ANTI-TAKEOVER DEVICES, BOARDS AND CORPORATE SOCIAL RESPONSIBILITY: TESTING THE ENTRENCHMENT HYPOTHESIS

(1)

ANTI-TAKEOVER DEVICES, BOARDS AND CORPORATE SOCIAL RESPONSIBILITY: TESTING THE ENTRENCHMENT HYPOTHESIS

Supervisor: dr. R.O.S. (Raymond) Zaal MBA Author: H.T. (Hendrik) de Boer

Study program: Master of Science in Finance Student number: 2208911

Contact: h.t.de.boer@student.rug.nl Report: Master thesis

Version: Final

(2)

ABSTRACT

(3)

I. INTRODUCTION

Over the past six decades corporate social responsibility (CSR) has increasingly received attention, that – with the exception of some interesting critiques – contributed to the currently dominant notion that firms have substantial obligations to society (Wang et al., 2016). One of these more critical theories, that stress the unintended effects of CSR, is the entrenchment hypothesis by Cespa and Cestone (2007).

According to this hypothesis, CSR is a means for high-powered managers of large corporations to collude with stakeholders, aimed at retaining their influence. The theory posits that the stakeholders of a corporation have the ability to endanger a managers’ position in the firm. In response, managers are incentivized to collude with these stakeholders to remain in power. Before turning to this mechanism in more detail, identification of the most relevant actors in the interplay is in place. The powerful decision makers that determine in which CSR projects their firms partake – and to which extend – are managers. Within firms their decisions are monitored by board members. These oversee whether corporate strategy is sufficiently in line with the interests of shareholders that hold the equity of the firm. In addition, interests of other stakeholders including communities, customers, suppliers and employees are considered. I hereafter use the term stakeholders to refer to the latter groups, excluding shareholders. Finally, raiders are investors that seek to takeover underperforming firms, with the aim of replacing managers to increase firm value (Shleifer and Vishny, 1997). Together, managers, board members, stakeholders and raiders make up for the main interactions in the entrenchment hypothesis.

In line with the underlying agency theory – that assumes individuals to be selfish actors aimed at exploiting private knowledge – the entrenchment hypothesis puts forward that managers are primarily interested in acting on private incentives. This entails that managers with decision-making power will – typically – try to maintain their influence. The ability of stakeholders to engage in media campaigns or even boycotts, thus poses a threat to managerial interests (Surroca and Tribó, 2008). The entrenchment hypothesis predicts that in response to this threat, managers collude with stakeholders by engaging in CSR (Cespa and Cestone, 2007).

(4)

These include the aforementioned internal governance function of boards and the external governance function of ‘the market for corporate control’ (Walsh and Seward, 1990).

The ‘market for corporate control’ is a concept that originates from financial economics. It portrays managers as competing for the position to lead the firm (Jensen and Ruback, 1983; Manne, 1965; Shleifer and Vishny, 1997). Driven by shareholders’ interest to increase financial performance, these select managers that are expected to perform in line with this goal. In case of underperformance, managers are replaced.

In addition to current owners, corporate raiders look for managerial underperformance across listed firms. After identifying such a firm, raiders acquire this corporation to sell it at a premium after managerial turnover. This motivates managers to perform well, to avoid being replaced. Therefore, the market for corporate control functions as an external governance mechanism. To protect themselves from this threat, entrenching managers will install anti-takeover devices (Cespa and Cestone, 2007).

Adoption of anti-takeover devices by entrenching managers is expected to be particularly widespread in the US. As a result of a well-functioning market for corporate control, managers are incentivized to obstruct this governance mechanism by installing the devices. This makes the US context a particularly suitable to test the entrenchment hypothesis.

The market for corporate control functions well in the US due to its market efficiency, that results from a high level of investor protection. This is caused by legal origin (La Porta et al., 1997; La Porta et al., 1998). Due to high investor protection in common law countries – such as the US – private benefits of control are relatively limited (Nenova, 2003; Dyck and Zingales, 2004). Therefore, ownership concentration is relatively unattractive (Bebchuk, 1999). One could argue that the high level of investor protection makes concentration of voting- and ownership rights to some extend redundant. The resulting limited ownership concentration boosts the functioning of the market for corporate control. Thus, a high level of investor protection is positively related to hostile takeover attempts (Rossi and Volpin, 2004). This results in a more prevalent role of the market for corporate control in the US.

(5)

Because installing anti-takeover devices and colluding with stakeholder groups are part of the typical entrenchment strategy, the actions are anticipated to be pursued in conjunction by opportunistic managers (Cespa and Cestone, 2007). Therefore, the number of anti-takeover devices of firms is expected to be positively related to CSR performance in the US.

Moreover, the moderating effect of the internal governance function of boards is studied on the relation between anti-takeover devices and CSR performance. The internal governance function of boards promotes effective oversight on managerial decision making. Therefore, the chance of mismanagement resulting in share prices falling below a point that incentivize corporate raiders to engage in a hostile takeover is reduced (Surroca and Tribó, 2008). Because board effectiveness, thus, prevents hostile takeovers that are often followed by replacing management, effective boards are expected to positively moderate the relation between anti-takeover devices and corporate social responsibility performance.

Contrary to an earlier international empirical test of the entrenchment hypothesis (Surroca and Tribó, 2008), my investigation concentrates on the US because of its well-functioning market for corporate control. This allows to better observe the corresponding hypothesized effect of anti-takeover devices, as an obstruction to the workings of the external governance mechanism. Although the increased market efficiency in the US context may increase the ability of raiders to distinguish costly from positive net-present-value CSR projects, the effect of a better functioning market for corporate control is expected to outweigh this effect. Another important addition to the Surroca and Tribó (2008) paper is analyzing the three pillars of CSR performance separately. This allows to examine interactions with specific stakeholder groups. As a result, examining which dimensions of CSR are used to collude with stakeholders is possible.

(6)

II. THEORY AND HYPOTHESES

This chapter has two main functions. First, it provides a synthesis of the main theories and concepts that underlie the entrenchment hypothesis. Subsequently, the six research hypotheses of this study are developed in the second part of this chapter.

2.1 Theories

With the aim of developing the hypotheses on the entrenchment hypothesis in the next subsection, a synthesis of the most relevant academic literature and theories is in place. To this end, an explanation of the two most relevant economic theories is provided, that – in conjunction – make up for the theoretical foundation of the pessimistic entrenchment hypothesis of Cespa and Cestone (2007). These are agency- and stakeholder theory.

Agency theory revolves around interactions of different individuals or groups that depend on each other, in the presence of an information asymmetry between them (Jensen and Meckling, 1976). Because the individuals are presumed to have dissimilar interest, those with more information are expected to exploit the information asymmetry to their own advantage. A common example of agency theory is that of a shareholder of a corporation and a manager in charge of the daily decision making in the organization. In this role, the manager is expected be well informed on current affairs of the firm in order to make decisions based on this knowledge on a daily basis. Shareholders, on the other hand, do not necessarily have specific knowledge of these corporations or the expertise to monitor managers. In addition to this lack of knowledge, shareholders often diversify their investment portfolios to reduce exposure to idiosyncratic risk. As a result, typical equity holders of listed corporations hold – with the exception of block holders – a small portion of the total amount of shares of the firm.

(7)

Adopting an agency theory perspective – that assumes exploitation of the information asymmetry between managers and shareholders – opens the possibility to put managerial discretion on CSR in a novel perspective. Through this lens, stakeholder interests become a powerful means for collusion. Before turning to this in more detail, a general overview of stakeholder theory is in place.

In financial economics the shareholder is assumed to be the most important stakeholder of the corporation (Jensen, 1986). Critics of this ‘shareholder value ideology’ argue that although managers have a fiduciary responsibility to the owners of the firm, a sole focus on shareholders forgoes a more inclusive assessment of interests of stakeholders (Crane, 2016: 509).

Stakeholder theory is a view that incorporates the interests of stakeholders of a corporation in determining its objectives and strategy. This view, that was coined in the ’60s, was popularized twenty years later (Freeman, 1984; Crane, 2016). The stakeholder view holds that all parties that are affected by corporations or affect these firms themselves should be considered a stakeholder, making all relevant interests determine corporate strategy conjointly. The interests of these are integrated, because the corporation is argued to have a responsibility that transcends the fiduciary responsibility to solely shareholders.

This so called corporate social responsibility (CSR) entails the idea that legal identities such as corporations have a moral obligation to take the interests of a broad range of stakeholders into account in determining corporate strategies. This supposition of non-human entities having a responsibility to stakeholders other than shareholders is protested by for instance Nobel-Price winner Milton Friedman in 1970, because addressing social issues is supposed a responsibility of elected politicians who can tackle social issues with public funds, thus relieving managers of the task to ‘misappropriate’ corporate funds to stakeholders (Crane, 2016: 58-59). With the increasing power of multinational corporations, the view that firms have a social responsibility, however, gained significance.

(8)

Socially responsible behavior is valued by a broad range of stakeholders, including customers, employees, suppliers and the communities. Finally, governance criteria address concerns including proper design of management structure. Also, executive compensation is an important factor. This includes examining bonusses of for instance senior employees and pay differences associated with gender. Other corporate governance criteria include the quality of internal controls and audits.

In prioritizing shareholders or different stakeholder groups, personal values of managers have a substantial influence (Chin et al., 2013). Managers could try to use this discretion to collude with certain stakeholders, in order to safeguard their own position (Cespa and Cestone, 2007). I recur to these stakeholders in the hypothesis development section, now turning to the mechanism behind the entrenchment hypothesis in more detail.

In line with the traditional agency theory, managers with a high level of discretionary power are considered the central actor in the entrenchment hypothesis. Their private incentive to maintain the top positions within firms is threatened media campaigns of stakeholders. Social activists, for instance, are found to have the ability to pressure firms through media attention (Dyck and Zingales, 2004). In this regard, the distinction by Friedman (1991) between media‐ oriented and marketplace‐oriented boycotts media campaign is relevant. While the former can be limited to publicly criticizing management, the later type is a more extensive campaign aimed at a substantial overhaul of corporate policies (Friedman, 1991). When firms decide not to give in to the pressure, they risk action of the latter category (Feddersen and Gilligan, 2001; John and Klein, 2003). In response to the threat, managers try to collude with stakeholders through CSR (Cespa and Cestone, 2007). This allows managers to protect their own positions by taking the interests of stakeholders into account.

This entrenchment hypothesis helps to bring the shared incentives of managers and stakeholders to light (Cespa and Cestone, 2007). In its simplest form the entrenchment hypothesis posits that the shared incentive, causes risk of collusion at the expense of shareholders. In this regard, considering the traditional means of safeguarding shareholder interest is helpful.

(9)

economics – can be adopted in combination with the organizational view of corporate governance, that emphasizes the salient monitoring function of boards. By using this combination, both internal- and external governance mechanisms are accounted for. I now turn to these mechanisms in more detail.

The ‘market for corporate control’ is a theory of an external governance mechanism that is dominant in financial economics literature (Manne, 1965; Jensen and Meckling, 1976, Fama, 1980; Fama and Jensen 1983a; 1983b). According to the theory, shareholders aim to optimize corporate financial performance, and thus, seek to hire managers that perform optimally in line with this target. Because corporate owners are on the lookout for potentially better candidates to lead the firm, managers are challenged by competing leaders. The concept is thus “best viewed as an arena in which managerial teams compete for the rights to manage corporate resources” (Jensen and Ruback, 1983: 5). Put differently, (potential) managers of large corporations are in competition on acquiring or maintaining the position to manage the firm. In order to keep their right or position to manage the firm, managers aim to keep performance on par, thus avoiding under-performance. Therefore, the market for corporate control functions as an external control mechanism on managerial performance (Manne, 1965).

An extension to the market for corporate control is the integration of corporate raiders. Integrating these, relaxes the assumption that current owners of the firm are able to identify managerial underperformance on their own. In addition to current shareholder, other investors analyze corporate performance. Raiders will take over firms when they identify that replacing management – and in turn corporate strategy – will be a profitable endeavor. Raiders are defined as affluent organizational entities that seek to financially exploit knowledge on suboptimal management teams. These profits are realized by acquiring a suboptimal performing firm through a hostile takeover. After the takeover, the raider changes corporate strategy by turning over management and exploiting the identified opportunities to sell the corporation at a premium. Managers seek to protect themselves from this threat of the market for corporate control by entrenching themselves. This is done by implementing anti-takeover devices, that only serve their private interest (Takeover Troubles, The Economist, January 31, 2002).

(10)

effective oversight on managers reduces the chance of strategic mistakes that lead to price drops, high quality boards reduce the chance of hostile takeovers and subsequent managerial turnover (Surroca and Tribó, 2008). Since high quality boards reduce the chance of hostile takeovers and subsequent managerial turnover, they are found to positively moderate the relation between managerial entrenchment metrics – including antitakeover devices – and corporate social responsibility performance (Surroca and Tribó, 2008). I now turn to the hypothesis development.

2.2 Hypothesis development

According to the entrenchment hypothesis, managers are inclined to use discretion over CSR opportunistically (Cespa and Cestone, 2007). Their main incentive is to protect their powerful position in firms. Bonding with stakeholder groups is a suitable means to this end. Collusion makes these powerful groups less inclined to pressure managers with lobbying, social activism and media campaigns (Cespa and Cestone, 2007). This allows managers to use their discretion over CSR to safeguard personal interests of staying in power.

In addition to colluding with stakeholder groups through CSR, entrenched managers are expected to safeguard themselves from the market for corporate control. This concept entails that managers are competing for the position to lead organizations. Profit seeking investors will thus be inclined to replace underperforming managers. In particular, corporate raiders who identify unseized opportunities for the firm are expected to acquire it, with the aim of installing a new management that will realize more profits. This mechanism motivates managers to perform well and to realize the possible opportunities, in order not to be replaced. By installing anti-takeover devices, managers reduce the chance of their firm being acquired by a raider. The anti-takeover devices, thus, complicate acquisitions by raiders that want to realize opportunities and profits by replacing shirking managers.

(11)

obstruction to a well-functioning market for corporate control (Rossi and Volpin, 2004). Thus, a high level of investor protection is positively related to hostile takeover attempts, resulting in a more prevalent role of the market for corporate control in the US, where 65,6 percent of traded firms are takeover targets in comparison to for instance the 6,4 percent of Japanese firms (Rossi, Volpin, 2004). As a result of this well-functioning market for corporate control, the US is a context in which entrenching managers are likely to install anti-takeover devices. Because installment of anti-takeover devices and colluding with stakeholder groups are part of the typical entrenchment strategy, the actions are anticipated to be pursued in conjunction by opportunistic managers (Cespa and Cestone, 2007). Therefore, the number of anti-takeover devices of firms is positively related to performance in each of three CSR pillars, environmental, social and governance in the US.

Hypothesis 1: The number of anti-takeover devices is positively related to performance in the environmental CSR dimension.

Hypothesis 2: The number of anti-takeover devices is positively related to performance in the social CSR dimension.

Hypothesis 3: The number of anti-takeover devices is positively related to performance in the governance CSR dimension.

In addition to the external governance function of the market for corporate control, internal governance mechanisms play a key role. In particular, boards perform a salient role in overseeing management (Zahra and Pearce, 1989). To prevent managers from costly opportunism, modern corporations employ a governance structure that separates decision management and decision control (Fama and Jensen, 1983b). This results in an organizational structure in which managers make the decisions that affect the operations of the firm, while board members oversee the decision-making. The responsibility of internal control is thus primarily delegated to boards that protect the interests of shareholders.

(12)

takeover profitability are reduced. This limits the governance function of the market for corporate control, since takeovers that give rise to managerial replacement are less likely to happen (Surroca and Tribó, 2008).

To examine the moderating effect of high-quality boards on the relation between anti-takeover devices and CSR performance, several aspects of board quality are studied. Previous literature shows that board independence (Harjoto and Jo, 2011), board size (Dalton et al., 1999; Hillman and Dalziel, 2003; John and Senbet, 1998) and gender diversity (Landry, Bernardi, and Bosco, 2016; Webb, 2004) are important board quality metrics. Within boards certain responsibilities are delegated to specialized committees. Important in this regard is the auditing committee that overlooks financial statements and appoints the independent auditor. In this paper the level of auditing committee independence is included as metrics for board quality. Since board size and gender diversity are important metrics for board quality, these are also included. Finally, the number of board meetings is included, to account for the intensity of monitoring.

The moderating effect of the internal governance function of boards is studied on the relation between anti-takeover devices and CSR performance. The internal governance function of boards promotes effective oversight on managerial decision making. Therefore, the chance of mismanagement resulting in share prices falling below a point that incentivize corporate raiders to engage in a hostile takeover is reduced. Because board effectiveness, thus, prevents hostile takeovers that are often followed by replacing management, effective boards are expected to positively moderate the relation between anti-takeover devices corporate social responsibility performance (Surroca and Tribó, 2008) in each of three CSR pillars, environmental, social and governance.

Hypothesis 4: Board metrics positively moderate the relation between the number of anti-takeover devices and performance in the environmental CSR dimension.

Hypothesis 5: Board metrics positively moderate the relation between the number of anti-takeover devices and performance in the social CSR dimension.

(13)

III. RESEARCH METHODS

With the aim of presenting the methodology to test the hypotheses that were laid out in the previous chapter, this chapter discusses the research methods. First, general information about the research method and statistical analysis are discussed, to than give an overview of the main research variables that are included in the investigation. Subsequently, the control variables are discussed. The chapter concludes with an overview of the regression equations that are used in empirically testing the research hypotheses.

3.1 Research design

To empirically test the entrenchment hypothesis, a dataset that includes all firms of the Standard and Poor’s 500 index is analyzed. This sample of large listed US firms is selected, because the efficient US market is – as argued in chapter two – a suitable context to test the entrenchment hypothesis in conjunction with anti-takeover devices and the market for corporate control.

The observations of the 2001 until 2017 timeframe are amassed from Thomas Reuters Datastream. Due to missing observations for the dependent variables before 2002 and using one-year-lagged dependent variables, the effective time span of the dataset is 2003 until 2017. The choice to use a time-lag for the dependent variables is motivated by endogeneity concerns. By linking independent variables with the more recent observations of the dependent variables, risk of reverse causality is reduced (Brooks, 2014).

The resulting panel-dataset comprises 506 firms – as a result of the constituent list used – and an effective timeframe of 15 years, making up a total of 8603 firm-year observations. To analyze this panel dataset, ordinary least squares (OLS) regression techniques are used. To control for industry- and time effects, dummy variables – for both year and the respective industry of the firms in the sample – are created incorporated in the regression equation. I now turn to an overview of the main variables that are incorporated in the panel-dataset.

3.2 Main variables

(14)

TABLE 1

Variable names, computation, source databases and data codes

Abbreviation Variable name Computation Data source Data code

Dependent variables

1 ENV Environmental CSR performance Available ASSET4 ENVSCORE 2 SOC Social CSR performance Available ASSET4 SOCSCORE 3 GOV Governance CSR performance Available ASSET4 CGVSCORE

Independent variables

4 ATD Number of anti-takeover Natural logaritm of value ASSET4 CGSRO06V devices

5 ACI Auditing committee Available ASSET4 CGBFDP018 independence

6 BSZ Number of board members Natural logaritm of value ASSET4 CGBSDP060 7 DIV Board diversity, as percentage Available ASSET4 CGBSO03V

of women on the board

8 NBM Number of board meetings Natural logaritm of value ASSET4 CGBFDP024 9 ACI*ATD Interaction ACI*ATD Var. '5' times var. '4'

10 BSZ*ATD Interaction BSZ*ATD Var. '6' times var. '4' 11 DIV*ATD Interaction DIV*ATD Var. '7' times var. '4' 12 NBM*ATD Interaction NBM*ATD Var. '8' times var. '4'

Independent control variables

13 CEX Capital Expenditure Natural logaritm of value Worldscope WC04601 14 LEV Leverage, measured as total debt Natural logaritm of value Worldscope WC03255 15 RDE Research & development Natural logaritm of value Worldscope WC01201

expense

16 ROA Return on assets Available Worldscope WC08326 17 SIZE Size, measured as total assets Natural logaritm of value Worldscope WC02999

This table presents the variables in their respective categories, listing their abbreviated name, full-name, computation, data source of origin and their unique Thomas Reuters Datastream code. Var. abbreviates variable.

(15)

The dependent variables are three CSR dimensions by ASSET 4, comprising ratings of corporate social responsibility performance on the pillars environmental, social and corporate governance. At the root of these ratings is a complex weighted rating of performance on 750 data points and more than 250 indicators. The three dimensions of CSR performance of ASSET 4 give a relatively complete picture of how a firm deals with its stakeholders. Examining the relations between the independent variables and these three separate CSR dimensions, allows to identify differences in the effects that the independent variables have on the different CSR dimensions.

The first main independent variables in this investigation is anti-takeover devices (ATD). This variable reflects the number of anti-takeover devices that are in place within the firm in excess of two. The variable is amassed from the ASSET 4 database. In addition, four other variables are incorporated, that are primarily used to test for interaction effects. These four variables reflect corporate board characteristics, including the percentage of independent board members on the audit committee, the total number of board members at the end of the fiscal year, diversity in terms of the percentage of women on the board and the number of board meetings during the year.

To test for moderating effects, interaction terms are computed for the anti-takeover devices with the four respective board metrics. The hypotheses predict that the relation between the number of anti-takeover devices and the separate CSR dimensions is moderated by board characteristics. To test this, interaction terms of the anti-takeover devices variable and the four board variables are computed and incorporated in the regression equation. This results in four moderating variables. More information on these variables is displayed in table one. To reduce the impact of outliers that typically threaten the results of analysis through OLS regression techniques, all continuous variables are winsorized at the third standard deviation from the mean. This entailed putting a cap on the lower bound of the variables auditing committee independence, board size and corporate social responsibility performance in the dimension governance.

(16)

relations. The standardization procedure is performed on all variables in the model, with the exception of the interaction terms themselves.

The interaction terms are influenced by the winsorizing, log transformation and standardization procedures. As a result of the observation adjustment procedures on the main variables, the robustness of the results is higher and the otherwise problematic influence of outliers in the fixed-effects pooled OLS estimation is reduced.

As toughed upon in the prior research design section, the dependent variables that reflect CSR performance in the three dimensions: environmental, social and governance are incorporated as one-year-lagged variables in order to prevent endogeneity issues. This safeguards the results of data interference from being based on reversed causality (Brooks, 2014).

3.3 Control variables

To control for effects of time and industry, the previously mentioned fixed-effects dummies are incorporated in the regression equation. To this end, all firms are manually categorized in their relevant industry through the Fama-French 12 industry classification on the basis of their respective primary General Industry Classification codes that were amassed from Orbis. In the sample, all firms in the S&P 500 are incorporated, meaning that firms operating in different industries are present in the sample. I choose to not exclude particular observations on the basis of their industries, to maintain a large representativeness of the sample for a wide range of sectors. As discussed above, dummy variables are incorporated to control differences caused by industry related deviations.

(17)

The five control variables are incorporated to control for firm level differences in the sample. Previous literature shows that capital expenditure is positively associated with CSR performance and thus controlled for (Harjoto and Jo, 2015). Furthermore, leverage is controlled for, since firms that are required to make interest payments on loans are restricted in using funds for CSR projects when interest and debt repayments and are due. Therefore, previous literature controls for leverage when investigating CSR (Reverte, 2009). Similarly, research and development outlay limits availability of funds for use in CSR projects, making these expenditures a control variable in CSR literature (Martínez-Ferrero et al., 2017). Firms with higher levels profitability, on the contrary, have more funds available for distribution amongst stakeholders. Therefore, return on assets is used as a control variable to control for corporate profitability in research papers (Waddock and Graves, 1997). Finally, I control for size of corporations to account for the ability of larger firms to have more CSR related impact on stakeholders, in line with previous literature (Barnea and Rubin, 2010).

Similarly, to the observation adjustment procedures that were conducted on the main research variables, control variables are adjusted to make the regression results more robust. To this end the control variable return on assets is winsorized at the third standard deviation below its mean. The control variables capital expenditure, leverage, research and development expense and total assets are transformed to their natural logarithms. As a result, some extremely low observations with a zero-value dropped out. The observation adjustment procedures increase the robustness of the results of the statistical analysis. The panel dataset is now ready to be analyzed, using the regression models that I now turn to.

3.4 Regression equations

(18)

(19)

IV. DATA

The aim of this chapter is to present the panel-dataset that is used to test the research hypotheses. As discussed in the previous chapter, the data is amassed through Thomas Reuters Datastream. An overview of the variables, their abbreviations, full names, computation, unique Datastream identification codes and original data sources including ASSET4 and Worldscope can be found in table one. I now turn to the descriptive statistics of the dataset.

The descriptive statistics of the panel-dataset are presented in table two. This table gives an overview of the number of observations, the minimum value, the maximum value, the mean value and the standard deviation of the distribution for the variables.

An important note is that table two presents the data after conducting the winsorizing procedure on several variables (ACI, BSZ, ROA and GOV) at the third standard deviation. This only had implications for the lower bounds of the distributions. The values reflected in table two are, however, prior to the log transformations. This makes the values in the table more intuitive and hence easier to interpret. The interaction terms are the product of selected standardized variables. As a result, these values are less intuitively interpretable. More information on the winsorizing- and log transformation procedure can be found in the previous chapter. Full-names of the abbreviated variable are presented in table one.

(20)

TABLE 2 Descriptive statistics

N Min Max Mean Std. Deviation

Dependent variables 1 ENV 6009 8,130 97,390 52,315 32,524 2 SOC 6009 3,580 99,320 55,725 28,980 3 GOV 6009 30,671 98,220 77,295 14,880 Independent variables 4 ATD 6028 0,000 12,000 4,224 3,150 5 ACI 5973 85,241 100,000 99,832 1,550 6 BSZ 6022 3,421 34,000 11,016 2,530 7 DIV 5991 0,000 62,500 15,907 9,154 8 NBM 5964 0,000 53,000 8,159 3,642 9 ACI*ATD 4796 -11,522 11,409 0,056 0,811 10 BSZ*ATD 4800 -7,208 7,689 -0,031 0,979 11 DIV*ATD 4798 -6,362 4,558 0,093 0,950 12 NBM*ATD 4767 -8,196 7,685 -0,017 1,012

13 CEX 8181 0,000 37985000,000 1007566,949 2393101,434 14 LEV 8164 0,000 810759000,000 12142793,301 50382897,371 15 RDE 4259 0,000 22620000,000 724729,070 1671962,265 16 ROA 8058 -19,720 104,520 7,022 7,680 17 SIZE 8181 1893,000 2573126000,000 49578765,695 175355499,144

This table presents the descriptive statistics of the variables. Displayed are the number, abbreviated name, the number of observations in the sample (N), the minimum value (Min), maximum value (Max), mean value (Mean) and the standard deviation (Std. Deviation) of the respective variables. Full names of the abbreviated variables are presented in table one. The reflected independent variables: auditing committee independence (ACI), board size (BSZ), return on assets (ROA) and the dependent variable corporate social responsibility performance in the dimension governance (GOV) are winsorized at the third original standard deviation from their original mean ex-ante. To make the table easier interpretable, the listed values are presented prior to log transformation. The moderating variables are computed from the standardized variables that were winsorized and log transformed where necessary ex-ante. A more extensive overview of the variables including their origins and Pearson-correlations are presented in tables one and three respectively.

(21)

The independent control variables are straightforward dollar values of these accounting metrics. The exception, however, is return on assets. This is the only independent control value that includes negative observations, as a result of certain firms making losses rather than profits. I conclude that the descriptive statistics of the dataset shows no strange or deviating observations that hint towards problems in the later stages of analysis. Therefore, I now turn to a discussion of the Pearson-correlations of the variables.

The Pearson-correlations of the variables are presented in table three. With the aim of discussing the most important correlations, I focus analyzing the most extreme values. A full overview that includes all correlations can be found in table three. I discuss correlations that are medium strong (> ,4), strong (> ,5) and very strong (> ,6).

In interpreting correlations, it is important to keep in mind that they do not proof causal structure, meaning that it cannot be deducted from the correlation alone which variable gives rise to the other. Furthermore, multiple potential explanations are possible for high correlations. The explanations presented here are not necessarily perfectly true or the only right one.

Medium strong correlations are observed between the variables environmental governance performance and capital expenditure and between the variables research and development expenditure and leverage. This implies that firms with high capital expenditure typically also perform well in the environmental CSR dimension. Although correlations give no information of the causal structure, a possible explanation could be that firms that invest more in their buildings – to make them for instance more energy efficient – score higher in environmental CSR ratings. The second medium high correlation is found between research and development expenditure and leverage. A possible explanation for this is that firms investing in development of products and processes typically finance this with debt.

(22)

TABLE 3 Correlations

This table presents the Pearson-correlations of the variables. Full-names of the abbreviated variables are listed in table one. The reflected independent variables ATD, ACI,

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Dependent variables 1 ENV 2 SOC ,767 *** 3 GOV ,540 *** ,540 *** Independent variables 4 ATD ,056 *** -,002 -,121 *** 5 ACI ,003 ,002 ,078 *** -,054 *** 6 BSZ ,256 *** ,301 *** ,128 *** -,032 ** ,020 7 DIV ,256 *** ,285 *** ,175 *** ,093 *** -,008 ,160 *** 8 NBM ,072 *** ,100 *** ,089 *** -,017 -,001 ,143 *** ,106 *** 9 ACI*ATD ,009 ,015 -,007 ,046 *** ,601 *** ,037 ** -,013 -,001 10 BSZ*ATD ,039 *** ,032 ** ,046 *** ,013 ,028 -,009 ,017 ,020 ,030 ** 11 DIV*ATD -,003 ,008 ,015 -,102 *** -,013 ,017 ,067 *** ,001 -,024 * ,135 *** 12 NBM*ATD -,028 * -,025 * -,022 ,021 -,001 ,020 ,002 -,012 ,000 ,139 *** ,096 ***

13 CEX ,446 *** ,380 *** ,259 *** -,066 *** -,011 ,299 *** ,159 *** ,136 *** -,005 -,011 ,009 -,003

14 LEV ,332 *** ,276 *** ,160 *** ,027 * -,007 ,399 *** ,171 *** ,264 *** ,021 -,076 *** -,007 -,030 ** ,646 ***

15 RDE ,347 *** ,323 *** ,181 *** -,043 * ,012 ,248 *** ,142 *** ,156 *** ,016 -,014 -,057 ** -,038 * ,562 *** ,456 ***

16 ROA ,011 ,045 *** ,036 *** -,029 ** ,004 -,137 *** ,018 -,139 *** -,005 ,032 ** -,008 -,016 -,017 -,199 *** ,061 ***

(23)

Another strong correlation is found between the total assets of firms and the size of their boards. This implies that larger firms, in terms of total assets, typically employ more directors on their board. This seems feasible, since larger firms will typically be faced with more complex decisions, making it likely that overseeing this decision-making requires more people on the board.

Furthermore, a strong correlation is found between research and development expenditure and capital expenditure. What follows from this is that firms that invest more in research and development of new products and processes are also faced with higher expenses relating to upgrades and expansion of their physical assets and equipment. A manufacturer that often introduces newly developed products will be likely to, for instance, build new production halls and assembly lines.

I now turn to the correlations that I define to be very strong (> ,6). The first of these is the correlation between the dependent variables environmental CSR performance and social CSR performance. This again, implies – similar to the above mentioned other medium strong correlations between the dependent variables – that firms in which CSR performance is aimed at, typically the emphasis will be not on one dimension exclusively, but CSR performance in multiple dimensions. These are dependent variables that are, thus, not integrated in the same regression model. As a result, no multicollinearity issues will arise from this high correlation. The interaction term of auditing committee independence and anti-takeover devices correlates very strong with the auditing committee independence variable itself. This is a result of the interaction term being computed on the basis of the latter variable. This does not have to be a problem, although a large set of variables and interaction terms that correlate very strongly with each other in the same regression model can create multicollinearity problems. I recur to this potential problem below.

(24)

The remaining very high correlations are found between the variable total assets and the three variables capital expenditure, leverage and research and development. The explanation is that larger firms, measured in total assets, invest more in their buildings and equipment, hold more debt and invest more in research and development. This is likely to be the case. These variables show high correlations that could potentially lead to multicollinearity problems, to which I now turn in more detail.

(25)

V. RESULTS

In this chapter the results of the ordinary least squares estimation with fixed-effects on the dataset are presented. The first subsection presents the main results for the CSR dimensions environmental, social and governance. Subsequently, the second subsection presents additional robustness tests.

5.1 Main results

The explanatory values of the first two regression equations – presented in table four and five – are higher, compared to the third regression equation, presented in table six. While the model of the first regression equation explains about 43 percent of the variance of the sample, the model testing the second equation accounts for around 36 percent. The model of the third regression equation is limited to an explanatory power of 18 percent of the variance. This implies that the regression results have to interpreted with care. In particular the relatively low explanatory power of the third model requires caution. Nevertheless, the results of the three regression models are suitable to be used in this investigation. I now turn to an overview of the research results for the six hypotheses respectively.

Hypothesis one posits that the number of anti-takeover devices is positively related to performance on the corporate social responsibility pillar environmental. No evidence is found for this relation. This is not in line with expectations. Only in the first two models, excluding fixed effects, a significantly positive relation is found. The models that including fixed-effects are, however, a more accurate depiction, since they account for the influence of time-series trends and industry differences. Because the models including fixed-effects provide no evidence for a relation between anti-takeover devices and environmental CSR performance, I conclude that no evidence is found for the first hypothesis.

(26)

TABLE 4

Main results of ordinary least squares estimation for the CSR dimension environmental

Corporate social responsibility performance

ENV ENV ENV ENV ENV ENV ENV

(t+1) (t+1) (t+1) (t+1) (t+1) (t+1) (t+1) Model 1 2 3 4 5 6 7 Independent variables ATD ,097 *** ,138 *** ,001 -,001 -,002 _-,002 _-,002 ACI ,003 ,009 ,012 -,019 -,019 _-,019 _-,019 BSZ ,375 *** ,191 *** ,183 *** ,182 *** ,184 *** ,184 *** ,185 *** DIV ,160 *** ,089 *** ,069 *** ,070 *** ,067 *** ,067 *** ,067 *** NBM ,030 _-,005 ,005 ,004 ,005 ,005 ,004 ACI*ATD ,045 * ,044 * ,044 * ,044 ** BSZ*ATD ,047 ** ,048 ** ,049 *** DIV*ATD -,003 _-,003 NBM*ATD -,013

CEX ,298 *** ,334 *** ,337 *** ,328 *** ,328 *** ,327 *** LEV ,137 *** ,082 *** ,080 ** ,083 *** ,083 *** ,082 *** RDE ,138 *** ,142 *** ,139 *** ,141 *** ,141 *** ,140 *** ROA ,082 *** ,051 *** ,051 *** ,053 *** ,053 *** ,053 *** SIZE -,062 -,037 -,035 -,031 _-,031 _-,030 Fixed effects

Industry Yes Yes Yes Yes Yes

Year Yes Yes Yes Yes Yes

N 1773 ₁₇₇₃ 1773 1773 1773 1773 1773 A. R sq. ,204 ,338 ,427 ,428 ,430 ,430 ,429 D. R sq. ,206 ,136 ,097 ,001 ,002 ,000 ,000 VIF 1,057 _7,390 8,226 8,231 8,237 _8,238 _8,250

The dependent variable is the one-year-lagged corporate social responsibility (CSR) performance score on the dimension environmental (ENV). Estimations are based on ordinary least squares techniques, with fixed-effects for industries and years from model three and onwards. N reflects the number of observed firm-years, A. R square abbreviates adjusted R square measure, D. R sq. reflects the change (delta) of the adjusted R square compared to the previous model and VIF is short for maximum variance inflation factor. Full names of the abbreviated variables are presented in table one. A more extensive overview of the variables, the descriptive statistics and Pearson-correlations are presented in tables one, two and three respectively. Symbols ***, ** and * indicate 1, 5 and 10 % significance levels respectively using two sided-testing.

(27)

TABLE 5

Main results of ordinary least squares estimation for the CSR dimension social

SOC SOC SOC SOC SOC SOC SOC

(t+1) (t+1) (t+1) (t+1) (t+1) (t+1) (t+1) Model 1 2 3 4 5 6 7 Independent variables ATD ,016 ,051 ** -,027 -,027 -,028 _-,028 _-,029 ACI -,014 _-,010 -,004 -,009 -,009 _-,009 _-,009 BSZ ,346 *** ,202 *** ,191 *** ,191 *** ,192 *** ,191 *** ,192 *** DIV ,208 *** ,149 *** ,103 *** ,103 *** ,102 *** ,101 *** ,102 *** NBM ,064 *** ,041 ** ,039 * ,039 * ,040 * ,040 * ,038 * ACI*ATD ,007 ,007 ,007 ,008 BSZ*ATD ,025 ,024 ,026 DIV*ATD ,006 ,006 NBM*ATD -,022

CEX ,207 *** ,266 *** ,266 *** ,262 *** ,262 *** ,261 *** LEV ,075 ** ,035 ,034 ,036 ,036 ,034 RDE ,098 *** ,138 *** ,137 *** ,139 *** ,139 *** ,138 *** ROA ,106 *** ,067 *** ,067 *** ,068 *** ,068 *** ,068 *** SIZE ,012 ,014 ,014 ,016 ,016 ,019 Fixed effects

N 1773 ₁₇₇₃ 1773 1773 1773 1773 1773 A. R sq. ,202 ,290 ,359 ,358 ,358 ,358 ,358 D. R sq. ,204 ,089 ,078 ,000 ,001 ,000 ,000 VIF 1,057 _7,390 8,226 8,231 8,237 _8,238 _8,250

The dependent variable is the one-year-lagged corporate social responsibility (CSR) performance score on the dimension social (SOC). Estimations are based on ordinary least squares techniques, with fixed-effects for industries and years from model three and onwards. N reflects the number of observed firm-years, A. R square abbreviates adjusted R square measure, D. R sq. reflects the change (delta) of the adjusted R square compared to the previous model and VIF is short for maximum variance inflation factor. Full names of the abbreviated variables are presented in table one. A more extensive overview of the variables, the descriptive statistics and Pearson-correlations are presented in tables one, two and three respectively. Symbols ***, ** and * indicate 1, 5 and 10 % significance levels respectively using two sided-testing.

(28)

TABLE 6

Main results of ordinary least squares estimation for the CSR dimension governance

GOV GOV GOV GOV GOV GOV GOV

(t+1) (t+1) (t+1) (t+1) (t+1) (t+1) (t+1) Model 1 2 3 4 5 6 7 Independent variables ATD -,113 *** -,093 *** -,330 *** -,327 *** -,328 *** -,328 *** -,329 *** ACI ,086 *** ,088 *** ,085 *** ,131 *** ,131 *** ,132 *** ,131 *** BSZ ,174 *** ,080 *** ,114 *** ,116 *** ,117 *** ,118 *** ,119 *** DIV ,140 *** ,098 *** ,104 *** ,103 *** ,101 *** ,102 *** ,103 *** NBM ,057 ** ,048 ** ,043 * ,044 * ,045 * ,045 * ,042 * ACI*ATD -,066 ** -,066 ** -,068 ** -,066 ** BSZ*ATD ,038 * ,040 *** ,044 ** DIV*ATD -,016 _-,015 NBM*ATD -,035

CEX ,081 * ,065 ,062 ,056 ,055 ,053 LEV ,063 * ,055 ,057 ,059 ,059 ,057 RDE ,008 ,009 ,013 ,015 ,014 ,012 ROA ,115 *** ,089 *** ,089 *** ,090 *** ,089 *** ,089 *** SIZE ,077 ,059 ,056 ,059 ,059 ,063 Fixed effects

N 1773 1773 1773 1773 1773 ₁₇₇₃ ₁₇₇₃ A. R sq. ,083 ,121 ,177 ,178 ,179 ,179 ,180 D. R sq. ,085 ,040 ,067 ,002 ,001 ,000 ,001 VIF 1,057 _7,390 8,226 8,231 8,237 _8,238 _8,250

(29)

Hypothesis number three posits that the number of anti-takeover devices positively relates to performance in the CSR dimension governance. Testing this hypothesis provides a statistically significant finding. Contrary to the anticipated relation, however, this evidence indicates a highly significant negative relation between the number of anti-takeover devices and performance in the governance dimension of CSR. This finding is robust across all seven research models in table six. I recur to this interesting finding in the discussion of the results of testing hypothesis six and in the subsequent conclusion and discussion chapter.

Before turning to the results for testing hypothesis four, five and six, a note on the explanatory value of the interaction terms is in place. The regression results presented in these three tables each comprise seven research models. The latter four models in each of these tables shows how the results are impacted as a result of adding each of the interaction terms. The delta of the adjusted R square measure – indicating the incremental change in explanatory power of the respective models – indicates that the additional explanatory value of incorporating the interaction terms is very low. Typically, simpler models are preferred over more complex models. Since the interaction terms explain a similar portion of the variance, the model without interaction terms has almost as much predictive power with less complexity. Nevertheless, the interaction terms provide valuable information on how the variables interact. Furthermore, these moderating interactions are important to test hypotheses four, five and six. I now turn to the discussion of the results of testing these respectively.

(30)

number of anti-takeover devices. This explains why no evidence is found of the main relation, while evidence is found of the positive moderating effect of audit committee independence, based on the results of testing hypothesis four.

Similarly, results of examining the moderating effect of board size on the relation between anti-takeover devices and environmental CSR performance reveals that board size positively moderates the effect of anti-takeover devices on environmental CSR performance. Although the overall main relation is not found, plotting the interaction reveals that firms with the smallest boards (in particular with five or less members) typically have increasingly lower environmental governance scores for a higher number of anti-takeover devices. Corporations with – the far more common – larger boards, typically indicate increasingly higher environmental governance scores in case of higher number of anti-takeover devices. This explains why no evidence is found of the main relation between the number of anti-takeover devices and CSR performance in the dimension environmental, while evidence is found of the positive moderating effect of board size. These findings, thus, partly provide evidence for the hypothesis that board metrics positively moderate the relation between the number of anti-takeover devices and environmental CSR performance.

According to hypothesis five, a positive moderating effect of board metrics is expected on the relation between the number of anti-takeover devices and social CSR performance. Testing the hypothesis delivers no evidence for such an effect. This is not in line with the expectation that is based on the entrenchment hypothesis of Cespa and Cestone (2007) and the earlier empirical international study by Surroca and Tribó (2008).

(31)

number of board meetings on the relation between the number of anti-takeover devices and CSR performance in the dimension governance.

The negative moderating relation of auditing committee independence on the relation between anti-takeover and performance in the CSR dimension governance can be interpreted as follows. This result implies that higher levels of auditing committee independence weaken the negative relation between the number of anti-takeover devices and governance CSR performance. Examination of plots of the interaction delivers consistent findings. I recur to these findings in the subsequent chapter that presents the conclusion and discussion, now turning to the robustness tests.

5.2 Robustness tests

(32)

TABLE 7

Robustness test with the average score on all three CSR dimensions as dependent variable

(t+1) (t+1) (t+1) (t+1) (t+1) (t+1) (t+1) Model 1 2 3 4 5 6 7 Independent variables ATD ,029 ,068 *** -,080 ** -,080 ** -,081 ** -,081 ** -,082 ** ACI ,014 ,019 ,022 ,015 ,014 ,015 ,014 BSZ ,361 *** ,193 *** ,192 *** ,192 *** ,193 *** ,193 *** ,194 *** DIV ,194 *** ,126 *** ,098 *** ,098 *** ,096 *** ,096 *** ,097 *** NBM ,053 ** ,025 ,028 ,028 ,028 ,028 ,027 ACI*ATD ,011 ,010 ,010 ,010 BSZ*ATD ,041 ** ,041 ** ,044 ** DIV*ATD -,003 _-,002 NBM*ATD -,023

CEX ,246 *** ,285 *** ,286 *** ,279 *** ,279 *** ,277 *** LEV ,110 *** ,065 ** ,065 ** ,067 ** ,067 ** ,066 ** RDE ,109 *** ,128 *** ,127 *** ,129 *** ,129 *** ,127 *** ROA ,108 *** ,071 *** ,071 *** ,073 *** ,073 *** ,073 *** SIZE -,009 ,001 ,001 ,004 ,004 ,007 Fixed effects

N 1773 ₁₇₇₃ 1773 1773 1773 ₁₇₇₃ ₁₇₇₃ A. R sq. ,204 ,318 ,398 ,397 ,399 ,398 ,399 D. R sq. ,206 ,116 ,087 ,000 ,002 ,000 ,000 VIF 1,057 _7,390 8,226 8,231 _8,237 _8,238 _8,250

This table presents the first robustness test. The dependent variable is the one-year-lagged average corporate social responsibility (CSR) performance score of the three dimensions environmental (ENV), social (SOC) and governance (GOV). Estimations are based on ordinary least squares techniques, with fixed-effects for industries and years from model three and onwards. N reflects the number of observed firm-years, A. R square abbreviates adjusted R square measure, D. R sq. reflects the change (delta) of the adjusted R square compared to the previous model and VIF is short for maximum variance inflation factor. Full names of the abbreviated variables are presented in table one. A more extensive overview of the variables, the descriptive statistics and Pearson-correlations are presented in tables one, two and three respectively. Symbols ***, ** and * indicate 1, 5 and 10 % significance levels respectively using two sided-testing.

(33)

The results of testing the main relation – between the number of anti-takeover devices and CSR performance – of the first robustness test are in line with the findings of the overall main research results.

No evidence is found of a moderating effect of auditing committee independence, board diversity and number of board meetings on the main relation. Not finding evidence for the latter two variables is in line with the findings in table four, five and six, presenting the main research results for the three CSR dimensions respectively. Not finding evidence for the moderating relation of auditing committee independence is also in line with the findings for social CSR performance. In the dimensions environmental and governance, however, significant findings indicate a positive and negative moderating effect respectively.

The first robustness test does find evidence for a positive moderating effect of board size. This is not in line with the findings for environmental and governance CSR dimensions. In the social CSR dimension no evidence is found for the moderating effect. The robustness test for general CSR performance is, thus, in line with the findings of the social CSR dimension for this variable. The findings of this first robustness test are in line with the main research findings. It should be noted that this was to be expected, given the fact that the general CSR performance robustness test is estimating an average of the three separate dimensions. Nevertheless, the test provides additional insight on the robustness of the findings when the dimensions are aggregated. Since the findings of the robustness test are in line with the main research results, I now turn to the second robustness test.

Table eight provides the results of the second robustness test. In this test the final main regression model is re-examined. Contrary to the original regression equation, the control variables are scaled for size using total assets. To avoid multicollinearity issues as a result of scaling, this variable is not included in the regression model as a separate control variable in the second robustness test.

(34)

TABLE 8

Robustness test for using controls scaled for total assets on CSR dimension governance

GOV GOV GOV GOV GOV GOV GOV

(t+1) (t+1) (t+1) (t+1) (t+1) (t+1) (t+1) Model 1 2 3 4 5 6 7 Independent variables ATD -,113 *** -,107 *** -,384 *** -,381 *** -,381 *** -,381 *** -,382 *** ACI ,086 *** ,088 *** ,087 *** ,135 *** ,135 *** ,136 *** ,135 *** BSZ ,174 *** ,164 *** ,182 *** ,183 *** ,185 *** ,186 *** ,186 *** DIV ,140 *** ,136 *** ,142 *** ,141 *** ,139 *** ,140 *** ,141 *** NBM ,057 ** ,074 *** ,065 *** ,065 *** ,066 *** ,066 *** ,063 *** ACI*ATD -,068 ** -,069 ** -,071 ** -,069 ** BSZ*ATD ,038 * ,041 * ,045 ** DIV*ATD -,022 _-,021 NBM*ATD -,038 *

CEX ,031 ,016 ,015 ,011 ,011 ,010 LEV ,015 ,019 ,024 ,027 ,026 ,023 RDE -,027 -,019 -,013 -,011 _-,013 _-,015 ROA ,107 *** ,085 *** ,084 *** ,086 *** ,084 *** ,084 *** Fixed effects

N 1773 ₁₇₇₃ 1773 1773 1773 ₁₇₇₃ ₁₇₇₃ A. R sq. ,083 ,094 ,163 ,164 ,165 ,165 ,166 D. R sq. ,085 ,013 ,080 ,002 ,001 ,000 ,001 VIF 1,057 _2,322 6,245 6,256 6,297 _6,311 _6,340

(35)

The differences between the findings of the second robustness test and those of the final main regression model (equation 3) are roughly similar. These are presented in table eight and table six respectively. The robustness test finds evidence of a negative moderating effect of the number of board meetings on the relation between anti-takeover devices and CSR performance in the dimension governance. Apart from not finding the evidence for control variable capital expenditure and leverage, that indicated a weak positively significant relation in just the first model that doesn’t account for the industry differences, the main results are similar.

This entails that consistent with the final main research model, the second robustness test show a highly significant negative relation between the number of anti-takeover devices and CSR performance in the dimension governance. Furthermore, evidence is found of a negative effect of auditing committee independence and a positive effect of board size on the relation between the number of anti-takeover devices and governance CSR performance. These findings are line with the main research results. No evidence is found of a moderating effect of board diversity on the main relation.

(36)

VI. CONCLUSION AND DISCUSSION

This paper tests the entrenchment hypothesis of Cespa and Cestone (2007) with a focus on the role of anti-takeover devices, the market for corporate control and boards in the efficient US market. The relation between anti-takeover devices and CSR performance is examined, since the latter aspect is expected to increase a result of collusion between managers and stakeholders. Because installment of anti-takeover devices and collusion with stakeholders are part of the typical entrenchment strategy, these actions are anticipated to be pursued in conjunction by opportunistic managers (Cespa and Cestone, 2007). Moreover, the moderating effect of the internal governance function of boards is studied on the aforementioned relation. Analyzing the model for three separate dimensions of CSR performance, was aimed at exploring which type of stakeholder interests are used for collusion. This can provide more insight on whether environmental-, social- or governance performance interested stakeholders are typically colluded with.

The results of the statistical tests using ordinary least squares techniques, provide mixed evidence. No evidence is found for an expected positive overall relation between the number of anti-takeover devices and performance in any of the three corporate social responsibility dimensions. Interestingly, a negative relation is found between the number of anti-takeover devices and performance in the CSR pillar governance. I now relate these findings to academic literature.

(37)

Another potential reason for not finding a positive relation between anti-takeover devices and performance in CSR dimensions in the US, is caused by market efficiency. Although this efficiency that results from the common law legal origins of the country, makes the market for corporate control more prevalent, it may reduce information asymmetry in the capital market (Surroca and Tribo, 2008). As a result of the higher efficiency of the market, investors may be better able to monitor whether CSR projects are costly collusion efforts, or investments with a positive net present value. The ability of investors to better assess the nature of CSR projects, makes collusion through CSR hard to maintain unobserved. According to the market for corporate control, corporate raiders that see through the nature of costly CSR collusion, are incentivized to takeover the firm to replace management and change strategy with the aim of realizing the profits. This means that in case markets are too efficient to hide collusion through CSR from investors, such a strategy may be unsustainable in this context. Since the US is characterized by very efficient markets, the inability to hide collusion may offset the benefits of testing the hypothesis in this context. These two explanations provide potential reasons why no evidence is found for a positive relation between takeover devices and performance in CSR dimensions. It does, however, not explain why anti-takeover devices relate negatively to performance in the CSR dimension governance. I recur to this below.

(38)

mechanism in line with the paper of Surroca and Tribó (2008), no evidence is found on the basis of examining the board metrics board diversity and the number of board meetings. Similarly, no evidence is found of an interaction effect of board quality metrics in the social CSR dimension. The latter results are not in line with previous literature (Cespa and Cestone, 2007; Surroca and Tribó, 2008). It is unclear why no evidence is found of this relation. I recur to this issue in my suggestions for future research below.

I now discuss the results for the CSR dimension governance in more detail, subsequently turning to the moderating effect of board quality metrics on the relation between the number of anti-takeover devices and performance in this CSR pillar. The expected positive relation between the number of anti-takeover devices and CSR performance in the dimension governance is not found. On the contrary, the results imply that there is a negative relation between the independent and dependent variable. This finding is not in line with previous research on the entrenchment hypothesis by Cespa and Cestone (2007) and Surroca and Tribó (2008). These papers, however, do not examine the relation for the separate CSR dimensions. Although the negative relation between the number of anti-takeover devices and governance CSR performance is not in line with the hypothesis, there is an intuitive potential explanation for this negative relation, that I recommend for future research.

This potential explanation resides in what CSR performance in the dimension governance actually entails. It is based on the notion of Surroca and Tribó (2008) that reduced discretion of managers may limit abilities for collusion. As discussed in chapter two of this paper, CSR governance criteria address concerns including proper design of management structure, matters concerning fairness of compensation and other corporate governance criteria such as the quality of internal controls and audits.

(39)

Since maintaining power motivates managers who install anti-takeover devices to entrench themselves and to collude with stakeholders in the first place, they will not seek score high in the governance CSR dimension. On the contrary, this potential explanation would imply that managers that entrench themselves would typically perform lower in this dimension. This could explain why the regression results imply a negative relation between the number of anti-takeover devices and CSR performance in the dimension governance. I recur to this in my suggestions for future research and now turn to the moderating relation of board quality metrics.

The negative moderating relation of auditing independence can potentially also be explained by the nature of the governance dimension of CSR. Managers that entrench themselves with anti-takeover devices typically seek to maintain their power, thus, seeking to score low in the governance dimension of CSR. Independent auditing committees, however, may promote high quality internal controls and audits. In this case, more independent audit committees reduce the ability of entrenching managers to limit internal controls and audits. In other words, more independent auditing committees could reduce the negative effect of managerial entrenchment on governance CSR performance. This is a potential explanation for a negative moderating relation of auditing committee independence. This negative moderating relation is not in line with earlier papers on the entrenchment hypothesis (Cespa and Cestone, 2007; Surroca and Tribó, 2008). I recur to this possibility in my suggestions for future research.

The findings indicate that board size positively moderates a relation between the number of anti-takeover devices and governance CSR performance. This implies that a higher number of members on a board typically increases the negative relation between anti-takeover devices and governance CSR performance. The positive moderating effect is in line with the hypothesis and previous literature (Surroca and Tribó, 2008). The results imply that – contrary to the hypothesis – the underlying main relation is negative. I recur to this finding in my suggestions for future research.

(40)

positively moderate the effect of the number of anti-takeover devices on performance in the CSR dimensions environmental and governance. Audit committee independence moderates the effect of the number of anti-takeover devices and performance in these CSR dimensions positive and negative, respectively. Additional robustness test imply that the findings are robust. These findings help to defragment the literature on anti-takeover devices, boards, CSR and the entrenchment hypothesis in a context with efficient markets. In particular, examining the CSR dimensions separately is an important addition to the Surroca and Tribó (2008) paper, since the results imply that entrenching managers perform differently on the certain CSR dimensions. In particular, these could potentially seek to avoid scoring high in the governance dimension to maintain their power in the firm.

Several limitations of this study can be identified. This paper tested the entrenchment hypothesis, focusing on entrenchment with anti-takeover devices. Therefore, other types of entrenchment are not investigated. Future studies may include a broader range of entrenchment proxies to examine alternative types of entrenchment. Furthermore, CSR is argued to be the only means for collusion with stakeholders, although in practice managers may find other ways to collude. The moderating relation of board quality metrics is studied, using four proxies for quality. The actual quality of boards may be hard to quantify and may require a far more inclusive set of proxies. All firms in the S&P 500 a were incorporated, operating in a wide range of industries. Excluding certain industries might deliver different outcomes. Large listed US firms were the object of study in this paper. As a result, the findings are not necessarily generalizable to other contexts, including smaller or unlisted firms and corporations in other countries.