larger companies tend to have more complex underlying operations which result in longer and more complex disclosures (Ajina et al., 2016). Consequently, I control for firm size as measured by the natural logarithm of total assets. Furthermore, Bloomfield (2002) suggests that less profitable companies exhibit a shift to less readable narrative disclosures as an attempt to cover up bad results. In other words, investors are more likely to question the viability of the business if profitability (PRFT) is lower. As such, I control for firm profitability as measured by the net income scaled by total assets. Similar to Nagar et al. (2019), I predict that political risks often arise from international tensions. Therefore, companies that realize more revenues abroad may have a greater exposure to political risks and react more strongly. As such, I control for foreign exposure (FOREX) as measured by the foreign pre-tax income scaled by total pre-tax income. At last, I winsorize variables at the top and bottom 1% to mitigate possible bias due to outliers in the data.

20

Analysis models and techniques

Given the characteristics of the sample, I use panel data analysis which includes both an individual dimension (company), as indicated by 𝑖, as well as a period dimension (quarter) noted as 𝑡 (Ajina et al., 2016). The Hausman test is used to determine whether a random or fixed effect model is more applicable to test the hypotheses. Based on the test outcomes, I conclude that for all hypotheses fixed effect models are desirable. This allows to control for changes in variables over time which may affect the test outcomes.

More specifically, the industry in which a company operates, or the year of disclosure, may influence the dependent variable: readability of narrative disclosures. For example, companies in a certain industry may reveal lower levels of readability due to the complex nature of the underlying operations. Likewise, readability may fluctuate over time due to regulatory or demographic trends. To control for industry or year fixed effects, I use year and industry dummies. First, I identify each firm’s primary digit SIC code and create 9 industry dummy codes. In a like manner, I group observations based on the 14 fiscal years of the sample to control for year-fixed effects. Accordingly, I test and confirm that industry and year fixed-effects have significant impact on the test outcomes and thus should be included as dummy variables.

Furthermore, I check whether heteroscedasticity or autocorrelation are present in the data. The Breusch-Pagan/Cook-Weisberg test demonstrates whether the variance in the measure for narrative disclosure readability is equal across the observations of political risk. I conclude that the variance is unequally distributed. In addition, I test and confirm that autocorrelation is present in the data by performing the Wooldrige test. To address both issues, I cluster the standard errors based on firm-level in the panel data analysis (Hoechle, 2019).

I argue that fluctuations in political risk explain the readability of narrative disclosures. Since literature is ambiguous about whether political risk and the readability of narrative disclosure are positively (informational perspective) or negatively (opportunistic perspective) correlated, the first hypothesis (H_1a and H_1b) tests whether there is an association between both variables using the following empirical model:

(𝑁𝑊𝑂𝑅𝐷𝑆; 𝐹𝑆𝐼𝑍𝐸)𝑖,𝑡 = 𝛽0+ 𝛽1𝑃𝑅𝐼𝑆𝐾𝑖,𝑡+ ∑ 𝛽𝑘 𝐶𝑂𝑁𝑇𝑅𝑂𝐿𝑆 + 𝜀𝑖,𝑡

[Model 1] where suppressing subscripts, CONTROLS is the set of control variables, and 𝜀 is the error term. Within model 1, the variables of interest are PRISK, the significance and the sign of 𝛽1. On the

21 one hand, a positive coefficient confirms the opportunistic perspective (H_1a). That is, increased PRISK is associated with higher levels of NWORDS and FSIZE. On the other hand, a negative coefficient corroborates the informational perspective (H_1b), implying that managers reduce the number of words and size when political risk increases.

To test the second (H_2a and H_2b) and third hypotheses (H_3a and H_3b), I examine the moderating effects of managerial ability and board independence on model 1 and use the following empirical models:

(𝐹𝑆𝐼𝑍𝐸; 𝑁𝑊𝑂𝑅𝐷𝑆)𝑖,𝑡= 𝛽0+ 𝛽1𝑃𝑅𝐼𝑆𝐾𝑖,𝑡+ 𝛽2𝑃𝑅𝐼𝑆𝐾𝑖,𝑡∗ 𝑀𝐴𝑁𝐴𝐵𝑖,𝑡+ ∑ 𝛽𝑘 𝐶𝑂𝑁𝑇𝑅𝑂𝐿𝑆 + 𝜀𝑖,𝑡

[Model 2] (𝐹𝑆𝐼𝑍𝐸; 𝑁𝑊𝑂𝑅𝐷𝑆)𝑖,𝑡= 𝛽0+ 𝛽1𝑃𝑅𝐼𝑆𝐾𝑖,𝑡+ 𝛽2𝑃𝑅𝐼𝑆𝐾𝑖,𝑡∗ 𝐼𝑁𝐷𝑖,𝑡+ ∑ 𝛽𝑘 𝐶𝑂𝑁𝑇𝑅𝑂𝐿𝑆 + 𝜀𝑖,𝑡

[Model 3] where suppressing subscripts, 𝑃𝑅𝐼𝑆𝐾_𝑖,𝑡∗ 𝑀𝐴𝑁𝐴𝐵𝑖,𝑡 and 𝑃𝑅𝐼𝑆𝐾𝑖,𝑡∗ 𝐼𝑁𝐷𝑖,𝑡 indicate the

moderating effects of managerial ability and board independence. Depending on the association found in model 1, I expect that MANAB and IND either strengthens or weakens the relationship between political risk and the readability of narrative disclosures. Specifically, I expect 𝛽2in

model 2 and 3 to be significant. In both cases, I expect the coefficients to be negative – implying that higher managerial ability and a more independent board of directors makes managers more inclined to act in accordance with the informational rather than the opportunistic perspective.

EMPIRICAL RESULTS Descriptive statistics

Table 2 presents the descriptive statistics of the variables for the full sample. The descriptive statistics of both proxies for readability (NWORDS; FSIZE) are consistent with prior evidence examining the readability of disclosures (Guay et al., 2016; De Souza et al., 2019). For example, on average, the narrative disclosures in the sample comprises of 27,971 words. Consistent with Hassan et al. (2017), the natural log of political risk (PRISK) spikes between 2008 and 2010 – during the financial crisis and U.S. presidential elections. The measure of managerial ability (MANAB) of the sample aligns with the study of Demerjian et al. (2012), indicating that on average 1.2% of the firm efficiency is due to managerial ability. The average board consists of 9 directors of whom 73% are independent (IND), this aligns with prior studies (Klein, 2002; Laksmana, 2008). Consistent with Gul et al. (2004), the average number of female

22

directors on the board is 2 (DIV). The CEO-characteristics show that the average CEO in my sample is 57 years old (AGE) and has been the head of the company for 8 years (TEN). Approximately 20% of a firm’s assets are financed with debt (LEV), and the profit amounts to 5% of the total assets (PRFT). Lastly, the foreign exposure (FOREX) indicates that companies realize on average 27% of their income outside the U.S.

Table 3 presents the Pearson correlation matrix for all pairwise combinations of variables in the overall sample. Besides the two proxies for readability (NWORDS; FSIZE), all variables have acceptable coefficients (less than 0.7). Nevertheless, I perform a more in-depth multi collinearity test to ensure that the tests are appropriate, and the results interpreted correctly. For all variables, I examine the variance inflating factors (VIFs) which varies between 1.01 and 2,54 – indicating that multi collinearity is not a problem.

23 Table 2. Descriptive statistics.

Full sample, N = 50,425

Variable Mean Std. dev. Q1 Median Q3

NWORDS 9.960 .723 9.400 9.863 10.487 FSIZE 12.004 .706 11.456 11.902 12.521 PRISK 4.544 1.766 3.776 4.935 5.751 MANAB .0121 .154 -.083 -.030 .060 IND .728 .150 .636 .750 .857 BSIZE 9.032 2.097 7.000 9.000 10.000 ACSIZE 4.682 3.455 3.000 4.000 6.000 DIV 1.814 1.706 1.000 1.000 3.000 COMP 8.153 1.011 7.489 8.232 8.867 AGE 56.620 7.670 51.000 56.000 61.000 TEN 8.127 7.905 2.000 6.000 11.000 LEV .204 .179 .0324 .185 .311 SIZE 7.626 1.550 6.490 7.517 8.643 PRFT .046 .0963 .0208 .055 .092 FOREX .268 .719 0.000 .075 .502

24

Table 3. Pearson correlation matrix.

Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) NWORDS 1.000 (2) FSIZE 0.998 1.000 (3) PRISK 0.062 0.063 1.000 (4) MANAB 0.065 0.063 0.046 1.000 (5) IND 0.048 0.048 0.010 0.024 1.000 (6) BSIZE 0.076 0.085 0.025 0.055 0.111 1.000 (7) ACSIZE 0.021 0.022 0.008 0.020 0.101 0.219 1.000 (8) DIV 0.085 0.087 -0.018 0.088 0.222 0.452 0.261 1.000 (9) COMP 0.147 0.154 0.046 0.176 0.204 0.403 0.170 0.310 1.000 (10) AGE -0.035 -0.033 0.001 -0.025 -0.070 0.086 0.042 0.023 0.028 1.000 (11) TEN -0.038 -0.039 0.014 0.022 -0.136 -0.085 -0.047 -0.124 -0.067 0.496 1.000 (12) LEV 0.120 0.126 -0.001 -0.113 0.034 0.209 0.104 0.158 0.231 0.036 -0.041 1.000 (13) SIZE 0.181 0.192 0.052 0.187 0.146 0.595 0.223 0.390 0.657 0.122 -0.051 0.335 1.000 (14) PRFT -0.102 -0.100 -0.024 0.190 0.007 0.089 0.030 0.072 0.160 0.022 0.018 -0.167 0.123 1.000 (15) FOREX 0.036 0.037 -0.008 0.009 0.041 0.052 0.029 0.027 0.092 0.006 -0.024 -0.016 0.107 0.059 1.000 Note: This table contains information about the pairwise correlation between variables of the panel regression on the readability of narrative disclosures. Correlation with p-values less than 0.01 are indicated in bold. The matrix comprises all number of observations, equals 50,425.

25 Test of H1

Table 4 reports the results of the panel data analysis which tests the three hypotheses of this study. For the first hypothesis, I examine whether the readability of narrative disclosures is significantly associated with the level of PRISK. Next, based on the sign of the coefficient, I determine which perspective (opportunistic or informational) prevails. Column 1 and 2 present the results of the first model of this study. That is, it examines whether several determinants of narrative disclosure readability are related with NWORDS or FSIZE. After controlling for other explanatory factors (i.e. governance-, CEO- and firm-characteristics), I find that higher levels of PRISK are significantly associated with higher NWORDS (0.0050, p<.01) as well as greater FSIZE (0.0051, p<.01). In contrast to the results found by Nagar et al. (2019), who use a comparable sample of U.S. firms and focus on voluntary disclosure, I find evidence corroborating the opportunistic perspective; managers tend to conceal information in narrative disclosures by lowering the readability in politically uncertain times. As suggested by Merkley (2014), firms may increase the disclosure quantity without providing extra relevant information. Hence, the difference in findings with the study of Nagar et al. (2019) may be explained by the fact that they do not include the readability component of the disclosures in their analysis.

The results also confirm prior literature’s findings regarding the association between firm-characteristics and the readability of narrative disclosures. Specifically, LEV is significant positively (0.2020, p<.01)3 _{correlated with NWORDS and FSIZE, corroborating the general}

notion that companies financed with more external financing have harder to read narrative disclosures (Ajina et al., 2016; De Souza et al., 2019). In a similar manner, SIZE is significantly positively (0.0882, p<.01) correlated with both proxies for readability, indicating that the complex underlying operations of bigger firms result in more difficult to read disclosures (Ajina et al., 2016). Moreover, I confirm Bloomfield’s (2002) suggestion that companies with lower profitability (PRFT) have a higher tendency to lower the readability of disclosures (-0.4050, p<.01). Besides firm-characteristics, I find a negative and significant (-0.0016, p<.10) correlation between AGE and NWORDS, demonstrating that older CEOs provide more readable narrative disclosures. In sum, these findings align with the notion that the circumstances in which narrative disclosures are produced influence the readability.

3_{To increase the readability of the text and given that both proxies for readability are strongly correlated, I only}

26

To summarize, I find that when conversations in conference calls are more centralized around political risks, all else being equal, the readability of narrative disclosures in the corresponding quarter decreases. In other words, the self-interested manager uses the narrative disclosures to mislead investors in political uncertain times – maximize the utility by making it harder to extract and process relevant information.

Test of H2 and H3

Subsequently, after verifying the prevailing perspective, I regress both MANAB and IND on the found association of model 1 to verify whether there is a moderating effect. Provided that this is the case, the sign of the coefficient determines if it enhances or reduces managers’ motivation to act conform the opportunistic perspective when confronted with political risks.

Firstly, to examine the effects of managerial ability, model 2 regresses the interaction (indicated as PRISK*MANAB) with both proxies for readability. Column 3 and 4 present the results of the second model of this study. Contrary to the expectation, I find that, after controlling for other determinants, higher levels of MANAB strengthen managers’ motivation to act in accordance with the opportunistic perspective (0.0294, p=.01). To properly interpret these results, Figure 1 and Figure 2 presents the interaction graphs. The figures demonstrate that when political risk increases, the readability of narrative disclosures decreases faster for companies with high-able managers than for companies with low-able managers. A potential interpretation of this interaction effect is the ‘rent extraction’ explanation. According to Hassan (2018), this entails that even though high-ability managers produce superior performance, they are over-compensated and tend to over-engage in uncertain activities. As a result, high-able managers lower the readability of disclosures to conceal this information to investors. In particular, this may be applicable when confronted with political risks; that is, investors place more emphasis on the disclosed information (Lev & Zarowin, 1999; Nagar et al., 2019). Consequently, this poses a bigger threat for high-able managers of being discovered.

Secondly, I examine if the presence of an independent board strengthens or weakens the tendency to act in accordance with the opportunistic perspective (indicated as PRISK*IND). Column 5 and 6 present the results of the third model of this study. In contrast to Laksmana (2008) suggestion that independent boards support the quality of disclosures, I conclude that IND has no interaction effect. That is, board independence does not moderate the tendency of managers to lower the readability of narrative disclosures when confronted with more political risks. Elaborating, Kim et al. (2014) posit that the effectiveness of independent directors is jeopardized by a lack of firm-specific knowledge. Prior studies (Carter & Lorsch, 2004) provide

27 corroborating evidence that in practice this notion is often applicable. Accordingly, independent directors may not be aware of firm-specific political risks, and therefore cannot effectively counteract managers’ tendency to act opportunistic.

In conclusion, the results show that high-ability managers, all else being equal, are more likely to act in accordance with the opportunistic perspective. Unlike literature suggestion, I find no moderating effect between board independence and the tendency to lower the readability of narrative disclosures when confronted with political risks.

Figure 1: The moderating effect of MANAB on the relation between PRISK and NWORDS.

Note: This figure shows the interaction effect of MANAB on the association between PRISK and

NWORDS as shown in column 3 of Table 4, with standardized control variables.

Figure 2: The moderating effect of MANAB on the relation between PRISK and FSIZE.

Note: This figure shows the interaction effect of MANAB on the association between PRISK and FSIZE as shown in column 4 of Table 4, with standardized control variables.

28

Table 4. Readability of narrative disclosure, political risk, managerial ability and board independence.

Model 1 Model 2 Model 3

Dependent variable

Independent variable NWORDS (1) FSIZE (2) NWORDS (3) FSIZE (4) NWORDS (5) FSIZE (6)

PRISK .0050 (2.69)*** (2.80)*** .0051 (2.63)*** .0049 (2.73)*** .0049 (0.20) .0016 (-0.05) -.0005 MANAB -.0742 (-1.16) -.0805 (-1.30) IND -.032 (-0.55) (-0.78) -.0446 PRISK * MANAB .0294 (2.52)** 0.0302 (2.65)*** PRISK * IND .0047 (0.43) 0.0076 (0.71) BSIZE .0002 (0.04) (0.26) .0008 (0.03) .0001 (0.26) .0008 (0.02) .0001 (0.24) .0007 ACSIZE -.0015 (-0.89) -.0010 (-0.60) -.0002 (-0.87) -.0010 (-0.58) -.0015 (-0.88) -.0010 (-0.59) DIV -.0010 (-0.24) (-0.14) -.0006 (-0.25) -.0010 (-0.16) -.0006 (-0.23) -.0009 (-0.13) -.0005 COMP -.0009 (-0.17) -.0008 (-0.09) -.0009 (-0.18) -.0005 (-0.10) -.0007 (-0.13) -.0002 (-0.05) AGE -.0016 (-1.66)* (-1.61) -.0015 (-1.68)* -.0016 (-1.62) -.0015 (-1.64)* -.0015 (-1.58) -.0014 TEN -.0007 (-0.68) -.0008 (-0.78) -.0007 (-0.70) -.0008 (-.80) -.0007 (-0.70) -.0008 (-0.80) LEV .2020 (5.80)*** .1950 (5.71)*** .2050 (5.90)*** .1980 (5.80)*** .2010 (5.77)*** .1940 (5.68)*** SIZE .0882 (13.15)*** .0894 (13.77)*** .0876 (13.08)*** .0888 (13.70)*** .0883 (13.12)*** .0894 (13.74)*** (Continued)

29

Model 1 Model 2 Model 3

Dependent variable

Independent variable NWORDS (1) FSIZE (2) NWORDS (3) FSIZE (4) NWORDS (5) FSIZE (6)

PRFT -.4050 (-9.07)*** -.388 (-9.00)*** -.4260 (-9.51)*** -.4080 (-9.45)*** -.4040 (-9.05)*** -.3880 (-8.99)*** FOREX .0023 (0.57) (0.51) .0020 (0.59) .0024 (0.53) .0021 (0.62) .0025 (0.56) .0022

INDUSTRY-FIXED EFFECTS YES YES YES YES YES YES

YEAR-FIXED EFFECTS YES YES YES YES YES YES

N 50,425 50,425 50,425 50,425 50,425 50,425

R-squared 0.0690 0.0717 0.0709 0.0735 0.0690 0.0717

Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Note: This table contains the results of the panel regression with fixed effects of political risks on the readability of narrative disclosures with the moderating effects of managerial ability and board independence. The t-statistic is based on standard errors clustered by firm levels. *, **, *** indicate significance at 1%, 5%, and 10% levels (two-tailed), respectively.

30

ROBUSTNESS CHECKS Alternative regressions

To add robustness to the results, I perform several additional analyses. Since the residuals of the variables may be correlated across firms or time, I run, in addition to the panel data analysis, a linear regression of all models with the standard errors clustered at both firm- and quarter-level (Petersen, 2009). After running the tests, I confirm that the found associations in model 1 (0.0204, p<.01) and 2 (0.0478, p<.01) still hold.

Alternative measures of readability

Considering the retrospective character of disclosures, political risks in a certain quarter may influence the readability of next quarter’s narrative disclosures (t+1). Controlling for this possibility, as well as to ease issues regarding reverse causality, I re-run model 1 with the lagged version of NWORDS and FSIZE and find evidence supporting the opportunistic perspective (0.0076, p<.01). However, using this surrogate measure of readability, the moderating interaction of MANAB in model 2 is unclear. Alternatively, high levels of PRISK may result in larger disclosures in general because bad news is more difficult to describe, and investors demand more information when news is bad (Bloomfield, 2008). If this is the case, differences in NWORDS and FSIZE do not capture the readability component of disclosures, they rather demonstrate the increase in information demand. Hence, I control for this possibility by regressing the gross file size (which includes pictures, tables, etc.), as proxy for general disclosure size, with PRISK. The results show no significant association, demonstrating that the readability of narrative disclosures specifically changes. To further alleviate concerns regarding Bloomfield’s (2008) proposition, I create two subsamples. One with observations in years with the highest PRISK (2008-2009), and the other with the lowest PRISK (2005-2006). In contrast to the alternative explanation which suggests that higher political risk in general determines the strength of the correlation, I find that in the latter subsample the correlation is stronger (low political risk: 0,0201 versus high political risk: 0,0166).

Alternative measure of political risk

I recognize that the measure of PRISK from Hassan et al. (2017) has not yet been frequently applied in empirical research. Because it is still unclear whether the measure effectively captures political risk in various contexts, I re-run all models using a surrogate