5.3.3 Hypotheses for bivariate analysis

The dependent variable, the level of compliance, has already been clarified above and three levels (I to III) can be distinguished. The independent variables are, as much as necessary, elucidated in more detail under the results and in section 5.4 on the data collection and methodology. The aim of this section is to embed the hypotheses as formulated in literature - i.e. the results of previous studies are quoted to clarify the choice for the specific hypotheses. Further specifics on how the following hypotheses are tested are provided in the subsequent sections; in this section the sole purpose is to clarify why these specific hypotheses were chosen and not the method of testing. Since the underlying research is the first of its kind, not all hypotheses have been researched as such before in previous studies. Therefore, previous research performed by univariate analysis is incorporated as well in the clarification of the underlying hypotheses. The hypotheses1 of the bivariate analysis are formulated as follows:

It is expected that, due to the fact that companies need time to adapt to the rules as laid down in the code and time to implement these, the level of compliance increases during the years in which the comply or explain principle is applicable and hence a positive relation is visible. Previous research on this matter shows various outcomes. Shabbir finds no motivation for the time hypothesis, since in his UK study a varying trend in compliance is visible over time. Moreover he finds that firms "tend to behave in an opportunistic manner so far as their compliance is concerned - becoming more complaint when their prior period stock market performance declines and less so with improvements in their operating performance" and "mergers and acquisitions tend to decrease compliance, whereas reorganizations/restructuring tend to increase compliance" (Shabbir 2008, p.1). Weir and Laing do not statistically research the relation between the level of compliance and time, but conclude that the adoption of the corporate governance structures recommended by the Cadbury code increased over time (Weir and Laing 2001, p. 279). In their review of 245 UK companies for the period 1998-2004 Arcot and Bruno find an annual increase in compliance (Arcot and Bruno 2006). Moreover, in chapter 4 (see section 4.7.2) several nationally conducted compliance studies show an increase over time as well, with sometimes a (temporary) small decrease or a point of saturation has been reached.

As explained before, the manner in which the comply or explain principle is judicially embedded varies per country, although five judicial corporate governance arrangements can be distinguished (pure self-regulation, self-regulation supported by non-statutory norm, self-regulation facilitated by statutory rules, regulation of self-regulation (meta-regulation) and pure regulation) (Voogsgeerd 2006, p. 21) (see chapters 3 and 4). It is expected that the stronger a comply or explain principle is embedded in a country the higher the level of compliance and/or quality of explanations is, since a strong embedding (e.g. the comply or explain principle is laid down in statutory norms) possibly shows faith in and consensus about the principle and, as a consequence, more compliance. Apart from the RiskMetrics Group no other international studies on code compliance (or compliance rates of other regulation related to its juridical embedding) have yet been conducted and can be reviewed here. The RiskMetrics Group only researched 15 companies per country for 2008 and did not relate the compliance results to the manner in which the comply or explain principle is embedded. The compliance results are therefore of interest for comparison purposes (Netherlands 65 deviations, Belgium 42, Italy 24, Germany 19 and the UK 13) (RiskMetrics Group 2009, p. 88), but provide no further support for the above hypothesis.

Hypotheses 3 a and 3b regard firm size, which is considered a common predictor for code compliance; as a result of relatively lower compliance costs (Dedman 2000, p. 134) and greater visibility (Pollock, Fischer et al. 2002, p. 1175) larger firms are expected to have a higher level of code compliance than smaller companies (Talaulicar and Von Werder 2008, p. 268). Several studies confirm that code compliance is positively associated with firm size. Arcot and Bruno show through univariate analysis that firms listed on the main stock index score higher on level of compliance than other companies (Arcot and Bruno 2006). The studies in table 5.3.3 below show various approaches to firm size. The underlying research opts for reasons ofcomparison (see section 5.4.5), for market capitalisation (hypothesis 3a) and for indices (hypothesis 3b) to relate to the levels of compliance.

Several studies show a set of recurring deviations dealing with a fixed specific set of issues, meaning that those issues are poorly complied with. Prior research finds that provisions on remuneration of the management board and its role and procedures are collectively deviated from (Hooghiemstra, Van Ees et al. 2008) (Van der Elst 2006) (Guberna and VBO 2009) (RiskMetrics Group 2009, p. 14) (Talaulicar and Von Werder 2008) (Von Werder, Talaulicar et al. 2005), as well as provisions on the length of the board members' contracts (Arcot and Bruno 2006) and board independence (Seidl, Sanderson et al. 2009). A relation is expected between the level of compliance and the specific issues that are not complied with. Of most interest are the specific issues that reveal the greatest relation with a low level of compliance, since then they typically do occur in the case of low compliance. It is expected that similar issues will be found as in the studies mentioned above.

In research as performed by Hooghiemstra and Van Ees (Hooghiemstra, Van Ees et al. 2008) and in the annual compliance studies of the Dutch monitoring committee since 2006 (Dutch Monitoring Committee 2006), a subdivision in reasons for non-compliance is made to be able to categorise the motives for the deviations from the code provisions. Eleven different categories of reasons for deviations were distinguished (Dutch Monitoring Committee 2006, p. 10). Taking a closer look at these categories and complying with the rules of content analysis (see section 5.4.3) 16 categories can in fact be distinguished, as is done in the underlying study as well (see Annex II). It is expected that a positive relation exists between the reasons provided and the level of compliance, since the higher the number of deviations, the more reasons for deviations occur. However, it is of interest to elaborate further on what kind of reasons provided have the highest relation with the level of compliance, since they typically occur in the case of low compliance.