5.3.2 Sample

In order to collect studies for the meta-analysis, 48 journals have been selected as potential journals in which articles of interest could be found. The list of journals is included in the appendix. Furthermore, working papers published at SSRN are included as potential sources of research. The potential sources were searched electronically by using keywords: outsider, outside, director, independence, independent and performance for the period January 2000 and March 2011. This search yielded approximately 4,000 articles, which have been scanned manually for the existence of a relationship between board independence or board structure and any form of performance measure. For the correlation results the relationship should be a Pearson correlation, but it is not relevant whether a variable is the dependent, independent or control variable in a study. For the regression results it is important that some kind of performance measure is the dependent variable in a study. This requirement is necessary as the influence of board independence on performance is the purpose of this meta-analysis and not the relationship in the other direction. Whether the board independence variable is an independent or control variable is not relevant. Neither does the study need to focus specifically on the relationship between board independence and some measure of performance; studies with other purposes are also included, as long as the size of the relationship is reported. If studies report relationships between board independence and more than one performance measure, all the relationships are filed.

The search and the following filter initially resulted in 57 studies that report at least one correlation (42 studies) or regression relationship (38 studies) between board independence and some measure of performance. The 42 correlation studies have 43 independent samples, which have a total sample size of 52,183 company year observations and 71 reported correlations. The 38 regression studies have 39 independent samples, which have a total sample size of 83,671 company year observations and 60 regression results. It must be noted that it is likely that there is some overlap between the companies in the different samples. However, the studies do not explicitly report the companies included, which results in an inability to measure the degree of overlap exactly. Due to the decision not to include the regression results in a meta-analysis, 15 studies have not been taken into consideration. Details about reported relationships, sample sizes, focus and the definition of independence used are given in Table 5-1 for correlation studies as well as regression studies.

Table 5-1: Overview of the studies included in the meta-analysis. Only correlation results are used in the meta-analysis; due to the many complications that arise when cumulating regression slopes the regression results are left out of the meta-analysis. The suffix i.a. stands for Industry adjusted.