Amsterdam Business School MSc Business Administration Spring semester 2017
What Are Friends For?
A quantitative study on social networks and CEO compensation
Author: Mette Mardbrant – 11386568 Supervisor: Nathan Betancourt
Abstract
With data from the database BoardEx, with a time span from 2010 to 2012, this thesis examines if the compensation level of CEOs on publicly listed U.S. firms is affected by their social network. The size of the social network is measured as number of social ties through three channels: shared alumni network, past and/or current employment, and other activities. The thesis also studies whether firm size moderates the relationship between ones social network and compensation level. The results are significant only when CEOs have social ties via other activities. According to this study, a CEO’s level of compensation is positively affected by her number of social ties via other activities and firm size negatively moderates this relationship. No evidence indicates that the other two types of social connections have significant impact on CEO compensation. Keywords: CEO compensation, Social ties, Large- and Small Cap, Regression Analysis.
Statement of Originality
This document is written by Student Mette Mardbrant who declares to take full responsibility for the contents of this document. No other sources than those mentioned in the text have been used and the text presented in this document is original. The faculty of Business and Economics is solely responsible for the supervision of the work, not for the content.
TABLE OF CONTENT
1. INTRODUCTION ... 5
1.1 Background ... 5
1.2 Motivation and Contribution ... 6
1.3 Structure ... 8
2. LITERATURE REVIEW ... 8
2.1 CEO Compensation ... 8
2.2 Social Networks ... 9
2.3 CEO Compensation and Social Networks ... 12
2.4 Firm Size as a Moderator ... 13
2.5 Research Gap ... 14
2.6 Hypotheses ... 16
3. METHODOLOGY ... 17
3.1 Sample and Data Collection ... 17
3.2 Variables ... 18 3.2.1 Dependent Variable ... 18 3.2.2 Independent Variables ... 18 3.2.3 Moderating Variable ... 19 3.2.4 Control Variables ... 19 3.3 Analytical Strategy ... 20 3.4 Limitations ... 21 4. EMPIRICAL RESULTS ... 22 4.1 Number of Institutions ... 22 4.2 Statistics Summary ... 23 4.3 Regression Analysis ... 25 4.4 Robustness Test ... 30 5. DISCUSSION ... 33 6. CONCLUSION ... 34 6.1 Conclusion ... 34 6.2 Limitations ... 34 6.3 Future Research ... 35 7. REFERENCES ... 36
1. INTRODUCTION
1.1 BackgroundIn recent years, there has been an increased attention by the media, politicians and academics on what, according to some people, appears to be overly generous executive compensation. CEO compensation is a controversial phenomenon and after the latest financial crises in 2008 the subject has been particularly debatable and authors have paid significant attention to the subject. Above all, the subject has gotten attention in the United States where across the S&P 500 companies the average compensation level of a CEO was 204 times higher than that of an average worker in 2013 (Bloomberg, 2015). A study by the Economic Policy Institute (2015) shows that on average CEO compensation keeps rising. After the latest financial crisis it is particularly interesting to investigate the circumstances around this and examine what factors that may have an impact on the CEOs’ compensation levels.
Critics argue that CEO compensations are too high and that the relationship between compensation and performance is inadequate. During recession the subject is particularly sensitive as it happens that generous salaries and high bonuses have been given to CEOs despite a lack of satisfactory results from the company (Dvorak, 2009). For example, the former CEO of Merrill Lynch, Stanley O’Neal, was given a pay raise by almost 50 percent and ended up with a salary larger than 28 million dollars during the same time that the company laid off more than 20,000 people over a three-year period (Beiser, 2004). Why the boards of directors approve these extensive compensation packages does not always make sense and can be questioned. A large amount of research has been done on this subject and above all researchers seek to explain the underlying factors behind the variance in CEOs’ compensation levels. Early research focused on the relationship between firm performance and CEO compensation, but
many found little or no evidence to support this relationship (Gomez-Meija et al, 1987). As a result, researchers have begun to move beyond established frameworks, to focus more on the social and political context in which the pay is set to search for explanations. This paper aims to further add to the understanding of the underlying factors behind these compensation packages.
Traditionally, the concept of CEO compensation has been predicted on economic theories such as agency theory, tournament theory and neoclassical notions of marginal product (Tosi & Gomez-Meija, 1989; Lazear, 1989; Jensen & Murphy, 1990). However, in later years the subject has been studied from numerous other perspectives. One theoretical field that has been commonly applied is that of Social Network Analysis, which focuses on “how the social structure of relationships around a person, group, or organization affects beliefs or behaviors” (Şule Erçetin & Banerjee, 2014, p. 108). In line with this emerging research (Hallock, 1997; Fich & White, 2003; Barnea & Guedj, 2006), this thesis will further seek to explain how a CEO’s social network may influence her level of compensation.
1.2 Motivation and Contribution
When considering prior research within this field, a crucial aspect that seems to have been overlooked is whether there is a difference in social networks’ influences on CEO compensation under varying circumstances. The relationship between a CEO’s social network and level of compensation may depend on a third factor, such as gender, tenure or age. This however, has rarely been studied.
The majority of previous research has been made on large firms whereas little attention has been drawn towards small or medium-sized businesses. Large and small firms differ in several ways. First of all, larger firms are more visible and subject to
stricter governance structures relative to smaller firms (Horton et al, 2012). They also receive more media coverage (Switzer, 2007) and usually have more experienced CEOs. Small companies, on the other hand, assumedly have limited access to resources relative to larger firms and may therefore be exposed to larger risks. A comparison between these makes it possible to examine whether social networks have a differential impact on CEO compensation, depending on whether the CEO is working for a large or small firm.
This paper attempts to extend existing research on CEO compensation and social networks. To do this I examine S&P 500, a stock market index of 500 large companies publicly listed on the New York Stock Exchange or NASDAQ, and S&P SmallCap 600, a stock market index that covers a broad range of small-cap stocks in the United States. By examining these I aim to explore whether the relationship between a CEO’s social network and her compensation differ depending on the size of the firm.
This paper makes several contributions to the existing literature on corporate governance and social networking – an area in which more research is needed (Geletkanycz, Boyd & Finkelstein, 2001). First of all, the paper shed new light on how social networks influences CEO compensation and thereby contributes to a broader understanding of the underlying factors behind the variance in CEOs’ compensation levels. The paper also provides insights to the Social Network Theory by extending the existing literature by making a comparative examination on social network effects, more specifically by using firm size as a moderator. By doing this, social networks’ impact on CEO compensation under varying circumstances is better understood. Finally, as most previous research on social networks has focused on board interlocks, i.e. the practice of directors sitting on each other’s corporate boards’ (Davis, 1991), this study will focus on the social ties that one develop outside of ones work place.
1.3 Structure
The paper is organized as follows. In section 2 the main insights of existing literature on CEO compensation and social networks are described. Research questions as well as the purpose of the paper will be presented. Finally hypotheses will be formulated with regards to the literature review. Section 3 presents the methodological approach of the paper. Moreover, the sample and the procedure for how data was collected will be described. In section 4 the results of the study will be presented and the hypotheses will be tested. Section 5 will discuss the results of the study and conclusions will be drawn. Lastly, limitations of the paper as well as recommendations for future research will be discussed.
2. LITERATURE REVIEW
2.1 CEO CompensationAs previously mentioned, a lot of research has been done on the topic of CEO compensation from a range of various perspectives, with mixed empirical findings. A significant amount of previous studies has been done on the determinants behind the variance in CEO compensation level with the purpose to find a comprehensive explanation. Traditionally, many researchers have tried to explain the variance in CEO compensation based on classical theoretical theories such as Agency Theory and Tournament Theory (Tosi et al, 2000; Murphy, 1999; Jensen & Murphy, 1990).
A large number of previous research focuses on the relationship between CEO compensation and various firm characteristics. Many show that the size and/or performance of the company, e.g. in terms of number of employees or revenue, have a positive effect on CEO compensation (Tosi et al, 2000). For example, Miller (1995) examined the influence of firm performance on CEO compensation and found, among
other things, a positive relationship between CEO compensation and ROE. Murphy (1985), who studied 501 managers in 72 companies, found a positive relationship between CEO compensation and firm performance in terms of shareholder return and growth in firm sale. On the contrary, Jensen & Murphy (1990) examined 1400 American firms between the years 1974 to 1988 and found that firm size and CEO compensation are positively correlated whereas the relationship between firm performance CEO compensation and is basically non-existing. These mixed empirical findings makes it difficult to draw any general conclusions. Even the studies with positive significant results show a weak correlation and many studies only explain 20 to 30 percent of the variance in CEO compensation (Crystal, 1991).
It appears as if traditional economic variables, usually in terms of size and performance, only can explain the variance in CEO compensation to a certain extent. Essentially, it seems like one have to extend the research beyond the traditional economic variables to understand something about our social human behavior that theses studies have not managed to capture. In later years, researchers have shifted focus towards more psychological, political, behavioral and social approaches, as these are known to shape individual- as well as group decisions, and are therefore likely to influence compensation levels (Baker et al, 1988).
2.2 Social Networks
The concepts of Social Network Analysis are originated from broad schools of thoughts such as psychology, anthropology and sociology but can be applied a wide range of areas and has been commonly applied to explain certain economic behavior. In one of the most cited articles on this subject, “Economic Action and Social Structure: The Problem of Embeddedness”, Granovetter (1985) argues that economic behavior is embedded in
networks of social relations. His ambition was to counter the economic idea of human behavior in isolation from its social context. By viewing companies as a set of social groupings, Social Network Analysis provides a tool to conceptualize the interactions within and between organizations.
Ganovetter is one of the main contributors within Social Network Analysis and in another of his commonly cited article “The Strength of Weak Ties” (1973) he developed the concept of tie strength and information flow. He argues that social ties between actors can be used to understand flow of information among social actors. According to Graonvetter, ones social network can channel information flow and can therefore be seen as a valuable source to the organization you work for. The author distinguishes between strong and weak ties. The strength of the ties reflects the closeness of the relationship between individuals. Strong ties are the ones you have with people that you are in frequent contact with, such as close friends and family. Weak ties are more distant and you have less frequent contact with these people. The author argues that strong ties tend to hold redundant information since they interact often and thus develop similarities whereas weak ties are more likely to contribute with novel information. Since then, a significant amount of research has been done on this subject. Engelberg et al (2012) found that social ties to a lender tend to lead to more beneficial conditions. Duchin and Sosyura (2013) argue that the strength of the tie a division manager has with her CEO has a positive effect on the investment capital awarded in terms of internal capital budgeting. The effect of social ties has also been studied on the external labor market and for example, Holzer (1987) found that your chance of being employed at a firm increases significantly if a person with whom you have a social tie with is already employed at the firm.
Social Network Analysis and the concept of social ties have been regularly applied on the subject of CEO compensation and have primarily focused on board interlocks, in which one analyzes the practice of directors sitting on each other’s corporate boards’ (Davis, 1991; Hallock, 1997; Fich and White, 2003; Barnea and Guedj, 2006). Davis (1991) article “Agents without Principles? The Spread of the Poison Pill through the Intercorporate Network” is one of the most cited articles on board interlocks, through which he provides a social context for corporate governance that supports the idea of adoption of common practices, such as CEO compensation, across interconnected firms. Hallock (1997) did a similar study and found that the compensation level of interlocked CEOs on average is higher than the compensation level of CEOs that do not sit in interlocked boards. Fich and White (2003) as well as Barnea and Guedj (2006) also found a strong positive relationship between interlocking boards and CEO compensation. In accordance to Granovetters argumentation, the bottom line in all these articles is that they find it likely that the interlocks serve as a device for transmission of information and knowledge.
Although the concept of board interlocks have been proven to explain some of the variance in CEO pay these studies excludes the social ties that may have been developed outside of ones work place. In a study by Chidambaran et al (2010) on CEO connections and corporate fraud they distinguish between professional connections and non-professional connections. Professional connections are the ones you have with people you have worked with and non-professional connections are the ones you have via other channels such as shared alumni network and other social activities. Less research has been drawn towards the latter category of social ties, which makes this particularly interesting to investigate.
In this paper, the size of a CEO’s social network is captured by her number of social ties. The number of social ties, i.e. the number of people with whom the CEO is connected to, is measured by the total number of people the CEO is acquainted to through her education, past and/or current employment, or other social activities (e.g. golf clubs). A link with another CEO will increase the size of her network by one.
2.3 Social Networks and CEO compensation
In previous literature, there seem two be two dominating views regarding social networks and CEO compensation. One view argues that CEO compensation follows a pay-for-performance model. As previously mentioned, a significant amount of previous research argues that a larger social network comes with benefits as they enhance the flow of knowledge and information (Adler and Kwon, 2002; Burt, 2000). Even if the information eventually reaches everyone, this may take time and individuals with greater social networks are likely to be informed earlier and more broadly, which gives them an advantage (Burt, 2000). Useem (1984) extends this argument further by stating that social networks gives you access to a broad range of business intelligence that enables managers to anticipate the latest business practices and the overall business environment, which may provide valuable access about new innovations. According to this view, a CEO’s social network may be viewed as an important resource that enables her to be a better decision-maker, and thus a valid criterion to take into consideration when determining the compensation level.
The second view argues that CEO compensation is unrelated to performance. Bebchuk and Fried (2004) argue that a large social network may enhance “managerial power”, which means that the CEO has control over the board of directors, and can use this power to negotiate a more favorable compensation package. This may create less critical
thinking and opportunities of pay-without-performance (Ibid.). Research has also found that a large social network may work as an indication of higher status (Belliveau et al, 1996). With limited possibilities to measure a CEO’s true value and marginal contribution at a given job (Finkelstein & Hambrick, 1988), the board may consider status as an important indication of your worth.
The two views provide different approaches to the relationship between a CEO’s social network and her level of compensation. Nevertheless, both theories indicate that the larger the size of a CEO’s social network, the higher the compensation level.
2.4 Firm Size as a Moderator
It has been argued that firm size can play an important role on various corporate governance decisions (Voulgaris et al, 2004), which makes it interesting to investigate whether firm size has a moderating effect in this specific relationship between ones social network and compensation level.
Regarding CEO compensation, the attention by the media as well as from academics and politicians has predominantly been on the largest publicly listed American companies and little attention has been given to smaller firms. One might raise concern about this since large and small companies differs in various ways and the findings on large companies may not always be applicable to the smaller ones.
First of all, the boards of directors of smaller companies are often faced with unique governance challenges with limited resources to help them. Access to resources provides strategic alternatives and financial flexibility and thus smaller companies may be exposed to larger risks (González & González, 2012). One could also argue that it may be easier for larger companies to hire third-party experts as advisors whilst smaller companies have to rely on existing social connections. This indicates that the access to
information that a large social network provides may be of bigger importance to smaller companies.
Secondly, CEO of larger firms usually has more experience and thus can be expected to have more power within the firm and the decision-making process. As previously described, a large social network can enhance “managerial power” and may be of bigger importance for less experienced CEOs in order to negotiate a favorable compensation package.
Finally, multiple studies show that larger firms receive more media coverage than smaller firms (Switzer, 2007). If smaller firms lack the finances to run advertising campaigns this may imply that they have to rely on their social network for support to get their organization’s name out there. This also points towards the view that ones social network is of greater importance for CEOs of smaller firms than for those of larger firms. Based on these differences it is interesting to investigate whether the impact of a CEO’s social network on her compensation level will vary between different firms depending on its size. 2.5 Research Gap Although past research on social networks and CEO compensation is quite widespread, this paper distinguishes itself from others as it do not only investigate the relationship between a CEO’s social network and her compensation level but it does so in a comparable setting. Prior research has investigated determinants of CEO compensation and are usually limited to large firms. Rather than focusing on CEO compensation and its determinants to present a new best model to explain as much as possible of the variance in CEO compensation, this paper differs from previous work as it focuses specifically on
the differences in the relationship between a CEO’s social network and her pay depending on whether she is CEO of a firm listed on Large- or Small Cap. After having examined previous research, a gap have been identified as many have focused on social networks and CEO compensation but fewer has examined social networks’ influence on CEO compensation in a comparable setting. No previous research has been found on the moderating effect of the size of the firm on the relationship between ones social network and compensation level. Moreover, this study focuses on the social ties a CEO has developed outside of the work place, as the majority of previous research has focused on the concept of board interlocks.
In other words, there are two purposes of the thesis. First of all, the purpose is to examine how much of the variation in total CEO compensation that can be explained by a number of selected variables as indicators of the CEO’s social network outside of ones workplace. The social network of a CEO can be measured in a number of different ways. In this thesis, three variables as indicators of ones social network are used: educational background, past and/or current employment, and other activities in terms of memberships in clubs, organizations or charities. In this paper, two CEOs are socially connected if they are connected through any of these three variables. Secondly, the paper aim to examine whether there is a difference in the relationship between a CEO’s social network and her compensation level depending on whether the firm is listed on Large- or Small Cap. The following research questions has been formulated: v What affect does a CEO’s number of social ties have on her compensation level? v What affect does firm size have on the relationship between the number of social ties a CEO has and her compensation level?
2.6 Hypotheses Two hypotheses have been formulated with regards to the literature review. As previously mentioned a CEO’s social connections can either be viewed as an indication of high status or as an important firm resource. Either way, these can both be viewed as criterion to consider when determining CEO pay. The large majority of previous research concludes that a CEO’s social network has a positive relationship with her level of compensation. In accordance to this, I propose that each of the three indicators of ones social network will have a positive influence on the compensation level you receive. Put differently, the larger number of social ties a CEO has, the higher will her level of compensation be. The first hypothesis that follows is: H1: A CEO’s level of compensation is positively affected by her number of social ties. As mentioned, one dominant difference between large and small firms is that small firms usually have limited resources, which typically means that they are exposed to larger risks. Previous research argue that getting the right information appears to be crucial for firm survival to be able to compete with larger firms with access to more resources. A larger social network may enhance finding relevant information more quickly and may increase the possibility to access resources that are not available in-house. I propose that social connections are probable to be of higher importance to smaller firms. In other words, I believe that the impact social connections have on CEO compensation weakens with firm size. The second hypothesis that follows is:
H2: Firm size negatively moderates the impact of a CEO’s number of social ties on her level of compensation.
3. METHODOLOGY
3.1 Sample and Data Collection
In general there seems to be a consensus that CEOs in the United States are paid a lot more than their foreign competitors (Fernandes et al, 2012), which makes it particularly interesting to investigate these companies. The research will be done on the companies and their corresponding CEOs listed on S&P 500, a stock market index of 500 large companies publicly listed on the New York Stock Exchange or NASDAQ, and S&P SmallCap 600, a stock market index that covers a broad range of small-cap stocks in the United States. CEO compensation for top executives at publicly traded U.S. companies is available at the ExecuComp database from COMPUSTAT, a division of Standard & Poor’s. ExecuComp calculates total compensation as the sum of base salary, bonus, stock options and compensation from long-term incentive plans, pension benefits and other benefits. Extensive biographical information on CEOs, including their educational background, employment history and other activities, will be acquired from WHARTON’s database BoardEx. Other sources have to be consulted too when certain data is missing or incomplete. This data has been acquired from Bloomberg or from the Annual Reports of respectively company.
In accordance to previous literature, financial firms will be excluded from the study as they differ significantly form other industries in several aspects related to corporate governance. In this study, 98 financial institutions will be excluded from S&P 500 and 118 financial institutions will be excluded from S&P SmallCap 600.
Not all companies were listed on Large- respectively Small Cap for the whole examined period. These companies were also excluded from the study. Finally, some companies were excluded from the study due to incomplete or missing data for one or more years. The final sample consists of 1026 observations on 342 companies.
Data on the variables was collected from three years, 2010 to 2012. This due to the fact that the BoardEx database coverage is limited or incomplete after 2012. These years however are particularly interesting to study since it was just a few years after the financial crisis and generous compensation packages were highly debated during this time. 3.2 Variables 3.2.1 Dependent Variable Total CEO compensation. Total CEO compensation is calculated as the sum of base salary, bonus, stock options and compensation from long-term incentive plans, pension benefits and other benefits. In accordance to previous studies (Fich & White, 2003; Murphy, 1985; Henderson & Fredrickson, 1996) the natural logarithm was calculated for total compensation. The benefits of using the natural logarithm are that it facilitates comparisons with previous studies and reduces skewness (Murphy, 1985). It should be noted that valuing compensation from long-term incentive plans is not a straightforward process and the pay one ultimately receive is uncertain at the time the compensation is awarded. Despite this, the choice has been made to include these in the study to increase the reliability since almost all CEOs receive some compensation from long-term incentive plans. 3.2.2 Independent Variables Education network. A tie exists between a two CEOs if they went to the same education institution. Employment network. A tie exists between two CEOs through their current or past job if they have worked at the same firm or served on the same board of directors.
Other activities network. A tie exists between two CEOs through other activities if they have memberships in the same clubs, organizations or charities.
In BoardEx, the variables are denoted by the institution name. For example, for social ties via education network the variables are denoted as e.g. Boston University, Yale University and Brooklyn College. These are transformed into continuous variables to count for the number of CEOs who went to the school with the same institution name. Finally, as a CEO can belong to several alumni networks, a new variable is generated which is the summation of the continuous variables.
3.2.3 Moderating Variable
Large Cap respectively Small Cap, which measures market capitalization, will be included in the regression as a representation of the size of the firm. Market capitalization is a measurement of the total value of all the shares of stock a company has issued (S&P Dow Jones Indices, 2017). In order to be listed on S&P 500 a company must have a market capitalization of more than 5,3 billion U.S. dollars and to be included on S&P SmallCap 600 a company must have a market capitalization between 400 million U.S. dollars to 1,8 billion U.S. dollars (Ibid.). 3.2.4 Control Variables The regression will include a number of control variables. To account for any possible relationship between CEO compensation and firm performance, several variables as indicators of firm performance will be included in the study. Revenue and market value are commonly used as control variables (Lilling, 2006; Miller, 1995) and will be included in this research too. Data for the performance measurements will be collected from the ExecuComp databse from CAMPUSTAT. Number of employees is another variable that
has been found to influence a CEO’s compensation level (Tosi et al, 2000). This information will also be gathered from the ExecuComp database from COMPUSTAT. Three variables as indicators of CEO characteristics will also be used as control variables. CEO tenure, i.e. number of years as CEO, age and gender are three variables that assumedly correlate to CEO compensation (Henderson & Fredrickson, 1996) and will be included in this study too. This data will be collected from the database BoardEx. Gender is a dummy variable and has the value 0 when the CEO is a male and 1 when the director is female. The schematic idea of the research model can be found in figure 1. Figure 1 Research Model. 3.3 Analytical Strategy
In examining the research question, the thesis is built on a deductive approach. Two hypotheses has been formulated based on a discussion on relevant theories and will be tested using a quantitative approach, namely the ordinary least square (OLS) regression
method. The thesis will be based on a quantitative method since the purpose of the study includes measuring and analyzing a large number of quantitative data (Bryman and Bell, 2013).
Three regression models, one for each independent variable, were done to test the relationship between ones social network and CEO compensation.
In a simple regression model, Y is the dependent variable and α is the intercept, which indicates the value of Y when X is equal to zero. β is the Beta coefficient for X, and X is the independent variable that is explaining the variance in Y. ε is the random disturbance term. The regressions have the following equation: Education Network: Log(TotalComp1) = α + β1ConnEDU + β2ConnEDU*Cap + β3Cap + Β4CEOTen + β5Age + β6Gender + β7Empl + β8Rev + β9MV + ε Employment Network: Log(TotalComp1) = α + β1ConnEMPL + β2ConnEMPL*Cap + β3Cap + Β4CEOTen + β5Age + β6Gender + β7Empl + β8Rev + β9MV + ε Other Activities Network: Log(TotalComp1) = α + β1ConnACT + β2ConnACT*Cap + β3Cap + Β4CEOTen + β5Age + β6Gender + β7Empl + β8Rev + β9MV + ε 3.4 Limitations Due to the scope and time frame of the paper, the thesis only focuses on observable and formal social ties although other social ties, such as friends of friends, may also have an impact on CEO compensation. In other words, we only look at direct ties between CEO A and CEO B, even though there might be a possibility that they have an indirect tie if they
are both connected to CEO C. This may imply that the size of ones social network may be underestimated. On the other hand, the study does not measure the possibility that some of the past ties may no longer exist, and thus the size of ones social network could might as well be overestimated.
Due to practical reasons, the companies on S&P 500 and S&P SmallCap 600 were examined, as there is most relevant data available on these companies. Based on market capitalization, these companies make out 1.100 of the 1.500 largest companies publicly listed in the United States and one could argue that there are no “small” companies included in the sample. In reality, the moderating effect of firm size on the relationship between a CEO’s social network and her compensation level may be even bigger.
4. EMPIRICAL RESULTS
4.1 Number of Institutions In this sample, approximately 41 percent have obtained one education qualification and around 49 percent have obtained 2 education qualifications. Roughly 9 percent have obtained 3 education qualifications and only around 1 percent has obtained more than 3 education qualifications. As a whole, the CEOs in this sample went to 419 different education institutions. The most common education institutions among the sample are Harvard University, Stanford University and University of Pennsylvania.In terms of employment, approximately 8 percent of the CEOs in this sample have had one previous employment and another 8 percent have had two previous employments. Roughly 10 percent have had three previous work places and the great majority, 73 percent, have had more than three previous work places. In total, the CEOs in this sample have been employed at 2647 different work places.
Lastly, in terms of other activities, approximately 20 percent of the CEOs are members in only one club, charity or organization and around 18 percent are members in two. Approximately 18 percent are members in three clubs, charities or organizations, and the great majority, around 44 percent, are members in more than three. In total, the CEOs in this sample are members of 1557 different clubs, charities or organizations.
A summary of number of institutions can be found in table 1.
Education Employment Other activities
One institution ,41 ,08 ,20
Two institutions ,49 ,08 ,18
Three institutions ,09 ,10 ,18
More than three
institutions ,01 ,74 ,44
Total 100% 100% 100%
Table 1, Number of institutions.
4.2 Statistics Summary
Table 2 shows a descriptive summary of the dependent variable, the independent variables, the moderator and the control variables. The variable “TotalComp1” is the dependent variable, which measures a CEO’s total compensation. On average, a CEO has a total compensation of approximately 7 million U.S. dollars per year. The variable “Cap”, our moderator, is a dummy variable with 0 being Large Cap and 1 being Small Cap. 52 percent of the CEOs works for companies listed on Small Cap. The independent variables
“ConnEDU”, “ConnEMPL” and “ConnACT” measures the social ties a CEO has develop via education network, employment network or other activities network. On average, a CEO has 5,5 social ties via education, 6,24 social ties via employment and 1,25 social ties via other activities. The control variable “CEOTen” measures CEO tenure and shows that on average, the CEOs in our sample has been the CEO of the company for approximately 8 years. The control variable “Age” indicates that on average, the CEOs in our sample are around 57 years. “Gender” is a dummy variable that has a value of 0 when the CEO is male and 1 when the CEO is female. Only 2 percent of the CEOs in the sample are female. The control variable “Empl” measures the number of employees at the company that the CEO works for. On average, the companies have roughly 20 thousand employees. “Rev” is another control variable that measures the revenue of the company that the CEO works for. Roughly 10 billion U.S. dollars is the average revenue of the companies. Finally, “MV” is the last control variable in the sample and indicates the market value of the company that the CEO works for. On average, the companies have a market value of approximately 12,5 billion U.S. dollars. The statistics summary can be found in table 2.
Note: “TotalComp1” is the dependent variable and is short for total compensation. “Cap” is a
dummy variable with 0 being Large Cap and 1 being Small Cap. “ConnACT” is the independent variable and represents social connections via other activities. “CEOTen” is CEO tenure, i.e. number of years as CEO. “Age” is the age of the CEO. “Gender” is a dummy variable with 0 being male and 1 being female. “Empl” is number of employees at the company that the CEO works for. “Rev” is the revenue of the company that the CEO works for. “MV” is the market value of the company that the CEO works for.
Variable Observations Mean Std. Dev.
TotalComp1 1026 7004,656 6907,371 Cap 1026 ,52 ,500 ConnEDU ConnEMPL ConnACT 1026 1026 1026 5,50 6,24 1,25 6,24 7,965 2,408 CEOTen 1026 7,81 6,942 Age 1026 56,58 7, 453 Gender Empl 1026 1026 ,02 20,046 ,148 38,409 Rev MV 1026 1026 10079,436 12434,182 30504,960 32049,474 Table 2, Statistics summary. 4.3 Regression Analysis When testing the hypotheses, the regression was done in SPSS using an ordinary least square (OLS) regression method. Before running our regression, the numerical independent variables as well as the numerical control variables were standardized. This is particularly useful if the variables are written in different scales, such as
percentages and absolute values, as it puts them all on a common scale, which makes them easier to compare.
Three regression models, one each for respectively independent variable, were done to test the relationship between ones social network and CEO compensation. The result shows us that only the number of social ties a CEO develop via other activities shows a significant impact on our dependent variable, compensation level. Essentially, in this study it appears as if the number of social ties one have developed through education and past and/or current employment does not have any impact on the CEOs’ compensation level. However, as mentioned, a positive significant relationship was found between number of social ties through other activities and compensation level, which means that the more social ties a CEO has through her memberships in clubs, organizations and charities, the higher will her level of compensation be, ceteris paribus. The regression results can be found in table 3.
Note: “Cap” is a dummy variable with 0 being Large Cap and 1 being Small Cap. “ZConnACT” is
the independent variable and represents social connections through other activities. “int_1” is the interaction term, i.e. ZConnACT*Cap. “ZCEOTen” is CEO tenure, i.e. number of years as CEO. “ZAge” is the age of the CEO. “Gender” is a dummy variable with 0 being male and 1 being female. “ZEmpl” is number of employees at the company that the CEO works for. “ZRev” is the revenue of the company that the CEO works for. “ZMV” is the market value of the company that the CEO works for. The Z means that the variable has been standardized.
Coeffficient Std. Error t Sig.
Intercept ,624 ,034 18,593 ,000 Cap -1,162 ,050 -23,125 ,000 ZConnACT ,046 ,027 1,687 ,032 int_1 ,151 ,057 2,643 ,008 ZCEOTen -,067 ,024 -2,762 ,006 ZAge -,019 ,024 -,777 ,437 Gender ZEmpl -,064 ,077 ,148 ,029 -,431 2,658 ,666 ,008 ZRev ZMV -,051 ,158 ,045 ,046 -1,145 3,417 ,252 ,001 R² = 0,527 p <0,001 Durbin-Watson 1,026 Table 3, Regression results. The interaction term tells us how much of the effect of social ties through other activities on total compensation is different between companies listed on Large respectively Small Cap. In this regression we can see that the result is significant, which means that the effect of social ties through other activities on CEO compensation level depends on the size of the firm.
Surprisingly, the regression demonstrates that the dummy variable Cap shows a negative relationship with CEO compensation, which means that CEOs of companies listed on Large Cap will receive a lower compensation level than CEOs of companies listed in Small Cap. As discussed, large and small firms differ in many ways and one of the most obvious differences is the media coverage and brand awareness. It is possible that smaller firms need to offer more generous compensation packages in order to attract the right talent, which would explain the results in the regression. In terms of social connections, the regression shows that two CEOs that differ by one unit through their social ties via other activities are estimated to differ by approximately 0,05 units on compensation level. The correlation is positive, meaning that those with relatively more social ties through other activities are estimated to have a higher compensation level. More specifically, in this study, the increase of one social tie via other activities will result in an increase of approximately 5 percent on the CEO’s pay, certeris paribus. The regression also shows us that our control variables number of employees and market value also has positive significant impact on a CEO’s compensation level. More specifically, two CEOs that differ by one unit of number of employees are estimated to differ by approximately 8 percent on compensation level and two CEOs that differ by one unit on market value are estimated to differ by approximately 16 percent on compensation level. An interesting observation is that the control variable CEO tenure shows a negative significant relationship with CEO compensation. Two CEOs that differ by one unit of CEO tenure are estimated to differ by approximately 7 percent on compensation level. Intuitively, this may sound strange but companies are always looking for skilled labor
and offering a pay raise may be a way to attract talent, which could result in a shorter tenure among CEOs. No other variables showed a significant relationship with our dependent variable. Nonetheless, it can still be mentioned that both gender and age presented a negative relationship with compensation level, meaning that in our sample, males will receive a lower compensation level than females, and the older you are, the lower will your compensation level be, ceteris paribus. As these variables show insignificant results I will not analyze the results in detail. Overall, the result shows a significant model with an Rsquare of approximately 0,53. This means that the model explains approximately 0,53 percent of the total variance in compensation level among the CEOs our sample. The Durbin-Watson Test is a measure of autocorrelation, where a value of 2 means no autocorrelation. The regression shows a value of approximately 1,03, which means that positive autocorrelation can occur. However, Field (2009) suggests that values less than 1 or more than 3 are a definite cause for concern and if the value is higher than 1, like in our case, it is only a possible cause for concern. This study also seeks to explore whether there is a difference in the relationship between ones number of social ties and compensation level depending on whether the CEO works for a firm listed on Large- respectively Small Cap. The moderating results can be found in table 4.
Cap Coeffficient Std. Error t Sig.
Large Small ,046 ,197 ,022 ,060 2,100 3,319 ,036 ,001 Table 4, Moderating results. By probing the significant interaction, we can see that the effect of number of social ties through other activities on level of compensation is significant for both large and small companies. In other words, social ties through other activities increase your level of compensation among both large and small companies, but the effect appears to be stronger among smaller companies. Two CEOs that differ by one unit through their social ties via other activities among large companies are estimated to differ by approximately 5 percent on compensation level, but two CEOs that differ by one unit through their social ties via other activities among small companies are estimated to differ by approximately 20 percent on compensation level. This supports the idea that social connections appears to be of higher importance among CEOs of small companies than those of large companies. 4.4 Robustness Test The robustness of the regression can be tested by adding or removing certain variables from the regression to see if this modifies the regression results. If the results remain plausible and robust that can be an indicator of structural validity (White & Lu, 2010). In table 5, the variable “ZEmpl”, i.e. number of employees at the company that the CEO works for, has been removed from the regression. When comparing the results from this
regression with the results from the original regression, which can be viewed in table 3, we notice that they only differ marginally. The coefficients in table 5 do not deviate significantly from their counterparts in table 3 and in particular, the significance level on all the variables remains the same. The results from the robustness test after removing the variable “ZEmpl” can be found in table 5. Note: “Cap” is a dummy variable with 0 being Large Cap and 1 being Small Cap. “ZConnACT” is
the independent variable and represents social connections through other activities. “int_1” is the interaction term, i.e. ZConnACT*Cap. “ZCEOTen” is CEO tenure, i.e. number of years as CEO. “ZAge” is the age of the CEO. “Gender” is a dummy variable with 0 being male and 1 being female. “ZRev” is the revenue of the company that the CEO works for. “ZMV” is the market value of the company that the CEO works for. The Z means that the variable has been standardized.
Coeffficient Std. Error t Sig.
Intercept ,638 ,033 19,200 ,000 Cap -1,194 ,049 -24,399 ,000 ZConnACT ,069 ,026 2,632 ,009 int_1 ,130 ,057 2,287 ,022 ZCEOTen -,065 ,024 -2,660 ,008 ZAge -,023 ,024 -,964 ,335 Gender ZRev -,088 -,039 ,148 ,044 -,592 -,887 ,554 ,375 ZMV ,182 ,045 4,018 ,000 R² = 0,524 p <0,001 Durbin-Watson 1,021 Table 5, Robustness test after removing the variable “ZEmpl”.
When the variable “ZMV”, i.e. market value of the company that the CEO works for, is dropped together with the variable “ZEmpl” the significance level for “ZRev”, i.e. the revenue of the company that the CEO works for, changes and the coefficient changes from negative to positive. This indicates that the original regression may not be robust, which should be taken into consideration when reviewing the results. However, the coefficients and significance levels of all non-control variables are still unchanged.
The results from the robustness test after removing the variables “ZEmpl” and “ZMV” can be found in table 6.
Note: “Cap” is a dummy variable with 0 being Large Cap and 1 being Small Cap. “ZConnACT” is
the independent variable and represents social connections through other activities. “int_1” is the interaction term, i.e. ZConnACT*Cap. “ZCEOTen” is CEO tenure, i.e. number of years as CEO. “ZAge” is the age of the CEO. “Gender” is a dummy variable with 0 being male and 1 being female. “ZRev” is the revenue of the company that the CEO works for. The Z means that the variable has been standardized.
Coeffficient Std. Error t Sig.
Intercept ,661 ,033 20,037 ,000 Cap -1,236 ,048 -25,693 ,000 ZConnACT ,071 ,026 2,722 ,007 int_1 ,127 ,057 2,224 ,026 ZCEOTen -,059 ,024 -2,411 ,016 ZAge -,029 ,024 -1,198 ,231 Gender ZRev -,140 ,113 ,149 ,023 -,939 4,802 ,348 ,000 R² = 0,516 p <0,001 Durbin-Watson 1,014 Table 6, Robustness test after removing the variables “ZEmpl” and “ZMV”.
5. DISCUSSION
Based on the results from our regression we find support for both of the hypotheses. It appears as if a CEO’s level of compensation is positively affected by her number of social ties. However, support could only be found on the relationship between compensation level and number of social ties via other activities. As mentioned in the literature review, Granovetter (1973) distinguishes between weak and strong ties. Granovetter argues that people with stronger ties tend to be more alike than they are with people with whom they have weak ties. In other words, novel information, which is probable to be of higher value to the firm, is more likely to come from weak ties. Social ties via other activities differ from social ties via education and employment as memberships in clubs, charities or organizations only are hobbies and not a full-time position. One could argue that these ties are probable to be weaker and thus of higher value for the company, which could explain why only the social ties developed via other activities have a positive significant impact on CEO compensation. The results also shows that firm size weaken the impact of social connections on compensation level. As discussed, larger firms usually generate higher and less volatile profits than small firms. Small firms also tend to have lower liquidity and limited access to resources, which indicates that smaller firms are exposed to larger risks. This suggests that access to information and knowledge might be of bigger importance to smaller companies, which would be consistent with the results of the study.6. CONCLUSION
6.1 ConclusionTo conclude, the research questions of the study was to investigate what affect a CEO’s number of social ties have on her compensation level, and what effect firm size have on this relationship. I analyzed social connections through three types of channels and found that social connections via other activities have a significant positive impact on CEO compensation. In other words, this supports our first hypotheses which said that a CEO’s level of compensation is positively affected by her number of social ties. Social ties via other activities are assumedly weak and are likely to provide more novel information than ties developed through education or employment, which can be a possible explanation why only these types of ties showed a positive significant result. Consistent with our second hypotheses, the results also indicates that firm size negatively moderates the impact of a CEO’s number of social ties on her level of compensation. Access to information and knowledge appears to be more crucial for smaller firms as they have limited access to resources and are exposed to larger risks. 6.2 Limitations Some of the information from the BoardEx database is inconsistent. For example, the database could tell us that one CEO graduated from Harvard Business School and another from Harvard University. When these variables are transformed into continuous variables to count for the number of CEOs who went to the institution with the same name they will count as two separate institutions, and will not be connected in the regression analysis. Due to the scope and time frame of the research, it was not possible to adjust these frictions manually. On the other hand, another of the limitations with the study is the assumption of social connections. As previously mentioned, it is
possibly that even though two CEOs for example went to the same education institution, they might as well not know each other at all.
6.3 Future Research
As mentioned, there are two dominating views regarding social networks and CEO compensation. One view argues that CEO compensation follows a pay-for-performance model as the social ties enhance access to valuable and information and knowledge. The other view suggests that the compensation level is unrelated to performance but may enhance “managerial power” and status. It would be interesting to take a deeper look into this to see if and how the size of ones social network affect firm value.
Moreover, as mentioned, little research has been done on the relationship between social connections and compensation level in a variable setting and the result of this study indicate that such an approach may be promising. It would be interesting if more studies would examine the affect of a third moderating variable on the relationship. For example, an interesting aspect would be to investigate how the relationship may differ in various industries. For firms with a large focus on R&D, having access to information and knowledge is crucial for firm survival and it is possible that they value the size of ones social network higher than other industries.
7. REFERENCES
Adler, P. S. and Kwon, S. (2002). Social Capital: Prospects for a New Concept. Academy of Management Review. Vol. 27, pp. 17-40.
Baker, G. Jensen, M. and Murphy, K. (1988). Compensation and Incentives: Practice vs. Theory. The Journal of Finance. Vol. 43, No. 3, pp. 593-616.
Barnea, A. and Guedj, I. (2006). But, Mom, all the other kids have one! CEO Compensation and Director Networks. McCombs School of Business Working Paper, University of Texas at Austin.
Bebchuk, L. and Fried, J. (2004). Pay Without Performance: The Unfulfilled Promise of Executive Compensation. Cambridge, MA: Harvard University Press.
Beiser, V. (2004). Rise of the Corporate Plutocrats. Los Angeles Times. Available online:
http://articles.latimes.com/2004/oct/17/magazine/tm-execclass42/5
Belliveau, M. O’Reilly, C. and Wade, J. (1996). Social capital at the top: Effects of social similarity and status on CEO compensation. Academy of Management Journal. Vol. 39, pp. 1568-1593.
Bryman, A. and Bell, E. (2013). Business Research Methods. Stockholm: Liber.
Burt, R. (2000). Structural Holes versus Network Closure as Social Capital. Social Capital: Theory and Research.
Chidambaran, N. K. Kedia, S. and Prabhala, N. R. (2012). CEO-Director Connections and Corporate Fraud. Not just whether you are connected but how.
Collins, M. and Hymowitz, C. (2015). CEO Pension Benefits: Bigger Than the Pay
Advantage? Bloomberg Businessweek. Available online:
https://www.bloomberg.com/news/articles/2015-01-08/ceo-pension-benefits-bigger-than-pay-advantage Davis, G. (1991). Agents without Principles? The Spread of the Poison Pill through the Intercorporate Network. Administrative Science Quarterly. Vol. 36, No. 4, pp. 583-613. Duchin, R. and Sosyura, D. (2013). Divisional Managers and Internal Capital Markets. The Journal of Finance. Vol. 68, No. 2, pp. 387-429.
Dvorak, P. (2009). Poor Year Doesn’t Stop CEO Bonuses. The Wall Street Journal. Available online: https://www.wsj.com/articles/SB123698866439126029
Engelberg, J. Sasseville, C. and Williams, J. (2012). Market Madness? The Case of Mad Money. Management Science. Vol. 58, No, 2, pp. 351-364.
Fernandes, N. Ferreira, M. Matos, P. and Murphy, K. (2012). Institutional Monitoring and REIT CEO Compensation. The Journal of Real Estate Finance and Economics. Vol. 40, No. 4, pp. 446-479.
Fich, E. and White, L. (2003). CEO Compensation and Turnover: The Effects of Mutually Interlocked Boards. Wake Forest Law Review. Vol. 38, pp. 935-959.
Field, A. P. (2009). Discovering statistics using SPSS: and sex and drugs and rock ‘n’ roll. London: Sage.
Geletkanycz, M. A. Boyd, B.K. and Finkelstein, S. (2001). The strategic value of CEO external directorate networks: Implications for CEO compensation. Strategic Management Journal. Vol. 22, pp. 889-898.
Gomez-Mejia, L. R. Larranza-Kintana, M. and Makri, M. (1987). The Determinants of Executive Compensation in Family-Owned Firms.
González, V. M. González, F. (2012). Performance, valuation and capital ctructure: survey of family firms. Corporate Governance. Vol. 12, No. 2, pp. 4745-4754.
Granovetter, M. (1973). The Strength of Weak Ties. American Journal of Sociology. Vol. 78, pp. 1360-1380.
Granovetter, M. (1985). Economic Action and Social Structure: The Problem of Embeddedness. American Journal of Sociology. Vol. 91, pp. 481-510.
Hallock, K. (1997). Reciprocally Interlocking Boards of Directors and Executive Compensation. Journal of Financal and Quantitative Analysis. Vol. 32, pp. 331-344.
Henderson, A. and Fredrickson, J. (1996). Information-Processing Demands as a Determinant of CEO Compensation. The Academy of Management Journal. Vol. 39, No. 3, pp. 575-606. Holzer, H. (1987). Hiring Procedures in the Firm: Their Economic Determinants and Outcomes. The National Bureau of Economic Research.
Horton, J. Millo, Y. and Serafeim, G. (2012). Resources or Power? Implications of Social Networks on Compensation and Firm Performance. Journal of Business Finance & Accounting. Vol. 39, No. 3-4.
Jensen, M. and Murphy, K. (1990). Performance Pay and Top-Management Incentives. Journal of Political Economy, Vol. 99, No 2.
Lazear, E. (1989). Pay Equality and Industrial Politics. Journal of Political Economy. Vol. 97, No. 3, pp. 561-80.
Lilling, M. (2006). The Link Between CEO Compensation and Firm Performance: Does Simultaneity Matter? Atlantic Economic Journal. Vol. 34, No. 1, pp. 101-114.
Miller, D. (1995). CEO Salary Increases May Be Rational after All: Referents and Contracts in CEO Pay. Academy of Management Journal. Vol. 38, No. 3.
Mishel, L. and Davis, A. (2015). CEO Pay Has Grown 90 Times Faster than Typical Worker Pay Since 1978. Economic Policy Institute. Available online:
http://www.epi.org/publication/ceo-pay-has-grown-90-times-faster-than-typical-worker-pay-since-1978/ Murphy, K. (1985). Corporate Performance and Managerial Remuneration: An Empirical Analysis. Journal of Accounting and Economics. Vol. 7, pp. 11-42.
Murphy, K. (1999). Executive Compensation. Handbook of Labor Economics. Vol. 3, pp. 2485-2563.
Şule Erçetin, Ş. and Banerjee, S. (2014). Chaos, Complexity and Leadership 2014.
Switzer, L. N. (2007). Corporate Governance, Sarbanes-Oxley, and small-cap firm performance. The quarterly review of Economics and Finance. Vol. 47, pp. 651-666.
S&P Dow Jones Indices. (2017). S&P 500. Available online:
http://us.spindices.com/indices/equity/sp-500
S&P Dow Jones Indices. (2017). S&P SmallCap 600. Available online:
http://us.spindices.com/indices/equity/sp-600
Tosi, H. Werner, S. Katz, P. and Gomez-Meija, L. (2000). How Much Does Performance Matter? A Meta-Analysis of CEO Pay Studies. Journal of Management. Vol. 26, No. 2, pp. 301-339. Useem, M. (1984). The Inner Circle: Large Corporations and the Rise of Business Political Activity. New York: Oxford University Press.
Voulgaris, F. Asterious, D. and Agiomirgianakis, G. (2007). Size and Determinants of Capital Structure in the Greek Manufacturing Sector. International Review of Applied Economics. Vol. 18, No. 2.
White, H. and Lu, X. (2010). Robustness Checks and Robustness Tests in Applied Economics. University of California, San Diego. Available online:
http://www.economics.uci.edu/files/docs/micro/s11/white.pdf
