Master in Business Economics, Track Finance
Master Thesis
The Efficiency of the Director Labor Markets:
Evidence from Europe.
July 2015
Alexandrina Pankovska
Statement of Originality
This document is written by Alexandrina Pankovska 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.
Alexandrina Pankovska, 5th of July, 2015
Abstract
This study assesses the relevance of the financial market returns for the election of non-‐ executive directors. The paper shows that the European director labor markets reward outsiders for their superior performance. The analysis exploits the unique database DirectorInsight, covering board movements in 1.054 European listed firms in the time frame 2008-‐2014. I identify the causal effect of the shareholder value creation on the directors’ rewards by constructing separate panel regression models. The positive effect of the abnormal portfolio returns is documented in the three different specifications with the number of directorships, the assets under control and the total granted remuneration. The empirical results provide evidence that the firms nominate directors with multi-‐area commitments hence those busy directors have a negative effect on the market returns after appointment. This thesis finds low significance for the after-‐appointment effect of outperforming directors on the investors’ wealth.
Keywords: corporate governance, nomination process, non-‐executive director, busy board, stock return
Table of Contents
I. Introduction ... 4
II. Related Literature ... 6
A. Institutional Background ... 6
B. Empirical Evidence ... 8
III. Data and Methodology ... 11
A. Sample Construction ... 11
B. Identification Strategy ... 13
C. Descriptive Statistics ... 17
IV. Empirical Results: Hypothesis 1 ... 19
A. Effect of Buy-‐and-‐hold Abnormal Returns on Number of Directorships ... 19
B. Effect of Buy-‐and-‐hold Abnormal Returns on Directorship of Total Assets ... 21
C. Effect of Buy-‐and-‐hold Abnormal Returns on Directorship Total Remuneration ... 22
V. Empirical Results: Hypothesis 2 ... 24
VI. Robustness checks and discussion ... 27
VII. Conclusion ... 29 VIII. References ... 31 IX. Appendix ... 43
I. Introduction
Building up a career as a non-‐executive director is frequently the case for retiring top managers. An example for such a professional path could be seen in the curriculum vitae of Dr. Paul Achleitner. In 2009, next to his Chief Financial Officer of Allianz he served as supervisory board member in three DAX listed companies-‐ RWE, Henkel and Bayer. In 2010 he acquired an additional directorship at Daimler. 2012 was the year, when he stepped down from his financial executive role, but was nominated for the reputable position of chairman of the Deutsche Bank.
It appears intriguing from a corporate governance perspective to establish the election criteria for outsiders. The recent governance code improvements call for an increased transparency in the directors’ appointment process, normally performed by the firms’ nominating committees. The majority of the empirical papers in the field use the share price movements as a performance measure for managerial skills (Fee and Hadlock (2003)). Therefore this paper focuses on the relevance of the stock returns for additional directorship nominations. Furthermore, I investigate the question from a rewards perspective and show that directors delivering high returns for their shareholders receive reputational and monetary benefits. Thus, director labor markets act efficiently if they manage to incorporate the information about superior performance into individuals’ promotions. A further implication of my research is the hypothesis, that outperforming non-‐executive directors manage to increase shareholder wealth after their appointment. Such investigation could be interesting for academics and investors and serve as a useful guideline for the participants on the directors labor market.
The common knowledge suggests that directors serving in successful companies are contributing to the prosperity of the firm with their expertise and professional experience. Therefore it seems that the probability for such effective directors to get nominations for other boards is higher than for those performing below the average. Following the argumentation by Adams and Ferreira (2007) the board of directors usually votes on important projects such as merger and acquisition proposals, divestments or capital structure changes. Moreover, the board members are responsible for the replacement of a CEO and thus affect the market returns. Such decisions
influence directly the shareholder value, which could be used ex-‐post as a proxy for the directors’ performance. Furthermore, in some European countries like Switzerland the directors receive part of their remuneration in shares, which will incentivize them to work harder and thus achieve higher rank among their peers.
On the other hand, prior research such as the paper by Fracassi and Tate (2012) pointed out that external networking and social connections in the past matter for corporate governance. Consequently such linked directors with more than one supervisory role on different boards could gain power and entrench themselves similar to the executives. In this case even directors, engaged in bad corporate governance and shareholder value destroying practices, will get nominations on other boards. A further implication of the new appointments could be the insignificance of the prior firms’ returns in comparison to other social relations and individuals’ characteristics. A feasible reason for this irrelevance might be the way the nomination committees evaluate the candidates. For instance, an individual with financial and accounting background would most likely replace a departing chairperson of the audit committee. In this sense a high shareholder value creation will be less desirable than a suitable education and a professional skillset.
Furthermore, this paper aims at providing evidence about the after-‐effect of directors´ appointments on shareholder value. Experienced, outperforming monitors are mostly valued by the investors due to governance improvements. However, the prior academic research suggests that too many appointments on different boards might distract the directors and decrease their attention and performance (Fich and Shivdasani (2006) and Jiraporn, Kim and Davidson (2008)). Thus this paper contributes to the existing literature on the busy boards with a fresh view on the European market.
My identification strategy is based on the unique European database DirectorInsight and covers the years 2008-‐2014. I begin the analysis by proving the hypothesis that the lagged buy-‐and-‐hold abnormal returns have positive causal effect on the number of directorships, the total assets under control and the remuneration per individual. European labor markets seem to reward directors for their investors’ wealth contribution, according to both the linear and non-‐linear panel regression models.
Furthermore, I examine the long-‐term effect on market returns of appointments of successful individuals. The analysis with an additional control for the importance of the assumed position delivers weak evidence for the contribution of outperformers in the third year after election. The second specification incorporates a busy factor and doesn’t succeed to establish positive effect of the nomination of an outperforming director. However, I find evidence that overcommitted individuals have negative impact on the returns.
This thesis is structured as follows. Section II provides information about the regulatory environment and the empirical papers in the field of board of directors. Section III starts with an explanation about the sample and its construction. It continues with the formulation of the hypothesis and the methodology and ends with the descriptive statistics. Section IV presents the results for the first Hypothesis and is split into three sub-‐sections, describing the different specifications. Section V contains the empirical implication of the second hypothesis. Section VI describes a robustness check and provides a discussion on the results. Section VII concludes.
II. Related Literature
A. Institutional Background
In order to understand the driving forces in the directors’ election process, the juridical environment should be considered. Firstly, it should be noticed that the majority of the European states are recognized having developed financial markets and actively participate in organization such as The Financial Stability Board and The World Bank. Among the 30 members of the Organization for Economic Co-‐operation and Development (OECD) mostly European countries could be found. Therefore the majority of the listed EU companies follow the non-‐binding guidelines and good corporate governance principles, developed initially by the OECD in 1999. The OECD rules appear to have a general advising character and serve as a foundation for the development of the regional corporate governance codes. According to the convention, the shareholders should be able to participate in the election and removal of board members. This basic shareholder right grants an access to the proxy material of the companies before AGMs and allows the owners to put own candidates on the ballot if
the necessary requirements such as size of the holding are satisfied. The OECD principles do not prescribe the establishment of a nominating committee with independent directors, but interpret this step as an example of good corporate governance. Furthermore, the regulation defines the main tasks of the board and includes the adequate return for shareholders as a responsibility of the directors. The next section moves the focus from the broad view of the OECD principles to the narrow local requirements.
Secondly, the European market includes countries both with common law and civil law practices, which ends up in different local corporate governance codes. Table A in the Appendix provides an outline of the requirements in Austria, Belgium, France, Germany, Italy, Sweden, The Netherlands and the United Kingdom. As shown in the summary, the codes distinguish from each other in their prescriptions for a unitary or dual board structure, independence criteria and cooling-‐off periods. Regarding the election process of the candidates, the maximal appointment period and the busy board restriction appear to have the highest relevance. In Sweden and UK the non-‐ executives are elected on an annual basis for a term of one year till the next annual shareholder meeting. In such cases the shareholders could evaluate the directors performance more frequently and replace bad performance with more suitable individuals. Therefore the possible higher frequency of directors’ rotation might result in incentive for the directors to please the owners and deliver high returns. Additionally, several European countries like Austria, Belgium, France and the Netherlands include an advisory provision for the maximal number of the directorships held by an individual. Such requirements aim at securing the full attention and commitment of the directors in their role as supervisors. Thus, even the best directors, measured by their contribution to the shareholder wealth, are bounded in their available engagement with different companies. In such situations it might be suggested that successful directors would be rewarded with positions in bigger firms and move across markets to gain more influence. The legal requirements have an additional link to the ownership structure in the particular countries. As demonstrated by La Porta et al (1999) a small fraction of the companies in countries like Austria, Belgium or Portugal are widely held. As an explanation factor the researchers point out the shareholder protection
mechanisms and show that in the Anglo-‐American financial markets include more widely held corporation. Thus, the circular relationship between regulation and ownership dispersion will lead to divergent governance regulations. Relevant examples could be found in the nomination process of directors. As summarized in the Report of Principles for Responsible Investment (PRI), a United Nations-‐ supported initiative, the election practices in countries such as Sweden and Italy differ strongly from the ones in France or Germany. The authors point out that the concentrated ownership of the first two states lead to greater rights of the shareholders. For instance the Swedish Code of Corporate Governance foresees a nomination committee, which consist of members of the Board of Directors and additional shareholder representatives, who are usually directly chosen by the largest four or five investors. As stated in the PRI Report, in Italy the Corporate Governance Code or Codice di Autodisciplina allows minority shareholders to submit a list of individuals, suitable for directors. All candidates are eligible for election at the Annual General Meeting. The active participation of the Swedish and Italian shareholders should lead to higher importance of the stock returns when evaluating the candidates for directors’ positions. In contrary, the German and the French legislative systems focus on the rights of the stakeholders. Both European states require an employee representation on the board of directors. Essentially, in France the workers get at least one board seat and the number of such directors is increasing with the board size. In Germany the staff members get from one-‐third to 50 % of the board seats depending on the number of employees in the particular company. In such cases it might be argued that the shareholder value is irrelevant criteria for the election, for the fact that the directors are chosen among the workforce. Nevertheless, it should be mentioned that some of the employee representatives come from workers unions. Such individuals serve on several boards in the same industry and the shareholder value could be seen as a performance measure for their directorships.
B. Empirical Evidence
The major part of the existing academic research on the board of directors is concentrated on the characteristics and the size of the supervisory venue. On the one hand, researchers such as Linck et al (2007) concentrated on the determinants of the board structure. They discovered evidence for the impact of the firms’ economic
conditions, such as growth opportunities and stock volatility and the legal environment, changed by the Sarbanes-‐Oxley Act on the supervisory venue. On the other hand, papers such as Coles et al (2008) and Hermalin and Weisbach (2003) investigate the other causality direction: the impact of specifics of the board compositions, for instance the proportion of independent non-‐executive’s, on the Tobin`s Q or the board actions. The research of Fahlenbrach et al (2013) contributes to the discussion about the regulation on independency of the boards with an alternative view. They argue that outside directors are more concerned with reputation than insiders and tend to abandon easily bad performing firms. These studies deliver important insights for the choice of directors and this thesis could be seen as a complement in the recruiting process.
Additionally, the papers by Fee and Hadlock (2003) and Brickley et al (1999) consider the stock performance and accounting measures as a signal for managerial skills. Fee and Hadlock (2003) investigate the executives’ jumps from one company to another. Similar to this empirical strategy, this thesis also aims to establish causality chain of the firm performance and new directors appointment. In contrast to the study by Fee and Hadlock (2003), the focus isn’t on the management team but on the non-‐executive suite. Brickley et al (1999) study the transition of executive to non-‐executive position and hence the relevance of the CEOs performance for appointments as non-‐executive positions after retirement. They find evidence that the high stock returns have positive effect on the probability that the CEO will stay in the company as supervisor and that the accounting figures matter for nominations outside the own company. Thus the paper by Brickley et al (1999) finds as a link between the shareholder wealth and the willingness of the owners to appoint the outperformers as non-‐executive supervisors. Likewise Kaplan and Reishus (1990) consider the companies’ dividends policy as performance metrics and discover that the chance that the CEO will get outside directorships decreases with around 50 % after dividend cuts. The researchers define the cuts as a bad performance metric and interpret the disapproval of the shareholders as fewer non-‐ executive directorships. Regarding the outside positions, the papers by Ferris et al. (2003) and Yermack (2004) include an analysis of the impact of firms’ performance on new appointments. Using a series of multivariate logistic regressions Ferris et al (2003)
find evidence that a higher operating margin causes more appointments. Moreover, they conclude a statistical insignificant but positive relationship between the market-‐to-‐ book ratio and the number of positions held by an individual. The main results appear to be robust in the short term hence their sample include positive abnormal returns around the day of announcement for a director with multiple positions. Similarly Yermack (2004) observes positive correlations between firm performance and additional board seats using binary models. In contradiction to these papers, the examination of the hypothesis in this thesis is built on more sophisticated models that allow more dynamic movements on the directors market. Additionally, all accounting profitability measures could be unimportant to the shareholders, if the earnings are not distributed in the form of dividends or share buyback programs. Therefore the methodology implemented in this paper uses the stock returns as more accurate performance measure from shareholders perspective than accounting figures. Thus this thesis could provide alternative view on the signaling argument and serve as extension on the empirical research in this area.
A third extensive corporate governance literature stream includes the research on the busy boards. In light of the invitation of outperforming director to serve on a new board, Fich and Shivdasani (2006) estimate lower CEO Turnover sensitivity, worse market to book ratio and weaker operating profitability for boards with majority of busy directors. Furthermore, Jiraporn, Kim and Davidson (2008) measure the excess value of a firm as the difference between the market value of the whole and its segments and report a diversification discount for companies with busy directors. The researchers’ findings contradict in this sense the reputation hypothesis that claims positive relationship between the multiple directorships and the companies’ outcomes. The negative effect of the directors’ busyness doesn’t support the networking and learning benefits from an additional position. Their results differ from the ones by Ferris et al. (2003) and Perry and Peyer (2005), who oppose the negative impact of multiple directorships on firms’ performance. In case that a company doesn’t suffer from agency problems, Perry and Peyer (2005) estimate positive announcement returns effect, if an executive is joining another company as supervisory board member. Controlling for the ownership of the executive and the independence of the board, the researchers
estimate more positive market reaction when the management member is joining a financial, same industry or high-‐growth company. An important insight from the paper by Perry and Peyer (2005) are the significantly higher stock returns of the “sender” companies compared to the “receiver” companies prior to the announcement. Regardless of the lack of a causality argument, it might be suggested that those outperforming directors were chosen to join the new boards for their contribution to the shareholder value creation. Although this thesis doesn’t intend to make an investigation on the busy boards concept, it considers it as a factor in the nominating process. In this way it contributes to the existing literature.
Finally, the one of the main differences to the papers mentioned above is the concentration of this paper on the European supply market for directors. Due to discrepancy in the shareholder rights, corporate practices and cultural models, there could be unexpected conclusions arising.
III. Data and Methodology
A. Sample Construction
The main data source used for this master thesis is the Director Insight platform, developed by AMA Partners. The database contains information on both director movements and remuneration policies. The compensation component covers information for the board of directors of European listed firms in the time frame 2008-‐ 2014. The data is hand-‐collected and extracted from the companies’ annual and remuneration reports. The movement section is updated on a daily basis according to firms’ press releases and general publicly available information. In total, the database consists of 1.054 companies from 26 indexes, traded on major European financial markets. Moreover, Director Insight incorporates the profiles of 7.322 individuals including characteristics such as gender, date of birth and nationality. Overall, 46.012 year-‐firm-‐individual combinations represent unique observations, in which data about stock return and accounting measures from S&P Capital IQ is merged into the original sheet. The simple total shareholder returns and the asset base of the companies over the previous three years are necessary for the construction of an averaged abnormal stock return measure. For the simple TSRs I use the stock price of two subsequent
periods, divide them and then subtract 1. In order to capture the effect of past stock performance on further nomination of the directors I adjust the data towards year-‐ individual observations. This step involves the accumulation of the quantitative measures per individual. With respect of the small minus big effect, documented by Fama and French (1992), I average the simple returns with the asset base of the firms, where a director already serves. In this way the issue with smaller firms outperforming the bigger ones could be avoided:
𝑇𝑆𝑅!" = 𝐴𝑠𝑠𝑒𝑡 𝐵𝑎𝑠𝑒 . 𝐴𝑛𝑛𝑢𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛
𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡 𝐵𝑎𝑠𝑒
For instance, the ex-‐CEO of SAP SE Prof. Henning Kagermann served as a non-‐ executive director in BMW AG, Deutsche Post AG and Nokia Co. in 2012. The firms had Asset Base parameters of 131.835, 33.857 and 29.984 million euro and simple annual returns of 46,9%, 45,8% and -‐15,0% respectively. Thus the TSR variable for Mr. Kagermann in 2012 was 37,2%.
As a control portfolio, the different European indexes are adopted. The benchmarking is necessary condition due to the fact that companies’ stock prices share common movement with the market and certain firms might have enjoyed gains or suffered losses without actually out-‐ or underperforming firms in other indexes. Thus, implementing the market-‐adjusted returns is a tool to overcome the issue with “lucky”/ ”unlucky” directors. Following the argumentation by Barber and Lyon (1997) I use the buy-‐and-‐hold abnormal returns as performance metric. The researchers document significant downsizes biases in event studies, relying on the cumulative abnormal returns especially when normalizing with a reference portfolio such as the market index. Moreover, the long-‐term horizon with annual returns over three years seems to be better captured in the BHARs rather than the CARs. For the exact estimation of the BHAR I use the suggested approach by Lyon et al (1999). The BHRs are computed through the multiplication of the shareholder returns of the three years prior. The estimated values represent the continuing scope of an investor holding the portfolio of companies, where the particular director serves. The normalizing procedure follows the
same steps for the index returns and ends up in a subtraction from the BHRs. The final version of the buy-‐and-‐hold abnormal returns takes the following form:
𝐵𝐻𝐴𝑅
!"=
𝑇𝑆𝑅
!" ! !!!−
𝐼𝑛𝑑𝑒𝑥_𝑇𝑆𝑅
!" ! !!!The adjusted sample consists of 27.645 unique observations. Next to the performance measures every single line includes the number of positions held by an individual, the sum of the asset base of the companies, where the director serves and the total remuneration, awarded in the particular year. Moreover, the variables are transformed using the winsorising statistical approach. The procedure appears a necessary step in order to account for outliers due for instance to stock splits and mergers that impact the returns. A description of the variables and the data sources is provided in Table B from the Appendix.
Besides this original sample, I prepare a second data panel including 1.730 new appointments for the methodology of the second hypothesis. The spreadsheet consists of Individual IDs, the name and ID of the company, where a director was appointed and the TSR measures of the firm and its index in the following 3 years after nomination. Furthermore, adjusting with the type of position that the individual assumes enriches the information about those appointment events. Hence the Lead Position variable equals 2 for a chairman role, 1,5 for a vice-‐chairman or a senior independent director and 1 otherwise, the influence strength of the individual in the company could be measured.
B. Identification Strategy
As shown in the theoretical paper by Raheja (2005) the composition of the board matters for the choice between a good and bad project. Therefore, investors should not only be concerned about the choice of top managers, but also examine carefully the election of non-‐executive directors. Considering the stock returns as a result from strategic and operational decisions, shareholders could use the share price development as a proxy for individual’s ability. Assuming that the European directors’ supply markets function efficiently, one should observe that high shareholder returns lead to more
appointments. Specifically, good directors, who signal their ability through the share price increase, are going to benefit in the form of more positions, directorships in bigger firms and higher compensation. Thus the first hypothesis investigated in this paper is:
H1: Non-‐Executive directors are rewarded for high stock returns.
In order to identify the statistical significance of the firm performance for the probability of being appointed to another board, panel regression model is implemented. The left-‐hand side variable in the estimation is the number of directorships (ND) held. The independent variable of the regression is represented by the average abnormal buy-‐and-‐hold returns of the companies, where a director is already serving.
𝑁𝑟. 𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟𝑠ℎ𝑖𝑝𝑠!" = 𝛽!+ 𝛽!. 𝐵𝐻𝐴𝑅!,!!!+ 𝛾𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜖!" (1)
A lagged value is used due to the fact that the recruiting companies need firstly to see performance of the directors and then make use of this knowledge in the selection process. Advisors will mostly likely access the performance of the individuals also looking at least three years in the past, therefore the buy-‐and-‐hold appear appropriate measure. In order to confirm the first Hypothesis, a significant positive coefficient 𝛽!is expected.
Another useful specification for the significance of the results would be a second regression with the sum of assets of the directors’ employers as a dependent variable. The use of this accounting measure would allow the identification of jumps from smaller to bigger companies and vice versa. The estimation is more precise and reveals the cases of rewarding successful directors, who give up their prior appointment. For instance directors who get a new appointment and resign from other board in the same year might have a constant number of directorships over several years. However those companies probably differ from each other so the directorship of total assets would have higher variation. This aspect appears to be especially important in countries where there is a restriction on the maximum number of positions held by in individual. There are several European Corporate Governance Codes that prescribe such limitations in order to secure the dedication of the directors to their supervisory role. Moving towards
the higher end of positions, there could be individuals that follow this substitution mechanism for reputational reasons.
𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟𝑠ℎ𝑖𝑝 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠!" = 𝛽!+ 𝛽!. 𝐵𝐻𝐴𝑅!,!!!+ 𝛾𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜖!" (2) Furthermore, a third specification represents a panel regression with the total annual remuneration as a dependent variable. This compensation measure has indeed a high correlation with the number of positions, but adds one more motivation reason for the individuals to try to gain more positions. Having a monetary incentive, occasionally even directly linked to shareholder value parameters though variable pay could stimulate the non-‐executives to perform their duties more accurate. The causality chain of achieving higher returns for the investors and getting higher compensation will be proven by a significant positive 𝛽! coefficent.
𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟𝑠ℎ𝑖𝑝 𝑇𝑜𝑡𝑎𝑙 𝑅𝑒𝑚𝑢𝑛𝑒𝑟𝑎𝑡𝑖𝑜𝑛!" = 𝛽!+ 𝛽!. 𝐵𝐻𝐴𝑅!,!!!+ 𝛾𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜖!" (3)
Additionally, the regressions include a set of controls for instance the age of the director and the busyness factor. According to prior empirical research such as the paper by Ferris et al. (2003) the age of the individuals can be used as a proxy for experience. Following their argumentation, the older directors participated actively longer on the directors’ market and are more valued by the shareholders. The rich background will result in higher probability for being invited to serve on more boards. The variable Age is winsorized at the 1% level in order to avoid extreme outlier on the high end of the data that are due to typos in the set. The regression (1), (2) and (3) might suffer from omitted variable bias in the form of overcommitted directors. Controlling for the busyness factor appears the most logical step to avoid biased results. The common approach, known from the most empirical papers on the busy board theory suggests the use of the number of directorships as a control. However, it is not feasible to use this measure, because it appears on the left-‐hand side of the first specification and is highly correlated with the dependent variables in the second and third analyses. Instead this research implements a different approach, particularly the number of countries where a director serves. It could be argued that individuals serving in different states substitute the time spent in effective monitoring with unnecessary traveling. The important debate on the effect of nominations in multiple boards ends up in dubious empirical results. A positive
coefficient of the control variable Busy Factor could be seen as a support for the networking and gaining experience argument, the negative one – evidence for the reduced attention.
The second part of this thesis focuses on the after-‐appointment period and investigates the contribution of outperformers to the shareholder value.
H2: Outperformers in the past increase firm value after joining a new company. While the first hypothesis could rather serve as guideline for individuals, who want to pursue career as non-‐executive directors, the second one investigates the perspective of the shareholders and their benefits from appointing outperformers. Being interested in value creation, the firms’ owners tend to seek objective valuation of the boards performance. Thus being able to proof that individuals, who achieved high rates of returns for their shareholders in the past, will continue their hard work, will give an additional reason to boards to search for such directors. Similarly to the first hypothesis the long-‐term perspective measured by the BHARs seems more robust shareholder wealth measure than the market reaction observed through the announcement returns. A simple OLS econometrics model using an Outperformer Dummy and the abnormal annual returns is a possible tool to capture the variation. However, this regression could be seen as rather weak and can suffer from reversed causality. For instance it could be argued that better firms attract high-‐skilled directors and these firms outperform their peers in general. Moreover, the causality chain could be threatened by omitted variable bias such as other firm characteristics that improve the firm performance. It might be suggested that in times that the company runs its business less efficiently and is taken over, not only the directors but also the CEO are replaced. There could be increased indicators and higher returns observed, but this growth would be rather attributable to the new management than to the non-‐executive directors. Furthermore, returns are sensitive to shocks such as oil prices disruptions, so that the volatility of the stock prices responds to general economic conditions rather than corporate governance characteristics. All these endogeneity problems would end up in biased estimates in the simple OLS econometric model. In order to minimize the effect of biases I run several regressions on different time windows from the year of appointment till three years
after that. An additional robustness check for the results is the measurement of the returns in case a lead role on the board was assumed. It might be argued that directors holding important positions such as the chairman of the board or of the audit, nomination or remuneration committee are having greater impact on the firms decisions and therefore should matter more for the firms stock price. Therefore, a positive sign for both the 𝛽! and 𝛽! coefficents will give an evidence to confirm the
second hypothesis.
𝐼𝑛𝑑𝑒𝑥 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑇𝑆𝑅! =
𝛽!+ 𝛽!. 𝑂𝑢𝑡𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑒𝑟! + 𝛽!. 𝐿𝑎𝑔𝑔𝑒𝑑 𝐼𝑛𝑑𝑒𝑥 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑇𝑆𝑅! + 𝛽!𝐿𝑒𝑎𝑑! + 𝛽!𝑂𝑢𝑡𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑒𝑟! ∗ 𝐿𝑒𝑎𝑑!+ 𝜖! (4)
Moreover, those individuals with many outside directorships could be considered “busy” and in reality have lower influence on the shareholder value. Assuming that the Busy Factor doesn’t capture the experience and the positive contribution of a director, one would suggest that the betas of the Busy Factor and the interaction term would have a negative sign.
𝐼𝑛𝑑𝑒𝑥 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑇𝑆𝑅! =
𝛽!+ 𝛽!. 𝑂𝑢𝑡𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑒𝑟! + 𝛽!. 𝐿𝑎𝑔𝑔𝑒𝑑 𝐼𝑛𝑑𝑒𝑥 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑇𝑆𝑅! + 𝛽!𝐵𝑢𝑠𝑦 𝐹𝑎𝑐𝑡𝑜𝑟! + 𝛽!𝑂𝑢𝑡𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑒𝑟! ∗ 𝐵𝑢𝑠𝑦 𝐹𝑎𝑐𝑡𝑜𝑟! + 𝜖! (5)
Regressions (4) and (5) are run separately on the periods t+1, t+2 and t+3. C. Descriptive Statistics
Table 1 presents information about the gender, age and nationality of the directors. As expected the majority of the directors (83.1 %) is male. The female representation quotas were recently introduced in many of the European countries, for instance Germany passed newly the 30%-‐rule at the beginning of 2015. Other European companies such as the Belgian ones have time to comply with the rule till 2018. Therefore, the 16,9% in the sample appears to be in line with the market conditions. The average age of the directors is 58,24 and the median is 59. The two figures are close enough to suggest that the observations are equally spread around the average and there is no skewness in the sample. Consistent with the findings by Ahern and Dittmar
(2012), the female directors in this sample are younger than their male colleagues. Considering the nationalities, the directors from UK, France and Germany are having the highest frequency in the data set. The three are seen as countries with well developed and big financial markets, which would reflect in a broad and deep directors labor market.
[Insert Table 1 here]
Figure 1 examines the difference between two groups if directors. The first one is called Outperformers and includes all the individuals that achieved above average buy-‐ and-‐hold abnormal returns and the second one – the underperformers contains of the rest. Already at this stage of the analyses one could notice that those belonging to the better directors hold a higher number of positions on average. The results provide support for the first hypothesis in the years 2009-‐2013. The beginning and the end of the sample remain rather puzzling. One possible explanation might be found in the corporate governance codes originally drafted in the beginning and mid 2000’s. The “comply or explain” principles defined an independent director as a non-‐executive who had served for the company under a certain amount of terms. For instance an outsider who have been with the company for more than 2 terms is considered to have developed tight relationship with the management and could not monitor the actions of the insiders objectively. This special milestone of the European best practices could have forced many companies to replace good directors with bad ones after the good ones served already for 8-‐10 years and reached their independence limit. This development will timely match the significant decrease in 2014.
[Insert Figure 1 here]
Next, Figure 2 provides provisional results for the second hypothesis. The histogram charts represent the distribution of the abnormal returns in the periods after the appointment t+1, t+2 and t+3. At this stage already it could be recognized that the good directors actually do not serve on boards of companies with high abnormal returns. The blue bars are rather left side skewed in t+1 and at least equally distributed to the underperformers’ returns in t+2. The situation changes slightly in t+3, when one could observe more Outperformer observations in the higher end of the returns scale.
[Insert Figure 2 here] IV. Empirical Results: Hypothesis 1
A. Effect of Buy-‐and-‐hold Abnormal Returns on Number of Directorships The analysis of the first hypothesis will start by presenting the empirical results from several regressions, testing the significance of the BHARs on the number of directorships. Table 2 summarizes the results from six different econometrical models. Firstly, it should be mentioned that those of the regressions not using fixed effects include clustered standard errors. As described in the paper by Petersen (2009) the cross-‐correlation in the panel might cause downsize or upwards bias in the beta estimations. In this thesis there are around 7 yearly observations per director, so it could be argued that there is time-‐series dependence. Using clustering procedure on the residuals according to the Individual ID, I aim at improving the efficiency of the econometrical models. The first two columns include tobit models without and with a control for the age of the director. In the first one the coefficient of the BHAR is positive but statistically insignificant. However, the second specification already provides an evidence to confirm H1 at 10% confidence level. The coefficient of the buy-‐and-‐hold abnormal returns of 0.05 could be denoted as the increase in the average number of positions in case that the director achieves 1% rise in the shareholder wealth. It might be even further elaborated so that the beta coefficient is interpreted as a 5 % probability increase for getting a new appointment. The third and the forth regressions are models that include nationalities dummies in order to control for those specific differences among the directors. This step is motivated by the fact that the companies might prefer individuals from specific countries and those directors are more likely to accumulate directorships without delivering high performance in the form of shareholder value. Such preferences might be the outcome of language barriers, cultural specifics or general local labor markets differences. Taken together, a British or an American director might be more valued by the recruiters and acquire positions faster than an individual from a smaller country. The procedure results in even higher and more significant beta coefficients of 0.0894 and 0.1132. Both estimations are having explanatory power at the 1% confidence level. The last two specifications account for
the individual-‐fixed effects and provide positive hence insignificant coefficients for the variable of interest. It appears that this panel data modeling is having too solid restrictions and the variation across the time is not enough to provide evidence even at the lowest significance line. Such results are not surprising due to the fact that the individual-‐fixed effect controls for all the measures that are not changing over time and leave the model with variables with maximum of 7 observations for the time span 2008-‐ 2014. Moreover, this specification rules out the individuals’ skills by controlling away part of the directors’ performance factors that are permanent over the event window. Overall, all these findings are consistent with the results by Ferris et al. (2003) and Yermack (2004). Considering the control for the age, the data establishes positive relationship between the variable and the number of directorships held. This fact is unsurprising and leads to the conclusion that older individuals are more likely to get more appointments. The coefficient of the Age measure is rather small, ranging between 0,0064 and 0,024 and implies only a tiny probability for more appointments with every single year passed. All of the regressions in Table 2 exclude the control Busy Factor mainly for the reason that the left-‐hand-‐side variable is highly correlated with this measure. The dependent variable is constructed very similarly to the Busy Factor so the higher significance of such coefficient would be primarily due to its conception rather than establishing causality. For a correlation matrix of the three dependent variables and the Busy Factor consult Table C in the Appendix.
[Insert Table 2 here]
In order to capture most efficiently the variation in the variables I construct three additional models. Considering the fact that some of the European countries such as Austria, Belgium, France and the Netherlands prescribe in their corporate governance codes a maximum amount of directorships that an individual can hold, the regressions in Table 3 include the square term of the BHARs. Thus, the models account for the possible non-‐linear relationship between the dependent and independent variable. Economically interpreted directors are rewarded with positions to a certain point. When they reach the maximum number of directorships the benefit of increase in the portfolio returns is not having a significant impact anymore. The results in Table 3 confirm this hypothesis, almost in all of the specifications. Firstly, the beta coefficient of all the buy-‐and-‐hold