The impact of social connections on promotions

THE IMPACT OF SOCIAL CONNECTIONS ON

PROMOTIONS

Author: Periklis Koutroumpis

Supervisor: Tomislav Ladika

iii

Contents

Acknowledgments………iv

Abstract………vi

CHAPTER 1: Introduction………...1

Chapter 2: Literature review……….……3

Chapter 3: Methodology ……….8

Chapter 4: Data and descriptive statistics……….11

Chapter 5: Results………..15

Chapter 6: Robustness and extensions………21

Chapter 7: Discussion………..25

Chapter 8: Conclusion……….27

iv

Acknowledgments

I would like to thank everyone who contributed to the successful completion of this thesis. I would first like to thank my supervisor, Tomislav Ladika, who was always available to offer me his knowledge and gave me the opportunity to deeply understand this work. Then, I would like to thank all the professors of the University of Amsterdam who guided me to the very interesting object of Finance.

At this point I would like to express my gratitude to some people outside the narrow academic environment, who greatly benefited me, giving the required balance. I would like to express my deepest appreciation to my family, which gave me the opportunity to be part of this master and supported me throughout my difficult moments. I owe my deepest gratitude to Mirto who has been greatly tolerant and supportive during the whole master and especially during the writing of this paper. Advices and comments given by Dila, Vasilis and Eirini have been a great help for the accomplishment of my thesis.

vi

Abstract

The aim of this study is to examine the relationship between social connections and promotions that a divisional manager receives, in a sample of firms listed in S&P 500 index. Multiple regression analysis is applied to identify the causal effect of social ties on promotions. The statistical analysis shows that social connections are positively correlated with promotions. Specifically, the more social connections a divisional manager has, the more promotions she receives. The findings of the research are discussed in the framework of Internal labor market and Social networks as well.

Key words: promotions, social connections, professional ties, educational ties, other

1

CHAPTER 1

Introduction

Social ties and relationships exist between people and affect their lives (McPherson et al., 2001). Through years and after numerous studies it is proved that social connections have a significant influence on both business productivity and profitability of firms. However, there is little consensus on the effect of social connections in firms. In literature, there are two major arguments which seek to demonstrate this effect. The first one is called the “dark side” and it suggests that social ties lead to favoritism and undermine the firm value. Duchin and Sosyura (2013) claim that divisional managers with more social ties to CEOs receive more capital allocation than those poorly connected. According to Chidambaran et al. (2012) non-professional connections between CEOs and directors can significantly elevate fraud. These ties reflect social interactions that can exert influence on board and also compromise monitoring. Those relationships, as a result, lead to a weaker corporate governance and discourage outside investors from financing the firm (La Porta et al., 2000).

On the other hand, there is the “bright side” which stresses that social connections can have a beneficial effect on firms. Montgomery et al. (1991) suggest that firms recruiting through employee referrals earn higher profits and explain why workers who are well connected, with those in high-paying jobs, fare better than those who are poorly connected. Moreover, Engelberg et al. (2012) find that having personal connections with the lender leads to a substantial lower interest rate, fewer covenants and larger loan amounts.

The aim of this thesis is to investigate the relationship between social connections and promotions. Specifically, we are interested in finding the causal effect of social ties on the number of the promotions that a divisional manager receives. Therefore, our research question states: What is the influence of social ties that a divisional

2

examines the effect of social ties in firms, there is no specific focus on the number of promotion that divisional managers receive depending on their social ties. The majority of the related studies focus on the internal capital allocation and not on the internal labor market, thus, with the findings of our research we attempt to bring a novel look on the issue of social ties in firms. This research topic is interesting for numerous reasons. Firstly, our findings give us a better understanding of internal labor market. Nowadays, more and more companies allocate their workers internally with the objective of maximizing their value. The results of this study reveal if they succeed their goals with their current allocation method. Furthermore, our results shed light on the determinants of promotion. We observe if social connections, an informal measurement of managerial skills, have an impact on promotions. Finally, humans as an integral part of a company affect the value of a firm. This research gives us an insight on how human networks are formed and affect the firm value.

We use cross sectional data to test four hypotheses which are related to social ties and promotions. Our data mainly consist of biographical data and managers’ individual networks. The combination of these elements requires an OLS regression and the analysis gives answers to several questions related to personal networks.

The rest of the paper is organized as follows. Chapter 2 presents the literature review. Chapter 3 describes the identification strategy, the theoretical model and the hypotheses which are tested. Chapter 4 describes the data, its sources and its descriptive statistics. Chapter 5 discusses the main results of the model and Chapter 6 discusses the additional results which are part of the robustness checks. Chapter 7 discusses the main findings of this study and Chapter 8 provides the conclusions of the research.

3

CHAPTER 2

Literature review

Social ties can be beneficial for firms in multiple ways. Human networks which play an important role in the flow of information are created from the social connections between people. Cross and Parker(2004) suggest that information does not flow through a human network in the same way as it does through electronic devices such as internet routers. People add interpretation, context and meaning to the information they receive and transmit. Therefore, they conclude that human networks can improve information flow, exchange creative ideas and promote cooperation. Engelberg et al. (2012) support this idea in their study which investigates the relationships between the employees at a firm and their lenders. They find that personal relationships lead to more favorable financing terms suggesting that social connections provide either better information flow or better monitoring. In fact, they point out that the interest rates are markedly reduced and the lending amount is on average or somewhat more when banks and firms establish interpersonal linkages through their employees’ previous shared job settings. It is also worthwhile to mention that after borrowing, the credits rating of personally connected firms improve and the stocks returns are higher compared to the unconnected counterpart borrowers.

The beneficial effects of social connections can also be observed in both external and internal labor markets. Holzer (1987) investigates the hiring procedures that firms follow in the external labor markets and provides evidence that these procedures do matter from the economic perspective. The results show that people who are recruited hired through referrals from current employees are more productive and the firms have lower staff turnover. In particular, newly hired employees who are referred by the current employees show better performance. This kind of procedures also seems to be the least costly hiring methods to the firm in terms of direct costs and foregone earnings. These results are in agreement with those of Montgomery (1991). His paper examines the external labor markets and the impact

4

of social networks on recruitment methods. He reinforces the previous argument by showing that referrals also serve as a useful screening device apart from being less expensive than other formal methods. He reaches the same conclusion that workers who are well connected fare better than poorly connected.

In internal labor markets, firms redeploy workers internally rather than recruiting others from outside. Thus, a firm takes into account the information that derives from the existing social and working relationships with the personnel and allocates the workers properly. Tate and Yang (2012) research the internal labor markets in diversified firms and they find that internal labor markets provide better corporate governance. The labor in diversified firms is more productive because firms exploit the option to redeploy workers internally and satisfy their current needs.

Although there are numerous papers providing evidence for the beneficial effect of social connections to the firms, there is little consensus on this argument in the existing literature. Fracassi and Tate (2012) study the external networking and the internal firm governance. Specifically, they focus on the network connections between CEOs and board members and the impact that these relationships has on corporate governance. Their results suggest that network ties with the CEO weaken the intensity of the board monitoring. These ties quite often reduce the firm value as there is evidence that firms with more CEO-director ties engage in more value-destroying acquisitions. Despite the fact that the majority of boards of directors consists of qualified independent directors, they observe that the board does not always act in favor of the firm’s claimholders. Such phenomena are opposed to the concept that directors should be more willing to vote against managerial initiatives that are harmful to shareholders. Chidambaran et al. (2012) further support this argument in their studies in which they analyzed the correlation between corporate fraud and CEO-director connections. They spotlight that the connections of non-business origin such as those from common education or non-profit institutions elevate fraud probability. They explain that these ties reflect social interactions that can also exert influence on board oversight function. While these types of ties can foster trust and aid teamwork when the task is to monitor the CEO the effects can be

5

less benign. McPherson et al (2001) raise the alarm and cite that connections can lead people to make favorable judgments in situations of ambiguity. Taking for granted that people tend to create networks with people structurally similar to them, homophily can be noticed almost in every social network. Notably, ‘’homophily affects the information that people receive, the behavior they form and the interactions they experience’’ (Mcpherson et al. 2001). Our thesis will deal with the internal labor market in diversified firms and will investigate the correlation between the social ties of divisional managers to CEO and the promotions they receive.

Promotion for most of the workers is simply the official upgrades of their current position or other formal amendments (Pergamit & Venum, 1999). Although most people mean no change of position, about 85% of promoted workers experience an increase in job responsibilities. Other consequences of promotion seem to be increased wages, training receipt and increased job satisfaction. In this paper, it is also mentioned that companies use promotions in order to motivate workers especially when the supervision is difficult. Pfeifer (2010) adds that firms use promotions for efficiently allocating the employees and it is also possible to benefit from incentive effects of promotions. The tournament theory states that firms use the prospect of promotion as an incentive for workers to exert greater effort. This theory implies that workers place significant value on the promotion itself (Kosteas, 2011). Even after controlling for wages and wage increases, Kosteas (2011) demonstrates that both promotion receipt and promotion belief have strong positive correlation with job satisfaction. This satisfaction derives from the fact that people enjoy having higher rankings among others. The strong negative correlation between quits and both promotion expectation and job satisfaction is another important aspect that should be stressed. Firms can convince a worker not to quit if they can maintain the worker’s belief that a promotion is possible (Kosteas, 2011). In our study we will assume that a person gets a promotion if her job title changes. We will count, for instance, a promotion if the job title changes from “Divisional manager” to “Vice president” without taking into account if more responsibilities were added to the new position.

6

Pheifer (2010) in his paper indicates the determinants of promotion in an internal labor market. His main findings are that more overtime, higher education, less absenteeism, longer contractual working and longer tenure are correlated with a higher probability of promotion. Moreover, short-term performance seems to be more important in the promotion process than long-term performance. Those determinants explicate why women might have a disadvantage for promotion. For instance, women might have more absent days, work less overtime and have lower educational level. Cobb-Clark’s (2001) research shed light as well on the determinants of internal promotion but she narrows down her interest in gender. Her results are in accordance with the previous authors suggesting that gender plays a significant role in the promotion process. Consequently, male employees are more likely to get a promotion (e.g., Kosteas, 2011; Pergamit &Venum, 1999; Pheifer, 2010). On the other hand, Booth et al. (2003), having controlled for observed and unobserved individual heterogeneity, find that women have the same promoted rate as men. We are going to comment about the gender gap in promotion rates in the Section III albeit it is not one of our primary hypotheses.

Existing literature in internal capital markets strengthen the idea that social ties play a crucial role within firms. Better connected and more powerful divisional managers obtain larger capital allocation (Glaser et al., 2013). This piece of evidence reflects inefficient allocation of internal resources and is consistent with the dark side view on internal capital markets. In the dark side stands also the paper by Duchin and Sosyura (2013). They support the idea that connections to CEO dominate formal measurers of managers’ influence and affect both capital allocation and managerial appointments. In other words, not only divisional managers with more social ties receive more capital but they are also placed more often in capital rich divisions. Graham et al (2010) stress that after the net present value of an investment the opinion that the CEO has for a divisional manager is the most important factor for capital allocation. Stein et al. (2000) claim that social connections have a negative effect on investment efficiency and firm value due to the fact that CEOs allocate more capital to divisional managers with whom they have better connections. Even

7

though CEO does not have social ties to divisional managers, she allocates more capital to these unconnected managers in an effort to win their support (Xuan, 2009). Nevertheless, a great amount of studies strongly support the bright side of internal capital market. The pre-existing connections or the current professional ties that formed within a firm provide a superior information flow which enables the CEO to make better allocation decisions (Gertner et al. 1994; Stein, 1997).

The present research is greatly influenced by the paper of Duchin and Sosyura (2013). Its contribution is remarkable for providing me the ideas for our data resources and our methodology. Furthermore, it gave me some taste of our imminent results and the estimation of our main variables. Their results, as we already mentioned before, show that the social ties of divisional managers to CEO affect the capital allocation. That is, the better connected a manager is, the greater amount she receives from CEO. We are expecting our results to follow the same logic. Consequently, the more social ties the manager has, the more promotions he receives. In addition, the paper by Engelberg et al. (2012) gave me insight into potential problems about social connections. Interpersonal connections are endogenous and as a result one should be careful while constructing network variables. Their paper helped me realize what adjustments we should do to our data in order to eliminate reverse causation. We will describe thoroughly the measurement of our variables and the adjustments of the data in Section III.

8

CHAPTER 3

Methodology

In the present thesis, we are going to test the causal relationship between the social connections of a manager and the number of the promotions she has received. For the accomplishment of this research we use several kinds of data related to managers’ lives. In particular, we use the job history of all managers in order to measure the promotions they have received in their primary firm and the number of years they have been working there. Biographical data concerning the age and the gender of the managers is also needed for our equation. Educational background is necessary for the estimation of one of our important control variables, Ivy League, as well as the individual network of any manager for the measurement of our independent variable.

In this study are tested four hypotheses which concentrated on different social ties. H1: Managers with more social connections receive more promotions.

H2: Professional connections of a manager increase the number of her promotions. H3: Ties formed in universities help a manager to get promoted more often.

H4: Social ties from other activities lead a manager to more promotion receipts In all of hypotheses it is assumed that social connections of all types are positively correlated with promotions. In other words, if a manager has more connections than another she has more promotion receipts. We derive these expectations from the existing literature although there is not a specific literature which studies the correlation between promotions and social connections. As already mentioned, better connected managers receive more capital allocation (Glaser et al. 2013; Stein et al., 2000; Duchin & Sosyura, 2013). Additionally, managers with more social ties are placed more often in rich capital divisions ( Duchin & Sosyura , 2013).

Before proceeding to the equations that are used, we would like to give some definitions about the types of social connections and how we measure them. First of

9

all, consistent with prior literature we define three types of social networks: connections via education, connections via previous employment and connections via nonprofit organizations. Two managers are connected via previous education if they both attended the same university and graduated within 2 subsequent years (e.g., Harvard Class of 1980 or 1981).

Professional connections formed when two people overlap through a common past job (e.g. both having worked for Coca-Cola). Lastly, two managers are connected via nonprofit organizations if they share membership in the same nonprofit. These organizations typically include social clubs, religious organizations, philanthropic foundations, industry associations, and other nonprofit institutions defined in BoardEx as manager’s other activities. We measure each type of social network by multiplying the number of a manager’s connections by the average years of these relationships. For instance we want to measure the professional network and we know that a manager knows 3 people for 10, 15 and 20 years respectively. The result of the variable of the professional network will be equal to 3 times 15 ((10+15+20)/3). We also create a variable which combines all these networks and is equal to the sum of all three connections. In specific, all social connections variable for a manager is equal to the sum of Professional, Educational and Other activities connections.

Interpersonal relationships are endogenous and the recognition of this fact is important for the construction of our equation. It is crucial to eliminate the reverse causation which indicates that a social connection is formed after the promotion receipt. We create 2 control variables to confront this issue. The first one is the Workexp variable which measures the number of years a manager has been working in her primary firm. The objective of this variable is to measure directly the managerial skills assuming that the more years in a working environment a manager is, the better skills he acquires. Our second proxy for measuring indirectly the managerial skills is the Ivy variable. This proxy has been used in prior studies which document positive correlation between Ivy League’s universities and managerial skills (Dachin & Sosyura, 2013; Li, Zhang, Zhao, 2011). We include the two other

10

control variables such as the age and gender of a manager. Existing literature suggests that gender and age are determinants of a promotion and as a result they are both included in our equations (Kosteas 2011). We explain in detail the formation of these variables in the next chapter, Data and descriptive statistics. For our econometric model we use cross sectional data and we run an OLS regression. The variables which are included in the regression are the following:

Table 1

Definition of Variables

The equations which used are the following:

1. Promotionsi 0 1SocialConnectionsAlli2Agei3Genderi4Workexpi5Ivyiui

2. Promotionsi 

 









3.

Pr

omotions

 

 

SocialConnectionsEdu





Age





Gender





Work

exp





Ivy u



Pr

omotions

 

 

SocialConnectionsOther





Age





Gender





Work

exp





Ivy u



Variable Definition

Age It shows how old a manager is

Gender It is a binary variable which is equal to 1 if the manager is male and 0 otherwise

Workexp It measures the number of years that a manager has been working in her primary firm

Ivy Equals to 1 if manager has attended a university which is member of Ivy League and 0 otherwise

SocialConnectionsAll It is the sum of the SocialConnectionsProf, SocialConnectionsEdu and SocialConnectionsOther

SocialConnectionsProf It is the product of the number of professional connections and the average years of these relationships

SocialConnectionsEdu It is the product of the number of educational connections and the average years of these relationships

SocialConnectionsOther It is the product of the number of other activities connections and the average years of these relationships

11

CHAPTER 4

Data and descriptive statistics

We begin constructing our sample with all firms included in S&P 500 in 2011. The selection of the year has been decided after acknowledging that the majority of the databases take approximately 2 years to incorporate all the information in their dataset. Since we are interested in studying the managerial connections and their effect on promotion, Boardex database seems ideal for this purpose. While browsing in the BoardEx database we realized that there is no sufficient information for all firms about divisions and subsidiaries and as a result, we narrowed down our list to 303 firms listed in S&P 500. In the beginning, we downloaded all firms’ files in which enlisted the most important managers with their job titles as well as other interesting information about companies. After exploring all these files, trying to identify the most prevalent job titles, we made a list of thirteen of them. In this list are included job roles like CEO, CFO, Vice President, Regional manager and so forth. The next step was to identify those titles in each one of the 303 firms and download the personal information of each manager plus her individual network that is provided by BoardEx. We are going to start with the personal information files, giving details on how we created our relevant variables and commenting at the same time on some important summary statistics. The easiest part was to create the variables for gender and age. Concerning the former, the relevant variable is called gender and we have data for 3748 observations. The most noteworthy mentioning thing about gender is that males are the predominant gender with a mean of 0.83 which in fact raises some questions about the non-existing gender analogy. Regarding the latter variable, in personal files there are the dates of birth for each manager and thus we easily created our control variable Age.

12

Table I: Summary Statistics

This table presents the variables in terms of their mean, median, standard deviation, minimum, maximum and number of observations.

Variable Mean Median Std. Dev. Min Max Obs

Age 54,31 54 6,52 34,00 88,00 2173 Male 0,83 1 0,37 0,00 1,00 3748 Promotions 1,47 1 1,95 0,00 14,00 3361 Working experience 9,15 7 7,95 0,50 51,00 3361 Ivy league 0,08 0 0,28 0,00 1,00 1593 Professional connections ( PC) 39,25 36 32,53 0,00 253,00 3748 Years of PC 5,28 5,02 4,32 0,00 29,00 3748

Educational connections ( EC) 0,05 0 0,44 0,00 11,00 3748

Years of EC 0,34 0 2,58 0,00 34,00 3748

Other activities connections (OC) 0,28 0 2,17 0,00 47,00 3748

Years of OC 2,13 0 5,17 0,00 36,50 3748

Social network of all connections 266,38 195 304,28 0,00 2443,50 3748

Social network of professional connections 262,57 192,25 297,87 0,00 2358,50 3748

Social network of education connections 0,76 0 7,24 0,00 200,00 3748

Social network of other activities connections

3,04 0 27,11 0,00 737,00 3748

From summary statistics we can notice that the average age of managers is almost 54 and the observations are less (2173) than the previously indicated since there were a few missing values. The variable “promotions”, which is the dependent variable in all of our hypotheses, measures the number of promotions that a manager has received in her primary firm. In the present research, as we already mentioned, we define a promotion as the change of the job title without taking into consideration possible changes in wages or responsibilities. As we can see the mean is 1.47 and the observations are 3361. We should add that for the estimation of this variable we dropped all missing data and the job title “various positions” considering it not important, otherwise BoardEx would provide more details. For the estimation of the working experience we subtracted the start year of managers’ primary firm from 2014 assuming simultaneously that managers who started working in a firm in 2014 have half a year (0.5) of experience. In the summary statistic table, we detect

13

that on average a manager has about nine years of experience in her primary firm and the maximum value in our data set is 51 years. The last interesting information that we drained from the personal files is the educational history. We created the control variable Ivy which is equal to 1 if a manager attended a university that is a member of the Ivy League and 0 otherwise. Although Ivy League is a collegiate athletic conference, it is consisted of prestigious American universities ranked among the top 15 in the world. It is obvious from Table I that only 8% of the managers attended one of these universities which is against our belief that firms listed in S&P 500 hire only high level educated people. However, there were a lot of missing values and now the observations are less than half of the full sample.

The second package of data which contributed to the formation of the rest of our variables comes from the merging data of each manager’s individual network. In this set of data appeared the different kinds of social connections that a manager has, as well as the duration of them. We make it clear once again that we focus only on the current relationships-not the old- for counting the number of connections that are in some way related to the manager’s primary firm. After calculating the sum of these connections we divided them into three categories by the type of relationship. The first category is “Professional connections” which includes the relationships that are formed from previous common working experience. This type of ties entails all the people who are connected to a manager without taking into account the job title of the connected individual. It can be seen from the Table I that the professional ties dominate among the other types with a mean of 39.5 and the maximum number 253. Although the average is 39.5, we expect great variation in our sample due to the fact that the standard deviation is 32.5. The second category that represents the lowest number of connections, with a mean of 0.05, is the “Educational connections”. This certain type reflects the connections that are formed during the attendance of the same university with the maximum overlapping time of 2 years. The third and the last category is the “Other social connections” which consists of the number of ties which were formed via co-membership in a nonprofit organization. Social clubs, religious organizations, philanthropic foundations,

14

industry associations, and other nonprofit institutions are all included in a manager’s other connections.

The same pattern follows the division of the average years of each manager’s connection. After we found the exact period of each relationship, we calculated the average number of all ties and we split them again into the same 3 categories. In specific, we notice that a manager is connected with previous co-workers for about 5.28 years on average. The lowest average number can be noticed in the “Educational connections” type (2.58) while the greatest deviation, which is 5.17, belongs to the “Other social connections” category. The important thing that should be stressed is that for relationships that started in 2014 we assume that the relationship exists for half a year and it is equal to 0.5. We try to give greater validity to our research by estimating for each hypothesis the interaction term between the number of connections and the average year of their existence. This multiplication is crucial in order not to neglect valuable information for the causal effect of our independent variable. Subsequently, each category of relationships has its own social connection variable. As expected, the variable of professional connections has the highest values for both mean and standard deviation (262.57, 297.87) followed by those of other activities and educational relationships (3.04, 27.11) and (0.76, 7.24) respectively.

15

CHAPTER 5

Results

In this part we present and discuss our main results of all of our equations. The first table, Table II, describes the coefficients of the first multiple regression analysis. This table presents five steps of this regression. We begin with the independent variable and in each step we add one more control variable.

Table II

Sum of social ties

This table presents the different cross sectional data regression stages of Promotions on Social network of all connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of all connections is the independent variable and is defined as the sum of professional, educational and other activities network. From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of all connections 0.0022*** 0.0005*** 0.0007*** 0.0007*** 0.0011*** [22.7] [5.74] [4.88] [4.85] [6.3] Working experience 0.1584*** 0.1630*** 0.1629*** 0.1718*** [46.15] [29.25] [29.23] [25.42] Ivy league 0.0234 0.0236 -0.1023 [0.16] [0.16] [0.6] Gender 0.1696 0.1581 [1.47] [1.03] Age -0.0514*** [6.11] Constant 0.8348*** -0.1191*** -0.0923 -0.2332** 2.5387*** [19.87] [3.07] [1.36] [1.99] [5.5] Summary Statistics R-squared 0.133 0.4696 0.468 0.4687 0.4842 Adj R-squared 0.1328 0.4692 0.4669 0.4673 0.4818 Observations 3361 3361 1485 1485 1110

16

As we can notice, in Table II our independent variable is the Social network of all connections and the dependent is the Promotions. In column 1, the coefficient of the independent variable has its highest value which is 0.0022. However, with the addition of the control variable working experience, column 2, the coefficient falls dramatically to 0.0005. After this vast decrease, it starts to get higher gradually and in column 5 the coefficient is 0.0011 which is the final value after the addition of all our control variables. It has to be pointed out that the coefficient of the independent variable is statistically significant at 1% level and positively correlated to the promotions. This correlation is in line with our expectations, documenting that an increase of one standard deviation in social network of all connections increases by 0.33 the number of promotions. In other words, we can conclude from this table that indeed social connections of all types do matter for the promotion receipt and a manager who is well connected is promoted more often than the poorly connected. Moreover, we should stress that the Ivy League variable is not only statistically insignificant in all columns but also it changes sign from column 4 to 5 and become from positive to negative. The last noteworthy finding in Table II is that both R squared and adjusted R square are in a relatively good level, that is 0.48.

Table III provides results for the relationships that are formed in a working environment. Our expectation is that professional ties help a manager to get promoted more frequently without taking into account the formal measures of managerial skills like productivity. In column 5 the coefficient of our independent variable, social network of professional connections, confirms this expectation. The coefficient is positively correlated with promotions and statistically significant at 1% level. A rise in professional connections by one standard deviation leads to an increase in promotions by 0.32. The magnitude of this change is approximately the same with that of all social connections indicating that professional connections are the most important ties in terms of their effect on promotions. People in all over the world spend at least one third of their day in a working environment. Consequently, the ties that are formed in a workplace tend to be stronger than other types as regarding the promotions. Living every day life’s experiences and sharing the same

17

working space appears to create founder bonds which exert great influence on promotions. The Ivy League variable follows the same order changing sign from positive to negative (columns 4 to 5). It is easily observed that with the addition of this variable (columns 2 and 3) the coefficient of the independent variable changes slightly and gets from 0.0005 to 0.0007.

Table III

Professional ties

This table presents the different cross sectional data regression stages of Promotions on Social network of professional connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of professional connections is the independent variable and is defined as the product of the number of professional ties and the average years of these relationships. From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of professional

connections 0.0023*** 0.0005*** 0.0007*** 0.0007*** 0.0011*** [22.64] [5.49] [4.65] [4.62] [6.15] Working experience 0.1587*** 0.1635*** 0.1633*** 0.1721*** [46.15] [26.27] [29.25] [25.38] Ivy league 0.0238 0.0241 -0.1023 [0.16] [0.16] [0.6] Gender 0.1708 0.1605 [1.48] [1.05] Age -0.0514*** [6.1] Constant 0.832*** -0.1169 -0.0902 -0.2322** 2.535*** [19.73] [3.01] [1.33] [1.98] [5.48] Summary Statistics R-squared 0.1324 0.4691 0.4672 0.468 0.4833 Adj R-squared 0.1322 0.4688 0.4661 0.466 0.481 Observations 3361 3361 1485 1485 1110

18

A control variable that makes the model more valid seems to be the Age variable. In column 5, the spot that we add the variable, the coefficient of our interest becomes higher and changes from 0.0007 to 0.0011. The last interesting thing which we should emphasize on is the control variable gender. The insertion of this control variable, column 4, not only does not seem to affect the coefficient of the independent variable but also is statistically insignificant.

Table IV

Educational ties

This table presents the different cross sectional data regression stages of Promotions on Social network of educational connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of educational connections is the independent variable and is defined as the product of the number of educational ties and the average years of these relationships. From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of education

connections 0.0351*** 0.0156*** 0.0137*** 0.0139*** 0.0184*** [8.06] [4.83] [3.59] [3.65] [4.2] Working experience 0.1654*** 0.1731*** 0.1728*** 0.1866*** [53.24] [34.67] [34.62] [30.05] Ivy league 0.0013 0.0014 -0.1175 [0.01] [0.01] [0.69] Gender 0.1946* 0.1896 [1.68] [1.22] Age -0.0505*** [5.95] Constant 1.4413*** -0.056 -0.0035 -0.1659 2.6088 [43] [1.5] [0.05] [1.42] [5.59] Summary Statistics R-squared 0.019 0.468 0.4641 0.4651 0.474 Adj R-squared 0.0187 0.4677 0.463 0.4637 0.4716 Observations 3361 3361 1485 1485 1110

19

Now, we proceed to the Table IV where the Educational connections are tested. Ties between two people who attended the same university and graduated within two subsequent years are called Educational. In column 5 of Table IV we discern that the coefficient of Social network of education connections is positive and statistically significant at 1% level. Undoubtedly, the educational ties affect the number of promotions that a manager receives. That is, the growth of educational network by one standard deviation leads to a rise of promotion by 0.13. Clearly, this type of ties does not affect the promotions in the same way as the professional ties do but still the effect is significant. The low impact on promotions can be justified as the relationships from universities may be either not so strong or not so profound to last through time and affect promotions as much. For the first time, we observe in column 4 that the control variable gender has a coefficient significant at 10% level which puts us in position to think is whether in fact gender affects the promotions or not. Nonetheless, the picture is changed in column 5 where we see that the coefficient of gender is again statistically insignificant and in the same line with previous results. The Ivy League variable keeps on reacting in an opposite way to what it was expected and once again the coefficient in column 5 is statistically insignificant and negatively related to promotions. Also, in column 5 we detect that with the addition of the variable Age the coefficient of our interest has the highest value which is 0.184.

The last type of social connections is being tested in Table V. This type consists of relationships that are formed via other activities. Membership in nonprofit associations, philanthropic foundations and religious organization are part of the other social ties. In column 5, the coefficient of the independent variable social network of other activities is positive and statistically significant at 10% level indicating that a change by one standard deviation affects promotion by a rise of 0.06. Although it is statistically significant, we observe that in column 4 the coefficient is merely higher and significant at 5% level.

20

Table V

Ties of other activities

This table presents the different cross sectional data regression stages of Promotions on Social network of other activities connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of other activities connections is the independent variable and is defined as the product of the number of other activitiesl ties and the average years of these relationships. From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of other activities

connections 0.004*** 0.0021** 0.0026** 0.0026** 0.0023* [3.42] [2.45] [2.05] [2] [1.68] Working experience 0.1668*** 0.175*** 0.1748*** 0.1897*** [53.86] [35.27] [35.23] [30.58] Ivy league 0.0209 0.0211 -0.0981 [0.14] [0.14] [0.57] Gender 0.1743 0.158 [1.5] [1.01] Age -0.0498*** [5.83] Constant 1.4574*** -0.0626* -0.0119 -0.1573 2.8567*** [43.1] [1.67] [0.18] [1.34] [5.51] Summary Statistics R-squared 0.0035 0.4653 0.461 0.4618 0.467 Adj R-squared 0.0032 0.465 0.4599 0.4603 0.4645 Observations 3361 3361 1485 1485 1110

21

CHAPTER 6

Robustness and extensions

Our main specification evaluates the professional connections as the sum of connections that are formed in a previous common working environment. To assess the robustness of our results to alternative measures of these connections, we test two alternative specifications.

In the first alternative specification we measure the number of professional ties focusing on the job title of the connected individual. Specifically, we are interested only in connections with the CEO and other top executives. CFO, Vice President and Director are the job titles that constitute the list of top executives. From now on, these connections will be referred as important professional connections assuming that they play a more important role than all of the professional connections let alone they affect promotions per se. Now, in order to create the independent variable we multiply again the number of these connections by the average years of these ties. After imposing this filter, we can see in Table VI the results of our alternative specification which are consistent with our main findings.

Column 4 demonstrates that the coefficient of our independent variable, Social network of important professional connections, is positive but relatively significant at 10% level. Nevertheless, Column 5 shows that the coefficient of the dependent variable is statistically significant at 1% level and positively correlated with promotions. That is, if we change the independent variable by one standard deviation the promotions will be increased by 0.15. Moreover, it can be seen that the value of coefficient is now 0.0015 which is higher than the previous test that measures all the professional ties. Our interpretation of this evidence is that these connections have more discretionary power to affect the promotions that a manager receives.

22

Table VI

Important professional ties

This table presents the different cross sectional data regression stages of Promotions on Social network of important professional connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of important professional connections is the independent variable and is defined as the product of the number of important professional ties and the average years of these relationships. Important professional connections include only ties of a manager to CEOs, CFOs, Vice Presidents and Directors From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of important professional

connections 0.0056*** 0.0004 0.0007* 0.0007* 0.0015*** [18.73] [1.59 [1.79] [1.77] [2.96] Working experience 0.1648*** 0.1711*** 0.1709*** 0.1818*** [48.19] [30.59] [30.57] [26.71] Ivy league 0.024 0.0242 -0.093 [0.16] [0.16] [0.54] Gender 0.178 0.1642 [1.54] [1.06] Age -0.0508*** [5.96] Constant 0.8605*** -0.0812*** -0.0518 -0.1997 2.5493*** [18.84] [2.02] [0.74] [1.67] [5.44] Summary Statistics R-squared 0.0945 0.4647 0.4606 0.4615 0.4698 Adj R-squared 0.0942 0.4644 0.4595 0.46 0.4674 Observations 3361 3361 1485 1485 1110

The second alternative specification concerns the measurement of all social networks. In our main results we calculate the variable of all social networks as the sum of all types of connections. In this specification, we measure the variable, Social network of important connections, as the sum of important professional connections, educational ties and other social relationships. The results are presented in Table VII in which the dependent variable is promotions and the independent is social connections of important connections.

23

Table VII

Sum of important ties

This table presents the different cross sectional data regression stages of Promotions on Social network of important connections. Promotions is the dependent variable of the regression and is defined as the change of a manager’s job title. The social network of important connections is the independent variable and is defined as the sum of important professional, educational, and other activities network. Important professional connections include only ties of a manager to CEOs, CFOs, Vice Presidents and Directors From column ( 1 ) to ( 5 ) the different control variables are inserted in the regression. Robust t-statistics are reported in parentheses. *, **, *** indicate significance at 10%, 5%, and 1% level, respectively.

Dependent variable: Promotions

Regressors (1) (2) (3) (4) (5)

Social network of important

connections 0.0051*** 0.0005*** 0.0001*** 0.0009** 0.0016*** [18.65] [2.46] [2.58] [2.55] [3.59] Working experience 0.1637*** 0.1693*** 0.1692*** 0.1805*** [48.3] [30.55] [30.53] [26.78] Ivy league 0.0245 0.0248 -0.0926 [0.17] [0.17] [0.54] Gender 0.1755 0.1603 [1.52] [1.03] Age -0.051*** [5.99] Constant 0.8897*** -0.0917** -0.0658 -0.2117 2.5515*** [19.91] [2.3] [0.95] [1.78] [5.46] Summary Statistics R-squared 0.0939 0.4653 0.4618 0.4627 0.4718 Adj R-squared 0.0936 0.465 0.4608 0.4612 0.4694 Observations 3361 3361 1485 1485 1110

In column 5, it is observed that the coefficient of our variable of interest is significant at 1% level and is equal to 0.0016. In other words, if we change the social network by one standard deviation the promotions will change by 0.18. Our findings are consistent with those of our main results and are in line with existing literature and our expectations. In addition, the coefficient is relatively bigger compared to one of the previous measurement suggesting that important connections exert totally more

24

influence on promotions. Column 4 and 5 depict another interesting finding. We realize that the control variables of gender and age are important additions to our equation. In particular, after the addition of gender variable the coefficient is nine times higher than column 3 albeit it is significant at a lower level. As far Age variable, we detect that it drives the coefficient of the independent variable to be approximately double (0.0016) and statistically significant at 1% level.

25

CHAPTER 7

Discussion

The main conclusion of our main results is that social ties play a role in promotions receipts. The more social ties a manager has, the more times she is being promoted in a firm. This is also consistent with our hypotheses which are mentioned in earlier parts of this thesis. One important finding which draws our attention to is that other social ties have the least effect in promotions compared to other two categories of social connections. Mcpherson et al (2001) stress that non professional connections endow individuals with common culture and shared experience foster team spirit and trust, however; when it comes to a promotion people seem to be more reluctant to succumb to favoritism and give the promotion to the connected ones.

The idea that prevails after all of these tests is that social ties, without taking the type into consideration, have a positive impact on promotion receipts. The results are in accordance with the dark side of social connections in other fields of interest such as capital allocation which supports that social ties lead in a way to homophily. People tend to act in favor of the one who is connected even if the formal means of evaluation turn to be against this action. . Glaser et al. (2013) support the idea that managers with more social connections receive more capital allocation by CEOs regardless of the net present value of a division’s investments. Additionally, Chidambaran et al. (2012) point out that certain type of social connections are responsible for compromising the monitoring of a CEO by the board of directors and as a consequence the firm is engaged in more value-destroying acquisitions. This favoritism leads to an inefficient allocation of labor and contradicts with the philosophy which braces the idea that firms redeploy their workers internally with the objective of satisfying their current needs. Furthermore, as far the variable Ivy League is concerned we see that the findings are in the same line with existing literature and particularly that of Duchin and Sosyura (2013). They find that Ivy League variable is negatively correlated with capital allocations supporting the idea that better managerial skills do not imply the receipt of more capital allocation. Finally, we are going to comment on the effect of the gender in promotions. Our

26

results are in the same line with that of Booth et al (2003) and suggest that men are promoted at the same rate as women. Throughout our research, Gender variable fails to support a different treatment to men as does the paper by Pergamit and Venum (1999).

27

CHAPTER 8

Conclusion

This thesis examines the role of divisional managers in internal labor market. In specific, we investigate the relationship between the promotion receipts and each divisional manager’s social network. Through their lives, divisional managers create social networks which include several types of relationships. We discern three main categories of such relationships. The first one involves the professional ties that are formed during the working life of a manager. A person is able to form either strong or profound relationships with colleagues which she maintains in her entire life. The second category refers to the ties that are formed in academic years of a manager. That is, relationships that are created during her attendance in a university for either bachelor or higher diploma. The last category of the ties that a manager forms is those through other social activities. This type of relationships includes all the connections of nonprofit activities such as membership in an industrial association or a philanthropic organization. Combining both the biographical data and the social networks we are able to test this relationship. Our results indicate a causal relationship between social connections and promotion. In other words, the social ties of a manager are able to affect the number of promotions that a manager receives. This effect is positive and statistically significant for all ties. However, this effect varies across the different types of relationships. We find that professional connections exert the most influence on promotions followed by those of educational and other activities connections respectively.

In interpreting the findings of the present study, several limitations must be acknowledged. In our methodology, as mentioned above, we divide the social categories into three categories. Nevertheless, in existing literature there are papers that use other types of relationships. For instance, Duchin and Sosyura (2013) measure professional connections only by those between a divisional manager and the current CEO. Moreover, we create a variable that reflects the social network by multiplying the number of connections by the average years of their existence.

28

Related articles use different measurements for creating a social network. Tate and Fracassi (2012) measure social network by the number of connections between a CEO and a board member. Furthermore, we use control variables as proxies for managerial skills in order to overcome the burden of reverse causality. These proxies are the working experience and the attendance in a prestigious university, member of the Ivy League. It is clear that other studies use different proxies for managerial skills. For instance, Duchin and Sosyura (2013) use SAT scores of a manager for measuring her skills. Concerning our methodology, we test our hypotheses with an OLS regression using cross sectional data. Engelberg et al. (2012) examine their hypotheses with panel data with a sample covering a period of ten years. Another important limitation that should be emphasized is the measurement of promotions. In our methodology, a promotion is defined only as the change in the job title. Kosteas (2011) uses a different measure which indicates that the definition of promotion may include either an increase in the wage or a change in the job title.

In conclusion, the present thesis provides some evidence on the social connections and their effects on promotions. However, different measures for both social networks and promotions can be used for analyzing this causal effect. Control variables in which the managerial skills are represented in a more efficient way can also be a part of future studies. Internal labor market is quite a new study field and as a result there is plenty of room for further research.

29

References

Booth, A.L., Francesconi, M., Frank, J. (2003). A Sticky Floors Model of Promotion, Pay, and Gender. European Economic Review , vol 47, no2, pp. 295–322.

Chidambaran, N.K., Kedia, S., Prabhala, N.R (2012). CEO-Director Connections and Corporate Fraud. Working paper. Fordham University, Rutgers Business School, University of Maryland.

Cobb-Clark, D.A. (2001). Getting Ahead: The Determinants of and Payoffs to Internal Promotion for Young U.S. Men and Women. Research in Labor Economics, vol 20, pp. 339–372.

Cross, R.L., Parker, A. (2004). The Hidden Power of Social Networks: Understanding how Work Really Gets Done in Organizations. Harvard Business Press.

Duchin, R., Sosyura, D. (2013). Divisional Managers and Internal Capital Markets. The

Journal of Finance, vol 68, no2, pp. 387–429.

Engelberg, J., Gao, P., Parsons, C.A. (2012). Friends with Money. Journal of Financial

Economics, vol 103, no1, pp. 169–188.

Fracassi, C., Tate, G. (2012). External Networking and Internal Firm Governance. The

Journal of Finance, vol 67, no1, pp. 153–194.

Gertner, R.H., Scharfstein, D.S., Stein, J.C. (1994). Internal Versus External Capital Markets. Quartely Journal of Economics, vol 109, pp.1211-1230.

Glaser, M., Lopez-De-Silanes, F., Sautner, Z. (2013). Opening the Black Box: Internal Capital Markets and Managerial Power. The Journal of Finance, vol 68, pp. 1577– 1631.

Graham, J., Harvey, R., Puri, M. (2010). Capital Allocation and Delegation of Decision-Making Authority within Firms. Working paper. Duke University.

Holzer, H.J. (1987). Hiring Procedures in the Firm: Their Economic Determinants and Outcomes. Working paper. National Bureau of Economic Research.

Kosteas, V.D. (2011). Job Satisfaction and Promotions. Industrial Relations: A Journal of

Economy and Society, vol 50, pp. 174–194.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R. (2000). Investor Protection and Corporate Governance. Journal of Financial Economics, vol 58, pp. 3–27.

30

Li, H., Zhang, X., Zhao, R. (2011). Investing in Talents: Managers Characteristics and Hedge Fund Performances. Journal of Financial and Quantitative Analysis, vol 46, pp. 59-82. McPherson, M., Smith-Lovin, L., Cook, J.M. (2001). Birds of a Feather: Homophily in Social

Networks. Annual Review of Sociology, vol 27, no1, pp. 415–444.

Montgomery, J.D. (1991). Social Networks and Labor-Market Outcomes: Toward an Economic Analysis. American Economic Review, vol 81, pp. 1407–18.

Pergamit, M.R., Veum, J.R. (1998). What Is a Promotion?. Industrial and Labor Relations

Review, vol 52, no4, pp. 581-601.

Pfeifer, C. (2010). Determinants of Promotions in an Internal Labor Market. Schmalenbach

Business Review (SBR), vol 62, no4, pp.342–358.

Scharfstein, D.S., Stein, J.C. (2000). The Dark Side of Internal Capital Markets: Divisional Rent-Seeking and Inefficient Investment. The Journal of Finance, vol 55, pp.2537– 2564.

Stein, J.C. (1997). Internal Capital Markets and the Competition for Corporate Resources.

The Journal of Finance, vol 52, pp.111–133.

Tate, G.A., Yang, L. (2012). The Bright Side of Corporate Diversification: Evidence from Internal Labor Markets. Working paper, Social Science Research Network, Rochester, NY.

Xuan, Y. (2009). Empire-Building or Bridge-Building? Evidence from New CEOs’ Internal Capital Allocation Decisions. Review of Financial Studies, vol 22, pp.4919-4948.