5.4.5 The sample composition

Within qualitative research various sampling techniques exist, such as convenience sampling, judgment sampling, also called purposeful sampling, and the second stage of purposeful sampling called theoretical sampling (Barbour 2008, p. 53). In this study the most common sampling technique is used, which is purposeful sampling: "the researcher actively selects the most productive sample to answer the research question. This can involve developing a framework of the variables (...) based on the researcher's practical knowledge of the area, the available literature and evidence from the study itself' (Marshall 1996, p. 523).

The characteristics of the population determine the selection, within which the different manifestations of the phenomenon must be present (phenomenal variation) (Sandelowski 1995, p. 181) (Boeije 2005, p. 50). The framework of the variables - or in other words the characteristics of the population in this study - involves i.a. the specific year, index and market capitalisation. These variables are of influence when selecting the sample as explained further below. By selecting companies with these variables probably the largest variation in the data can be found without losing the ability to spot trends and developments and the ability to illuminate patterns, concepts, categories and dimensions of the phenomenon (the point of saturation). In addition to purposeful sampling, theoretical sampling is also used. Theoretical sampling involves 'second-stage' sampling: based on previous research and sampling, the researcher performs purposeful sampling for the remainder of his study, taking the results of the previous research into account in selecting the sample (Boeije 2005, p. 51). Hence, based on literature and the study previously performed on the Dutch companies, it was decided to take the year, index and market capitalisation into account in sampling for the underlying study.

This study chooses for a period ofthree years to detect the possible trends and developments in the application of the comply or explain principle. The years under research are 2005, 2006 and 2007, since the comply or explain principle was applied in all the countries under research in these financial years.1,2 This period is of extra interest due to the fact that during this period Directive 2006/ 46/EC, making the comply or explain principle mandatory for listed companies, was adopted by the EU Parliament and EU Council. Outdated research data is thus not a direct threat, since other (national) studies on code compliance with more recent data come to similar conclusions, although only based on one country and often one year (see section 5.2 and in chapter 4 the eighth key question per country). A code compliance study with a research period of longer than one year and with more than one country under research has not been performed before. Because this research is mainly focused on and performed in the Netherlands, all the companies listed on the three main indexes in the Netherlands are reviewed for the period 2004-2007. The year 2004 was the first year that the principle was implemented and therefore the period 2004 to 2007 provides a good overview of the developments in the application in the Netherlands. It is argued - based on the results of the previous research involving the Netherlands for the period 2004-2006 (the pilot study) -that a research period of three years is adequate since in-depth research with the most possible variation is expected. The 50 companies selected per country are all listed on the three most important stock exchanges, since the comply or explain principle is only applicable to listed companies.3 Other studies and literature on corporate governance focus on the main indexes as well and in the underlying study the aim is to incorporate those studies and literature in this research to contribute further to the research on corporate governance. Table 5.4.5 below shows the stock exchanges on which the companies under research are listed as well as the number of annual accounts reviewed.

For the purpose of making relevant comparisons within the collected data per year and per index, the annual accounts analysed concern companies listed for three consecutive years on the same index. Therefore, the items in the dataset are comparable with each other in a year-based comparison, which is not the case for an index-based comparison. Since each index has different admission criteria, one cannot for example compare the results of the AMX index with the FTSE250, the companies are therefore distinguished based on their market capitalisation5in line with Euronext Notice N°6-02 of 23 March 2005. In this Euronext Notice three compartments are distinguished: the large caps (compartment A), the mid caps (compartment B) and small caps (compartment C). Therefore, article 2 states that:

The companies to be analysed have the same market capitalisation compartment in all three years under research, which is calculated on the same date each year.6 By doing so, the companies sampled can be compared country-wise and year-wise per compartment for trends/developments. Another possibility would have been to select, per country, the 50 companies with the largest market capitalisation since a scandal involving these companies probably has the largest impact. However, the market capitalisation of the 50 largest companies differs greatly between the countries and would bias the dataset. For instance, the market capitalisation of BP, the largest UK company selected, with a market capitalisation of EUR 156,639,200,000 in 2007, is four times as large as the market capitalisation of Fortis, the largest Belgian company in 2007 (EUR 39,827,910,000). Because of these large differences in size the fifty largest companies per country cannot be compared with each other, which is possible7 when distinguishing per compartment and thereafter comparing per compartment (and not per company) the same number of companies per country.

To summarise, for reasons of comparability companies are selected per country whenever they show a stock quotation of three consecutive years on the same index within the same market capitalisation compartment. Of the companies that remain in the sample after this selection,8 50 companies are selected per country. They are first arranged per market capitalisation compartment and then sampled at random (stratified random sampling), as promoted in the research of Short and Ketchen et al. (Short, Ketchen et al. 2002, p. 382). Per compartment a number of companies are selected with - if possible - 25 compartment A companies, 15 compartment B companies and 10 compartment C companies. This partition has been opted for since - as explained above - for the largest companies (compartment A companies) good corporate governance and application of the comply or explain principle is essential to prevent further corporate scandals. For some countries, such as Germany and the Netherlands the dataset does not contain 10 compartment C companies, therefore more A and B companies are selected. If possible first an A company is selected if there are insufficient C companies because of the importance of good corporate governance in large companies and the considerable impact of a scandal. In the Netherlands only 37 companies were quoted on the same stock exchange (AEX, AMX or AMS) for three consecutive years. Apparently, compared to the other countries, there are many shifts within the companies with a stock quotation in the Netherlands and therefore only 37 companies could be selected for the country-wise comparisons. Only in the case of the UK, Italy and Belgium can the C companies be compared 'country-wise' for market capitalisation (see table 5.4.5a below and Annex I for an overview of all the companies arranged by name).9

A criterion for the companies to be selected was that the annual accounts of 2005 to 2007 (for the Netherlands 2004 to 2007) had been published.10 Since this was not the case for a few companies, these were removed from the dataset and randomly replaced by others. In several studies on corporate governance the financial firms are excluded from the dataset since they had a different company structure or the regulatory environment differs (Arcot and Bruno 2006, p. 9) (Goncharov, Werner et al. 2006, p. 435), which might interact with the application of the comply or explain principle (Arcot and Bruno 2007, p. 11). Although acknowledging these differences, this study also includes financial firms, as some other studies do (Akkermans, Van Ees et al. 2006) (Otten 2007). All five national corporate governance codes under review are applicable to financial firms. Moreover, financial firms are often large firms or banks for which a good corporate governance practice is especially of importance because of the many stakeholders having an interest in the well-being of the company. There are few reasons not to select financial firms and more reasons to select and research them further, since others do not.

To conclude, in total 898 annual accounts are reviewed of 294 different companies (4 times 50 plus 94 NL companies gives a total of 294 companies). As stated before, because this research is mainly focused on and performed in the Netherlands, all the companies listed on the three main indexes in the Netherlands are reviewed for the period 2004-2007. For the four remaining countries 50 companies are selected that have a stock quotation for three consecutive years on the same index within the same market capitalization compartment. Although reviewed, the extra Dutch companies under research are solely part of the pilot study and for the reason given below not incorporated in the underlying research results. In comparing the countries as regards year, index and market capitalisation, as explained above, not all the Dutch companies can be taken into account, but only the 37 companies selected, which has resulted in 237 companies researched for comparison purposes (see table 5.4.5a). In the remainder of this study it is indicated each time that the study involves the 37 companies selected for comparison purposes ('NL (selected)'). In the sections 5.5 (univariate results), 5.6 (bivariate results) and 5.7 (regression models) only the Dutch selected companies are part of the results.