Audit committees’ social capital and financial reporting quality

Accounting and Business Research ISSN: 0001-4788 (Print) 2159-4260 (Online) Journal homepage: http://www.tandfonline.com/loi/rabr20 Audit committees’ social capital and financial reporting quality Nieves Carrera, Tashfeen Sohail & Salvador Carmona To cite this article: Nieves Carrera, Tashfeen Sohail & Salvador Carmona (2017): Audit committees’ social capital and financial reporting quality, Accounting and Business Research, DOI: 10.1080/00014788.2017.1299617 To link to this article: http://dx.doi.org/10.1080/00014788.2017.1299617 Published online: 16 May 2017. Submit your article to this journal Article views: 35 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=rabr20 Download by: [Cornell University Library] Date: 22 May 2017, At: 13:20 Accounting and Business Research, 2017 https://doi.org/10.1080/00014788.2017.1299617 Audit committees’ social capital and financial reporting quality NIEVES CARRERAa*, TASHFEEN SOHAILb AND SALVADOR CARMONAa a Department of Accounting and Management Control, IE Business School – IE University, Madrid, Spain; bDepartment of Accounting, Goodman School of Business, Brock University, St. Catharines, ON, Canada We draw on social capital theory to examine the relationship between audit committee (AC) members’ social capital and financial reporting quality. Using US data for the period 2001– 2010, our results suggest that non-AC directors’ social capital does not appear to be relevant to financial reporting quality. As far as AC members are concerned, our findings show a negative relationship between their social capital and financial reporting quality, suggesting a ‘dark side’ to social capital. Specifically, we find that sitting in multiple ACs (centrality) has a negative impact on reporting quality only for those AC members designated as financial experts. When other proxies for social capital are considered (connectedness, brokerage position and strong ties), our results show that the quality of financial reporting significantly decreases with the social capital of non-financial experts sitting in the AC. We contribute to prior research by: (i) relying on social capital theory, which is widely neglected in accounting research, (ii) using multiple metrics to capture the complex dimensions of social capital, and (iii) discriminating between the effects of financial and non-financial experts’ social capital on reporting quality. Our results suggest policy-makers might wish to limit financial experts’ multiple directorships as well as assess the actual contribution of non-financial experts to AC effectiveness. Keywords: audit committee; social capital; financial reporting quality; financial experts 1. Introduction The relationship between audit committee (AC) effectiveness and financial reporting quality has attracted significant research interest (Bédard et al. 2004) and regulatory attention (SOX 2002, SEC 2003, NYSE 2010). Accounting research drawing on agency theory indicates that directors sitting on multiple boards (busy directors) have less time to devote to one particular firm and, hence, cannot conduct effective monitoring (Sharma and Iselin 2012, Andres et al. 2013). Accordingly, the social connections of AC members affect their effectiveness as well as their ability to oversee a firm’s financial reporting processes. Although agency theory is a valid perspective to *Corresponding author. Email: [email protected] © 2017 Informa UK Limited, trading as Taylor & Francis Group 2 N. Carrera et al. examine AC effectiveness, Bédard and Gendron (2010, p. 181) also note that ‘in order to broaden our understanding of ACs, researchers need to adopt other theoretical perspectives’. Importantly, research drawing on social capital theory provides many valuable insights into the effects of a board of directors’ connections on organisations (e.g. Kim and Cannella 2008, Kor and Sundaramurthy 2009). Therefore, this perspective may advance understanding about the impact of AC members’ social connections on financial reporting quality (Horton et al. 2012). In particular, accounting research drawing on social capital theory may enhance our knowledge of the extent to which subgroups within ACs (financial and non-financial experts, see Johnson et al. 2013, Tanyi and Smith 2015), and the specific nature of AC members’ social ties, over and beyond the mere counting of cross-board seats, may affect financial reporting quality. According to Burt (2005, p. 5), social capital is ‘[t]he advantage created by a person’s location in a structure of relationships’. More specifically, an individual’s social capital is defined as ‘the sum of actual and potential sources embedded within, available through, and derived from, the network of relationships possessed by that individual’ (Nahapiet and Ghoshal 1998, p. 243). Social capital is positively related to the relevance and diversity of information available within the network as well as to the transfer of knowledge among network members (e.g. Reagans and McEvily 2003, Anderson 2008). As noted by Javakhadze et al. (2016, pp. 39– 40), a social network is the medium through which social capital is created, maintained, and used, and mainstream social capital theory focuses on the beneficial effects of social connections and social relations in achieving goals (Lin 2001). However, social capital does not necessarily bring about positive effects. As noted by Portes (1998, p. 18), ‘sociability goes both ways’, and this implies that social capital also has a ‘dark side’ (Deth and Zmerli 2010). From this perspective, finding a negative association between social capital and financial reporting quality would be considered an ‘indicator’ of a dark side of social capital (Deth and Zmerli 2010, p. 632), and a contribution to social capital theory. The ‘busyness hypothesis’ postulates that serving on multiple boards overcommits individuals and, as a result, they shirk their responsibilities (Ferris et al. 2003, p. 1088). Therefore, directors serving on multiple ACs face time constraints, and this in turn affects AC effectiveness and exerts a negative impact on financial reporting quality (Fich and Shivdasani 2006, Sharma and Iselin 2012). As there is no perfect measure of a director’s busyness (Andres et al. 2013, p. 1244), accounting research drawing on social capital theory may complement the ‘busyness hypothesis’. First, as noted by Field et al. (2013, p. 71) ‘while all busy directors face the problems of time constraints, the quality of experience and depth of networks are not equal across all busy networks’. Furthermore, the impact of time pressures on AC members’ effectiveness may differ across individuals (Tanyi and Smith 2015), so that busy directors may be more effective and beneficial to the firm than their non-busy counterparts (Field et al. 2013). Second, as a proxy for busy directors/boards, extant research uses metrics based on the number of board seats (though see Andres et al. 2013 for a significant exception). This operationalisation, in turn, is typically interpreted as a director’s time-commitment. However, these metrics have also been considered as a proxy for informational resources available to individuals and firms (e.g. Dhaliwal et al. 2010, Larcker et al. 2013). From the latter perspective, the number of social ties is considered a proxy for social capital, even though researchers in accounting and finance ‘are reluctant to use the term “social capital” and refer to network effects’ (Javakhadze et al. 2016, p. 44). We complement existing research using the number of board seats, or the number of connections to other directors, by drawing on metrics that intend to capture the complexity of directors’ social capital. We draw on social network analysis (SNA) to operationalise the complexity and multiple dimensions of social capital (Horton et al. 2012, Javakhadze et al. 2016). Therefore, we consider four constructs: the number of direct connections of an AC member (degree centrality), the ‘quality’ of AC members’ connections (connectedness), and their structural position within the Accounting and Business Research 3 network (brokerage and strong ties). According to SNA, degree centrality is close in spirit to estimating cross-board appointments of directors or, as noted by Andres et al. (2013), this measure is theoretically close to the notion of ‘busy boards’. As far as social capital theory is concerned, degree centrality is a contentious variable in the sense that some studies consider it to be a proxy for social capital (e.g. Javakhadze et al. 2016) while others do not (Horton et al. 2012). As for the latter three metrics, several studies consider them as measures of different facets of social capital within the network (e.g. Burt 2000, Krackhardt 1998). As our primary proxy for financial reporting quality, we adopt an accruals model (see Walker 2013). We gather data from the US Corporate Library Database for the period 2003–2010. Our final sample consists of 13,668 firm-year observations. We investigate the US because it has been particularly influential in corporate governance worldwide (Beattie et al. 2012, p. 371). Our study is likely to be of interest for several reasons. First, drawing on SNA, we revisit the association between AC social connections and financial reporting quality. Social capital theory has been extensively used in management research and organisational theory to address how a board of directors’ social connections affect various organisational outcomes, such as firm performance. However, this framework has not been employed to examine ACs and their effectiveness, and we argue that this theoretical lens may enhance understanding of such social connections and their effects on financial reporting quality. Our findings suggest that AC members’ social capital is negatively associated with financial reporting quality. Therefore, this result may contribute to research examining the negative consequences of social capital, which is an underexplored area of research (Field 2008, Deth and Zmerli 2010). Second, we respond to calls for further research on directors’ characteristics (Johnson et al. 2013). In particular, we focus on a neglected aspect: the extent to which being designated as a financial expert influences the relationship between social connections and financial reporting quality (see Tanyi and Smith 2015 as an exception). According to our results, the negative association between social capital and reporting quality is mainly due to individuals who are not designated as financial experts and this, we argue, adds to existing agency-theory-based research examining the ‘busyness hypothesis’. Third, our results suggest that social capital measures based on counts of the number of connections among directors do not capture the richness of an individual’s social ties. Therefore, our findings highlight the importance of using various theoretically driven and well-established social network metrics to enhance understanding of the impact of directors’ social connections on organisations in general, and on accounting outcomes in particular. Furthermore, social capital can be approached from an individual- or group-level perspective. By using an individual-level approach, we add to extant research focusing on groups’ social capital (see Jha and Chen 2015). Finally, our study has implications for the ongoing policy debate about AC members’ social connections (i.e. multiple directorships). In particular, given the differential effects of social capital for experts and non-experts within an AC, our findings suggest revisiting policy recommendations that indiscriminately restrict interlocked directorships to all AC members (Sharma and Iselin 2012). This paper is organised as follows: Section 2 provides some regulatory background. Section 3 reviews the literature and develops our hypotheses. Section 4 details our dataset and outlines the research design. The main results and robustness checks are presented in Section 5. Finally, Section 6 discusses our results and notes some limitations and implications of our findings. 2. Regulatory background Since the enactment of the Securities and Exchange Act of 1933 and the Securities and Exchange Act of 1934, US stock markets have supplemented federal and state laws with further 4 N. Carrera et al. requirements to enhance the corporate governance of listed firms. In this regulatory context, ACs have a long tradition; in June 1978, ACs became a listing requirement of the NYSE, and in 1987 the National Commission on Fraudulent Reporting recommended the establishment of ACs by public companies (Treadway Commission 1987). In February 1999, the Blue Ribbon Committee (BRC 1999) addressed firms’ distortions of their financial performance by proposing a new structure of ACs, which included recommendations for proper functioning. The Sarbanes–Oxley Act (SOX) describes an AC’s purpose as ‘overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer’ (2002, p. 116, STAT. 747). Specific duties of ACs include reviewing the financial statements, discussing financial reporting issues with external auditors and management, discussing matters related to internal control, meeting privately with internal and external auditors, and overseeing the scope of internal and external audits (see Sections 301 and 302 of SOX; also, the final rule of the SEC on Standards Relating to Listed Company Audit Committees, SEC 2003). To tighten AC independence, SOX established that ACs should be solely formed by independent directors and enforced the requirements related to the background and expertise of AC members. Specifically, Section 407 directed the SEC to adopt rules requiring a company to disclose whether its AC includes at least one member who is a financial expert as well as a definition of ‘audit committee financial expert’.1 3. Literature review and hypotheses development 3.1. Literature review Accounting research has examined the association between AC members’ social ties and the quality of financial reporting. Dhaliwal et al. (2010) found that directors holding many social connections may degrade their monitoring work, leading to lower accruals quality. Bruynseels and Cardinaels (2014) showed that firms whose ACs hold ‘friendship’ ties to the CEO engage more in earnings management than their counterparts. Tanyi and Smith (2015) found that both busy AC chairs and AC financial experts have a negative effect on their firms’ financial reporting quality. However, this finding does not hold for non-AC chairs and non-AC financial experts. Sharma and Iselin (2012) suggested a positive association between financial misstatements and multiple directorships in the post-SOX period and argued that AC members with multiple appointments may be ‘stretched too thinly to effectively perform their monitoring responsibilities’ (Sharma and Iselin 2012, p. 149), and concluded that it may be desirable to limit the number of directorships held by AC members. This stream of research has primarily relied on agency theory and elaborated on the ‘busyness hypothesis’ with the notable exception of Dhaliwal et al. (2010), who incorporate both agency theory and resource dependence theory. In addition, while a growing number of studies have drawn on social capital theory to study the relationship between boards of directors’ networks and organisational performance (Horton et al. 2012, Field et al. 2013, Johansen and Pettersson 2013, Johnson et al. 2013), this perspective has not been applied to the study of the social connections of AC members. Mainstream social capital theory focuses on how individuals rely on their relationships, or social ties, to better mobilise their resources and get higher returns (Burt 2001, 2007, Adler and Kwon 2002). In this study, we draw on this stream of research for our hypotheses development and, hence, we focus on the beneficial effects of AC members using their social connections and social relations to improve organisational effectiveness (Lin 2001). Nonetheless, social connections are not necessarily beneficial in all cases as there are risks stemming from social capital. For example, Putnam (2000, p. 14) refers to cases of malevolent cooperation (e.g. criminal gangs). Therefore, an eventual rejection of our hypotheses would enable us to look into the Accounting and Business Research 5 negative consequences of social ties in the case of AC members and, hence, contribute to the growing research examining the ‘dark side’ of social capital (Deth and Zmerli 2010). The development of our hypotheses draws on the understanding of social capital as a multifaceted construct (Johnson et al. 2013). Prior research in this area suggests four metrics to approach the multiple dimensions of social capital: centrality, connectedness, brokerage position and strong ties (Simmel 1950, Burt 1997, 2000, Borgatti and López-Kidwell 2010). Centrality describes how active individuals are in the network (e.g. working to maintain and manage contacts; Lee 2013). Connectedness captures the quality of the connections (Andres et al. 2013, Larcker et al. 2013). Brokerage position measures individuals connected to people who are not connected to each other and, hence, how an AC member becomes central to her social network (Burt 1997). Finally, strong ties shows how individuals interrelated in groups of three or more, that is, when two people are reciprocally and strongly tied to each other and strongly tied to at least to a common third party, see their personal behaviour affected by group membership (Simmel 1950, Krackhardt 1998, Andres et al. 2013). 3.2. Hypotheses development 3.2.1. Social capital of AC members and financial reporting quality Social ties offer an opportunity to exchange knowledge and information through relationships, which enhances individuals’ abilities, competencies and skills, which in turn benefits organisations (Burt 1997). From this perspective, networks act as efficient information channels (Coleman 1988, p. S14); prior research shows that connections within a network facilitate individuals’ ‘access to broader sources of information and improve information’s quality, relevance, and timeliness’ (Adler and Kwon 2002, p. 29, see also Haunschild 1993, Westphal et al. 2001). Drawing on social capital theory, we identify two mechanisms that relate AC members’ social capital to financial reporting quality: (i) AC members’ monitoring job may be positively influenced by the flow of information available in their social networks, and (ii) social capital is intimately related to reputation (e.g. Nahapiet and Ghoshal 1998) and may operate as a disciplinary mechanism; individuals are careful to behave well to preserve their reputation among colleagues (Burt 2005, p. 8, Javakhadze et al. 2016, p. 40). According to mainstream social capital theory, individuals draw on their social ties to improve their performance. As argued by Johansen and Pettersson (2013, p. 289), board colleagues are considered a ‘trusted source’ and it is acceptable for them and their firms to share information. In the case of ACs, this information sharing enables AC members to learn about other firms’ accounting and governance practices, which may not be observable by outsiders (Reppenhagen 2010) in a detailed, timely manner (Burt 1992, Horton et al. 2012). Therefore, a firm’s governance structure and the effectiveness of its AC may capitalise on information transfers obtained through social networks, which are not otherwise available (Kim 2005, Stuart and Yim 2010, p. 175). Furthermore, AC members sitting on several boards are exposed to diverse accounting, governance and strategic experiences (Vafeas 1999, Kor and Sundaramurthy 2009, p. 985), and this ‘cosmopolitan view’ of management issues (Useem 1984, p. 84) allows them to mimic actions and transfer them across networks (Beckham and Haunschild 2002). In this context of collaboration and open communication, Hoitash (2011, p. 405) suggests that social ties bring about improvements in internal controls and financial reporting quality. The second mechanism relates to the cost of reputation loss (Javakhadze et al. 2016). Social networks filter and legitimise information about their members and forward strong signals to the market about an individual’s reputation (Burt 1992, p. 14, 2005, Johnson et al. 2013, p. 14). Such high reputation costs make it difficult for board and AC members to collude with management (Vafeas 2001, 2005, Yermack 2004, Yang and Krishnan 2005, Fich and Shivdasani 2006). 6 N. Carrera et al. 3.2.2.1. Centrality. The social capital of AC members depends on their position in the social network (Burt 2000). An individual is said to be central if her ties make her visible to other individuals in the network (Zaheer et al. 2010, p. 66), and if she is expected to play a more important role in their social structure than her counterparts with fewer direct contacts (Wasserman and Faust 1994). Therefore, individuals with a high degree of centrality may gather information about accounting practices from sources that are not available to their counterparts with a lower degree of centrality and, hence, we expect a positive association between the number of direct contacts (centrality) of AC members and financial reporting quality.2 Thus, we hypothesise: H1: The greater the centrality of AC members, the higher the quality of financial reporting. 3.2.2.2. Connectedness. An AC member’s social capital also depends on the quality of her connections (Horton et al. 2012, Andres et al. 2013). In the case of two AC members with the same number of connections, the one linked to powerful individuals will outperform those with access to less powerful individuals (Borgatti and López-Kidwell 2010). Therefore, the higher the quality of the connections of an AC member (connectedness), the higher will be the quality of information available to her, and this in turn will positively affect her social capital and her effectiveness within the AC. Consequently, we expect a positive association between the quality of ACs members’ connections and financial reporting quality. Therefore, we hypothesise: H2: The greater the connectedness of AC members, the higher the quality of financial reporting. 3.2.2.3. Brokerage position. Burt (1997, p. 340) argues that social capital is ‘a function of brokerage opportunities in a network’. Individuals holding a brokerage position have access to people who are not interconnected and, hence, obtain access to information that is not available to disconnected individuals, and this arguably lets them exert a prominent position over the network (Burt 1992, 2001, Horton et al. 2012). Therefore, a brokerage position provides individuals with high social capital as well as marginally significant and timely access to information, which in turn leads them to exert a positive effect on reporting quality. Therefore, we hypothesise: H3: The greater the AC members’ brokerage position, the higher the financial reporting quality. 3.2.2.4. Strong ties. Strong ties inculcate commitment, offer mentoring and constitute a source of trust, which in turn affects an individual’s attitudes, opinions, and beliefs (Kilduff and Krackhardt 1994). In terms of knowledge and organisational culture, strong ties are effective at transferring knowledge that is otherwise difficult to codify and explain (Hansen 1999). AC members having this form of social capital (strong ties) may use their sources of trust to enhance the transfer of knowledge as well as organisational effectiveness and, ultimately, contribute to financial reporting quality. Thus, we hypothesise: H4: The stronger the ties of AC members to each other and to third parties in common, the higher the financial reporting quality. 3.2.2. Social capital of AC financial experts and financial reporting quality Johnson et al. (2013) suggest that the impact of social capital on organisational outcomes may be moderated by some characteristics of individual directors. The case of financial experts has attracted research and regulatory research interest. Tanyi and Smith (2015) examine the social ties of two well-defined subgroups within the AC: individuals designated as ‘AC financial Accounting and Business Research 7 experts’ and those lacking such designation. Tanyi and Smith (2015) found a negative association between the ‘number of positions’ held by an AC member designated as a financial expert and/or as a chair of the AC and the quality of financial reporting. Furthermore, SOX requires firms to disclose whether or not, and if not, the reasons therefore, its AC includes at least one member designated as a financial expert (SOX 2002, Section 407). AC financial experts are competent to provide management judgement in key accounting areas and to appraise financial reporting quality (DeFond et al. 2005). Chiu et al. (2013) argue that those more directly responsible for monitoring financial reporting are more likely to be ‘opinion leaders’ on accounting choices and financial reporting decisions. From a social capital perspective, these individuals act as effective brokers in their networks and receive relevant information on a timely basis. Arguably, the experience, knowledge, and skills of individuals designated as AC financial experts make them likelier to become involved in financial reporting decisions than their non-expert counterparts. The transfer of critical information related to financial reporting and accounting across firms is more likely to occur through the social ties of those designated as financial experts compared with the social ties of non-experts. In addition, the role of social capital as a disciplinary mechanism is expected to be even more important for AC financial experts as compared with non-financial experts. As experts in financial reporting, these individuals are expected to face high reputational losses in the event of a financial scandal or news about poor AC performance. Taking into account these considerations, we re-examine hypotheses H1–H4 for those individuals designated as experts within the AC (H1_Ex-H4_Ex) with the aim of providing evidence on whether the hypothesised association between AC members’ social capital and financial reporting quality is contingent on individuals’ professional expertise. 4. Sample, network measures and methods 4.1. Sample The Corporate Library Board dataset provides coverage for the period 2001–2010 and has a total of 24,995 firm-year observations (see Table 1). As the number of firms covered by the dataset in 2001–2002 is small, these observations may limit the number of observable network measures. Therefore, we use the firm-year observations starting from 2003, which also mitigates the potential bias that might arise from the implementation of SOX, in 2002. As noted above, SOX enforced requirements regarding the independence of ACs (Section 301, SOX 2002) as well as disclosure procedures for ACs’ financial experts that may affect the composition of ACs. For the period 2003–2010, the total number of unique firm years is 21,695. To generate our network measures, we first identify firms for which we are able to compute our financial reporting quality measures. This procedure ensures that we are not inflating our social measures by using a large network; in fact, we are biasing our measures towards the low end, and therefore biasing against finding statistical significance. The matching procedure Table 1. Sample selection. Initial number of observations: years 2001–2010 Less number of observations for 2001–2002 Number of observations for 2003–2010 Less number of observations missing discretionary accruals Sample of firms with discretionary accruals Less number of observations with missing control variables/regulated industries Final sample 24,995 −3300 21,695 −4268 17,427 −3739 13,668 8 N. Carrera et al. reduces our sample to 17,427 firm years (see Table 1). We match these observations with firms’ financial variables gathered from the Compustat database. After we eliminate the observations for firms operating in regulated industries (financial firms [2-digit SIC (Standard Industrial Classification) codes 60–69] and utilities firms [2-digit SIC code 40–49]), our final sample consists of 13,668 firm-year observations. 4.2. SNA and network measures A social network ‘consists of a set of actors or nodes along with a set of ties of a specified type (such as friendship) that link them’ (Borgatti and Halgin 2011, p. 2). In this study, we focus on ego networks, that is, on how individuals are embedded in local social structures and, hence, we examine the nodes (individuals) around a given node (the ‘ego’) and all the ties among them. Burt (1992) argues that the shape of the ego network around an individual creates advantages for that individual and the organisation. In SNA analysis, the researcher defines a network by choosing a set of nodes and a type of tie between the nodes (Borgatti and Halgin 2011, p. 2). As our study focuses on the social ties formed by AC members, we split the board members in our sample into two groups: directors who serve on at least one AC (AC members) vis-à-vis their counterparts who do not join any ACs (non-AC members). To further investigate the effects of AC members’ social connections, we partition our subsample of AC members into experts and non-expert members. Thus, we generate four separate subgroups: AC members, non-AC members, AC members who are designated as financial experts (AC experts), and AC members who are not designated as financial experts (AC non-experts), wherein each subgroup is treated as a separate network. The network metrics are then estimated at the director level (e.g. the individual’s network) for each of the four subgroups, which in turn are aggregated at the firm level by averaging individual directors’ scores. We use concepts and methods derived from SNA to account for the multi-dimensional character of social capital (Andres et al. 2013, Johnson et al. 2013). The first proxy of social capital is centrality (H1 and H1_Ex) (Freeman 1979, Scott 1991, p. 82). The most basic measure of centrality is degree centrality, which is based on an individual’s number of relationships (ties). The variable degree centrality (DEGREE) is defined as the number of ties incident upon a node (Borgatti 2005, p. 62) and, hence, it is a simple count of the number of ties a person has in a network (Krackhardt 2010, p. 819). That is, if L(i, j) indicates the linkage between nodes ‘i’ and ‘j’, the variable for director ‘i’ in the network is defined as follows: DEGREE ;
j =1 L(i, j) (n − 1) . (1) We normalise the variable DEGREE by dividing it by n − 1 (where n is the total number of directors in the network). In spirit, degree centrality is the closest to the traditional metric employed in the literature on ‘busy boards’, a count of the number of directors holding cross-board appointments (Andres et al. 2013). To test the influence of connectedness (H2 and H2_Ex), we use the metric eigenvector centrality (variable CONNECTEDNESS). This metric considers the number of the ego’s own ties as well as the centrality of the ego’s ‘alters’, where ‘alter’ refers to the nodes (individuals) that are directly connected to ego. Bonacich (1972) uses a coefficient to take into account the quality of an individual’s social ties – i.e. how well connected are her alters. The variable CONNECTEDNESS takes into account whether a node’s direct contacts are themselves highly connected (Freeman 1979, Larcker et al. 2013). This metric captures the ‘quality’ of connections, because two directors could theoretically be connected to the same number of people, but their Accounting and Business Research 9 contacts, in turn, might not themselves be multiply connected. Eigenvector centrality is a summary metric of overall connectedness and has also been used as a proxy for a director’s power and prestige (Larcker et al. 2013, p. 232). This measure is a weighted sum of the degree centrality for the other nodes (directors) to which a node is connected. For directors ‘i’ and ‘j’, CONNECTEDNESS is calculated as follows: C(ni ) =
j =i Wij . C(nj ) lmax , (2) where λ is the proportionality factor and Wij = 1 if directors ‘i’ and ‘j’ are interconnected, while C is the eigenvector. Like DEGREE, this measure is also normalised. In H3 and H3_Ex, we predict a positive association between the brokerage position of an AC member and of an AC expert respectively and a firm’s financial reporting quality. According to Burt (1992), a structural hole between two individuals exists when they are connected to the same other actor but are not connected to each other. They become effective brokers in the network (metric BROKERAGE). This variable attempts to capture the positional consequences for individuals that result from how they are embedded in their neighbourhood. This notion is similar in spirit to the brokerage measure discussed in Horton et al. (2012). Furthermore, it constitutes a measure of power as it accrues to those actors who enjoy exclusive relations with others who are themselves disconnected (Burt 1992). The measure is calculated as total network size of the focal director minus the average network size of alters (connected directors). Brokerage ;  j  mij 1 −  q  piq m jq ; q = i, j, (3) where m is the number of ties, ‘i’ the ego, ‘j’ the alters, and ‘p’ the effort (Burt 1992, p. 52). The three metrics introduced above refer to forms of social capital arising from dyadic relations. However, research has shown that when people are interrelated in groups of three or more, individual behaviour is affected by group membership (Simmel 1950, Krackhardt 1998, Andres et al. 2013). Both Simmel (1950) and Krackhardt (1998) document that three individuals is the minimum size of the group where the dynamics are decidedly different, and having more than three members is not relevant. Two nodes are Simmelian tied to one another if they are reciprocally and strongly tied to each other and strongly tied to at least one third party in common; that is, the tie is embedded in a closed triad (Krackhardt 1998). In a study of ties across organisations, Tortoriello and Krackhardt (2010) suggest that Simmelian ties help an organisation more than simple ties do. The variable STRONG_TIES, which accounts for the number of Simmelian ties, captures the number of ties embedded in a closed triad (Krackhardt 1998) and is a proxy for the triadic relationships that a director belongs to in a network. This measure is also normalised. For a director, the values of STRONG_TIES range from zero (if neither of the director’s contacts is connected) to one (if all directors’ contacts are connected). In H4 (H4_Ex) we hypothesise a positive association between STRONG_TIES of AC members (STRONG_TIES of AC experts) and financial reporting quality. We use the four metrics above (DEGREE, CONNECTEDNESS, BROKERAGE, and STRONG_TIES) as proxies to compute an AC member’s position within the network and, in turn, their social capital. We measure the different proxies for the social capital of a firm’s AC by computing the average social capital of the individuals forming the AC. For an illustration of how the metrics are calculated, see the appendix. 10 4.3. N. Carrera et al. Research methods Prior research has used different proxies to capture the quality of earnings (for a review, see Dechow et al. 2010). Earnings management, defined as ‘the use of managerial discretion over (within-GAAP) accounting choices, earnings reporting choices, and real economic decisions to influence how underlying economic events are reflected in one or more measures of earnings’ (Walker 2013, p. 446), is typically considered as an indicator of poor quality of financial reporting (Carcello et al. 2011, p. 4). As Walker (2013, p. 543) notes, the use of within-GAAP discretion over the level of accruals is by far the most frequently studied earnings management method. Its popularity relies on the fact that discretionary accruals, whether resulting from bias or error, undermine decision implementation. A lower level of discretionary accruals suggests a higher quality of financial reporting. Many studies examining AC effectiveness have used accruals models as a proxy for financial reporting quality (Klein 2002, Xie et al. 2003, Geiger and North 2006, Dhaliwal et al. 2010, Bruynseels and Cardinaels 2014, Habib and Bhuiyan 2016). Accordingly, we use discretionary accruals as our primary measure, and accounting conservatism (e.g. Krishnan and Visvanathan 2008) as an alternative proxy for financial reporting quality. Among the different models for computing abnormal accruals, we adopt the modified Jones model (Dechow and Sloan 1995). While Jones (1991) defines accruals as a function of sales growth and property, plant and equipment (PPE), the modified Jones model is adjusted by the growth in credit sales, which are frequently manipulated (Dechow and Sloan 1995). Our first dependent variable is the absolute value of discretionary accruals (ABSDA), as we are interested in the magnitude of the impact. To compute the variable ABSDA we first estimate the modified Jones model for each two-SIC digit year grouping as follows:    TAi,t 1 (DREVi,t –DARi,t ) + a2 ∗ = a1 ∗ ASSETS1,t – 1 ASSETS1,t –1 ASSETSi,t –1 PPEi,t + a3 ∗ ASSETS1,t – 1 + 1i,t , (4) where for fiscal year t and firm i, TAi,t represents total accruals defined as earnings before extraordinary items and discontinued operations minus operating cash flows from continuing operations, ASSETS indicates total assets, ΔREVi,t represents change in revenues from the previous year, ΔARi,t is the change in accounts receivable from the previous year, and PPEi,t indicates the gross value of property, plant, and equipment. The discretionary accruals (ABSDA) are computed as the absolute value of the difference between total accruals (TA) deflated by lagged assets and the fitted values of Equation (4). We employ a host of audit, governance, and financial controls and estimate the following multivariate model: ABSDAi,t =b0 + b1a ∗SC ACi,t + b1b ∗SC NonACi,t + b2 ∗LNDIRTOT + b3 ∗BDMTGi,t + b4 ∗PCTDIROUTi,t + b5 ∗EXPERTi,t + b6 ∗CEOCHAIRi,t + b7 ∗CEOAGEi,t + b8 ∗BIG4i,t + b9 ∗AUD ACi,t + b10 ∗LEVi,t + b11 ∗ZSCOREi,t + b12 ∗INSTMAJi,t + b13 ∗LNMVi,t + b14 ∗CFi,t + b15 ∗DDCFi,t + b16 ∗LROAi,t –1 + b17 ∗DDNIi,t   + b18 ∗LOSSi,t + b19 ∗ACQi,t + b20 ∗BMi,t + bk Yeari,t + bj Indusi,t + 1i,t , (5) Accounting and Business Research 11 where SC_AC represents our variables of interest, namely one of the proxies for social capital computed for the subnetwork resulting from the social ties of individuals serving on ACs. Given that we predict a positive association between social capital and financial reporting quality, we expect the coefficients for the social capital proxies to be negatively associated with accruals. The model also includes the social capital metrics of board members who do not belong to the AC (SC_NonAC). Financial reporting quality assurance is entrusted to the whole board of directors (e.g. SOX 2002). Thus, arguably, all board members’ relational and social capital (Mallin and Michelon 2011) could potentially influence the quality of financial reporting. The inclusion of the social capital variables for non-AC members helps us to determine whether the social capital of different subgroups within the board affects financial reporting quality differently. We make no predictions regarding the sign of the social capital variables for non-AC members. Table 2 describes the variables included in the model. Our model relies on a number of control variables that are traditionally used in studies examining AC effectiveness and discretionary accruals (e.g. Klein 2002, Geiger and North 2006, Bruynseels and Cardinaels 2014). Therefore, we control for some AC and board characteristics (e.g. Klein 2002, Bédard et al. 2004, Krishnan and Visvanathan 2008, Bruynseels and Cardinaels 2014). We control for the size of the board (LNDIRTOT) but make no prediction for its sign because some scholars suggest problems arising from large boards (e.g. Hermalin and Weisbach 2003) while others such as Klein (2002) find a positive correlation between board size and independent ACs. We also include the number of board meetings for a particular time period (BDMTG). Beasley et al. (2000) find that companies with fewer AC meetings have more fraud, so we expect the coefficient of BDMTG to be negative. Assuming that a high percentage of outside directors indicates independence (Klein 2002), we expect the percentage of outside directors (PCTDIROUT) to be negatively associated with accruals. The percentage of experts in the AC is captured by the variable (EXPERT). Xie et al. (2003) and Bédard et al. (2004) suggest that AC members’ financial expertise is associated with lower abnormal accruals, and this in turn lessens the likelihood of aggressive earnings management. Dhaliwal et al. (2010) find that accruals quality is positively related to the presence of accounting experts in the AC. We also include a dummy variable for CEO duality (CEOCHAIR). As separation of the roles of CEO and chair of the board is considered an indicator of good corporate governance, we expect the coefficient for CEOCHAIR to be negative (Krishnan and Visvanathan 2008). We also control for the age of the CEO (CEOAGE). In accord with Huang et al. (2012), we expect a positive association between CEO age and financial reporting quality. We include two control variables related to auditing. First, we consider whether the company has been audited by one of the big audit firms (BIG4). Big auditors are usually considered ‘high quality providers’ (Krishnan 2005, p. 654). However, some studies do not find significant differences between big firms and smaller ones (e.g. Bruynseels and Cardinaels 2014). Thus, we do not predict any sign for the variable BIG4. We also add a dummy variable to control for auditor interlocks between firms that have cross-appointed AC members (AUD_AC). That is, if two firms are connected via common AC members and they have the same auditor, then the variable takes the value 1, and 0 otherwise. We include a number of variables to proxy for alternative monitoring mechanisms (Krishnan et al. 2011). As creditors use financial information to assess the financial health of a firm, we include the variable LEV, measured by long-term debt divided by total assets, to proxy for such monitoring. We expect a positive association between financial leverage and discretionary accruals (Krishnan et al. 2011). We also include the variable ZSCORE to control for the financial condition of the firm (Altman 2000). Smaller values of the variable indicate higher levels of financial distress. We expect a negative association between a firm’s financial health and its discretionary accruals (Reynolds and Francis 2000). Krishnan et al. (2011) noticed that institutional ownership also serves as an alternative monitoring mechanism. We include the variable 12 N. Carrera et al. Table 2. Description of variables. Variable Definition ABSDA Absolute value of discretionary accruals (cross-sectional modified Jones model) Discretionary accruals (cross-sectional modified Jones model) Conservatism measure, C-Score as detailed in Khan and Watts (2009) Set of social capital measures for individuals serving on at least one AC. Specifically: Total number of ties of directors serving on at least one AC, normalised Eigenvector centrality of directors serving on at least one AC, normalised Effective Network Size of directors serving on at least one AC Simmelian Ties of directors serving on at least one AC Set of social capital measures for members of the board of directors (directors) who do not serve on an AC. Specifically: Total number of ties of directors who do not serve on an AC, normalised Eigenvector centrality of directors who do not serve on an AC, normalised Effective Network Size of directors who do not serve on an AC Simmelian Ties of directors who do not serve on an AC Set of social capital measures (see above) for members of the AC designated as financial experts Set of social capital measures (see above) for members of the AC who are not designated as financial experts Percentage of AC members who are designated as financial experts Percentage of total directors who are considered as outside board members Number of board meetings Dummy variable that takes value 1 if the majority of outstanding shares are held by institutional investors, 0 otherwise Logarithm of the total number of directors on the board Dummy variable that takes value 1 if the CEO chairs the board, 0 otherwise Age of the CEO at year t Dummy variable that takes value 1 if the firm is audited by one of the big audit firms (Deloitte, Ernst & Young, KPMG, PricewaterhouseCoopers), 0 otherwise Dummy variable that takes value 1 if two interlocked firms (connected by one director who serves at least in one AC) have the same auditor, 0 otherwise Dummy variable that takes value 1 if a firm was involved in merger and acquisitions activity, 0 otherwise Lagged return on assets (operating income before extraordinary items over total assets) Change in net income between t − 1 and t Dummy variable that takes value 1 if net income is negative, 0 otherwise Cash flow from operations scaled by total assets Change in cash flow from operations between t − 1 and t scaled by total assets Financial leverage (long-term debt divided by total assets) Ratio of book-value of the firm to the market value of equity Log of market value of equity of the firm Financial distress measure, calculated as Altman’s Z-score = 1.2*X1 + 1.4*X2 + 3.3*X3 + 0.6*X4 + 1.0*X5 where X1 = Working Capital/Total Assets, X2 = Retained Earnings/Total Assets, X3 = Earnings before Interest and Taxes/Total Assets, X4 = Market Value of Equity/Book values of Total Liabilities, and X5 = Sales/Total Assets Dummy variable for each year in the panel Dummy variable for each two-SIC digit industry group DA C_SCORE SC_AC DEGREE_AC CONNECTEDNESS_AC BROKERAGE_AC STRONG_TIES_AC SC_NonAC DEGREE_NonAC CONNECTEDNESS_NonAC BROKERAGE_NonAC STRONG_TIES_NonAC EXPERT_SC_AC NEXPERT_SC_AC EXPERT PCTDIROUT BDMTG INSTMAJ LNDIRTOT CEOCHAIR CEOAGE BIG4 AUD_AC ACQ LROA ΔNI LOSS CF ΔCF LEV BM LNMV ZSCORE YEAR INDUS Accounting and Business Research 13 INSTMAJ to measure whether the majority of outstanding shares are held by an institutional investor. We expect INSTMAJ to increase accruals because institutional investors may focus on short-term rather than medium/long-term performance and therefore be more likely to encourage earnings management (Krishnan and Visvanathan 2008). As Krishnan et al. (2011) explain, accruals are associated with factors driven by a firm’s business model and operating cycle. Accordingly, we control for firm size measured by the logarithm of the market value of equity (LNMV), which is a firm’s estimated market value of equity at the end of the fiscal year. In accord with Xie et al. (2003), we expect LNMV to have a negative impact on discretionary accruals. Research has shown that operating cash flow and changes in operating cash flow are negatively associated with abnormal accruals (Ashbaugh et al. 2003, Geiger and North 2006, Krishnan et al. 2011). We scale operating cash flow and changes in operating cash flow by total assets (variables CF and ΔCF). We also control for past performance by including the ratio ‘return on assets’ lagged one period (LROA). As in Geiger and North’s (2006) study, we expect a positive association between LROA and discretionary accruals. In addition, as in Bruynseels and Cardinaels (2014), we incorporate a variable that captures changes in net income (ΔNI) and a dummy variable that takes the value 1 if the firm’s net income in a given year is negative (LOSS). We expect a positive association between ΔNI and discretionary accruals (e.g. Klein 2002, Bruynseels and Cardinaels 2014). Ex ante, the sign of the association between LOSS and discretionary accruals is not unequivocal (Klein 2002). Following Geiger and North (2006), we include a variable to control for firms that entered into an acquisition, because this event may affect accounting accruals and management’s ability to influence the level of accruals (ACQ). We expect acquisitions to increase abnormal accruals. The firm’s growth opportunities are captured by the ratio of book-to-market equity (BM), which has been found to be negatively associated with abnormal accruals (e.g. Ashbaugh et al. 2003, Menon and Williams 2004). Finally, we control for the impact of industry characteristics and the time period by introducing industry dummies (INDUS) and year dummies (YEAR). 5. Results 5.1. Descriptive statistics Summary statistics for the full sample of firms are shown in Table 3. All financial variables are winsorised at the top and bottom 1% to mitigate any effects of extreme outliers. The mean (median) numbers of all the social capital measures of AC members are higher than those for the non-AC members. For example, the mean (median) of degree centrality for AC (DEGREE_AC) and non-AC members (DEGREE_NonAC) are 0.018 (0.016) and 0.010 (0.010), respectively. AC members occupy a more central position in the social network (DEGREE), the quality of their connections is higher (CONNECTEDNESS), they are more effective brokers in a network (BROKERAGE), and they have a higher number of triadic relationships (STRONG_TIES). Furthermore, AC members designated as financial experts are more central than non-experts – mean (median) of DEGREE of experts is 0.025 (0.020) compared with 0.017 (0.015) for non-experts. Financial experts also have a higher quality of their connections and are more effective brokers in the network vis-à-vis their non-expert counterparts. Among the corporate governance variables, the mean (median) of the log of total number of directors on board (LNDIRTOT) is 2.141 (2.197) and the mean (median) percentage of outside board members (PCTDIROUT) is 82.3% (85.7%). The mean (median) of annual board meetings (BDMTG) is 8.1 (7.0). On average, 51% of AC members are classified as experts. The mean (median) age of the CEO is 54.5 (54.0), and approximately 53% of these CEOs chair the board (CEOCHAIR). Around 90% of the companies included in the sample are audited by a big firm (mean BIG4 = 0.902). 14 Table 3. Descriptive statistics. Percentile Mean SD 25th 50th 75th Min Max N DA ABSDA C-SCORE LNDIRTOT BDMTG PCTDIROUT EXPERT CEOCHAIR CEOAGE BIG4 AUD_AC LEV ZSCORE INSTMAJ LNMV CF ΔCF LROA ΔNI LOSS ACQ BM 0.080 0.316 0.086 2.141 8.095 0.823 0.510 0.527 54.472 0.902 0.620 0.199 3.411 0.710 7.113 0.078 –0.002 0.005 –0.003 0.113 0.199 0.475 0.452 0.361 0.088 0.250 3.791 0.091 0.250 0.499 7.357 0.298 0.486 0.208 5.015 0.454 1.718 0.143 0.099 0.177 0.090 0.317 0.399 1.233 −0.071 0.041 0.035 1.946 6.000 0.778 0.333 0.000 49.000 1.000 0.000 0.006 1.596 0.000 5.979 0.048 –0.036 0.000 –0.024 0.000 0.000 0.265 0.013 0.134 0.086 2.197 7.000 0.857 0.500 1.000 54.000 1.000 1.000 0.163 3.052 1.000 7.033 0.092 –0.001 0.041 0.001 0.000 0.000 0.445 0.203 0.494 0.136 2.303 10.000 0.889 0.667 1.000 59.000 1.000 1.000 0.303 5.113 1.000 8.194 0.141 0.032 0.081 0.023 0.000 0.000 0.687 −1.000 0.000 –0.271 1.386 2.000 0.429 0.091 0.000 35.000 0.000 0.000 0.000 –37.731 0.000 1.374 –1.267 –0.769 –1.582 –0.824 0.000 0.000 –30.521 1.000 1.000 1.209 2.773 28.000 1.000 1.000 1.000 79.000 1.000 1.000 1.439 24.965 1.000 12.245 0.423 0.766 0.342 0.744 1.000 1.000 10.003 13,688 13,688 11,066 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 13,688 N. Carrera et al. Variable 0.018 0.008 7.369 0.001 0.011 0.015 6.281 0.001 0.010 0.000 2.689 0.001 0.016 0.002 5.581 0.001 0.024 0.008 10.142 0.002 0.000 0.000 0.000 0.000 0.100 0.239 51.315 0.007 13,688 13,688 13,688 13,688 Social capital measures for non-AC members DEGREE_NonAC 0.010 CONNECTEDNESS_NonAC 0.000 BROKERAGE_NonAC 3.810 STRONG_TIES_NonAC 0.001 0.007 0.003 3.863 0.000 0.003 0.000 1.000 0.000 0.010 0.000 2.250 0.001 0.013 0.000 4.947 0.001 0.000 0.000 0.000 0.000 0.050 0.253 33.501 0.003 13,688 13,688 13,688 13,688 Social capital measures for AC financial experts EXP_DEGREE_AC 0.025 EXP_CONNECTEDNESS_AC 0.008 EXP_BROKERAGE_AC 4.000 EXP_STRONG_TIES_AC 0.001 0.022 0.021 4.199 0.001 0.010 0.000 1.000 0.000 0.020 0.001 2.714 0.001 0.035 0.007 5.884 0.002 0.000 0.000 0.000 0.000 0.300 0.388 38.247 0.010 13,688 13,688 13,688 13,688 Social capital measures for AC non-financial experts NEXP_DEGREE_AC 0.017 NEXP_CONNECTEDNESS_AC 0.006 NEXP_BROKERAGE_AC 3.079 NEXP_STRONG_TIES_AC 0.001 0.013 0.022 3.186 0.001 0.010 0.000 1.000 0.000 0.015 0.000 2.000 0.001 0.023 0.002 4.191 0.001 0.000 0.000 0.000 0.000 0.130 0.485 30.556 0.008 13,688 13,688 13,688 13,688 Accounting and Business Research Social capital measures for AC members DEGREE_AC CONNECTEDNESS_AC BROKERAGE_AC STRONG_TIES_AC Note: Variables are defined in Table 2. 15 16 Table 4. Correlation matrix. ABSDA C-SCORE DEGREE_AC CONNECTEDNESS_AC BROKERAGE_AC STRONG_TIES_AC DEGREE_NonAC CONNECTEDNESS_NonAC BROKERAGE_NonAC STRONG_TIES_NonAC EXP_DEGREE_AC EXP_CONNECTEDNESS_AC EXP_BROKERAGE_AC EXP_STRONG_TIES_AC NEXP_DEGREE_AC NEXP_CONNECTEDNESS_AC NEXP_BROKERAGE_AC NEXP_STRONG_TIES_AC LNDIRTOT BDMTG PCTDIROUT EXPERT CEOCHAIR CEOAGE BIG4 AUD_AC LEV ZSCORE INSTMAJ LNMV CF ΔCF LROA ΔNI LOSS ACQ BM (2) (3) (4) (5) (6) (7) (8) 1.000 –0.030*** 0.029** 0.114*** 0.001** 0.073*** 0.029** 0.033*** 0.121*** 0.092*** 0.035*** 0.075*** 0.041*** 0.065*** 0.066*** 0.067*** 0.054*** 0.013** –0.074*** 0.019* –0.015 0.002 –0.004 –0.053*** –0.014 –0.025** –0.050*** –0.026*** 0.005 –0.017 –0.031*** 0.030*** –0.086*** 0.039*** 0.053*** 0.042*** –0.054*** 1.000 –0.324** –0.346*** –0.322*** –0.361*** –0.151*** 0,002 –0.282*** –0.271*** –0.277*** –0.196*** –0.284*** –0.319*** –0.205*** –0.198*** –0.233*** –0.225*** –0.252*** –0.001 –0.106*** –0.0905*** –0.0898*** –0.002 –0.133*** –0.189*** 0.130*** –0.121*** –0.0819*** –0.609*** –0.213*** –0.0208** –0.177*** –0.110*** 0.150*** –0.0326*** 0.342*** 1.000 0.605*** 0.669*** 0.618*** 0.463*** 0.017* 0.240*** 0.381*** 0.634*** 0.401*** 0.491*** 0.593*** 0.610*** 0.378*** 0.609*** 0.622*** 0.473*** 0.009 0.278*** 0.119*** 0.131*** 0.045*** 0.230*** 0.452*** 0.099*** –0.052*** 0.158*** 0.514*** 0.085*** 0.009 0.101*** 0.043*** –0.159*** 0.012 –0.053*** 1.000 0.637*** 0.660*** 0.302*** 0.044*** 0.372*** 0.373*** 0.590*** 0.568*** 0.560*** 0.611*** 0.526*** 0.569*** 0.459*** 0.505*** 0.368*** 0.026** 0.189*** 0.141*** 0.103*** 0.028*** 0.127*** 0.299*** 0.062*** –0.031*** 0.119*** 0.435*** 0.070*** 0.006 0.099*** 0.013 –0.117*** –0.000 –0.046*** 1.000 0.664*** 0.134*** 0.028** 0.452*** 0.340*** 0.655*** 0.510*** 0.661*** 0.616*** 0.457*** 0.366*** 0.644*** 0.581*** 0.392*** 0.068*** 0.291*** 0.136*** –0.043*** 0.023** 0.172*** 0.497*** 0.099*** –0.065** 0.077*** 0.423*** 0.055*** 0.012 0.042*** 0.018* –0.097*** –0.007 –0.042*** 1.000 0.342*** 0.034*** 0.431*** 0.442*** 0.644*** 0.500*** 0.608*** 0.661*** 0.677*** 0.414*** 0.693*** 0.629*** 0.509*** 0.049*** 0.333*** 0.154*** 0.084*** 0.036*** 0.237*** 0.535*** 0.118*** –0.063*** 0.150*** 0.534*** 0.078*** 0.010 0.095*** 0.029*** –0.149*** 0.005 –0.053*** 1.000 0.035*** 0.362*** 0.696*** 0.295*** 0.160*** 0.088*** 0.178*** 0.320*** 0.156*** 0.128*** 0.254*** 0.498*** –0.010 0.162*** 0.099*** 0.114*** 0.028** 0.193*** 0.148*** 0.103*** –0.060*** 0.119*** 0.388*** 0.065*** 0.012 0.073*** 0.046*** –0.135*** 0.024** –0.057*** 1.000 0.040*** 0.076*** 0.012 0.025** 0.020* 0.026** 0.028*** 0.019* 0.031*** 0.035*** 0.026** 0.016 0.015 0.026** 0.016 –0.019* 0.011 0.013 0.015 –0.007 0.003 0.018* –0.001 0.003 0.003 0.004 –0.010 0.002 0.005 N. Carrera et al. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37) (1) BROKERAGE_NonAC STRONG_TIES_NonAC EXP_DEGREE_AC EXP_CONNECTEDNESS_AC EXP_BROKERAGE_AC EXP_STRONG_TIES_AC NEXP_DEGREE_AC NEXP_CONNECTEDNESS_AC NEXP_BROKERAGE_AC NEXP_STRONG_TIES_AC LNDIRTOT BDMTG PCTDIROUT EXPERT CEOCHAIR CEOAGE BIG4 AUD_AC LEV ZSCORE INSTMAJ LNMV CF ΔCF LROA ΔNI LOSS ACQ BM (9) 1.000 0.636*** 0.274*** 0.306*** 0.450*** 0.432*** 0.202*** 0.194*** 0.256*** 0.229*** 0.326*** 0.045*** 0.231*** 0.141*** 0.074*** 0.002 0.137*** 0.277*** 0.125*** –0.078*** 0.053*** 0.331*** 0.035*** 0.012 0.042*** 0.008 –0.060*** 0.002 –0.052*** (10) (11) (12) (13) (14) (15) (16) (17) 1.000 0.313*** 0.254*** 0.304*** 0.344*** 0.276*** 0.198*** 0.235*** 0.285*** 0.577*** 0.058*** 0.249*** 0.148*** 0.077*** 0.011 0.189*** 0.252*** 0.146*** –0.082*** 0.100*** 0.423*** 0.057*** 0.016 0.060*** 0.032*** –0.120*** 0.014 –0.061*** 1.000 0.544*** 0.610*** 0.633*** 0.314*** 0.208*** 0.358*** 0.350*** 0.375*** 0.035*** 0.244*** 0.210*** 0.068*** 0.037*** 0.169*** 0.388*** 0.072*** –0.037*** 0.108*** 0.423*** 0.075*** 0.011 0.083*** 0.029*** –0.118*** 0.006 –0.040*** 1.000 0.597*** 0.616*** 0.193*** 0.154*** 0.234*** 0.208*** 0.247*** 0.034*** 0.135*** 0.095*** 0.037*** 0.011 0.086*** 0.210*** 0.034*** –0.025*** 0.070*** 0.285*** 0.049*** 0.005 0.067*** 0.007 –0.081*** –0.002 –0.030*** 1.000 0.635*** 0.182*** 0.188*** 0.383*** 0.252*** 0.338*** 0.070*** 0.261*** 0.193*** –0.041*** 0.019* 0.142*** 0.425*** 0.075*** –0.052*** 0.073*** 0.368*** 0.055*** 0.009 0.053*** 0.007 –0.086*** –0.003 –0.030*** 1.000 0.232*** 0.208*** 0.374*** 0.285*** 0.400*** 0.068*** 0.287*** 0.258*** 0.008 0.033*** 0.176*** 0.440*** 0.087*** –0.054*** 0.101*** 0.436*** 0.073*** 0.012 0.077*** 0.020* –0.116*** 0.003 –0.036*** 1.000 0.503*** 0.618*** 0.612*** 0.376*** 0.006 0.211*** –0.104*** 0.093*** 0.024** 0.171*** 0.337*** 0.077*** –0.044*** 0.119*** 0.347*** 0.042*** 0.006 0.068*** 0.016 –0.107*** 0.003 –0.037*** 1.000 0.480*** 0.503*** 0.201*** 0.006 0.096*** 0.016 0.043*** 0.010 0.065*** 0.149*** 0.045*** –0.018** 0.049*** 0.218*** 0.030*** 0.001 0.044*** 0.003 –0.062*** 0.003 –0.014 1.000 0.659*** 0.317*** 0.048*** 0.214*** –0.077*** –0.035*** 0.021* 0.136*** 0.370*** 0.076*** –0.055*** 0.051*** 0.320*** 0.031*** 0.013 0.010 0.024** –0.074*** –0.013 –0.028** Accounting and Business Research (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37) 17 (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37)) Continued. NEXP_STRONG_TIES_AC LNDIRTOT BDMTG PCTDIROUT EXPERT CEOCHAIR CEOAGE BIG4 AUD_AC LEV ZSCORE INSTMAJ LNMV CF ΔCF LROA ΔNI LOSS ACQ BM LEV ZSCORE INSTMAJ LNMV CF ΔCF LROA ΔNI LOSS ACQ BM (27) 1.000 –0.334*** –0.029*** 0.026** –0.067*** 0.003 –0.099*** –0.008 –0.061*** 0.019* –0.121*** Note: Variables are defined in Table 2. (18) 1.000 0.396*** 0.023** 0.239*** –0.125*** 0.053*** 0.038*** 0.165*** 0.371*** 0.082*** –0.054*** 0.100*** 0.367*** 0.046*** 0.012 0.046*** 0.029*** –0.112*** –0.007 –0.031*** (19) (20) (21) (22) (23) (24) (25) (26) 1.000 –0.013 0.308*** 0.077*** 0.007 0.080*** 0.240*** 0.322*** 0.152*** –0.052*** 0.149*** 0.507*** 0.093*** 0.017* 0.125*** 0.030*** –0.199*** 0.011 –0.025** 1.000 0.096*** 0.005 –0.100*** –0.088*** –0.017* 0.032*** 0.088*** –0.130*** –0.033*** –0.023** –0.099*** 0.021* –0.144*** 0.022* 0.102*** 0.007 0.009 1.000 0.053*** 0.066*** –0.048*** 0.171*** 0.237*** 0.083*** –0.086*** 0.083*** 0.184*** 0.009 0.020* 0.005 0.023** –0.056*** 0.011 –0.019* 1.000 0.037*** –0.006 0.080*** 0.102*** 0.036*** 0.002 0.060*** 0.134*** 0.038*** –0.003 0.057*** –0.003 –0.037*** 0.039*** –0.016 1.000 0.182*** 0.037*** –0.005 0.056*** 0.009 0.056*** 0.106*** 0.035*** –0.021* 0.094*** 0.016 –0.073*** 0.034*** –0.017* 1.000 –0.016 0.005 0.028** 0.020** 0.024** 0.075*** 0.039*** 0.004 0.068*** 0.002 –0.060*** –0.033*** 0.029*** 1.000 0.343*** 0.110*** 0.016* 0.096*** 0.293*** 0.108*** 0.006 0.104*** 0.030*** –0.120*** 0.029*** –0.045*** 1.000 0.087*** –0.040*** 0.082*** 0.318*** 0.037*** 0.007 0.029*** 0.020* –0.069*** 0.013 –0.028** (28) (29) (30) (31) (32) (33) (34) (35) (36) 1.000 0.082*** 0.177*** 0.399*** 0.074*** 0.368*** 0.115*** –0.242*** 0.009 0.044*** 1.000 0.231*** 0.145*** –0.001 0.160*** 0.000 –0.138*** 0.036*** 0.010 1.000 0.349*** 0.057*** 0.354*** 0.100*** –0.366*** 0.084*** –0.031*** 1.000 0.396*** 0.541*** 0.234*** –0.580*** 0.036*** 0.024** 1.000 –0.093*** 0.531*** –0.102*** –0.037*** –0.001 1.000 –0.176*** –0.531*** 0.081*** 0.048*** 1.000 –0.204*** –0.039*** –0.021* 1.000 –0.094*** 0.011 1.000 0.011 N. Carrera et al. (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37)) 18 Table 4. Accounting and Business Research 19 Our sample consists of large firms with relatively low ROAs during our observation period (LROA). The current mean (median) leverage of firms is 19.9% (16.3%) (LEV), and the mean (median) of Altman’s Z-score (ZSCORE) is 3.411 (3.052). Around 71% of firms in our sample have an institutional majority (INSTMAJ) on their boards, as one would expect for large firms. The mean equity market value of the full sample (LNMV) is roughly 7.113, with a range of 1.718 to approximately 12.245. The mean (median) of lagged return on assets (LROA) is 0.5% (4.1%). The means (medians) of firms with negative earnings (LOSS) and those engaged in mergers and acquisitions (ACQ) are 11.3% (0.0%) and 19.9% (0.0%), respectively. Finally, the mean (median) of the BM ratio is 0.48 (0.45). Table 4 shows the correlation matrix of the variables. Most of the correlations exhibit the usual sign and significance. For example, LEV is negatively and statistically significantly related (p-value < .01) to ZSCORE, meaning that firms with higher leverage are more likely to face bankruptcy than their counterparts with lower leverage. Apart from the social capital measures, all correlations greater than 0.016 are statistically significant at p-value < .10, and no correlation between any two independent variables is greater than 0.58. These correlations are in line with those reported by Bruynseels and Cardinaels (2014). However, we also check for multicollinearity, as noted below. The network metrics measure different but related forms of social capital and, as expected, they are highly correlated. For example, degree centrality for the AC members (DEGREE_AC) has correlations of 0.605, 0.669, and 0.618 with CONNECTEDNESS_AC, BROKERAGE_AC and STRONG_TIES_AC, respectively (see Table 4). Accordingly, we introduce each of the social capital measures individually during our multivariate regressions to test our hypotheses (Andres et al. 2013). 5.2. Main findings Table 5 shows the multivariate tests for hypotheses H1–H4. Model 1 is the baseline that includes all the control variables. White’s test rejected the null hypothesis that the error terms are homoscedastic (White 1980). We use Petersen’s (2009) method for all models to pool cross-sectional data to estimate robust standard errors that are consistent for both the firm and year effects. The adjusted R2, the overall fit, of our model is 0.43, in line with results shown by research examining discretionary accruals models (Klein 2002), and all variance inflation factors (VIF) scores are below 3.15. Columns 2–5 (Table 5) show the multivariate results to test hypotheses H1 to H4; the maximum value of all the VIF scores for these regressions is below 3.16. First, our results suggest that irrespective of the social capital measures in use, social capital only affects discretionary accruals (ABSDA) when we consider connections of individuals serving on ACs. Therefore, AC members have a positive impact on ABSDA and the results are all statistically significant at p-value < .01.3 These results lead us to reject the four hypotheses, which predict a positive association between AC members’ social capital and financial reporting quality. Model (2) reports a significant but positive coefficient for DEGREE_AC (0.954) suggesting that greater centrality of AC members deteriorates financial reporting quality. Specifically, we reject the hypothesis that the greater the centrality of AC members, the higher the quality of financial reporting (H1). Model (3) leads us to reject H2 (significant and positive coefficient for CONNECTEDNESS_AC). The results indicate that the greater the quality of AC members’ connections, the lower the financial reporting quality. Similarly, we reject H3: Model (4) shows that the better AC members’ brokerage position, the lower the financial reporting quality will be (BROKERAGE_AC). Finally, H4 is also rejected as the coefficient for STRONG_TIES_AC is significant but positive in Model (5). For non-AC members, the coefficients of the social capital metrics are all statistically insignificant at p-value < .01 (Models 2–5, Table 5). The reported differences between the results for 20 Table 5. Discretionary accruals and AC members’ social capital. ABSDAi,t = b0 + b1a ∗SC ACi,t + b1b ∗SC NonACi,t + b2 ∗LNDIRTOT + b3 ∗BDMTGi,t + b4 ∗PCTDIROUTi,t + b5 ∗EXPERTi,t + b6 ∗CEOCHAIRi,t b7 ∗CEOAGEi,t + b8 ∗BIG4i,t + b9 ∗AUD ACi,t + b10 ∗LEVi,t + b11 ∗ZSCOREi,t + b12 ∗INSTMAJi,t + b13 ∗LNMVi,t + b14 ∗CFi,t   b15 ∗DCFi,t + b16 ∗LROAi,t – 1 , +b17 ∗DNIi,t + b18 ∗LOSSi,t + b19 ∗ACQi,t + b20 ∗BMi,t + bk ∗Yeari,t + bj ∗Indusi,t + 1i,t . Exp. Sign – DEGREE_NonAC ? CONNECTEDNESS_AC – CONNECTEDNESS_NonAC ? BROKERAGE_AC – BROKERAGE_NonAC ? STRONG_TIES_AC – STRONG_TIES_NonAC ? LNDIRTOT ? BDMTG – PCTDIROUT – EXPERT – CEOCHAIR – CEOAGE + Model (2) Model (3) Model (4) Model (5) 0.954** (0.014) 0.153 (0.547) 0.739*** (0.000) 0.009 (0.977) 0.002*** (0.001) –0.000 (0.839) 0.016 (0.106) –0.001 (0.270) –0.047 (0.236) 0.007 (0.448) 0.008 (0.164) 0.000 (0.661) 0.007 (0.516) –0.001 (0.217) –0.061* (0.096) 0.004 (0.672) –0.007 (0.199) 0.000 (0.672) 0.011 (0.349) –0.001 (0.227) –0.053 (0.181) 0.004 (0.630) 0.007 (0.232) 0.000 (0.653) 0.0012 (0.263) –0.001 (0.234) –0.057 (0.124) 0.003 (0.717) 0.008 (0.196) 0.000 (0.652) 13.559*** (0.003) –8.174 (0.194) 0.013 (0.204) –0.001 (0.232) –0.061* (0.097) 0.004 (0.603) 0.007 (0.208) 0.000 (0.676) N. Carrera et al. DEGREE_AC Model (1) – AUD_AC ? LEV + ZSCORE – INSTMAJ + LNMV – CF – ΔCF – LROA + ΔNI + LOSS ? ACQ + BM – INTERCEPT ? Obs Adj. R2 –0.014 (0.214) 0.004 (0.452) 0.047** (0.036) –0.001 (0.100) 0.002 (0.730) –0.010*** (0.000) –0.012 (0.286) –0.004 (0.507) 0.046** (0.043) –0.001 (0.144) 0.001 (0.770) –0.012*** (0.000) –0.011 (0.309) 0.001 (0.874) 0.047** (0.035) –0.001 (0.145) 0.001 (0.757) –0.012*** (0.000) –0.012 (0.249) –0.002 (0.742) 0.046** (0.046) –0.001 (0.135) 0.001 (0.774) –0.012*** (0.000) –0.011 (0.297) –0.004 (0.463) 0.046** (0.044) –0.001 (0.137) 0.001 (0.797) –0.012*** (0.000) 0.193*** (0.001) –0.041 (0.472) –0.100** (0.021) 0.013 (0.873) –0.020 (0.190) –0.003 (0.601) –0.005*** (0.001) 0.214*** (0.000) 13,688 0.428 0.194*** (0.001) –0.042 (0.464) –0.099** (0.024) 0.015 (0.858) –0.021 (0.184) –0.002 (0.710) –0.005*** (0.001) 0.236*** (0.000) 13,688 0.428 0.195*** (0.001) –0.042 (0.464) –0.099** (0.023) 0.015 (0.856) –0.021 (0.182) –0.002 (0.731) –0.005*** (0.001) 0.241*** (0.000) 13,688 0.428 0.195*** (0.001) –0.043 (0.460) –0.100** (0.021) 0.016 (0.850) –0.021 (0.189) –0.002 (0.691) –0.005*** (0.001) 0.242*** (0.000) 13,688 0.428 0.196*** (0.001) –0.042 (0.460) –0.101** (0.022) 0.015 (0.856) –0.021 (0.187) –0.002 (0.708) –0.005*** (0.001) 0.232*** (0.000) 13,688 0.428 Accounting and Business Research BIG4 Notes: Variables are defined in Table 2. Coefficients for year and industry variables are suppressed. P-values in parentheses are heteroscedastic firm- and year-corrected values (Petersen 2009). *Significant at 10%. **Significant at 5%. ***Significant at 1%. 21 22 N. Carrera et al. AC and non-AC members highlight the importance of separating out subgroup and director effects when examining boards of directors (see Johnson et al. 2013). Therefore, we do not find support for the contention stating a potential benefit of AC members’ social capital on financial reporting quality. Opposite to predictions drawing on mainstream social capital theory, our findings suggest that directors’ social capital is associated with higher discretionary accruals when the cross-board links involve individuals serving on ACs. None of the control variables for corporate governance is statistically significant except for PCTDIROUT. PCTDIROUT is negative and statistically significant in Models 2 and 5, at p-value < 0.10, suggesting that a larger percentage of outsiders in the board is associated with lower ABSDA. Contrary to our expectations, although in line with recent studies (Bruynseels and Cardinaels 2014), EXPERT is not associated with ABSDA. These results are consistent through all the models (Models 1–5, Table 5), except for AUD_AC, whose coefficient is negative in Models 2, 4 and 5 and positive in Model 3. Among the proxies for other monitoring mechanisms (INSTMAJ, LEV, and ZCORE), only LEV is statistically significant, at p-value < .05. In line with previous studies (Krishnan et al. 2011, Cohen et al. 2014), the coefficient is positive, suggesting that debt-holders are not active monitors of financial reporting quality. Our proxy for company size (LNMV) is negatively associated with ABSDA and statistically significant at p-value < .01 (see Cohen et al. 2014). CF is positive and significant in all models at p-value < .01, supporting the notion that high cash flows increase abnormal accruals. Unlike previous research (Ashbaugh et al. 2003, Menon and Williams 2004), we find that lagged ROA is negatively associated with ABSDA at p-value < .05. Contrary to previous studies (e.g. Geiger and North 2006, Dhaliwal et al. 2010), the coefficients for changes in net income (ΔNI), negative earnings (LOSS), and acquisitions (ACQ) are not statistically significant. Finally, the BM variable is negative and statistically significant at p-value < .01 as in prior research (e.g. Geiger and North, 2006). As noted above, it could be the case that an overlapping characteristic such as different features between ACs’ financial and non-financial experts confound the findings regarding the effects of ACs’ social capital on financial reporting quality. Accordingly, we further explore our hypotheses by separating the group of AC members into two subgroups: (a) AC members who are designated as financial experts (EXP_AC); and (b) AC members who are not designated as experts (NEXP_AC), and we compute the social capital metrics for the two subgroups. In Table 6 we report the models introduced in Table 5 including the social capital metrics for the two subgroups of the AC. The adjusted R2 of our model is 0.43, in line with results shown by research examining ABSDA models (e.g. Klein 2002). Model 1 in Table 6 is a baseline that includes all the control variables. The results for expert and non-expert AC members after controlling non-AC members are shown in Table 6 (Models 2–5). The maximum value of all the VIF scores is lower than 4 for all models reported in Table 6. Model 2 tests the variable degree centrality (DEGREE). The results show that the variable EXP_DEGREE_AC is positive and statistically significant at p-value < .05. The variables NEXP_DEGREE_AC and DEGREE_NonAC are not significant. This finding suggests that improvements in the centrality of AC financial experts increase discretionary accruals (ABSDA). This is not the case for non-experts serving on the AC or other directors. Therefore, and in contrast to our predictions, we find that the centrality of AC experts is negatively associated with financial reporting quality. Models 3–5 detail the results for CONNECTEDNESS, BROKERAGE, and STRONG_TIES for expert and non-expert AC members. When compared with Model (2), we find two relevant results: First, AC experts’ social capital as proxied by CONNECTEDNESS, BROKERAGE and STRONG_TIES is not statistically significant vis-à-vis discretionary accruals. Second, Accounting and Business Research 23 NEXP_CONNECTEDNESS_AC is positive and statistically significant (p-value < .01), as also are NEXP_BROKERAGE_AC and NEXP_STRONG_TIES_AC (p-value < .05). These results suggest that when social capital measures take into consideration the quality of the connections (whether the individual’s contacts are themselves extensively connected), the social capital of non-designated financial experts serving on the AC is positively associated with ABSDA (NEXP_CONNECTEDNESS, Model 3). Furthermore, the higher the degree to which such non-experts can serve as effective brokers in the network (NEXP_BROKERAGE_AC, Model 4), or the higher their number of ‘strong ties’ (NEXP_STRONG_TIES_AC, Model 5), the higher the level of discretionary accruals will be. Separating the group of individuals sitting in the ACs based on whether or not they are designated as financial experts produces some interesting findings. First, the results vary depending on the form of social capital captured by the network metrics. When social capital is measured by the raw number of ties that an individual has in a network (DEGREE), Model (2) reports a positive and significant coefficient for AC members designated as financial experts. That is, the larger the number of direct contacts of AC members designated as financial experts, the lower the quality of financial reporting will be. However, once we take into account the quality of the social connections of individuals and their structural position in the network, we find that financial experts’ social capital is not associated with discretionary accruals. The opposite happens with non-experts: except DEGREE, all social capital variables are positive and statistically significant. The results for the control variables are similar to those presented in Table 5. Second, based on the results reported in Table 6, we suggest that the social capital of AC experts, as measured by CONNECTEDNESS, BROKERAGE, and STRONG_TIES is not associated with financial reporting quality; the negative impact of social capital metrics on financial reporting quality is explained by the connections of non-experts (Models 3–5). 5.3. Additional analysis Using a measure of firm conservatism as an alternative proxy for financial reporting quality (see Tables 7 and 8), we re-estimated all models reported in Tables 5 and 6. Since we are interested in a firm-year level measure that captures both cross-sectional as well as time-series measures of asymmetric timeliness of earnings, we employ the C-Score metric developed by Khan and Watts (2009), hereafter KW. We first estimate the C-Score using the procedure of KW and then match the C-Score to our sample of firm-years, noting that higher C-Score means a higher degree of conservatism. To estimate the C-Score, we first run regressions on a yearly basis using the KW model: Xi = b1 + b2 Di + Ri (m1 + m2 Sizei + m3 M /Bi + m4 Levi ) + Di Ri (l1 + l2 Sizei + l3 M /Bi + l4 Levi ) + (d1 Sizei + d2 M /Bi + d3 Levi + d4 Di Sizei + d5 Di M /Bi + d6 Di Levi ) + 1i , where X is the earnings before extraordinary income/lagged market value of equity, R is the annual returns compounded from monthly returns beginning fourth month after the end of fiscal year, t − 1, Size is the natural log of market value of equity, M/B is the ratio of market value of equity to book value of equity at the end of year, and D is dummy variable equal to 1 if return < 0, and 0 otherwise. The firm-level C-Score is estimated using KW equation: C Scorei = l1 + l2 Sizei + l3 M /Bi + l4 Levi . (6) 24 Table 6. Discretionary accruals and AC members’ social capital (experts vs. non- experts). ABSDAi,t = b0 + b1a ∗SC EXP ACi,t + b1b ∗SC NEXP ACi,t + b1c ∗SC NonACi,t + b2 ∗LNDIRTOT + b3 ∗BDMTGi,t + b4 ∗PCTDIROUTi,t + b5 ∗EXPERTi,t + b6 ∗CEOCHAIRi,t + b7 ∗CEOAGEi,t + b8 ∗BIG4i,t + b9 ∗AUD ACi,t + b10 ∗LEVi,t + b11 ∗ZSCOREi,t + b12 ∗INSTMAJi,t  + b13 ∗LNMVi,t + b14 ∗CFi,t + b15 ∗DCFi,t + b16 ∗LROAi,t – 1 + b17 ∗DNIi,t + b18 ∗LOSSi,t + b19 ∗ACQi,t + b20 ∗BMi,t + bk ∗Yeari,t  bj ∗Indusi,t + 1i,t . + Exp. Sign – NEXP_DEGREE_AC – DEGREE_NonAC ? EXP_CONNECTEDNESS_AC – NEXP_CONNECTEDNESS_AC – CONNECTEDNESS_NonAC ? EXP_BROKERAGE_AC – NEXP_BROKERAGE_AC – BROKERAGE_NonAC ? EXP_STRONG_TIES_AC – NEXP_STRONG_TIES_AC – STRONG_TIES_NonAC ? LNDIRTOT ? BDMTG – PCTDIROUT – Model (2) Model (3) Model (4) Model (5) 0.266** (0.013) 0.372 (0.145) 0.190 (0.457) N. Carrera et al. EXP_DEGREE_AC Model (1) 0.042 (0.712) 0.503*** (0.000) 0.042 (0.891) 0.000 (0.623) 0.003** (0.013) –0.000 (0.850) 0.016 (0.106) –0.001 (0.270) –0.047 0.009 (0.432) –0.001 (0.231) –0.058 0.012 (0.263) –0.001 (0.259) –0.050 0.011 (0.281) –0.001 (0.235) –0.055 2.491 (0.110) 9.461** (0.027) –6.807 (0.271) 0.013 (0.199) –0.001 (0.245) –0.057 – CEOCHAIR – CEOAGE + BIG4 – AUD_AC ? LEV + ZSCORE – INSTMAJ + LNMV – CF – ΔCF – LROA + ΔNI + LOSS ? ACQ + BM – INTERCEPT ? Obs Adj. R2 (0.236) 0.007 (0.448) 0.008 (0.164) 0.000 (0.661) –0.014 (0.214) –0.004 (0.452) 0.047*** (0.036) –0.001 (0.100) 0.002 (0.730) –0.010*** (0.000) 0.193*** (0.001) –0.041 (0.472) –0.100** (0.021) 0.013 (0.873) –0.020 (0.190) –0.003 (0.601) –0.005*** (0.001) 0.214*** (0.000) 13,688 0.428 (0.118) 0.006 (0.527) 0.007 (0.203) 0.000 (0.643) –0.012 (0.286) –0.002 (0.730) 0.0467** (0.040) –0.001 (0.140) 0.002 (0.740) –0.011*** (0.000) 0.195*** (0.001) –0.042 (0.459) –0.100** (0.024) 0.015 (0.859) –0.021 (0.185) –0.002 (0.682) –0.005*** (0.001) 0.229*** (0.000) 13,688 0.428 (0.212) 0.008 (0.389) 0.007 (0.187) 0.000 (0.634) –0.012 (0.258) 0.002 (0.655) 0.046** (0.039) –0.001 (0.115) 0.002 (0.724) –0.011*** (0.000) 0.194*** (0.001) –0.042 (0.464) –0.099** (0.023) 0.015 (0.857) –0.020 (0.190) –0.002 (0.656) –0.005*** (0.001) 0.225*** (0.000) 13,688 0.428 (0.143) 0.009 (0.343) 0.007 (0.207) 0.000 (0.638) –0.013 (0.239) –0.001 (0.915) 0.046** (0.044) –0.001 (0.134) 0.002 (0.756) –0.011*** (0.001) 0.196*** (0.001) –0.043 (0.454) –0.100** (0.021) 0.015 (0.858) –0.021 (0.189) –0.002 (0.670) –0.005*** (0.001) 0.229*** (0.000) 13,688 0.428 (0.126) 0.010 (0.323) 0.007 (0.200) 0.000 (0.680) –0.012 (0.268) –0.002 (0.759) 0.046** (0.041) –0.001 (0.132) 0.001 (0.785) –0.011*** (0.000) 0.195*** (0.001) –0.043 (0.456) –0.101** (0.022) 0.015 (0.859) –0.020 (0.194) –0.002 (0.675) –0.005*** (0.001) 0.226*** (0.000) 13,688 0.428 25 Notes: Variables are defined in Table 2. Coefficients for year and industry variables are suppressed. P-values in parentheses are heteroscedastic firm- and year-corrected values (Petersen 2009). *Significant at 10%. **Significant at 5%. ***Significant at 1%. Accounting and Business Research EXPERT 26 Table 7. Accounting conservatism and AC members’ social capital. C Scorei,t = b0 + b1a ∗SC ACi,t + b1b ∗SC NonACi,t + b2 ∗LNDIRTOT + b3 ∗BDMTGi,t + b4 ∗PCTDIROUTi,t + b5 ∗EXPERTi,t + b6 ∗CEOCHAIRi,t + b7 ∗CEOAGEi,t + b8 ∗BIG4i,t + b9 ∗AUD ACi,t + b10 ∗LEVi,t + b11 ∗ZSCOREi,t + b12 ∗INSTMAJi,t + b13 ∗LNMVi,t + b14 ∗CFi,t + b15 ∗DCFi,t   bk ∗Yeari,t + bj ∗Indusi,t + 1i,t . + b16 ∗LROAi,t – 1 + b17 ∗DNIi,t + b18 ∗LOSSi,t + b19 ∗ACQi,t + b20 ∗BMi,t + Model (1) DEGREE_AC Model (2) Model (3) Model (4) −0.778*** (0.003) −0.172 (0.423) DEGREE_NonAC CONNECTEDNESS_AC BROKERAGE_AC −0.001** (0.028) −0.000 (0.344) BROKERAGE_NonAC STRONG_TIES_AC STRONG_TIES_NonAC BDMTG PCTDIROUT EXPERT CEOCHAIR CEOAGE 0.005* (0.071) −0.000 (0.724) −0.008 (0.202) −0.002 (0.621) −0.000 (0.883) −0.000 (0.753) 0.012*** (0.000) −0.000 (0.945) 0.003 (0.570) 0.001 (0.727) 0.000 (0.700) −0.000 (0.713) 0.010*** (0.001) −0.000 (0.845) −0.003 (0.626) 0.001 (0.851) 0.001 (0.455) −0.000 (0.678) 0.008*** (0.001) −0.000 (0.767) −0.000 (0.944) 0.001 (0.766) 0.000 (0.793) −0.000 (0.689) −10.215*** (0.002) −4.236 (0.287) 0.014*** (0.000) −0.000 (0.950) 0.005 (0.476) 0.001 (0.727) 0.001 (0.595) −0.000 (0.664) N. Carrera et al. −0.670*** (0.000) −0.020 (0.899) CONNECTEDNESS_NonACC LNDIRTOT Model (5) BIG4 AUD_AC LEV ZSCORE INSTMAJ LNMV ΔCF LROA ΔNI LOSS ACQ BM INTERCEPT Obs Adj. R2 0.003 (0.188) 0.006*** (0.001) 0.106*** (0.001) 0.000 (0.295) 0.000 (0.975) −0.034*** (0.000) −0.030*** (0.005) 0.016*** (0.002) 0.018** (0.025) −0.010 (0.447) −0.004 (0.538) −0.001 (0.717) 0.042*** (0.010) 0.311*** (0.000) 11,730 0.611 0.002 (0.386) 0.003** (0.032) 0.105*** (0.002) 0.000 (0.269) 0.000 (0.949) −0.033*** (0.000) −0.032*** (0.003) 0.016*** (0.001) 0.019** (0.016) −0.010 (0.449) −0.004 (0.535) −0.001 (0.622) 0.042*** (0.010) 0.306*** (0.000) 11,730 0.615 0.004 (0.131) 0.004** (0.028) 0.106*** (0.001) 0.000 (0.223) 0.000 (0.985) −0.034*** (0.000) −0.031*** (0.003) 0.016*** (0.001) 0.018** (0.016) −0.011 (0.438) −0.004 (0.522) −0.000 (0.841) 0.041** (0.010) 0.308*** (0.000) 11,730 0.609 0.003 (0.205) 0.006*** (0.001) 0.107*** (0.001) 0.000 (0.313) 0.000 (0.950) −0.034*** (0.000) −0.031*** (0.003) 0.016*** (0.001) 0.018** (0.028) −0.010 (0.444) −0.004 (0.536) −0.001 (0.663) 0.042*** (0.010) 0.302*** (0.000) 11,730 0.611 Accounting and Business Research CF 0.004* (0.083) 0.000 (0.822) 0.105*** (0.001) 0.000 (0.110) −0.000 (0.940) −0.036*** (0.000) −0.027*** (0.009) 0.014*** (0.002) 0.019** (0.012) −0.009 (0.503) −0.005 (0.495) 0.000 (0.886) 0.041** (0.012) 0.331*** (0.000) 11,730 0.606 Notes: Variables are defined in Table 2. Coefficients for year and industry variables are suppressed. P-values in parentheses are heteroscedastic firm- and year-corrected values (Petersen 2009). *Significant at 10%. **Significant at 5%. ***Significant at 1%. 27 28 Table 8. Accounting conservatism and AC members’ social capital (experts vs. non_experts). C Scorei,t = b0 + b1a ∗SC EXP ACi,t + b1b ∗SC NEXP ACi,t + b1c ∗SC NonACi,t + b2 ∗LNDIRTOT + b3 ∗BDMTGi,t + b4 ∗PCTDIROUTi,t + b5 ∗EXPERTi,t + b6 ∗CEOCHAIRi,t + b7 ∗CEOAGEi,t + b8 ∗BIG4i,t + b9 ∗AUD ACi,t + b10 ∗LEVi,t + b11 ∗ZSCOREi,t + b12 ∗INSTMAJi,t   + b13 ∗LNMVi,t + b14 ∗CFi,t + b15 ∗DCFi,t + b16 ∗LROAi,t – 1 + b17 ∗DNIi,t + b18 ∗LOSSi,t + b19 ∗ACQi,t + b20 ∗BMi,t + bk ∗Yeari,t + bj ∗Indusi,t + 1i,t . Model (1) EXP_DEGREE_AC Model (2) Model (3) Model (4) −0.130** (0.016) −0.344 (0.101) −0.213 (0.320) NEXP_DEGREE_AC DEGREE_NonAC EXP__CONNECTEDNESS_AC CONNECTEDNESS_NonAC EXP_BROKERAGE_AC −0.000 (0.709) −0.002* (0.079) −0.000 (0.126) NEXP_BROKERAGE_AC BROKERAGE_NonAC EXP_STRONG_TIES_AC NEXP_STRONG_TIES_AC STRONG_TIES_NonAC BDMTG PCTDIROUT EXPERT N. Carrera et al. −0.024 (0.828) −0.258* (0.059) −0.061 (0.686) NEXP_CONNECTEDNESS_AC LNDIRTOT Model (5) 0.005* (0.071) −0.000 (0.724) −0.008 (0.202) −0.002 (0.621) 0.011*** (0.001) −0.000 (0.870) −0.001 (0.932) −0.003 (0.431) 0.007** (0.019) −0.000 (0.743) −0.007 (0.282) −0.002 (0.479) 0.009*** (0.001) −0.000 (0.801) −0.001 (0.854) −0.003 (0.314) −0.978 (0.401) −7.065** (0.026) −5.446 (0.155) 0.014*** (0.000) −0.000 (0.897) 0.001 (0.914) −0.004 (0.210) CEOCHAIR CEOAGE BIG4 AUD_AC LEV ZSCORE INSTMAJ CF ΔCF LROA ΔNI LOSS ACQ BM 0.000 (0.712) −0.000 (0.662) 0.003 (0.163) 0.004*** (0.000) 0.106*** (0.001) 0.000 (0.210) −0.000 (0.908) −0.035*** (0.000) −0.030*** (0.008) 0.016*** (0.001) 0.019** (0.020) −0.009 (0.470) −0.004 (0.536) −0.000 (0.867) 0.041** (0.011) 0.000 (0.886) −0.000 (0.697) 0.004 (0.130) 0.001 (0.409) 0.106*** (0.001) 0.000 (0.121) −0.000 (0.949) −0.035*** (0.000) −0.029*** (0.006) 0.015*** (0.001) 0.018** (0.013) −0.010 (0.449) −0.005 (0.492) −0.000 (0.972) 0.041** (0.012) 0.000 (0.674) −0.000 (0.638) 0.004 (0.109) 0.004** (0.036) 0.106*** (0.001) 0.000 (0.230) −0.000 (0.931) −0.034*** (0.000) −0.032*** (0.003) 0.017*** (0.000) 0.019** (0.011) −0.010 (0.432) −0.004 (0.517) −0.000 (0.865) 0.041** (0.011) 0.000 (0.682) −0.000 (0.681) 0.003 (0.143) 0.004*** (0.002) 0.106*** (0.001) 0.000 (0.228) −0.000 (0.988) −0.034*** (0.000) −0.030*** (0.005) 0.016*** (0.000) 0.018** (0.023) −0.010 (0.447) −0.005 (0.510) −0.000 (0.821) 0.041** (0.010) 0.331*** (0.000) 11,730 0.606 0.320*** (0.000) 11,730 0.609 0.324*** (0.000) 11,730 0.610 0.316*** (0.000) 11,730 0.610 0.311*** (0.000) 11,730 0.610 Accounting and Business Research LNMV −0.000 (0.883) −0.000 (0.753) 0.004* (0.083) 0.000 (0.822) 0.105*** (0.001) 0.000 (0.110) −0.000 (0.940) −0.036*** (0.000) −0.027*** (0.009) 0.014*** (0.002) 0.019** (0.012) −0.009 (0.503) −0.005 (0.495) 0.000 (0.886) 0.041** (0.012) INTERCEPT Obs Adj. R2 29 Notes: Variables are defined in Table 2. Coefficients for year and industry variables are suppressed. P-values in parentheses are heteroscedastic firm- and year-corrected values (Petersen 2009). *Significant at 10%. **Significant at 5%. ***Significant at 1%. 30 N. Carrera et al. Specifically, we merge annual Compustat data with monthly Compustat security prices dataset for the years 2002–2010. We drop all firm years with negative book values, price per share less than $1 and with missing data for any variable required to run the estimation model. We also delete firm years in the top and bottom 1% of all the variables that are used in the estimation model, that is earnings, returns, size, and market-to-book ratio (Khan and Watts 2009, p. 136). Annual returns are estimated by cumulating monthly returns starting four months after the end of the firm’s previous fiscal year (Hayn 1995, Basu 1997, Khan and Watts 2009). The resulting estimation sample is 40,680 observations for the time-period 2003–2010, where we lose the first year due to requirement of lagged value of market value of equity. The C-Score from the estimation dataset is then merged with our sample including social capital and other control metrics, which gives us a final sample of 11,730 firm-year observations. Our matched firm-year sample over the 2003–2010 has mean C-Score of 0.086, which is slightly less than that of KW mean of 0.105. The median, however, is higher at 0.086, than KW median of 0.044. It seems, thus, that the degree of conservatism has increased over the time period 2003–2010. Our finding of increased value of C-score (higher degree of conservatism) is in line with Francis, Hasan, and Wu (2013), who document a C-score mean (median) values for 2006 as 0.209 (0.188). One possible reason for this could be the fact that firms, in general, have become more conservative in their financial reporting with the passage of time (Givoly and Hayn 2000). Using the explanatory variables as before, we re-estimate our models using C-Score as dependent variable. Since C-Score is measure of degree of conservatism of a firm, the predicted signs of the coefficients of social capital metrics should be negative. Output detailed in Tables 7 and 8 show that the same social capital metrics, as before, are now negative and statistically significant. These results provide further empirical support for our main findings. 5.4. Robustness checks We also conducted several robustness tests. First, we used other specifications for discretionary accruals. In particular, as in Bruynseels and Cardinaels (2014), we used Ball and Shivakumar’s (2006) specification, which takes into account asymmetric timeliness of gains and losses. We also considered the effects of future sales, as in Dechow et al. (2003). Our untabulated results showed that our findings remained unchanged. Our proxies for financial reporting quality are based on accounting figures as reported by firms. Arguably, a few extreme cases of accounting irregularities could be driving our results. Accordingly, we identified those firms that restated their accounting statements during 2003– 2010 and excluded these observations from our sample. We re-estimated all our models using a reduced sample of 12,047 firm years. Our untabulated results showed that our results were qualitatively the same. All empirical models suffer from potential ‘omitted-variable’ bias.4 As suggested by Prawitt et al. (2009) and Brazel et al. (2009), we included change in sales, abnormal employee change and standard deviation of sales (based on a rolling five-year window) as additional independent variables. Our untabulated results showed that none of these variables were statistically significant. In our setting, either firms or directors could have selected themselves, causing network measures to be endogenous. To evaluate such a possibility, we employed the Durbin–Wu– Hausman test to check for endogeneity (Wooldridge 2002). Our untabulated results showed that none of our variables of interest (e.g. DEGREE, BROKERAGE) are endogenous to the model specification. Nevertheless, we drew on prior research examining endogeneity in board interlocks and financial reporting (e.g. Fich and Shivdasani 2006, Krishnan et al. 2011, Accounting and Business Research 31 Bruynseels and Cardinaels 2014), and re-ran all our models using lagged values of the variables. Our untabulated results showed no qualitative change from our findings in the main section. Finally, recent research in social capital points towards firm-related geographic and demographic variables (Jha and Chen 2015) that, if not included, might result in ‘omitted-variable’ bias. In this regard, we re-ran our models using a firm fixed-effects technique in addition to Petersen (2009) two-level clustering. Introducing individual firm-level dummy variables in the model (firm fixed-effects) controls for any firm specific variables, such as a potential location and/or local-demographics effect, especially because these additional variables are controls in our setting (Wooldridge 2002). Finally, we re-ran our models using feasible generalised least square regression models for panel datasets (Wooldridge 2002). Employing Newey–West’s (1987) covariance matrix to account for serial correlations as well as heteroscedasticity (unreported), we found that the tenor of our findings remained unchanged. 6. Discussion and conclusions In this study, we have drawn on social capital theory to examine the association between AC directors’ social connections and financial reporting quality. According to mainstream social capital theory, AC directors embedded in social networks have access to information and resources that are not otherwise available (Anderson 2008). Accordingly, the social capital of these directors would exert beneficial effects on organisational effectiveness (Burt 1997, Adler and Kwon 2002) and, arguably, on financial reporting quality. Contrary to the expectations arising from this stream of research, our results suggest a significant and negative association between AC members’ social capital and financial reporting quality. More specifically, our results indicate that financial reporting quality deteriorates as a consequence of AC members’ positioning within social networks with respect to: centrality, quality of their connections, brokerage position, and strong ties (i.e. the triadic relationships within a network). Our examination of the relationship between directorships and financial reporting quality reveals that this only affects directors sitting on the ACs. This finding provides support for the regulators’ viewpoint that the oversight and monitoring of the financial reporting process is the domain of the AC members of the board (e.g. see SOX 2002). Unlike the strategic role of board of directors, ACs’ primary goal is to improve financial reporting practice and, hence, our results suggest that further research on firm’s financial reporting quality should focus on ACs. Our findings suggest that the social capital of non-AC members does not exert a significant influence on financial reporting quality. However, AC members’ social capital has a negative effect on the quality of financial reporting. Therefore, these results provide support to research examining the ‘dark side’ of social capital (Deth and Zmerli 2010). As noted by Portes (1998), social ties may pose restrictions on independent behaviour, as demands for conformity to a network’s norms do not necessarily promote efficiency and effectiveness but result in ‘solidarity benefits’ (Brass et al. 1998), that is, ‘the ties that bind may also turn into ties that blind’ (Powell and Smith-Doerr 1994, p. 393). As far as ACs are concerned, the social capital hypothesised to be beneficial for AC members in monitoring the financial reporting of a firm, appears to be more a case of a ‘contagion effect’ of ‘questionable’ accounting and disclosure practices via directors’ connections (Bouwman 2011, Cai et al. 2014). In this respect, Chiu et al. (2013) show that earnings management practices are more likely to be adopted by firms that share directors with firms that manage earnings. In a similar vein, Bouwman (2011) suggests that governance practices propagate across firms by virtue of the network effects of the overlapping directors. Specifically, cross-appointed directors generate an ‘influence effect’: ‘directors acquainted with the governance practices at other firms influence the firm’s governance to move towards the practices of those other firms. These network effects cause governance practices to converge’ 32 N. Carrera et al. (Bouwman 2011, p. 2358). In the case of ACs members, interlocked directorships could make them knowledgeable about questionable accounting and reporting practices, which could potentially be transferred across organisations. Furthermore, the negative association of STRONG_TIES and financial reporting quality, applying Krackhardt’s (1999, p. 187) interpretation of Simmelian ties,5 suggests that AC members have less autonomy and less independence, which in turn makes them more vulnerable to accepting doubtful accounting and reporting practices when facing conflicting goals. Finally, this adverse association between ACs’ social capital and effectiveness provide support for stakeholders’ negative view of board interlocks (Bizjak et al. 2009, Hillman et al. 2011). The subgroup analysis of ACs (financial experts vs. non-financial experts) reveals interesting insights. Based on the degree centrality variable, our results show that AC members designated as experts have a negative impact on reporting quality. However, this is not the case for the centrality of AC members designated as non-financial experts. Regarding the social capital metrics connectedness, brokerage, and strong ties, our results suggest that the higher the metrics for AC members designated as non-financial experts, the lower financial reporting quality will be. In particular, the higher the quality of their connections (connectedness), the higher the degree of their brokerage services within the networks (brokerage), and the higher their number of strong ties (strong ties), the lower the quality of financial reporting quality will be. In the case of experts’ social capital, conversely, there is no significant impact on the quality of financial reporting. Taken together, our results regarding degree centrality for AC experts are in line with Tanyi and Smith’s (2015) finding that busy AC financial experts are associated with lower financial reporting quality. That is, we replicate the same relationship that has previously been documented under the ‘busyness hypothesis’ (Ferris et al. 2003, Fich and Shivdasani 2006, Andres et al. 2013). Under the current research design, it is nearly impossible to disentangle whether degree centrality captures a time-commitment effect or a knowledge transfer effect. In fact, it is a joint measure for both. This discussion leads us to question to what extent all metrics based on SNA are a good proxy for social capital. As noted by Horton et al. (2012, p. 403), degree centrality is not always considered as a measure of social capital. Furthermore, Horton et al. (2012, p. 403) argue that connectedness, brokerage, and strong ties capture relevant dimensions of social capital. Furthermore, these three metrics measure an individual’s power within a network (see also Burt 1992, Krackhardt 2010). By using these dimensions in our analysis, we focus on the social capital embedded within the network and depart from Tanyi and Smith’s (2015) reliance on the ‘number of positions’ (degree centrality). AC members designated as experts arguably engage in networks formed by professionals with a technical background and this does not bring about negative externalities of social capital. Therefore, financial experts’ social capital, if measured by the quality of their connections and not by the number of seats in ACs, does not affect financial reporting quality. AC non-financial experts, on the other hand, are arguably appointed as AC directors as a consequence of the quality of their social connections and not because of their professional background and technical expertise; they are powerful individuals. As noted by Adler and Kwon (2002, p. 33), non-financial experts’ entrepreneurial brokering may be problematic for their performance as AC members, which may bring about a dominant role over the ACs’ decisions that would result in questionable accounting and reporting practices. The joint effect of AC financial experts sitting on many committees (degree centrality), who can transfer doubtful accounting and reporting practices across organisations, and AC non-financial experts with high social capital and powerful connections, measured by connectedness, brokerage, and strong ties, exerts a negative impact on financial reporting quality. Accounting and Business Research 33 Our results also have policy implications. Research drawing on agency theory suggests that busy directors are less diligent (Fich and Shivdasani 2006, Sharma and Iselin 2012, Tanyi and Smith 2015). We qualify Tanyi and Smith’s (2015) findings by demonstrating that the quality of the connections of non-financial experts is relevant for financial reporting quality, irrespective of how ‘busy’ they are. These individuals are connected to other key individuals, they bridge networks and join close subgroups, making them influential and powerful within the network. Therefore, we suggest that policy-makers might wish to consider that the effects of multiple directorships are mediated by the financial expertise of AC members. Accordingly, regulators could impose restrictions on multiple directorships of financial experts, as well as requiring financial expertise for all members of ACs. Our study has limitations that may encourage further work. First, we focus on firm ties through shared board members based on data gathered from the Corporate Library Board dataset. Although our dataset covers a large number of firms and their directors, some individuals may arguably have social connections with out-of-sample firms. In our setting, however, the measures of social capital will be biased downwards and will preclude the finding of any statistical significance. Second, our dataset does not provide observations on informal social connections, though research has shown that informal and voluntary networks among AC members influence AC effectiveness (Turley and Zaman 2007; see also Beasley et al. 2009). Therefore, this makes our social capital measures conservative and any increase in capturing the informal ties would only strengthen our findings. Third, our data do not allow segmenting on types of social ties (e.g. educational ties, friendship ties) and/or detailed refining of subgroups of different types of human capital (e.g. accounting, financial, legal, or industry expertise) as done by some prior studies (Krishnan et al. 2011, Bruynseels and Cardinaels 2014, Cohen et al. 2014). However, the subgroup analysis is a first step in disentangling the effects of social/human capital in line with Johnson et al.’s (2013) recommendation. Fourth, our database does not allow us to track some internal processes within the networks (e.g. contagion effect of questionable accounting and reporting practices), and hence further research may look into the transfer of accounting knowledge across organisations and how this is mediated by AC members’ social capital. Future research could further theorise on the reasons why the impact of social capital may differ across organisational units (e.g. boards versus ACs). Furthermore, we still have much to learn about a ‘dark side’ of social capital in the context of board of directors. Specifically, future research could explore further the determinants of non-experts’ social capital (e.g. power and prestige) that explain its negative effects on financial reporting quality. Finally, we suggest that policy-makers might wish to limit the number of multiple directorships for financial experts and assess the actual contribution of non-financial experts to AC effectiveness; arguably, non-financial experts are appointed as AC members for their social connections rather than for their technical knowledge. Acknowledgements Previous versions of this paper have been presented at the European Accounting Association (EAA) Congress, the American Accounting Association (AAA) Meeting, and at the Workshop Raymond Konopka. We are grateful to the participants at these conferences for their many helpful suggestions, as well as to Mark Clatworthy (editor), Macario Cámara and Luis Fernández-Revuelta. Disclosure statement No potential conflict of interest was reported by the authors. 34 N. Carrera et al. Funding Financial support from the Spanish Ministry of Economy and Competitiveness [ECO2013-48392-P] is gratefully acknowledged. Notes 1. 2. 3. 4. 5. According to the SEC, an AC financial expert is a person who has, through education and experience as a public accountant, auditor, principal financial officer, controller or principal accounting officer, of a company […] or experience in one or more positions that involve the performance of similar functions […] the following attributes: (1) An understanding of generally accepted accounting principles and financial statements; (2) Experience applying such generally accepted accounting principles in connection with the accounting for estimates, accruals, and reserves that are generally comparable to the estimates, accruals and reserves, if any, used in the registrant’s financial statements; (3) Experience preparing or auditing financial statements that present accounting issues that are generally comparable to those raised by the registrant’s financial statements; (4) Experience with internal controls and procedures for financial reporting; (5) An understanding of audit committee functions. (SEC 2003) As noted above, degree centrality is the closest metric to the measures of ‘busy boards’ used in previous studies (Andres et al. 2013). We use this metric as a proxy for social capital because it is, relatively, the most suitable centrality measure for capturing an individual’s access to information and her potential communication activity (Freeman 1979, Tsai 2000, p. 931), and because it allows us to compare our findings with previous literature on the social ties of AC members. We report two-tailed p-values. Previous research has also addressed the percentage of shares held by AC members and auditors’ expertise as possible explanatory variables. As our sample is post-SOX, we follow Bruynseels and Cardinaels (2014), who suggest that AC members’ shareholdings are not a valid independent variable under the SOX guidance. Furthermore, Minutti-Meza (2013) shows that the current method for estimating auditor expertise (based on percentage of audit firm market share) is not a valid proxy for expertise. Hence, we do not include these variables. Krackhardt (1999, pp. 187–8) argues that the formation of a group, a triad or larger, restricts a member’s public behaviour vis-à-vis other members, as groups develop rules that members must accept if they want to join them. Therefore, ties bound by a third party could reduce individuals’ autonomy, power and independence in their relationship with other members. References Adler, P.S. and Kwon, S.W., 2002. Social capital: prospects for a new concept. Academy of Management Review, 27 (1), 17–40. Altman, E., 2000. Predicting the financial distress of companies: revisiting the Z-score and Zeta models. Working Paper, New York University. Available from: http://pages.stern.nyu.edu/∼ealtman/ PredFnclDistr.pdf [Accessed 18 March 2017]. Anderson, M.H., 2008. Social networks and the cognitive motivation to realize network opportunities: a study of managers’ information gathering behaviors. Journal of Organizational Behavior, 29 (1), 51–78. Andres, C., Bongard, I., and Lehmann, M., 2013. Is busy really busy? Board governance revisited. Journal of Business Finance and Accounting, 40 (9–10), 1221–1246. Ashbaugh, H., Lafond, R., and Mayhew, B.W., 2003. Do nonaudit services compromise auditor independence? Further evidence. Accounting Review, 78 (3), 611–639. Ball, R. and Shivakumar, L., 2006. The role of accruals in asymmetrically timely gain and loss recognition. Journal of Accounting Research, 44 (2), 207–242. Basu, S., 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics, 24 (1), 3–37. Beasley, M.S., Carcello, J.V., Hermanson, D.R., and Lapides, P.D., 2000. Fraudulent financial reporting: consideration of industry traits and corporate governance mechanisms. Accounting Horizons, 14 (4), 441–454. Accounting and Business Research 35 Beasley, M.S., Carcello, J.V., Hermanson, D.R., and Neal, T.L., 2009. The audit committee oversight process. Contemporary Accounting Research, 26 (1), 65–122. Beattie, V., Fearnley, S., and Hines, T., 2012. Do UK audit committees really engage with auditors on audit planning and performance? Accounting and Business Research, 42 (3), 349–375. Beckham, C.M. and Haunschild, P.R., 2002. Network learning: the effects of partners’ heterogeneity of experience on corporate acquisitions. Administrative Science Quarterly, 47 (1), 92–124. Bédard, J. and Gendron, Y., 2010. Strengthening the financial reporting system: can audit committees deliver? International Journal of Auditing, 14 (2), 174–210. Bédard, J., Chtourou, S.M., and Courteau, L., 2004. The effect of audit committee expertise, independence, and activity on aggressive earnings management. Auditing: A Journal of Practice & Theory, 23 (2), 15–37. Bizjak, J., Lemmon, M., and Whitby, R., 2009. Option backdating and board interlocks. Review of Financial Studies, 22 (11), 4821–4847. Blue Ribbon Committee, 1999. Report and Recommendations of the Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Committee. New York: BRC. Bonacich, P., 1972. Factoring and weighting approaches to status scores and clique identification. Journal of Mathematical Sociology, 2 (1), 113–120. Borgatti, S.P., 2005. Centrality and network flow. Social Networks, 27 (1), 55–71. Borgatti, S.P. and Lopez-Kidwell, V., 2010. Network theorizing. In: P. Carrington and J. Scott, eds. The Sage Handbook of Social Network Analysis. London: Sage Publications, 40–54. Borgatti, S.P. and Halgin, D.S., 2011. On network theory. Organization Science, 22 (5), 1526–5455. Bouwman, C.H.S., 2011. Corporate governance propagation through overlapping directors. Review of Financial Studies, 24 (7), 2358–2359. Brass, D.J., Butterfield, K.D., and Skaggs, B.C. 1998. Relationships and unethical behaviour: a social network perspective. Academy of Management Review, 23 (1), 14–31. Brazel, J.F., Jones, K.L., and Zimbelman, M.F., 2009. Using nonfinancial measures to assess fraud risk. Journal of Accounting Research, 47 (5), 1135–1166. Bruynseels, L. and Cardinaels, E., 2014. The audit committee: management watchdog or personal friend of the Ceo? Accounting Review, 89 (1), 113–145. Burt, R.S., 1992. Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard University Press. Burt, R.S., 1997. The contingent value of social capital. Administrative Science Quarterly, 42 (2), 339–365. Burt, R.S., 2000. The network structure of social capital. Pre-print of a chapter in Research in Organizational Behaviour, vol. 22, by R.I. Sutton and B.M. Staw. Greenwich: JAI Press, 345–423. Burt, R.S., 2001. The social capital of structural holes. Pre-print of Chapter 7. In: M.F. Guillén, R. Collins, P. England, and M. Meyer, eds. New Directions in Economic Sociology. New York: Russell Sage Foundation, 201–247. Burt, R.S., 2005. Brokerage and Closure: An Introduction to Social Capital. New York: Oxford University Press. Burt, R.S., 2007. Second-hand brokerage: evidence on the importance of local structure for managers, bankers, and analysts. Academy of Management Journal, 50 (1), 119–148. Cai, Y., Dhaliwal, D.S., Kim, Y., and Pan, C., 2014. Board interlocks and the diffusion of disclosure policy. Review of Accounting Studies, 19 (3), 1086–1119. Carcello, J.V., Neal, T.L., Palmrose, Z.-V., and Scholz, S., 2011. CEO involvement in selecting board members, audit committee effectiveness, and restatements. Contemporary Accounting Research, 28 (2), 396–430. Chiu, P.C., Teoh, S.H., and Tian, F., 2013. Board interlocks and earnings management contagion. Accounting Review, 88 (3), 915–944. Cohen, J.R., Hoitash, U., Krishnamoorthy, G., and Wright, A.M., 2014. The effect of audit committee industry expertise on monitoring the financial reporting process. Accounting Review, 89 (1), 243–273. Coleman, J.S., 1988. Social capital in the creation of human-capital. American Journal of Sociology, 94 (Supplement), S95–S120. Dechow, P.M. and Sloan, R.G., 1995. Detecting earnings management. Accounting Review, 70 (2), 193–225. Dechow, P.M., Richardson, S., and Tuna, I., 2003. Why are earnings kinky? An examination of the earnings management explanation. Review of Accounting Studies, 8 (2/3), 355–384. Dechow, P.M., Ge, W., and Schrand, C., 2010. Understanding earnings quality: a review of the proxies, their determinants and their consequences. Journal of Accounting and Economics, 50 (2/3), 344–401. 36 N. Carrera et al. DeFond, M., Hann, R., and Hu, X., 2005. Does the market value financial expertise on audit committees of boards of directors? Journal of Accounting Research, 43 (2), 153–193. Deth, J.W. and Zmerli, S., 2010. Introduction: civicness, equality, and democracy – a ʻdark side’ of social capital? American Behavioural Scientist, 53 (5), 631–639. Dhaliwal, D.S., Naiker, V., and Navissi, F., 2010. The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees. Contemporary Accounting Research, 27 (3), 787–827. Ferris, S.P., Jagannathan, M., and Pritchard, A.C., 2003. Too busy to mind the business? Monitoring by directors with multiple board appointments. Journal of Finance, 58 (3), 1087–1111. Fich, E.M. and Shivdasani, A., 2006. Are busy boards effective monitors? Journal of Finance, 61 (2), 689–724. Field, J., 2008. Social Capital. 2nd ed. London: Routledge. Field, L., Lowry, M., and Mkrtchyan, A., 2013. Are busy boards detrimental? Journal of Financial Economics, 109 (1), 63–82. Francis, B., Hasan, I., and Wu, Q., 2013. The benefits of conservative accounting to shareholders: evidence from the financial crisis. Accounting Horizons, 27 (2), 319–346. Freeman, L.C., 1979. Centrality in social networks conceptual clarification. Social Networks, 1 (3), 215–239. Geiger, M.A. and North, D.S., 2006. Does hiring a new CFO change things? An investigation of changes in discretionary accruals. Accounting Review, 81 (4), 781–809. Givoly, D., and Hayn C., 2000. The changing time-series properties of earnings, cash flows and accruals: has financial reporting become more conservative? Journal of Accounting and Economics, 29 (3), 287–320. Habib, A. and Bhuiyan, M.B.U., 2016. Problem directors on the audit committee and financial reporting quality. Accounting and Business Research, 46 (2), 121–144. Hansen, M.T., 1999. The search-transfer problem: the role of weak ties in sharing knowledge across organization subunits. Administrative Science Quarterly, 44 (1), 82–111. Haunschild, P., 1993. International imitation: the impact of interlocks on corporate acquisition activity. Administrative Science Quarterly, 38 (4), 564–592. Hayn, C., 1995. The information content of losses. Journal of Accounting and Economics, 20 (2), 125–153. Hermalin, B.E. and Weisbach, M.S., 2003. Board of directors as an endogenously determined institution: a survey of the economic literature. Part 1: a review of the literature on corporate governance. Federal Reserve Bank of New York, 9 (1), 7–26. Hillman, A.J., Shropshire, C., Certo, S.T., Dalton, D.R., and Dalton, C.M., 2011. What I like about you: a multilevel study of shareholder discontent with director monitoring. Organization Science, 22 (3), 675– 687. Hoitash, U., 2011. Should independent board members with social ties to management disqualify themselves from serving on the board? Journal of Business Ethics, 99 (3), 399–423. Horton, J., Millo, Y., and Serafeim, G., 2012. Resources or power? Implications of social networks on compensation and firm performance. Journal of Business Finance and Accounting, 39 (3/4), 399–426. Huang, H.W., Rose-Green, E., and Lee, C.C., 2012. CEO age and financial reporting quality. Accounting Horizons, 26 (4), 725–740. Javakhadze, D., Ferris, S.P., and French, D.W., 2016. Social capital, investments, and external financing. Journal of Corporate Finance, 37, 38–55. Jha, A. and Chen, Y., 2015. Audit fees and social capital. Accounting Review, 90 (2), 611–639. Johansen, T.R. and Pettersson, K., 2013. The impact of board interlocks on auditor choice and audit fees. Corporate Governance: An International Review, 21 (3), 287–310. Johnson, S.G., Schnatterly, K., and Hill, A.D., 2013. Board composition beyond independence: social capital, human capital, and demographics. Journal of Management, 39 (1), 232–262. Jones, J.J., 1991. Earnings management during import relief investigations. Journal of Accounting Research, 29 (2), 193–228. Khan, M. and Watts, R.L., 2009. Estimation and empirical properties of a firm-year measure of accounting conservatism. Journal of Accounting and Economics, 48 (2), 132–150. Kilduff, M. and Krackhardt, D., 1994. Bringing the individual back in: a structural analysis of the internal market for reputation in organizations. Academy of Management Journal, 37 (1), 87–108. Kim, Y., 2005. Board network characteristics and firm performance in Korea. Corporate Governance: An International Review, 13 (6), 800–808. Kim, Y. and Cannella, A.A., 2008. Toward a social capital theory of director selection. Corporate Governance: An International Review, 16 (4), 282–293. Accounting and Business Research 37 Klein, A., 2002. Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics, 33 (3), 375–400. Kor, Y.Y. and Sundaramurthy, C., 2009. Experience-based human capital and social capital of outside directors. Journal of Management, 35 (4), 981–1006. Krackhardt, D., 1998. Simmelian ties: super strong and sticky. In: R.M. Kramer and M. Neale, eds. Power and Influence in Organizations. Thousand Oaks, CA: Sage, 21–38. Krackhardt, D., 1999. The ties that torture: Simmelian tie analysis in organizations. In S.B. Andrews and D. Knocke, eds. Research in the Sociology of Organizations (16), 183–210. Greenwich, CT: JAI. Krackhardt, D., 2010. Social networks. In: J.M. Levine and M. A. Hogg, eds. Encyclopedia of Group Processes & Intergroup Relations. Los Angeles: Sage, 17–821. Krishnan, G.V., 2005. The association between Big 6 auditor industry expertise and the asymmetric timeliness of earnings. Journal of Accounting, Auditing and Finance, 20 (3), 209–228. Krishnan, G.V. and Visvanathan, G., 2008. Does the SOX definition of an accounting expert matter? The association between audit committee directors’ accounting expertise and accounting conservatism. Contemporary Accounting Research, 25 (3), 827–858. Krishnan, J., Yuan, W., and Wanli, Z., 2011. Legal expertise on corporate audit committees and financial reporting quality. Accounting Review, 86 (6), 2099–2130. Larcker, D.F., So, E.C., and Wang, C.C.Y., 2013. Boardroom centrality and firm performance. Journal of Accounting and Economics, 55 (2–3), 225–250. Lee, S.H.M., 2013. Ethics and expertise: a social networks perspective. Journal of Business Ethics, 118 (3), 607–621. Lin, N., 2001. Social Capital: A Theory of Social Structure and Action. Cambridge, UK: Cambridge University Press. Mallin, C.A. and Michelon, G., 2011. Board reputation attributes and corporate social performance: an empirical investigation of the U.S. best corporate citizens. Accounting and Business Research, 41 (2), 119–144. Menon, K. and Williams, D.D., 2004. Former audit partners and abnormal accruals. Accounting Review, 79 (4), 1095–1118. Minutti-Meza, M., 2013. Does auditor industry specialization improve audit quality? Journal of Accounting Research, 51 (4), 779–817. Nahapiet, J. and Ghoshal, S. 1998. Social capital, intellectual capital, and the organizational advantage. Academy of Management Review, 23 (2), 242–266. Newey, W.K. and West, K.D., 1987. A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55 (3), 703–708. NYSE, 2010. Report of the New York Stock Exchange Commission on Corporate Governance. Available https://www.nyse.com/publicdocs/nyse/listing/NYSE_Corporate_Governance_Guide.pdf from: [Accessed 28 August 2014]. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Review of Financial Studies, 22 (1), 435–480. Portes, A., 1998. Social capital: its origins and applications in modern sociology. Annual Review of Sociology, 24, 1–24. Powell, W.W. and Smith-Doerr, L., 1994. Networks and economic life. In: N. Smelser and R. Swedberg, eds. Handbook of Economic Sociology. Princeton, NJ: Princeton University Press, 368–402. Prawitt, D.F., Smith, J.L., and Wood, D.A., 2009. Internal audit quality and earnings management. Accounting Review, 84 (4), 1255–1280. Putnam, R.D., 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster. Reagans, R. and McEvily, B., 2003. Network structure and knowledge transfer: the effects of cohesion and range. Administrative Science Quarterly, 48 (2), 240–267. Reppenhagen, D.A., 2010. Contagion of accounting methods: evidence from stock option expensing. Review of Accounting Studies, 15 (3), 629–657. Reynolds, J.K. and Francis, J.R., 2000. Does size matter? The influence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics, 30 (3), 375–400. Scott, J., 1991. Social Network Analysis: A Handbook. London: SAGE Publications. SEC, 2003. Standards Relating to Listed Company Audit Committees. Available from: http://www.sec.gov/ rules/final/33-8220.htm [Accessed 2 August 2014]. Sharma, V.D. and Iselin, E.R., 2012. The association between audit committee multiple-directorships, tenure, and financial misstatements. Auditing: A Journal of Practice & Theory, 31 (3), 149–175. 38 N. Carrera et al. Simmel, G., 1950. The Sociology of Georg Simmel. New York: Free Press. SOX, 2002. Sarbanes-Oxley Act of 2002. Available from: http://fl1.findlaw.com/news.findlaw.com/cnn/ docs/gwbush/sarbanesoxley072302.pdf [Accessed 28 August 2014]. Stuart, T.E. and Yim, S., 2010. Board interlocks and the propensity to be targeted in private equity transactions. Journal of Financial Economics, 97 (1), 174–189. Tanyi, P.N. and Smith, D.B., 2015. Busyness, expertise, and financial reporting quality of audit committee chairs and financial experts. Auditing: A Journal of Practice & Theory, 34 (2), 59–89. Tortoriello, M. and Krackhardt, D., 2010. Activating cross-boundary knowledge: the role of Simmelian ties in the generation of innovations. Academy of Management Journal, 53 (1), 167–181. Treadway Commission, 1987. Report of the National Commission on Fraudulent Financial Reporting. Washington, DC: Government Printing Office. Available from: http://www.coso.org/publications/ ncffr.pdf [Accessed 24 August 2014]. Tsai, W., 2000. Social capital, strategic relatedness and the formation of intraorganizational linkages. Strategic Management Journal, 21 (9), 925–939. Turley, S. and Zaman, M., 2007. Audit committee effectiveness: informal processes and behavioural effects. Accounting, Auditing & Accountability Journal, 20 (5), 765–788. Useem, M. 1984. The Inner Circle: Large Corporations and the Rise of Business Political Activity. New York: Oxford University Press. Vafeas, N., 1999. Board meeting frequency and firm performance. Journal of Financial Economics, 53 (1), 113–142. Vafeas, N., 2001. On audit committee appointments. Auditing: A Journal of Practice & Theory, 20 (1), 197– 207. Vafeas, N., 2005. Audit committees, boards, and the quality of reported earnings. Contemporary Accounting Research, 22 (4), 1093–1122. Walker, M., 2013. How far can we trust earnings numbers? What research tells us about earnings management. Accounting and Business Research, 43 (4), 445–481. Wasserman, S. and Faust, K., 1994. Social Network Analysis: Methods and Applications. New York and Cambridge, UK: Cambridge University Press. Westphal, J.D., Seidel, M.D., and Stewart, K., 2001. Second-order imitation: uncovering latent effects of board networks ties. Administrative Science Quarterly, 46 (4), 717–747. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48 (4), 817–838. Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press. Xie, B., Davidson III, W.N., and Dadalt, P.J., 2003. Earnings management and corporate governance: the role of the board and the audit committee. Journal of Corporate Finance, 9 (3), 295–316. Yang, J.S. and Krishnan, J., 2005. Audit committees and quarterly earnings management. International Journal of Auditing, 9 (3), 201–219. Yermack, D. 2004. Remuneration, retention, and reputation incentives for outside directors. Journal of Finance, 59, 2281–2308. Zaheer, A., Gözübüyük, R., and Milanov, H., 2010. It’s the connections: the network perspective in Interorganizational research. Academy of Management Perspectives, 24 (1), 62–77. 39 Accounting and Business Research Appendix Social capital – network measures estimation The social capital dimensions used in this study are based on network measures. Network measures are individual-level constructs representing the social ties of the focal director (node) to other directors in the network. Conceptually, if one of a company’s board members holds multiple directorships, then there also exist social ties between the firms involved. To illustrate, consider the eight-director network A–H detailed in Figure A1. Figure A1. Director network. SNA uses an adjacency matrix to represent the structure of Figure A1, which numerically depicts the connections of the directors. The adjacency matrix for Figure A1 is shown in Table A1: Table A1. Adjacency matrix for network in Figure A1. A A B C D E F G H 1 1 1 1 1 1 B C D E F G 1 1 1 1 1 1 1 1 H 1 1 1 1 1 1 In Table A1, the first thing to note is that only the non-diagonal elements of the matrix can have a value, that is, the diagonal elements between two directors are 0 by definition (Andres et al. 2013). Second, the matrix is symmetric around the diagonal, because if director A knows director B, then director B must know director A. Table A2 details the estimated network measures for each of the directors in Figure A1. The measures for this particular example are estimated as follows: DEGREE is measured as the total number of ties that exist on a given node. For example, director A has 6 ties, while directors G and H have 2 and 1 ties, respectively. This score is then normalised by n − 1, where n is the total number of nodes in a given network. Therefore, the normalised values of DEGREE for A, G and H will be 0.85 (6/7), 0.29 (2/7), and 0.14 (1/7), respectively. Column 2 of Table A2 details the values of DEGREE used in the analysis. CONNECTEDNESS is measured as the eigenvector centrality for each node in the network. Let xi be the eigenvector of the ith node and let A = [aij] be the adjacency matrix. Then mathematically, the eigenvector is represented as xi = k  j=Ci xj = k N  j=1 aij xj = N 1 1 aij xj ; AX , l j=1 l 40 N. Carrera et al. where k is the proportionality constant, Ci is the set of nodes connected to the ith node, and A and X are matrix representations. Typically, there are many values of λ, so additional constraints are imposed, which require that all solutions of the vector X must be positive and that the greatest value of λ be used in solving the equation. The eigenvector for each node in Figure A1 is listed in column 3 (Table A2). BROKERAGE, what Burt (1992) calls effective network size, is measured as non-normalised DEGREE (ego) – average non-normalised DEGREE (alters), where ‘ego’ refers to the node of interest while ‘alter’ refers to the nodes that are directly connected to ego. For example, if A in Figure A1 is the focal node, then its alters will be B, C, D, E, F, and G. The total number of ties among the alters (not counting their ties to ego) is 6 (B is connected to 1 alter [C], C to 2, D to 1, E to 1, and F to 1). The average DEGREE of alters, then is 1 (6/6), and the effective network size (BROKERAGE) of ego A is 5 (6–1). For node C as ego, the alters would be A, B, and D. The total number of ties for alters excluding ego would be 4 (A is connected to 2 alters, B to 1, and D to 1). The average DEGREE of alters is 1.33 (4/3), and the effective network size (BROKERAGE) for node C is 1.67 (3 − 1.33). Column 4 of Table A2 details the measure for each node. STRONG_TIES are estimated using normalised values of Simmelian ties (Simmel 1950, Krackhardt 1998). This measure is estimated by identifying all triads (interlocks of three or more people, each of whom is connected to all the others, e.g. ABC) and then estimating the total number of unique ties that the focal node has within its triads. For example, node A in Figure A1 belongs to three triads (ABC, ACD, and AEF). In these three triads, the number of unique ties that A has is 5 (the tie AC appears in two different triads), which is the raw count for Simmelian ties. This value is normalised by n − 1, where n is the total number of nodes. Hence, the Simmelian ties value (column 7, Table A2) for node A is 0.71 (5/7). For nodes G and H, the value of Simmelian ties is 0, because they do not belong to any triad (Table A2, column 5). Table A2. Measures for director’s network in Figure A1. (1) (2) (3) (4) (5) (6) (7) DEGREE CONNECTEDNESS BROKERAGE STRONG_TIES Node (Raw DEGREE (Eigenvector (Effective Triad Simmelian (Simmelian ties title count) (Normalised) centrality) network size) count ties normalised) A B C D E F G H 6 2 3 2 2 2 2 1 0.86 0.29 0.43 0.29 0.29 0.29 0.29 0.14 0.87 0.47 0.59 0.47 0.42 0.42 0.31 0.10 5 1 1.67 1 1 1 2 1 3 1 2 1 1 1 0 0 5 2 3 2 2 2 0 0 0.71 0.29 0.43 0.29 0.29 0.29 0.00 0.00