Does social capital provide the
power to extract rents?
Coupling social network theory and managerial power approach
Lisa Stein s1323520
ljustein@gmail.com
ABSTRACT
How come that even with a compensation committee which is supposed to protect shareholders’ interests, executives still manage to negotiate outrageous compensation packages? This thesis researches the managerial power approach to executive compensation, examining how powerful executives extract more rents. The concept of power is linked to social network theory, which hinges on the concept that the structure of a person’s social network can provide the individual with a better bargaining position. This concept was tested using 2004 data on executive compensation and board structure. Results are limited in significance but do support the hypothesis that the structure of social networks result in bargaining power that can be used to extract rents.
INTRODUCTION
Executive compensation is on the rise, and as it continues to do so, more and more people voice their disagreement. The ratio of executive pay compared to worker pay in the United States has increased sevenfold between 1982 and 2003 (Chan, 2008). While the mentioned ratio of 301:1 is tremendous, this is nothing more than an average ratio. Therefore, there are instances for which this ratio is even higher. Only in excessive cases, the public makes itself heard, sometimes even leading to repayment of (part of) the compensation. A recent example was the bonus that was granted to ING management even though they received state support to deal with the crisis. In this particular case, the CEO succumbed to the public’s pressure and repaid the bonus.
If executive compensation is higher than optimal, why is there no mechanism that prevents such compensation packages from being awarded? In the current structure of determination of executive compensation, the shareholders of a company, being the owners and therefore the ultimate decision maker in the company, are ultimately responsible for the compensation packages. However, as they are not involved in the daily routine of managing the company, they delegate this responsibility to the board of directors, or even a compensation committee within the board, of the company. The compensation committee prepares a proposal package which is presented in the annual meeting of shareholders. In some countries, the shareholders have a say with regards to executive compensation, but this is by no means a definitive decision. In fact, their role is an advisory one.
Even though there are regulations that govern the independence of the compensation committee, judging by current discussions, some argue that executives manage to extract more money from the companies than would be expected given their performance (Conyon, 2006). Apparently, executives are somehow able to extract rents from the company for their personal gain. How come that even with a compensation committee consisting fully of independent directors, executives still manage to negotiate compensation packages that are less than optimal from the shareholders’ point of view?
Aim and added value of the research
through their compensation package. It also describes, and tests mechanisms that explain the occurrence, methods that can be used to obscure the total value of the compensation from the general public in order to prevent public outrage. In this research, the concept of power is formalized through social network analysis. Social network theory argues that the structure of a network can present opportunities to broker information. These brokerage activities can result in power than can be used to influence the board during compensation-setting negotiations.
Both the fields of executive compensation and social network analysis are thoroughly researched. However, to my knowledge there have not been any studies linking the concept of power through social networks and CEO compensation on the general level. Salvaj and Farraro (2006) have researched this concept in terms of the presence of golden parachutes and accelerated cash-outs, which are phenomena typically explained by the managerial power approach, but they did not look at traditional executive compensation packages, i.e. when the CEO remains employed by a company. Similarly, the composition and level of compensation in ‘normal’ situations has been researched largely, but these studies usually look at the influence of interlocking directorates or other factors such as institutional investors and independence of directors. In the field of managerial power approach, Bebchuk, Fried & Walker (2002) have also pointed to the influence of tenure, resulting in familiarity threats, and using compensation consultants in order to justify the structure and height of the compensation package. Although social network analysis would include interlocking directorates as a source of power, the scope of these networks is much larger than this and also includes indirect linkages to other directors and executives. Given the mixed evidence that is found in studies that aim to explain levels or structure of executive compensation (Devers, Cannella, Reilly, & Yoder, 2007), one could conclude that there may be other sources of influence. This research seeks to show that executives use the power derived from their business networks to influence the level and structure of executive compensation. This study will add to the vast amount of empirical studies in the field of executive compensation. More specifically, it will add to the understanding of the role of social networks, limited to business networks, in the relationship between boards and executives.
Research questions
align incentives, but it can also be structured to maximize executive wealth or to obscure the height of the compensation package by using complex payment structures (Bebchuk et. al., 2002).
Structure
The following section will provide an overview of literature on executive compensation. It will then proceed to outline the underlying assumptions of managerial power approach and the conceptualization of executive power through social network theory. The third section will describe the theoretical model that is the core of this research. The following section translates the theoretical model into a testable format by describing the collected data and the final sample that was used to empirically test the theoretical model. The thesis will then proceed to describe the empirical results of the research. Finally, the conclusions and limitations of the research, as well as recommendations for future research will be discussed.
LITERATURE
Many theories of executive compensation originate from an agency problem. With the growth of corporations, control became separated from ownership (Berle & Means, 1932). This led to differing interests between those owning the firm (principals) and those managing it (agents). The principals want the firm to earn as much profit as possible, while agents may pursue their own goals. This discrepancy between the interests of principals and agents leads to agency costs. These are the costs that are incurred by the principals in the process of making the agents pursue the goals of the principals. In addition, agency costs also include the costs that arise from efforts that are made by the principals to monitor their agents’ behavior. As the principals are not involved in the daily management of the firm, they do not possess the same information with regards to the agents’ efforts and goals as the agents do. Agents can exploit this informational asymmetry to shirk (exert less effort than desired), as the principals do not have sufficient information to conclude whether the firm’s performance has suffered as a result of shirking on the executives’ part (Bebchuk, Fried, & Walker, 2001). Alternatively, they could use this advantage to pursue their own goals rather than shareholders’ best interests. It is often assumed that the main goal of CEOs is to maximize their own personal wealth, rather than maximizing shareholder value, but it can also include other, less quantifiable goals, such as the desire to manage a large firm or to attract media attention.
opportunistic behavior by executives that may harm the interest of the firm’s owners. One way to align the executives’ interests with those of the shareholders is through their compensation package.
Aligning incentives between executives and shareholders
The bulk of research into the field of executive compensation concentrates on the theory of optimal contracting (Bebchuk et. al., 2002). This theory argues that contracts are devised to provide executives with contracts that minimize agency costs. It assumes that employers try to find suitable managers, and bargain with them at arm’s length to obtain an agreement that satisfies both parties (Dorff, 2005). Thus, the compensation package must be attractive enough to attract the best executive. In addition, the costs of monitoring the executive’s performance, as well as the residual costs (stemming from the fact that the executive could still develop a policy that deviates from the shareholders’ preferred policy), should be minimized.
Research has indicated many ways to induce CEOs to adopt policies that are in line with shareholders’ interests, the most important way being to link the executives’ financial incentives to their performance (see, for example, Chan, 2008). This can be accomplished by offering executives a bonus when they meet certain performance criteria. This is a direct linkage between pay and performance. As it is the board’s role to set the executives’ compensation, they can construct the goals such that executives obtain a bonus when shareholders’ interests are best served. A commonly used method is to tie bonuses to performance. This can be done through, for example, tying bonuses to profits, or to changes in share price. Murphy (2000) notes that using accounting measures as the target for a bonus could induce an executive to influence these measures. His research shows that firms that use accounting measures as a target show more indications of earnings management.
As a result, executives can no longer be disciplined by the market. The market is aware of this fact, leading to decreasing value of the firm. Since it is uncommon for externally hired managers to own that large a fraction of their firms, insider ownership can be considered a good measure to align management’s interests with those of shareholders.
From the perspective of control, there are more indirect ways to reduce agency costs. Yermack (1996) argues that smaller boards are more efficient in monitoring of the executives’ behavior. This leads to increased market value of firms. He also provides empirical evidence that smaller boards rely more on financial incentives to induce executives to adopt a policy that is beneficial to the shareholders. Thus, smaller boards are more likely to include bonuses as part of the compensation package. The compensation of executives at companies with smaller boards was found to be (Tosi, Werner, Katz, & Gomez-Mejia, 2000) more responsive to changes in shareholder wealth.
The structure of the board can also reduce agency costs. Rosenstein and Wyatt (1990) have studied the influence of outside board directors on shareholder wealth. They define an outside director as a person that is not, neither currently nor in the past, employed by the company, and the only connection that the person has with the company are the duties as a director. They conclude that announcements of the appointment of an outside director generally lead to an increase in shareholder wealth. This indicates that shareholders believe that outside directors are better able to protect their interests. This is in line with corporate governance theories that argue that outside directors provide better monitoring of executives, and that they provide management with complementary knowledge.
Some researchers argue that shareholders are not the only agents that are relevant to take into account when managing a firm. Coombs and Gilley (2005), for example, argue that executives should not only focus on shareholders, but, more general, on serving the interest of all stakeholders involved with the company. They reason that executives that adhere to a policy of stakeholder management should receive a higher compensation due to the superior performance of their firm. However, their empirical tests reveal the opposite result: executives that pursue stakeholder management receive a lower compensation than those that focus on financial results.
The level of executive compensation
and if goals are chosen wisely, it predicts the height of these variable components. Other theories are more concerned with the level of executive compensation. Baumol (1967) already mentioned that “executive salaries appear to be far more closely correlated with the scale of operations of the firm than with its profitability”1.
Sanders and Carpenter (1998) argue that executives are compensated for a measure of productivity. Their argument is based on the assumption that the bulk of an executive’s work is based on processing large amounts of information in a complex environment. Not only are executives better at processing such large amounts of information, they must also be compensated for the increased complexity, and risks, that they face. Therefore, the more complexity, stemming from, for example, the size of the firm or the degree of internationalization, the higher the executive’s compensation will be. In a meta-analysis of CEO pay studies, Tosi et. al. (2000) conclude that approximately 40 % of the variance in executive compensation is explained by firm size.
The screening hypothesis links personal characteristics to salary (see Hogan and McPheters, 1980, for references). As it is very hard to determine whether a person will succeed as an executive only by looking at current performance, the board will take into account a person’s characteristics, such as educational background, age, and experience, to determine whether he will be able to provide good management to the company. Ultimately, these give an indication of the capabilities of the executive, and this will also determine an executive’s compensation. Hogan and McPheters (1980) provide empirical evidence for some variables. Specifically, age, tenure, and years as an executive have significant effects on compensation, whereas having an administrative background or graduate experience did not.
Limitations to theories of optimal contracting and other theories
The aforementioned mechanisms to reduce agency costs are all methods to align executives’ goals with those of the principals. However, some researchers (such as Bebchuk, Fried, & Walker) argue that these methods do not work efficiently. For example, even though most boards have a nomination committee for appointing directors, and the approval of shareholders is needed to appoint a new director, effectively, most new directors are proposed by the CEO himself
1 As quoted by David H. Ciscel (1974). Full reference: Baumol, William J., 1967. Business behavior, value and growth, rev. ed.
(Rosenstein & Wyatt, 1990). In addition, as noted by Devers et. al. (2007), empirical studies show mixed evidence with regards to drivers of executive compensation as mentioned above.
Several other practices do not coexist very well with the agency theory and optimal contracting. Many of these concern practices surrounding option grants. While in itself, granting options can be a very good measure to align management’s incentives with those of shareholders, by making managers themselves shareholders, and by serving as a bonus that can only be received when a certain goal is attained, some of the attributes that are commonly included in option contracts are puzzling from the optimal contracting view (Bebchuk et. al., 2002). For example, nearly all options are granted at-the-money, i.e. with an exercise price which is equal, or close, to current share price (Lie, 2005). Because share prices tend to move in line with the market, managers may be rewarded even if the increase in value has nothing to do with their performance, but is just a result from a general market movement. As a result, the incentive that is provided by the options is weakened. This problem could easily be solved by using options that are tied to, for example, a peer group of companies in the same industry or facing similar risks, that need to be outperformed in order for the options to vest and become exercisable. However, there are few companies that include such a provision in their option contracts (Bebchuk et. al., 2002). Including such a provision in the option contract would reward performance rather than market movements by correcting the change in share price for the general market trend.
In addition, when (general) market prices fall, it is not uncommon to reset the exercise price in option contracts. This effectively removes the incentive from the option contracts, as executives know that when their company’s stock performs poorly, the exercise price may very well be reset to current market price, meaning that they can still make a profit on the option contract (Bebchuk et. al., 2002).
Another practice that is common also does not coexist very well with the concept of the board setting executive compensation in the agents’ best interest is a practice called option backdating. This results in the grant date of the options lying in the past. As the exercise price of the options granted is usually equal to the prevailing price at grant date, this results in possibilities to select more favorable exercise prices. (Lie, 2005) found that abnormal returns on shares are negative in the period prior to awarding stock option grants, and positive in the period afterwards. This suggests that “[…] unless executives possess an extraordinary ability to forecast the future market
The managerial power approach
In response to all this puzzling evidence, Bebchuk et. al. (2002) have developed an approach to executive compensation they call managerial power approach. Essentially, this theory argues that management does have considerable power over their company and its board, and that they use this power to extract excessive rents from the company. The term excessive rent refers to the amount of compensation that the executive receives on top of what is needed to persuade him to work in the company and to provide him with the incentives that will make him manage the company in the shareholders’ best interest (Conyon, 2006). This managerial power approach acknowledges the agency problem that exists between shareholders and management, but rather as seeing executive compensation as the solution to the problem, it is viewed as part of the problem itself (Bebchuk et. al., 2002). The ability of executives to extract excessive rents constitutes an agency cost. After all, it is not in the best interest of the shareholders to transfer as much money as possible to the executive.
The managerial power approach rests on the assumption that the executives have considerable power over the board of directors. This power extends even to those directors that are nominally independent (Bebchuk et. al., 2001), i.e. people whose only relation to the company is the fact that they serve on its board. As pointed out by Rosenstein and Wyatt (1990), new directors are commonly proposed by the CEO, which means that he influences the selection of people that are favorable towards him. In addition, once directors are on the board, they will be inclined to agree with the CEO rather than oppose him, because opposing the CEO and not reaching consensus would mean that they should either fire him or step down as a director. To fire the CEO, the director needs to get at least the majority of other directors on his side, which is unlikely to happen over the executive’s compensation, but the alternative, to step down, is also not very attractive to most directors. Often, they would feel stepping down as a director means a loss of reputation for themselves (Bebchuk et. al., 2002).
To prevent outrage from occurring, CEOs and boards of directors use a technique that Bebchuk, Fried & Walker call camouflage. As long as the magnitude of a compensation package can be justified, legitimized, or obscured from the public, it will not cause outrage. An example of camouflage techniques is the desire of companies to compensate their executives above the industry average, as a justification of the level of remuneration. However, this also leads to ever-increasing amounts of compensation (Bebchuk & Fried, 2005). As a result of the benefits of camouflage, executives, and their boards, prefer compensation structures to be complicated, as this can make it harder for outsiders to recognize that excessive rents are extracted. Another way to obscure the total magnitude of the compensation package is by using option plans. As mentioned before, option plans are not necessarily a bad thing, as they can provide very good incentives to executives, but “the devil is in the details” (Bebchuk et. al., 2002). Research has indicated that executives generally hold too many options, decreasing the marginal benefit of each option2. In addition, some other practices that do not make sense include uniform use of at-the-money options, reload options, and the freedom for executives to unwind incentives by selling and/or hedging options and shares (Bebchuk et. al., 2002).
Thus, the managerial power approach addresses a number of limitations to agency theory. By using their power in the bargaining process with the board, executives can extract excessive rents, or outrageous compensation, from the company. These excessive rents can be defined as the surplus that is being paid to executives in excess of what would be necessary to get the executive to work for the firm (Conyon, 2006). This constitutes an agency cost that the board is supposed to prevent.
CEO power and social network theories
The power of CEOs can stem from many sources. One source of power for CEOs lies in CEO duality, a situation in which one person is CEO as well as chairman of the board. Finkelstein and D’Aveni (1994) argue that this situation leads to increased power for the CEO for several reasons: as chairman of the board, the CEO is responsible for organizing meetings as well as setting the agenda of these meetings; and as chairman the CEO has additional influence over nomination of new directors.
2 As quoted by Bebchuk et. al., 2002. Full reference: Michael A. Habib and Alexander P. Ljungqvist, 2002. Firm value and
Another factor that increases the executives’ power over the board of directors is tenure. Van Essen, Otten, & Carberry (2012) argue that executives that have retained their position for a longer period of time have more experience with the company and can therefore exert more influence over the board. Larger boards provide the executives with more power over the compensation-setting process due to the increased complexity of social dynamics within such groups (Yermack, 1996). Finally, much discussion is centered on the independence of the board of directors, such as Rosenstein & Wyatt (1990) who argue that outside directors are believed to be better able to protect shareholders’ interests. Therefore, the higher the percentage of independent directors, the less power the executives will have over their board.
Note that the managerial power approach argues that even directors that are nominally independent, can be influenced by the executives, for example as familiarity increases as executives and directors work together for a prolonged period of time (Van Essen et. al., 2012). Another factor that results in executives gaining power over the board of directors is interlocking directorates (Bebchuk et. al., 2001). This is the situation in which directors sit on more than one board together. This increases feelings of reciprocity and interdependence.
The old boys’ network in the popular press is related to the previous two arguments; a term that reflects the sentiment that the world is governed by a group of people that allocate jobs amongst themselves and look out for each other. Or, as one website puts it: “They sit on the largest boards
of the country. Many sit on government committees. They make decisions that affect our lives. They rule.”3 This argument is formalized in the theories of social network.
Sociologists have long argued that, in addition to human capital, there is such a thing as social capital. The fact that some people do better than others is often thought to be the result of better qualifications and ability (i.e. human capital), but Burt (2000) argues that this phenomenon does not explain all variation. Rather, Burt says, “the individuals who do better are somehow better connected”. They can reap the benefits of their network by having access to information at the right time. Burt’s argument is formalized in the following paragraph.
People (actors in sociology terms) are generally grouped in clusters; these are composed of actors that frequently interact with each other. Actors within the same cluster tend to know about the same things around the same time (Burt, 1997), while actors in other clusters learn about these things later. Generally, these people are not unaware of each other’s existence, but they are busy
doing their own things, such that they do not have sufficient time to monitor the activities in the other cluster (Burt, 1997).
Thus, each cluster has its own flow of information. One cluster is unaware of the information flowing in the other cluster as long as there are no connections between the two clusters. The absence of connections between two clusters is called a structural hole (Burt, 1992). A person that happens to bridge this structural hole can direct the information that flows from one cluster to another. Blyler and Coff (2003) argue that these people will have more bargaining power that they can use to appropriate rent from an organization to themselves. This bargaining power originates in the fact that the broker (the person that spans the structural hole) can outplay individual demands against each other (Burt, 2000). In addition, they control who obtains which information, and when.
The network can constrain a broker’s opportunities for brokerage to the extent that the network is concentrated, directly or indirectly, in a single contact. Several characteristics of a network can influence a person’s social capital. These characteristics are size, density, and hierarchy of the network (Burt, 1997). A larger network provides the person with more bits of information that can be used to broker between individuals’ demands (Burt, 2000), and therefore increases social capital. Dense networks provide less social capital, because there are less structural holes and therefore fewer opportunities to broker between individuals (Burt, 2000). There are so many ties amongst the actors that the individual actors do not have unique connections. The third characteristic refers to the extent to which a network is concentrated in a single contact. When the network is more hierarchical, it provides less social capital, because the central contact in the person’s network can obtain the same information from others (Burt, 2000). This is caused by the fact that hierarchical networks are concentrated around a single dominant person (Burt, 1997). In the event of a highly hierarchical network, a single tie with the dominant person would provide the same information flow as investing in ties with all individuals in the network. I will explain these three characteristics using an example at the end of this section.
All three characteristics are summarized in a measure called network constraint. This is an index measure that is based on the extent to which a person’s network is invested in a single contact, j:
2 qj iq q ij ij p p p c
, where pij is the proportion of i’s relations that is invested in person j, and
CEO would invest all his time in a single contact, the measure would be one. This would result in a network consisting of only one tie (Burt, 1997). As such, this measure also gives an indication of the dispersion of the person’s network. The larger the network, the smaller will be the amount of time invested in a single contact.
The squared sum of all these measures is the network constraint index C jcij (Burt, 1997).
The constraint index sums all individual constraints, resulting in a measure that shows the constraints of the person’s entire network rather than the constraint of the relation between the person and one of his contacts. The index is influenced by all factors relevant to network theory (network size, density, and hierarchy). Network size has a positive impact on social capital, whereas density and hierarchy have a negative impact. Note that the effects are opposite for the network constraint measure, as less constraint indicates more social capital.
As this research is focused on executive compensation the relationship between executives and the board of directors is crucial. Translating Burt’s concept to this setting leads to the following observations. Within a single board of directors, there are no structural holes as all directors have ties to each other. This also means that each person invests in each other person directly as well as indirectly, and this constraint measure (cij) is equal for each of the individuals on the board. The
of the compensation company from the public, in order to prevent the public to conclude that excessive rents are being extracted from the company.
A graphic depiction of an executive network is included in Figure 1. The individual of interest, denoted by the red dot, sits on the board of three companies. This can be seen from the green dots that represent the executive’s direct contacts; these are clustered in three groups. Two of these direct contacts are on the board of two companies of which the executive is a member. As such, the executive is not the only one to span the hole between the companies. However, the executive
is the only one sitting on all three boards. Therefore, the executive can gain advantage by
determining the flow of information among the three boards. The blue dots in the figure are the executive’s second degree contacts, the ones that his network provide access to.
FIGURE 1: EXAMPLE OF EXECTUIVE NETWORK
an executive would be on the board of one company, and none of his fellow board members would be on the board of another company, the network’s effective size would be 1, as all board members provide access to the same source of information. Density in this network is 0.409, where the maximum density in a network is 1. A density of 1 indicates no structural holes and no brokerage opportunities. Thus, a lower density indicate more structural holes, resulting in less social constraint. Density is calculated as the percentage of the network that is not relevant to the effective size4. Finally, hierarchy of this network is low, due to all individuals in each board
having a link to each other. As a result, none of the individuals are a better connection in terms of hierarchy. The executive whose network is depicted in Figure 1 shows very little social constraint; 0.093. The size of the network result in the executive having access to many sources of information, and the size and density of the network provide the executive with opportunities to broker this information amongst agents. This power can be used in the negotiation process in which executive compensation is set, to extract excessive rents from the company.
FIGURE 2: PROPERTIES OF SOCIAL NETWORKS
In Figure 2, I have included two simple networks for explanatory purposes. The upper figure represents a network in which the individual of interest, Peter, sits on the board of one company. He has three direct contacts, but the effective size of his network is only 1.67, as some of the
contacts’ contacts overlap. This results in less separate sources of information. Peter’s constraint is high, 0.84, as there are few broker opportunities; most actors in the network have access to the same ties. This is also reflected in the density of Peter’s network, which has a value of 0.67. The hierarchy indicator in Peter’s network is low, 0.07, indicating that there are no dominant contacts. James’ network only differs from Peter’s because he sits on two boards. This has a direct impact on all three factors. James has a total of 5 direct contacts, that provide access to 3.80 sources of information. Adding two contacts to the network thus results in a larger increase in network size. James has a social constraint of 0.48, much lower than Peter. James’ network shows a density of 0.30, indicating that there are more broker opportunities in James’ network. This results mostly from James being the only person to bridge the structural hole between the two companies. Finally, James’ network also does not have a dominant contact ,this can be seen from the hierarchy indicator of 0.03.
Research hypotheses
The social network theory provides a measure of power, however, this is a proxy as power cannot be quantified. The theory indicates that people with less constraint should be able to benefit from the access to information originating in different clusters. This power can be used in the bargaining process between the board and the executive, as well. By having access to more information, less constrained executives will be able to negotiate more favorable compensation packages.
Executives can use this power to influence both the total dollar value of their compensation package, as well as the structure of the compensation package. Following the managerial power approach, executives will wish to increase the total level of compensation in order to maximize their personal wealth; thereby extracting excessive rents from the company (Bebchuk et. al., 2002). Because it is not possible to measure the minimum compensation that is needed to attract and retain an executive, I assume that a higher dollar value of compensation indicates more rent extraction.
Hypothesis 1: The less constrained an executive is, the larger the total dollar value of his compensation package will be.
compensation forms, such as options, will be used to camouflage the level of compensation. These camouflage techniques are most important when the executive has been able to extract sufficient rents to induce outrage by the public. The board of directors will be more inclined to improve of a compensation package when this outrage can be prevented. Therefore, it is likely that executives that hold the power to obtain an exorbitant compensation package will resort to option plans to camouflage the dollar value of their compensation. By using option plans rather than dollar-value bonuses, the executive can impede the public’s ability to assess the size of the compensation package, as the costs and the value of option contracts are difficult to measure.. Please note that this argument does not conclude that option contracts are unfavorable per se, as they can be an efficient means to align executives’ interests with those of shareholders.
Hypothesis 2: The less constrained an executive is, the more likely it is that variable compensation is paid in options.
Finally, the manner in which the option contracts are structured provide an insight into whether the option contracts were granted to camouflage excessive rents. One of the features of option contracts that can indicate camouflage is granting options that are closer to the money at grant date (Bebchuk et. al., 2001). Therefore, executives that are in the position to extract excessive rents are more likely to be granted options of which the strike price is close to the share price prevailing at grant date.
Hypothesis 3: The less constrained an executive is, the smaller the positive gap between strike price and the prevailing price at grant date.
The following section will provide a theoretical model that can be used to test the validity of these hypotheses.
THEORETICAL MODEL
In this section, I will outline a theoretical model that is appropriate to address the research hypotheses that were identified in the previous section. Apart from the variables of interest, the control variables will also be defined, resulting in a testable mathematical expression. A description of the dataset is provided in the section Data and methods.
increase this person’s power by providing the person with influence over the information flow between people. The structure of a person’s network can be measured by a measure called network constraint index (as discussed in the section CEO power and social network theories). Social constraint is expected to have a negative correlation to the height of the executive’s compensation.
(1.1) = , where EC = executive compensation C = network constraint
In order to properly assess the influence of network constraint, the influence of other factors on executive compensation needs to be taken into account. The factors that are most commonly cited as being a determinant of executive compensation, are firm size, firm performance, and personal characteristics of the executive, therefore, these variables will be included in the theoretical model.
(1.2) + , where
EC = executive compensation C = network constraint FP = firm performance FS = firm size
PC = personal characteristics, defined as sex, age, and tenure
The next section will proceed to describe the dataset that was used to analyze the above model.
DATA AND METHODS
The data that were used in to analyze the proposed model of executive compensation were collected in a data project initiated by the University of Groningen. The data collection project was based upon a website called Theyrule5; which provides a tool to analyze the connections
between the most influential people in the United States: those that sit on the boards of the largest companies and participate in government organizations. By joining the efforts of several students, a dataset comprising all companies in the Fortune 500 as per 2005, supplemented by a number of companies that were included in the S&P 500, was created. The dataset consists of two levels of information, on the company level and on the level of the directors of the company. The company-level set includes data on each of the companies in the sample, including data such as industry, stock price, firm size, and sales. The director-level set includes data on each of the directors in the company, including data such as age, tenure, membership of audit and / or compensation committee, compensation, etc. All data were collected for the years 2003 and 2004, and with the majority of the data having been retrieved from SEC filings (through EDGAR filings6), the
Theyrule website, and S&P databases.
Definition of each of the variables
This section will provide a definition of each of the variables to be included in the data analysis, based upon the literature review and the theoretical model described in previous sections.
Executive compensation: Executive compensation consists of four components: base salary, bonus, options, and other compensation. The base salary is defined as the total fixed compensation paid during the year, measured in dollars. The bonus is defined as the total variable compensation paid during the year, measured in dollars. The options are measured by retrieving the potential realizable value / grant date value as reported in the proxy statements as filed by the company. In the event that the option value is represented by potential realizable value, the potential value using the 5 % stock appreciation term was used. The other compensation, finally, consists of all other forms of compensation, measured in dollars. This includes forms of compensation such as relocation compensation or pension plans.
5 Web address: https://www.theyrule.net
Constraint index: The constraint index is calculated using the network data analysis tool Ucinet. This tool is based upon the social network theories advocated by Burt (see for example Burt. (2000)), and takes network matrices as its input. Such a matrix is an overview of the ties between persons in a network. For purposes of this study, a network is defined as all people a person sits on a board with, regardless of their roles. The constraint index serves as a proxy for power and has an inverse relationship to power. As such, less constraint indicates more power.
Firm performance: Firm performance is measured as the change between the stock price ultimo 2003 and ultimo 2004 in percentage points.
Firm size: Firm size is measured as total sales in millions of dollars.
Personal characteristics: Sex will be measured using a dummy variable, distinguishing between male (1) and female (0). Age will be computed from the year of birth, taking 2004 as the year of reference. Tenure will be computed as the difference between the year of appointment and 2004.
Final sample
The dataset that was collected through the university project consisted of 544 firms with a total of 6,094 individuals holding one or more positions on the boards of those firms. As a result of missing data, some observations were dismissed from the final data set. These exclusions will be discussed below.
Network constraint was calculated using all observations, including those that were later dismissed. This was done because networks were defined narrowly to begin with; as the only individuals that are included are people with whom a person sits on a board. Other individuals that may be relevant to a person’s network, such as university friends or sports buddies, are not taken into account.
To calculate network constraint, the data were arranged in a manner that links each individual to all other individuals with whom they sit on a board. This file was then analyzed using Ucinet software. The software calculated constraint measures, taking into account the individuals’ ties to other individuals, and the ties of their ties. In other words, an individual’s network includes his connections, and his connections’ connections.
regarding the frequency of interaction, for example. However, if two individuals share more than one board, each additional tie adds weight to the relationship. If, for example, two individuals sit on three boards together, their tie will have a value of three, whereas their tie would have a value of one if they only sat together on one board.
After calculating the network constraint, the dataset was reviewed in order to filter it from any incomplete observations. First of all, the relevant individuals were filtered as the starting point to determine whether the observation was complete. Because this research aims to explain executive compensation, non-executive directors were excluded as observations. This results in a sample size of 902 individuals. An overview of the sample split by position is provided in Table 1.
TABLE 1: SAMPLE BY POSITION
Description Nr of observations
Chief executive officer 592 Chief financial officer 310
Executives 902
Other insiders 1,461
Outsiders 5,046
Undetermined 32
Total observations 7,441
Subsequently, I deleted all individuals that changed position during the sampling period (i.e. 2003 and 2004), as this could distort the data due to compensation periods of less than a year, hiring bonuses, and compensation for termination of employment. Finally, I verified whether all data needed for the analyses to be performed were present. All observations for which information was incomplete, were deleted from the sample. Missing information included items such as birth dates, option values, and firm information. A summary of reasons of deletion is included in Table 2.
TABLE 2: DELETIONS
Reason # of deletions
Changed position during sampling period 14 Missing date of birth or date hired 75
Incomplete option data 15
Incomplete compensation data 10
Incomplete firm data 40
Taking into account the above described deletions, the final sample consists of 748 executives for which data is complete.
Statistical methods
The model described in section 3, will be estimated using regression analysis. The regression equation that will be estimated is specified as follows:
(2.1) , where
C = network constraint
FP = firm performance, measured as change in stock price between 2003 and 2004 FS = firm size, measured as total sales in millions of dollars
S = sex, defined as 1 = male, 0 = female
A = age in years, computed using the year of birth and 2004 as the year of reference T = tenure, defined as the number of years between first appointment and 2004 For each of the hypotheses that will be tested, y will take on a different meaning.
Hypothesis 1 is concerned with total compensation. In this model, y = total dollar value of the compensation package. This includes base salary, variable compensation, and option plans granted.
Hypothesis 2 is concerned with the probability that options are used in the variable component of the total compensation package. In this model, y represents a choice variable that can take the values 1 (options are used) or 0 (options are not used).
Hypothesis 3 is concerned with the properties of the option contracts. In this model, y = market price of the share at the closing of the financial year -/- exercise price of the option.
EMPIRICAL RESULTS
In this section, regression result will be discussed per hypothesis as set out in section 2. I will first present several descriptive statistics relating to the sample that was used for the analysis. The section then continues by describing that regression analysis is the most appropriate manner to address the hypotheses. Table 3 summarizes the sample’s properties.
TABLE 3: DESCRIPTIVES
The summary statistics for the sample are included in Table 3. From this table, it can be observed that of the 748 observed executives, 654 received a bonus during 2004. The bonus ranges from $ 3,000 to $ 17.2 million. 441 executives received some form of irregular compensation. Finally, 555 executives received option grants. The average firm performance improved by almost 15 percentage points, but there were 223 firms that had a negative performance, measured as the increase in stock price.
Appropriateness of regression analysis
To test whether none of the assumptions relating to regression analysis are violated, I have used SPSS to calculate several diagnostics.
N Minimum Maximum Mean Standard
deviation Age 748 35 84 55 7.11 Tenure 748 1- 55 8 7.86 Constraint 748 0.06 0.50 0.24 0.09 Total compensation 1 748 1 56,532,552 5,475,578 6,261,433 Salary 2 748 - 4,973,073 822,999 445,845 Bonus 2 654 3,000 17,273,290 1,695,155 2,018,952 Irregular 2 441 90 6,869,842 228,435 511,287 Options 2, 3 555 16 48,563,243 4,099,839 5,404,342 Firm performance 4 749 52- 1,019 15 61 Firm size 5 748 1 7,392,113 25,878 277,934 1: measured as the sum of Salary, Bonus, Irregular, and Options
2: as disclosed in the 2004 financial reports
FIGURE 3: SCATTERPLOT OF STANDARDIZED RESIDUALS AND STANDARDIZED PREDICTED VALUE
To determine whether the data are heteroskedastic, I have plotted the residuals against the expected values of the regression analysis, refer to Figure 3. Since there is no pattern, in the sense that residuals increase as the predicted value increases or similar, I conclude that heteroskedasticity is not present.
In order to test for presence of serial autocorrelation, SPSS calculates the Durbin-Watson statistic. From Table 4, it can be seen that this statistic is 1.967, close to 2, therefore the conclusion can be drawn that autocorrelation is not present.
TABLE 4: MODEL SUMMARY
Finally, I have examined correlations between the variables included in the model to determine whether the variables are related to each other. In Table 5 the correlations between each of the variables are displayed. Most correlations are between -0.2 and + 0.2, indicating little correlation between the variables. The variables Age and Tenure show a higher correlation, which is not surprising as both are dependent on the year of birth. However, correlation is less than 0.5, as such, this correlation does not affect the predictability of the variables.
R R square
Standard error
of estimate Durbin-Watson
0.19 0.04 6,177,160 1.97 a, b
TABLE 5: CORRELATIONS
The analysis of appropriateness as presented above was made for the dataset used to analyze Hypothesis 1, regarding the relation between social constraint and total compensation. The datasets for the other two hypotheses were analyzed in a similar manner.
Hypothesis 1: social constraint and the total compensation value
The first hypothesis is concerned with the relationship between social constraint and total compensation in dollars. To control for other factors that influence total compensation, control variables were added to filter for their influence on the total compensation and to isolate the influence of social constraint. Table 6 presents the outcome of this regression analysis. When examining the signs of the coefficients, it can be seen that all signs except firm performance are in the expected direction. Constraint and gender have a negative correlation to total compensation, which is expected based upon the analysis of theory presented in section two. The more constrained executives have a lower total compensation package, measured in dollars. This also holds for the females in the sample. Age and tenure both have a positive effect on total compensation, which is in line with observations that people earn more as they age. Larger firms also indicate more compensation. I expected firm performance, finally, to have a positive impact on total compensation, as the executive is to be compensated for his efforts to improve firm performance. However, in Table 6, it can be seen that firm performance has a negative impact on total compensation. This impact is highly insignificant, with a p-value of 0.876, therefore I will not discuss this unexpected outcome further. As can be seen from the output table, only two variables have a significant influence on total compensation, being social constraint and tenure.
TABLE 6: REGRESSION RESULTS H1
From Table 6, it can be seen that the model as estimated explains 2.7 % of the variance in total compensation. This indicates a low explanatory power of the model. This may be caused by omitted variables that explain part of the variance that was not explained by the social constraint measure. It may also be caused by having used inaccurate proxies as input variables. These limitations will be discussed in more detail in section 6.
From Table 6, it can be concluded that constraint and tenure are the two significant variables influencing total compensation. The remaining variables add to the model’s explanatory ability but are not of significant influence. When estimating the model using only the significant variables, the model’s explanatory power increases slightly to 2.8 %. From this observation I conclude that the variables that were regressed insignificant do have an impact on total compensation, but they were possibly not modeled correctly. Perhaps they have a non-linear effect or they may be related to each other. The significance of the variables constraint and tenure increases and effects become more pronounced.
Hypothesis 2: social constraint and the presence of option contracts
In the second hypothesis, I am concerned with the question whether less social constraint, indicating more power, makes it more likely that options are being used as part of the total compensation package. From Table 3, it can be read that out of the 748 executives, 556 persons received option grants as part of their compensation package. The average value of such a grant was $ 4,092,465.
Variable Coefficient Standard error
TABLE 7: REGRESSION RESULTS H2
I have used logistic regression analysis to determine whether social constraint is a predictor of the use of option contracts as part of executive compensation. From Table 7, I conclude that the estimated model does not predict the use of options very well. The explanatory power of the model is 3.7 %. In the classification table in Table 8, it is shown that of the 748 observations, the model predicted 745 observations correctly.
TABLE 8: CLASSIFICATION TABLE H2
When examining the coefficients in Table 7, it can be concluded that only two variables, Firm performance and Age, are of significant influence on the probability that options are used as part of the compensation plan. When examining the coefficients for social constraint, it can be concluded that the sign for constraint is negative as would be expected, but it is not significant enough to conclude that social constraint is a determinant of using options as part of executive compensation.
Hypothesis 3: social constraint and option contract features
The third hypothesis is concerned with one of the characteristics of options: their exercise price. According to Bebchuk et. al. (2002), one of the details that does not make sense in terms of using option contracts to align executives’ interest with those of shareholders, is the granting of options with exercise prices close to the current market price. To test whether executives can use their social capital to negotiate better option contracts, I have estimated a regression using the difference between market price prevailing at the grant date, and the option’s exercise price as a dependent variable. As options may be granted throughout the year, I have used the prevailing
Variable Coefficient Standard error
share price at year end as a proxy for the price at grant date. A positive number, in this respect, would imply that the strike price is below the market price and, as such, that the options were granted in the money. The main explanatory variable of interest is social constraint, which is expected to have a negative impact on the difference between strike price and market price. The other variables remain as control variables. Table 9 shows a summary of the dataset. Note that the dataset has been reduced to 556 observations as only the executives that were granted options have been taken into account.
TABLE 9: DESCRIPTIVES
The mean difference between strike and market price is positive, indicating that the majority of the option grants are made so at ‘in the money’ levels. This is an extraordinary observation given the fact that even if shareholders’ value would not improve (assuming that this is the objective of the shareholders and, as such, the executive’s task), the executive would benefit from the option grant. As such, it resembles more closely a general compensation component rather than a compensation component used to align the executive’s interest with those of shareholders. This can be interpreted as a camouflage technique, used to obscure the magnitude of the compensation package to the public (Bebchuk et. al., 2002).
TABLE 10: REGRESSION RESULTS H3
Examining scatter plots of residuals and correlations between variables do not indicate violations of the assumptions to regression, therefore a linear regression model is appropriate. In Table 10, it
N Minimum Maximum Mean
Standard deviation Strike vs. Market 556 51.66- 81.66 4.20 10.62 Constraint 556 0.07 0.50 0.23 0.08 Firm size 556 1 7,392,113 30,935 321,943 Firm performance 556 52.00- 235.00 11.37 27.80 Age 556 36.00 80.00 54.15 6.81 Tenure 556 1.00 55.00 7.61 7.48
Variable Coefficient Standard error
can be seen that Constraint and Firm performance are the only variables that have a significant impact on the dependent variable, although the magnitude of the effect of firm performance is limited. As can be read from Table 10, constraint has a negative impact on the difference between strike and market price, as was expected. It can be concluded that the less constrained the executive is, the more likely it is that options are granted at favorable prices. This can then be considered an indication of the notion brought forward by Bebchuck et. al. (2002) that executives with more power use options to camouflage the outrageousness of their compensation package. From Table 10, it can be concluded that this model has a higher explanatory power than the ones for Hypotheses 1 and 2. When regressing using only the significant variables Constraint and Firm performance, overall significance increases slightly, from 20.1 % to 20.4 %. As such, I conclude that the insignificant variables, just as with Hypothesis 1, do have an impact on the dependent variable, but they have perhaps not been modeled correctly.
Alternative measures for power
In the literature section, the theory around social networks and constraint was discussed and accepted as a proxy for power. Earlier literature has focused on other measures to determine the independence of the board of directors, where an independent board supposedly curbs the executives’ power. One measure to determine the independency of the board is the percentage of outside directors, defined as being directors that have no connection to the firm other than their directorship (Rosenstein & Wyatt, 1990). In order to determine whether constraint does a better job of explaining variance in total compensation than the percentage of outside directors, I have performed a regression analysis using the percentage of outside directors as the explanatory variable to address Hypothesis 1.
TABLE 11: REGRESSION RESULTS FOR H1 USING PERCENTAGE OF OUTSIDERS AS PROXY FOR POWER
As can be seen from Table 11, the percentage of outsiders, and Tenure, are the only two variables with a significant influence on Total compensation. However, I would expect the influence of
Variable Coefficient Standard error
outsiders on total compensation to be negative, which is the opposite of the found effect. Other variables, except Firm performance, are insignificant but in the expected direction. Firm performance is highly insignificant but the sign is in the opposite direction of what would be expected.
CONCLUSIONS
The research question as set out in the introduction of this thesis was “How can the power derived from social networks be used to extract rents from the employing company?”. This was subsequently split into three hypotheses:
Hypothesis 1: The less constrained an executive is, the larger the total dollar value of his compensation package will be.
Hypothesis 2: The less constrained an executive is, the more likely it is that variable compensation is paid in options.
Hypothesis 3: The less constrained an executive is, the smaller the gap between strike price and the prevailing price at grant date.
Using a sample of 6,094 directors, sitting on the boards of 544 firms, I started the empirical evaluation of these questions by calculating measures of social constraint. I then reviewed the sample’s data and deleted all observations that were incomplete or otherwise not valid for the analysis. The final sample of executives that was analyzed consisted of 748 executives.
In the first hypothesis, I argued that executives with more social capital, indicated by less social constraint, will be able to extract more excessive rents from their companies. Because it is impossible to estimate excessive rents, I used total compensation as a proxy for excessive rent, reasoning that, on average, a higher level of total compensation is an indication of more excessive rents. The results of the regression show a significant negative effect of constraint on total compensation, which is in line with the proposed theory.
camouflage the height of their compensation package. As such, it is expected to be more likely that executives that are less constrained negotiate option contracts as part of their compensation package. The regression results show the expected sign for the impact of constraint on the use of option contracts, but the effect was found to be insignificant. One of the reasons of this insignificant result may be caused by the fact that option contracts also serve as a means to align executives’ goals with those of shareholders (Devers et. al., 2007), and, thus, option contracts can be granted as part of executive compensation for other reasons than camouflage alone.
The third hypothesis focused on a feature of options granted that is difficult to reconcile with classic agency theory: the granting of at the money options. I argued that executives that are less constrained are better able to negotiate favorable options contracts, given that option contracts are part of their compensation package. As a result, the exercise price of the options is closer to the prevailing share price at grant date. Regression results support this notion.
For the first and third hypothesis, the regression analyses showed a significant effect of social constraint. As such, I conclude that the sample provides support for the managerial power approach as advocated by Bebchuk et. al. (2002). Specifically, the analyses’ outcomes indicate that social constraint is an appropriate way to measure power in this context. Unfortunately, the regression results for hypothesis 2 showed no significant effect of social constraint on the use of option contracts as part of executive compensation. With regards to the regression analyses I also want to draw attention to the limited explanatory power that the models were found to exhibit, reflected in low levels of R2. This indicates that there may have been variables of influence that were omitted in this research.
A related problem is the insignificance of variables that were found to be of significant influence in other studies, such as firm size, age, and gender. This may be caused by having chosen imprecise proxies for the underlying variables. For example, for firm performance, I used the change in share price as a proxy for change in firm performance. Another appropriate proxy could have been the growth in earnings or profits. The limited significance of variables may also have been caused by incorrect modeling. For example, effects are assumed to be linear, as I used a linear regression model, but possibly the effects are curved.
Limitations
A major limitation is the definition of social network that was used in this research, being all the individuals that an executive sits on a board with. Intuitively, one could argue that the network that is used to benefit from businesswise is much larger than that, and contains, for example, former university friends, ex-colleagues, and even such non-business like people such as family. This definition is a difficult one to decide upon, especially when considering availability of data. For this type of data to be meaningful, quantity of persons included in the sample is important. However, to obtain the most reliable data, it would be most convenient to interview people directly, which is highly time-consuming for large samples. As such, a careful balance is to be sought for when deciding on the definition of social network.
Another major limitation lies in the definition of total compensation. The managerial power approach is concerned with excessive rents. As discussed in section two, excessive rents consist of compensation received by the executive exceeding the minimum required to convince the executive to work for the company (Bebchuk et. al., 2001). As such, in order to determine whether excessive rent extraction has occurred, it is necessary to determine what the minimum required amount would be for the specific person to concur to work for the specific company. As these data are obviously not freely available, I have used total compensation as a proxy for excessive rent, thereby assuming that a higher total compensation indicates a higher rent extraction. This assumption, although convenient, is unlikely to hold true, as there may be other factors, such as prestige, involved when deciding to accept a job.
Suggestions for future research
BIBLIOGRAPHY
Bebchuk, L. A., & Fried, J. M. (2005). Pay without performance: overview of the issues. Journal
of aaplied corporate finance , 17 (4): 8 - 23.
Bebchuk, L. A., Fried, J. M., & Walker, D. I. (2001). Executive compensation in America: optimal contracting or extraction of rents? The Berkeley law & economics working papers , Article 10.
Bebchuk, L. A., Fried, J. M., & Walker, D. I. (2002). Managerial power and rent extraction in the design of executive compensation. The University of Chicago Law Review , 69: 751 - 846.
Berle, A. A., & Means, G. C. (1932). The modern corporation and private property. New York: Macmillan.
Blyler, M., & Coff, R. W. (2003). Dynamic capabilities, social capital, and rent appropriation: ties that split pies. Strategic management journal , 24: 677 - 686.
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for Windows: Software for social
network analysis. Harvard, MA: Analytic Technologies.
Burt, R. S. (1992). Structural holes. Cambridge, MA: Harvard University Press.
Burt, R. S. (1997). The contingent value of social capital. Administrative science quarterly , 42: 339 - 365.
Burt, R. S. (2000). The network structure of socail capital. In R. I. Sutton, & B. M. Staw,
Research in organizational behavior (Vol. 22). Greenwich CT: JAI Press.
Chan, M. (2008). Executive compensation. Business and society review , 113 (1): 129 - 161. Ciscel, D. H. (1974). Determinants of executive compensation. Southern economic journal , 40 (4): 613 - 617.
Conyon, M. J. (2006). Executive compensation and incentives. Academy of management
perspectives , 20 (1): 25 - 44.
Coombs, J. E., & Gilley, K. M. (2005). Stakeholder management as a predictor of CEO compensation: main effects and interaction with financial performance. Strategic management
Devers, C. E., Cannella, A. A., Reilly, G. P., & Yoder, M. E. (2007). Executive compensation: A multidisciplinary review of recent developments. Journal of management , 33: 1016.
Dorff, M. B. (2005). Does one hand wash the other? Testing the managerial power and optimal contracting theories of exectuive compensation. The journal of corporation law , 30 (2): 255 - 308.
Essen, M. v., Otten, J., & Carberry, E. J. (2012). Assessing mangerial power theory: a meta-analytic approach to understanding determinants of CEO compensation. Journal of management , published online 5 January 2012.
Finkelstein, S., & D'Aveni, R. A. (1994). CEO duality as double-edged sword - How boards of directors balance entrenchment avoidance and unity of command. Academy of management
journal , 37, 1079 - 1108.
Hill, R., Griffiths, W. E., & Judge, G. G. (1997). Undergraduate econometrics. Hoboken: John Wiley & Sons, Inc.
Hogan, T. D., & McPheters, L. R. (1980). Executive compensation: performance versus personal characteristics. Southern economic journal , 46 (4): 1060 - 1068.
Lie, E. (2005). On the timing of CEO stock option awards. Management science , 51 (5), 802 - 812.
McConnel, J. J., & Servaes, H. (1990). Additional evidence on equity ownership and corporate value. Journal of financial economics , 27: 595 - 612.
Murphy, K. (2000). Performance standars in incentive contracts. Journal of accounting &
economics , 30 (3), 283 - 309.
Rosenstein, S., & Wyatt, J. G. (1990). Outside directors, board independence, and shareholder wealth. Journal of financial economics , 26: 175 - 191.
Salvaj, E., & Ferraro, F. (2006). CEO constraint and the dark side of social capital. Best paper
Sanders, W. G., & Carpenter, M. A. (1998). Internationalization and firm governance: the roles of CEO compensation, top team composition, and board structure. Academy of management journal , 41 (2): 158 - 178.
Tosi, H. L., Werner, S., Katz, J. P., & Gomez-Mejia, L. R. (2000). How much does performance matter? A meta-analysis of CEO pay studies. Journal of management , 26 (2), 301 - 339.
Yermack, D. (1996). Higher market valuation of companies with a small board of directors.
