Student: Simon Dettenkofer Supervisor: Dr. Torsten Jochem
University of Amsterdam: Faculty of Economics and Business Business Economics: Finance
Master Thesis:
Checking Buddies - Do new CEOs favor their society after getting
appointed? A review of executive hiring and its consequences
over the period of 2002-2013
Statement of Originality
This document is written by Student Simon Dettenkofer 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.
Content
I. Rationale and aim of the research: ... 5
II. Literature review: ... 6
III. Thesis objectives and suggested approach: ... 9
a. Hypotheses ... 9
b. Data collection ... 10
c. Methodology ... 11
CEO ED Matrix and SNI built up ... 11
Control Variables ... 13 Testing procedures ... 15 IV. Results ... 16 V. Conclusion ... 21 Bibliography: ... 22 Appendix ... 25
Glossary and Assumptions ... 25
Figure 1:Histogram of sample SNI ... 16
Figure 2: U test insider ... 17
Figure 3: U test Gender ... 18
Figure 4: Regression Results Return on Equity ... 19
Figure 5: Robustness Test group 0-2 ... 20
I. Rationale and aim of the research:
Since John Chen took over as chief executive officer (CEO) at the struggling smartphone maker BlackBerry, he has hired four old colleagues as executives. “The experience that the majority of the new leadership team has in working together previously will drive change within the organization at a faster pace” (Chen 2014). The newly appointed CEO is surrounding himself with people he can trust. Naturally this course of action might be very common in such detrimental situations. After a CEO turnover, the new leader tries to get control over his executives, or simply reason that it is easier to start with a team he already is acquainted with. Therefore he is placing executives in order to reach his aim, the preservation or recovery of the firm’s performance. The question arises if, as in John Chen’s case, newly appointed CEOs generally put their buddies in charge. This move would probably strengthen their position, especially in a new company (where presumably one has no allies), and maybe even with a very strong board e.g. led by the CEO’s predecessor. Blackberry did not really recover so far. Is this a coincident, was the company already too far behind the curve or did the jobs for the boys tip the company over the edge? Favoritism is probably not the explanation for Blackberry’s years of struggle.
There are a lot of kinds of actions taking place in Corporate Governance. Previous research on CEOs’ networks has primary focused on the consequences for effectiveness and performance of companies they are strong within. We do know from other work that ties bear tremendous power on Corporate Governance decisions and financial outcomes.
The main focus of this thesis’ research contains the following questions: After a CEO turnover, do new CEOs preferably advance or hire executives they have professional, educational or social ties to? Is there a difference in these actions when a newly appointed CEO is from among inside or outside the firm? This thesis attempts to better answer the question if, regarding their personnel decisions, CEO’s networks are important and what might be the consequential, positive or negative, impact for their companies.
II. Literature review:
The appointment of a new CEO often results out of turmoil and is then mostly followed by an immediate management shakeup. Pan and Wang (2012) state that top management shakeup, particularly after CEO turnover (as answer to a crisis), is a crucial element of a firm’s effective error correction. Using a sample of 2,221 CEO turnover events from 1992 to 2007, they find that these turnovers are pervasively followed by significant corrective actions and these actions positively affect firm value. Given these positive results, little is known on how executive positions are actually filled in these situations and what kind of an impact a CEO’s network has on final decisions. My research could help to explain these outcomes.
This thesis builds on to the still growing literature examining professional, educational and social ties as well as their importance within Corporate Governance. The existence of shared networks provides an advantage because of superior information within the same (Cohen, Frazzini & Malloy 2010, Engelberg, Gao & Parsons 2012). Cohen et al. (2010) study the impact of social networks on analysts’ ability to gather superior information about the firms they cover. They establish that if analysts have educational links to their covered companies, their stock recommendations outperform those of analysts that are not connected to the firms observed by them. A key finding in Engelberg et al. (2012) is that personal relationships between executives or directors at firms and their lending banks lead to more favorable financing terms. These results show that networks can bias outcomes. Other advantages of high profile professionals, originated in being part of a coterie, can be that Alumni networks are helpful to find jobs; some buddies even provide each other with more inside investment opportunities or better future predictions on price movements (Cai, Walkling & Yang 2012, Cohen, Frazzini & Malloy 2012).
Managerial decision making is affected by social networks (Fracassi 2012, Shue 2013). As information is common through these webs, Shue (1013) points out the fact that executive social interactions can affect firm policies and managerial decision making. To explore this, she looks at a historical random assignment of MBA graduates from Harvard Business
School. Using information on 30,860 key executives and directors of 2,059 companies, Fracassi (2012) assembles a set of social, educational and professional network matrices. He provides evidence that managers rely on their networks when making Corporate Finance policy decisions. These findings on networks and their influence on Corporate Finance decisions even more support my aim to find evidence on how much networks effect the appointments of executives.
Following these conclusions, there is literature on how ties (professional, educational and social) influence corporate outcomes. Hwang and Kim (2009) state that directors on corporate boards are classified as independent if they neither have financial nor familiar ties to their firm’s CEO (or to their firm). They add the variable of social ties to control the board’s independence. Using hand-collected data on the Fortune 100 firms they find that including their new measure, the independence of the boards decreases from 87% (without the measure social ties) to 62%. The independence even decreases when new CEOs have influence on selecting directors. Their results show that boards with stronger ties to the CEO perform inferior in monitoring and disciplining the same. Furthermore, it looks as if these findings might have a negative impact on firm’s value. Hence, like Subrahmanyam (2008) finds by analyzing a dataset obtained from the IRRC database (now RiskMetrics database) and Execucomp, networks can interfere with the quality of Corporate Governance and influence executive board’s decision making. He witnesses that this phenomenon lowers firm value. Nguyen (2011) examines the impact of social ties on the effectiveness of boards of directors and finds out, that if the CEO and a number of directors belong to the same network, the same is less likely to be dismissed for poor performance. This result is discovered by looking at a sample of the largest publicly-traded firms in France. Landier, Sauvagnat, Sraer, and Thesmar (2012) look at the independence of top executives to their firm’s CEO and show that a lower fraction of independent executives is associated with a significantly lower level of a firm’s profitability. As strong CEOs can appoint their peers as directors, Fracassi and Tate (2012) use panel data on S&P 1500 companies to identify external network connections between directors and CEOs. They suggest that network ties between directors and the CEO undermine the effectiveness of internal governance and even destroy shareholder’s value. These
numerous negative outcomes point out the fact that it is important to look at CEOs’ networks and their influence on personnel decisions in the executive level. Further literature shows that CEOs’ compensation is higher if they hold strong ties to the board of directors in their company (Butler & Gurun 2012, Engelberg, Gao & Parsons 2013, Hwang and Kim 2009). My research will fit into this literature, as it explores further where CEO networks can also cause privileging.
Duchin and Sosyura (2013), as well as Glaser, Lopez-de-Silanes and Sautner (2013) provide that capital allocation within companies vary with the connectedness of CEOs and their divisional managers. Duchin and Sosyura (2013) use a hand-collected dataset of divisional managers at S&P 500 firms and study the effect of ties between divisional managers and CEOs to obtain their results. On the contrary, Glaser et al. (2013) prove their findings by analyzing the internal capital markets of a multinational conglomerate. These results clearly provide evidence that CEOs favor their society and underline the relevance to test the hypotheses of my thesis. Berger, Kick, Koetter and Schaeck (2013) check if an appointee for the executive board is an out- or insider (respectively worked or did not work at the company before). They use an extensive dataset of all executive appointments in the German banking industry within 15 years. Their results show that people from outside the company have better chances to get the position, if there were stronger ties between them and the board. Hence, Berger et al. (2013) prove that networks matter in the German banking industry. As they are studying the ties an executive has to the board of directors, my research could add to this by showing the influence of ties to the CEO in particular. I will also look at different industries by exploring appointments at the executive level of large American corporations.
The existing literature has proven interconnectedness between networks and CEOs’ actions. This thesis therefore contributes by empirically document on the following question. Is there is a connection between top management appointments, CEOs’ professional, educational and social networks and their firms performance? As ties among CEOs and executives, appointed by the same, have not been checked sufficiently for
significance in literature, it will try to add to research on networks in Corporate Governance decisions and to reduce that gap.
III. Thesis objectives and suggested approach:
a. Hypotheses
This thesis aims to give evidence on how newly appointed CEOs take their network, respectively persons they have stronger ties with, into account, when it comes to hire or advance someone as an executive in their firm. It will highlight whether companies with strong connected directors perform better or worse than non-connected ones. As the first problem looks at kind of the same events, these central research questions can be manifested into the three specific hypotheses stated below.
Hypothesis 1a: CEOs prefer to hire or advance executives they have stronger ties
with.
After getting appointed, a new CEO’s personnel decisions are biased, because he hires professionals he has ties to as executives. Further the newly appointed CEO will advance persons he has ties to over people he has less to none ties to if it comes to promote them as executives.
To conclude the research questions, the following hypothesis has to be tested.
Hypothesis 1b: Outcomes differ if an executives is appointed from among inside or
outside the company.
CEOs already know the executives they will work with, when hiring them from inside the company.
Hypothesis 1c: Gender doesn’t matter as it comes to social ties. It is irrelevant whether the hired executive is male or female.
The actual hires of the CEO ties will be tested for the intensity of their ties:
Hypothesis 2: Firms with a strong connected management perform better than
weak connected ones.
As the management team can better gauge the performance of each other. That lead to more efficient and effective outcomes.
b. Data collection
To obtain the dates of CEO turnover events and the ties between divisional managers and CEOs, data on boards and managers at S&P 500 firms will be used. The core of the data set will contain biographical information on the directors of each company (which belonged to the S&P 500 at some point in the research’s horizon) and comes from the BoardEx database. BoardEx contains biographic data on executive and non-executive directors including their affiliation at private, public, and non-profit organizations. The sample contains information on firms between 2002 and 2013. Since BoardEx coverage in earlier years is very limited, it starts in 2002. To complete missing and double check the BoardEx data, the Thesis uses following databases: ExecuComp, CapitalIQ and Bloomberg Executive Profiles, which are also derived from Standard & Poors databases.
Any company fundamentals like balance sheet, income statement or cash flow items are obtained from the Compustat database, which is provided by WRDS business research. This data is later used to establish the return on assets (ROA) and return on equity (ROE) growth rates of each company. According to the literature, like Baker and Wurgler (2002, p. 4) and Duchin and Sosyura (2013, p. 394), financial firms (SIC codes 6000-6999) and utilities (4900-4949) are excluded for the performance measurement.
Next to information about the composition and character of boards of directors, the BoardEx database contains details on the professional as well as personal development of its observed directors and parts them into five different sub-categories. These allow this thesis to define the ties between managers and include Board- and Other Employment, as well as Activities, Achievements and Education. The information is obtained from director’s CVs and is therefore as complete, as disclosed. If there is e.g. no birth date available, it can’t be found on other databases.1 The first two categories contain most data, as they
show all companies the directors worked for. Board Employment features positions as executive and non-executive directors, henceforth ED and NED, of worldwide public and private companies. A NED is a board member who is not employed by the company, hence
a so called supervisory director (SD). An ED is a full time employed individual, which is also part of the executive management team.2 Other Employment lists senior managers (SM) of
worldwide companies. It is obviously possible that one person can be on these tables multiple times. If e.g. the director worked some years as a SM in different companies and then changed into the board level, as an ED and probably NED for other companies as well.
Actvities itemizes church, sport club, cultural, nonprofit organization, foundation and
similar memberships. Achievements and Education are self-explanatory, they feature honors like top 100 influential people in media or county citizen of the year and graduations from elementary schools to PhD together with CFA titles on the educational career. Out of this bulk of information, this thesis forms the SNI and tries to evaluate its impact on firm’s success.
The SNI includes all prior employment connection between a new appointed CEO and a particular executive, whom he/she advances or hires. It is important to define precisely, as e.g. graduating from the same university does automatically imply a personal connection. In about 50 percent of the cases, a missing date would eliminate the information. The available dates for Achievements and Education are the actual dates of qualification; while the other three categories feature start- and end dates. As most of the end dates are missing, it is complicated to find the right measure. This thesis consequently chose to use assumptions based on the former research of Duchin and Sosyura (2013) as the main lack that averts exact overlaps for director attributes of the BoardEx database are missing dates.
c. Methodology
CEO ED Matrix and SNI built up
The original BoardEx tables had to be cleaned out of incomplete data, without thinning the data out and jeopardize the results. The database is the most thorough for events from 2002 on, which was then defined as start off for the observations. All spreadsheets were controlled for start dates, as end dates are rarely available. To identify the CEO
appointment, this thesis deployed the “BoardEmployment” table. The sample was reduced to only American public companies, only ED callings were looked at, since NEDs are indeed members of the board, albeit not executives of the firm. To point out the relevant data all different CEO titles where identified and then replaced with “CEO”3. Only the first CEO
appointment date was used, as there can be multi mentions. All personnel changes (at the executive level) after a CEO’s appointment were taken into account. As all CEOs where defined, the residual hiring was therefore EDs only. These were obviously cleared out for recurrences as well.
The prior described procedure results in a CEO-ED-Matrix, which illustrates all CEO appointments in publicly traded US companies in the timeframe of 2002 to 2013. It further features all EDs who have been recruited after the CEO started at the firm. As soon as a new CEO takes over, the following recruitments are accounted to the same. This matrix scheme represents the basis for the SNI measure. This spreadsheet consists of 3,200 EDs, 2,250 CEOs and 2,262 companies from 2002 to 2013.
First an independent variable to reveal the ties among directors has to be set. It results out of a modified version of the Social Network Index introduced in Fracassi and Tate (2012) and is from now on identified as SNI.
This measure SNI is defined by all connections between the CEO and the appointed executive. The introduced method is derived from the one used in Duchin and Sosyura (2013) and combined with the SNI of Fracassi and Tate (2012). This Thesis looks at the five different classes of BoardEx information. Board Employment and Other Employment are the most comprehensive tables. They are used to measure previous employment ties. If a CEO has worked in a firm previous to his appointment, this firm will be looked at. If the designated executive in question has also worked at that entity before the CEO appointment date, a tie can be verified. As all positions in these tables are at least senior management level, up to the board of directors, it is likely that managers know each other, even as they didn’t work at the company for the same period.
This assumption is supported by Duchin and Sosyura (2013, p. 398), where the authors find that educational connections “foster a sense of belonging to a common group”. They find evidence with “alumni clubs, donations to the home school, and college sports”. Accordingly two directors share an educational tie if they belong to the same alumni network, which the paper equals with acquiring a degree at the same university.
Consequently having worked in the same company before, at least at a senior level, is also comparable to an alumni network. People stay in contact, respectively know each other from the businesses mixers or other social events. The time restrictions prevent that the SNI will get too high. Only past work relationships are accounted for the independent variable.
As set out above in the prior paragraph, Activities, Achievements and Education are treated like employment data. All connections that can be established before the CEO’s initiation date are incorporated in the SNI. The CEO ED Matrix is overlapped by a spreadsheet that calculates all connections for each CEO/ED combination. The formula to derive such ties can be seen disentangled in the appendix.4
The new Matrix shows all SNI possibilities for the CEO-ED connections. This thesis will employ to show their impact on companies.
Control Variables
The other independent variables like boardtenure, boardconnection, executiveeducation and executiveage as well as firm fixed effects are also built with the aid of the BoardEx data.
The variable Boardtenure includes the period a candidate has already worked as an executive in a public board, before getting appointed by the CEO. It shows the candidate’s experience on public boards. It is derived from the Average Yrs on Public Boards variable that is supported by BoardEx in the spreadsheet Characteristics as the total number of
years a Director has sat at board level on quoted boards, divided by the number of them. Even though a director might have been on different boards, this variable can count as a cardinal experience measure. In this case it is twice as much experience, if a director sits on two boards in one year.
The dummy variable boardconnection will include the period a candidate has already worked as an executive in the firm, before getting appointed by the CEO. The combination of the ED’s birth date in Characteristics and his changing age every year in the table Boards reveals the actual starting year to the variable Time on Board. This variable is described as the length of time, stated in years, that the Director has been on the Board. The Time on
Board variable is used to derive the EDs first year on the company’s board. If that is before
getting appointed by the CEO, Boardconnection will take the value of one and shows that the candidate has connections to the board (e.g. if he already served at the board or has strong professional ties to board members).
Executiveeducation accounts the level of education an ED has. The BoardEx variable
Number of Qualifications describes the number of qualifications held by the Director. It is a
count of all qualifications of degree level including all professional qualifications (e.g. PhD or master’s degree), the executive holds. Each qualification has the count one.
The variable executiveage will consider the age of the professional by the time he gets appointed. It can be calculated using the BoardEx Date of Birth variable.
Time in Company shows the length of time, stated in years, the ED spends in the company.
Similar to executivetenure, a dummy variable can be calculated. This value called
companyinsider shows, if a candidate was already in the firm, before the CEO got
appointed.
Gender and nationality of the directors is disclosed in the database as well. To measure the executives’ team performance, the Return on Assets (ROA) and Return on Equity (ROE) ratios are calculated. As, among others, Execucomp is used to combine as much data as
possible, the ROA is only calculated using the net income divided by total assets. Execucomp uses a ROA without adjustment to interest payments on debt.
The CUSIP needs to be calculated from the companies ISIN numbers to merge S&P’s Compustat and Execucomp with BoardEx data. This is an easy procedure, as this thesis only observes US firms. The first two letters, as well as the last digit of the ISIN have to be abandoned to create the CUSIP. If the CUSIPs doesn’t match, the firms’ ticker is used.
Testing procedures
To answer Hypothesis 1a, descriptive statistics will be used. We look at the distribution plotted in a histogram. It links the SNI to ED hires.
To investigate if there are significant differences between the two groups of hired EDs, a Mann-Whitney U test will be applied as testing procedure. The U test allows to compare two samples without requiring them to be normally distributed. One sample will include all EDs who came from inside their firm, while the other will contain the ones who did not work for their companies before. The same will be done with male and female directors. Like a t-test this nonparametric test controls whether the two samples statistically differ or just randomly. This method will answer the Hypotheses 1b and 1c.
To find the consequential positive or negative impact of CEOs’ executive personnel decisions for their companies, a performance check is done. Return on Assets (ROA) and Return on Equity (ROE) are controlled in different regressions. Firms with the fewest connected executives (connections to the CEO) are compared to ones with the most connected executives. The outcome of this comparison should support the rationale of this thesis and addresses Hypothesis 2. The following regressions are conducted:
1. Δ𝑅𝑂𝐴𝑖= 𝛽0+ 𝛽1𝑡𝑖𝑒𝑠𝑖+ 𝛽2𝑏𝑜𝑎𝑟𝑑𝑡𝑒𝑛𝑢𝑟𝑒𝑖+ 𝛽3𝑏𝑜𝑎𝑟𝑑𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖+ 𝛽4𝑒𝑥𝑒𝑐𝑢𝑡𝑖𝑣𝑒𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽5𝑒𝑥𝑒𝑐𝑢𝑡𝑖𝑣𝑒𝑎𝑔𝑒𝑖+ 𝑢𝑖
2. Δ𝑅𝑂𝐸𝑖= 𝛽0+ 𝛽1𝑡𝑖𝑒𝑠𝑖+ 𝛽2𝑏𝑜𝑎𝑟𝑑𝑡𝑒𝑛𝑢𝑟𝑒𝑖+ 𝛽3𝑏𝑜𝑎𝑟𝑑𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑖+ 𝛽4𝑒𝑥𝑒𝑐𝑢𝑡𝑖𝑣𝑒𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛𝑖+ 𝛽5𝑒𝑥𝑒𝑐𝑢𝑡𝑖𝑣𝑒𝑎𝑔𝑒𝑖+ 𝑢𝑖
This thesis controls performance with the absolute change in percentage points in the two performance indicators over different time horizons since CEO appointment. These time horizons are one, two and three calendar years commencing with the first full fiscal year after the CEO has been hired.
IV. Results
SNI ranges between 0 and 11 with a mean of 1.020 and a median of 1. The total number included in the sample is 2,338. The average board tenure is 5.95 years while the
executive’s average age is 54 years when getting appointed. The youngest are 29 and the most senior is 88. An executive gained on average two educational degrees. 95% of the directors are male and 96% of the executives are Americans.
The following histogram as shown in exhibit 1 is used to underpin Hypothesis 1a. A social tie is given if there is one or more connections.
Figure 1:Histogram of sample SNI
0% 20% 40% 60% 80% 100% 0 200 400 600 800 1000 1200 1400 1600 1800 0 1 2 3 4 5 6 7 8 9 10 11 >11 Cu m u lativ e % Fr e q u e n cy SNI
CEOs prefer to hire or advance executives they have stronger ties with. As 45.4% of the directors have no ties to the CEO, the majority of the hired executives have at least one connection to the hiring CEO. Thus one can conclude a support of the data to confirm Hypothesis 1a.
The research question Outcomes differ if an executives is appointed from among inside or outside the company.
Figure 2: U test insider
Mann-Whitney Test for Two Independent Samples Hyp median/mean: 0 Outsider Insider count 324 3042 median 0 1 rank sum 512554 5154107 U 525704 459904
one tail two tail
alpha 0.05 U 459904 mean 492804 std dev 15592.9229 ties z-score 2.10993155 effect r 0.03636732 U-crit 467155.924 462242.433 p-value 0.01743213 0.03486425 sig (norm) yes yes
One can conclude from the above analysis that executive directors from among inside a company statistically differ from outsiders at the 5% significant level. The insider has a median of one tie whereas the outsiders show median of no ties. This difference between the groups is not random and therefore supports the Hypothesis 1b.
Figure 3: U test Gender
Mann-Whitney Test for Two Independent Samples Hyp median/mean: 0 Female Male count 163 3253 median 1 1 rank sum 270645 5565591 U 272960 257279
one tail two tail
alpha 0.05 U 257279 mean 265119.5 std dev 11514.7228 ties z-score 0.68091088 effect r 0.01165015 U-crit 246179.466 242551.058 p-value 0.24796394 0.49592789 sig (norm) no no
The figure above supports the Hypothesis 1c that Gender does not matter as it comes to social ties. The Mann-Whitney test is not significant at the 5% level.
Finally the results for the regression analysis to answer Hypothesis 2, firms with a strong c Connected management perform better than weak connected ones.
Figure 4: Regression Results Return on Equity
(1) (2) (3)
VARIABLES
deltaroe1 deltaroe2 deltaroe3 t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1 sni 157.5 115.6 289.8 (0.342) (0.213) (0.610) boardtenure -121.9 -31.88 -97.32 (-1.201) (-0.273) (-0.952) executiveeducation 772.0 746.3 704.4 (1.174) (0.979) (1.055) boardconnection 5,770*** 5,215*** 5,309*** (3.511) (2.710) (3.148) executiveage 177.9** 174.3** 165.1** (2.444) (2.055) (2.221) gender -236.8 -758.5 -2,160 (-0.0811) (-0.223) (-0.723) constant -15,615** -14,358* -11,034 (-2.310) (-1.817) (-1.593) observations 1,833 1,756 1,756 R-squared 0.012 0.009 0.010 R2_ajd 0.00864 0.00570 0.00707 Prob > F 0.0013 0.0137 0.0053
The return on equity figures for the time horizons of one, two and three years after the CEO has been appointed can be explained by the regression model. At a significance level of between 0.0013 and 0.0137. Although the significance of the explanatory power, the R2 is comparingly low at 0.9% and 1.2%. This means that other factors have larger impact on the performance of the firm. When we look at the indicator of the social ties, the SNI is not significant. Against this, the general board connection and the age of the executive are significant to explain performance of the company.
As robustness check the SNIs are clustered into two groups separating low and high network indexes. The cutoff point was chosen at an SNI of 2 and 3 respectively.
Figure 5: Robustness Test group 0-2
(1) (2) (3)
VARIABLES
deltaroe1 deltaroe2 deltaroe3 t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1 sni_012 1,054 963.3 1,191 (0.548) (0.430) (0.606) boardtenure -121.0 -31.55 -94.15 (-1.196) (-0.271) (-0.924) executiveeducation 769.2 745.2 698.2 (1.170) (0.978) (1.046) boardconnection 5,797*** 5,240*** 5,348*** (3.528) (2.723) (3.171) executiveage 178.1** 174.7** 165.1** (2.447) (2.059) (2.221) gender -239.3 -756.6 -2,178 (-0.0820) (-0.222) (-0.729) constant -15,590** -14,382* -10,865 (-2.313) (-1.825) (-1.573) observations 1,833 1,756 1,756 R-squared 0.012 0.009 0.010 R2_ajd 0.00874 0.00578 0.00707 Prob > F 0.0012 0.0130 0.0053
Figure 6: Robustness test Group 0-3
(1) (2) (3)
VARIABLES
deltaroe3 deltaroe2 deltaroe3 t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1 sni_0123 1,191 963.3 1,191 (0.606) (0.430) (0.606) boardtenure -94.15 -31.55 -94.15 (-0.924) (-0.271) (-0.924) executiveeducation 698.2 745.2 698.2 (1.046) (0.978) (1.046) boardconnection 5,348*** 5,240*** 5,348*** (3.171) (2.723) (3.171) executiveage 165.1** 174.7** 165.1** (2.221) (2.059) (2.221) gender -2,178 -756.6 -2,178 (-0.729) (-0.222) (-0.729) constant -10,865 -14,382* -10,865 (-1.573) (-1.825) (-1.573) observations 1,756 1,756 1,756 R-squared 0.010 0.009 0.010 R2_ajd 0.00707 0.00578 0.00707 Prob > F 0.0053 0.0130 0.0053
Figure five and six support the above findings. Therefore the chosen regression model is not influenced by the linearity or nonlinearity of the SNI and performance. The second performance measure ROA turned out not to be relevant. This holds true for the two robustness szenarios.
V. Conclusion
The relation between social ties in the context of executive hiring and firm performance is only partially supported by the analyzed dataset covering US sample of the S&P500 between 2002 and 2013. A higher explanatory power of firm performance have the board connection and the age of the executive. Gender has no impact on the social ties and hiring, but hired executives from inside the company tend to show stronger ties then directors from outside, which have on average no ties to the CEO. The sample demonstrates the majority of hirings have social ties. This interesting research topic could be amended including salary ranks and a match of the birth years of connected directors.
Bibliography:
Baker, M. and Wurgler, J., 2002, Market Timing and Capital Structure, The Journal of Finance,
57, 1-32.
Berger, A.N., Kick, T., Koetter, M. and Schaeck, K., 2013, Does it pay to have friends? Social ties
and executive appointments in banking, Journal of Banking & Finance, 37, 2087– 2105.
Butler, A.W. and Gurun, U.G., 2012, Educational Networks, Mutual Fund Voting Patterns, and CEO
Compensation, The Review of Financial Studies, 25, 2533-2562.
Cai, J., Walkling, R.A. and Yang, K., 2012, The Price of Street Friends: Social Networks, Informed
Trading, and Shareholder Costs, Working Paper, Drexel University.
Chen, J., 2014, Blackberry CEO hires another former colleague to drive turnaround, Thomson
Reuters, viewed 13. January 2014, from
http://reuters.com/article/idUSL2N0KN1D420140113?irpc=932.
Cohen, L., Frazzini, A. and Malloy, C., 2010, Sell-Side School Ties, The Journal of Finance, 65, 1409-
1437.
Duchin, R. and Sosyura, D., 2013, Divisional Managers and Internal Capital Markets, The Journal of
Engelberg, J., Gao, P. and Parsons, C.A., 2012, Friends with Money, Journal of Financial Economics, 103, 169–188.
Engelberg, J., Gao, P. and Parsons, C.A., 2013, The Price of a CEO’s Rolodex, The Review of Financial Studies, 26, 79-114.
Fracassi, C. and Tate, G., 2012, External Networking and Internal Firm Governance, The Journal of
Finance, 67, 153-194.
Fracassi, C., 2012, Corporate Finance Policies and Social Networks, Working Paper, University of
Texas.
Glaser, M., Lopez-de-Silanes, F. and Sautner, Z., 2013, Opening the Black Box: Internal Capital
Markets and Managerial Power, The Journal of Finance, 68, 1577-1631.
Hwang, B.-H. and Kim, S., 2009, It pays to have friends, Journal of Financial Economics, 93, 138-
158.
Landier, A., Sauvagnat, J., Sraer, D. and Thesmar, D., 2012, Bottom-Up Corporate Governance,
Review of Finance, 17, 161-201.
Nguyen, B.D., 2011, Does the Rolodex Matter? Corporate Elite’s Small World and the Effectiveness
Pan, Y. and Wang, T.Y., 2012, First Year in Office: How Do New CEOS Create Value?, Working
Paper, University of Utah.
Shue, K., 2013, Executive Networks and Firm Policies: Evidence from the Random Assignment of
MBA Peers, The Review of Financial Studies, 26, 1401-1442.
Subrahmanyam, A., 2008, ‘Social Networks and Corporate Governance’, European Financial Management, 14, 633–662.
Appendix
Glossary and Assumptions
Average Yrs on Public Boards
Total number of years a Director has sat at Board Level on Public Boards, divided by the number of them
CEO
Chief Executive Officer CUSIP
Committee on Uniform Security Identification Procedures
Executive Director
A full time employed individual who is on the company Board ISIN
International Securities Identification Number Net Income
The company's total earnings: revenues adjusted for costs of doing business, including Interest, depreciation and taxes; also referred to as net earnings.
Non-Executive Director
Any member of a company’s Board who is not an employee of the company Return On Assets
RoA = Net income / Total Assets Return On Equity
RoE = Net Income / Equity, with: Equity = Total Assets – Total Liabilities SD
Supervisory Director SM
Senior Manager SNI
Time (Yrs) in Company
The length of time, stated in years, spent by the Director in the company * If the date of joining the Company is not available, then the assumed date is the start of the financial year when the Director is first Quoted in the Annual Report * If the date left the Role is not known then the assumed date is the end of the financial year
Time (Yrs) on Board
The length of time, stated in years, that the Director has been on the Board If the date of joining the Board is not available, then the assumed date is the start of the financial year when the Director is first Quoted in the Annual Report * If the date of leaving the Board is not known then the assumed date is the end of the financial year
Total Assets
Total value of company's assets. Total assets include Current Assets; Fixed Assets such as buildings and equipment; and other assets such as licenses and good will.
Total Liabilities
Description algorithm and formula of the file “SNI Model incl Dates”
The SNI Model (SNI = Social Network Index) shows how many intersections between
Directors and CEOs exist (Director being hired by the respective CEO). The formula looks for intersections in the CompanyIDs each of them lists in their CVs.
The Worksheet “Matrix” is created for this model. It shows a list of CEOs and the Directors they hired. The original Matrix is derived from the BoardEx data and its generation is understandably explained by a separate Excel file. The colours separate the different formulas used in the calculation.
I. Original formula direct from Excel (Worksheet “FoundActivities”, cell “F3”) before adding the “timing” component:
=WENN(ISTZAHL(Matrix!F3);SUMME(WENN(ISTFEHLER(VERGLEICH(WENN(ISTZAHL(INDIRE
KT("Activities!$C"&VERGLEICH(FoundActivities!$A3;Activities!$A:$A;0)&":$L"&VERGLEICH(F oundActivities!$A3;Activities!$A:$A;0)));INDIREKT("Activities!$C"&VERGLEICH(FoundActiviti es!$A3;Activities!$A:$A;0)&":$L"&VERGLEICH(FoundActivities!$A3;Activities!$A:$A;0));"x");
INDIREKT("Activities!$C"&VERGLEICH(Matrix!F3;Activities!$A:$A;0)&":$L"&VERGLEICH(Mat rix!F3;Activities!$A:$A;0));0));0;1));"")
II. Breakdown and description of the formula:
=WENN( Check: ISTZAHL(Matrix!F3); Then: SUMME( WENN( Check: ISTFEHLER( VERGLEICH(
Search criterion: WENN( Check: ISTZAHL(
INDIREKT("Activities!$C"&VERGLEICH(FoundActivities!$A3;Activities! $A:$A;0)&":$L"&VERGLEICH(FoundActivities!$A3;Activities!$A:$A;0))) ;
Then:
INDIREKT("Activities!$C"&VERGLEICH(FoundActivities!$A3;Activities! $A:$A;0)&":$L"&VERGLEICH(FoundActivities!$A3;Activities!$A:$A;0))
[We build the row, starting with the column C to end column L. The row number is determined with the aid of MATCH. MATCH locates the row of the CEO (first column of the similar row in the worksheet “FoundActivities”) in the worksheet “Activities”]
Otherwise: ;"x");
[IF it is ISNUMBER, then we look for it]
Search matrix:
INDIREKT("Activities!$C"&VERGLEICH(Matrix!F3;Activities!$A:$A;0)&":$L"&V ERGLEICH(Matrix!F3;Activities!$A:$A;0))
[We build the row, starting with the column C to end column L. The row number is determined with the aid of MATCH. MATCH locates the row of the Director (from the similar located cell in the worksheet “Matrix”) in the worksheet “Activities”]
MATCH type: ;0))
[If there is no accordance, the output is FALSE]
Then: ;0 Otherwise: ;1))
[Sum of the calculated zeros and ones]
Otherwise: ;"")
[Sum, if Director is a digit, otherwise the cell is empty]
The formula calculates the quantity of connections between a CEO and the Director hired by this CEO.
There is no time factor involved. Therefore we add a simple limitation to our model: The CEO must have joined the CompanyID before the Director, while the Director must have joined the CompanyID before being hired by the CEO.
(The colours “CEO” and “Director” now highlight the characters of the formula which will be replaced in the next step)
1. [Variable end column for the CEO’s CompanyIDs]
L=WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>$C3:$L3)=WAHR();SPALTE($C3:$L3);0)); 4);1;)
[We replace the fixed column L with a formula to include time restrictions]
$C3:$L3=INDIREKT("ActivitiesDate!$C"&VERGLEICH(FoundActivities!$A3;ActivitiesD ate!$A:$A;0)&":$L"&VERGLEICH(FoundActivities!$A3;ActivitiesDate!$A:$A;0)) [1. Step to replace the fixed end of the row]
L=WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>INDIREKT("ActivitiesDate !$C"&VERGLEICH(FoundActivities!$A3;ActivitiesDate!$A:$A;0)&":$L"&VERGLEICH(FoundAc tivities!$A3;ActivitiesDate!$A:$A;0)))=WAHR();SPALTE(INDIREKT("ActivitiesDate!$C"&VERG LEICH(FoundActivities!$A3;ActivitiesDate!$A:$A;0)&":$L"&VERGLEICH(FoundActivities!$A3; ActivitiesDate!$A:$A;0)));0));4);1;);““) L=INDIREKT("EndActivities!B"&VERGLEICH(FoundActivities!$A3;ActivitiesDate!$A:$A ;0))
[2. Step: We use the Worksheet “EndActivities” to locate the last populated cell for the CEO’s CompanyIDs of the Worksheet “Activities” and replace the L]
L=WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>INDIREKT("ActivitiesDate !$C"&VERGLEICH(FoundActivities!$A3;ActivitiesDate!$A:$A;0)&":$"&INDIREKT("EndActiviti es!B"&VERGLEICH(FoundActivities!$A3;ActivitiesDate!$A:$A;0))&VERGLEICH(FoundActiviti es!$A3;ActivitiesDate!$A:$A;0)))=WAHR();SPALTE(INDIREKT("ActivitiesDate!$C"&VERGLEIC H(FoundActivities!$A3;ActivitiesDate!$A:$A;0)&":$"&INDIREKT("EndActivities!B"&VERGLEI CH(FoundActivities!$A3;ActivitiesDate!$A:$A;0))&VERGLEICH(FoundActivities!$A3;Activitie sDate!$A:$A;0)));0));4);1;);"")
2. [Variable end column for the Director’s CompanyIDs (same steps as above)]
L=WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>$C3:$L3)=WAHR();SPALTE($C3:$L3);0)); 4);1;)
$C3:$L3=INDIREKT("ActivitiesDate!$C"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A; 0)&":$L"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)) L=WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>INDIREKT("ActivitiesDate !$C"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)&":$L"&VERGLEICH(Matrix!F3;Activities Date!$A:$A;0)))=WAHR();SPALTE(INDIREKT("ActivitiesDate!$C"&VERGLEICH(Matrix!F3;Acti vitiesDate!$A:$A;0)&":$L"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)));0));4);1;);““) L=INDIREKT("EndActivities!B"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)) L=WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C3>INDIREKT("ActivitiesDate !$C"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)&":$"&INDIREKT("EndActivities!B"&VE RGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0))&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0) ))=WAHR();SPALTE(INDIREKT("ActivitiesDate!$C"&VERGLEICH(Matrix!F3;ActivitiesDate!$A: $A;0)&":$"&INDIREKT("EndActivities!B"&VERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0))&V ERGLEICH(Matrix!F3;ActivitiesDate!$A:$A;0)));0));4);1;);"")
IV. New formula including timing component:
(The colours “CEO” and “Director” now represent the changes made in the original formula)
=WENN(ISTZAHL(Matrix!E2);SUMME(WENN(ISTFEHLER(VERGLEICH(WENN(ISTZAHL(INDIRE KT("Activities!$C"&VERGLEICH(FoundActivities!$A2;Activities!$A:$A;0)&":$L"&VERGLEICH( FoundActivities!$A2;Activities!$A:$A;0)));INDIREKT("Activities!$C"&VERGLEICH(FoundActivi ties!$A2;Activities!$A:$A;0)&":$"&WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Mat rix!$C2>INDIREKT("ActivitiesDate!$C"&VERGLEICH(FoundActivities!$A2;ActivitiesDate!$A:$ A;0)&":$"&INDIREKT("EndActivities!B"&VERGLEICH(FoundActivities!$A2;ActivitiesDate!$A: $A;0))&VERGLEICH(FoundActivities!$A2;ActivitiesDate!$A:$A;0)))=WAHR();SPALTE(INDIREK T("ActivitiesDate!$C"&VERGLEICH(FoundActivities!$A2;ActivitiesDate!$A:$A;0)&":$"&INDI REKT("EndActivities!B"&VERGLEICH(FoundActivities!$A2;ActivitiesDate!$A:$A;0))&VERGLEI CH(FoundActivities!$A2;ActivitiesDate!$A:$A;0)));0));4);1;);"")&VERGLEICH(FoundActivities !$A2;Activities!$A:$A;0));"x");INDIREKT("Activities!$C"&VERGLEICH(Matrix!E2;Activities!$A :$A;0)&":$"&WENNFEHLER(WECHSELN(ADRESSE(1;MAX(WENN((Matrix!$C2>INDIREKT("Ac tivitiesDate!$C"&VERGLEICH(Matrix!E2;ActivitiesDate!$A:$A;0)&":$"&INDIREKT("EndActivi ties!B"&VERGLEICH(Matrix!E2;ActivitiesDate!$A:$A;0))&VERGLEICH(Matrix!E2;ActivitiesDa te!$A:$A;0)))=WAHR();SPALTE(INDIREKT("ActivitiesDate!$C"&VERGLEICH(Matrix!E2;Activiti
esDate!$A:$A;0)&":$"&INDIREKT("EndActivities!B"&VERGLEICH(Matrix!E2;ActivitiesDate!$ A:$A;0))&VERGLEICH(Matrix!E2;ActivitiesDate!$A:$A;0)));0));4);1;);"")&VERGLEICH(Matrix! E2;Activities!$A:$A;0));0));0;1));"")
