What determines the informativeness of firms' explanations for deviations from the Dutch corporate governance code?

Abstract

The comply-or-explain principle is a common feature of corporate governance codes. While prior studies investigated compliance with corporate governance codes as well as the effects of compliance on firm behaviour and performance, explanations for deviations from a corporate governance code remain largely unexamined. This paper intends to fill that gap. The paper draws on the voluntary disclosure literature and agency theory to examine the association between firm characteristics and the informativeness of explanations for deviations from the Dutch corporate governance code. Applying content analysis to corporate governance statements for a sample of Dutch listed firms for the period 2005–2009, the study finds that ownership concentration and number of analysts following the firm are positively associated with informativeness. Furthermore, there is indicative evidence that board strength and informativeness are positively associated. The study also finds a negative association between leverage and informativeness. Institutional investors, however, do not seem to affect this type of disclosure. Taken together, the findings suggest that certain firm characteristics are associated with a firm's choice to provide either generic and uninformative explanations or more firm-specific and informative explanations. On the basis of the study's findings, I argue that firms having weaker boards, firms followed by fewer analysts, firms having more dispersed ownership and firms relying more on debt finance tend to approach comply-or-explain more symbolically than substantively.

Keywords:

• agency theory
• comply-or-explain
• content analysis
• corporate governance codes
• voluntary disclosure

Acknowledgements

The author gratefully acknowledges Niels Hermes, Carel Huijgen, Abe de Jong, Teye Marra, Bo Qin, Bob Scapens and Martijn van der Steen for useful comments. Furthermore, the author wishes to thank the anonymous reviewers and the editor for their helpful comments. The usual disclaimer applies.

Notes

Examples of corporate governance codes that contain best practice provisions include the British Combined Code, the German corporate governance code (also known as the Cromme Code) and the Dutch corporate governance code (also known as the Tabaksblat Code).

A recent study on monitoring and enforcement practices in corporate governance in EU member states finds overwhelming support for the comply-or-explain principle from regulators, listed firms and institutional investors (RiskMetrics 2009). However, in the same study there was a wide consensus among the participants that the comply-or-explain principle does not function properly. For instance, that study reported that only a quarter of respondents considered corporate governance-related disclosure in general and explanations for non-compliance in particular to be of sufficient quality.

For example, in 2005 Dutch company law was changed to reflect the requirements of the EU Company Law Directives.

For more information regarding the Dutch corporate governance system see Hooghiemstra and Van Manen (2004), De Jong et al. (2005), Groenewald (2005) and Cools and Van Praag (2007).

Firms that have more than 100 employees and have a (book value of) shareholders' equity in excess of 16 million euros are required to meet the requirements of the Two-Tier Structure Reform Act of 2004.

The fraud in the construction industry in the Netherlands and the discussion about a corporate governance code at an EU level were other important factors contributing to the establishment of the Dutch Corporate Governance Committee.

The first section, ‘compliance with and enforcement of the code’, deals with the provision in the annual report of information on the firm's corporate governance structure. It also prescribes that firms should report on their compliance with the code's provisions on a comply-or-explain basis. The second section, ‘the management board’, covers issues such as the responsibilities of the management board (e.g. for internal control), the determination and disclosure of executive remuneration, and conflicts of interest. The third section, ‘the supervisory board’, covers provisions relating to the composition of the supervisory board, its members' responsibilities (especially the chairman and the company secretary), its independence and the formation of board committees. The fourth section, ‘the shareholders and the general meeting of shareholders’ deals with, among others, the powers of shareholders and the simultaneous provision of information to all shareholders. The fifth section includes provisions regarding financial reporting and the appointment and roles of the internal and external auditors.

German law also refers to a corporate governance code. More specifically, the German Corporate Governance Code, which is incorporated into the Aktiengesetz. Similar to the situation in the Netherlands, German listed firms are required to publish annually a compliance statement in their annual reports (von Werder et al. 2005, Wymeersch 2005, Goncharov et al. 2006). Unlike the Dutch/German mode of enforcement, the British Combined Code is not part of the legal framework. Instead, it is annexed to the Listing Rules (Wymeersch 2005, MacNeil and Li 2006). The consequence, however, is similar and listed firms are required to report on their compliance with the provisions of the Combined Code on a comply-or-explain basis.

One year after the Dutch Corporate Governance Committee presented its final report a committee, better known as the Frijns Committee, was set up to monitor the operation of the Dutch corporate governance code, its implementation by listed firms and to indicate whether there are any gaps or lack of clarity in the Dutch corporate governance code (see www.corpgov.nl).

In the Netherlands, explanations of deviations from the best practice provisions of the Dutch corporate governance code are subject to monitoring by market-wide regulators to only a limited extent (see, e.g. Deumes and Knechel 2008, RiskMetrics 2009). First, the main task of the committee charged with monitoring the Dutch corporate governance code (the Frijns Committee) is to identify areas for improvement that could lead to revisions to the code. Second, the role of the Netherlands Authority for the Financial Markets (Autoriteit Financiële Markten – AFM) is limited to verifying that corporate governance statements are present in the annual reports of listed firms and it does not assess the quality of the information contained in the corporate governance statements.

In a one-tier context, the board of directors typically includes non-executive directors and the CEO. However, the literature on boards generally refers to the board of directors as the supervising body and, thus, excludes the CEO (see Adams et al. 2010). Even though the Dutch structure is a two-tier system in which the non-executive (or supervisory) directors reside in a supervisory board and senior management in a separate management board, for matters of parsimony and in line with the literature, I equate the board of directors in a one-tier structure with the supervisory board in a two-tier structure.

The ‘board strength’ measure in the Hoitash et al. (2009) study comprises board meeting frequency, board size, proportion of independent board members, board tenure and outside board membership.

An additional benefit of using a less refined categorisation scheme is that coders have fewer options to choose from which should benefit the intercoder reliability (Neuendorf 2002).

Qualitate qua similar results are found on the basis of Arcot and Bruno's classification.

I am indebted to one of the reviewers for this suggestion.

The provisions were equally weighted because the Dutch corporate governance code considers each provision to be equally important. Obviously, this does not rule out the possibility that users do not perceive differences. However, incorporation of possible differences in the importance of the various provisions would require a survey among users, which was outside the scope of the present study.

Eight of the 331 firm-year observations involve observations in which there was full compliance with the Dutch corporate governance code. Excluding these observations does not alter the main results reported in this study.

Pursuant to the Act on the Disclosure of Major Holdings and Capital Interests in Securities-Issuing Institutions of 2006, any person or institution that acquires or disposes of shares in a Dutch listed firm and as a result, the percentage holding in the capital and/or voting rights reaches, exceeds or falls below a threshold must notify the AFM of such acquisition or disposal. The thresholds are 5, 10, 15, 20, 25, 30, 40, 50, 60, 75 and 95%. The AFM keeps a public register on its website that can be consulted free of charge. The register presents information on identity, percentage shares owned and type of shareholder (i.e. individual, financial institution, corporation, etc.).

Untabulated results show that the overall mean (median) number of explanations for non-compliance is 5.78 (6.00). The mean number of explanations per year are 5.92 (2005), 6.11 (2006), 5.68 (2007), 5.73 (2008) and 5.46 (2009). A non-parametric Kruskal–Wallis test does not reveal a significant difference in mean number of explanations found in the individual years. With respect to the various types of explanations the means (medians) are as follows (not tabulated): No explanation: 0.78 (0.00); Generic explanation: 2.67 (2.00); Firm-specific explanation: 2.23 (2.00). Non-parametric tests do not reveal differences between the years.

An analysis of non-compliance with the code's principles (not tabulated) indicates that a limited set of best practice provisions represent a considerable share of all reported non-compliance. Consistent with prior research among British and German listed firms (von Werder et al. 2005, Arcot and Bruno 2006), I find that Dutch firms frequently deviate from best practice provisions with respect to executive compensation.

The Herfindahl index is the cumulative squared shareholdings of individuals or institutions classified as a major shareholder (i.e. a 5% stake or larger). Specifically, the Herfindahl index can be calculated as follows:

By using the Herfindahl index, it is possible to take into account simultaneously differences in number of major shareholders as well as the size of block holdings. For instance, an untransformed measure would not distinguish one major shareholder owning 20% of the firm's shares from four major shareholders each owning 5% of the firm's shares. The Herfindahl index reflects that in the first case ownership is, relatively speaking, more concentrated than in the second case (the Herfindahl index for the first case would be 0.04, while in the second case it would be 0.01). However, in a sensitivity test I show that the results are robust to the use of raw percentages instead of the Herfindahl index.

As a robustness check, I replaced the issue dummy, which assumed the value of 1 if the firm issued share capital in the year of the annual report, by an alternative issue dummy. The alternative issue dummy assumed the value of 1 if the firm issued share capital in the year following the year of the annual report (e.g. when looking at the annual report 2008, the dummy assumed the value of 1 if the firm issued capital in the period 1 January–31 December 2009). A regression analysis using this alternative measure yields similar results.

I also re-estimated model (4) using the data before I imposed the restriction of I/B/E/S coverage. This yields a larger sample of 448 firm-year observations. In this model, I proxy for analyst following by including a dummy variable for AEX membership indicating that a firm is included in the most important index of Euronext Amsterdam and, therefore, is expected to attract the attention of more analysts. The results based on this larger sample are similar to the results reported in model (4), Table 5.

The r-squares for the various models range from 0.120 to 0.165. The model that includes the dummy variable AEX membership (i.e. model (4) in Table 5) has the highest r-square and may suggest that index membership is an important predictor of informativeness of explanations for non-compliance (for a similar observation see Arcot and Bruno 2009).