5.4.4 Determining the sample size in the underlying research

Quantitative research has very clear probability sampling techniques. Nevertheless, those techniques are rarely appropriate when conducting qualitative research. Qualitative research has its own techniques such as convenience sampling, purposeful sampling or theoretical sampling (Marshall 1996, p. 522) (Boeije 2005, p. 50). The aim of quantitative sampling is to draw a representative sample from the population to such an extent that the results of the sample can be generalised back to the population; the size of the sample is determined by the optimum number necessary to enable valid inferences to be made about the population (Marshall 1996, p. 522). One wants to test pre-determined hypotheses and produce generalisable results. In qualitative studies the aim is to provide illumination and understanding of complex issues (how is the comply or explain principle applied and why in that specific manner?), therefore qualitative research has its own sampling techniques. However, as is the case for quantitative research, the sample size also matters. An appropriate sample size for qualitative studies is one that adequately answers the research questions (Marshall 1996, p. 523). The sample must be large enough to ensure that almost all or the most possible variation in the data is likely to be found. Furthermore, the sample size ought to be large enough to study the data more in depth. Usually the required number to select becomes obvious during the study.

A sample size is adequate when the data collected can be studied in depth and has the most possible variation. In the underlying study the required size of the sample became obvious during the study; since, as a pilot study, all the Dutch companies quoted on the three main indexes were first researched for the period 2004-2006.1 From this large amount of collected data it could be derived that, on average, five best practice provisions of the Dutch corporate governance code were not complied with per company, which meant that the dataset of the pilot study consisted of 1129 deviations (400 deviations in 2004, 362 deviations in 2005, 367 deviations in 2006). For the current study the average number of deviations is three per company and the total number of deviations collected is 2672 deviations, which is presumed to be a sufficient dataset to answer the research questions. With this large amount of data - 50 companies per country for three years2 - in-depth research is possible and the most possible variation likely. In previous studies fixed datasets from rating agencies were often used or only one country was subject to research (Vander Bauwhede and Willekens 2008, p. 104), which makes determining the sample size unnecessary. This is not the case in this study. An important consideration for taking a sample size of 50 companies per country involved the fact that, in literature on qualitative research, a sample size of 50 is considered sufficient (Sandelowski 1995, p. 182). Moreover, it is believed that, with a sample size of 50 per country, the point of saturation is reached, i.e. "a point of diminishing return where increasing the sample no longer contributes new evidence" (Barbour 2008, p. 54) (Boeije 2005, p. 52).

The previous research of the Dutch dataset revealed that there was very little variation in the best practice provisions not complied with since the top 10 of the deviations, compared to all the deviations, were between 49% and 62%.3 It is expected that in the other four countries quite a fixed set of provisions in the national code will often not be complied with as well, as can already be seen in previous studies (Huse 2005, p. 486) (MacNeil and Li 2006, p. 486) (Von Werder, Talaulicar et al. 2005, p. 185). Therefore, once again, a sample size of 50 companies per country for a period of three years is considered sufficient.