5.3.4 Application of moderators

If variation between studies exists, moderator variables might be able to explain the differences between studies. In order to find the impact of these moderator variables, the total sample is split into relevant subgroups, which have a characteristic in common. An analysis of literature in the previous chapter showed two interesting factors, which are used as possible moderators in the meta-analysis: geographic focus of the studies and the nature of the performance measure.

Bell et al. (2010) conjectured and found evidence to substantiate that the relationship in the United States between board independence and performance is negative, relative to the United Kingdom. Since the United States is rule-based with strong supervisors, investors have more confidence in the healthy corporate governance practices of American companies than in their British counterparts. Therefore, investors are prepared to pay a premium for good governance in the United Kingdom, whereas this is less necessary in the United States. Hence, the moderator United States versus non-United States is especially expected to influence market-based performance measures. A distinction between performance measures in general – accounting-based versus market-based – might have influence as a moderator too. Although board independence does not have a direct impact on performance, investors might have more confidence in companies with healthy corporate governance and value these companies higher than what would be expected based on their accounting results. Furthermore, Cornett et al. (2008) demonstrate that the impact of board independence is stronger for accounting results that are corrected for earnings management. Market-based performance measures do not suffer from earnings management and therefore the distinction between market-based versus accounting-based might be a promising moderator in this field.

Dalton et al. (1998), Wagner et al. (1998) and Rhoades et al. (2000) have included the type of performance measure as moderator as well, but they have also investigated the impact of the type of board composition measure as moderator variable. This meta-analysis also applies this moderator, but in this respect also focuses on the definition of independence applied in the study. The distinction of Dalton et al. between large versus small companies is not made, because the absolute majority of companies is large.

The total sample of studies consists of articles with different performance measures, different geographical focus areas, and different definitions of independence. These differences are dealt with by applying the attenuation factor, because they might distort the results. Besides calculating a weighted average of all the correlations over all the articles in the sample, meta-analytic procedures are used to investigate whether the relationship between board independence and performance measure depends on other factors. To test whether the differences between the moderators are significant, the Fisher transformations of the average sample correlations are used in the following formula to calculate the critical ratio (Wolf 1986: 36):