Boardroom Diversity and its Effect on Social Performance: Conceptualization and Empirical Evidence

Abstract

In this paper, we seek to answer two questions: (1) what does boardroom diversity stand for in the strategic management literature? And, (2) is there a significant relationship between boardroom diversity and corporate social performance. We first clarify the boardroom diversity concept, distinguishing between a structural diversity of boards and a demographic diversity in boards, and then we investigate its possible linkage to social performance in a sample of S&P500 firms. We find a significant relationship between diversity in boards and social performance. This relationship is moderated by diversity of boards. Our results also reveal the effects of the specific variables that make up the diversity of boards and diversity in boards constructs. In particular, gender, and age have a significant effect on corporate social performance. Some important measurement issues are raised and discussed.

Appendices

Appendix 1

Procedure for Diversity of Boards Index (I _DoB)

Recall that, in our data set, each data point (i.e., company board), is represented with four variables, or features, in terms of diversity of boards . We discussed our operationalization of variables in detail in the measures section. Our calculation is, in short, as follows: the first variable is the number of directors sitting in a given board; the second and third variables are the percentage of outside directors of a given board and the percentage of company shares owned by all directors of a given board, respectively; the fourth variable represents whether CEO also acts as chairman or not. As such, our data set is composed of heterogeneous variables, in which the first variable is discrete, the second and third variables are continuous, and the fourth variable is of a dichotomous nature.

Physically, distance is a numerical description of how far apart objects are. In mathematics, distance is a reflection of physical distance. Using a dissimilarity matrix, a collection of proximities that are available for all pairs of n objects can be stored. In measuring inter-sample distances, the distance-measurement tool compares data samples in a matrix and provides a metric to assess how (dis)similar they are. Measured difference, or dissimilarity between two objects, or d(i, j), is a non-negative number that is close to 0 when objects i and j are highly similar or “near” each other, and becomes larger as they differ more. Several measures of inter-sample distances are formulated depending on the types of variables (i.e., binary, nominal, interval-scaled, and ratio-scaled). For instance, the Euclidean distance metric previously used in management studies (e.g., Roth and O’Donnell 1996; Thatcher et al. 2003) measures paired distances only in interval-scaled variables.

However, as discussed earlier, since our data set is made of different types of variables, we use an aggregated distance function that enables us to combine all types of variables in a single dissimilarity matrix, and, hence, assess them together (Han et al. 2011). That is, the dissimilarity between data points can be computed even when the variables describing these data points are of different types. This function is defined by Han et al. as:

$$ d\left( {i,j} \right) = \frac{{\sum\nolimits_{f = 1}^{p} {\delta_{ij}^{\left( f \right)} d_{ij}^{\left( f \right)} } }}{{\sum\nolimits_{f = 1}^{p} {\delta_{ij}^{\left( f \right)} } }} $$

(1)

where i and j are two p-dimensional data points represented as (x _i1, x _i2,…, x _ip ) and (x _j1, x _j2,…, x _jp ) respectively, and d(i, j) is a distance function (metric) used to express the (dis)similarity between two data points (i.e., i and j in this case). Then, the contribution of variable f to the dissimilarity between i and j (i.e., d ^(f) _ij ) is computed dependent on its type:

• 1. If f is binary or nominal: d ^(f) _ij  = 0 if x _if  = x _if , or otherwise d ^(f) _ij  = 1

• 2. If f is interval-scaled: $d_{ij}^{\left( f \right)} = \frac{{\left| {\mathop x\nolimits_{if} - \mathop x\nolimits_{jf} } \right|}}{{\max_{h} x_{hf} - \min_{h} x_{hf} }}$

• 3. If f is ordinal or ratio-scaled: compute ranks r _if and $z_{if} = \frac{{r_{if} - 1}}{{M_{f} - 1}}$, and treat z _if as interval-scaled (r _if  ∈ {1,…, M _f })

In this function, the contribution of all different types of variables to the dissimilarity (i.e., d ^(f) _ij ) are normalized, and hence expressed on a common scale of (0, 1).

In our analysis, we individually compute the distance of each data point (i.e., company board) to all other data points in our data set using the above mentioned metric. Here, we have given equal weights to the relative contributions of each variable to the distance function (i.e., δ ^(f) _ij  = 1). Then, we average the computed distances of each data point to all other data points using the formula below:

$$ d\left( {i,j} \right) = \frac{{\sum\nolimits_{f = 1}^{p} {d_{ij}^{(f)} } }}{s} $$

(2)

And for the average distance to all the other boards we use:

$$ D\left( i \right) = \frac{{\sum\nolimits_{z = 1}^{k} {d\left( {i,z} \right)} }}{k - 1} $$

(3)

where, x _i1, the number of directors sitting in company board i; x _i2, the percentage of outside (non executive) directors of company board i; x _i3, the percentage of company shares owned by all directors of company board i; x _i4, the fact that whether CEO of company board i also acts as chairman or not; s, the number of features representing diversity of boards (i.e., board size, outsiders, ownership and duality); k, the number of company boards; d ^(f) _ij , the distance of company board i to company board j with respect to the variable f; δ ^(f) _ij , the relative contribution of the variable f to the distance between the company board i and the company board j; d(i, j), the distance of company board i to company board j; D(i), the average distance of company board i to all other boards.

The output of this distance-measurement metric provides information on how (dis)similar a given board, taken into consideration four variables at the same time, from all other boards in our sample. This information represents the nature of diversity of boards.

Appendix 2

Calculation of Director Experience

Since we did not have a direct measure that represent director experience, we used committee experience as a proxy and calculated whether, or not, a given director served continuously through three consecutive years in one or more of the four common board committees (nomination, compensation, audit, and governance). We calculated experience using the following formula:

$$ \varepsilon_{k} = \frac{1}{{\sum_{j = 1}^{3} {n_{jk} } }}\sum\limits_{j = 1}^{3} {\sum\limits_{i = 1}^{{n_{jk} }} {\left( {x_{jik} + y_{jik} + z_{jik} + t_{jik} } \right)} } $$

(4)

where, n _jk , number of directors in company board k in year j; x _jik , nomination committee membership of director i in company board k in year j (0: No, 1: Yes); y _jik , compensation committee membership of director i in company board k in year j (0: No, 1: Yes); z _jik , audit committee membership of director i in company board k in year j (0: No, 1: Yes); t _jik , governance committee membership of director i in company board k in year j (0: No, 1: Yes); j, a given year (i.e., 2005, 2004 and 2003); i, a given director (i.e., of all 1,028 directors in our sample); k, a given company board (i.e., of all 95 company boards in our sample); ε _k , overall director experience of company board k.