Abstract
This research examines the relationship among Board Diversity, Social Capital, and Governance Effectiveness by asking, “does board ethno-racial diversity moderate the relationship between Social Capital and Governance Effectiveness, and if so, how?” Exploring the direct and interacting effects of demographic diversity and Social Capital, and their relation to governing-group effectiveness using a two-sample field survey design, we illustrate whether heterogeneous or homogeneous group compositions amplify or attenuate Governance Effectiveness, and to what degree. Primary analyses find no support for Board Diversity moderating the Social Capital-Governance Effectiveness relationship, with secondary analysis revealing a more complex interaction for Governance Effectiveness, albeit inconsistently, across samples. Our investigation points to the value of social resources in understanding governance as an inherently socially complex activity or capability, predicated on truce or mutual agreement and shaped by the composition and connections of boards.
Nonprofit governance is frequently viewed through the lens of board process, adherence to rules and regulations, and formalized structures and systems of control, yet the work of governing groups is highly interdependent and inherently social, factors that are often overlooked and underspecified. This social underbelly is both a strength (e.g., cohesion, solidarity, togetherness, trust, cooperation) and an area of vulnerability (e.g., homophily, isolation, groupthink, blindspots, exclusion) as many governance challenges have no codified or standardized response and therefore rely on mutual agreement, consensus, and coalitioning. Social Capital, thought to reflect real and potential resources mobilizable through the social relationships of members of a group or unit (Nahapiet & Ghoshal, 1998), affords an important means of characterizing the social value, strength, vitality, and resilience of groups. It has also been predicated on a tension between outward facing and inward facing dynamics (Leana & Pil, 2006; Weisinger & Salipante, 2005). This tension presents boards of directors with a paradox when it comes to Social Capital and Board Diversity. Evidence exists to suggest that Board Diversity and Social Capital each may improve governing performance, with Social Capital potentially conferring enriched decision-making and resilience by virtue of interpersonal trust, shared understanding, and information sharing (Fredette & Bradshaw, 2012; Leana & Pil, 2006). Diversity, alternatively, may similarly improve decision-making and organizational agility by affording broader access to otherwise unheard perspectives, novel and unanticipated opportunities, and economic or legitimacy-conferring resources (Brown, 2002; Harris, 2014; Siciliano, 1996). However, it remains unclear how diversity and Social Capital interact in boardroom settings, and whether their combined potential benefits outweigh the potential risks associated with their interaction, which could result in difference-driven conflict and group deterioration, attenuating Governance Effectiveness. Indeed, attention has recently turned to theorizing that in the highly interdependent work of governance (Johnson et al., 2013), the direct and interacting effects of diversity and Social Capital hold potentially unanticipated consequences (Tasheva & Hillman, 2019).
Defining what does or does not constitute Board Diversity, Social Capital or Governance Effectiveness is frequently contested and often rife terrain for dispute (and rightly so), and in this research, we define our terms up front. Notwithstanding the possible competing traditions and perspectives of greater or lesser scope, for the purposes of this research, we herein define the following:
Board Diversity in terms of ethno-racial demography, reflecting “the representation, in one social system, of people with distinctly different group affiliations of cultural significance” (Cox, 1993, p. 5), emphasizing variety in the representation of underrepresented minority community members. Adapting criteria used in Canadian employment law and the collection of census information (Statistics Canada, 2016), we define ethno-racial categories in this research to include Aboriginal, Arab, Black, Chinese, Filipino, Japanese, Korean, Latin American, South Asian, Southeast Asian, West Asian, and White.
Social Capital as the pool of potential and actual structural, cognitive, and relational resources available through and embedded in the relationships of group members (Adler & Kwon, 2002; Nahapiet & Ghoshal, 1998; Oh et al., 2004), which Leana and Pil (2006) use as the basis for internal Social Capital. They emphasize the relationships among group members that influence information sharing, collective visioning, and trust, which drive the flows of intellectual and social resources.
Governance Effectiveness as emphasizing the capacity of boards to engage in a series of strategic and oversight activities, including providing strategic guidance in long-term planning and safeguarding mission fulfillment, as well as ensuring executive succession or performance management and performing fiduciary accountability (Bradshaw et al., 1992; Fredette & Bradshaw, 2012).
These definitions inform this research, in which we ask, “does board ethno-racial diversity moderate the relationship between Social Capital and Governance Effectiveness, and if so, how?” Alternatively put, is there a performance-benefit or a penalty-cost in the relationship between Social Capital and Governance Effectiveness associated with greater ethno-racial variety in the nonprofit boardroom? Exploring the direct and interacting effects of demographic diversity and Social Capital, we illustrate whether more-diverse or less-diverse board compositions amplify or attenuate Governance Effectiveness, and to what degree. This research contributes to the leadership diversity and governance literature by attending to an underexamined aspect of the interdependence in board composition, social dynamics, and board performance, by demonstrating how Board Diversity interplays with Social Capital and Governance Effectiveness. This research uses a two-sample field survey design to determine the pattern of relationships among our constructs, following which we test the replicability of our results with an independent second data set. We conclude by discussing findings and implications for research, practice, and policy.
Social Capital, Board Diversity, and Governance Effectiveness
In the management literature, Social Capital has been argued to be an essential marker of group performance in a variety of contexts (Inkpen & Tsang, 2005; Prashantham & Dhanaraj, 2010), including among boards of directors’ capacity to share knowledge and information, agree on a common vision, and demonstrate trust (Fredette & Bradshaw, 2012; Harris & Helfat, 2007). It is a construct frequently associated with relational strength, social solidarity, and group togetherness needed for coordinated collective action (Nahapiet & Ghoshal, 1998), often drawing comparison to other forms of capital which are subject to gains if deployed wisely, or losses if deployed ineffectively (Adler & Kwon, 2002). Management scholars have developed an integrative three-dimension construct based on structural, cognitive, and relational aspects of Social Capital (Nahapiet & Ghoshal, 1998; Tsai & Ghoshal, 1998), used in the study of groups and teams (Fredette & Bradshaw, 2012; Oh et al., 2004), within departments or business activities (Gargiulo & Benassi, 2000), and among organizations and their environments (Sundaramurthy et al., 2014).
Here, we conceptualize Social Capital in the managerial tradition (Nahapiet & Ghoshal, 1998), as the latent reflection of structural, cognitive, and relational factors that in combination afford the capacity to share and integrate information, develop and retain collective cognitive schemas, and exhibit sufficient trust and team comradery needed to support the functional work of boards as governing groups. The structural aspect of Social Capital is reflective of both the strength of connection among board members and the quality of those connections (Leana & Pil, 2006). For information sharing, these connections are indicative of structural Social Capital by demonstrating the ability and willingness to communicate and update others of pending challenges, concerns, vulnerabilities, or uncertainties. The ability of boards to collectively generate insight or appropriate value in relationships by reaching the disparate sources or knowledge pools of board members relies on brokering member information flows (Burt, 1997), whether connections are less-known to each other (Granovetter, 1973) or deeply embedded in trusted relationships (Uzzi, 1997). For governance, Fredette and Bradshaw (2012) emphasize the impact of information-sharing characteristics upon which boards rely to develop, integrate, and distribute knowledge and insights needed to develop clear, reliable understanding of pending events and circumstances.
The cognitive aspect of Social Capital is reliant in many ways on the effective information sharing among board members, as the frequent updating and transparent communication among members improves the probability of developing and retaining an accurate understanding of a common vision or collective understanding (Weick & Roberts, 1993). For governance, building a shared vision of organizational purpose and strategy among board members is essential (Fredette & Bradshaw, 2012; Hawkins, 2014), because without collective agreement on the scope and direction of the organization’s activities, mission and focus may become obscured or contested, thereby increasing the chances of organizational drift (Bennett & Savani, 2011). This suggests that the association between board Social Capital, and the collective capacity to govern effectively, may rely on the board sustaining an uncontested—or shared—collective vision of the organization’s goals and the paths to be followed to reach them.
The relational dimension of Social Capital most frequently takes the form of trust (Leana & Pil, 2006; Nahapiet & Ghoshal, 1998), where trust characterizes the relational qualities that facilitate risk-taking and vulnerability (Granovetter, 1985; Tsai & Ghoshal, 1998), encourage reciprocal sharing (De Vries, 1999), and inform decisions about whether board members are safe to continue investing in relationships with others in groups (Ferrin et al., 2006). The ability to confidently rely on the trustworthiness of teammates, particularly in uncertain and consequential circumstances such as those found in governance settings, is paramount to coordinated collective action (Enjolras, 2009), which we anticipate to improve Governance Effectiveness.
The structural, cognitive, and relational dimensions of Social Capital, and our characterization of them in the form of information sharing, shared vision, and trust, are unique in composition and effect but also interdependent and reflective of a broader latent concept. Consistent with the work of others (Fredette & Bradshaw, 2012; Leana & Pil, 2006), we argue that board Social Capital contributes positively to Governance Effectiveness:
We take the position that the arguments underlying the business case for diversity have been well-made by others (including Cox et al., 1991; Jayne & Dipboye, 2004), notwithstanding the serious and continuing challenges posed by such factors as recruitment, selection, and retention biases, or tokenism, alienation and marginalization, nor those internal to boards and organizations that frequently dissuade engagement, inhibit participation, and impede empowerment. Our intuition is that greater Board Diversity ought to enhance the capacity to govern. Recently, Fredette and Bernstein (2019) explored the impact of ethno-racial Board Diversity on three areas of governance activity. Relying on a proportional Board Diversity measure, they found some support for a relationship between Board Diversity and facets of board performance. By increasing the representation and participation of people with distinct and culturally relevant group affiliations (Cox, 1993), boards have the potential to grow the pool of informational and financial resources available to them (Harris, 2014), as well as enhance innovative capacity of the organization to better identify and respond to the needs and interests of stakeholders in the external environment (Brown, 2002). The organizational benefits of greater Board Diversity are not limited to objective resource or process gains; however, boards also serve as a signaling mechanism expressing values, aspirations, and commitments to internal and external audiences that may demonstrate and confer legitimacy (Brown et al., 2002; Siciliano, 1996).
We posit here that governing effectiveness is enhanced by more ethno-racially diverse board compositions by way of information, economic, and decision-making benefits not otherwise available to less diversely composed boards. These benefits might include the capacity to better understand and reflect the needs and interests of community stakeholders in developing responsive service delivery programs, the ability to identify and access funding, clients, and workforce resources that might otherwise remain inaccessible, and the opportunity for richer ideation and deliberation in the course of strategic planning and decision-making activities:
Paradox in Board Diversity and Social Capital for Governance Effectiveness
We noted in the introduction that the interaction of Social Capital and Board Diversity poses a unique paradox in terms of governing effectiveness. Implicit in diversity and group performance studies is the notion of underlying fault-lines and the potential for disruption to board performance and group Social Capital as diversity of the group increases (Jehn & Bezrukova, 2010; Lau & Murnighan, 1998). Kanter (1977) was among the first to suggest a relationship between proportions of ingroup and outgroup composition and group function, with scholars subsequently testing thresholds and critical mass arguments in group diversity (Fredette & Bernstein, 2019; Konrad et al., 2008). In many respects, this echoes assertions that as minority membership increases, the majority group may perceive a threat to their existing positions and associated power (Blalock, 1967), or the greater likelihood of conflict and group fracture (Jehn et al., 1999), or an increase in board member turnover and decline in collective performance (Pelled, 1996). This anticipates that the potential gains to performance afforded by group diversity are mitigated by a looming social cost. Differentiating between research that links increased group diversity to constructive conflict and destructive social penalty (Jehn et al., 1999), in the absence of intervening activities and processes, more diverse boards of directors are believed to become more fragmented and conflictual, underperforming their less-diverse peers in areas of board performance, group cohesiveness, and commitment (Fredette et al., 2016).
Leana and Pil (2006) examined Social Capital as dualistic, simultaneously external in its orientation to stakeholders and constituents, and internal in its structural, cognitive, and relational influence on organization leaders and members, with unique but largely independent implications. The nature of nonprofit governance and decisions about board composition, however, place boards of directors at the nexus of this internal-organization and external-community divide, where the implications of the internal/external duality often result in less independence than might be anticipated. This view of Social Capital was picked up by Fredette and Bradshaw (2012); however, they stopped short of addressing compositional considerations.
Governing activities, accomplished by debate and discussion, held in the purview of boards of directors, speak to the raison d’être for separating the responsibilities of boards from those of managers in the sector (Brown & Guo, 2010; Herman & Renz, 1998). That is, they serve to preserve and coordinate the future of the organization in the face of present-day challenges and opportunities encountered by managers (Drucker, 1990). Board members, as organizational leaders, deal with complex decision-making which often involves debate, discussion, and dissent (Brown & Guo, 2010), which through compromise or reconciliation result in truces that establish a mutually agreed upon set of governing parameters (Nelson & Winter, 1982).
Board Diversity potentially extends the board’s reach, connecting to resources and information, clients and constituents, and funding and markets not currently accessible, thereby improving decision-making, particularly when board members engage in greater information sharing, possess a shared vision of the organization’s purpose, and demonstrate interpersonal trust (Tasheva & Hillman, 2019). At the same time, Board Diversity has the potential to introduce differing perspectives, governing assumptions, decision-making routines, and operating habits potentially attenuating the board’s cohesion or cognitive and affective bonds (Tasheva & Hillman, 2019), degrading truces at least in the short term. That is, conflict degrades the value of Social Capital stocks, leaving truces or “agreed-upon areas of legitimate activity” essential for social function weakened. In the absence of basic boundaries, values, and pathways for discussion, it seems trust is unlikely to hold, let alone develop among board members:
Pivoting to this alternative perspective, it is unclear whether the potential benefits of diversity to Social Capital outweigh the potential risks associated with difference-driven conflict and group deterioration often perceived to accompany the addition of visibly-diverse board members, which would attenuate the strength of the Social Capital–Governance Effectiveness relationship. This suggests that the association between board Social Capital and the collective capacity to govern effectively may, in part, be a product of the underlying composition of the board, such that boards with more ethno-racial variety differ from less-diverse ones.
Research Design, Method, and Analysis
This research uses a two-sample design to test hypotheses using comparable dependent and independent construct measures and testing protocols, only differing the control variables to account for differences in the context in which the data were collected. In what follows, we describe sampling characteristics, measurement approaches, testing protocols, and results for each study in turn, using structural equation modeling (SEM) with AMOS 25.
Study 1—Data Collection and Sampling
We first examine data collected from a nationally sampled survey of Canadian not-for-profit organizations conducted in 2008, using a 14-page mail-in survey which was pretested and refined with focus groups of experienced professionals prior to distribution. These efforts involved outreach to 825 organizations, of which 236 responded with sufficient data to be considered complete, yielding an effective response rate of 28.6%. Each participant represented a nonprofit board and answered questions for individual members of each board, resulting in 3,011 board members being represented in this sample. All data included in this study were provided by respondents characterized as either Executive Director/Chief Executive (78.4%), Board Chairperson (3.8%), or other board representative (17.8%) with firsthand knowledge of board members, practices, and outcomes.
Our sample, approximately 60% of which is composed of organizations based in the province of Ontario, reflects an organizational population that is relatively large in terms of budget (M = $9.8 m CDN; standard deviation = $33.3 m CDN) and full-time equivalent staff (median = 11). In addition, this set of organizations tended to be relatively mature, with an average age of 42.5 years, and drawn from a variety of sectors. It would be reasonable to characterize our data as consisting of the larger, older, more-established end of the nonprofit organizational spectrum operating primarily in central Canada’s health and social welfare sectors.
Study 2—Data Collection and Sampling
The data in our second study are drawn from an examination of Board Diversity among arts, sports, and environmental organizations conducted in 2012 in the Greater Toronto Area (GTA). The GTA is estimated to include slightly more than 6.1 million residents, with approximately 42.9% of the population self-identifying as a member of a visible minority community (Statistics Canada, 2016).
Survey data were collected using a brief online survey of senior nonprofit organizational leaders, where responses were sought and received from Chief Executives, Executive Directors, or Board Chairpersons on the impact of board room diversity on their organization’s governance and performance. A total of 903 organizations were contacted, yielding 269 responses, 247 of which we consider substantially complete (response rate of 27.4%). Each participant represented a nonprofit board and answered questions for individual members of each board, resulting in 2,789 board members being represented in this sample. We provided respondents a unique direct link to the survey, later confirming no redundancy among organization or respondent names.
Characterizing the responding organizations included in our second study, most were comparatively smaller and younger than those in the first study, with an average budget of $1.4 m CDN (standard deviation = $3.6 m CDN), an average number of full-time staff of 12.5 (median = 3), and an average age of the organizations was 30.1 years (SD = 25 years). Over half of those that replied to the survey were involved in arts and culture (63.5%), followed by sports and recreation (29.3%), and environmental (7.2%) sectors.
Missing Data Protocols
In the conduct of field surveys, missing data are not uncommon, but do raise two types of concerns: first, whether the omissions constitute a threat to the data collection effort by virtue of the scale and scope of missingness; and second, whether the randomness of the missing data can be qualified as random (Newman, 2014). We performed missing value analysis to identify the scope of concern, following which we assess the randomness of the missing data by estimating Little’s (1988) Missing Completely at Random (MCAR) test statistic to determine whether the data reflect a MCAR standard, before proceeding with a corrective strategy.
Study 1
Missing value analysis conducted at the item level reveals that no variable is missing more than 10% of its values, with 84.75% of cases and 98.85% of the values complete. Testing randomness, we estimate Little’s chi-square statistic where a significant statistic (i.e., p < .05) is suggestive of data not missing at random. These data meet the condition of MCAR (χ2 = 5,810.754, df = 11,915, p = 1.0).
Study 2
In this second sample, item-level missing value analysis demonstrated 55.87% of cases and 74.86% of values complete, with each item included in our modeling missing between 10.1% and 31.6% of its values. We estimate Little’s MCAR test statistic (χ2 = 2,262.705, df = 2,949, p = 1.0), revealing that these data too meet the MCAR standard.
Although rigidly defined standards of exclusion are difficult to trace, Newman (2014) suggests using a maximum likelihood data imputation strategy when more than 10% of the respondent pool is composed of partial respondents and the data meet MCAR assumptions. Given our analytic approach to testing moderation hypotheses, a full information maximum likelihood (FIML) estimation approach was employed (Enders, 2001; Graham, 2009; Newman, 2014).
Variables and Construct Measurement
We take a consistent approach to measurement and assessment of construct validity in each study, beginning by introducing our dependent variable, followed by our independent variables, and finally, explaining the factors we chose to control. Subsequently, the results of confirmatory factor analysis of our Social Capital and Governance Effectiveness measurement models are presented, as are the results of discriminant validity tests.
Governance Effectiveness
Assessing board performance or Governance Effectiveness is both complex and contested terrain in the literature (Herman & Renz, 1998). Our measure draws on a previously published five-item Likert-type scale (Bradshaw et al., 1992; Fredette & Bradshaw, 2012), in which we requested respondents to consider their board’s capacity to meet governance challenges by assessing (a) the overall performance of the board, (b) the quality of fiduciary and financial oversight, (c) the capacity to safeguard and fulfill the organization’s mission, (d) the provision of performance evaluation and feedback to the Executive Director or Chief Executive, and (e) the undertaking of annual strategic planning activities (Appendix). In Study 1, these items were measured with a 10-point scale as it had been elsewhere (Bradshaw et al., 1992; Fredette & Bradshaw, 2012), demonstrating a reliability (Cronbach’s α = .881), with removal of any single item not significantly improving the scale reliability. In Study 2, a 5-point scale was employed and demonstrated strong reliability, with a Cronbach’s alpha of .856. Alterations to the scale range resulted from survey design constraints centering on aesthetic appearance, software limitations, and considerations related to consistency in the presentation of other measures. Again, removal of any single item did not significantly improve the measurement reliability.
Social Capital
We adopted a three-factor model of Social Capital based on information sharing, shared vision, and trust, which in prior research demonstrated reliability and validity (Leana & Pil, 2006, p. 364). The item wording and scale content were adapted for a nonprofit governance context in prior research (Fredette & Bradshaw, 2012). In Study 1, each Social Capital dimension was measured with a six-item scale using a 7-point Likert-type rating system. Scale reliability statistics for information sharing (α = .861), shared vision (α = .908), and trust (α = .872) suggested a reasonable basis for further investigation.
In Study 2, we pared down from the number of items used to measure Social Capital in Study 1, in response to concerns about respondent fatigue resulting from the length of the survey instrument. To do so, an examination of the factor structure of items used in Study 1 was undertaken to discern which items would retain face validity while also providing statistical reliability and validity. Each dimension of Social Capital was measured using a three-item scale using a 5-point Likert-type rating system (Appendix). Pared down first-order constructs for information sharing (α = .829), shared vision (α = .898), and trust (α = .847) demonstrated reliability.
Board Diversity
We operationalized the concept of diversity in terms of the ethno-racial variety in the composition of the boards of directors by calculating Blau Index scores for each organization (Blau, 1977). 1 Variety speaks to the scope of difference or degree of heterogeneity among board members that might attenuate homophilistic tendencies commonly associated with homogeneous groups, making Blau indices appropriate measures of diversity for our purposes (Harrison & Klein, 2007). Our measure of diversity is a board-level measure in which one respondent per organization identified the composition of their board of directors based on Canadian census categories of ethnic origin and visible minority status. These counts were translated into a single board-level score derived as a proportion of total board membership ranging from 0 to 1, with higher scores indicating a greater variety of groups represented. In Study 1, for example, the composition of board membership from our boards of directors was dominantly white, with boards having 2.17 visible minority groups represented, yielding a mean Blau score of .191 (SD = .211). Whereas in Study 2, the composition of board membership was dominantly White but slightly more varied, with Blau Index scores averaging .251 (SD = .216), representing 1.83 (SD = 3.03) racialized groups on average.
Controls
We controlled for four characteristics, of which three were organizational—sector, board size, and organizational age, and one was individual—respondent diversity, in both studies and a control for community size in Study 1, which included responses from across the country, in contrast to the geographic concentration of Study 2. The first control variable, organization sector, coded each organization based on respondent perception of primary activity, providing a multi-categorical variable (11 subcategories in Study 1; three in Study 2) intended to capture underlying sector-level factors that might influence demographic participation rates and governance activities. The second control variable captured board size to control for differences in the availability of board positions and potential barriers to entry (Goodstein et al., 1994). The third control variable, organization age, was measured by the number of years in operation. This is important because the date of founding and age of an organization have been shown to influence the formalization with which organizations approach aspects of diversity and governance (Bradshaw et al., 1996; Bradshaw & Fredette, 2013). Finally, we include respondent diversity as a control variable to address potential sources of bias resulting from the ethno-racial background of respondents as it seemed reasonable that their influence might underlie an organization’s interest, ability, and sensitivity in addressing issues of compositional diversity. These variables add control for confounding factors, even though not all control variables are statistically significant on their own.
In the Canadian context, communities with larger populations tend to exhibit more diversity than smaller or more isolated ones (Statistics Canada, 2017). Community size grouped the population of the location in which the organization’s main office resides into categories, 1 = Big city (over 200,000); 2 = Small city (100,000–200,000); 3 = Town (25,000–100,000); 4 = Small town/rural (under 25,000). Because the second sample was drawn from a well-documented ethno-racially diverse metropolis (Statistics Canada, 2016), we only controlled for community size in the first study to address confounding factors and improve the accuracy of our models.
Measurement Models, Confirmatory Factor Analysis, and Structural Modeling
Our testing is conducted in two phases in each study, first examining the factor structure and validity of measurement models, and second, testing structural models for a Social Capital–Board Diversity interaction on Governance Effectiveness. Our tests involve assessing measurement models for the latent second-order construct of Social Capital as a product of three first-order factors (information sharing, shared vision, and trust), assessing the measurement fitness of our dependent variable Governance Effectiveness, and establishing convergent and discriminant validity among our predictor and dependent variables. These measurement models establish a foundation to develop and test structural models which incorporate covariate controls, Board Diversity, and interaction terms.
Measurement models
To test the appropriateness of constructing Social Capital as a second-order latent variable, we begin by loading the measurement items associated with each subordinate or first-order construct. Next, we assess the factor loading scores, following which we constrain the covariance parameter to null and then to unity, finally removing the constraint to freely estimate the parameter. For Social Capital to be rightly identified as a common latent factor, each first-order factor must load significantly, with freely estimated covariance parameters providing superior model fit compared with the alternatives (Balasubramanian et al., 2003). If the null model fit is preferred, it suggests the first-order factors reflect independent or unrelated constructs, where a preference for the unity model fit statistics suggests the data support a single factor first-order construct (Balasubramanian et al., 2003). As Governance Effectiveness is proposed as a single first-order factor, our measurement modeling centers on ensuring model fit statistics based on acceptable factor loading scores.
Standardized regression weights for first- and second-order factors are presented in Table 1, with corresponding measurement items included for Study 2 as identified in the appendix. Composite reliability (CR), average variance extracted (AVE), maximum shared variance (MSV), and correlations between Social Capital and Governance Effectiveness presented in Table 2 with the square root of AVE reported on the diagonal. Analyses demonstrate strong reliability (CR scores > .7), convergent validity (AVE scores > .5), with correlations among second-order latent variables below the threshold for concern (Henseler et al., 2015; Weston & Gore, 2006).
Standardized Regression Weights for Measurement Models.
Measurement Model Reliability and Validity.
Note. The bolded values reflect the AVE measurement and are not subject to a significance test. Significance has been added where appropriate. Square root of AVE reported on the diagonal. CR = composite reliability; AVE = average variance extracted: MSV = maximum shared variance.
Significance of correlations: †p < .100. *p < .050. **p < .010. ***p < .001.
Measurement model fit statistics outlined in Table 3 demonstrate a preference for the hypothesized second-order three-factor model of Social Capital (Study 1: χ2 = 212.880, df = 129, χ2/df =1.65; Tucker–Lewis index [TLI] = .966; comparative fit index [CFI] = .971; standardized root mean square residual [SRMR] = .038; root mean square error of approximation [RMSEA] = .053, pClose = .355; Study 2: χ2 = 45.715, df = 21, χ2/df = 2.177; TLI = .975; CFI = .985; SRMR = .022; RMSEA = .069, pClose = .116), and a single first-order model of our dependent variable Governance Effectiveness (Study 1: χ2 = 7.424, df = 4, χ2/df = 1.856; TLI = .986; CFI = .995; SRMR = .022; RMSEA = .060, pClose = .327; Study 2: χ2 = 5.779, df = 4, χ2/df = 1.445; TLI = .991; CFI = .997; SRMR = .021; RMSEA = .043, pClose = .483).
Convergent and Discriminant Validity of Social Capital and Governance Effectiveness Measurement Models.
Note. TLI = Tucker–Lewis index; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.
Table 3 summarizes results of the confirmatory factor analyses and discriminant validity testing, demonstrating excellent fit in Study 1 (χ2 = 393.288, df = 119, χ2/df = 1.796; TLI = .947; CFI = .954; SRMR = .045; RMSEA = .058, pClose = .074) and an adequate fit in Study 2 (χ2 = 156.233, df = 129, χ2/df = 2.298; TLI = .951; CFI = .963; SRMR = .043; RMSEA = .073, pClose = .008). Further examining discriminant validity, where the square root of AVE for Governance Effectiveness was smaller than its correlation with Social Capital in Study 1 (Hair et al., 2010; Malhotra & Dash, 2011), we tested and found a statistically significant preference for the freely estimated model representing Social Capital and Governance Effectiveness as distinct constructs (Study 1: Δχ2/df = 5.447, p < .02; Study 2: Δχ2/df = 71.728, p < .001).
To address common method bias, a common latent factor was added to the measurement models and subsequently constrained and unconstrainted to compare goodness of model fit using a chi-square difference test. Results demonstrate a nonsignificant chi-square change (Study 1: Δχ2/df = 2.95; Study 2: Δχ2/df = 2.09), providing no evidence of underlying common method bias (Podsakoff et al., 2003).
These results provide a basis to proceed with structural modeling by demonstrating that the construction of each measurement model provides a good reflection of the underlying data in each study, that the measurement items converge on the appropriate first-order factors, and that the test of discriminant validity demonstrates that Social Capital and Governance Effectiveness are distinct, but related constructs.
Structural Modeling Results
To examine our hypotheses, we model our control variables, our predictors, and hypothesized interaction, standardizing the interaction terms before adding them to the structural models to improve interpretability of moderation effects (Aiken & West, 1991). Structural models provide an excellent fit for the data, estimating a squared multiple correlation of 73.6% for Governance Effectiveness (Goodness of Fit [GFI] = .862) in Study 1 (Table 4: χ2 = 584.934, df = 381, χ2/df = 1.535; TLI = .948; CFI = .947; SRMR = .058; RMSEA = .048, pClose = .682) and a squared multiple correlation of 58.5% for Governance Effectiveness (GFI = .899) in Study 2 (Table 5: χ2 = 284.852, df = 156, χ2/df = 1.826; TLI = .936; CFI = .947; SRMR = .061; RMSEA = .058, pClose = .109).
Study 1: Structural Model Statistics With Standardized Estimates.
Note. TLI = Tucker–Lewis index; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.
Study 2: Structural Model Statistics With Standardized Estimates.
Note. TLI = Tucker–Lewis index; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.
We find evidence in support of a positive association between Social Capital and Governance Effectiveness (H1a—Study 1: β = .824, p < .001; Study 2: β = .739, p < .001), but no support for a direct effect of Board Diversity (H1b—Study 1: β = .014, p = .749; Study 2: β = .007, p = .895) or for an interaction to support a moderation hypothesis that Board Diversity attenuates this relationship (H2—Study 1: β = −.030, p = .496; Study 2: β = .068, p = .182). Although our structural models reflect the underlying patterns in our data, we find statistical support for only one of our hypothesized relationships, a positive main effect of Social Capital for Governance Effectiveness, with no association found for Board Diversity.
Whereas our theorizing suggests homogeneous groups would differ from more heterogeneous groups in terms of Social Capital (information sharing, shared vision, and trust), we find no support for this argument.
A Secondary Analysis Probing for Nonlinear Relationships
To ensure that we had not overlooked the presence of a more complex relationship among the variables by limiting analysis to linearity rather than probing for underlying complexity (Aiken & West, 1991), a post hoc analysis was undertaken to examine potential curvilinear relationships lying latent in the data that might be otherwise overlooked (Fredette & Bernstein, 2019). We evaluated potential curvilinear associations in our data by creating a quadratic main effect term for Board Diversity (Board Diversity2) and a standardized higher order interaction term (Social Capital × Board Diversity2), repeating our moderation tests with main effect and interaction terms.
Structural models provide a good estimation of the data, estimating a squared multiple correlation of 75.7% for Governance Effectiveness (GFI = .854) in Study 1 (Table 6: χ2 = 666.454, df = 439, χ2/df = 1.518; TLI = .945; CFI = .954; SRMR = .062; RMSEA = .047, pClose = .754) and retained a squared multiple correlation of 58.5% for Governance Effectiveness (GFI = .895) in Study 2 (Table 7: χ2 = 329.756, df = 192, χ2/df = 1.717; TLI = .953; CFI = .961; SRMR = .060; RMSEA = .054, pClose = .246).
Study 1: Structural Model Statistics With Standardized Estimates.
Note. TLI = Tucker–Lewis index; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.
Study 2: Structural Model Statistics With Standardized Estimates.
Note. TLI = Tucker–Lewis index; CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.
Our results are inconsistent in their support for a nonlinear interaction effect, with support for a quadratic main and moderation effect, in data associated with Study 1 (Table 6). Table 6 illustrates separate positive main effects for Social Capital (β = .844, p < .001) and Board Diversity2 (β = .272, p = .027), with the interaction reaching significance (Social Capital × Board Diversity2 β = −.364, p = .007). Data associated with Study 2 showed no such effect. Plot 1 illustrates the standardized interaction (Social Capital × Board Diversity2) in Study 1 with lines representing mean and ±1 standard deviation values.
Study 1: Interaction of Social Capital and Board Diversity2 for Governance Effectiveness (standardized).
Discussion
This study examines governing-group effectiveness, testing linear, nonlinear, and interacting predictions through the lens of demographic variety and Social Capital. This is unique, not only because group diversity and Social Capital rarely get examined together in the consequential and complex context of governance activities but also because few studies have assessed the potential influence of Board Diversity via an attenuation of Social Capital (Labianca & Brass, 2006; Tasheva & Hillman, 2019). We find little evidence of a linear association that Board Diversity either helps or harms the social or performative success of their board in either our broader national or more-concentrated regional sample of organizations. In one case, a nonlinear interaction suggesting a more complex, albeit attenuating, association was present. For these reasons, our findings are academically and practically significant as we might infer from our analysis that Social Capital, with its benefits of information reach, collective purpose, and interpersonal trust, is more consistently associated with positive governing outcomes, demonstrating a persistent positive linear association. Board Diversity does not demonstrate a similar linear association in either of our studies, nor does it appear to interact with Social Capital in its linear transformation. Alternatively, when examined using a quadratic transformation, Board Diversity demonstrates a positive main effect on Governance Effectiveness, superseded by an interaction with Social Capital, such that more diverse boards exhibit higher performance than less-diverse ones (Table 6), where the greatest performance difference among more diverse boards and their less-diverse counterparts is reported at lower levels of Social Capital. This interaction suggests positive performance benefits for more diverse boards across the range of Social Capital observation, albeit at an attenuated rate of incline. Could it be that when Social Capital is too strong, the performance benefits of diversity are less able to take root or are crowded out as truces set in around shared patterns of belief, practice, routine, and habit? In comparison, in boards exhibiting lower Social Capital, is there more room for contestation, adjustment, improvisation, and revising patterns of governing? This is an important area of future inquiry warranting examination with alternate methods of study.
Our investigation points back to the value of social resources in understanding governance as an inherently socially complex activity or capability (Collis, 1994), shaped by the composition of boards. Nelson and Winter (1982) hinted at the importance of truce-coordinated collective action, recognizing the social and relational underpinnings of collective performance emphasizing the need for stability among participants engaged in highly interdependent decision-making activities. In this case, consequential governance decisions such as those dealing with strategic planning, directing and evaluating executive action, or allocating capital and resources are often contested by group members with differing interests (Kaplan, 2015). When overcoming differences among members, Social Capital is built, and truces develop as points of convergence in position or understanding until a future resetting of the balance of power occurs (Zbaracki & Bergen, 2010). In our view, altering the composition of governing groups is consequential to understanding board Social Capital and therefore the fragility or durability of governing truces, as these decisions tend to be subject more to consensus-building and majority voting rules than to conventional managerial fiat.
Our post hoc analyses speak to the complex relationship among composition, social dynamics, and performance outcomes, particularly as Board Diversity reaches compositional thresholds or tipping points (Fredette & Bernstein, 2019). Nonprofit organizations face external pressures from government mandates and funders to respond to new demographic characteristics of many communities (Bradshaw & Fredette, 2013). Moving toward diverse, inclusive, and equitable governing bodies might reasonably result in the renegotiation, and perhaps dissolution, of prior truces resulting from the reordering and redistribution of decision-making power among diversely composed boards that include a greater number of members of traditionally marginalized communities.
If we consider the underlying difference in samples, the first more nationally representative where the population self-identifies as approximately 84% White and the second drawn four years later from a large multicultural urban environment where approximately half the population self-identify as White, it may be that the salience of demographic differences are more contextually sensitive in some environments and less so in others. Perhaps the passage of time and the effects of locality inform these sensitivities. In this respect, we note the relevance of time, in the form of organization age, as predictive of Governance Effectiveness. Although we do not test for associations among organization age and Board Diversity or Social Capital, we do note that age is a significant control in Study 1 and a near-significant control in Study 2, which is consistent with the suggestions of having reached a state of organizational formalization posited in the work of others (Bradshaw & Fredette, 2013).
Limitations and Implications for Future Research
An avenue for extending this research lies in examining the constituent elements of Social Capital and their relationship to Governance Effectiveness vis-à-vis diversity. In addition, replications of our findings in other localities would overcome the limitations associated with using only Canadian data. We acknowledge potential limitations presented by common method and social desirability biases in our studies, which to the degree possible, we have sought to address statistically. Combining responses from Chief Executives and board chairpersons may present a limitation given their differences in perception of board governance (Bernstein et al., 2016). Future studies could explore other regions and examine whether CEOs and board chairpersons’ responses were significantly different. Similarly, the implications of organizational sector warrant further consideration, particularly in samples that exclude social services and health care emphasis such as our second study, where participation in arts or environmental and recreational or sporting organizations may be perceived as a luxury and historically exclusive, whether by intention or not.
Another limitation of this study is that we define Board Diversity very narrowly, in terms of ethno-racial demography. By focusing only on visible diversity, we presume that racial/ethnic differences would necessarily imply different cultural and social approaches to governing. Further research should be conducted to tease out these lesser-studied facets of diversity and their intersections in similar leadership and governance contexts.
We are optimistic that the ethno-racial diversity of the board members was well known by the survey taker but note a potential limitation in having a single survey respondent reporting on the ethno-racial diversity of the other board members. We make no claims of causality given the design of our research and note the need for longitudinal examination of our hypotheses before such claims can be made. On a practical level, our findings suggest a complex relationship among Board Diversity, Social Capital, and Governance Effectiveness that warrants care, attention, and active management.
Implications for Practice
Our approach to replication and generalizability across a range of subsectors, organizations, and points in time demonstrate consistencies in our first-order associations if not our nonlinear ones. We encourage practitioners and policy makers to consider the interplay between Board Diversity or composition and Social Capital and governing effectiveness, but not to be overwhelmed by anxieties that improving representational diversity will undermine the performance benefits of healthy social dynamics. This is particularly relevant for those engaged in shaping the diversity of leadership teams and governing groups as they work on issues of executive or board succession planning, struggle to enact improved stakeholder outreach and integration strategies, or wrestle with retaining relevance and legitimacy in the face of social and economic reformation.
Footnotes
Appendix
Survey Items of Dependent and Independent Variable Measurement.
| Dependent variable: Governance Effectiveness | |
| All in all, how satisfied are you with your board’s performance? For each of the following statements, please write in the number that comes closest to your opinion. | |
| Study 1: 1 (Totally dissatisfied) to 10 (Completely satisfied in every way) | |
| Study 2: 1 (not at all) to 5 (a very great extent) | |
| GE_1 a | Overall board effectiveness |
| GE_2 a | Fiduciary and financial oversight |
| GE_3 a | Safeguarding and fulfilling the mission of the organization |
| GE_4 a | Providing regular feedback on the performance of the Chief Executive Officer or Executive Director |
| GE_5 a | Ensuring that strategic planning takes place annually |
| Independent variable: Social Capital | |
| Still thinking about how strongly you agree or disagree with the following statements about how board members INTERACT with one another, please write the answer that best represents your opinion in the space provided. | |
| Study 1: 1 (Strongly disagree) to 7 (Strongly agree) | |
| Study 2: 1 (not at all) to 5 (a very great extent) | |
| Inf_Shr_1 a | The board members engage in open and honest communication with one another |
| Inf_Shr_2 a | The board members at this organization have no hidden agendas or issues |
| Inf_Shr_3 | The board members share and accept constructive criticisms without making it personal |
| Independent variable: Social Capital | |
| Inf_Shr_4 | The board members discuss personal issues if they affect board performance |
| Inf_Shr_5 a | The board members willingly share information with one another |
| Inf_Shr_6 | The board members keep each other informed at all times |
| Shr_Vis_1 | Each board member shares the same ambition and vision as other members of the board |
| Shr_Vis_2 a | People in our board are enthusiastic about pursuing the collective goals and mission of the whole organization |
| Shr_Vis_3 a | There is a commonality of purpose in the board of my organization |
| Shr_Vis_4 a | All board members are committed to the goals of this organization |
| Shr_Vis_5 | Board members view themselves as partners in charting the direction of the organization |
| Shr_Vis_6 | There is a total agreement on our organizational vision across all members of the board |
| Trust_1 a | Each member can rely on the others they work with in this board |
| Trust_2 a | The board members in this organization are usually considerate of one another’s feelings |
| Trust_3 a | Board members have confidence in one another at this organization |
| Trust_4 | Board members in this organization show a great degree of integrity |
| Trust_5 | There is no “team spirit” among board members in this organization |
| Trust_6 | Overall, the board members at this organization are trustworthy |
Items included in Study 2.
Acknowledgements
The authors thank the many thoughtful participants of a mid-morning session of the Governance section of ARNOVA 2018 and Julie Ann Israel for insightful comments and feedback. This research is a collaborative effort resulting from the equal contribution of both authors.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We gratefully acknowledge the support of The Maytree Foundation, CivicAction, and the Social Science and Humanities Research Council of Canada.
