Board structure and earnings persistence: The mediation effect of board processes

Abstract

Prior research has overlooked board processes and examined direct links between board structure and monitoring financial reporting with mixed results. We use an input–process–output heuristic model to argue that board processes mediate the association between board structure and earnings persistence. Our results support this assertion by showing the mediating effect of board processes and the impact on managerial behaviour in relation to monitoring financial reporting. We contribute to the accounting and finance literature by demonstrating the importance of examining alternative theoretical models and board-process variables, along with structure. This article also highlights the importance of establishing best-practice guidelines around board processes in governance regulations.

JEL Classification: M41, G34

Keywords

Board processes cognitive conflict earnings persistence effort norms knowledge and skills mediation

1. Introduction

Boards of directors have a fiduciary duty to provide oversight of activities undertaken by managers to ensure the integrity of financial reporting (Cohen et al., 2004) and attributes of earnings (Dechow et al., 2010). Consistent with this view, extant accounting research has examined the effect of the board of directors on earnings management and earnings quality (Bradbury et al., 2006; Davidson et al., 2005; Kent et al., 2010; Mather and Ramsay, 2006; Peasnell et al., 2000; Strydom et al., 2017; Xie et al., 2003); the ex-post consequences of low-quality financial reporting (Abbott et al., 2012; Chen and Zhou, 2007) and perceived financial-reporting quality and earnings informativeness (Ahmed and Duellman, 2007; Petra, 2007). These large-sample archival studies have used board-structure variables (e.g. board size, independent directors, presence of female directors and presence of financial experts) to measure the effectiveness of the board. However, most studies that link board characteristics with various financial-reporting outcomes have overlooked the inner-working behaviour of the board (i.e. board processes).¹ Thus, the predictive power of these input–output parsimonious models has failed to emerge, even in well-researched areas (Johnson et al., 1996). In contrast, examining the processes that influence board structure and firm-level outcomes has demonstrated high-quality explanatory power in the literature on small-group performance (LePine et al., 2008). Motivated by the importance of board processes, a body of research has examined the effect of board processes on firm performance.² Furthermore, McNulty et al. (2013) examined the effect of board processes on financial risk taking during the global financial crisis of 2008–2009 and found that board processes are important determinants of financial risk taking. However, prior research has not examined the mediation effect of board processes on the association between board structure and aspects of financial-reporting integrity or attributes of earnings.

Similarly, corporate scandals and/or failures that have occurred in some well-known companies around the world (e.g. Enron in the United States, HIH Insurance in Australia and Paramalat in Italy) have drawn attention to deficiencies in corporate governance. Although some of these cases had problems with board structure such as a lack of director independence, this has not been the reason for all ineffective boards (Sonnenfeld, 2002). For example, in a landmark and well-publicised case, the Federal Court of Australia held the directors of Centro Australia liable for failing to render due diligence in reviewing financial statements despite the board meeting most of the required structural attributes (Australian Securities and Investments Commission v Healey 717: Federal Court of Australia, 2011). The current study adds to the literature by empirically examining the mediation effect of board processes on the association between board structure and earnings persistence, which is one attributes of earnings quality.³

Forbes and Milliken (1999) proposed an input–process–output heuristic model that can be applied to gaining better understanding of the board processes that may explain the association between board structure and various firm-level outcomes. Following this approach, some studies have examined internal board behaviour to assess board effectiveness (Huse, 2005; Minichilli et al., 2009; Van Ees et al., 2008; Wan and Ong, 2005; Zona and Zattoni, 2007). As boards of directors are ‘elite, and episodic decision-making groups that face complex tasks pertaining to strategic issue processing’ (Forbes and Milliken, 1999: 492), they confront interaction difficulties that can prevent them from fulfilling their tasks (Hambrick et al., 2008). In these circumstances, board effectiveness largely depends on board processes, which are defined as social-psychological processes pertaining to critical discussion, group participation and interaction and the exchange of information (Forbes and Milliken, 1999). Relying on these assertions, we examine the meditating effect of three board processes, namely, effort norms (i.e. the level of the effort that each board member is expected to contribute towards group tasks), cognitive conflicts (i.e. differences in viewpoints, ideas and opinions about board tasks) and use of knowledge and skills (i.e. awareness, collaboration and integration of each board members’ knowledge and skills) on the association between board structure (i.e. board size, board independence, presence of female directors and presence of financial experts) and earnings persistence.

Combining survey responses and archival data from Australian listed firms, we find that board processes mediate the association between board structure and earnings persistence. Based on partial least squares–structural equation modelling (PLS-SEM), we find that monitoring financial reporting significantly mediates the association between board processes and earnings persistence. We also find that board independence significantly affects monitoring financial reporting via board processes that in turn affect earnings persistence. However, the other structural variables (i.e. board size, presence of female directors and presence of financial experts) are not significantly associated with board processes. In addition, we use a parsimonious input–process–output model that examines the mediation effect of individual board processes on the association between board structure and earnings persistence and find that two of the process variables (i.e. effort norms and dealing with cognitive conflicts) fully mediate the association between board structure and earnings persistence.

We make the following contributions to literature and practice. First, we extend the conventional agency-theory input–output approach to the effects of boards in contemporary accounting and finance research. We draw on the management literature to apply Forbes and Milliken’s (1999) input–process–output framework to examine the mediation effect of board processes on earnings persistence. Accordingly, we provide a group-process-oriented view of board-monitoring effectiveness on financial reporting. The combination of the use of survey and archival data while maintaining the anonymity of survey respondents with its acknowledged limitations extends the methodology typically used in the financial-reporting literature and provides opportunities for extension of the research. Thus, we demonstrate the possible application of these techniques in accounting research. Finally, the Australian Securities Exchange (ASX) provides corporate-governance guidelines on board structure but does not provide adequate guidelines on the inner workings of the board (i.e. on board processes). Our findings emphasise that board-structure attributes influence effective monitoring through board processes. Thus, while structure is important, the inclusion of best-practice guidelines around processes is also essential.

The remainder of this article is structured as follows. Section 2 discusses the theoretical background; Section 3 develops the hypothesis; Section 4 outlines the research method; Section 5 presents the empirical results and Section 6 concludes the paper.

2. Theoretical background

Corporate collapses and financial scandals around the world have created the necessary conditions for developing and implementing corporate-governance mechanisms to constrain managerial opportunistic behaviour and to ensure the credibility of financial information. Effective board monitoring has thus become one of the most important mechanisms in constraining managerial opportunistic behaviour. In relation to the credibility of financial information, prior research has examined the effect of board-structure variables such as board size, board independence, chief executive officer (CEO) duality, financial expertise, and presence of female directors on various financial-reporting outcomes such as discretionary accruals, total accruals, earnings quality, earnings informativeness, earnings timeliness and earnings persistence (for reviews, see Brown et al., 2011; García-Meca and Sánchez-Ballesta, 2009). While many studies find relationships between board-structure variables and various measures of financial-reporting quality, some claim that board structure and the quality of financial disclosure are not significantly related (Matolcsy et al., 2012). This body of research has predominantly used agency theory to test the association between board structure and various measures of earnings quality and attributes of earnings (Eisenhardt, 1989). Accordingly, prior research has examined the effect of board structure on various firm-level outcomes such as performance and financial-reporting quality to understand the effectiveness of board monitoring. These studies were based on the assumption that board structure somehow captures the processes that influence performance. However, the parsimonious models used in such research provide equivocal conclusions. This narrow view of agency theory coupled with the overemphasis on input–output links creates the need for research to use alternative mechanisms to better understand board structure and firm-level outcomes (i.e. input–output links).

According to Cohen and Bailey (1997), inputs are the factors that can be manipulated to change processes and outcomes. Within this framework, inputs can be at the individual level (e.g. skills and attitudes) or at the group level (e.g. group size and structure) (McGrath, 1964). Processes are defined as the group behaviours that can be observed and are influenced by different input factors, which then effect the outcomes. These processes have been identified as internal activities (e.g. effort, time spent together, communication, encouragement among group members) (McGrath, 1964) and conflict and strategy discussions (Gladstein, 1984). Outputs or outcomes refer to the results of group processes and have been conceptualised in various ways, including as performance effectiveness, attitudes of group members and behavioural consequences (Cohen and Bailey, 1997).

Given that the board is a group at the apex of the decision-control structure (Fama and Jensen, 1983), the input–process–output framework is applicable. Forbes and Milliken (1999) argue that board effectiveness depends on board processes, which can be defined as social–psychological processes related to group participation and interaction, exchange of information and critical discussion.⁴ In general, the mediation effect assesses how or by what means an independent variable affects a dependent variable (Preacher and Hayes, 2008). As stated, we derive the theoretical premise for the mediation hypothesis from the group-effectiveness literature and the seminal work of Forbes and Milliken (1999). We combine this group inner-working behaviour with the agency-theory perspective of the board’s fiduciary duty in monitoring managerial financial-reporting behaviour. Thus, we examine the potential mediation effect of board processes on the association between board structure and earnings persistence.

Several board processes were identified in the literature based on group-process theories. However, only ‘effort norms’, ‘cognitive conflicts’ and ‘use of knowledge and skills’ have been identified consistently as having significant associations with board monitoring. The relevance of these three processes for the effectiveness of board monitoring has been advanced theoretically by Forbes and Milliken (1999) and supported empirically by some other studies (e.g. McNulty et al., 2013; Minichilli et al., 2009, 2012; Wan and Ong, 2005; Zona and Zattoni, 2007). Therefore, we focus on these three board processes because they are consistently found to be important in determining the effectiveness of board monitoring.

3. Hypothesis development

This section develops our hypothesis. First, we present a brief discussion on the association between board structure and earnings persistence (i.e. direct link between input and output). Second, we discuss the arguments for a mediation effect in relation to effort norms, cognitive conflicts and use of knowledge and skills (i.e. the link between input, process and output).

3.1. Board structure and earnings persistence

We focus on board independence, presence of financial experts, presence of female directors and size because a significant number of studies in the accounting and finance literature have examined the direct effect of these structural variables on various financial-reporting outcomes with inconsistent results (for reviews, see Brown et al., 2011; García-Meca and Sánchez-Ballesta, 2009).

From a theoretical perspective, agency theory suggests that independent board members are more likely to act in the interests of shareholders by overseeing managerial financial-reporting behaviour (Fama and Jensen, 1983). Financial experts possess the required technical knowledge and experience to identify concerns and issues about financial reports (Dhaliwal et al., 2010). Reputation and litigation risk also motivate independent directors and financial experts to monitor managerial financial-reporting behaviour. Thus, independent board members and financial experts are expected to increase monitoring of financial-reporting issues. Competing views exists in relation to the effect of board size on monitoring management. Larger boards have been found to suffer from greater lack of coordination, free-rider issues and lack of engagement than smaller boards. However, larger boards are conducive to director specialisation. From a resource-dependency perspective, larger boards increase effective monitoring because of the diverse experience of board members. Prior research provides mixed evidence on this association (e.g. Davidson et al., 2005; Kent et al., 2010; Mather and Ramsay, 2006; Peasnell et al., 2000). Female presence on the board has been found to increase the board’s ability to maintain an attitude of independence and reduces the extent of ‘groupthink’. Consequently, the presence of female directors on a board enhances the board’s ability to monitor financial reporting (Abbott et al., 2012; Adams and Ferreira, 2009). Thus, board structure is associated with earnings persistence, which is an attribute of earnings quality.

3.2. Board structure, effort norms and earnings persistence

In this section, we first define effort norms. We then present arguments for the link between variables of board structure and effort norms and then present arguments for the link between effort norms and earnings persistence.

Effort is an individual-level construct, which is a product of motivation, and represents the intensity of individuals’ task-performance behaviour (Forbes and Milliken, 1999). However, norms is a group-level construct, which represents standards of behaviour for groups (Ong and Wan, 2008). Therefore, ‘effort norms’ are a group-level construct that reflect the group’s shared beliefs about the level of the effort that each individual is expected to contribute towards group tasks (Wageman, 1995). We argue that it is necessary to set high effort norms at the board level to engender individual effort (e.g. time allocation), and that board members must be attentive to perform their roles effectively.

Independent directors and financial experts are expected to set high effort norms in relation to monitoring financial-reporting behaviour. Independent directors are outsiders to the firm and thus need to exert considerable time and effort to become familiar with the financial reports, accounting policies and accounting treatments of the firm to deploy their duties to the board and firm effectively. Consequently, independent directors are likely to set challenging standards of efforts within the boardroom. This sets a norm within the board, which motivates all members of the board to exert a higher level of effort. Financial experts are expected to show particular interest in the financial-reporting aspects of the board’s decision-making and are thus likely to engage in a higher level of effort and set higher expectations in relation to board financial-reporting activity. Furthermore, the participation of females in teams has been found to create synergies and improves the efforts of individual group members (Thomas and Ely, 1996). Given that female directors are more likely to make informed decisions, gender diversity increases information search (Hillman et al., 2007). Therefore, gender diverse boards are more likely to set higher effort norms in relation to obtaining more information before making decisions. Thus, boards as teams are able to set higher effort norms when there are female members on the board. However, board size and effort norms may have a negative association because lack of participation (Judge and Zeithaml, 1992) and social loafing (Dalton et al., 1999) can occur in larger teams. Thus, we argue that board independence, financial experts and female board members are likely to increase effort norms while board size has a negative association with effort norms.

However, most independent directors, financial experts and female directors encounter competing demands for their time and expertise. Thus, devotion of time and effort by directors considerably differ in different boards. Some boards simply go through agendas and documents without engaging in issues faced by the firm, while other boards require members to perform diligent research on board issues and expect members to engage actively in board activities (Forbes and Milliken, 1999). A high level of board expectations in relation to effort norms (which consists of attentive listening, applying intuitive and moral judgement and intense interest in financial reporting) increases monitoring effectiveness through directors’ well-informed judgements. Thus, well-prepared and attentive board members are more likely to ask critical questions from actors such as management, the audit committee and external auditors about any unusual reporting practices or similar issues. Accordingly, it is argued that boards that set higher standards of effort norms from their members promote behaviours that elevate efforts that result in effective monitoring of managerial financial-reporting behaviour.

Thus, a high level of effort norms in the boardroom plays an important role in exerting effective monitoring of financial reporting, and therefore earnings persistence. We argue that the presence of independent directors, financial experts and female directors on the board, and an appropriate board size affect the processes of effort norms, which in turn affect monitoring financial reporting, and thus earnings persistence.

3.3. Board structure, cognitive conflicts and earnings persistence

In this section, we first define cognitive conflicts. We then present arguments for the link between variables of board structure and cognitive conflicts and then present arguments for the link between cognitive conflicts and earnings persistence.

Cognitive conflicts are defined as ‘disagreements about the content of the task being performed, including differences in viewpoints, ideas and opinions’ (Jehn, 1995: 258). These conflicts foster learning and the development of insight and understanding and increase the ability to solve complex problems (De Dreu, 2006). According to the literature on group effectiveness, divergent viewpoints and critical discussions have positive effects on team effectiveness (Amason, 1996; Amason and Sapienza, 1997).

Independent directors and financial experts are expected to provoke debate and discussion about monitoring financial reporting in the boardroom. From an agency-theory perspective, independent directors show objectivity in their monitoring function (Fama and Jensen, 1983) and be willing to consider diverse views before making decisions, thereby initiating constructive conflicts (Finkelstein and Mooney, 2003). For example, when the independent directors ask critical questions from management and demonstrate their interest in and vigilance of monitoring matters, the inside directors are coerced into participating actively in board meetings (Wan and Ong, 2005). Financial experts possess necessary skills and expertise to review financial reports and challenge (through questioning and requiring answers) accounting treatments and policies used by the managers. Such questioning and answering provokes higher cognitive conflicts within the board. Furthermore, boards that are gender diverse adhere to more informed deliberations and discuss tougher issues that are often considered challenging by all-male boards (Huse and Solberg, 2006). Thus, female directors are more likely to provoke cognitive conflicts in the boardroom. However, female leadership attributes such as accepting others’ positions, supporting and soothing others and speaking tentatively (Eagly and Carli, 2003) may hinder critical discussion, and thus have a negative effect on the level of cognitive conflicts. Nielsen and Huse (2010) found that female participation on the board has a negative effect on cognitive conflict. However, board size is likely to have a nonlinear relationship with cognitive conflicts. Larger groups have greater potential for diverse cognitive capabilities. For example, Amason (1996) suggests that an increase in team size intensifies cognitive capability, which is considered an antecedent for effective group decisions. Thus, larger teams generally have greater cognitive resources than smaller teams. However, some evidence suggests that larger groups may have a negative effect on group effectiveness (Lipton and Lorsch, 1992). For example, Dalton et al. (1999) found that larger boards would tend to be more diverse and be more contentious than smaller boards, and therefore develop factions and coalitions that may lead to group conflict, and therefore have a negative effect on group effectiveness (Goodstein et al., 1994). Thus, larger boards may not harness the benefits of their members’ divergence. We argue that board independence, presence of financial experts, female directors and board size are associated with cognitive conflicts.

Studies that examine direct links between structure and outcomes implicitly assume that the presence of independent directors, financial experts and female directors enhances critical discussion in the boardroom. However, there are considerable differences across different boards (with similar structures) in relation to the degree of critical discussion (Forbes and Milliken, 1999). The process of cognitive conflicts in the boardroom encourages managers to explain, justify and possibly modify their financial-reporting behaviour. Given that cognitive conflicts result in consideration of alternatives and evaluation of alternatives (Forbes and Milliken, 1999), managers are more likely to choose the best financial-reporting treatments and policies to meet the interests of shareholders. In relation to reporting, cognitive conflicts are useful in assessing the viability of alternative accounting treatments and addressing concerns about estimates, judgements and issues around managerial financial-reporting behaviour. For example, critical discussion in the boardroom may discourage managers from engaging in inappropriate financial-reporting behaviour that negatively affects earnings persistence.

Thus, critical discussion in the boardroom plays an important role in exerting effective monitoring of financial reporting, thus earnings persistence. We argue that the presence of independent directors, financial experts and female directors and an appropriate board size affect the processes of cognitive conflicts, which in turn affect monitoring financial reporting, and thus earnings persistence.

3.4. Board structure, use of knowledge and skills and earnings persistence

In this section, we first explain use of knowledge and skills as a board process. We then present arguments for the link between variables of board structure and use of knowledge and skills, and then present arguments for the link between use of knowledge and skills and earnings persistence.

Board members’ knowledge and skills are an important factor in effective board functioning (Forbes and Milliken, 1999; Jackson, 1992). Cohen and Bailey (1997) suggest that the use of knowledge and skills is associated with the behavioural dimension of social integration, and thereby affects a group’s ability to cooperate. A board must be aware of each member’s knowledge and expertise and collaborate and integrate this knowledge to achieve better performance (McNulty et al., 2013). Wageman (1995) provides empirical evidence to suggest that use of knowledge and skills affects group effectiveness.

It is expected that financial experts enhance the use of knowledge and skills, particularly in financial-reporting matters. For example, board members who have accounting qualifications possess technical knowledge that must be shared to ensure that board members who do not have accounting backgrounds understand managerial behaviour around financial reporting. Independent directors who possess accounting qualifications (i.e. Certified Practising Accountant (CPA) or Chartered Accountant (CA)) and financial-accounting professional experience (e.g. chief financial officer, treasurer or auditor) in other firms have experience with accounting practices and trends in their industry. These independent directors bring outside expertise to the board. This is utilised to educate other board members, and the shared knowledge may then be used to scrutinise financial statements. Shared knowledge and skills, well-informed judgements and careful attention to detail enable board members to ask critical questions of a firm’s managers. Based on resource-dependence theory, it can be asserted that directors bring a wide variety of knowledge, skills and experience to the board (Pfeffer and Salancik, 1978). This creates an opportunity to integrate a wide variety of knowledge and skills into the boardroom when the board is large. In relation to gender, Adams and Ferreira (2009) find that members’ attendance at board meetings increases when there is higher gender diversity in the board. Thus, presence of female directors increases member participation in boardroom activities, which in turn leads to an increase in the use of knowledge and skills. Thus, we argue that board independence, presence of financial experts, presence of female board members and board size are associated with the use of knowledge and skills.

Thus, use of knowledge and skills is an integral process through which well-structured boards can deliver effective monitoring mechanisms on financial reporting and consequently, earnings persistence. Boards need to integrate their knowledge of the firm’s internal affairs with their expertise, must elicit and respect each other’s expertise and must build on each other’s contributions to exert effective monitoring mechanisms (Forbes and Milliken, 1999). Thus, we argue that the process of engagement, identification, integration and coordination of relevant expert knowledge and skills in the boardroom facilitates the positive effect of a well-structured board on monitoring financial reporting. Thus, the board process of use of knowledge and skills plays an important role in effective monitoring of financial reporting, thus earnings persistence. We argue that the presence of independent directors, financial experts and female directors as well as an appropriate board size positively affects the board process of use of knowledge and skills, which in turn increases monitoring financial reporting, and thus earnings persistence.

While it is expected that board members will be judicious when monitoring managerial financial-reporting behaviour, the mixed empirical evidence on direct links between board-structure variables and financial-reporting outcomes suggests that this expectation may not be achieved. Board processes play an important intervening role in exerting effective monitoring of financial reporting. We believe that board structure (i.e. board independence, presence of financial experts and female directors, and board size) affects monitoring financial reporting, thus earnings persistence via board processes (i.e. effort norms, cognitive conflicts and use of knowledge and skills). More specifically, we argue that board processes mediate the relationship between board structure and monitoring financial-reporting behaviour of managers, and thus earnings persistence.

Accordingly, we propose the following hypothesis:

Hypothesis: The association between board structure and earnings persistence is mediated by board processes.

4. Research method

To test our hypothesis, we require variables of board structure, board processes and earnings persistence. We used a questionnaire to collect data on board processes, and archival sources to collect data on board structure and earnings persistence.

4.1. Measuring board processes: questionnaire administration

We adapted the questionnaire from prior literature on board processes (McNulty et al., 2013; Minichilli et al., 2012; Van Ees et al., 2008; Wan and Ong, 2005; Zona and Zattoni, 2007). Board processes (i.e. latent constructs) were measured using a multi-item 7-point Likert-type scale in a questionnaire. As presented in Table 1, effort norms (Cronbach’s α = 0.915) were measured using a three-item scale (en1, en2, en3) and cognitive conflicts were measured using five-item scale (cc1, cc2, cc3, cc4, cc5). The criteria extracted two factors to measure cognitive conflicts. Based on the qualitative nature of the questions, we identified those two factors as existence of cognitive conflicts (with an α value of 0.748) and dealing with cognitive conflicts (with an α value of 0.791). Use of knowledge and skills (Cronbach’s α = 0.868) was measured using a four-item scale (uks1, uks2, uks3, uks4). Before administering the survey, we conducted several interviews with directors to ascertain the applicability of the survey in the Australian context. A draft of the revised survey was sent to practitioners and academics and used their feedback to further enhance the measurements of the instrument.

Table 1.

Measurements of latent constructs, means and standard deviations.

Variable/latent construct	Items used	Mean	SD
Effort norms	The board members devote all the time needed to complete the task assigned (en1)	5.65	1.36
	The board members participate actively during meetings (en2)	5.98	1.24
	The board members carefully scrutinise the information provided by the firm prior to meetings (en3)	5.59	1.29
Cognitive conflicts	There is constructive challenge and debate between the members at board meetings (cc1)	5.65	1.35
	The Board is able to reach collectively shared decisions (cc2)	5.84	1.12
	Differences of opinion arise at board level regarding the key decisions to be taken (cc3)	4.12	1.32
	Differences of opinion arise at board level regarding the roles and responsibilities of the board (cc4)	2.88	1.50
	Differences of opinion arise at board level regarding the company results and performance (cc5)	3.12	1.49
Use of knowledge and skills	The nonexecutive directors of this Board use their skills and knowledge to contribute to board tasks (uks1)	5.90	1.21
	The tasks of this Board are delegated in a way that ensures the best fit between assigned tasks and each director’s skills and knowledge (uks2)	5.47	1.21
	When an issue is discussed at board meetings, the most knowledgeable members on that subject generally have the most influence (uks3)	5.22	1.38
	The chair of the Board runs the meetings to enable all members to join in the discussions (uks4)	5.96	1.39
Board structure	Board size: number of directors in the board (takes a value from 1 to 4 based on respective industry-sector quartile, see note 6)	2.36	1.10
	Independent directors on board: number of independent directors in the board, ordered categorical variable (takes a value from 1 to 4 based on respective industry-sector quartile, see note 6 footnote)	2.49	1.10
	Financial experts on board: financial experts in the board, measured as a dichotomous variable, 1, if at least one member in the board has CA or CPA, 0 otherwise	0.65	0.48
	Existence of female directors: female representation on the board, measured as a dichotomous variable, 1, if the board has at least one female director, 0, otherwise	0.30	0.46
Monitoring financial reporting	The board members keep themselves informed about the financial position of the firm	5.30	1.60
	The Board oversees audit committee’s appraisal of the CFO’s performance	5.27	1.69
	Constructive challenge and debate takes place at meetings when the audit committee presents its findings to the Board	5.42	1.61
	The Board oversees the activities and operations of the audit committee to ensure its effectiveness	5.01	1.65

CA: Chartered Accountant; CPA: Certified Practising Accountant; CFO: chief financial officer.

Table 1 presents measurements of all latent theoretical constructs and board-structure variables, their, means and standard deviations.

Our descriptive statistics and hypothesis testing are based on financial year 2011/2012. Thus, respondents were instructed to answer the questions on board processes based on the financial year 2011/2012 (see Appendix 1 for a summary of instructions given to the respondents). We randomly selected a nonexecutive director from each firm as the respondent and mailed the survey. A nonexecutive director is suitable to answer questions on board processes in an independent manner. Due to difficulties in gaining access to process data in studies on board of directors (e.g. Daily et al., 2003; Pettigrew, 1992), primary survey data are often based on a single respondent (e.g. Minichilli et al., 2012; Pearce and Zahra, 1992). Board members are part of a group that meets periodically, which suggests potential downsides of having multiple respondents. Kumar et al. (1993) emphasise that having multiple respondents sometimes can increase the risk of constructing averaged measures that reflect divergence across respondents rather than representing the constructs being investigated. The selected respondents’ contact details were obtained from the FinAnalysis database or the company website. A cover letter was prepared inviting the respondents to participate in the survey. It was clearly indicated in the cover letter why and how the respondents were selected to participate in the survey. The survey package, which consisted of the cover letter, project-information statement, survey and reply-paid envelop, was posted. A reminder letter with the same documents sent in the first round was sent to all respondents 1 month later. The letter also stated to ignore this if responded in the first round. Given that assurance of anonymity in surveys substantially increases response rates (Dillman, 2011) and reduces social-desirability bias (Podsakoff et al., 2003), we conducted the surveys anonymously.

However, in doing so, we could not match the returned surveys with the earnings persistence and board-structure variables collected from archival sources. Thus, we developed an ex-ante mechanism to overcome this problem. To ensure the anonymity of the respondents as required by the university ethics committee, we followed the procedure outlined in Section 4.2 to code the board-structure variables and earnings persistence figures and pre-populate them on the respective survey instrument for each sample firm. Our initial sample consisted of 1942 firms listed in the ASX in 2012. Following the practice in capital-market research, 262 firms in the financial and utilities sectors were excluded due to their unique working-capital structures, which may affect cash flow, and consequently earnings persistence. Furthermore, these firms have an extra layer of governance imposed through regulation (Davidson et al., 2005). Another 487 firms were excluded due to missing data on corporate governance, which gave us a final total of 1193 sample firms.

4.2. Measuring board-structure variables

Board structure was measured using the variables board size, board independence, presence of financial experts and presence of female directors. Board size was measured as the number of board members and board independence was measured as the proportion of independent directors on the board. As stated, to ensure the anonymity of the respondent, as required by the university ethics committee, we coded the board-structure variables and earnings persistence figures and pre-populated those codes on the respective survey instruments for each sample firm.⁵ In coding these variables, we converted the variables of raw board size and independence into ordinal variables, from 1 to 4, 1 being the lowest board size (board independence) and 4 being the highest board size (board independence). When determining the 1 to 4 categories, we used industry-sector quartiles to better reflect the board size and board independence relative to peer firms. The quartiles were based on the Global Industry Classification Standard (GICS) of industry sectors.⁶ Accordingly, the lowest 25% of board size (board independence) of each industry sector was assigned number 1; 25%–50% of board size (board independence) was assigned number 2; 50%–75% of board size (board independence) was assigned number 3 and the highest 25% of board size (board independence) was assigned number 4. Thus, sample firms were assigned to their respective industry-sector quartile for the variables of board size and board independence. The first quartile was assigned ‘1’, the second quartile was assigned ‘2’, the third quartile was assigned ‘3’ and the fourth quartile was assigned ‘4’. Presence of financial experts was measured as a dichotomous variable, taking a value of 1 if the board had at least one financial expert. Similarly, presence of female directors was also measured as a binary variable, which took the value of 1 if the board had at least one female director and 0 otherwise. All variables are defined in Table 1.

4.3. Measuring earnings persistence

In the input–process–output model, the output variable we examined was earnings persistence. We examined the mediation effect of board processes on the association between board structure and monitoring financial reporting. Earnings persistence is facilitated by effective financial-reporting monitoring and is an attribute of earnings quality. Prior literature asserts that more persistent earnings will yield better inputs to equity-valuation models; thus, a more persistent earnings number is of higher quality than a less persistent earnings number (Dechow et al., 2010). Earnings persistence has been used as a proxy for earnings quality in previous accounting and finance literature (Oei et al., 2008). Based on the literature (Dechow et al., 2010; Richardson et al., 2005; Sloan, 1996), the requirements of this study, and the notion of parsimony, we used a first-order autoregressive AR(1) model with ordinary least squares (OLS) for estimation as the operational measure of persistence. We used earnings from 2004 to 2011 (time series) to measure earnings persistence for the financial year 2011/2012.

The following formula ascertains the level of earnings persistence

E a r n i n g s_{t + 1} = α + β E a r n i n g s_{t} + e_{t} \dots

(1)

In equation (1), based on Sloan (1996), we define earnings as the operating income after depreciation, scaled by total assets excluding nonrecurring items (e.g. extraordinary items, discontinued operations, special items and nonoperating income). As explained earlier, β coefficients were calculated for all of the sample firms using AR(1). A higher β implies a more persistent earnings stream and vice versa. As stated, to ensure the anonymity of the respondents, earnings persistence figures were also coded and pre-populated on the respective surveys for each sample firm to maintain anonymity. Prior to doing that, we removed outlier firms from the sample, that is, the data items above the ninety-ninth percentile and below the first percentile of each industry sector (57 firms). We also removed data items that were approximately +1% or −1% of the industry-sector median to mitigate the lack of differentiation between the closest and farthest data points relative to the cut-point between the groups, mitigating bias that would have been created due to misclassification (53 firms). Thus, the final mailing-list population consisted of 1083 firms (1193 – 53 – 57 = 1083). We then dichotomised the absolute (β) earnings persistence figures. To achieve this, we categorised the firms that had β coefficients above the industry-sector median as the high earnings quality group and the firms that had a β coefficient below the industry-sector median as the low earnings quality group. We assigned ‘1’ to the firms in the high earnings quality group and ‘0’ to the firms in the low earnings quality group.

4.4. Survey responses and nonresponse bias

We received 184 usable responses, a 17% response rate, of which of 89 high earnings quality firms and 95 were low earnings quality firms. Response rates in surveys where the respondents were directors or executives have been relatively low in prior literature. For example, Minichilli et al. (2009) who surveyed CEOs in Italy had a response rate of 15%. In addition, Brooks et al. (2009) who surveyed independent directors in Australia had a response rate of 12.3%. Thus, a 17% response rate is comparable with or better than similar surveys. Thus, a low response rate for this type of survey is usual.

We initiated some measures a priori to reduce response bias. For example, multiple response formats were used and items measuring each construct were dispersed throughout the survey (DeVellis, 2003). However, nonresponse bias may still have contaminated the survey results. Therefore, posteriori, we compared the mailing-list population and the respondent sample in relation to industry-sector representation, firm and board characteristics. In untabulated results, we found that there was no significant difference between the population firms and the respondent firms. Thus, we concluded that the respondents provided a fair representation of the population of mailing-list firms. Furthermore, we compared responses to all constructs of early (i.e. respondents who answered after the initial mail-out) and late responses (i.e. respondents who answered after the second mail-out). The χ² and two-sample t-tests revealed no systematic differences (p > 0.05). Therefore, nonresponse bias is unlikely to have affected the results.

4.5. Addressing other biases

There is a potential for creating bias in results arising from the self-report data in the questionnaire, known as ‘common method bias’ (CMB). We took both a priori and a posteriori measures to minimise the effect of CMB. First, for the a priori measures, we used two different sources to collect data for variables. The independent variable (i.e. board structure) and the dependent variable (i.e. earnings persistence) were collected from archival sources, while the mediating variables of board processes and monitoring financial reporting were collected from the questionnaire. Podsakoff et al. (2003) suggest that the use of multiple sources in collecting variables minimises CMB. Second, we took stringent measures to reduce the potential effects of CMB by safeguarding the respondents’ anonymity. Third, a posteriori, we conducted Harman’s one-factor test on the conceptually crucial variables in our theoretical model. A multiple factor solution rather than a single factor emerged, indicating that CMB was unlikely to have contaminated the empirical findings (Podsakoff et al., 2003). Finally, potential response bias was investigated by manually examining each individually completed questionnaire and inspecting the variance across all items. We did not identify any patterns in the responses that raised serious concerns. Response bias was also examined based on other descriptive statistics. For example, we determined the industry-sector distribution across the respondent sample and found that the distribution was somewhat similar to the population.

4.6. Mediation-hypothesis testing using PLS-SEM

Our hypothesis development is mainly informed by agency theory, which suggests that effective monitoring mechanisms constrain managerial opportunistic financial-reporting behaviour, which results in higher earnings persistence. However, earnings persistence can also result from effective board engagement in strategic decision-making, which may lead to an increase in firm performance, and in turn lead to higher earnings persistence. To rule out this alternative explanation, we utilised an extended version of the input–process–output framework by including monitoring financial reporting, which is a board-role latent construct, as an additional mediating variable into the model. By empirically testing this model, we expect to rule out the alternative explanation that earnings persistence is a result of the board’s active engagement in processes of strategic decision-making.⁷ The extended input–process–output model thus examines three links: board structure leads to board processes (Link 1), which leads to better monitoring of financial reporting (Link 2), which then leads to higher earnings persistence (Link 3). This sequel mediation is presented in Figure 1. That is, our empirical model demonstrates that board processes and monitoring financial reporting sequentially mediate the association between board structure and earnings persistence. By demonstrating this, we partially, but not entirely, intend to rule out an alternative plausible explanation that earnings persistence is a result of the board’s active engagement in strategic decision-making. We measured monitoring financial reporting as a multi-item latent construct in the questionnaire along with the board-process variables (see Table 1 for the measurements of monitoring financial reporting).

Figure 1.

Mediation effect of board processes using PLS-SEM, the conceptual model.

We empirically test this model using PLS-SEM. SEM is able to simultaneously assess the quality of measurement and examine causal relationships among constructs by considering various forms of measurement errors (Kline, 2011). Covariance-based structural equation modelling (CB-SEM) uses a maximum likelihood estimation procedure while PLS-SEM uses a regression-based OLS estimation method (Hair, 2011; Hair et al., 2017b). Use of PLS-SEM in social-science research has gained momentum recently (Hair et al., 2017a). While several studies have utilised PLS-SEM relative to CB-SEM in accounting research (Lee et al., 2011), the full potential of the PLS-SEM technique has not yet been utilised in accounting studies (Nitzl et al., 2016).⁸

We use PLS-SEM to test our extended input–process–output model with sequel mediation for several reasons. First, our sample size is small, with only 184 responses. CB-SEM is a large-sample technique (Kline, 2011), but PLS-SEM is suitable for smaller sample sizes. Second, PLS-SEM can be applied when the distributional properties of constructs are nonnormal (Hair et al., 2017b). Third, we use a complex model with sequel mediation, which can be reliably estimated using PLS-SEM. Our model becomes further complicated by having multiple-indicator variables in the measurement model (i.e. outer model) and several path coefficients in the structural model (i.e. inner model). Our outcome variable is a dichotomous variable that presents difficulties in applying CB-SEM.⁹ PLS-SEM has been recognised as a useful tool for estimating complex mediations (Nitzl et al., 2016). Fourth, PLS-SEM is capable of accommodating single indicator variables along with multiple-indicator variables. Our outcome variable, earnings persistence, is a single indicator variable, while our latent constructs are measured using several indicators. We evaluate the statistical significance of the resulting parameter estimates of the structural model using the bootstrapping method (1000 samples with replacement; all samples have the same size as the original sample). We present the empirical estimates of PLS-SEM in testing the extended input–process–output model in Section 5.1.

The algorithm used to estimate a PLS-SEM model consists of three sequential stages of estimation (see Lohmöller, 1989). In the first stage, latent-variable scores are estimated for each case. Using these scores, in the second stage, measurement-model parameters (weights/loadings) are estimated. In the same manner, in the third stage, structural-model parameters (path coefficients) are finally estimated. The first stage is what makes PLS-SEM a novel method, in that the second and third stages conduct a series of regression analysis using the OLS method (for further detail about these three stages, see Lee et al., 2011). The PLS algorithm is designed to obtain the optimum weight estimates for each component of indicators corresponding to each theoretical construct. PLS-SEM yields a component or composite variable representation of the theoretical construct and aims to maximise the variance of the dependent variables that is explained by the independent variables (Lee et al., 2011). PLS-SEM path models are estimated through an integration of techniques from OLS regression, principal component analysis and path analysis. However, PLS-SEM allows for estimating a model with a binary-dependent variable using a logistic-regression model. We used the PLS-SEM package for STATA v.15 to estimate the model.

4.7. Further analysis: mediation-hypothesis testing using binary mediation

In the main analysis, we tested the mediation hypothesis using PLS-SEM. In this further analysis, we tested our hypothesis using a binary-mediation model. We used binary mediation because our outcome variable is dichotomous, that is, high earnings persistence and low earnings persistence. We used a two-stage approach to test our mediation hypothesis. Following McNulty et al. (2013), in the first stage, the questionnaire responses were coded and transformed into meaningful factor scores on board processes using principal component analysis. In the second stage, these factor scores were used in the binary mediation to ascertain the mediation effect. First, to maintain consistency with the main analysis (i.e. PLS-SEM model), we used a mediating variable (i.e. composite board processes) to ascertain the mediation effect of board processes. Second, we examined the mediation effect of the separate board processes (i.e. effort norms, cognitive conflicts and use of knowledge and skills) on the association between board structure and earnings persistence. We used principal component analysis for dimension reduction and extracted one factor to explain the variances of the board-structure variables in this model for two reasons. First, the binary-mediation model does not allow us to incorporate multiple independent variables in the model; thus, we could not use separate board-structure variables in this model. Second, Larcker et al. (2007) claim that the mixed findings in the literature on the associations between various corporate-governance attributes and earnings quality may be a result of poor reliability and construct validity in the governance measures used. They argue that principal component analysis can be used to find factors that retain most of the variance in the original board-structure indicators, and so characterise the dimensionality of the individual governance indicators.

We used the bootstrapping method to conduct this binary-mediation analysis (MacKinnon et al., 2002; Preacher and Hayes, 2004, 2008). This method has a lower Type II error rate and greater statistical power than the traditionally used causal-steps approaches. Furthermore, the bootstrapping method does not assume normality of the distribution of the indirect effects and thus provides stronger protection against Type II errors (Field, 2013; Tabachnick and Fidell, 2007). The bootstrapping method was performed following the recommendations by Preacher and Hayes (2008), with k = 5000 resamples and 95% confidence interval (CI) used to evaluate indirect effects. We conducted binary bootstrapped mediation for combined board processes, effort norms, cognitive conflicts and use of knowledge and skills separately to test our mediation hypothesis. Although not common, several studies in accounting research have used mediation with bootstrapping (e.g. De Baerdemaeker and Bruggeman, 2015; Parker et al., 2011). We present the empirical results of the binary-mediation testing of the input–process–output model in Section 5.2.

5. Empirical results

5.1. Empirical results of mediation-hypothesis testing using PLS-SEM

The means and standard deviations of measurements of the latent constructs and board-structure variables are presented in Table 1. Table 2 presents the correlation matrix; the Pearson coefficients estimates are presented below the diagonal, and the Spearman estimates are presented above the diagonal.

Table 2.

Correlations.

		1	2	3	4	5	6	7	8	9
1	Earnings persistence	1.000	0.388*	0.080	0.386*	0.415*	0.134	0.069	0.135	0.124
2	Effort norms	0.465*	1.000	−0.027	0.908*	0.969*	0.110	0.114	0.270*	0.081
3	Existence of cognitive conflicts	0.110	0.033	1.000	−0.004	−0.010	0.125	0.144	0.139	−0.019
4	Dealing with cognitive conflicts	0.457*	0.965*	0.044	1.000	0.884*	0.135	0.135	0.298*	0.102
5	Use of knowledge and skills	0.477*	0.988*	0.050	0.959*	1.000	0.019	0.065	0.132	0.050
6	Board size	0.134	0.093	0.123	0.089	0.032	1.000	0.326*	0.619*	0.117
7	Female directors	0.069	0.049	0.173*	0.058	0.020	0.321*	1.000	0.354*	0.192*
8	Board independence	0.134	0.221*	0.135	0.200*	0.123	0.615*	0.352*	1.000	0.135
9	Financial experts	0.124	0.088	−0.006	0.089	0.073	0.110	0.192*	0.133	1.000

Table 2 shows Pearson correlation coefficient below the diagonal and Spearman correlation coefficient above the diagonal, significance at *p < 0.05. The correlations of effort norms, existence of cognitive conflicts, dealing with cognitive conflicts and use of knowledge and skills are based on the factor scores generated ex-post principal component analysis.

As explained in Section 4.6, we used PLS-SEM to test our mediation hypothesis. We present the empirical estimates of measurement and structural models derived from PLS-SEM in the following sections.

5.1.1. Measurement model

The extended input–process–output model has a sequel mediation with several indicator variables in the measurement model and path coefficients in the structural model. The measurement model includes constructs used to measure board processes and constructs used to measure monitoring financial reporting. With the intention of having a parsimonious model and increasing model efficiency, we combined effort norms (en1 to en3), cognitive conflicts (cc1 to cc5) and use of knowledge and skills (uks1 to uks4) into one board-process factor variable, a composite board processes. Several iterations of the measurement model were executed, and we removed three items that had higher cross-loadings or weaker factor loadings to maintain validity and reliability, leaving nine indicator variables to explain the variance of the meditating variable composite board processes. The second mediating variable, monitoring financial reporting, was measured using four indicators (mfr1 to mfr4), as shown in Table 3, Panel A.

Table 3.

Mediation effect of board processes with PLS-SEM estimation.

Panel A: latent constructs and measures based on measurement model
Construct	Measures	Estimate	Construct reliability	AVE
Composite board processes (CBP)	The board members devote all the time needed to complete the task assigned	0.881	0.947	0.704
	The board members participate actively during meetings	0.902
	The board members carefully scrutinise the information provided by the firm prior to meetings	0.868
	There is constructive challenge and debate between the members at board meetings	0.823
	The Board is able to reach collectively shared decisions	0.842
	The nonexecutive directors of this Board use their skills and knowledge to contribute to board tasks	0.850
	The tasks of this Board are delegated in a way that ensures the best fit between assigned tasks and each director’s skills and knowledge	0.781
	When an issue is discussed at board meetings, the most knowledgeable members on that subject generally have the most influence	0.753
	The chair of the Board runs the meetings to enable all members to join in the discussions	0.843
Monitoring financial reporting (MFR)	The board members keep themselves informed about the financial position of the firm	0.850	0.899	0.768
	The Board oversees audit committee’s appraisal of the CFO’s performance	0.900
	Constructive challenge and debate take place at meetings when the audit committee presents its findings to the Board	0.909
	The Board oversees the activities and operations of the audit committee to ensure its effectiveness	0.845
Panel B: results of the structural model
Structural model				Estimates
Path coefficients: Link 1
Path between board size and CBP				−0.077
Path between board independence and CBP				0.225**
Path between board female directors and CBP				0.029
Path between board financial experts and CBP				0.074
Path coefficient: Link 2
Path between CBP and MFR				0.718***
Path coefficient: Link 3
Path between MFR and earnings persistence				0.603**
Indirect effects
Board size → CBP → MFR				−0.055
Board independence → CBP → MFR				0.161*
Board female directors → CBP → MFR				0.021
Board financial experts → CBP → MFR				0.053
CBP → MFR → earnings persistence				0.100***

CFO: chief financial officer; AVE: average variance extracted.

Panel A of Table 3 shows composite board processes (CBP) construct measurements and monitoring of financial-reporting (MFR) construct measurements. Estimates show the loadings of each construct on the extracted factor and construct reliability is also depicted for each factor. Average variance extracted (AVE) is also shown.

Panel B of Table 3 shows path coefficients of causal links 1, 2 and 3. CBP is composite board processes and MFR is monitoring financial reporting. See Table 3, Panel A for the theoretical construct measurements, estimates and reliability measures of CBP and MFR variables. The bottom panel shows the indirect effect size. Significance levels ***p < 0.000, **p < 0.05, *p < 0.10. All other variables are defined in Table 1.

The first stage of the PLS-SEM analysis involves assessing the psychometric properties of composite board processes and monitoring financial reporting, the latent constructs. These two variables were measured using multiple indicators and evaluated for reliability and validity. The psychometric properties of the multi-item constructs were assessed simultaneously in the measurement model. Discussing item reliability, Hair et al. (2017b) suggest that significant item loadings greater than 0.50 are sufficient to establish reliability. As shown in Table 3, Panel A, the factor loadings of all indicators of composite board processes and monitoring financial-reporting range from 0.75 to 0.91 (p < 0.05) suggesting good reliability with all loadings being significant (p < 0.05) and above 0.50. Then, convergent validity was assessed by examining the construct reliability (CR) and average variance extracted (AVE). As shown in Table 3, Panel A, the CR score for composite board processes is 0.95 and for monitoring financial reporting is 0.90. The CR index of all constructs is good because all scores equal or exceed the threshold value of 0.70 (Hair et al., 2017b). Moreover, the AVE scores are higher than the threshold value of 0.50, which indicates that at least 50% of the indicator variance is accounted for by construct variance rather than by noise (Fornell and Larcker, 1981). The discriminant validity of the latent constructs composite board processes and monitoring financial reporting was assessed by comparing the square roots of the AVEs with the correlations between the constructs. We found that the square roots of the AVEs are larger than the reported correlations of latent constructs, which provides evidence of discriminant validity (Fornell and Larcker, 1981). In addition, the results of the heterotrait-monotrait (HTMT) analysis also substantiate discriminant validity of the latent constructs. Overall, the results from the measurement model indicate adequate construct validity for all constructs in the model.

5.1.2. Structural-model results

We used the structural model to assess the validity of causal structures among variables. The input variable(s) in the model are board size, board independence, presence of female directors and presence of financial experts (see Table 1 for definitions and Figure 1 for graphical representation). The firm-level output variable in the model is earnings persistence, measured as explained in Section 4. The mediating variables, composite board processes and monitoring financial reporting, were measured as discussed. We examined each set of predictors in the structural model for possible collinearity without bootstrap. The empirical findings are not affected by collinearity because the variance inflation factor (VIF) scores for all predictors are well below the threshold of 5. We used the bootstrapping method with 1000 samples with replacement to generate standard errors and t-values (Hair et al., 2017b). We estimated path relationships among the latent variables in the model through the sign and magnitude of path coefficients (see Table 3 and Figure 1 for results). As shown in Table 3, Panel B (and Figure 1), board independence is positively and significantly associated with composite board processes (0.225, p < 0.05). However, board size, presence of female directors and presence of financial experts are not significantly associated with composite board processes. The variable composite board processes is positively and significantly associated with monitoring financial reporting (0.718, p < 0.000). The relationship between monitoring financial reporting and earnings persistence is also positive and significant (0.603, p < 0.05).

We used a nonparametric bootstrapping method to evaluate the significance of the mediating effect (Hair et al., 2017b; Preacher and Hayes, 2008). The requirement that is commonly accepted as an indicator for mediation is that the indirect effect must be significant (Ali and Park, 2016). The significant indirect effect absorbs some of the direct effect. The indirect effects of paths are presented in Table 3, Panel B. The mediation effect of composite board processes on the association between board independence and monitoring financial reporting is significant (Sobel test statistic: 2.31, p < 0.05). The bias-corrected bootstrapping (1000 reps) also suggests that the indirect effect is significant (IE = 0.161, standard error (SE) = 0.058, BCCI = 0.024–0.298). The mediation effect of monitoring financial reporting on the association between composite board processes and earnings persistence is also statistically significant (Sobel test statistic: 3.7, p < 0.000). The indirect effect is significant (IE = 0.100, SE = 0.030, BCCI = 0.046–0.152). Our results provide support for the conditions necessary for mediation only for board independence in both parts of the model. In the first part of the model (i.e. the relationship among board independence, composite board processes and monitoring financial reporting), our results show that composite board processes significantly mediate the relationship between board independence and monitoring financial reporting. For the second part of our model (i.e. the relationship among composite board processes, monitoring financial reporting and earnings persistence), our results show that monitoring financial reporting mediates the relationship between composite board processes and earnings persistence. Accordingly, using composite board processes and monitoring financial reporting as the mediators, the PLS-SEM estimates support the mediation hypothesis for board independence. While Link 2 (i.e. association between composite board processes and monitoring financial reporting) and Link 3 (i.e. association between monitoring financial reporting and earnings persistence) are significant, the relationships among the other board-structure variables and composite board processes (i.e. Link 1) are not statistically significant.

5.2. Empirical results of mediation-hypothesis testing using binary-mediation model

In this section, we present the results of the hypothesis testing based on the binary-mediation model. In the binary-mediation model, board structure is the input variable, board processes are the mediation variable and earnings persistence is the outcome variable. Following McNulty et al. (2013), in the first stage, the questionnaire responses were coded and transformed into meaningful scores on board processes based on principal component analysis. In untabulated results, we find that factors extracted to explain variances of all board processes were loaded significantly high with respect to all the measurements. The predicted factor scores for processes were then used in the binary-mediation model.

To be consistent with our main model, which was tested using PLS-SEM, we first tested the mediation of composite board processes. Figure 2, Panel A shows the mediation of composite board processes. We found that board structure is positively associated with earnings persistence (B = 0.235, z (182) = 2.19, p < 0.05). Board structure is also positively associated with composite board processes (B = 0.256, t (182) = 1.98, p < 0.05). The mediator, composite board processes, is positively and significantly associated with earnings persistence (B = 0.563, z (182) = 5.38, p < 0.000). The total effect of board structure on earnings persistence was significant (TE = 0.177, SE = 0.080, CI = 0.011 to 0.337) and the direct effect was not significant (DE = 0.088, SE = 0.067, CI = –0.041 to 0.215). Thus, composite board processes mediate the relationship between board structure and earnings persistence (IE = 0.088, SE = 0.045, CI = 0.005–0.180), suggesting a full mediation.

Figure 2.

The mediation effect of board processes using binary mediation.

We also decompose composite board processes into its constituent parts. Figure 2, Panel B, shows the mediation analysis with respect to effort norms. Board structure is positively and significantly associated with earnings persistence (B = 0.235, z (182) = 2.19, p < 0.05) and effort norms (B = 0.168, t (182) = 2.06, p < 0.05). The mediating variable, effort norms, is positively and significantly associated with earnings persistence (B = 0.790, z (182) = 5.22, p < 0.000). The total effect of board structure on earnings persistence is significant (TE = 0.172, SE = 0.084, CI = 0.009 to 0.338), the indirect effect is significant (IE = 0.085, SE = 0.041, CI = 0.006 to 0.166) and the direct effect is insignificant (DE = 0.087, SE = 0.077, CI = –0.063 to 0.242). Thus, effort norms fully mediate the association between board structure and earnings persistence, which is consistent with the argument that boards that set high effort norms are more likely to be effective in monitoring managerial financial-reporting behaviour, thus resulting in higher earnings persistence.

We also examined the mediation effect of cognitive conflicts (presented graphically in Figure 2, Panel C). In our factor analysis (in untabulated results), we derived two factors, dealing with cognitive conflicts and existence of cognitive conflicts. Thus, we use multiple mediation to test the mediation effect in this scenario. We present results in Figure 2, Panel C. Board structure is positively and significantly associated with earnings persistence (B = 0.235, z (182) = 2.19, p < 0.05) and dealing with cognitive conflicts (B = 0.146, t (182) = 2.25, p < 0.05). However, board structure is not significantly associated with existence of cognitive conflicts. The mediating variable, dealing with cognitive conflicts, is positively and significantly associated with earnings persistence (B = 0.986, z (182) = 5.41, p < 0.000). The total effect of board structure on earnings persistence is significant (TE = 0.187, SE = 0.083, CI = 0.029–0.354), the indirect effect of dealing with cognitive conflicts is significant (IE = 0.092, SE = 0.041, CI = 0.010–0.180) but the indirect effect of existence of cognitive conflicts is insignificant (IE = 0.007, SE = 0.010, CI =−0.005 to 0.039). The total indirect effect (of both dealing with cognitive conflicts and existence of cognitive conflicts) is positive and significant (TIE = 0.099, SE = 0.042, CI = 0.027 to 0.184). The direct effect is insignificant (DE = 0.088, SE = 0.070, CI = –0.047 to 0.241). These findings show that dealing with cognitive conflicts fully mediates the association between board structure and earnings persistence. This supports the argument that boards that engage in challenging debates and collectively reach shared decisions are more likely to be effective in monitoring managerial financial-reporting behaviour, thus resulting in higher earnings persistence. However, our results do not provide support for the mediation effect of the existence of different views and opinions in the boardroom.

The mediation analysis in respect of the use of knowledge and skills is shown in Figure 2, Panel D. Board structure is positively and significantly (B = 0.235, z (182) = 2.19, p < 0.05) associated with earnings persistence and insignificantly associated with use of knowledge and skills. The mediator is positively and significantly associated with earnings persistence (B = 0.739, z (182) = 5.31, p < 0.000). The total effect of board structure on earnings persistence is significant (TE = 0.184, SE = 0.083, CI = 0.128 to 0.324), the direct effect is not significant (DE = 0.123, SE = 0.76, CI =−0.031 to 0.267) and the indirect effect is not significant (IE = 0.061, SE = 0.040, CI = –0.015 to 0.141). Our empirical evidence does not suggest that the relationship between board structure and earnings persistence is mediated by the use of knowledge and skills.

6. Discussion and conclusion

We follow Forbes and Milliken’s (1999) board-process framework and apply an input–process–output heuristic model to the board structure–processes–outcome setting. In an extended input–process–output model, we examine whether board processes mediate the association between board structure and monitoring financial reporting, thus earnings persistence.

We use PLS-SEM to test whether board structure affects board processes, which then affect monitoring financial reporting and earnings persistence in a sample of ASX listed companies. Merging survey data with financial data, we find that monitoring financial reporting significantly mediates the association between board processes and earnings persistence. We also find that board independence significantly affects monitoring financial reporting via board processes, which in turn affect earnings persistence. However, the other structural variables are not significantly associated with board processes. We also tested the mediation hypothesis using binary mediation, with bootstrapping. Our evidence from the binary-mediation model also suggests that combined board processes mediate the association between board structure and earnings persistence. Binary mediation was also used to examine the mediation effect of separate board processes. The results show that effort norms and dealing with cognitive conflicts fully mediate the association between board structure and earnings persistence. However, the evidence did not support the predicted mediation effect of use of knowledge and skills. Notwithstanding the aforementioned caveat, the PLS-SEM and binary-mediation analysis taken together suggest that the effect of board structure on earnings persistence is mediated by the inner workings of the board (i.e. board processes).

Although we took all possible measures to mitigate certain study limitations, some caveats remain. For example, we pre-populated surveys with board structure and other control-variable data after grouping them into industry-sector quartiles to protect respondent anonymity. We recognise this as a limitation due to lack of variation introduced by the method. As is the case with most surveys, this study also suffers from a small sample size, and the cross-sectional research design in this study does not capture variations in processes over time. Furthermore, despite using an extended model to add robustness to our main findings, the design of this study, which necessitated anonymity, limits further robustness tests. Thus, we cannot rule out all possible alternative explanations.

Future empirical research could utilise a similar research design to examine other behavioural dynamics of boards that are important in determining other financial-reporting issues. The focus of this article was on the mediating effect of board processes on the association between board structure and earnings persistence. Future research could examine any potential moderation effects of other variables in the model. Furthermore, this study could be extended to other institutional settings, in particular in emerging economies. Finally, the scope for qualitative studies should not be overlooked. While we find that there are significant associations between board processes and earnings persistence, qualitative work could provide a more in-depth examination of why and how these associations exist.

Board processes have been largely overlooked by accounting researchers and governance regulators, and cases such as Centro Australia exemplify the need to have sound board processes in addition to well-structured boards. The overriding proposition of this article is that in addition to board structure, board processes are important for effective monitoring financial reporting. Current ASX corporate-governance regulations emphasise board structure, with only limited provisions relating to board processes. While the importance of structure is acknowledged, this article provides insights for regulators by showing that board processes are also an important factor in monitoring financial reporting.

Footnotes

Appendix 1

Final transcript accepted 18 April 2019 by Millicent Chang (AE Finance).

Authors’ note

The authors appreciate the comments on earlier versions of this article made by seminar participants at La Trobe University, University of Otago, as well as participants at the British Accounting and Finance Association Conference, Manchester, the Quantitative Accounting Research Symposium, University of Auckland and the AFAANZ Conference, Gold Coast. The authors are also grateful to the anonymous survey respondents and appreciate the helpful comments on the draft survey instruments made by a number of academics and practitioners.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

Notes

References

Abbott

Parker

Presley

(2012) Female board presence and the likelihood of financial restatement. Accounting Horizons 26: 607–629.

Adams

Ferreira

(2009) Women in the boardroom and their impact on governance and performance. Journal of Financial Economics 94: 291–309.

Ahmed

Duellman

(2007) Accounting conservatism and board of director characteristics: An empirical analysis. Journal of Accounting and Economics 43: 411–437.

Ali

Park

(2016) The mediating role of an innovative culture in the relationship between absorptive capacity and technical and non-technical innovation. Journal of Business Research 69: 1669–1675.

Amason

(1996) Distinguishing the effects of functional and dysfunctional conflict on strategic decision making: Resolving a paradox for top management teams. Academy of Management Journal 39: 123–148.

Amason

Sapienza

(1997) The effects of top management team size and interaction norms on cognitive and affective conflict. Journal of Management 23: 495–516.

Blanthorne

Jones-Farmer

Almer

(2006) Why you should consider SEM: A guide to getting started. Advances in Accounting Behavioural Research 9: 179–207.

Bradbury

Mak

Tan

(2006) Board characteristics, audit committee characteristics and abnormal accruals. Pacific Accounting Review 18: 47–68.

Brooks

Oliver

Veljanovski

(2009) The role of the independent director: Evidence from a survey of independent directors in Australia. Australian Accounting Review 50: 161–177.

10.

Brown

Beekes

Verhoeven

(2011) Corporate governance, accounting and finance: A review. Accounting and Finance 51: 96–172.

11.

Carter

Lorsch

(2013) Back to the Drawing Board: Designing Corporate Boards for a Complex World. Boston, MA: Harvard Business Press.

12.

Chen

Zhou

(2007) Audit committee, board characteristics, and auditor switch decisions by Andersen’s clients. Contemporary Accounting Research 24: 1085–1117.

13.

Cohen

Krishnamoorthy

Wright

(2004) The corporate governance mosaic and financial reporting quality. Journal of Accounting Literature 23: 87–152.

14.

Cohen

Bailey

(1997) What makes teams work: Group effectiveness research from the shop floor to the executive suite. Journal of Management 23: 239–290.

15.

Daily

Dalton

Cannella

(2003) Corporate governance: Decades of dialogue and data. Academy of Management Review 28: 371–382.

16.

Dalton

Daily

Johnson

, et al. (1999) Number of directors and financial performance: A meta-analysis. Academy of Management Journal 42: 674–686.

17.

Davidson

Goodwin-Stewart

Kent

(2005) Internal governance structures and earnings management. Accounting and Finance 45: 241–267.

18.

De Baerdemaeker

Bruggeman

(2015) The impact of participation in strategic planning on managers’ creation of budgetary slack: The mediating role of autonomous motivation and affective organisational commitment. Management Accounting Research 29: 1–12.

19.

De Dreu

(2006) When too little or too much hurts: Evidence for a curvilinear relationship between task conflict and innovation in teams. Journal of Management 32: 83–107.

20.

Dechow

Schrand

(2010) Understanding earnings quality: A review of the proxies, their determinants and their consequences. Journal of Accounting and Economics 50: 344–401.

21.

DeVellis

(2003) Scale Development: Theory and Application. 2nd ed. Thousand Oaks, CA: SAGE.

22.

Dhaliwal

Naiker

Navissi

(2010) The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees. Contemporary Accounting Research 27: 787–827.

23.

Dillman

(2011) Mail and Internet Surveys: The Tailored Design Method – 2007 Update with New Internet, Visual, and Mixed-mode Guide. New York: John Wiley & Sons.

24.

Eagly

Carli

(2003) Finding gender advantage and disadvantage: Systematc research integration is the solution. The Leadership Quarterly 14: 851–859.

25.

Eisenhardt

(1989) Agency theory: An assessment and review. Academy of Management Review 14: 57–74.

26.

Fama

Jensen

(1983) Agency problems and residual claims. Journal of Law and Economics 26: 327–349.

27.

Federal Court of Australia (2011) Australian Securities and Investments Commission v Healey and Others, FCA 717; Australian Securities and Investments Commission v Healey and Others (No. 2), FCA 1003. Federal Court of Australia, Middleton J, 27 June 2011, 31 August 2011.

28.

Field

(2013) Discovering Statistics Using IBM SPSS Statistics. Thousand Oaks, CA: SAGE.

29.

Finkelstein

Mooney

(2003) Not the usual suspects: How to use board process to make boards better. Academy of Management Executive 17: 101–113.

30.

Forbes

Milliken

(1999) Cognition and corporate governance: Understanding boards of directors as strategic decision-making groups. Academy of Management Review 24: 489–505.

31.

Fornell

Larcker

(1981) Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research 18: 39–50.

32.

Gabrielsson

Winlund

(2000) Boards of directors in small and medium-sized industrial firms: examining the effects of the board’s working style on board task performance. Entrepreneurship & Regional Development 12: 311–330.

33.

García-Meca

Sánchez-Ballesta

(2009) Corporate governance and earnings management: A meta-analysis. Corporate Governance: An International Review 17: 594–610.

34.

Gladstein

(1984) Groups in context: A model of task group effectiveness. Administrative Science Quarterly 29: 499–517.

35.

Goodstein

Gautam

Boeker

(1994) The effects of board size and diversity on strategic change. Strategic Management Journal 15: 241–250.

36.

Hair

Hollingsworth

Randolph

, et al. (2017a) An updated and expanded assessment of PLS-SEM in information systems research. Industrial Management and Data Systems 117: 442–458.

37.

Hair

Hult Tomas

Ringle

, et al. (2017b) A Primer on Partial Least Squares Structural Equation Modelling (PLS-SEM). Thousand Oaks, CA: SAGE.

38.

Hair

Ringle

Sarstedt

(2011) PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice 19: 139–151.

39.

Hambrick

Werder

Zajac

(2008) New directions in corporate governance research. Organization Science 19: 381–385.

40.

Hillman

Shropshire

Cannella

, et al. (2007) Organizational predictors of women on corporate boards. Academy of Management Journal 50: 941–952.

41.

Huse

(2005) Accountability and creating accountability: A framework for exploring behavioural perspectives of corporate governance. British Journal of Management 16: S65–S79.

42.

Huse

Solberg

(2006) Gender-related boardroom dynamics: How Scandinavian women make and can make contributions on corporate boards. Women in Management Review 21: 113–130.

43.

Jackson

(1992) Consequences of group composition for the interpersonal dynamics of strategic issue processing. Advances in Strategic Management 8: 345–382.

44.

Jehn

(1995) A multimethod examination of the benefits and detriments of intragroup conflict. Administrative Science Quarterly 40: 256–282.

45.

Johnson

Daily

Ellstrand

(1996) Boards of directors: A review and research agenda. Journal of Management 22: 409–438.

46.

Judge

Zeithaml

(1992) Institutional and strategic choice perspectives on board involvement in the strategic decision process. Academy of Management Journal 35: 766–794.

47.

Kent

Routledge

Stewart

(2010) Innate and discretionary accruals quality and corporate governance. Accounting and Finance 50: 171–195.

48.

Kline

(2011) Principles and Practice of Structural Equation Modeling. New York: Guilford Press.

49.

Kumar

Stern

Anderson

(1993) Conducting interorganizational research using key informants. Academy of Management Journal 36: 1699–1651.

50.

Larcker

Richardson

Tuna

(2007) Corporate governance, accounting outcomes, and organizational performance. The Accounting Review 82: 963–1008.

51.

Lee

Petter

Fayard

Robinson

(2011) On the use of partial least squares path modeling in accounting research. International Journal of Accounting Information Systems 12: 305–328.

52.

LePine

Piccolo

Jackson

, et al. (2008) A meta-analysis of teamwork processes: Tests of a multidimensional model and relationships with team effectiveness criteria. Personnel Psychology 61: 273–307.

53.

Lipton

Lorsch

(1992) A modest proposal for improved corporate governance. The Business Lawyer 48: 59–77.

54.

Lohmöller

(1989) Latent Variable Path Modeling with Partial Least Squares. Heidelberg: Physica-Verlag.

55.

MacKinnon

Lockwood

Hoffman

, et al. (2002) A comparison of methods to test mediation and other intervening variable effects. Psychological Methods 7: 83–104.

56.

Mather

Ramsay

(2006) The effects of board characteristics on earnings management around Australian CEO changes. Accounting Research Journal 19: 78–93.

57.

Matolcsy

Tyler

Wells

(2012) Is continuous disclosure associated with board independence? Australian Journal of Management 37: 99–124.

58.

McGrath

(1964) Social Psychology: A Brief Introduction. New York: Holt, Rinehart and Winston.

59.

McNulty

Florackis

Ormrod

(2013) Boards of directors and financial risk during the credit crisis. Corporate Governance: An International Review 21: 58–78.

60.

Minichilli

Zattoni

Zona

(2009) Making boards effective: An empirical examination of board task performance. British Journal of Management 20: 55–74.

61.

Minichilli

Zattoni

Nielsen

, et al. (2012) Board task performance: An exploration of micro-and macro-level determinants of board effectiveness. Journal of Organizational Behavior 33: 193–215.

62.

Nielsen

Huse

(2010) The contribution of women on boards of directors: Going beyond the surface. Corporate Governance: An International Review 18: 136–148.

63.

Nitzl

Roldán

Cepeda

(2016) Mediation analyses in partial least squares structural equation modeling: Helping researchers discuss more sophisticated models. Industrial Management & Data Systems 116: 1849–1864.

64.

Oei

Ramsay

Mather

(2008) Earnings persistence, accruals and managerial share ownership. Accounting and Finance 48: 475–502.

65.

Ong

Wan

(2008) Three conceptual models of board role performance. Corporate Governance 8: 317–329.

66.

Parker

Nouri

Hayes

(2011) Distributive justice, promotion instrumentality, and turnover intentions in public accounting firms. Behavioral Research in Accounting 23: 169–186.

67.

Pearce

Zahra

(1992) Board composition from a strategic contingency perspective. Journal of Management Studies 29: 411–438.

68.

Peasnell

Pope

Young

(2000) Accrual management to meet earnings targets: UK evidence pre- and post-Cadbury. British Accounting Review 32: 415–445.

69.

Petra

(2007) The effects of corporate governance on the informativeness of earnings. Economics of Governance 8: 129–152.

70.

Pettigrew

(1992) On studying managerial elites. Strategic Management Journal 13: 163–182.

71.

Pfeffer

Salancik

(1978) The External Control of Organizations: A Resource Dependence. Redwood City, CA: Stanford University Press.

72.

Podsakoff

MacKenzie

Lee

(2003) Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology 88: 879–903.

73.

Preacher

Hayes

(2004) SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, Instruments, and Computers 36: 717–731.

74.

Preacher

Hayes

(2008) Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods 40: 879–891.

75.

Pugliese

Bezemer

Zattoni

Huse

Van den Bosch

Volberda

(2009) Boards of directors’ contribution to strategy: A literature review and research agenda. Corporate Governance: An International Review 17: 292–306.

76.

Richardson

Sloan

Soliman

, et al. (2005) Accrual reliability, earnings persistence and stock prices. Journal of Accounting and Economics 39: 437–485.

77.

Sloan

(1996) Do stock prices fully reflect information in accruals and cash flows about future earnings? Accounting Review 71: 289–315.

78.

Sonnenfeld

(2002) What makes great boards great. Harvard Business Review 80: 106–113.

79.

Strydom

Yong

HHA

Rankin

(2017) A few good (wo)men? Gender diversity on Australian boards. Australian Journal of Management 42: 404–427.

80.

Tabachnick

Fidell

(2007) Multivariate Analysis of Variance and Covariance Using Multivariate Statistics. Boston, MA: Allyn & Bacon, pp. 243–310.

81.

Thomas

Ely

(1996) Making differences matter: A new paradigm for managing diversity. Harvard Business Review 74: 79–90.

82.

Van Ees

Van der Laan

Postma

(2008) Effective board behavior in the Netherlands. European Management Journal 26: 84–93.

83.

Wageman

(1995) Interdependence and group effectiveness. Administrative Science Quarterly 40: 145–180.

84.

Wan

Ong

(2005) Board structure, process and performance: Evidence from public-listed companies in Singapore. Corporate Governance: An International Review 13: 277–290.

85.

Xie

Davidson

III DaDalt

(2003) Earnings management and corporate governance: The role of the board and the audit committee. Journal of Corporate Finance 9: 295–316.

86.

Zona

Zattoni

(2007) Beyond the black box of demography: Board processes and task effectiveness within Italian firms. Corporate Governance: An International Review 15: 852–864.