Monica V. Wout

Master Thesis MSc Accountancy Supervised by Wilfred Kevelam 18-01-2021

11.832 words

Governance and disclosure quality

A study examining the influence of the board aspects ‘type of internal

supervision, risk attitude, female representation, and board size’ on the quality

of disclosure by small to large Dutch pension funds

Monica V. Wout

ABSTRACT

In this study, the influence of several board aspects on 77 Dutch pension funds’ disclosure quality, from 2015 to 2019, after the introduction of new regulations, is examined. This is done using multiple regression analyses and a self-constructed disclosure index. Since the pension funds’ disclosures are of significant importance to the capital markets and society, and little research is done on its determinants, this study closes a large gap in the literature and sheds some new light on the matter. The results show a significant positive association between the type of internal supervision and disclosure quality, suggesting that establishing of a supervisory board, as opposed to a visitation committee or non-executive directors, is

beneficial for the disclosure quality. The risk attitude, female representation, and board size are not significantly related to the disclosure quality and cannot be considered reliable in explaining changes in the disclosure quality. Furthermore, in a one on one situation, a significant positive association is found between the disclosure quality and both control variables pension fund size and type, suggesting that larger pension funds, and industry pension funds have a higher disclosure quality. This study can serve as a stepping stone for future research on disclosure quality determinants, for which is much need, or for more in-depth research on governance aspects and their influence on pension funds’ disclosure quality.

Keywords: Pension fund – Disclosure/Reporting quality – Financial reporting transparency –

Governance – Board of directors – Supervisory board – Internal supervision – Gender diversity – Risk attitude/behavior – Investment portfolio – Board size – Disclosure index

Kerklaan 41a, 9717 HB, Groningen m.v.wout@student.rug.nl

+31 640197603 S-2719894

1

Table of content

1. INTRODUCTION ... 3

2. THEORETICAL BACKGROUND AND HYPOTHESES ... 7

Dependent variable: Disclosure quality ... 7

Independent variables ... 8

Type of internal supervision ... 8

Risk attitude ... 10

Female representation ... 12

Board size ... 13

Control variables ... 14

3. METHODOLOGY ... 16

Sample and population ... 16

Data collection method and measurement ... 16

Dependent variable: Disclosure quality ... 17

Independent variables ... 17 Control variables ... 19 Data analysis ... 20 4. RESULTS ... 20 Descriptive statistics ... 20 Correlation analysis... 23 Regression analysis ... 24 Control variables ... 24 Independent variables ... 24

Additional analyses and robustness checks ... 27

Data reliability and validity ... 30

5. DISCUSSION AND CONCLUSION ... 31

Practical and theoretical implications ... 31

Limitations and future research ... 33

6. REFERENCES ... 35

7. APPENDIX ... 44

2

7.2 Disclosure index ... 46

7.3 Algemene kenmerken pensioenfonds ... 48

3

1. INTRODUCTION

Corporate disclosure is of great importance to the capital markets and their functioning (Healy & Palepu, 2001). The provision of value-relevant, high-quality information of organizations is something that capital markets depend upon in, for example, facilitating business practices that are transparent, ethical, and sustainable (Zhou et al., 2017). Since pension funds are considered one of the most important institutional investors in the capital markets, these markets depend on their provision of information (Klumpes, 2000). Furthermore, as pension funds are responsible for providing pensions to their participants (Shi et al., 2017; De Lange, 2019), their functioning and information provision are of importance to them, and as this concerns all citizens, in the end also to society and economy in general. Additionally, due to their involvement, all participants have a justified interest in being well informed on the pension fund’s financial position, for which the disclosures can serve (De Lange, 2019).

The Dutch pension funds system is praised for its large pension capital built up for financing the pensions, and its integrity, adequacy, and sustainability, including supervision and transparency (Pensioenfederatie, 2020; MMPGI, 2019). However, the introduction of new regulations, such as the ‘Wet versterking bestuur pensioenfondsen’ (hereafter: Wvbp) in 2013 and the ‘Code Pensioenfondsen’ in 2014, suggest the need for improvement concerning governance, as codes of conduct, for example, are often introduced after abuse in a sector has occurred (Shi et al., 2017; DNB, 2013; De Lange, 2019). Governance is an essential aspect concerning disclosures as governing boards are ultimately responsible for decision-making, including decisions on disclosures (Kakabadse et al., 2003; Khan et al., 2019). This is

supported by Manita et al. (2018), who suggest that corporate governance’s effectiveness is of influence on the disclosures. Furthermore, good governance contributes to the trust in pension funds and can help maintain relationships with stakeholders (De Lange, 2019). The latter is important, particularly for Dutch pension funds, as their participants are subject to a

mandatory nature of participation (Broeders et al., 2014; De Lange, 2019). For example, citizens who work in specific sectors, like the construction industry, do not have the freedom to choose between pension funds in the Netherlands and can only join the one that executes the pension scheme of the industry or sector in which they work (Shi et al. 2017). Since the participants can only join another pension fund if they switch from the sector or industry in which they work, the board may lack the incentive to act in their interests (Kowalewski, 2012). Consequently, this could be of influence on the board’s decisions on disclosures.

4 Moreover, the mandatory nature of participation does not only concern the Dutch industry pension funds, but also the corporate and professional group pension funds (De Lange, 2019). Therefore, all are subject to the same aforementioned possible lack of incentive by the board to act in the participants’ interests.

Furthermore, good governance is expected to ensure that all stakeholders’ interests are taken into account and can serve as an instrument to control excessive risk-taking behavior (Jizi et al., 2014; Akande et al., 2020). The internal supervision in place at a pension fund can be considered such an instrument, with the responsibility to monitor the board’s functioning and ensure that the board does not engage in excessive risk-taking (De Lange, 2019). As the board is ultimately responsible for decision-making on disclosures and internal supervision is in place to monitor the board, it can be assumed that one influences the other. This is in line with the study of Rahman et al. (2016), in which it is suggested that a supervisory board aids in achieving higher levels of disclosure. Moreover, as Dutch pension funds are non-profit organizations without the involvement of shareholders, their internal supervisory body differs from the traditional one that primarily guards shareholders’ interests, and of which members are elected or approved by the shareholders after nomination by the board (Shi et al., 2017; Renneboog & Zhao, 2020). The Dutch pension fund’s internal supervision focuses on

ensuring that the board takes the environment and all stakeholders into account when

performing their tasks, among which disclosure decisions, with its members being appointed by either the board or the body of stakeholders, depending on the board structure (De Lange, 2019). Moreover, the board’s requirement of taking stakeholders’ interests into account is incorporated in Article 105 of the Dutch Pension Act (Lutjens & Akkermans, 2020). Due to the aforementioned possible lack of incentive by the board to act in the participants’ interests, and the responsibilities of internal supervision, it can be assumed that internal supervision is, either directly or indirectly, of great importance to the disclosures.

Regarding risk-taking of the board, taking risks in investing premiums is something pension funds engage in to provide pensioners their pension (De Lange, 2019). However, there are frameworks within which the investment policy must be shaped, including the requirement of diversification and investing in such a way that safety, liquidity, quality, and return of the portfolio are guaranteed. These requirements are prescribed by the so-called ‘Financieel Toetsingskader’ (hereafter: FTK) and supervised by an external party, namely ‘De

5 can still occur, possibly resulting in financial issues, and consequently hinder pension funds in meeting their obligations and providing pensions. Therefore, boards engaging in excessive risk-taking may be less inclined to be transparent concerning disclosures, influencing the disclosure quality. This assumption is supported by the research of Wang et al. (2015) and Kanagaretnam et al. (2019), who suggest that higher disclosure quality and levels are

associated with lower risk-taking levels by banks. Although these studies concern banks, the results may be generalizable to pension funds as both are similar in the sense that they are of great importance to the capital market and provide financial support to citizens.

Other board aspects that are often, in general, suggested to be of influence on disclosures are gender diversity (Baalouch et al., 2019; Manita et al., 2018; Katmon et al., 2017) and board size (Xu et al., 2019; Jizi et al., 2014; Lorenzo et al., 2018). In some of these studies, it is suggested that more diverse or larger boards provide additional resources that can be beneficial for the disclosure quality, such as a high stakeholder orientation of women and a broad range of knowledge, expertise, and considerations that come with larger boards. Furthermore, the studies by Nekhili and Gatfaoui (2013), Abdullah (2014), and Shi et al. (2017) indicate that board size is also positively associated with female representation.

Moreover, involving risk behavior, it is also often suggested that females are, in general, more risk-averse than men (Qu, 2020; Enkel & Grossmann, 2008; Sapienza et al., 2009). This would suggest that more women on the board may result in a board taking fewer risks.

Due to the possible influence of the aforementioned aspects of Dutch pension funds’ boards on the disclosures, this study aims to examine, using regression analyses, to what extent these aspects are of influence on the disclosure quality, leading to the following research question: “To what extent are the board aspects ‘the type of internal supervision, risk attitude, female representation and board size’ of influence on the quality of the disclosures of Dutch pension funds? It will also be examined whether the type of internal supervision and female representation is of influence on the relationship between the risk attitude and the disclosure quality, and whether board size is of influence on the relationship between the female representation and the disclosure quality. To provide an answer to the research question and the possible moderating effects, empirical research is conducted whereby data is collected through an individual examination of the annual reports of 80 randomly selected small to large Dutch pension funds out of a total population of 200, of the years 2015, 2017 and 2019.

6 Furthermore, the quality of disclosure will be determined by the use of a self-constructed disclosure index.

This paper contributes to the existing literature by providing a comprehensive way of measuring the quality of disclosure as a whole instead of focusing on specific areas like corporate social responsibility (hereafter: CSR), risk, or voluntary disclosure, which is often the case with prior literature (Khan et al. 2019; Miihkinen, 2012; Lorenzo et al., 2018; Baalouch et al., 2019). Therefore, in practice, this study provides insight into disclosures to a wide array of stakeholders as it covers the quality of disclosure as a whole. Moreover, in theory, it provides a comprehensive measurement of disclosure quality, by using a self-constructed disclosure index, for which still is much need according to existing literature (Beattie et al., 2004; Baalouch et al., 2019; Beretta & Bozzolan, 2008; Healy & Palepu, 2001). Second, despite the importance of pension funds’ disclosure quality to the capital markets, and society and economy in general (Healy & Palepu, 2001; Zhou et al., 2017; Klumpes, 2000), little research has been conducted on pension funds’ disclosure quality and determinants. For example, Klumpes (2000) examined incentives and disincentives for voluntary disclosure by pension funds. Other studies that involve pension funds as a

population examine determinants of performance dimensions, such as investment or pension performance (Broeders et al., 2019; Ammann & Zingg, 2010; Xu et al., 2020). Furthermore, Xu et al. (2020) recommend exploring the effects of board aspects such as diversity in determining the quality of disclosure. This paper seems to be, to my knowledge, the first to examine the influence of board aspects on the disclosure quality by specifically Dutch pension funds, closing a gap in the literature on pension funds’ disclosure quality and its determinants. In practice, the results will provide pension funds knowledge on the influence of certain board aspects on the disclosure quality and give all stakeholders insight into the disclosure quality of pension funds and its determinants. Third, with several changes in law and regulations on governance being introduced from 2014 on, among which the ‘Code Pensioenfondsen’ (De Lange, 2019; Shi et al., 2017), this research is of high value in providing insight in to what extent these changes have been of influence on the disclosure quality in the period 2015 to 2019. For example, as research indicates that gender diversity still is an issue concerning Dutch pension funds’ boards, with 35% having no women on the board as of 2014 after the introduction of self-regulation (Shi et al., 2017), this study contributes to finding out to what extent self-regulation on gender diversity has been effective from 2015 to 2019.

7 This research paper is organized in the following order. First, in section two, the relevant theory is presented, accompanied by hypotheses development. Next, in section three, a description of the used methodology is provided, including the data sample, collection, measurement, and analysis. In section four, the results of all the conducted analyses are presented, among which the descriptive statistics, correlations, regressions, and additional analyses and robustness checks. Last, in section five, a conclusion and discussion on the research results, including implications, recommendations, and limitations, is presented.

2. THEORETICAL BACKGROUND AND HYPOTHESES

According to Singhvi and Desai (1971), the quality of disclosure refers to the reliability, accuracy, and completeness of the information disclosed. Additionally, according to Barth and Schipper (2008), elements like transparency and understandability are also involved. The disclosed information by organizations can be distinguished in several ways, such as financial or non-financial, and voluntary or non-voluntary (Beattie et al., 2004; Khlif & Hussainey, 2016). Financial disclosure is defined by Gibbins et al. (1990) as all financial information that is deliberately disclosed, whether qualitative or quantitative, voluntary or non-voluntary. Robb et al. (2001) define non-financial disclosure as all qualitative information disclosed in annual reports, excluding the financial statements and related footnotes. Due to the nature, it is more likely for large variations to occur within voluntary disclosures, either financial or non-financial, as opposed to the non-voluntary ones (Khlif & Hussainey, 2016). However, the extent of voluntary disclosure is difficult to measure due to its nature (Healy & Palepu, 2001). Furthermore, firms can have various incentives to disclose information voluntarily. For example, to lower their cost of capital, thus a financial reason, or as it is considered a social responsible thing to do (Gelb & Strawser, 2001).

For explaining reasons to disclose information in one’s interests, the agency theory can be applied. This theory suggests that issues arise due to agents and principals’ conflicting interests, often resulting in the parties acting in their interests, possibly at the expense of the other parties’ interests (Hoppmann et al., 2019). In accordance with this theory, it may be that pension funds’ decisions regarding disclosures are in their interests, possibly at the expense of their stakeholders. However, it is difficult to determine the exact motive due to pension funds being foundations without a profit motive (Shi et al., 2017). For example, if much risk is taken with investments, one can claim that this is irresponsible towards the participants of the

8 pension fund due to the involved financial risks with all their consequences, but one can also claim that taking these risks are necessary to provide the pensioners their pensions (Akande et al., 2020; De Lange, 2019). Due to this difficulty in determining the motive, the legitimacy

and stakeholder theory seem more insightful in explaining why pension funds decide to

disclose information. Disclosing information because it is socially responsible is in line with the stakeholder theory that suggests firms carry the responsibility to disclose information so that their stakeholders can make appropriate decisions, and because it enables them to

discharge their accountability (An et al., 2011). This is also in line with the legitimacy theory, which expands the stakeholder theory and considers society as a whole, not just a firm’s stakeholders (Laan, 2019; An et al., 2011). Since pension funds are one of the most important investors in the capital markets, making the whole society depend on their information

provision, society can be considered an important stakeholder (Zhou et al., 2017).

Furthermore, the legitimacy theory suggests that firms should adhere to the bounds and norms of the society in which they operate in order to obtain social legitimacy, which is the case with especially Dutch pension funds since they are non-profit firms that derive their existence from society, and their participants (An et al., 2011). For obtaining legitimacy in the eyes of society, they can use disclosures as an instrument (Hassan, 2009).

Independent variables

Regarding the independent variables, which all can be considered governance aspects, the following definition of governance by De Lange (2019) is used: “the process of management and control, involving the roles, tasks, and responsibilities of the board and the control on it.” The possible influence of the board aspects on the disclosure quality are explained in the next subsections with the use of the following four theories: stakeholder theory, legitimacy

theory, agency theory, and the resource-based view (hereafter: RBV).

Type of internal supervision

Dutch pension funds have a supervisory board, a visitation committee, or non-executive directors for internal supervision (De Lange, 2019). In general, a supervisory board’s presence to monitor the board is considered beneficial in terms of clear delineation of responsibilities and quick decision-making, as opposed to having a board of which the non-executive directors are responsible for internal supervision (Millet-Reyes and Zhao, 2010). This is supported by Jaffar et al. (2013), who suggest that members of a supervisory board are more

9 independent and objective in supervising compared to non-executive directors. As for the Dutch visitation committees, no academic information can be found. However, Article 103 of the Dutch Pension Act does indicate that a visitation committee is an independent body that pays a yearly visit to a company for which they perform the internal supervisory function (Lutjens & Akkermans, 2020). This information and the general assumption on a supervisory board and non-executive directors will be elaborated on within this study. Therefore, it can be assumed that a supervisory board is the most beneficial for the internal supervision, followed by a visitation committee, due to their independent nature, and non-executive directors are considered the least beneficial due to their least independent nature.

Regarding explaining pension funds’ motivation to disclose information, the stakeholder and

legitimacy theory can be used. Those responsible for internal supervision within a pension

fund have partial responsibility to ensure that the board takes stakeholders’ interests into account (De Lange, 2019). This responsibility is in line with the stakeholder and legitimacy theory, suggesting that stakeholders, including society, should be taken into account as it is their social responsibility to do so (Zhou et al., 2017; An et al., 2011; Laan, 2019). Since Dutch pension funds derive their existence mainly from their participants, these participants have a justified interest in being informed on its financial position, for which the disclosures can serve (De Lange, 2019). Taking stakeholders’ interests into account regarding disclosures, among which the participants, enable pension funds to obtain legitimacy (An et al., 2011; Hassan, 2009). Considering the just mentioned characteristics, role, and responsibility of those responsible for the internal supervision, it is likely that internal supervision in general, and the type of internal supervision, is of influence on the board’s decisions concerning disclosures, and thus the disclosure quality. This is supported by Rahman et al. (2016), who suggest that a supervisory board aids in achieving higher levels of disclosure.

The next theory in explaining pension funds’ reason for disclosing is the agency theory. It can be suggested that, in accordance with the agency theory, disclosures can help mitigate the information asymmetry that exists between the board and the stakeholders due to the board possessing inside information that the outsiders, the stakeholders, do not (An et al., 2011). In this sense, the disclosures provide monitoring (At et al., 2011). Since those responsible for the internal supervision monitor the board, and the board decides on what to disclose, it can be assumed that one influences the other. Elaborating on this, it can be suggested that little or insufficient supervision could make the agents, the board members, more inclined to act in

10 their interests, rather than the principals’, stakeholders’, interests, and disclose certain or less information (Hoppmann et al., 2019). As Dutch pension funds’ are foundations without a profit motive, and board members have fixed salaries (Shi et al., 2017; De Lange, 2019), this is an unlikely motive for acting in their interests. However, for example, a board subject to little supervision may be more focused on the short-term or inclined to take more risks to achieve higher returns and not disclose this to uphold a reputation or maintain specific job prospects. These possible agency issues may be enhanced by the absence of shareholders and the Dutch pension funds’ mandatory participation nature (Kowalewski, 2012; Shi et al., 2017). Based on all the aforementioned, the following directional hypothesis is established: H1: The presence of a supervisory board is positively associated with disclosure quality.

Moreover, since internal supervision can be considered an instrument to ensure that

stakeholders’ interests are taken into account and mitigate the board’s excessive risk-taking (Akande et al., 2020), one would expect the better the internal supervision, the less the board engages in excessive risk-taking. Consequently, this may be of influence on the disclosure quality. This is supported by Barakat and Hussainey (2013), who find that independent and effective supervision positively influences the risk reporting quality of banks. Based on the aforementioned, the following moderating effect is expected:

H1a: The presence of a supervisory board has a positive moderating effect on the relationship between the board’s risk attitude and the disclosure quality.

Risk attitude

For any organization to achieve success, it is essential to optimally use resources for which appropriate risk-taking is required (Akande et al., 2020). To provide pensioners their

pensions, pension funds undertake risks with investing, as far as the requirements of the FTK allow (De Lange, 2019). Although achieving higher yields often requires taking more risk, a trade-off must be made between risk and return regarding the investment portfolio selection (Cortés et al., 2013). Campbell and Viceira (2005) define risk as “the conditional variances and covariances per period of asset returns.” Fixed-income securities, such as corporate and government bonds, are considered secure, low-risk investments that offer great stability (Cortés et al., 2013; Liu, 2016; Alexander, 1997). Government bonds are suggested to provide a degree of certainty that no other asset is able to provide (Blundell et al., 2005). Although fixed-income securities are considered low-risk investments, they also involve lower returns than, for example, common stocks (Alexander, 1997). Equities and alternative investments,

11 like private equity, hedge funds, real estate, and commodities have the potential to provide higher returns, but typically involve more risk (Yeoh, 2007; De Lange, 2019; Gao & Nardari, 2018; Liang & Park, 2010; Blundell et al., 2005). This distinction of fixed-income securities being low-risk and equities and alternative investments being high-risk, is supported by the generally accepted notion that debt is considered low-risk and equity high-risk, as in case of payment problems and bankruptcy, all the creditors are the first to get their money back since they have the first claim. Furthermore, for absorbing risks, such as possible disappointing returns, Dutch pension funds are required to keep a reserve, called the ‘capital requirement’ (hereafter: CR). This requirement is prescribed by the FTK, whereby the CR amount is calculated by using a model in which risk factors are included and valued, namely a so-called ‘solvency test’, which involves ten types of risks used to determine the pension fund’s risk profile level. The board may decide to deviate from this standard model, but only with the DNB’s approval (De Lange, 2019). Furthermore, in general, the following applies with CR: the higher the risk profile level, the higher the CR needs to be for absorbing risks.

Although risky investments can be profitable, the involved risks can also threaten the pension funds’ goal to provide pensioners their pensions. For example, it can be assumed that the more risk is taken, the higher the likeliness of financial issues to occur, and the higher the likelihood of the pension fund not being able to meet its obligations and provide pensioners their pension. This could cause huge societal issues as the Dutch citizens are obligated to join a pension fund, and the pension funds are essential for the capital markets’ functioning (Shi et al., 2017; Healy & Palepu, 2001; Klumpes, 2000). Moreover, in line with the stakeholder

and legitimacy theory, all stakeholders, including society, should be considered concerning

risk-taking as it is socially responsible to do so (An et al., 2011; Barakat & Hussainey, 2013). This also includes being transparent about risk-taking and risk management in the disclosures (Barth & Schipper, 2008), aiding in obtaining legitimacy (Hassan, 2009).

As the board is responsible for determining the extent to which the pension fund is prepared to take risks concerning investments (De Lange, 2019), the aforementioned possible lack of incentive by the board to act in the stakeholders’ interests may result in a greater tendency to engage in risky investments. According to Barth and Schipper (2008), financial reports should include information about risks that are faced, suggesting that when a board decides to make investments involving more risk, this should be mentioned, including how the risks will be managed. As taking excessive risks, possibly at the expense of stakeholders, conflicts with the

12 assumptions of the stakeholder and legitimacy theory that a pension fund is expected to adhere to, it can be assumed that a board then may be less inclined to be transparent regarding disclosures. This is in line with the agency theory that suggests issues arise due to conflicting interests between agents and principals, in this case, the board and the stakeholders, with the board taking risks possibly at the expense of stakeholders and not being transparent about it in the disclosures (Akande et al., 2020). This assumption is supported by prior studies indicating that higher financial reporting quality and higher disclosure levels are associated with lower risk-taking levels by banks (Wang et al., 2015; Kanagaretnam et al., 2019). Another example is provided by the study of Hunton et al. (2006), in which it is suggested that more

transparency in reporting aids in detecting earnings management. Elaborating on this and generalizing it to this study, it could be assumed that more transparency regarding disclosures results in easier detection of an excessive risk-taking board, which the board members may not desire and make them inclined to be less transparent in the disclosures. Based on all the aforementioned, the following directional hypothesis is established:

H2: The risk attitude of the board, measured by the risk profile of the investment portfolio, is negatively associated with the disclosure quality.

Female representation

Various papers state that females tend to be more concerned with society’s wellbeing and are more socially responsible and stakeholder-oriented (Hofstede et al., 2010; Adams et al., 2015; Setó-Pamies, 2013; Ben-Amar et al. 2015; Adams and Ferreira, 2009). Therefore, women may be more inclined to focus on building trust and reducing information asymmetry (Al-Shaer & Zaman, 2016). All these aspects fit the mentioned concept of disclosing as it is socially responsible to do so, which is in line with the stakeholder and legitimacy theory. Moreover, according to the RBV, a more diverse board has access to more resources which is beneficial for identifying and fulfilling stakeholder demands (Khan et al., 2019), and provides more varied perspectives that may influence the decision-making (Adams et al., 2015). Since having females on the board may provide extra resources such as the mentioned stakeholder-oriented attitude, they might aid in higher disclosure quality. This assumption, and the fact that multiple other studies suggest a positive association between board gender diversity and the quality of disclosure (Khan et al., 2019; Katmon et al., 2017; Al-Shaer & Zaman, 2016; Liao et al., 2015) led to the following expected association:

13 Furthermore, prior studies suggest that women, as opposed to men, are more risk-averse, more averse to reputation loss and litigation, and are able to reduce opportunistic financial reporting behavior (Qu, 2020; Al-Shaer & Zaman, 2016; Abbott et al., 2012). However, Adams and Funk (2012) find that females are also more tolerant towards aggressive financial reporting practice, which detracts from the effect of them being more risk-averse than men and reducing opportunistic reporting behavior. Nevertheless, due to the majority of studies indicating that women are in general more risk-averse than men (Qu, 2020; Eckel & Grossmann, 2008; Sapienza et al., 2009), the following moderating effect is expected:

H3a: Female representation of the board has a positive moderating effect on the relationship between the risk attitude of the board, measured by the risk profile of the investment portfolio, and the disclosure quality.

Board size

Board size has frequently been included in prior research on governance as it is considered an essential element concerning decision-making and governance (Xu et al., 2020; Al-Shaer & Zaman, 2016; Singh et al., 2018). According to the RBV, a large board comes with a broad range of opinions and considerations, resulting in more expertise, knowledge, and resources crucial for decision-making (Adams & Mehran, 2012; Dalton et al., 1999). This is supported by Jizi et al. (2014) and Al-Shaer and Zamand (2016) who find a positive association between board size and CSR disclosure, and sustainability disclosure quality, respectively.

On the contrary, smaller boards are claimed to be more effective as they are more decisive and cohesive in terms of the information exchange and decision-making than larger boards (Singh et al., 2018; Jensen, 1993; Yermack, 1996). Yermack (1996) supports this assumption as he finds a negative association between board size and firm performance, suggesting that smaller boards are more effective than larger boards. This is also supported by Xu et al. (2020), who find a negative association between board size and pension plan performance, suggesting that larger boards are less effective in decision-making, causing poorer

performance. Therefore, it can be assumed that larger boards may negatively impact the decisions on disclosure and consequently, the disclosure quality. Due to multiple papers providing evidence that is in conflict with evidence provided by various other papers, with one showing a positive association between disclosure quality and board size, and the other showing a negative association, the following unidirectional hypothesis is drafted:

14 Furthermore, according to several studies, female representation is suggested to be positively associated with board size (Shi et al, 2017; Abdullah, 2014; Nekhili & Gatfaoui, 2013). This indicates that board size possibly has a moderating effect on the relationship between female representation on the board and the disclosure quality, leading to the following hypothesis: H4a: Board size has a positive moderating effect on the relationship between female representation on the board and the disclosure quality.

In figure 1, a conceptual model is presented, including all the expected associations.

Control variables

Various studies provide evidence on firm size positively influencing a firm’s disclosures, such as that it positively affects the risk and sustainability reporting quality, the incentives to disclose voluntarily, or to disclose social information (Al-Shaer & Zaman, 2016; Miihkinen, 2012; Klumpes, 2000; Scott, 1994; Belkaoui & Karpik, 1989; Robb et al., 2001). Moreover, the RBV suggests that having more resources at ones disposal is beneficial for the disclosure quality, which is the case with a larger as opposed to a smaller firm (Palmer, 2008). Due to this assumption and the evidence provided by prior studies, a positive association between pension fund size and disclosure quality is expected, and will thus be controlled for.

15 Second, a pension fund’s funding ratio, reflecting the actual financial position, is suggested to influence the disclosures (Klumpes, 2002). A pension fund’s funding ratio, namely the equity to pension liabilities (De Lange, 2019), can be considered equal to an entity’s solvency, which is the ratio of equity to debt. An entity’s leverage is the ratio of total debt to assets. Therefore, it can be assumed that high leverage equals low solvency. Bradbury (1992), Lan et al. (2013), and Hossain et al. (1995) provide evidence of a positive association between leverage and disclosure, suggesting a negative association between solvency and disclosure. A possible explanation for the association could be that entities with high leverage, or low solvency, thus relatively high debt, are inclined to disclose more information to try to decrease the firm’s risk level that is perceived by the market and decrease the pressure of society (Höring & Gründl, 2011). Another possible explanation is that they disclose more in order to keep regulatory authorities at a distance. This explanation is based on the suggestion of Helbok and Wagner (2008) that leveraged banks, thus with low solvency, tend to disclose more in order to try to ward off unwanted attention by supervisors. However, Spiegel and Yamori (2004) find that banks with low solvency issues disclose less. Dobler (2008) supported this, suggesting that threats within an entity, such as bad solvency, may provide incentives to withhold from disclosing information. Although the latter shows some evidence for a positive association, due to many other studies providing evidence of a negative association, this is the one expected between the funding ratio and disclosure quality, and will thus be controlled for.

Last, the type of pension fund. As mentioned, there are three types of Dutch pension funds: industry, corporate, and professional group pension funds (De Lange, 2019; Shi et al., 2017; Broeders et al., 2014). In the case of industry pension funds, all firms and their employees operating in a sector, like the construction industry, are obligated to join the corresponding pension fund. With professional group pension funds, specific professions, like veterinarians and doctors, are obligated to join the corresponding pension fund (Broeders et al., 2014). A corporate pension fund is one that certain companies join, such as the ING pension fund, to which all the employees of ING firms are affiliated (De Lange, 2019). Concerning industry pension funds, these typically have the most participants as they are the pensions providers for entire industries. The mean values of the pension funds’ participants, presented in table 3B, appendix ‘7.4’ provide support for this notion. Furthermore, Elshandidy et al. (2013) find that, in general, firms with high outside ownership have higher levels of risk disclosure. Brown et al. (2011) suggest that firms with more external owners have a higher risk disclosure quality. Since participants of Dutch pension funds can be considered external owners (Shi et al.,

16 2017), it can therefore be suggested that industry pension funds may have higher disclosure levels and quality as they typically have the most participants. Moreoever, since industry pension funds are required to have a supervisory board in place (Lutjens & Akkermans, 2020), which is, as mentioned, suggested to be beneficial for the disclosure quality, it can also be suggested that industry pension funds have a higher disclosure quality due to the presence of a supervisory board. Because of these assumptions of industry pension fund having higher levels and quality of disclosure, the possible influence of the type of pension fund on the disclosure quality will be controlled for in this study.

3. METHODOLOGY

The initial sample amounted to 80 small to large Dutch pension funds, randomly selected out of a total population of 200. However, the final sample consists of 77 instead of 80 pension funds, as three pension funds were excluded due to missing data or liquidation, causing an unbalanced data set. This resulted in a final sample of 77, amounting to 231 observations, 77 of the years 2015, 2017, and 2019 each. The list of Dutch pension funds included in the final sample is presented in appendix ‘7.1’. With the elimination of three pension funds, the sample covers 38.5% of the total population. Additionally, the number of participants can be used to determine the coverage. Considering the population, the Dutch pension funds, the total of participants amounted to about 19 million in 2019, 18 million in 2017, and 18 million in 2015 (DNB, 2019). Dividing the sample’s average total of participants by the average total of participants of the population over the years comes with coverage of 42%, as is illustrated in appendix ‘7.4’, table 3C. With a coverage of 38.5% to 42% and a random selection, the sample can be considered representative, and one that provides an unbiased reflection.

Data collection method and measurement

All the data used for this study is hand-collected from the annual reports of the initial 80 Dutch pension funds from the sample, of 2015, 2017, and 2019, covering the period 2015 to 2019. These three years were selected due to the occurrence of legislation changes in the years before, such as the aforementioned ‘Wvbp’ and ‘Code Pensioenfondsen’, which were introduced in 2013 and 2014, respectively. Since most pension funds typically need some time to adapt to newly introduced regulations, changes are most likely to become visible in the following years, not immediately. Therefore, by analyzing the following years, 2015,

17 2017, and 2019, it can be examined to what extent certain aspects of the disclosure quality have changed over the years. The use of annual reports as the data source to analyze the disclosure quality is considered one of the best as they are comprehensive information-wise and widely available (Beretta & Bozzolan, 2004; Marston & Shrives, 1991). Due to the large extent of work, the task is carried out jointly with three other students, who each individually hand-collected the data of the 20 pension funds allocated to them.

Dependent variable: Disclosure quality

To determine the disclosure quality, a self-constructed disclosure index is used, comprising of 62 items incorporated in the form of a question. The complete index is included in appendix ‘7.2’. The choice to include these 62 items is based on the RJ 610 pension fund reporting guidelines designed to guarantee the quality of reporting. For measurement, a dichotomous scale is used, which is a two-point scale presenting options that are opposite from each other. In this case, an item, in the form of a question, is answered with either a “yes”, “no”, or “not applicable”. For example, when it is explained that a recovery plan applies for the year (item #49), a “yes” is assigned to this item. The total amount of times a “yes” is filled in, divided by the total points that can be achieved, amounts to the disclosure quality score. The maximum score to be achieved is generally 62 points, corresponding with a disclosure quality score of 100%. However, for each item that is considered ‘not applicable’, a point is deducted from the total points that can be achieved. For example, when it is explained in an annual report that a recovery plan does not apply (item #49), the next items involving the measurements and results of the recovery plan (items #50 and 51) are not applicable, decreasing the maximum score from 62 to 60. In that case, achieving 60 points would amount to a score of 100%. Regardless of the maximum score that can be achieved, the following applies: the higher the score, the higher the disclosure quality. Moreover, the used scale is unweighted: all items are considered equally important in measuring the disclosure quality, and weigh the same.

Independent variables

Next to collecting the data for the disclosure index, general data required for measuring all the independent variables included in the study was also obtained from the annual reports. This concerned 42 items, of which the list is included in appendix ’7.3’. The way of measuring the independent variables included in this study is explained in the next section.

18

Type of internal supervision

As mentioned, Dutch pension funds have a supervisory body, a visitation committee, or non-executive directions for internal supervision. Since, in general, a supervisory body, which is called a ‘Raad van Toezicht’ (hereafter: RvT) in the case of Dutch pension funds, is

considered the strongest, pension funds who have an RvT in place are given a score of “1”. If there is something other than an RvT in place for internal supervision, such as a visitation committee or non-executive directors, the pension fund will be given a score of “0”.

Risk attitude

The risk attitude of the board, in other words, the extent to which the board engages in risky behavior, will be determined by the risk profile of the investment portfolio. This risk profile is calculated by summing up high-risk investments, divided by the total amount of investments, thus the low and high-risk investments together. Based on information derived from studies, the generally accepted notion that debt is considered low-risk and equity high-risk, all the fixed-income investments, like corporate and government bonds, are considered low-risk, and everything else, the equity and alternative investments, among which private equity, stocks, real estate, hedge funds, and commodities are considered high-risk investments.

Additional

Furthermore, as mentioned before, it can be assumed that a pension fund with a high-level risk profile is required to keep a high CR to absorb the risks. Therefore, it can also be

expected that a pension fund’s investment portfolio risk profile should add up to the CR, trend wise, since an increase in the risk profile of the investment portfolio would also cause the CR to increase. To see whether the CR adds up to the risk profile of the investment portfolio, the percentage of the CR will be used, which is a ratio provided in the annual reports. This will be further explained in the section “additional analyses”.

Female representation and board size

The female representation, thus the proportion of female directors on the board, will be measured by taking the ratio of women to men on the board. This will be calculated by summing up the number of women on the board, divided by the total of board members. The total number of members present on the board is how board size will be measured.

19 Control variables

The first control variable, firm size, thus pension fund size, is measured by the balance total, namely the total of assets per year-end. Since the total of assets varies largely within the observations, which is illustrated in table 2, appendix ‘7.4’, the natural logarithm (hereafter: LN) of total assets will be used in order to make the variable more suited for the application of regression analyses. In the additional analyses, the pension fund size will be measured by the total of participants. Again, due to high variation, which can be viewed in table 2,

appendix ‘7.4’, the LN of total participants will be used. For measuring the funding ratio, the provided ratio in the annual reports will be used. Regarding the measurement of the last control variable, the pension fund type, industry pension funds will correspond to a “1” as it is expected that this type will most likely be of influence on the disclosure quality, and corporate or professional group pension funds will correspond to a “0”.

All variables, their description, and measurement are presented in table 1. In this table, the additional variables, which are not included in the initial regression model presented in the next section, are also already mentioned. The application of these variables will be further explained later on, in the section ‘Additional analyses and robustness checks’.

TABLE 1 Summary variables

Variable name Description / Proxy Measurement

Dependent variable DISCL_SC Independent variable TYPE_IS RISK_INV FEM_REP B_SIZE Control variables LN_TA

Disclosure quality: the score of the disclosure index (%)

Type of internal supervision

Risk attitude of the board: risk profile of the investment portfolio Proportion of females on the board

Total numbers of board members

Firm size: balance total

Points scored, divided by the maximum points to be achieved

Supervisory board (RvT) is present = 1, not present = 0

Total of high risk investments divided by the total of investments No. of women on the board divided by the total of board members Total number of board members

20 FUND_R TYPE_PF Additional LN_TP RISK_CR Funding ratio

Type of pension fund (PF)

Firm size: participants total Risk attitude of the board: the CR (capital requirement)

Funding ratio per year-end

Industry PF = 1, professional group or corporate PF = 0

LN of total participants per-year end CR per year-end

Data analysis

To test the established hypotheses and determine the extent and direction of the expected associations between the disclosure quality and the type of internal supervision, risk attitude, female representation, and board size, data analyses were conducted using STATA. First, descriptive statistics are provided, followed by a correlation and regression analysis. The type of regression analysis that is used, is an ordinary least squares (OLS) regression. The

following regression model was used for testing the hypotheses:

DISCL_SC = β0 + β1 TYPE_IS + β2 RISK_INV + β3 FEM_REP + β4 B_SIZE + β5 LN_TA +

β6 FUND_R + β7 TYPE_PF + β8 MOD_1 + β9 MOD_2 + β10 MOD_3 + εi

In this model, βi corresponds to the coefficients and εi to the error term. Moreover, MOD_1

represents the expected positive moderating effect of type of internal supervision on the relationship between risk attitude and disclosure quality, MOD_2 the expected positive moderating effect of female representation on the relationship between risk attitude and the disclosure quality that is expected, and MOD_3 the expected positive moderating effect of board size on the relationship between female representation and disclosure quality.

4. RESULTS

In table 2, the descriptive statistics of all the variables, over the years 2015, 2017, and 2019 together are presented. Additionally, in appendix ‘7.4’, tables 2A, 2B, and 2C are included in order to provide the descriptive statistics for the three years separately. All the mean values are derived from these tables and presented per year in table 3 for a clear overview.

21 Starting with the disclosure score (DISCL_SC), table 2 shows thaty, on average the pension funds have a disclosure quality of 76.2%, with the lowest being 50.8% and the highest 92.0%. Furthermore, table 3 shows that disclosure quality has increased over the years. It increased the most, in proportion, from 2015 to 2017, namely 5.6%. From 2017 to 2019, the disclosure quality increased by 1.8%, showing a decline in the increase. Furthermore, in appendix ‘7.1’ the average disclosure scores of each pension fund included in the final sample are presented. Next, the type of internal supervision (TYPE_IS) is shown in table 2 to be on average 0.437, meaning that on average less than half of the pension funds had a supervisory board in place over the years 2015 to 2019. However, it is more interesting to look at the means per year in table 3, which shows a yearly increase, indicating that more pension funds are establishing a supervisory board as the type of internal supervision over the years. Third, regarding the risk attitude (RISK_INV), measured by the risk profile of the investment portfolio, table 2 shows that on average, 42.9% of the portfolio involves high risk, with the lowest being 14.8% and the highest 89.4%. The large difference between the minimum and maximum value shows that there is a great variety. However, table 3 shows that there is not much change over the years, with the mean values staying between 41% and 49%. Fourth, table 3 illustrates that the female representation (FEM_REP) increased over the years. However, table 2 shows that with the average being 17.3%, the minimum being 0%, and the maximum being 60%, that there are still Dutch pension fund boards that do not include women, only men. In order to get a good overview of how many pension funds’ boards include no females, table 3A is included in appendix ‘7.4’, showing that as of 2019, out of the 77 pension funds, 22 had no females on their board, 23 had one, and 32 had more than one. Fifth, the descriptive statistics of board size (B_SIZE) in tables 2 and 3 do not show much interesting. The means over the years in table 3 show that board size varied from 7.7 to 7.9. Considering the control variables, both the total of assets (TOT_ASSETS) and participants (TOT_PART), which measure the pension fund size, involve a large standard deviation, as shown in table 2. To make them both more applicable for regression, the LN was used, after which the standard deviations for both (LN_TA; LN_TP) decreased to below 2.0, as shown in table 2, and the means varying from 14.2 to 14.5, and 9.44 to 9.46, as shown in table 3. Table 3 also shows that the type of pension fund (TYPE_PF) involves a constant mean value over the years of 0.299. Lastly, both the funding ratio (FUND_R) and CR (RISK_CR) involve small standard deviations below 0.1.

22 TABLE 2

Descriptive Statistics

Variable Mean Median St. Dev. Min. Max.

DISCL_SC TYPE_IS RISK_INV FEM_REP B_SIZE TOT_ASSETS LN_TA FUND_R TYPE_PF TOT_PART LN_TP RISK_CR 0.762 0.437 0.429 0.173 7.814 9341120 14.368 1.111 0.299 102469 9.451 1.174 0.770 0 0.413 0.167 8 1331973 14.102 1.103 0 10186 9.229 1.175 0.087 0.497 0.122 0.149 2.296 2.93e+07 1.714 0.097 0.459 343420 1.923 0.054 0.508 0 0.148 0 3 30665 10.331 0.903 0 340 5.829 1.008 0.920 1 0.894 0.6 15 2.64e+08 19.393 1.390 1 2708400 14.812 1.283 N = 231 TABLE 3

Descriptive Statistics: Means per year

Variable 2015 2017 2019 DISCL_SC TYPE_IS RISK_INV FEM_REP B_SIZE TOT_ASSETS LN_TA FUND_R TYPE_PF TOT_PART LN_TP RISK_CR 0.719 0.377 0.410 0.137 7.818 7298076 14.213 1.081 0.299 100727 9.451 1.178 0.775 0.390 0.439 0.173 7.909 1.04e+07 14.394 1.144 0.299 103608 9.448 1.178 0.793 0.545 0.438 0.210 7.714 1.03e+07 14.497 1.108 0.299 103073 9.455 1.165 N = 231

23 Correlation analysis

In table 4, the results of the pairwise correlation between the variables are presented. Correlation values above 0.700 provide an indication for multicollinearity, which can be described as a strong correlation between certain variables that can affect the regression results. There is no correlation value above 0.700, suggesting that this is not the case, and all variables can be included in the regression analysis. However, some values can be suggested to be considerably high. For example, between the type of pension fund (TYPE_PF) and type of internal supervision (TYPE_IS), board size (B_SIZE), and pension fund size (LN_TA), there are significant correlations on a 0.1% level: 0.493, 0.425, and 0.457, respectively. This can be explained by the aforementioned aspects, such as that industry pension funds are obligated to establish a supervisory board and that these types of pension funds, due to their nature, are large of size and often go together with a larger board. Another notable correlation value is between the pension fund size (LN_TA), and board size (B_SIZE) of 0.668 that is significant on a level of 0.1%. This correlation value can be explained by the general notion that larger firms, thus also pension funds, typically involve a larger board.

TABLE 4 Correlation Matrix Variable 1 2 3 4 5 6 7 1. DISCL_SC 2. TYPE_IS 3. RISK_INV 4. FEM_REP 5. B_SIZE 6. LN_TA 7. FUND_R 8. TYPE_PF 1.000 .232*** (.009) .116 .163* .150* (.035) .175** 1.000 .116 .075 .270*** .368*** (.223)*** .493*** 1.000 .094 .148* .126 .014 .037 1.000 .069 .200** .065 .139* 1.000 .668*** (.054) .425*** 1.000 .015 .457*** 1.000 (.282)*** Significance level (two-tailed): *** p < .001, ** p < .01, * p < .05

Since later on, in the additional analyses section, the LN of total participants (LN_TP) will be included for measuring pension fund size, and the CR (RISK_CR) for measuring the risk attitude, two additional correlation analyses are executed. The results of these are presented in appendix ‘7.4’, tables 4A and 4B. Both tables show no values above 0.700 suggesting too strong correlations. Considering table 4A and 4B, the strongest correlations that exist are similar to the ones of the initial correlation analysis presented in table 4, namely between the

24 type of pension funds and type of internal supervision (.049), board size (.425), and pension fund size (.665), and between board size and pension fund size (.610). Therefore, the

explanations for these correlations are similar to the ones provided previously.

Regression analysis

Table 5 shows the regression results, which are further explained in the following subsections. Control variables

In models 1, 2, and 3, only the control variables are included, showing that in a one on one regression, there is a significant positive association between the disclosure quality and both pension fund size (β = .008, p < .05), and pension fund type (β = .033, p < .01). This suggests that both larger pension funds and industry pension funds have a higher disclosure quality. As there was no significant association between the funding ratio and the disclosure quality in a one on one situation, this was not included in the table. In models 4 to 10, after including the independent variables, the significant association between both control variables and the disclosure quality ceases to occur. A notable thing is that after including board size in model 7, the positive association between the pension fund size and the disclosure quality becomes negative (β = -.002, p > .05). However, this association is not significant, and relatively weak. To ensure that the LN of total assets is an accurate way of measuring the pension fund size, another way of measurement, namely the total of participants, will be used to run another regression, which will be further explained in the section “Additional analyses and robustness checks”. Lastly, overall the funding ratio seems to be positively associated with disclosure quality, but these associations are never significant.

Independent variables Type of internal supervision

The models 4 to 10, except from 8, illustrate that there is a significant positive association between the type of internal supervision and disclosure quality (β = .033 to .035, p < .05), suggesting that having a supervisory board in place for the internal supervision is beneficial for the disclosure quality. Therefore, support is found for hypothesis 1: The presence of a supervisory board is positively associated with disclosure quality. Furthermore, the fact that model 8 is not showing the significance illustrated in the other models (β = .024, p > .05), is most likely due to the moderator (MOD_1) that also included the type of internal supervision.

25 Risk attitude

The models 5 to 10 indicate a negative association between the risk attitude, measured by the risk profile of the investment portfolio, and the disclosure quality (β = -.029 to -.050, p > .05). This suggests that a high risk profile of the investment portfolio negatively influences the disclosure quality. However, as this association is not significant in any of the models, no support is found for hypothesis 2: The risk attitude of the board, measured by the risk profile of the investment portfolio, is negatively associated with the disclosure quality.

Female representation

Models 6 to 9 show that a positive association exists between female representation and disclosure quality (β = .031 to .057, p > .05), suggesting that a higher proportion of females on the board has a positive effect on the disclosure quality. However, since the associations are not significant, no support is found for hypothesis 3: Female representation of the board is positively associated with the disclosure quality. Furthermore, model 10 showing a negative association between female representation instead of the expected positive one (β = -.073, p > .05), is most likely due to the moderator (MOD_2) that also included female representation. Board size

Models 7 to 10 show board size to have a weak, positive influence on the disclosure quality (β = .001 to .004, p > .05). However, as the associations are not significant, no support is found for hypothesis 4: Board size is of influence on the disclosure quality.

Interactions, moderating effects

The results show a positive association for all three moderators. First, model 8 shows a positive association between MOD_1 and disclosure quality (β = .026; p > .05), suggesting that having a supervisory board as the type of internal supervision has a positive effect on the relationship between the risk attitude and disclosure quality. However, as this association is not significant, no support is found for hypothesis 1A: The presence of a supervisory board has a positive moderating effect on the relationship between the risk attitude of the board and the disclosure quality. The same goes for the second moderator. Model 9 indicates a positive association between MOD_2 and disclosure quality (β = .062; p > .05). This suggests that having females on the board positively influences the relationship between risk attitude and disclosure quality. However, since the association is not significant, no support is found for

hypothesis 3A: Female representation of the board has a positive moderating effect on the

26 investment portfolio, and the disclosure quality. Lastly, model 10 shows that MOD_3

positively influences the disclosure quality (β = .018; p > .05), suggesting that large boards positively affect the association between female representation and disclosure quality.

However, as this association is not significant, no support is found for hypothesis 4A: Board size has a positive moderating effect on the relationship between female representation on the board and the disclosure quality.

Lastly, all models include an R-squared value below 10%, in the case of model 7, with all independent variables included, a value of 7.1%. This suggests that the independent variables explain less than 10% of the variance in the disclosure quality, and in the case of model 7, 7.1%. The F-values that vary between 2.50 and 8.87 indicate that the models’ fit varies between 2.50 and 8.87. For both the R-squared and F-value, the same general assumption applies: the higher, the better. Since, apart from the variables included in this study, there may be many other variables that could explain the changes in disclosure quality, both the R-squared value and the F-value can be considered acceptable.

TABLE 5 Regression Analysis N=231 1 2 3 4 5 6 7 8 9 10 Intercept Control LN_TA FUND_R TYPE_PF Main eff. TYPE_IS RISK_INV FEM_REP B_SIZE Interact. MOD_1 MOD_2 MOD_3 R2 ΔR2 F-value VIF 0.653 .008* - - .022 .018 5.24* 1.00 0.752 - - .033** .031 .026 7.25** 1.00 0.689 .004 .002 .026 .037 .024 2.89* 1.28 0.683 .003 .022 .013 .033* .062 .045 3.70** 1.37 0.691 .003 .023 .012 .033* (.029) .063 .042 3.03* 1.31 0.702 .002 .016 .011 .034* (.033) .052 .071 .046 2.83* 1.27 0.720 (.002) .018 .007 .035* (.040) .057 .004 .078 .049 2.68* 1.48 0.728 (.002) .016 .007 .024 (.050) .055 .004 .026 .078 .045 2.35* 5.24 0.726 (.002) .018 .007 .034* (.049) .031 .004 .062 .078 .045 2.34 5.11 0.747 (.001) .017 .004 .034* (.048) (.073) .001 .018 .083 .050 2.50* 4.52 Significance level (two-tailed): *** p < .001, ** p < .01, * p < .05

MOD_1 (H1a): the positive interaction between TYPE_IS and RISK_INV MOD_2 (H3a): the positive interaction between FEM_REP and RISK_INV

27

MOD_3 (H4a): the positive interaction between B_SIZE and FEM_REP

Additional analyses and robustness checks

Three additional regression analyses have been executed to expand the research, to aid in understanding the results of the initial regression, and increase the reliability of the results. The first one concerns the way of measuring the pension fund size. To check whether the way of measuring the pension fund size is accurate, another regression is run with size being measured by the LN of total participants, of which the results are presented in table 5A. These results show that in a one on one regression, again, there is a significant positive association between pension fund size and disclosure quality (β = .009, p < .01). Measuring size this way shows significance on a higher level, namely on a 1%-level as opposed to the 5%-level, which was the case with size being measured by the LN of total assets. However, again, the only other significant relationship among the main variables and interactions is the one between the type of internal supervision and the disclosure quality, on a 5%-significance level, which was also the case in the initial regression. Due to these similarities between the results of this regression and the initial one, indicating that both ways of measuring provide overall similar regression results, suggest that both can be considered acceptable and usable ways of

measuring the size of pension funds.

TABLE 5A Regression Analysis N=231 1 2 3 4 5 6 7 8 9 Intercept Control LN_TP FUND_R TYPE_PF Main eff. TYPE_IS RISK_INV FEM_REP B_SIZE Interact. MOD_1 MOD_2 MOD_3 R2 ΔR2 F-value VIF 0.676 .009** .041 .037 9.83** 1.00 0.672 .007 .018 .015 .044 .032 3.52** 1.58 0.663 .005 .031 .004 .031* .067 .051 4.07** 1.59 0.672 .006 .033 .003 .032* (.030) .069 .048 3.33** 1.48 0.681 .005 .024 .002 .032* (.034) .049 .075 .051 3.04** 1.42 0.686 .003 .018 .001 .032* (.040) .052 .003 .079 .050 2.73** 1.54 0.694 .003 .016 .001 .021 (.050) .049 .003 .028 .079 .046 2.39* 5.32 0.690 .003 .018 .001 .032* (.045) .038 .003 .033 .079 .046 2.38* 5.17 0.714 .004 .018 (.003) .032* (.048) (.092) (.001) .020 .085 .052 2.57** 4.64

28

Significance level (two-tailed): *** p < .001, ** p < .01, * p < .05 MOD_1 (H1a): the positive interaction between TYPE_IS and RISK_INV MOD_2 (H3a): the positive interaction between FEM_REP and RISK_INV MOD_3 (H4a): the positive interaction between B_SIZE and FEM_REP

The second additional regression that was executed concerns the risk attitude, measured by the investment portfolio’s risk profile (RISK_INV). As mentioned before, the CR of a pension fund provides an indication of the general risk profile of the pension fund. Although the CR is not necessarily directly related to the board’s risk attitude, it can be used to test to what extent the measured risk profile of the investment portfolio is in line with the pension funds’ CR. One would expect that both would provide somewhat similar results. Therefore, a regression is conducted in which the risk attitude that was measured by the investment portfolio’s risk profile (RISK_INV) is now measured by the pension fund’s CR (RISK_CR). The results are presented in table 5B.Since the regression results of the associations between the disclosure quality and the control variables are similar to those presented in table 5, the one on one effect of each is not presented again. Furthermore, the results overall show a resemblance to

previous results with only a significant relationship existing between the type of internal supervision and disclosure quality (β = .038, p < .05). One notable difference is that model 7 shows a significant relationship regarding the second moderator, the interaction between female representation, CR, and disclosure quality (β = 2.205, p < .001). However, this is caused by the occurrence of multicollinearity between female representation and the CR since the VIF value is 82, which is above the threshold of 10 (O’Brien, 2007). When female

representation is removed from the mix, the VIF decreases to 1.52, and both moderator two and CR do not show a significant association anymore. The only significant relationship that remains then is the one between the type of internal supervision and disclosure quality. Since this finding does not add anything new to the initial results, it is not presented.

TABLE 5B Regression Analysis N=231 1 2 3 4 5 6 7 8 Intercept Control LN_TA FUND_R TYPE_PF 0.689 .004 .002 .026 0.683 .003 .022 .013 0.626 .002 .026 .013 0.623 .001 .020 .012 0.672 (.002) .020 .008 0.694 (.003) .015 (.048) 1.300 (.002) (.007) .003 0.721 (.001) .017 .006

29 Main effects TYPE_IS RISK_CR FEM_REP B_SIZE Interactions MOD_1A MOD_2A MOD_3 R2 ΔR2 F-value VIF .037 .024 2.89* 1.28 .033* .062 .045 3.70** 1.37 .032* .051 .062 .042 3.00** 1.34 .032* .062 .051 .070 .045 2.80** 1.30 .033* .030 .055 .004 .075 .046 2.58** 1.52 .032* .098 .059 .005 (.025) .082 .048 2.46* 4.91 .027* (.468)** (2.552)*** .004 2.255*** .137 .106 4.42*** 81.09 .033* .007 (.056) .001 .015 .078 .045 2.36* 4.65

Significance level (two-tailed): *** p < .001, ** p < .01, * p < .05 MOD_1A (H1a): the interaction between TYPE_IS and RISK_CR MOD_2A (H3a): the interaction between FEM_REP and RISK_CR MOD_3 (H4a): the interaction between B_SIZE and FEM_REP

The third additional analysis included is a regression in which winsorization is applied with the pension funds size, measured by the total of assets and participants. These results are presented in appendix ‘7.4’, table 5C, and are meant to illustrate that outliers of the total of assets or participants influence the results to such a small extent, leading to the choice of not applying winsorization. When winsorizing the total of assets at the 1th and 99th percentile, the only significant association that occurs is the one between disclosure quality and type of internal supervision. Some of these associations’ significance level is now 1% as opposed to 5% before winsorization. However, the fact that the expected signification association between pension fund size and disclosure quality is no longer present, and the occurrence of similar significant associations between type of internal supervision and disclosure quality, may indicate that applying winsorization is of little value. Winsorizing the total of participants at the 1th _{and 99}th _{percentile resulted in the same regression results as when no winsorization}

was applied, which is why these results are not presented. Therefore, the total of participants was winsorized at the 5th and 95th percentile, of which the regression results are presented in appendix ‘7.4’, table 5D. However, as the only significant associations that occurred were again between pension fund size and disclosure quality, and type of internal supervision and disclosure quality, winsorizing seemed to be of little added value.