Do social connections with CEOs help a director's promotion?

(1)

Do social connections with CEOs help a director's promotion?

Zhichuan Jia (10897097) Supervisor: Dr. Ilko Naaborg

(2)

2

Abstract

By using the BoardEx database with the time span from 2000 to 2012, I study the social connections between CEOs and directors, and whether these connections have effectively promoted the directors to better positions. Directors are connected with the CEOs through three types of channels: shared alumni network, previous employment overlaps and shared social service. The results are significant only when directors have social connections with the CEOs via previous employment overlaps. No evidence indicates that the other two types of social connections have impacts on director's promotion significantly. The results are robust when drop other variables, such as gender and remaining tenures.

JEL Classification: G3, M12, M51

(3)

3

Statement of Originality

This document is written by Student Zhichuan Jia who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(4)

4

1. Introduction

Promotion is the primary mean for a company to reward the employees, it is a recognition for an employee's previous work performance. Meanwhile, promotion enables corporations to preserve talents and keep them committed (Ruderman & Ohlott, 1993). Even though promotion is important for both individual employees and corporations, only a few research has been done in this field in terms of how and why promotion is done. Pfeifer (2010) collected data from three large German companies and analyze the determents which attribute to promotions. Among several hypothesis, Pfeifer suggests a number of individual attributes which play important roles in employee's promotion, such as gender and education. Yet, no studies have been devoted to investigate whether interpersonal relation has an impact on the probability of promotion, in particular, the personal connection of the employee with top executives in the management team of a company. This paper intends to fill this gap in the existing literature, and aims to answer the central research question: Does social connections with CEO helps with director's promotion?

Abundant evidence has shown that having a social connection with a CEO is beneficial to the directors in a number of ways. For example, Hilger, et al. (2013) find that senior executives who have strong interpersonal ties with CEO are more likely to have an upward career move, and their careers will not be affected even the connected CEO is dismissed or fired. Some other literature have been devoted to the CEO-Manager relationship in terms of internal capital market. Duchin & Sosyura (2013) find that divisional managers with social connections to the CEO receive more capital in their divisions. Gaspar & Massa (2011) report that divisional managers with social connections to the CEO receive more investments and are less sensitive to cash flow shortfall. Glaser, et al (2013) find that when firms have unexpected cash windfalls,

(6)

6

more powerful (connected) managers obtain more capital allocations in their own business units than their less connected peers. Besides internal capital markets, CEO-Manager relations have been used for different other research fields. For example, Chidambaran, et al. (2010) use the social connections between CEO and managers to explain the incidence of corporate fraud, they argue that social connections may increase or decrease the incidence of corporate fraud depending on how the social connections are established (through which channel).

Having viewed the literature which suggest social connections with the CEO in most cases have a positive impact on the managers, it is probably true also that the social connections between CEO and manager will have a positive influence on manager's promotion perspective. A model is built in order to test the research question. In the model, I collect data of the directors in the US who have experienced promotions between the time span of 2000 and 2012. The time length for the promotion to be taken place is used as dependent variable in the model. Meanwhile, I record the number of social connections the director has with CEOs via alumni network, previous employment overlaps and shared social work respectively, and use them as the three main explanatory variables. The empirical results suggest that social connections can have a positive impact on directors promotion only when the director has shared working experience in one company before. The other two channels do not have significant impacts on the promotions for the directors. The results are robust when other independent variables are dropped, such as gender and remaining tenure.

My study contributes to the existing literature on the corporate governance, social networking and internal labor market. Although there is a substantial amount of studies that have been conducted related to corporate governance, many of them focused on the area of CEO-board relationships and the principal-agent problem or

(7)

7

the so-called “control vs ownership” (Rose, et al., 2014). This study improves the understanding on the corporate internal social network and the CEO’s execution decision. Additionally, this study should bring the interest to board committee and decision makers within the firms, because management promotion is in line with firm's long-term strategy, every profit maximizing firm will take position movements into consideration when planning the future of the firm. Furthermore, to promote a connected director could have potential impacts on firm's value, whether positive or negative, it is interesting and worth investigating for future research. The rest of the paper will be organized as follows. Section 2 discusses the relevant literature of this field. Section 3 describes the data and methodology. Section 4 presents the empirical results and robustness check. Finally, section 5 discusses limitations, concludes and makes some suggestions for further research.

2. Literature review

Many researchers who have studied the subject of CEO-director connections point out that CEO-director connections have a positive impact on director's career. Hilger, et al. (2013) use matched-pair design to analyze data on 77 CFO-turnover events in large, stock-marked-quoted companies in Germany between 1999 and 2006. By using a logit regression model, they find that a strong interpersonal connection to the CEO helps a top management team member to stay in office, when the CEO leaves the office, it is likely that top management team member will experience an internal promotion within that company or leave the current company for a better position elsewhere. However, if the connected CEO is dismissed, the strong connection with that CEO increases the top management team member's risk of dismissal.

(8)

8

Besides, many related studies have been conducted in the area of internal capital markets in recent years. Gaspar & Massa (2011) studies the role of commonalities between the CEO and divisional managers on the resource allocation process within the firm. By using a large sample of multisegment U.S. corporations from 1996 to 2004, the results show that segments run by connected managers receive more investment and exhibit lower sensitivity. They also discuss the intuitions why this would happen potentially, the connections make it more likely that CEO and the divisional managers have similar beliefs, also the connections are associated with the ability of the division managers to lobby the CEO for resource allocations. Glaser, et al. (2013) find similar results, additionally, they suggest that when firms have unexpected cash windfalls, more powerful (connected) divisional managers obtain more capital allocations than their less connected peers. Two types of data are used in their study, the first one is a five years of quarterly data on planned and actual capital allocations for each of the firm's 20 business unit. The second one is a formal and informal measures of managerial power and connections for the 40 business-unit CEOs.

Duchin & Sosyura (2013) use hand-collected data on divisional managers at S&P 500 firms to study the impact of interactions between CEO and managers on capital budgeting. The results show that divisional managers who have social connections with CEOs receive more capital in terms of internal capital budgeting. Differently than the previous two studies, Duchin & Sosyura associate the CEO-manager relation with its impact on the firm value and investment efficiency. Interestingly, they find that the results will decrease the firm value and investment efficiency if the firm has a weak corporate governance, and have positive impact on firms with high information asymmetry.

(9)

9

Existing literature has not yet covered the CEO-manager relation on the perspective of how this relation would affect manager's promotion. This paper intends to contribute to the existing literature and fill the gap. Chidambaran, et al. (2010) use data on CEO-director connectedness from the "BoardEx" database compiled by Management Diagnostics Limited to study the impact of CEO-director connections on corporate fraud. They classify that CEO and director have professional connections and nonprofessional connections. CEO and director can have professional connections when they worked together in the same company before, CEO and director can have nonprofessional connections through shared alumni network and shared social service work such as charity. The way they classify the CEO-director connections is in line with my research, except that I further separate the two type of connections in terms of nonprofessional connections, and seek for their impact on director's promotion individually. By using a logit model, Chidambaran, et al. argue that social connections between CEO and director have ambiguous impacts on the incidence of corporate fraud depending on how they are connected. Nonprofessional connections (such as shared non-business service and alumni network), increase the incidence of fraud. While professional connections through the employment overlaps decrease the probability of fraud. Interestingly, the major finding in my research shows that professional connections between CEO and director have a positive impact on the director's promotion. While nonprofessional connections do not affect the director's promotion process. Associated with motivations developed by Chidambaran, et al, I speculate why professional connections and nonprofessional connections could have different effect. When the CEO and the director have past working experience, the director has the opportunity to observe the CEO's work settings, norms and actions. Because of this, the director is likely to develop professional skills that help him/her

(10)

10

to build mutual trust with the CEO, thus increasing the likelihood of the director's promotion opportunity. In terms of the nonprofessional connections via the same alumni network and shared non-business services, they are less likely to endow the type of business skills which could help the directors do a better job.

It is also worth mentioning literature on internal corporate promotions. Pfeifer (2010) collect data from three large German companies and analyze the determents in employee's promotion process. He derive a number of hypotheses in his paper regarding to the determinants of promotion, such as education and remaining tenure. By using a pooled sample and random effects probit models, they find that less absenteeism, more overtime, longer contractual working time, higher education, higher entry age and longer remaining tenure are correlated with a higher promotion probability, but female employees are less likely to get promoted. Besides that, he also argues that short-term performances are more important than long-term performances in the promotion process. Some of the target determinants Pfeifer include also appear in the model I constructed, namely gender and remaining tenure year. Surprisingly, the results in my model are quite contrary to his findings. For example, I find that remaining tenure year has a negative effect on director's promotion process, also gender does not play a significant role in director's promotion process.

3. Methodology & Data

3.1 Methodology

An Ordinary Least Square model is used to test the impact of CEO-director social connections on director's promotion process. To do so, a sample of CEO pool is created which consists of the directors who serve as Chief Executive Officer in a

(11)

11

company from 2000 to 2012. Having social connections with more CEOs in the CEO pool is presumed to accelerate directors' promotion process. Therefore, variables are created to count the number of CEOs who went to the same school, work and participated in the same social work with each of the non-CEO director respectively.

Take social connection via education as an example, the variables which

represent the educational institutions are denoted as InstitutionName1,

InstitutionName2, and InstitutionName3 (the first three are taken into consideration).

Corresponds to the institution name variables, three continuous variables are generated (denoted as countedu1, countedu2 and countedu3) to count the number of CEOs who has the same InstitutionName1, InstitutionName2, InstitutionName3 with each of the non-CEO directors. Then another new variable is generated which is the summation of the three continuous variables as follows:

countedutot = i

Similar to the previous process, the other two dependent variables for previous employment overlaps and other activities are as follows:

countotheremploymenttot = i

countotheractivitiestot = i

3.2 Hypotheses

Three mutually exclusive hypotheses are considered to explain the central question in this study:

Hypothesis1: CEO-director connections help with a director's promotion.

Hypothesis2: CEO-director connections do not have an impact on a director's promotion.

(12)

12

Hypothesis3: CEO-director connections have a negative impact on a director's promotion.

The first hypothesis posits that CEO-director connections have a positive impact on a director's promotion. This hypothesis predicts that social connections forester mutual trust between the director and the CEO, in particular, a social connection through previous employment overlaps helps the director to understand the CEO's work settings, which endows the director skills to do a better job. The second hypothesis posits that CEO-director connections have no impact on a director's promotion. This hypothesis suggests that the commonalities and social connections between the CEO and the director play little role in directors' promotion process. The third and the last hypothesis posits that CEO-director connections have a negative impact on a director's promotion. This hypothesis is supported by the "bridge building" hypothesis mentioned by Duchin & Sosyura (2013). "bridge building" hypothesis predicts that a CEO allocated more capital resources to unconnected directors in order to win their support.

3.3 Data

I collect the relevant information of the directors for the study from BoardEx database. BoardEx database covers the directors' education background, previous employments and non-business service through the worksheets of "education", "otheremployment" and "otheractivity". BoardEx database also encodes basic information of the directors such as gender, nationality, time to retirement, and number of qualifications under the worksheet of "BoardEmployment". The identification for director individually is denoted as DirectorID, and the identification for firm basis is denoted as CompanyID. These 4 worksheets are downloaded and imported separately into Stata. Before I

(13)

13

discuss the settings for director's promotion, it is necessary to first introduce the basic characteristics of these 4 datasets and the merge of them by using DirectorID.

3.3.1 Sample

BoardEx is a large database, with 533,371 observations in terms of board employments and 72,729 DirectorID. The sample is based on the observations from year 2000 till year 2012, because the BoardEx database coverage is incomplete and limited before the year of 2000. Furthermore, only directors with US nationality are taken into consideration. This has to do with two reasons. First, the US is the major nationality among the population of the database, therefore choosing US directors as the sample will enhance its integrity and yield a more precise research regression. Second, considering that different countries endow different corporate cultures, the internal governance mechanism varies from one to another. For instance, the criteria for personnel promotion in a Chinese corporate will not be the same within a US corporate. Hence, keeping director's nationality constant can partially eliminate some of the endogeneity issues. Lastly, I drop observations whose employment type is "NED", "NED" is short for Non-executive Director. Another employment type opposed to "NED" is "ED" which is short for Executive Director. The reason for dropping the Non-executive directors is that they are also called the outside directors who participate in the board of directors of a company, however, they are not employees and they are not involved in the management or execution matters. Therefore, this group of directors is of the interest of the study.

After sorting out the sample from the raw database based on the time span, the nationality of the directors and the role type of the directors, the sample is settled down to 6,267 directors. In the later sections, I will discuss the roles of the directors

(14)

14

and their promotions, I will also discuss the key variables and other variables, and there will be a summary of statistics presented.

3.3.2 Board Employment

The "BoardEmployment" dataset consists of various biographical information of the directors as well as their roles within one or different firm(s). Each observation represents one board employment for one director, however, the majority of the directors hold multiple board positions in more than one firm, which suggests that one director might match with a number of observations. In order to have one director match with one observation exclusively, the data is reshaped in Stata. After the reshape, every observation corresponds to one director, and contains a overview information of this director's board employments. The positions of employments are denoted as Role1, Role2, Role3... Each role has a start date (denoted as StartDate1,

StartDate2, StartDate3...), end date (denoted as EndDate1, EndDate2, EndDate3...) and company's identification (denoted as CompanyID1, CompanyID2, CompanyID3...).

3.3.2.1 Position

There are hundreds of kinds of position titles for director's role in the company, it is time-consuming and unrealistic to rank all these positions from top to bottom. For convenience, all the positions are clustered into three big categories. From the top to bottom, they are "CEO", "Chief officers" and "others" (denoted as 3, 2, 1 respectively). Any titles with CEO included are regarded as CEO, for example, in BoardEx, one director's role in the company might be written as Chairman/CEO or

(15)

15

CIO and COO. "others" are directors whose positions are neither CEOs nor Chief

officers, for example: sales director.

Over the course of career, directors may experience promotions or demotions, because of this, the roles of the directors are not consistently the same over the time. It is somewhat difficult to determine a director's position due to this inconsistency, therefore, I keep only the first two roles for each director on the data sheet. There are two reasons for this, firstly, approximately 80% of the directors who have two roles over the time. Secondly, with two roles for each director, it is possible to determine every director's position and monitor their promotion situation. To illustrate, I create a new variable called position = Role1 *Role2, because the possible values for Role1 and Role2 are both ranging from 1 to 3. Therefore, the multiplication of Role1 and

Role2 can maximum generate 6 results: 1, 2, 3, 4, 6 and 9. Each of the 6 numbers

represents the director's position status. (see table1)

When position=9, it means the director has both two roles as CEO(3*3=9), which makes this director a CEO position. When position=6, it means the director either go from Chief position(2) to CEO (3) as a promotion, or go down from CEO(3) to Chief position(2) as a demotion. When position=4, it means the director is at a chief position (2*2=4). When position=3, it means the director goes from others(1) to CEO(3) as a promotion or CEO(3) to others(1) as a demotion. When position=2, it means the director goes from others(1) to Chief position(2) as a promotion or Chief position(2) to others(1) as a demotion. When position=1, it means the director belongs to other position(1*1=1).

Total 6,267 100.00 9 2,138 34.12 100.00 6 583 9.30 65.88 4 521 8.31 56.58 3 1,305 20.82 48.27 2 394 6.29 27.45 1 1,326 21.16 21.16 position Freq. Percent Cum.

Table 1: Distribution of positions

(16)

16

As a result, there are 2,138 CEOs (position=9) and 4,129 non-CEO directors. In the non-CEO directors, 521 are Chief officers (position=4) and 1,326 are neither CEOs nor Chief officers (position=1). The rest of the directors in the non-CEO directors have experienced either a promotion or demotion over the time span of Role1 to Role2, with a number of 2,282 (the sum of position=2, 3, 6).

3.3.2.2 Promotion

There are 3 possibilities of promotions. First, when a director goes from position "others" to a Chief position (1-2). Second, when a director goes from position "others" to a CEO position (1-3). Finally, when a director goes from a Chief position to a CEO position (2-3) is another promotion. A new variable is generated to measure the unit of promotions for directors: promotion=Role2-Role1, "promotion" has a value of 1 when the first and third situation occur and has a value of 2 when the second situation happens. Furthermore, another new variable is generated and denoted as DOP which is short for "days of promotion", DOP= StartDate2-StartDate1. Obviously, variable "DOP" is meant to measure the time length which the director takes to get promoted. Note, however, that both "DOP" and "promotion" can be negative, which poses a problem that the start date and role are not consistent. To solve this issue, a variable called "depvar" = DOP/promotion is created which will be also used in the regression analysis as dependent variable. Therefore, when "depvar" is a positive number, it means DOP and promotion has consistent sign (+/-) and the start date and role corresponds. A positive number of "depvar" for a director means the director experienced a promotion, and it measures the days it takes for him/her to reach one unit of promotion. If the denominator promotion equals 2, the days will be cut in half, which suggest that it takes half time for him/her get one unit of promotion. If the denominator promotion equals 1, then "DOP" is equivalent to "depvar", and in this

(17)

17

case "DOP" measures the speed of promotion by per unit of promotion. When "depvar" has a "-" sign, the start date of the role and the role does not correspond, which means the director undergo a demotion. The smaller the "depvar" is the faster the director gets demoted. If a director does not undergo any promotions nor demotions, this will yield a value of zero for the denominator "promotion", thus the "depvar" will be not defined and generate a missing value in Stata, I replace the missing values into number zero.

3.3.3 Social connections

A non-CEO director has social connections with a CEO via three channels: alumni network, previous employment and non-profit organization.

3.3.3.1 Social connection via alumni network

Manager and CEO have social connection via alumni network when they went to the same education institution. The top 3 most popular education institutions among the sample are Harvard Business School, Stanford University and Wharton School, University of Pennsylvania. The "education" worksheet in the BoardEx database records extensively the institution's name, type of qualification and date of qualification for each director. Each row in the "education" worksheet records one education qualification for one of the directors exclusively. However, 36.62% of the directors have more than one education qualification, which means one director could appear multiple times in the worksheet. In order to have the directors' name occur once only and to have each entry of the worksheet representing only one director, the long data is reshaped into a wide data with rows displaying different directors and have columns displaying different education qualifications for the directors. Furthermore, maximum three education qualifications are taken considerations for

(18)

18

each director, because only a negligible 2% of the directors has more than three education qualifications.

Table 2 presents the percentages of directors who have 1, 2, and 3 education qualification(s) in the sample. In the sample, there are 4,129 non-CEO directors and 2,138 CEOs. Besides that, the percentages of directors who have shared the same alumni network are given along with the percentage of non-CEO directors who have shared alumni network with the CEOs.

Non-CEO CEO Total No. of observations 4,129 2,138 6,267 with one education qualification 63.67% 63.70% 63.38% with two education qualifications 28.27% 28.96% 28.50% with three education qualifications 8.07% 7.34% 7.82%

Total 100% 100% 100%

same alumni network 39.67% 36.88% 44.65% CEO alumni network size 83.39% N/A N/A

Table 2 shows the distribution of non-CEO directors and CEOs in terms of the number of education qualifications they obtained in the past years, and their previous education overlaps. Approximately two thirds of the directors have obtained one education qualification, little under 30% of the directors have two education qualifications, only around 8% of the directors who have obtained three education qualifications which is substantially smaller. I also calculate the percentages of directors who share the same alumni network. In terms of the non-CEO directors, they went to 983 different education institutions as a whole, in which 593 education institutions have one and only one director in their alumni network. Therefore, 390 institutions (983-593=390) have two or more than two directors in their alumni network, which means 39.67% (390/983) of the non-CEO directors at least share the same alumni network with another non-CEO director. The same calculation is applied

Table 2: Directors' social connections via education

(19)

19

to CEOs and whole sample, and the percentages of same alumni network for CEOs and non-CEOs and CEOs combined are 36.88% and 44.65% respectively. Additionally, I find that the ratio among the non-CEO directors who share the same alumni network with the CEOs is considerably large: 83.39%, this is due to the fact that the CEO pool is relatively large compared to the whole sample.

3.3.3.2 Social connection via previous employment

Manager and CEO have social connections via previous employment when they worked in the same company in the past. The worksheet "otheremployment" in BoardEx dataset records the previous employments for each director, one row corresponds to one job entry. For convenience, I reshape the data from long to wide. In that way, every column in the data is equivalent to a simple "CV" for a director. Maximum of four previous employments are taken considerations for each director, because only a few directors have more than five previous employments. The identification of previous employment is CompanyID, two directors have employment overlaps when the same CompanyID occurs both in their "CV".

Table 3 shows the distribution of non-CEO directors and CEOs in terms of the number of previous employments as well as non-CEO directors' previous employment overlaps with CEO. Unlike the fact that every director has at least obtained one education qualification, a small fraction (2.06%) of the directors has no previous job experiences. The percentages in terms of one, two and three previous employments are still small with approximately 5%, 8%, 10% respectively. Most of the directors have four previous employments, the ratios are 73.24% and 76.38% for non-CEO directors and CEOs respectively. The reason for this is that the average age of the directors is over 50, the senior directors naturally have more job experiences

(20)

20

comparatively. For non-CEO directors, the ratio of previous employment overlaps with CEO is 41.12%.

Table 3 presents the percentages of directors who have 0, 1, 2, 3 and 4 previous employment(s) in the sample. The sample consists of 6,267 directors in total, with 4,129 non-CEO directors and 2,138 CEOs. Besides that, the last row in the table presents the percentage of the non-CEO directors who have previous employment overlaps with the CEOs.

Non-CEO CEO Total No. of observations 4,129 2,138 6,267 with zero previous employments 2.06% 2.06% 2.06% with one previous employments 5.91% 4.91% 5.57% with two previous employments 7.87% 7.67% 7.8% with three previous employments 10.92% 8.98% 10.26% with four previous employments 73.24% 76.38% 74.31% Total 100% 100% 100% overlaps with CEO 41.12% N/A N/A

3.3.3.3 Social connection via other activity

The third and the last social tie between non-CEO directors and CEO is via non-profit organizations. The BoardEx database also records the non-profit organizations that the directors had been involved with under the "other activity" worksheet. Non-profit organizations are social services that directors engage besides their work, for example: charity. Similar to previous section, I reshape the data and keep only four non-profit organizations. The identification of non-profit organization is denoted as

OrganizationID.

Table 3: Directors' social connections via previous employment

(21)

21

Table 4 presents the percentages of directors who have participated in 0, 1, 2, 3 and 4 non-profit organization(s) in the sample. The sample consists of 6,267 directors in total, with 4,129 non-CEO directors and 2,138 CEOs. Addition to that, the last row in the table presents the percentage of the non-CEO directors who have participated in the same non-profit organizations with the non-CEOs.

Non-CEO CEO Total No. of observations 4,129 2,138 6,267 with zero NPO 68.71% 61.55% 66.27% with one NPO 13.80% 16.65% 14.77% with two NPO 5.62% 7.58% 6.29% with three NPO 2.96% 4.12% 3.35% with four NPO 8.91% 10.10% 9.32% Total 100% 100% 100% same NPO with CEO 24.05% N/A N/A

Table 4 shows the distribution of non-CEO directors and CEOs in terms of the number of non-profit organizations they have been engaged with as well as the percentage of non-CEO directors who have been in the same non-profit organization with the CEOs. The participation rate for non-profit organization is relatively lower compared to "education" and "previous employment". Approximately two thirds of the directors have no other activities being recorded in the database. Around 14% percent of the CEO directors and 17% of the CEOs have participated in one non-profit organization. The percentages of directors who have participated in two or more than two non-profit organization are almost all below 10% except that 10.10% of the CEOs have been involved in four non-profit organization. Moreover, 24.05 % of the non-CEO directors have been participated in the same non-profit organization with the CEOs, this figure is much lower as compared to its counterparts in terms of the alumni network and previous employment.

Table 4: Directors' social connections via other activities

(22)

22

3.3.4 Other independent variables

Besides the social connections via alumni network, previous employments and non-profit organizations. Three other independent variables are considered that may affect the time of promotion for non-CEO directors. They are gender, number of qualifications, and time to retirement.

Pfeifer (2010) find a number of determinants that affect employee's probability of promotion. Unfortunately, it is not possible to find all the determinants mentioned in the paper from BoardEx database, nevertheless, the characteristics that are accessible in the database and also mentioned by Pfeifer are gender and time to retirement. According to his hypothesis, female employees are less likely to be promoted. Various reasons can be explained why female employees have a lower probability of getting promoted, one of the main reason is that female employees have more days on absent on average. Furthermore, Pfeifer hypothesize that employees with more remaining tenure will be more likely to be promoted, because companies are more willing to invest in young human capitals. The last variable which might have an impact on the speed of promotion for directors is the number of qualifications.

3.3.5 Summary statistics

Table 5 shows the descriptive summary of the dependent variable and independent variables. Variable "depvar" is the dependent variable, as mentioned in section 3.3.2.2, measures the days of promotion per unit of promotion. On average, it takes 31.28 days for a non-CEO director to get one unit of promotion, which is approximately 1 month. Variable "male" is a gender dummy variable, it has a value of 1 when the director is male and 0 otherwise, the majority of the non-CEO directors are males (94%). Variable "NumberOfQualifications" indicates the number of

(23)

23

*Note: "depvar" is short term for dependent variable, which is the division of days of promotion by units of promotion (YOP/Promotion). "male" is the gender dummy variable with 1 being male and 0 being female. "NumberOfQualifications" is the number of qualifications a non-CEO director obtained. "TimeToRetirement" is the remaining tenure. "countedutot" is the total number of social connections a non-CEO director has with the CEOs via education (alumni network). "countotheremploymenttot" is the total number of social connections a non-CEO director has with the CEOs via previous employment. "countotheractivitiestot" is the total number of social connections a non-CEO director has with the CEOs via non-profit organizations.

variable Obs Mean Std. Dev depvar 4,129 31.28 787.52 male 4,129 0.94 0.23 NumberOfQualifications 4,129 2.25 0.97 TimeToRetirement 4,129 13.22 9.31 countedutot 4,129 36.23 50.02 countotheremploymenttot 4,129 1.70 3.75 countotheractivitiestot 4,129 2.42 9.11

educations qualifications that the directors obtained over the past years, the more qualifications means more schooling, and more schooling is more likely to increase the probability of getting promoted for directors, the average number of qualifications directors obtained is 2.25. Variable "TimeToRetirement" measures the remaining tenures for the directors, as discussed "in section 3.4", directors with longer remaining tenures will be more likely to receive promotions because firms are more willing to invest in young employees, the average time to retirement is 13.22 years for the directors. Variable "countedutot" is the total number of social connections a non-CEO director has with CEOs via education, on average, every non-CEO director is connected with 36 CEOs in the CEO pool via previous education. Variable "countotheremploymenttot" measures the total number of social connections a non-CEO director has with non-CEO via previous employment. On average, every director has

Table 5: summary statistics

(24)

24

employment overlaps with 1.7 CEOs from the CEO pool. Variable "countotheractivitiestot" is the total number of social connections a non-CEO director has with CEO via non-profit organization. Every non-CEO director, on average, participated in the same non-profit organization with 2.4 CEOs from the CEO pool.

4. Empirical results

4.1 Regression analysis

*Note: "depvar" is short term for dependent variable, which is the division of days of promotion by units of promotion (YOP/Promotion). "male" is the gender dummy variable with 1 being male and 0 being female. "NumberOfQualifications" is the number of qualifications a non-CEO director obtained. "TimeToRetirement" is the remaining tenure. "countedutot" is the total number of social connections a non-CEO director has with the CEOs via education (alumni network). "countotheremploymenttot" is the total number of social connections a non-CEO director has with the CEOs via previous employment. "countotheractivitiestot" is the total number of social connections a non-CEO director has with the CEOs via non-profit organizations.

* p<0.05, ** p<0.01, *** p<0.001 Standard errors in parentheses

R-sq 0.009 0.011 0.008 N 1270 1270 1270 (94.68) (95.13) (95.09) _cons 657.1*** 665.8*** 657.4*** (2.551) countotheractivitiestot -0.621 (4.854) countotheremploymenttot -11.23* (0.406) countedutot 0.408 (2.506) (2.507) (2.513) TimeToRetirement 6.515** 6.406* 6.576** (18.57) (17.94) (17.98) NumberOfQualifications -29.78 -23.22 -25.18 (75.51) (75.72) (75.36) male 53.11 65.43 58.25 depvar depvar depvar (1) (2) (3)

Table 6: Regression results

(25)

25

Table 6 presents the regression results, the variable "depvar" is the dependent variable which measures the speed of promotion for the non-CEO directors. The smaller the "depvar" is the faster the non-CEO director is promoted. The first column of the table lists a number of independent variables in the regression equation. The three regression equations are as follows:

1. depvar = β0 + β1*male +β2*NumberOfQualifications + β3*TimeToRetirement

+β4* countedutot + µ

2. depvar = β0 + β1*male +β2*NumberOfQualifications + β3* TimeToRetirement

+β4* countotheremploymenttot + µ

3. depvar = β0 + β1*male +β2* NumberOfQualifications + β3*

TimeToRetirement +β4* countotheractivitiestot + µ

For "male", the coefficient β1 is positive in all three regression specifications.

A positive sign of β1 indicates that female directors get promoted faster than male

directors, however the result is insignificant. The gender of the directors does not explain the speed of their promotion. For "NumberOfQualifications", the coefficients β2 is negative, which indicates that the more education qualifications that the director

obtained the faster he/she will get promoted. This result is within my expectation, however, it is again not significant. The Number of qualification does not explain the speed of promotion either. When it comes to "TimeToRetirement", it is positively related to the speed of promotion, with the significance level at 1% , 5% and 1% for the three regressions respectively. The interpretation is that for every one more year to retirement, it delays 6.5 days of promotion, which means the young directors will have slower promotions compared to the elder ones. This is contradicted to what was expected. On the other hand, someone may make a argument that senior and sophisticated directors are more experienced and more likely to have a shorter period

(26)

26

of promotion. While young directors need more time to learn, thus their promotion is relatively prolonged.

I separate the three channels of social connections, and put them into each of the regression analysis respectively. As a result, only the connection via previous employment shows up with a significant impact on the dependent variable. Keep other variables constant, one more CEO in the network of the director in terms of employment overlaps will reduce the days of promotion per unit of promotion by 6 days. The other two types of connections do not have any impacts on the speed of director's promotion. Why would the previous employment overlaps have an impact on the promotion while the other two types do not? Chidambaran, et al., (2010) study the correlation between CEO-director connections and corporate fraud, and in their paper, they use the same type of social connections as I do: alumni network, previous employment and non-profit organization. They define connections via alumni network and non-profit organization as Non-professional social connections, while they define the social connection via employment overlaps as professional network. Because when a CEO and a director have employment overlaps, they must work together in the past, and this will give the director precious experience in terms of acknowledging the work attitudes and work ethics of the CEO. In other words, the director will be familiarized with the CEO's personality, and this gives the director advantages over other directors who have the same conditions except the work experience with the CEO. While the social connections via same education institutions and same social work are non-professional ties which are less likely to give the advantages when it comes to work.

(27)

27

4.2 Robustness check

The robustness of the social connections is tested by dropping the non-social connection variables in the initial regressions, namely gender, number of qualifications and time to retirement. The results show that the impact of social connection through previous employment overlaps on the speed of director's promotion is robust.

In table 7, specification (1) to (3) exhibit the regression from the original model. In specification (4) to (6), the dummy variable "male" is dropped from the original model. As a result, the coefficients in specification (4) to (6) do not deviate significantly from their counterparts in specification (1) to (3), in particular, the significance level of effect of variable "TimeToRetirement" and social connection via previous employment overlaps on the dependent variable is unchanged.

Furthermore, the variable "NumberOfQualification" is dropped together with the variable "male" in Table 8. By comparing the specification (1) to (3) and specification (4) to (6), the significance level and the magnitude of coefficients are more or less the same, except that the significance level in terms of the impact that variable "NumberOfQualification" has on the dependent variable is improved from 10% to 5% in specification (5). This means that with variable "NumberOfQualification" and "male" being dropped, the number of qualifications that a director obtains has a greater impact on the speed of his/her promotion process as far as the social connection in the regression is via previous employment overlaps is concerned.

In table 9, all the three other variables are excluded which are included in the original model. The specifications (1) to (3) are remained as the original three models. In specification (5), the coefficient of variable "countotheremployments" is decreased

(28)

28

from -11.23 to -12.24. The economic interpretation is that when exclude the other three variables, every one more CEO a director is connected via previous work will reduce his/her days of promotion per unit of promotion by one day approximately. This is considered to be a minor change, since the significance level remains to be 10%. Additionally, the coefficient of "countotheractivities" in specification 6 is decreased from -0.621 to -1.132, however, the significance level remains unchanged, and the results are robust.

Table 7: Robustness Check: Variable "male" dropped

Specification (1) to (3) are the regression equations from the original model, specification (4) to (6) are the regressions which excludes variable "male" compared to the original model. Note: *,**,*** represent significance level 0.05, 0.01, 0.001.

Table 8: Robustness Check: Variable "male" and "NumberOfQualifications" dropped

Specification (1) to (3) are the regression equations from the original model, specification (4) to (6) are the regressions which excludes variable "male" and "NumberOfQualificationscompared to the original model. Note: *,**,*** represent significance level 0.05, 0.01, 0.001.

R-sq 0.009 0.011 0.008 0.008 0.010 0.008 N 1270 1270 1270 1270 1270 1270 (94.68) (95.13) (95.09) (60.15) (61.29) (60.43) _cons 657.1*** 665.8*** 657.4*** 707.6*** 727.8*** 712.9*** (2.551) (2.553) countotheractivitiestot -0.621 -0.617 (4.854) (4.841) countotheremploymenttot -11.23* -11.05* (0.406) (0.405) countedutot 0.408 0.424 (2.506) (2.507) (2.513) (2.501) (2.503) (2.508) TimeToRetirement 6.515** 6.406* 6.576** 6.457** 6.343* 6.516** (18.57) (17.94) (17.98) (18.54) (17.91) (17.95) NumberOfQualifications -29.78 -23.22 -25.18 -29.84 -23.12 -25.05 (75.51) (75.72) (75.36) male 53.11 65.43 58.25 depvar depvar depvar depvar depvar depvar (1) (2) (3) (4) (5) (6)

(29)

29

Table 9: Robustness Check: Variable "male", "NumberOfQualifications and "TimeToRetirement" dropped

Specification (1) to (3) are the regression equations from the original model, specification (4) to (6) are the regressions which excludes variable "male", "NumberOfQualifications" and "TimeToRetirement" compared to the original model. Note: *,**,*** represent significance level 0.05, 0.01, 0.001.

R-sq 0.009 0.011 0.008 0.007 0.009 0.006 N 1270 1270 1270 1270 1270 1270 (94.68) (95.13) (95.09) (41.52) (41.72) (40.47) _cons 657.1*** 665.8*** 657.4*** 639.0*** 671.7*** 651.6*** (2.551) (2.551) countotheractivitiestot -0.621 -0.722 (4.854) (4.833) countotheremploymenttot -11.23* -11.42* (0.406) (0.393) countedutot 0.408 0.288 (2.506) (2.507) (2.513) (2.481) (2.490) (2.495) TimeToRetirement 6.515** 6.406* 6.576** 6.907** 6.676** 6.880** (18.57) (17.94) (17.98) NumberOfQualifications -29.78 -23.22 -25.18 (75.51) (75.72) (75.36) male 53.11 65.43 58.25 depvar depvar depvar depvar depvar depvar (1) (2) (3) (4) (5) (6)

R-sq 0.009 0.011 0.008 0.000 0.004 0.000 N 1270 1270 1270 1270 1270 1270 (94.68) (95.13) (95.09) (25.52) (22.95) (21.09) _cons 657.1*** 665.8*** 657.4*** 736.3*** 767.5*** 750.0*** (2.551) (2.587) countotheractivitiestot -0.621 -1.132 (4.854) (4.868) countotheremploymenttot -11.23* -12.24* (0.406) (0.393) countedutot 0.408 0.301 (2.506) (2.507) (2.513) TimeToRetirement 6.515** 6.406* 6.576** (18.57) (17.94) (17.98) NumberOfQualifications -29.78 -23.22 -25.18 (75.51) (75.72) (75.36) male 53.11 65.43 58.25 depvar depvar depvar depvar depvar depvar (1) (2) (3) (4) (5) (6) Source: BoardEx

(30)

30

5. Limitations, Future research and Conclusion

5.1 Limitations

One of the limitations for my study has to do with the assumption of social connections. Even if one CEO and one director are somehow connected through one of the three connection channels, it is possible that they might as well do not know each other at all. For example, even two directors have shared with one alumni network in the database, in reality they may have no knowledge towards one another. Therefore, it is important for me to make a crucial assumption that suppose two directors have social connections via any one of the three connection channels, they have certain knowledge about each other.

Additionally, there are small frictions in the data. The BoardEx company updates the BoardEx database in a regular time basis, because of this, some of the information have the problem of inconsistency. For example, one director is graduated from Harvard Business School, and another director is graduated from Harvard University, they actually share with the same alumni network, but because of the different expressions, they are not connected in the regression analysis. The reasons why I did not clean these frictions manually are first of all, it is extremely time consuming and not practical to check thousands of thousands cells in the database one by one. Secondly, I do not have the perfect knowledge to fix every frictions. These limitations may deviate the actual results from the true or accurate results, the results are more or less expected nevertheless.

(31)

31

5.2 Suggestions for future research

The results of this paper have shown that social connection with the CEO via previous employment overlaps will accelerate director's promotion process. As for future research on this topic, prospective researchers may want to take a deep look into on how this relationship could affect firm's value and other various elements. This may be interesting to decision markets as well as shareholders in a firm.

5.3 Conclusion

To conclude, I use the BoardEx database with the time span from 2000-2012 to study the relationship between non-CEO directors' promotion time and their social connections to the CEOs. I analyzed the three types of social connections: same alumni network, previous employment overlaps and same non-profit organization, and the impacts that these connections have on their speed of the promotions. I find that among the three social connections, the social connections via previous employment is the only channel through which the director have benefits to his/her promotion. Besides that, I find senior directors are more likely to increase their speed of promotion as opposed to junior directors.

(32)

32

Bibliography

Berger, P. G. & Ofek, E., 1995. Diversification's Effect on Firm Value. Journal of Financial Economics, 37(1), pp. 39-65.

Chidambaran, N., Kedia, S. & Prabhala, N. R., 2010. CEO-Director Connections and Corporate Fraud. Fordham University Schools of Business Research Paper, Issue 2010-009.

Denis, D. J., Denis, D. K. & Sarin, A., 1997. Agency Problems, Equity Ownership, and Corporate Diversification. The Journal of Finance, 52(1), pp. 135-160.

Duchin, R. & Sosyura, D., 2013. Divisional Managers and Internal Capital. The Journal of Finance, pp. 397-428.

Gaspar, J.-M. & Massa, M., 2011. The Role of Commonality between CEO and Divisional Managers in Internal Capital Markets. Journal of Financial and Quantitative Analysis , 46(3), pp. 841-869.

Glaser, M., Lopez de Silanes, F. & Sautner, Z., 2013. Internal Capital Markets and Managerial Power. The Journal of Finance , pp. 1577-1631.

Hilger, S., Richter, A. & Schaffer, U., 2013. Hanging Together, Together Hung? Career Implications of Interpersonal Ties Between CEOs and Top Managers. BuR-Business Research , 6(1), pp. 8-32.

Mills , D., 1985. Seniority versus Ability in Promotion Decisions. Industrial and Labor Relations Review, 38(3), pp. 421-425.

Pfeifer, C., 2010. Determinants of Promotions in an Internal Labour Market. Schmalenbach Business Review, Volume 62, pp. 342-358.

Rajan, R., Servaes, H. & Zingales, L., 2000. The Cost of Diversity: The Diversification Discount and Inefficient Investment. The Journal of Finance, 55(1), pp. 35-80.

(33)

33

Rose, J. M., Rose, A. M., Norman, C. S. & Mazza, C. R., 2014. Will Disclosure of Friendship Ties Between Directors and CEOs Yield Perverse Effects?. The Accounting Review , 89(4), pp. 1545-1563.

Ruderman, M. N. & Ohlott, P. J., 1993. The Realities of Management Promotion. s.l.:Center for Creative Leadership.

Scharfstein, D. S. & Stein, J. C., 2000. The Dark Side of Internal Capital Markets: Divisional Rent-Seeking and Inefficient Investment. The Journal of Finance, 55(6), pp. 2537-2564. Schoar, A., 2002. Effects of Corporate Diversification on Productivity. The Journal of Finance, 57(6), pp. 2379-2403.

Wulf, J., 2009. Influence and Inefficiency in the Internal Capital Market. Journal of Economic Behavior & Organization , 72(1), pp. 305-321.

Do social connections with CEOs help a director's promotion?