The influence of outside directors on M&A activity
Name: Robin Janssen Student number: 11152540
Thesis supervisor: V. R. O’ Connell Date: 19th of June, 2017
Word count: 13,019
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Robin Janssen who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This study examines the relationship between outside directors and M&A activity. More specifically, I focus on the influence of the proportion of outside directors in a firm on deal occurrence, relative deal size and subsequent goodwill impairments. Outside directors are defined as affiliated and non-affiliated outside directors, as well as a combined measure of both. The sample is drawn from the S&P 1500 for the years 1998 through 2006 and consists of 7,307 firm-year observations and 3,661 related M&A transactions. The results show that relative deal size of M&A transactions decreased following the adoption of SEC regulations in 2004. These regulations require for NYSE and Nasdaq firm that a majority of the board of a listed firm is a non-affiliated outside director. However, there is no indication that the proportion of outside directors also influences the occurrence of completed M&A deals. Furthermore, it seems that there is a positive relationship between the proportion of outside directors and goodwill impairments within two years of a completed deal. Boards with a greater proportion of outside directors tend to recognize more goodwill impairments. This study contributes to research on outside directors, regulatory shocks and goodwill impairment.
Acknowledgements
I would like to express my gratitude to my supervisor prof. d.r. V.R. O’Connell, who provided great feedback during the process and helped me developing and structuring the core elements of this thesis. I would also like to thank my supervisor at KPMG, Thijs Ravesloot. He guided me during my internship at Audit FS. Conducting scientific research has been quite tough, but when I look back I learned a lot during the process. I hope this will be useful to me in my future professional career.
Table of Content
Abstract ... 3
Acknowledgements ... 4
1 Introduction ... 6
2 Theory and literature ... 9
2.1 Agency theory ... 9
2.2 Mergers and acquisitions ... 10
2.3 Outside directors ... 11
2.4 Goodwill and goodwill impairment ... 14
3 Data ... 16
3.1 Sample size and time period ... 16
3.2 Dependent variables ... 17 3.2.1 M&A activity... 17 3.2.2 Goodwill impairment ... 17 3.3 Independent variables ... 18 3.3.1 Outside directors ... 18 3.3.2 Regulatory shock ... 18 3.4 Control variables ... 19 3.5 Data collection ... 20 4 Research methodology ... 22 4.1 Hypothesis 1 ... 22 4.2 Hypothesis 2 ... 23 4.3 Hypothesis 3 ... 24 5 Results ... 25 5.1 Descriptive statistics ... 25 5.2 Correlation ... 27 5.3 Regression results ... 29 5.3.1 OLS regression ... 30 5.3.2 Probit regression ... 32
5.4 Summary of main findings ... 34
6 Conclusion and discussion ... 35
1 Introduction
How valuable are outside directors on a board? CEOs have the opportunity to choose their own outside directors, while these directors are hired to monitor the CEO in the first place (Hermalin & Weisbach, 2001). This study examines the relationship between outside directors on a board and M&A activity. More specifically, I focus on the influence of the proportion of outside directors in a firm on deal occurrence, relative deal size and subsequent goodwill impairments.
This study builds on agency theory (Jensen & Meckling, 1976). Agency theory consist of the theory of agency, theory of property rights and the theory of finance. Jensen & Meckling (1976) combined them in order to develop a theory of the ownership structure of the firm. Agency theory focuses on the relationship between the group that delegates work and the group that accepts to perform the work, respectively the principal and the agent. In a firm setting, the agent is the manager, whereas the principal is the shareholder. Individuals try to maximize their own self-interest (Jensen & Meckling, 1976) and prior research argues that managers pursue their own interest through mergers and acquisitions (Gormley & Matsa, 2011). One of the constraints of agency theory is that monitoring is assumed to be costly and ineffective (Perrow, 1986).
Dey (2008) argues that a board with more outside directors is considered to be more effective in monitoring executive management. Outside directors have a board seat, but are supposed to perform only supervisory tasks. These directors can be either non-affiliated outside directors or affiliated outside directors. A distinction between these two is made based on their relationship with the firm. Affiliated outside directors are not employees and cannot exercise control over the firm, but they have a material relationship with the firm due to former employment, a family relationship with the CEO or any other financial relationship. Non-affiliated outside directors have no material connection to the firm other than a board seat. Furthermore, Weisbach (1988) argues that a corporate board with a higher proportion of outside directors is more likely to replace a firm’s CEO after a period of poor firm performance. These findings seem to suggest that firms with more outside directors on the board are able to improve monitoring abilities and thus, decrease agency problems.
Armstrong, Core and Guay (2014) found a positive relationship between outside directors and corporate transparency. They argue that increased corporate transparency results in a better information environment for outside directors. This could help them perform their monitoring tasks better. However, the causality of this direction is still unclear. A growing literature documents that when information asymmetry is high and information transfer costly, firms tend to choose to have less outside directors (Linck, Netter & Yang, 2008; Maug, 1997).
Prior literature found evidence of managers acquiring other firms for various reasons not aligned with shareholders’ interests (Gormley & Matsa, 2011; Jensen & Meckling, 1976; Roll, 1986; Trautwein, 1990). If managers can be monitored more effectively with a higher proportion of outside directors on the board, it will be less likely that they will make M&A decisions not beneficial to the firm. This leads to the research question: is there a negative relationship between M&A activity and the proportion of outside directors on the board?
I measure M&A activity with deal occurrence, relative deal size and subsequent goodwill impairments. I also examine this relationship prior- and post to a regulatory shock. The regulatory shock in this study is the approval of new regulations proposed by the NYSE and Nasdaq in 2003. These regulations affect board composition and board committees of listed companies, because firms were forced to increase the amount of non-affiliated outside board members to a majority. I use this regulatory shock to measure whether the relationship between outside directors and M&A activity is different after the adoption of these regulations is 2004.
The results shows that relative deal size of M&A transactions decreased following the adoption of SEC regulations in 2004. However, there is no indication that the proportion of outside directors also influences the occurrence of completed deals. This indicates that there is no robust evidence of a significant relationship between outside directors and M&A activity. In addition, I found a positive relationship between the proportion of outside directors and goodwill impairments within two years of a completed deal. It seems that boards with a greater proportion of outside directors tend to recognize more goodwill impairments. This is not consistent with the expectations in this study and may be the case due to unforeseen limitations.
The purpose and influence of outside directors and other board characteristics has been studied extensively in the past, but not in relation to mergers and acquisitions. This study supports the assumption that is common in the financial reporting and disclosure literature that outside directors can and do influence corporate transparency (Armstrong et al., 2014). Duchin, Matsusaka, and Ozbas (2010) suggest that future research should focus on the influence of the information environment on the relationship between board behavior and decision-making.
This study provides evidence on the influence of outside directors on M&A activity. Furthermore, it adds to the existing literature on the effectiveness of the monitoring role of outside directors, with regards to merger and acquisition activity. This study might be helpful as well to regulators and standard setters, because of the incorporation of a regulatory shock. Lastly, it also explains the relationship between outside directors and goodwill impairments, which is a relatively new research area.
The study is structured as follows. The next section explains the underlying theory and literature. The data is discussed in Section 3. The research methodology is included in Section 4. Results are reported in Section 5. Section 6 contains the conclusion and discussion. The references are included in the last section.
2 Theory and literature
This study builds primarily on agency theory (Jensen & Meckling, 1976). I will also discuss prior literature on managerial behavior, outside directors, M&A activity and regulatory shocks.
2.1 Agency theory
During the 1960s and 1970s, economists explored risk sharing among individuals or groups (Jensen & Meckling, 1976; Wilson, 1968). These studies explain the risk-sharing problem as one that arises when cooperating parties have different attitudes toward risks. Agency theory broadened this point of view by including the agency problem that occurs when cooperating parties have different goals and division of labor (Jensen & Meckling, 1976). More specifically, agency theory is directed at the relationship between the group that delegates work and the other group that accepts to perform the work. There are three assumptions underlying to this theory. The first assumption is the most common to economists: individuals try to maximize their own self-interest. The second assumption is more specific to agency theory: social life is a series of exchanges, governed by competitive self-interest. The third assumption applies to costly and ineffective monitoring contracts (Jensen & Meckling, 1976; Perrow, 1986).
Furthermore, agency theory is concerned with resolving two problems that occur in agency relationships. The first problem arises when the goals of the principal and agent are not aligned and it is costly for the principal to verify what the agent is actually doing. The second problem occurs when the principal and the agent have different attitudes toward risk (Eisenhardt, 1989).
Jensen & Meckling (1976) use the theory of agency, theory of property rights and the theory of finance to develop a theory of the ownership structure of the firm. The theory helps explain why a manager will choose a set of activities for the firm such that the total value of the firm is less than it would be if he were the sole owner. Although it is being used in total different research settings and practices, it is still surrounded by controversy. The detractors call it trivial, dehumanizing and even dangerous. Perrow (1986) argues that contracts can be violated because of the ineffectiveness of monitoring and will be violated because of the ineffectiveness of monitoring. This gives managers, who are prone to cheating, a chance to do so.
The principal delegates decision rights to the agent. This separation of ownership and control introduces agency problems (Jensen & Meckling, 1976). This is caused by misalignment of interests between principal and agent, where the agent may prefer other activities over productive effort. This productive effort can be triggered by aligning incentives and monitoring. This results
in agency costs, which can be divided into two types. Firstly, agency costs of equity are the sum of monitoring costs of the principal, bonding costs by the agent and the so-called residual loss. Secondly, the agency costs of debt include monitoring costs that follow from incentive effects associated with high leverage and bankruptcy costs (Jensen & Meckling, 1976).
This study will focus on the incentive effects associated with leveraged firms. Managers in such firms have the incentive to undertake risky projects with high pay-off and low likelihood of success. They can benefit from the upside potential of their decisions and are limited exposed to the downside, because of their limited liability. This is caused by the separation of ownership and control. This behavior has also been linked to mergers and acquisitions by Trautwein (1990), arguing that managers try to maximize their own utility through mergers. This approach assumes that managerial goals are the explanatory factor behind a merger.
2.2 Mergers and acquisitions
Mergers and acquisitions are transactions in which the assets of one firm are transferred to or combined with the assets of another firm. A substantial amount of research has been conducted in the area of mergers and acquisitions, where most studies focus on managers’ incentives to undertake acquisitions (Miller, Le Breton-Miller & Lester, 2010; Roll, 1986; Trautwein, 1990). Other scholars have examined M&A from a wealth perspective (Healy, Palepu & Ruback, 1992; Jensen & Ruback, 1983; Owen & Yawson, 2010).
Roll (1986) introduced the overconfidence approach to M&A with his hubris theory of acquisitions. Hubris theory is the unrealistic perception by managers that they can manage the assets of a target firm more efficiently than firms' current management (Roll, 1986). Further research is also unmistakable on CEO behavior, stating that CEOs are overconfident and undertake mergers and acquisitions that are value-reducing (Black, 1989; Jensen, 1986; Malmendier & Tate, 2004). They argue that managers overpay for targets because they are overly optimistic and because there exists a divergence of interests between them and the firm’s shareholders.
For instance, Gormley and Matsa (2011) show that managers respond to increased risk arising from employees’ exposure to newly identified carcinogens. These carcinogens increase the probability of a bankruptcy of the firm. They respond by acquiring large, unrelated businesses with relatively high operating cash flows, which can be assumed to be inefficient acquisitions with a negative net present value (Gormley & Matsa, 2011). This form of M&A activity is driven by the agency conflict and also associated with negative abnormal returns, because the goals of the principal and the agent are not aligned. This is imperfectly observable for the principal, which
allows the agent to pursue his self-interests. This action by the agent has negative consequences for the principal, but also leads to risk reduction for the manager himself (Gormley & Matsa, 2011).
Motis (2007) sums up a list of other motives and determinants of M&A activity. His view is consistent with the earlier mentioned approach that makes a distinction between M&A activity motivated by either shareholder gains or managerial gains. He refers to shareholder gains as mergers that lead to an increase in the market value of the firm, whereas managerial gains is referred to as the incentive effects associated with the separation of ownership and control. Most of the motives associated with managerial gains are discussed above. Motis (2007) argues that M&A activity motivated by shareholder gains consists of four types, namely (a) efficiency gains (b) synergy gains (c) cost savings (d) financial cost savings. Efficiency gains mainly focus on economies of scale and economies of scope, as well as economies of vertical integration. Synergy gains arise due to the diffusion of knowledge and R&D. Cost savings can be achieved through the creation of internal capital markets, as well as through purchasing power of the newly combined firm. Lastly, financial cost savings are either tax advantages or lower interest rates.
2.3 Outside directors
As mentioned earlier, a board with a greater proportion of outside directors is considered to be more effective in monitoring management (Dey, 2008). Outside directors have a board seat, but are supposed to perform only supervisory tasks. These directors can be either affiliated outside directors or non-affiliated directors. Affiliated outside directors are not employees and cannot exercise control over the firm, but they have a material relationship with the firm due to former employment, a family relationship with the CEO or any other financial relationship. Non-affiliated outside directors have no material connection to the firm other than a board seat. They are considered to be independent from management and free from any business or other relationship that could materially interfere with the exercise of their independent judgement.
The importance of the information and advising function of outside directors has been emphasized widely in recent literature (Balsmeier, Buchwald & Stiebale, 2011; Faleye, Hoitash & Hoitash, 2011). Inside directors provide valuable information about the firm’s activities, while outside directors may contribute both expertise and objectivity in evaluating managers’ decision making (Byrd & Hickman, 1991). Winter (1977) argues that outside directors can insist on proper auditing procedures and review corporate decisions to manage risks. They can also provide a different perspective on certain topics and can ask tough questions of managers. Furthermore, outside directors may have family and/or business relationships with the company. Such
relationships with a firm provide incentives to directors to endorse certain decisions that they otherwise wouldn’t (Dey, 2008). For example, Hermalin and Weisbach (1998) argue that outside directors are more independent than other board members, because their personal future careers do not depend on successes of their direct board colleagues and the CEO. Ferris, Jagannathan and Pritchard (2003) also finds that more experienced outside directors reflect more intensive monitoring. Other scholars report positive abnormal stock returns on the announcement of outside director appointments (Fich, 2005). Weisbach (1988) finds that a board with a greater proportion of non-affiliated outside directors is more likely to replace a firm’s CEO after a period of poor firm performance. This indicates that non-affiliated outside directors do have a significant influence on decision making by applying their independent monitoring abilities. These findings do not apply to affiliated outside directors.
Armstrong et al. (2014) found a positive relationship between outside directors and corporate transparency. Increased corporate transparency results in a better information environment for outside directors. This will help them perform their monitoring tasks more appropriately. The direction of this causality is unclear, because the relationship between these two variables can be explained in two different ways.
At first, corporate transparency increases after the proportion of outside directors on the board increases, because these outside directors fulfil their monitoring tasks and therefore encourage management to be more transparent on corporate activities. Another reason is that outsiders to the firm generally have to acquire and process a substantial amount of firm specific information to effectively perform their duties (Armstrong et al., 2014). Hence, the result of the increased transparency is due to the information needs of the outside director him/herself.
On the other hand, one can argue that higher corporate transparency leads to a higher proportion of outside directors on the board. A growing literature documents that when information asymmetry is high and information transfer and processing costs are high, these firms generally choose to have relatively few outside directors (Linck et al., 2008; Maug, 1997). In that case, firms with an increasing corporate transparency will choose to appoint more outside directors. Nevertheless, Armstrong et al. (2014) found a positive relationship between corporate transparency and outside directors. Therefore, managers will be monitored more effectively with a greater proportion of outside directors on the board.
According to Helland and Sykuta (2005), the proportion of outside directors is also associated with firm performance. They use the probability of a firm being sued by their
shareholders as a measure of firm performance. Their results suggest that boards with a greater proportion of outside directors do a significantly better job of monitoring management. Also Chen, Cussatt and Gunny (2017) examined the disciplinary role of outside directors in overseeing the CEO. They find that more independent boards are more effective monitors of real earnings management. McDonald, Westphal and Graebner (2008) found evidence that supports that outside directors deliver great benefits to independent boards. They focus on outside director expertise in acquisition decision making and its implications for the performance of these acquisitions (McDonald et al., 2008). These findings seem to imply that CEOs need the disciplinary role of outside directors in making M&A decisions that are beneficial to the firm and the firm’s performance.
As we have seen earlier, prior literature found evidence of managers acquiring other firms for various reasons not aligned with shareholders’ interests (Gormley & Matsa, 2011; Jensen & Meckling, 1976; Roll, 1986; Trautwein, 1990). If managers can be monitored more effectively with a higher proportion of outside directors on the board, it will be less likely that they will make mergers and acquisition decisions not beneficial to the firm. Therefore, I expect outside directors to have a negative relationship with M&A activity and I will examine this with a one-tailed test. H1: There is a negative relationship between M&A activity and the proportion of outside directors on a board.
During this research, I will also observe the effects of a regulatory shock on the relationship between outside directors and M&A activity. The Securities and Exchange Commission (SEC) approved new regulations proposed by the NYSE and Nasdaq in 2003. These regulations affect board composition and board committees of listed companies, because firms were forced to increase the amount of non-affiliated outside board members to a majority. Similar to Duchin, Matsusaka and Ozbas (2010) and Armstrong et al. (2014), I use these new regulations as an exogenous shock event that caused significant differences in the proportion of outside directors on boards. Prior research predicts that the regulatory shock should lead to an increase in outside directors on a firm’s board (Armstrong et al., 2014; Duchin et al., 2010). Possibly, firms already had a majority of non-affiliated directors on the board. I examine whether this exogenous shock has a negative impact on the relationship between M&A activity and outside directors.
H2: The adoption of SEC regulations in 2004 had a negative influence on the relationship between M&A activity and outside directors.
2.4 Goodwill and goodwill impairment
Goodwill is an intangible asset on the balance sheet that arises when firms acquire other businesses for a premium value. Goodwill consists of the value of the firm’s brand name, customer base, patents and any other assets that are not capable of being separated or divided from the entity and sold. The amount the acquiring firm pays over the target’s book value is assumed to be the value of the acquired goodwill.
Goodwill impairment is a charge that firms record when a reduction in fair value of goodwill from acquired business units occurs. This goodwill impairment test is a relatively new phenomenon and replaces goodwill amortization. Under the amortization approach, a periodic amortization charge is recorded on a straight-line basis over the assumed useful life of the intangible asset. Goodwill is no longer amortized since the issuance of SFAS 142 in 2001. It is now subject to annual testing to determine if any impairment indicators occurred. Firms are required to determine the fair value and carrying value of the goodwill on the balance sheet. If the fair value drops below carrying value, an impairment loss should be recognized in the income statement and the goodwill value needs to be reduced. The changes included in SFAS 142 will improve financial reporting, because the financial statements will reflect the underlying economics of the goodwill better.
Events that may trigger goodwill impairment include deterioration in economic conditions, increased competition, loss of key personnel and new regulations. According to prior literature, goodwill impairments mainly result from poor acquisition decisions in the past (Gu & Lev, 2011; Hayn & Hughes, 2006). Therefore, I will focus on goodwill impairments in the third hypothesis to assess to what extent M&A deals are considered to be value-reducing, within the following two years of a deal. I'm wondering if there is a declining trend observable in the amount of goodwill impairments with a higher proportion of outside directors on a board. I will measure this with a one-tailed test.
H3: Subsequent goodwill impairments will occur less with a greater proportion of outside directors on a board.
However, management could also have incentives to maximize or minimize goodwill impairment losses for various reasons. There is some evidence that managers use the discretion allowed by standards to decide about recording losses, by taking big-bath charges or by smoothing the earnings figures when they have the intention to do so (Beatty & Weber, 2006; Giner & Pardo, 2015; Riedl, 2004). Other research also finds that recognition of goodwill impairment losses significantly increases when firms have stronger corporate governance (Guler, 2007). It is also
expected that highly leveraged firms are less likely to record goodwill impairment losses in order to avoid violations of debt covenants.
A study by Schatt, Doukakis, Bessieux-Ollier and Walliser (2016) examined the usefulness of goodwill impairments provided by European firms. They argue that goodwill impairment is useful in one case and less adequate in another. The first case is the private information disclosure, because goodwill impairments may provide useful information to investors when they are not able to formulate precise expectations about future earnings. Private information is defined as information which is only known by management of the firm and not by investors and other stakeholders. Investors revise their expectations only if there is new information included in the goodwill impairment. Thus, only unexpected goodwill impairments contain useful information to investors. In order to understand which goodwill impairments are useful, investors should make a distinction between expected goodwill impairments and unexpected goodwill impairments.
The second case in Schatt et al. (2016)’s study is earnings management. As mentioned earlier, disclosed information might be useful to investors only when it is reliable. However, managers may also use the discretion allowed by the accounting standards to manage earnings. In that case, goodwill impairment will be considered as unreliable and useless information. The underlying reason for this is that managers have incentives to delay or not to impair goodwill, because it might imply that they paid an excessive price for an M&A deal. It also signals to investors that future earnings will be lower, which could have negative consequences for management in terms of their remuneration or employment (Schatt et al., 2016).
3 Data
In this section the data will be discussed. I explain the choices I made during the process, the measurement of the variables and the data collection.
3.1 Sample size and time period
The sample is restricted to all S&P 1500 firms between 1998 until 2006, regardless of whether they have been part of the S&P 1500 during the entire time period or not. I chose this time period to include a regulatory shock and to isolate the data from any influences due to the financial crisis.
During this research, I will also observe the effects of a regulatory shock on the relationship between non-affiliated outside directors and M&A activity. The Securities and Exchange Commission (SEC) approved new regulations proposed by the NYSE and Nasdaq in 2003. These regulations affect board composition and board committees of listed companies, because firms were forced to increase the amount of non-affiliated outside board members to a majority. Similar to Duchin, Matsusaka and Ozbas (2010) and Armstrong et al. (2014), I use these new regulations as an exogenous shock event that caused significant differences in the proportion of non-affiliated outside directors on boards. Therefore, this study will be free from endogeneity concerns.
These regulations require for Nasdaq and NYSE that a majority of the board of directors of a listed firm is non-affiliated. Firms were required to comply with these regulations by the earlier of: (1) the listed firm’s first annual shareholder meeting after January 15, 2004; or (2) October 31, 2004. The regulations do not apply to closed-end and open-end management investment companies that are registered under the Investment Company Act of 1940.
There are minor differences in the requirements between NYSE and Nasdaq. NYSE states that there should not be a material relationship between the firm and the outside director. This means that an outside director will not be considered non-affiliated if within the preceding three years (a) the director is an employee (b) an immediate family member holds a board seat (c) director has a material relationship with a former external auditor of the listed company (d) the director has an executive board seat of another firm that has a material relationship with the listed firm itself. Nasdaq only requires that there will be no interference with independent judgement. Their rules state that a non-affiliated outside director is a person other than an officer or employee or an individual with a relationship that is considered by the firms’ board to interfere with independent judgement. The requirements for both stock exchanges are more or less the same, except for three aspects. Firstly, outside director compensation should not be higher than $ 60,000 for Nasdaq firms, compared to $ 100,000 for NYSE. Secondly, under NYSE listing standards foreign private
issuers are not required to have a majority of non-affiliated outside directors on their boards of directors. Thirdly, firms with securities quoted on NYSE are required to compose audit, nominating and compensation committees entirely of non-affiliated outside directors. Nasdaq only requires this for audit committees. Although there are some minor differences, I will consider the requirements of both stock exchanges to be the same.
I measure the initial proportion of outside directors as of 1998. The ending point of the sample is 2006, to ensure that sufficient data is included prior and post to the regulatory shock and that influences of the global financial crisis are ruled out. This regulatory shock took place in 2003 and firms were required to adopt the new regulations in 2004. This divides the sample into two groups, namely prior to adoption (1998-2003) and post to adoption (2004-2006).
3.2 Dependent variables
This section will discuss the dependent variables in this study. M&A Activity will be measured in three ways, as explained in Sections 3.2.1 and 3.2.2.
3.2.1 M&A activity
Data on mergers and acquisitions has been collected from ThomsonReuters SDC, a database for global deal activity. In line with prior research, M&A activity includes the total value of all completed deals of majority interests above $ 1 million. A majority interest is an acquisition that leads to an interest of at least 50 percent of the total equity in the target firm. Group firms frequently use an incorporated subsidiary to complete acquisitions. For the purposes of this study, such acquisitions will be attributed to the ultimate listed firm (Tolmunen & Torstila, 2005). I measure this with a binary variable in the regression analysis. This dummy equals a value of 1 (one) if a deal has been executed and a 0 (zero) otherwise, similar to Boschma, Marrocu and Paci (2016). A second measure for M&A activity is relative deal size (DEALSIZE). Relative deal size is defined as total deal value divided by total assets in a specific year. By using two measures, I am able to examine whether the presence of outside directors influences the occurrence of deals, but also whether this presence influences the relative size of an M&A deal.
3.2.2 Goodwill impairment
Goodwill and goodwill impairment will be used to assess whether poor acquisition decisions are recognized earlier with a higher proportion of outside directors on the board. Goodwill impairment will be measured with a binary variable (DUMMYIMPAIR) which equals 1 for goodwill impairments greater than 10% of total goodwill on the beginning of the year balance
sheet within the same (year t), (year t + 1) and/or (year t + 2). I can examine then to what extent M&A deals appear to be value-reducing within the two years of a deal, similar to Hayn and Hughes (2006) and Gu and Lev (2011).
Dependent Variables Measurement Database
Deal activity (DUMMYDEAL) value of 1 if a deal > $ 1 million has been executed and zero otherwise
ThomsonOne SDC
Relative deal size (DEALSIZE) Total deal value divided by total
assets ThomsonOne SDC &
Compustat Goodwill impairment
(DUMMYIMPAIR) value of 1 if an impairment of >10% of total goodwill on balance sheet occurred in (year t), (year t + 1) and/or (year t + 2) following a deal, zero otherwise
ThomsonOne SDC & Compustat
3.3 Independent variables
This section will discuss the independent variables that will be used in this study.
3.3.1 Outside directors
As mentioned earlier, a distinction between affiliated and non-affiliated outside directors will be made. In this study, the proposed relationship will be tested using a combination of these two variables, but also as separate variables to examine whether being a non-affiliated outside director has a (more) significant influence on M&A activity. Therefore, I created three independent variables related to the presence of outside directors. Non-affiliated outside directors will be defined as %NA_OUTDIR and represents the portion of non-affiliated outside directors on the board. Affiliated outside directors are defined as %A_OUTDIR and will be measured in the same way as non-affiliated outside directors. Subsequently, %OUTDIR represents the total portion of outside directors, which is the sum of the earlier mentioned variables. This data will be retrieved from ISS/RiskMetrics, a database which contains information on the nature and characteristics of directors in S&P 1500 firms.
3.3.2 Regulatory shock
The second hypothesis focuses on the relationship between M&A activity and outside directors after a regulatory shock. As explained in Section 3.1, this regulatory shock forced firms to adopt these regulations – at the latest – in 2004. Therefore, I divide the total sample into two groups by
using a binary variable DUMMYREG. This variable defines whether the firm-year observation is before or after the adoption of the SEC regulations. The binary variable will equal a value of 1 in the years 2004 until 2006 and a zero in the years between 1998 and 2003.
Independent Variables Measurement Database
Non-affiliated outside directors
(%NA_OUTDIR) Non-affiliated outside directors divided by total board size ISS / RiskMetrics Affiliated outside directors
(%A_OUTDIR) Affiliated outside directors divided by total board size ISS / RiskMetrics Outside directors (%OUTDIR) Total outside directors divided by
total board size ISS / RiskMetrics Regulatory shock (DUMMYREG) value of 1 in the year 2004 until
2006 and a zero in the years between 1998 and 2003
ISS / RiskMetrics
3.4 Control variables
In the regression analysis, I will include several control variables. These variables have received empirical support in prior literature. I assume these variables to have an influence on the relationship between outside directors and M&A activity.
Board size (BSIZE) represents the total number of directors on a board. Higher board size is related to increased monitoring performance and therefore I control for this variable in the regression analysis (Duchin et al., 2010). The size of a firm is a significant determinant of acquisition likelihood, because larger firms tend to make more acquisitions (Harford, 1999). Therefore, I include the market-to-book-ratio (M2BOOK), as well as the natural logarithm of total assets (LNASSETS) (Kumar & Rajib, 2007; Armstrong et al., 2014). To control for other factors affecting monitoring abilities, also analyst coverage is included. Higher analyst coverage is associated with increased monitoring and therefore adds noise to the possible outcomes (Irani & Oesch, 2013). Firms with higher analyst coverage exhibit less characteristics of overinvesting. Analyst coverage (LNCOVERAGE) will be measured as the natural logarithm of the total amount of analyst forecasts in the specific year (Armstrong et al., 2014; Duchin et al., 2010) and will be retrieved from the I/B/E/S database. Furthermore, free cash flow theory argues that firms with a higher cash ratio are assumed to undertake more mergers and acquisitions (Jensen, 1986). Cash ratio (CRATIO) will be included as a control variable and defined as cash divided by total assets. Kumar & Rajib (2007) also use return on equity (ROE) in their regression model for M&A activity, hence I also add this variable to the model (Tolmunen and Torstila, 2005).
Control Variables Measurement Database
Board size (BSIZE) Total count of board
members ISS/Riskmetrics
Analyst coverage (LNCOVERAGE) Natural logarithm of count of EPS estimates
per year per firm I/B/E/S Market-to-book ratio (M2BOOK) Market value divided by
book value Compustat
Firm size (LNASSETS) Natural logarithm of
total assets Compustat
Cash ratio (CRATIO) Cash and equivalents
divided by total assets Compustat Return-on-equity (ROE) Net income divided by
total equity Compustat
3.5 Data collection
Director data has been collected from the ISS database between 1998 and 2006. All S&P 1500 firms within this sample period are included. These 2,495 different firms led to a total of 10,375 firm-year observations. The data rows consist of the Ticker code, year, CUSIP code and the board composition per firm, split into employee, affiliated and non-affiliated. First, I deleted the rows with missing CUSIPs and missing board data. This resulted in a minor loss of 197 observations, with 10,178 firm-year observations remaining.
The second step was retrieving M&A data from the Thomson SDC One database. I searched for all completed mergers and acquisitions between 1998 and 2006 with a US bidder firm. This identified 40,759 deals. Subsequently, I deleted all completed deals with missing transaction values and transaction values below 1 million dollar. A major part of the remaining 19,403 completed deals were executed by firms outside of S&P 1500 and by firms that had already been deleted during the sample selection process. This led to a loss of 15,742 completed deals after merging this data with the director data from ISS/Riskmetrics.
From this point, I used Compustat database to retrieve data for goodwill, goodwill impairment and control variables. This splits the data into two separate files, since there are a lot of missing goodwill values. These values are not necessary for the first and second hypothesis. For the first and second hypothesis, I included control variables like total assets, net income, total equity, current liabilities, market value and cash. Removing missing values and outliers reduced the initial sample size for the first and second hypothesis to 7,307 firm-year observations with 3,661
linked deals. I only removed the most severe outliers, which lead to a minor loss of 10 firm-year observations.
The third hypothesis also includes goodwill and goodwill impairment, next to the aforementioned control variables. Since I have a lot of missing goodwill values, I decided to create a separate data file for this hypothesis. This resulted in a sample for the third hypothesis of 4,705 firm-year observations and 137 subsequent goodwill impairments. After including control variables, deleting missing values and removing outliers I have 3920 firm-year observations remaining with 125 subsequent goodwill impairments.
Table 1A. Sample selection
Firm-year observations H1 and H2 H3
Total firm-year observations S&P 1500
1998-2006 board characteristics 10,375 10,375
Less: missing CUSIPs and incomplete board data (197) (197)
Remaining observations 10,178 10,178
Less: missing control variables and outliers (2,871) (2,871)
Remaining observations 7,307 7,307
Less: missing goodwill values and outliers (3,387)
Total 3,920
Table 1B. Sample selection
M&A data H1 and H2
All completed M&A deals 1998-2006 with US bidder 40,759 Less: missing deal values and values below $ 1 million (21,356)
Remaining observations 19,403
Less: firms that are not part of S&P 1500 and merging losses (15,742)
Total 3,661
Table 1C. Sample selection
Goodwill impairments H3
All goodwill impairments S&P 1500 1998-2006 Compustat 511
Less: merging losses (346)
Remaining observations 165
Less: impairments that did not meet dummy requirements (28)
Remaining observations 137
Less: missing values and outliers (12)
4 Research methodology
To examine whether M&A activity is influenced by the proportion of outside directors on the board, the following empirical models are derived. Each model for H1 and H2 uses two dependent variables, as well as two different treatments of the independent variables. H3 uses two models with one dependent variable and two different measures of outside directors.
4.1 Hypothesis 1
The two dependent variables aim to examine the relation between outside directors and M&A activity. M&A activity is measured in two ways, namely deal occurrence and the relative size of a deal. I conducted a probit regression and an Ordinary Least Squares regression to examine the relationship between outside directors and M&A activity. A probit model is a type of regression where the dependent variable can only take two values. The nature of the dependent variable for the H1 probit model (III) and (IV) is a binary variable. This binary variable can only take a value of 1 or zero, so a probit regression is suitable for these models. An OLS regression estimates the intercept and slope coefficients of the independent variables. R2 indicates the variance in M&A activity which is explained by the model. A random error component is included, which makes the relationship stochastic. The OLS regression will be performed for model (I) and (II) to examine the relationship between outside directors and relative deal size. First, I use relative deal size to examine the relationship with outside directors as a combined measure. Control variables are included in all models.
I. DEALSIZE = β0 + β1 * %OUTDIR + β2 * BSIZE + β3 * LNASSETS + β4 * M2BOOK + β5 * CRATIO
+ β6 * LNCOVERAGE + β7 * ROE + ε
Outside directors will be measured differently in the second model. The combined independent variable %OUTDIR is now separated into two variables, namely affiliated outside directors (%A_OUTDIR) and non-affiliated outside directors (%NA_OUTDIR).
II. DEALSIZE = β0 + β1 * %A_OUTDIR + β2 * %NA_OUTDIR + β3 * BSIZE + β4 * LNASSETS + β5
* M2BOOK + β6 * CRATIO + β7 * LNCOVERAGE + β8 * ROE + ε
The third and fourth model measure M&A activity with a second measure namely deal occurrence. This binary variable (DUMMYDEAL) is included to examine the relationship between outside directors and M&A activity with a probit model. Again, control variables are included.
III. DUMMYDEAL = β0 + β1 * %OUTDIR + β2 * BSIZE + β3 * LNASSETS + β4 * M2BOOK + β5 *
With the distinction between the treatment of independent variables, I try to assess whether there also exists a different relationship between outside directors as a combined measure and as outside directors separated into non-affiliated and affiliated outside directors. Therefore, model IV uses the separated measure. The results for hypothesis 1 are reported in Tables 7 and 8A.
IV. DUMMYDEAL = β0 + β1 * %A_OUTDIR + β2 * %NA_OUTDIR + β3 * BSIZE + β4 * LNASSETS
+ β5 * M2BOOK + β6 * CRATIO + β7 * LNCOVERAGE + β8 * ROE + ε 4.2 Hypothesis 2
In addition, the model for H2 incorporates an extra independent variable, namely DUMMYREG. The aim is to examine if the relationship between M&A activity and outside directors is different after adding this variable to the regression analysis. Control variables are included. The first model uses the combined measure %OUTDIR. The sample is divided into two groups with the binary variable DUMMYREG and I compare the group means by conducting a t-test. Furthermore, I use an OLS regression for model (I) and (II). A probit regression will be conducted for model (III) and (IV), since the dependent variable can only take two values. The results are shown in Tables 3, 7 and 8A.
I. DEALSIZE = β0 + β1 * %OUTDIR + β2 * DUMMYREG + β3 * BSIZE + β4 * LNASSETS + β5 *
M2BOOK + β6 * CRATIO + β7 * LNCOVERAGE + β8 * ROE + ε
Model (II) uses %A_OUTDIR and %NA_OUTDIR, instead of the combined measure of model (I). Again, control variables are included. Hence, the only difference between model (I) and (II) is the measurement of outside directors.
II. DEALSIZE = β0 + β1 * %A_OUTDIR + β2 * %NA_OUTDIR + β3 * DUMMYREG + β4 * BSIZE +
β5 * LNASSETS + β6 * M2BOOK + β7 * CRATIO + β8 * LNCOVERAGE + β9 * ROE + ε
Also for the second hypothesis, I use two measures of M&A activity. Model (III) and (IV) use deal occurrence as dependent variable. This will be tested with a probit regression. The third model uses the combined measure of %OUTDIR.
III. DUMMYDEAL = β0 + β1 * %OUTDIR + β2 * DUMMYREG + β3 * BSIZE + β4 * LNASSETS + β5
* M2BOOK + β6 * CRATIO + β7 * LNCOVERAGE + β8 * ROE + ε
Model (IV) also uses deal occurrence as the dependent variable, but defines outside directors as two separated measures, namely affiliated outside directors and non-affiliated outside directors. With this distinction, I assess whether there is a different relationship between the two measures of outside directors.
IV. DUMMYDEAL = β0 + β1 * %A_OUTDIR + β2 * %NA_OUTDIR + β3 * DUMMYREG + β4 * BSIZE
+ β5 * LNASSETS + β6 * M2BOOK + β7 * CRATIO + β8 * LNCOVERAGE + β9 * ROE + ε 4.3 Hypothesis 3
The third hypothesis includes goodwill impairment, which will be measured with a binary variable. Also here, the treatment of independent variables is defined as a separated and a combined measure. I focus on goodwill impairments in the third hypothesis to assess to what extent M&A deals are considered to be value-reducing, within the following two years of a deal. The dummy takes a value of 1 if an impairment of at least 10 percent of total goodwill on balance sheet occurred in (year t), (year t + 1) and/or (year t + 2) following a deal, zero otherwise. For the third hypothesis, I use two models. The first model combines both outside director measures. Except for DUMMYREG, all control variables are included. I test this relationship with a probit regression, which is typical to use in an empirical model with a dependent binary variable.
I. DUMMYIMPAIR = β0 + β1 * %OUTDIR + β2 * BSIZE + β3 * LNASSETS + β4 * M2BOOK + β5 *
CRATIO + β6 * LNCOVERAGE + β7 * ROE + ε
The second model separates %OUTDIR into affiliated outside directors and non-affiliated outside directors. Again, control variables are added to the model, so the only difference between model (I) and (II) is the measurement of outside directors. The results are shown in Table 8B.
II. DUMMYIMPAIR = β0 + β1 * %A_OUTDIR + β2 * %NA_OUTDIR + β3 * BSIZE + β4 * LNASSETS
5 Results
In this section, I will describe and explain the statistic tests I used to examine the proposed relationships.
5.1 Descriptive statistics
Table 2A shows the descriptive statistics for the first sample. The mean value of %NA_OUTDIR was 0.6703 and implies that 67 percent of board members were considered independent. This is a bit higher than the 60,36 percent reported by Duchin et al. (2010) when studying board characteristics in S&P 1500. However, that study focused on the time period 1996-2005 which is slightly earlier than this study’s time period. They also report an average board size of 9.55, which is higher than the average of 9.0562 in this study. This could imply that firms dismissed their affiliated outside directors in order to comply with board independence requirements, so board size decreases and percentage of independent directors increases. The DUMMYREG value of .3700561 indicates that most firm-year observations occur in the first period of this sample, prior to the adoption of new regulations.
Table 2A. Summary statistics H1+2
Variable Observations Mean p25 Median p75 SD DEALSIZE 7,307 .0530601 0 0 .0122076 .2405123 DUMMYDEAL 7,307 .3114821 0 0 1 .4631311 %OUTDIR 7,307 .8019406 .75 .83 .89 .1163708 %A_OUTDIR 7,307 .1314479 0 .11 .2 .1418023 %NA_OUTDIR 7,307 .6703134 .57 .7 .8 .0306001 DUMMYREG 7,307 .3700561 0 0 1 .4828525 BSIZE 7,307 9.056247 7 9 11 2.514777 LNASSETS 7,307 3.238778 2.764157 3.175685 3.645772 .6500721 M2BOOK 7,307 1.721404 .6778366 1.180587 2.057756 1.934624 CRATIO 7,307 .5678464 .0807401 .2384013 .63125124 1.24788 LNCOVERAGE 7,307 1.556035 1.30103 1.591065 1.857332 .4271341 ROE 7,307 .1102434 .0521497 .1194449 .1834163 .9575037
Table 2B shows descriptive statistics for the second sample including goodwill impairments. This sample contains less observations due to missing values in the goodwill data. However, the summary statistics in Table 2A and 2B do not differ significantly. For example, non-affiliated directors (%NA_OUTDIR) only increased by 0.1155 to .6818673. The dependent variable DUMMYIMPAIR takes a value of 1 if an impairment of >10% of total goodwill on balance sheet occurred in (year t), (year t + 1) and/or (year t + 2) of a deal, zero otherwise. In
Table 2B a very low mean of 0.0319 is observable, which is caused by the relatively low amount of subsequent goodwill impairments. This low mean value of DUMMYIMPAIR can be explained by two factors. Firstly, it is binary variable composed of goodwill impairments within one or two years after the deal. This means that all goodwill impairments that were not recognized within two years of a deal are excluded. Secondly, the standard that introduced goodwill impairment (SFAS 142) was only effective since 2001. Under certain circumstances it was possible to allocate some goodwill impairments to either 1998, 1999 or 2000, but not many firms made use of this.
Table 2B. Summary statistics
Variable Observations Mean p25 Median p75 SD DUMMYIMPAIR 3,920 .0318878 0 0 0 .1757237 %OUTDIR 3,920 .8128291 .75 .83 .89 .106554 %A_OUTDIR 3,920 .1306939 0 .11 .2 .1395563 %NA_OUTDIR 3,920 .6818673 .57 .71 .8 .1680909 BSIZE 3,920 9.393112 8 9 11 2.435135 LNASSETS 3,920 3.326031 2.859877 3.279016 3.719615 .6362628 M2BOOK 3,920 1.520337 .7345357 1.165898 1.875844 1.285106 CRATIO 3,920 .4669017 .0841784 .2196112 .5354866 .8498405 LNCOVERAGE 3,920 1.566144 1.322219 1.60206 1.863323 .4290401 ROE 3,920 .1154384 .0592975 .1239008 .1880122 .5583104
Table 2C shows summary statistics for outside directors. Total outside directors (%OUTDIR) is the sum of affiliated outside directors (%A_OUTDIR) and non-affiliated outside directors (%NA_OUTDIR). These variables increase and decrease in the expected direction. Proportion of outside directors on the board increases over the time period, as well as the proportion of non-affiliated outside directors. The proportion of affiliated outside directors on the board decreased, probably indicating that firms substituted their affiliated outside directors for non-affiliated outside directors in order to meet regulatory requirements. This is also supported by the data in Table 2C, indicating that the proportion of affiliated outside directors became lower over time.
Table 2C. Summary statistics
Year %OUTDIR %A_OUTDIR %NA_OUTDIR 1998 .771046 .1722176 .598159 1999 .7879 .1702667 .6174 2000 .7884466 .1477346 .6405825 2001 .8094678 .1376751 .6713725 2002 .800249 .1320124 .6682365 2003 .8129057 .1245472 .6882075 2004 .8256786 .1157679 .7095714 2005 .8316725 .115 .7163908 2006 .8377391 .1138957 .7234957 Average .8128291 .1306939 .6818673
Table 3 shows a t-test to assess whether there is a difference in group means observable in the period prior to and post-adoption. The results are significant at the 5% level for the variable DEALSIZE in the hypothesized direction, which indicates that relative deal size of M&A transactions decreased following the adoption of SEC regulations in 2004. Furthermore, the test highlights that there is no decrease observable in the amount of deals undertaken after the regulatory shock. There is no significant difference in the mean value of DUMMYDEAL between the two groups.
Table 3. T-test to compare group means
Variable 1998-2003 2004-2007 T-statistic P-value
DEALSIZE .0576418 .0452607 2.1251 .0336**
DUMMYDEAL .3091462 .315586 -.5625 .5738
*, **, *** significant at the respectively 0.01, 0.05 and 0.1 level 5.2 Correlation
Table 4A contains the Pearson correlation matrix for the first sample. It measures the strength and direction of the association that exists between the variables. The bold values in the correlation matrix are significant at the 1% level. The three independent variables %OUTDIR, %A_OUTDIR and %NA_OUTDIR are highly correlated among each other, which is expected since %OUTDIR is constructed from the two other independent variables. The correlation between relative deal size and the binary dependent variable (DUMMYDEAL) is positive and significant at the 1% level. This was also expected, because the binary variable takes a value of zero if the relative deal size is zero as well. There exists a negative and significant association between return on equity and
and the dependent variable DEALSIZE. However, this association is not significant. The binary variable (DUMMYREG) is included for the second hypothesis to assess the correlation of the regulatory shock with other variables. In accordance with the expectations, this variable is correlated to the independent director variables and significant at the 1% level. It has a significant positive correlation with outside directors (%OUTDIR) as well as with non-affiliated outside directors (%NA_OUTDIR) and a negative and significant correlation with affiliated outside directors (%A_OUTDIR).
Table 4A. Pearson correlation matrix H1+2
Bold values are significant at the 1% level
Table 4B shows the Pearson correlation matrix for the second sample and yields quite similar results. For example, the independent variables are also correlated among each other. The dependent variable in this correlation matrix is the binary variable DUMMYIMPAIR. This binary variable shows a negative and significant association with M2BOOK and ROE. This is contrary to Table 4A, where M2BOOK is significant and positively correlated with both dependent variables.
Variable DEALSIZ E DU MM YDE AL %O UTD IR %A_ OUT DIR %N A_O UTD IR DU MM YRE G BSIZ E LNA SSET S M2B OO K CRA TIO COV ERA GE R O E DEALSIZE 1 DUMMYDEAL .3280 1 %OUTDIR -.0195 .0316 1 %A_OUTDIR -.0054 .0035 .0935 1 %NA_OUTDIR -.0082 .0183 .5887 -.7495 1 DUMMYREG .-.0249 .0066 .1632 -.1092 .1967 1 BSIZE -.0252 .0676 .2334 .0195 .1379 .0210 1 LNASSETS -.0239 .1603 .2389 -.0279 .1810 .0747 .5793 1 M2BOOK .2203 .0681 -.1185 -.0088 -.0714 -.0290 -.1686 -.149 1 CRATIO .0280 -.0285 -.1034 -.0432 -.0334 .0160 -.2181 -.225 .1925 1 COVERAGE .0570 .1523 .0699 -.0239 .0659 .0952 .2026 .5148 .1479 -.0353 1 ROE -.0630 -.0024 -.005 .0017 -.0048 .0194 .0327 .0230 .0551 -.0203 .0228 1
Table 4B. Pearson correlation matrix H3
Bold values are significant at the 1% level
5.3 Regression results
This section will discuss the results of the regression analyses. Table 5 summarizes the earlier mentioned empirical models. Control variables are included for all models. The second column states the dependent variable for each model, whereas the third column explains the measurement of the outside director variables per model. For the second hypothesis, the extra control variable DUMMYREG is included. The last column states the type of test. Furthermore, section 5.3.1 will explain the OLS regression results. Section 5.3.2 discusses the Probit regression results.
Table 5. Empirical models
Model Dependent variable Director variables Control variable Type
H1 model (I) DEALSIZE Combined OLS
H1 model (II) DEALSIZE Separated OLS
H1 model (III) DUMMYDEAL Combined Probit
H1 model (IV) DUMMYDEAL Separated Probit
H2 model (I) DEALSIZE Combined DUMMYREG OLS
H2 model (II) DEALSIZE Separated DUMMYREG OLS
H2 model (III) DUMMYDEAL Combined DUMMYREG Probit
H2 model (IV) DUMMYDEAL Separated DUMMYREG Probit
H3 model (I) DUMMYIMPAIR Combined Probit
H3 model (II) DUMMYIMPAIR Separated Probit
Variable DUMMYIM PAIR %O UTD IR %A_ OUT DIR %NA_ OUTD IR BSIZ E LNASSETS M2B OO K CRA TIO LNCO VERA GE ROE DUMMYIMPAIR 1 %OUTDIR .0358 1 %A_OUTDIR -.0045 .0887 1 %NA_OUTDIR .0271 .5599 -.7754 1 BSIZE -.0138 .1749 0.220 .0911 1 LNASSETS -.0299 .2103 -.0359 .1626 .5575 1 M2BOOK -.0618 -.1180 -.0067 -.0694 -.1035 -.0900 1 CRATIO .0208 -.1169 -.0619 -.0220 -.2362 -.2273 .2463 1 LNCOVERAGE .0349 .0955 -.0227 .0791 .2740 .6169 .1244 -.0843 1 ROE -.0803 .0251 -.0204 .0322 .0766 .0652 .1004 -.0293 .0390 1
Table 6 shows a multicollinearity test. Multicollinearity focuses on the presence of correlation among the independent variables. It indicates that the independent variables are not independent, which leads to issues for individual coefficient testing. If there exists multicollinearity among the independent variables, it is very difficult to disentangle the contribution of each variable. I conduct the multicollinearity test for the first and second model of hypothesis 2. These are the most extensive models available in the sample, because they also include the regulatory shock, as well as both measures for outside directors. The reported values in Table 6 are the Variance Inflation Factor (VIF). The VIF indicates how much of the variance in one independent variable is explained by the other independent variables. A value below 5 means that there is no collinearity among the independent variables. Table 6 reports only values below 3, which indicates that there are no multicollinearity issues in this study.
Table 6. Multicollinearity tests
Variable H2 model (I) VIF H2 model (II) VIF
%OUTDIR 1.11 %A_OUTDIR 2.40 %NA_OUTDIR 2.55 DUMMYREG 1.04 1.05 BSIZE 1.57 1.57 LNASSETS 2.13 2.14 M2BOOK 1.14 1.14 CRATIO 1.10 1.10 LNCOVERAGE 1.50 1.50 ROE 1.01 1.01 Mean VIF 1.32 1.61 5.3.1 OLS regression
Table 7 reports Ordinary Least Squares (OLS) regression results for hypothesis 1 and 2 for the relative deal size measure. Model (I) of each hypothesis defines outside directors combined, model (II) of each hypothesis defines outside directors separated as affiliated outside directors and non-affiliated outside directors. OLS is a method for estimating the unknown parameters in a linear regression model, with the goal of minimizing the sum of the squares of the differences between the observed responses. The coefficient R2 is the portion of the total variation in the dependent variable that is explained by variation in the independent variables. The R2 for the models in Table 7 is respectively .0553, .0553, .0557, .0558. H2 model (II) could be considered as the most reliable model, but the increase in R2 is probably due to the extra independent variables in that model.
Hypothesis 1 focuses on the impact of the proportion of outside directors on M&A activity. The results for both H1 models seem consistent, which indicates that there is no major difference between the combined measure (%OUTDIR) and the separated measure (%A_OUTDIR & NA_OUTDIR) of outside directors. Therefore, only the combined model H1 will be discussed. The coefficient for outside directors is .007371, which is not in the hypothesized direction and neither significant. Only market-to-book ratio and return-on-equity is significant at the 1% level, which might indicate that bigger firms tend to execute more deals, as well as that firms undertaking M&A deals have a lower return on equity. However, there is no support for hypothesis 1 in this OLS regression.
The second hypothesis examines the same relationship, but also incorporates the variable DUMMYREG to control for the regulatory shock. This variable focuses on the regulatory shock and takes a value of 1 post-adoption, zero otherwise. In accordance with hypothesis 2, the binary variable exercises a negative influence on the dependent variables and is significant at the 10% level. This indicates that relative deal size is lower after the regulatory shock, which was already a significant result in Table 3. However, it was expected that outside directors would also have a negative influence on the dependent variable DEALSIZE, but this is not the case. The coefficients of outside directors even increased after including DUMMYREG. The control variable LNCOVERAGE is significant at the 1% level as well in both H2 models. Based on the H2 model findings in this regression, there is support for hypothesis 2. The findings indicate that the regulatory shock had a negative influence on the relative deal size. This means that the relative deal size decreased after the adoption of the SEC regulations. In the next section, relative deal size will be replaced by deal occurrence to determine whether the findings are robust in a probit model with a different measure of M&A activity.
Table 7. OLS regression H1+2
Variable H1 model (I) H1 model (II) H2 model (I) H2 model (II)
%OUTDIR .007371 .0140667 %A_OUTDIR .0016653 .0048851 %NA_OUTDIR .0092981 .0167979 DUMMYREG -.0099165*** -.0103842*** BSIZE .0014021 .0014121 .0013182 .0013337 LNASSETS -.0066295 -.0067973 -.0064991 -.0067373 M2BOOK .0277505* .0277569* .0276607* .0276642* CRATIO -.0030555 -.0030918 -.0029104 -.0029617 LNCOVERAGE .017562 .017576 .0185688* .0186325* ROE -.0191835* -.0191803* -.0190738* -.0190643* Intercept -.0153239 -.0154246 -.0181928 -.0180873 R2 .0553 .0553 .0557 .0558
*,**,*** significant at the respectively 0.01, 0.05, 0.1 level
5.3.2 Probit regression
A probit regression is typical to use in a model where the dependent variable can only take two values, such as a dummy variable. The dependent variable in Table 8A is DUMMYDEAL, which takes a value of 1 if a deal has been executed and a zero otherwise. Pseudo R2 for all models is respectively .0311, .0312, .0311 and .0312, which is less than the 0.13 reported by Tolmunen and Torstila (2005) in a similar probit model with a dummy variable to measure deal occurrence. Table 8A shows the results for the models.
Also in this regression both models for H1 seem consistent, which indicates that there is no difference between the combined and separated measure of outside directors. The coefficient for outside directors is .0755433, which is not in the hypothesized direction and also insignificant. Similar to the results reported in Table 7, control variables M2BOOK and LNCOVERAGE are significant at the 1% level. Furthermore, not many differences occur between the OLS regression and the probit regression. The findings do not provide any support for the first hypothesis.
The H2 models are also consistent for both the combined and separated measure of outside directors. However, the binary variable DUMMYDEAL is not significantly related to DUMMYREG, which indicates that deal occurrence is not affected by the regulatory shock. The coefficient for %OUTDIR is .0904256 and insignificant. These results seem to indicate that the proportion of outside directors does not affect the occurrence of M&A deals, but does have a significant influence on the relative size of the deal. Therefore, there is some support for the second hypothesis, but the evidence is not robust.
Table 8A. Probit regression H1+2
Variable H1 model (III) H1 model (IV) H2 model (III) H2 model (IV)
%OUTDIR .0755433 .0904256 %A_OUTDIR .1496761 .1556758 %NA_OUTDIR .0682616 .08221 DUMMYREG -.0216665 -.0190256 BSIZE -.0116942 -.0118187 -.0119013 -.0119865 LNASSETS .2861531* .2874094* .2864432* .2875335* M2BOOK .0525321* .0525155* .0523* .0523144* CRATIO -.0179027 -.0174031 -.0175009 -.0170997 LNCOVERAGE .2394752* .2394228* .2416548* .2413422* ROE -.0214761 -.0216797 -.0210473 -.0212798 Intercept -1.839151 -1.847102 -1.845395 -1.851934 Pseudo R2 .0311 .0312 .0311 .0312
*,**,*** significant at the respectively 0.01, 0.05, 0.1 level
Table 8B shows a probit regression for the empirical models in the third hypothesis. The dependent variable is the binary variable DUMMYIMPAIR and takes a value of 1 if an impairment of at least 10 percent of total goodwill on balance sheet occurred in (year t) or (year t + 1) or year (t + 2) following a deal, zero otherwise. The results for both models seem consistent, although the H3 model (II) has a slightly higher pseudo R2, respectively .0698 and .0704. However, this is caused by the incorporation of more independent variables.
The coefficient for outside directors is 1.082896 in model (I), which is not in the hypothesized direction, but significant at the 5% level. This indicates that outside directors have a positive influence on the occurrence of goodwill impairments of at least 10 percent within two years following a completed M&A deal. This might be the result of an ambiguous relationship between outside directors and subsequent goodwill impairments after a deal, because a higher proportion of outside directors on a board might also imply that those firms are more likely to recognize goodwill impairments, due to increased monitoring. Also LNCOVERAGE is positive and significant at the 1% level, which might indicate that increased analyst coverage is associated with increased monitoring (Armstrong et al., 2014; Duchin et al., 2010). Furthermore, control variables LNASSETS and M2BOOK are significant at the 1% level as well.
H3 Model (II) uses the separated measure of outside directors. The coefficients for the separated model are respectively .9545984 and 1.139726 and significant at the respectively 10% and 5% level. However, these results are also not in the hypothesized direction. This might be caused by the earlier explained ambiguous relationship between the variables, which will be further discussed in the Section 6. Based on the findings, there is no evidence supporting hypothesis 3.
Table 8B. Probit regression H3
Variable H3 model (I) H3 model (II)
%OUTDIR 1.082896** %A_OUTDIR .9545984*** %NA_OUTDIR 1.139726** BSIZE .0166879 .0173173 LNASSETS -.471909* -.4770797* M2BOOK -.2650697* -.2657958* CRATIO .0767857 .0748544 LNCOVERAGE .7031577* .704575* ROE -.1762386 -.1759945 Intercept -1.847102 -2.172839 Pseudo R2 .0698 .0704
*,**,*** significant at the 0.01, 0.05, 0.1 level 5.4 Summary of main findings
In order to clarify the main findings of this study a summary table is included. It provides a quick overview of all findings. Based on the findings, there is no support for hypotheses 1 and 3. Most feasible results were found for hypothesis 2, regarding the influence of the regulatory shock variable.
For the second hypothesis, Table 3 shows a t-test to assess whether there is a difference in group means observable in the period prior to and post-adoption. The results are significant at the 5% level for the variable DEALSIZE in the hypothesized direction, which indicates that relative deal sizes decreased following the adoption of SEC regulations in 2004. Furthermore, Table 7 shows that the binary variable DUMMYREG exercises a negative influence on the dependent variable DEALSIZE and is significant at the 10% level. This indicates that relative deal size is lower after the regulatory shock. However, I found no support for a negative relationship between outside directors and deal occurrence. See Table 9 for a summary of the main findings.
Table 9. Summary of main findings
Hypothesis Tests Findings H1 OLS, Probit
Results are insignificant and not in the hypothesized direction, no support for this hypothesis.
H2
T-test, OLS, Probit
Dummy variable for regulatory shock is significant in Model (I) and (II) in the hypothesized direction, for both the OLS model and the t-test. No robust support for hypothesis 2.
H3 Probit
Results are significant, however not in the hypothesized direction. No support for hypothesis 3.
