12.3.4 Cluster analysis

The results are used to perform a cluster analysis. This analysis is performed to investigate whether people with comparable answers to the questions have common characteristics. The results from the Ward’s hierarchical cluster analysis show two clusters, with 288 and 129 respondents respectively. The answers to the questions of both clusters are provided in Table 12-5. Shareholders are more important for the respondents of cluster one than for the respondents of cluster two. 66.7% of the respondents in cluster one regard monitoring on behalf of shareholders as a task of the supervisory board on average, whereas only 30.2% of cluster two thinks that the supervisory board should look after the interests of shareholders. The difference between these two scores is significant on all levels. This is not the case for employees, suppliers, customers, the state, society and debt providers. Relatively more respondents of cluster two than the respondents of cluster one look after the interests of these particular stakeholders. The differences are significant at all levels of significance, with the exception of debt providers (p = 0.057).

Table 12-5: Results from the cluster analysis. The group of respondents is divided into two clusters based on their answers to the questions regarding whose interests should be looked after and which suggested potential threats to independence are a real threat to independence. Differences between the clusters were tested with an F-test. The P-values are given in the last column.

A closer look at the threats to independence shows that cluster one is more concerned about share ownership than the respondents in cluster two (significant at a 5% level). 50.7% of cluster one and 38.0% of cluster two regard the possession of any share as a threat to independent supervision. The difference is significant at a level of 5%. Share ownership that does not exceed a certain level is a greater concern for cluster two (16.3%) than for cluster one (9.4%). A long tenure, a business relationship, a position as former executive, and family or social relationships are, for 40% or more of the respondents of both groups, a threat. Cluster two is more concerned about these relationships than cluster one. The differences are significant at all levels.

Although social relationships are conjectured in this study to be a very important threat to independence, it is not the most important threat according to the respondents. In both clusters, business and family relationships are regarded to be a larger obstacle to independent supervision. But cluster two is significantly more concerned about social relationships than cluster one.

Table 12-6: Results from the discriminant analysis. The table gives the characteristics of the two clusters from Table 12-5. Age, gender, experience, the age at the time of the first supervisory board position, education, background, number of supervisory board positions, time spent on these positions, remuneration, type of organisation, number of employees and characteristics of their supervisory board, such as independence, number of meetings and the subcommittees. Differences between the clusters were tested with an F-test. The P-values are given in the last column.* Significant at a 90% confidence level, ** 95%, *** 99%.

As the two clusters differ as regards the answers to the above-mentioned questions, the differences in characteristics are described here. The results are given in Table 12-6. The table shows that the respondents in cluster one are slightly older (59.3 years) in comparison to the respondents of cluster two (57.9 years). This difference is significant at a 10% level. Cluster one also has significantly more male respondents (82.9 % versus 72.9%) than cluster two. The respondents in cluster one accepted their first supervisory board position 2.5 years later than the respondents in cluster two. This result is significant at a 1% level. Another significant difference is the percentage of respondents that receive less than 5,000 euros per year for their position. In cluster two this percentage is almost twice as high as in cluster one (27.9% versus 14.2%). Another difference is that the respondents of cluster one are more active in non-listed and family-owned companies, while the respondents of cluster two are more active in healthcare institutions and housing associations. Thus the respondents of cluster two are more active as supervisory directors in nonprofit organisations, while the respondents in cluster one are significantly more active in profit organisations. This fact can also be linked to lower remuneration, because supervisors in non-profit organisations are, on average, paid less than in profit organisations.

From the table it can also be derived that the respondents in cluster two operate on a board that is significantly more independent than the boards of the respondents in cluster one (90.0% versus 78.5%). A closer look at the presence of board committees shows that the respondents in cluster two operate on supervisory boards with significantly more audit and remuneration committees than the respondents in cluster one, which might indicate higher independence. The higher independence and the presence of more board committees might indicate that the respondents in cluster two are more aware of good governance.

Although both clusters differ as regards certain characteristics, they also have some similarities. Supervisory directors in both clusters do not differ significantly with respect to experience, measured in numbers of years of experience. The same holds for their educational background, number of positions and time spent, which do not differ significantly. Earlier in this subsection it was mentioned that remunerations of less than 5,000 euros are observed more often in cluster two than cluster one. Remuneration between 10,000 and 25,000 euros is observed more in cluster one. This difference is significant as well. The other remuneration categories are not observed more often in one cluster in comparison to the other cluster. Although the occurrence of non-listed and family-owned companies is higher in cluster one, and the occurrence of healthcare institutions and housing associations is higher in cluster two, the occurrence of listed companies does not differ between the two clusters. The relative number of listed companies in each cluster is the same. Finally, the size – in terms of number of employees – does not differ significantly either.

Therefore, in short, cluster one is more inclined to monitor for shareholders and cluster two is more inclined to monitor for the other stakeholders. Respondents of cluster two are also more concerned about independence (i.e. they regard more relationships as a threat to independence) and operate on a board that has a higher percentage of independent members and has more board committees. Cluster two is also significantly more concerned about social relationships than cluster one. With respect to the characteristics it can be concluded that cluster two is significantly younger and consists of more females and relatively more internal supervisors that earn less than 5,000 euros. This can also be explained by the fact that these three variables – percentage male, age and remuneration below 5,000 euros – are significantly correlated with each other, as becomes clear from Table 12-8 in the Appendix. However, the occurrence of listed companies in both clusters is relatively equal and the organisations in both clusters are not significantly different with respect to size.

Table 12-7: The answers for profit and non-profit organisations to the questions regarding whose interests should be looked after and which suggested potential threats to independence are a real threat to independence. The group of respondents is divided into two groups of internal supervisors: internal supervisors in profit and non-profit organisations. Differences between the two groups were tested with an F-test. The P-values are given in the last column. * Significant at a 90% confidence level, ** 95%, *** 99%.

In the introduction it was conjectured that the attitude towards independence would be different between respondents from non-profit and profit organisations. To research whether there is a difference between those two sectors, the differences in answers between these two groups were analysed. The results are given in Table 12-7. A similar pattern can be observed. The non-profit sector, which is larger in our sample than the profit sector (221 versus 196 respondents) is more inclined to monitor for suppliers, customers, the state and society, while the respondents in the profit sector consider monitoring for shareholders, employees and debt providers to be of higher importance. Employees were expected to be of higher importance to the non-profit supervisory directors, but the survey gives a different result. These results are all significant.

With respect to the threats of independence, the difference between the groups in their answer to the question whether shareholding is a threat to independence is not significant. However, this question is not relevant for the comparison, since non-profit organisations do not have shareholders. A long tenure, a business, family or social relationship and a position as former executive is a larger threat for the respondents from the non-profit sector than it is for the respondents from the profit sector. This greater concern about independence for supervisory directors in non-profit organisations might be caused by the fact that other mechanisms to keep management in check are smaller in these types of organisations. The absence of shareholders and the threat of takeovers make independent supervision more important as it functions as a substitute.