Impact of Directors’ Network on Corporate Social Responsibility Disclosure: Evidence from China

Abstract

Using listed firms in China over the period 2010–2018, we investigate the association between directors’ network and quality of corporate social responsibility (CSR) disclosure from the lens of resource-based view. We find a significantly positive effect of directors’ network centrality on the CSR disclosure quality, and the effect is more pronounced when the firm (1) invests less in advertising; (2) is followed by less analysts; (3) is less financially constrained; and (4) has no assurance of sustainability report. Furthermore, we document that independent directors’ network centrality is positively associated with CSR disclosure quality. Our findings have important implication for practitioners, policy makers, and regulators.

Appendices

Appendix A

See Table

Table 16 Variable Definitions

16

Appendix B Details of Centrality Measures and Process of Generating Networking Data

Details of Centrality Measures

Following literature (Freeman, 1978; Larcker et al., 2013; Wasserman & Faust, 1994), we focus on the three commonly used measures of centrality: degree, betweenness, and closeness centrality. We describe each dimension conceptually and how they are measured as follows (Freeman, 1978; Larcker et al., 2013; Wasserman & Faust, 1994):

• (1) Degree Centrality

  If a director possesses relatively many channels of communication or resource exchange, this director is well-connected. This concept is measured by degree centrality, which assumes that directors know each other if they sit on the same board of a firm (Freeman, 1978; Larcker et al., 2013; Wasserman & Faust, 1994). Degree centrality is computed as follows:

  $$Degree_{i} = \frac{{\Sigma _{j} X_{{ij}} }}{{g - 1}},$$

  (1)

  where $g$ is the total number of directors of all the listed firms in the same year, i is a director, j is a director other than i. If directors i and j both serve on the same board, ${X}_{ij}$ is 1, and otherwise 0.

• (2) Betweenness Centrality

  If a director lies on relatively more paths between pairs of other directors, which is vital in connecting directors to each other and a key broker of information or resource exchange, this director is well-connected. This concept is measured by betweenness centrality, which represents how important a director is in connecting other directors to each other. Betweenness centrality is defined to be the average proportion of paths between two directors on which a director lies (Freeman, 1978; Larcker et al., 2013; Wasserman & Faust, 1994). Betweenness centrality is computed as follows:

  $$Betweenness_{i} = \frac{{\Sigma _{{j < k}}\,g_{{jk(n_{i} )}}/g_{{jk}} }}{{(g - 1)(g - 2)/2}},$$

  (2)

  where i, j, and k represent different directors. ${g}_{jk}$ is the total number of shortest paths between director k and director j. ${g}_{jk({n}_{i})}$ denotes the total number of paths that point i falls on the shortest distance line connecting j and k.

• (3) Closeness Centrality

  If a director possesses relatively closer ties to other directors, making information or resource exchange quicker, this director is well-connected. This concept of connectedness is measured by closeness centrality, which represents how easily or quickly a director can reach other directors. It is defined as the inverse of the average distance between a director and any other director (Freeman, 1978; Larcker et al., 2013; Wasserman & Faust, 1994). Closeness centrality is computed as follows:

  $$Closeness_{i} = \left[ {\frac{{\sum\nolimits_{{k = 1}}^{g} {d(i,k)} }}{{g - 1}}} \right]^{{ - 1}},$$

  (3)

  where i refers to a given director and k is all the directors other than i in the same year. d(i, k) is the number of steps in the shortest path between director i and director k.

Process of Generating Networking Data through Software

To map the directors’ network, for each annual volume of the data during the year 2010–2018, we construct the entire boardroom network and compute each of the three centrality measures (degree, betweenness, and closeness centrality) for every firm. We obtain information on board of directors from CSMAR and generate data through network analysis software Pajeck. The sample contains 93,008 directors and 432,786 director-year observations. Each director is assigned a unique identifier that was used in the measurement of centrality.

Following prior literature (Freeman, 1978; Goergen et al., 2019; Homroy & Slechten, 2017; Larcker et al., 2013; Omer et al., 2019; Wasserman & Faust, 1994), we construct our annual networks based on the individual director’s board memberships. For each year, two directors are connected if they sit on at least one board (Omer et al., 2019). We use Stata’s “Stata2Pajek” command to transfer the data into a network of “company–director” that Pajek can identify. We use Pajek to transfer the network of “company–director” into a network of “director–director,” and then generate the value of three centrality measures.

Larger firms tend to have better-networked boards, giving rise to a mechanical positive relationship between firm size and board connectedness (Larcker et al., 2013). To separate the effects of size and board connectedness on CSR disclosure quality, we take ranked versions of the centrality measures that attempt to purge the “size” effect (Larcker et al., 2013). Specifically, in each year, all firms are ranked into tenths based on firm size. Within each size tenths, firms are sorted into tenths based on the maximum (Score_max), mean (Score_mean), median (Score_med), and minimum (Score_min) values of the three centrality measures of each firm, where highest (lowest) values of centrality assume a value of ten (one). The centrality measures are calculated at the director level.

To obtain centrality measures at the board-level, we calculated the sum of the maximum (Score_max), mean (Score_mean), median (Score_med), and minimum (Score_min) values of the three centrality measures of each firm as independent variables.