1 CEO DECISION-MAKING AUTONOMY AND FIRM PERFORMANCE:
A STEWARDSHIP THEORY PERSPECTIVE June 2017
Master Thesis International Economics & Business Daan Best
S2204681
e-mail: d.best@student.rug.nl University of Groningen Faculty of Economics and Business
Groningen
Supervisor: Prof. dr. H. van Ees Co-assessor: dr. D. Soudis
ABSTRACT
Drawing on a framework from stewardship theory, I examine the relation between the decision-making autonomy of Chief Executive Officers (CEOs) and the financial and non-financial firm performance using robust fixed effect, probit and dynamic panel data estimation methods for a sample of 1441 US firms in a period of five years 2010-2014. In addition, I examine the moderating effect of performance based compensation on the financial and non-financial performance of firms. The results reveal that even when controlling for board and firm characteristics, two CEO decision-making autonomy proxies, Duality and Tenure, are positively related to financial firm performance. Furthermore, the results indicate a moderating effect of performance based compensation on the relation between CEO decision-making autonomy and non-financial firm performance. The results are in line with stewardship theory.
2
ABSTRACT ... 1
1. INTRODUCTION ... 3
2. THEORY DEVELOPMENT AND HYPOTHESES ... 4
2.1. Perspectives... 4
2.1.1. Agency theory ... 4
2.1.2. Stewardship theory ... 5
2.2. Hypotheses development ... 7
2.2.1. Autonomy and financial performance ... 7
2.2.2. Autonomy and non-financial performance ... 8
2.2.3. Autonomy and compensation ... 8
3. RESEARCH DESIGN ... 10
3.1. Data ... 10
3.1. Variables ... 11
3.1.1. Financial firm performance ... 11
3.1.2. Non-financial firm performance ... 12
3.1.3. Decision-making autonomy ... 13 3.1.4. Moderator variable ... 14 3.1.5. Control variables ... 14 3.2. Methodology ... 15 3.2.1. Methodology Hypothesis 1A ... 15 3.2.2. Methodology Hypothesis 1B ... 16
3.2.3. Methodology Hypothesis 2A and 2B ... 16
3.3. Diagnostic tests and econometric issues ... 17
3.3.1. Multicollinearity ... 17
3.3.2. Heteroskedasticity ... 17
3.3.3. Fixed and Random effects model ... 17
3.3.4. Probit and Logit model ... 18
3.3.5. Endogeneity and GMM ... 18
4. EMPERICAL RESULTS ... 19
4.1. Descriptive statistics ... 19
4.2. Correlation coefficients ... 20
4.1. Results Robust Fixed Effects regressions ... 20
4.1.1. Hypothesis 1A ... 20
4.1.1. Hypothesis 2A ... 23
4.2. Results Probit regressions ... 27
4.2.1. Hypothesis 1B ... 27
4.2.2. Hypothesis 2B ... 27
4.3. Results Generalized methods of moments regressions ... 28
5. LIMITATIONS AND FUTURE RESEARCH... 29
6. CONCLUSION ... 30
7. APPENDIX ... 32
3
1. INTRODUCTION
Recent studies have shown that the number one indicator of job satisfaction is job autonomy (e.g. Zunker, 2006). Multinationals and smaller firms are increasingly experimenting with awarding autonomy to their employees and the results are predominantly positive. Famous examples are Google, where employees can spend 20% of their time at work that directly interests them, and Netflix, where they are allowed to take as much vacation as they want. The outcomes are in line with the predictions of stewardship theory, the equivalent to agency theory. The stewardship theory assumes that people are intrinsically motivated to achieve goals set by the organization because they value cooperation more than defection (Davis, Schoorman & Donaldson, 1997). Increasing the autonomy therefore improves firm performance. This is in contrast with the assumptions of agency theory that describe people as opportunistic and will choose own goals above the goals of the organization. The need for extrinsic motivational factors is therefore needed to align the interests of the agent with those of the organization.
The decision to grant autonomy to CEOs and other top managers is an important decision due to the direct relation between CEOs and firm performance: CEOs are held responsible in times of bad firm performance and are rewarded in times of good firm performance. The decision to grant the CEO more autonomy with the intention to improve firm performance therefore depends on whether the expectations are in line with stewardship or agency theory. When CEOs are expected to be stewards, firm performance will improve when autonomy is granted. This is in contrast to the agency perspective, which assumes opportunistic behaviour and therefore states to limit autonomy.
4 I use a panel dataset of 1441 US firms in the period 2010-2014 resulting in 7205 observations. The dataset is created by combining data from The Wharton Research Data Services Database (WRDS) and the Tomson Reuters Datastream. WRDS is used for CEO related data and financial and non-financial data is retrieved from Datastream.
Robust fixed effects estimation, probit estimation and generalized methods of moments (GMM), a dynamic panel data regression method, are used to answer the research questions. The remainder of this paper is organized as follows. The next section reviews agent and stewardship literature and based on this literature hypotheses will be developed. The research design of this study will be discussed in section 3, where the sampling method, methodology, the variables and econometric issues will be addressed. The empirical results will be presented in section 4 and section 5 focusses on the limitations and ideas for future research. The final section concludes with a summary.
2. THEORY DEVELOPMENT AND HYPOTHESES
In the corporate governance literature, two theories have different perspectives on autonomy. The first theory is the agency theory, which is the dominant paradigm underlying most governance research (Davis et al., 1997). The other theory is stewardship theory, which has its founding in the sociology and psychology research and runs counter to the agency (Davis et al., 1997; Donaldson & Davis, 1991).
2.1. Perspectives
2.1.1. Agency theory
5 opportunistic behaviour of agents. According to the agency theory, it is therefore necessary to align the interests between agent and principals to prevent opportunistic behaviour (Bosse & Phillips, 2016; Davis et al., 1997).
However, Argyris (1973: 253) argues that many of the assumptions made by agency theory are “non-relevant to many critical problems of the organization”. Humans have a need to improve their current state and reach higher levels of achievement (Davis et al., 1997). Job tasks that are not externally rewarded could still be performed because self-actualization motivates humans to move beyond a current state of affairs (Argyris, 1973). Argyris (1973: 253) therefore argues for a “more complex and humanistic model of man”.
Madison et al. (2015) agree with Argyris and argue that the need for an alternative model of human decision-making is relevant as the underlying assumption of behaviour of CEOs in the complex organizational life is different from the situation described by agency theory. Stewardship theory describes an alternative model in which the stewards, in this paper CEOs, are not opportunistic but are rather motivated by intrinsic factors (Francoeur, Melis, Gaia & Aresu., 2017; Davis et al., 1997).
2.1.2. Stewardship theory
In contrast with the agency theory, the stewardship theory assumes that collectivistic behaviour, oriented toward the collective and focused on the social system (Moorman & Blakely, 1995), yields higher utility to stewards than individualistic behaviour. Hence, stewards value cooperation more than defection (Davis et al, 1997). Individuals like CEOs identify themselves with the mission of the organization and, as a result, align their own goals with the goals of the principals and their organization (Francoeur et al., 2017). The behaviour of stewards can be described as self-motivated and non-opportunistic (Francoeur et al., 2017). The behaviour of stewards creates trust within the organization. Stewards make decisions in the best interest of the company because the gain in firm performance will increase the utility of the steward itself.
6 consequence”. Stewards are motivated and value aspects like determination, self-efficacy and feelings of purpose (Francoeur et al., 2017; Manz, 1990).
Both theories have different implications for the mechanisms of alignment and the amount of decision-making autonomy the CEO needs to effectively guide the firm towards performance. Agency theory argues that the interest alignment is important. This is done through incentive schemes and monitoring. Only when the CEO is monitored and interests are aligned, should the CEO be awarded with decision-making autonomy. Stewardship theory believes the opposite is true. Stewards need high decision-making autonomy to effectively guide and control the organization. There is no need for monitoring and alignment because stewards identify themselves with the mission of the organization. They realize the trade-off between personal and organizational objectives and believe personal needs are met by working toward the organizational objectives (Davis et al., 1997).
Studies on agency theory or stewardship theory found contradicting evidence. For example, Veprauskaite & Adams (2013) found evidence for the agency theory that CEO decision-making power is negatively related to firm performance in the UK. On the other hand, Dedman (2016) investigated the claim by UK regulators who recommended that the CEO should not become Chairman of the board. In line with agency theory, the regulators argued that this would prevent CEOs from misusing their power. However, Dedman (2016) found no evidence that supports the agency view. Mamatzakis and Bermpei (2015) found that CEO power has a positive effect on US investment banks performance, consistent with the stewardship hypothesis. Francoeur et al. (2017) studied the relation between CEO compensation and environmental commitment and found that environmental friendly firms pay their CEO less compensation and rely less on incentive-based compensation.
7 2.2. Hypotheses development
2.2.1. Autonomy and financial performance
Because autonomy is a difficult concept to measure, most studies use power measures as a proxy for autonomy. In line with previous studies, the autonomy measures described in this study are power as autonomy measures. Power as autonomy is “a form of power that allows one person to ignore and resist the influence of others and thus to shape one’s own destiny” (Lammers, Stoker, Rink & Galinsky, 2016: 499). In other words, the increase in power of the CEO leads to an increase in the decision-making autonomy of the CEO (Lammers, 2016). Hence, the CEO can make decision without the influence of the board. The goals of the steward are in line with the goals of the organization as the stewardship theory assumes CEOs can be awarded with autonomy to decide which actions should be taken in order to achieve the goals set by principals (Davis et al., 1997).
Because stewards will not behave opportunistically, the governance structures must be set in such a way the organizational actions taken by the CEO are best facilitated. This is done by giving them high authority and discretion (Donaldson & Davis, 1991). From a stewardship perspective, the CEO’s authority to make autonomous decisions is positively associated with the performance of firms and “should be deliberately extended to maximize the benefits of stewards” (Block, 1996: 25). In contrast with agency theory where decision-making autonomy increases the need for monitoring and bonding costs, is the amount of resources needed to guarantee pro-organization behaviour from stewards diminished (Davis et al., 1997). Furthermore, control can diminish the CEO its productivity because it undermines the pro-organization behaviour of the CEO due to lowering the steward’s motivation (Argyris, 1964).
As a result, I argue that CEOs who identify themselves with the organization and therefore align their goals with the goals of the organization will increase the financial performance of the company when they have higher decision-making autonomy. Stewards will receive autonomy because they do not need to be monitored due to their intrinsic motivation and the alignment of goals with the company. Following this reasoning, I can formulate our first hypothesis:
8 2.2.2. Autonomy and non-financial performance
Previous papers investigating stewardship theory mainly investigated the presence of stewards in family firms (Madision et al., 2015). Family firm ownership and management is often unified, resulting in an environment of aligned goals (Madison et al., 2015). Family firms were chosen because the reasoning is that especially family businesses desire to pursue noneconomic goals like corporate social responsiblitiy. They value future prospects of the firm higher than financial performance today. However, Klein, Shapiro & Young (2005) found evidence that contradicts this statement. As a result, no distinction between family and non-family firms will be made.
Noneconomic goals are in line with stewardship theory as stewards not only value financial performance (Madison et al., 2015). Stewardship theory can explain the commitment towards other stakeholders such as the natural environment (Driscoll & Starik, 2004), besides the commitment towards the shareholders (Godos-Diez, Fernandez-Gago & Martinez-Campillo, 2011). Stewards believe it is important to act social responsible because humans are the dominant species on this earth and therefore have the moral duty to protect it (Francoeur et al., 2017). Moreover, human beings might value non-financial performance because they enjoy the intrinsic rewards like moral satisfaction or social approval (Andreoni, 1989).
Arnaud, Wasieleski & Bus (2014) argue that autonomy nurtures the level of social responsibility. When the CEO is awarded with autonomy, the decision to act social responsible can be made by the CEO without the interference of principals who value financial goals over non-financial goals. Hence, the ability of CEOs to strive for non-financial goals increases when their autonomy increases. As a result, the probability that CEOs improve non-financial performance increases when they receive more autonomy.
Accordingly, I will try to find evidence that CEOs who are stewards will improve the non-financial performance of firms, regardless the firm characteristics. In line with stewardship theory, I expect that the probability of non-financial performance will increase when autonomy is given to the CEO. This results in the following hypothesis:
Hypothesis 1B: decision-making autonomy of CEOs is associated with a higher probability of non-financial firm performance
2.2.3. Autonomy and compensation
9 will improve the performance of firms but this relation may be weaker for CEOs receiving performance based compensation.
As previously mentioned, stewardship theory argues that stewards are intrinsically motivated and already focused on organizational objectives. Stewards always seek to achieve goals in line with the company instead of falling into self-interested behaviour. As a result, there is no need to use explicit incentives to align the goals of the steward with the goals of the organization. Financial rewards that are aimed to direct the CEOs attention towards shareholders can detract them from serving other principals (Francoeur et al., 2017). The incentives may decrease the willingness to allocate time and efforts towards objectives that are not covered by rewards. This can cause overinvestment of time and resources into certain objectives and underinvestment in objectives not covered by rewards. Furthermore, bonuses and long-term compensation are extrinsic motivational factors to align goals but are unnecessary due to the intrinsic motivation of CEOs. Therefore, it is expected that CEOs feel no higher level of incentive to improve the firm performance when they are rewarded with bonuses or other sorts of performance-based compensation because performance compensation can be an indicator of extrinsic motivation, a characteristic in conflict with stewardship theory. Furthermore, bonuses can detract the CEO its attention towards objectives valued by the principals. These objectives may be not in accordance with the objectives set by the steward him or herself. Performance compensation may therefore decrease the performance of the firm.
It is therefore expected that autonomous CEOs who receive performance based compensation moderate the financial firm performance. In addition, the moderator variable might reverse the expected positive relation between autonomy and financial performance because performance compensation distracts the CEO from its initial objectives towards the objectives of the principals.
The same relation holds for the non-financial performance of firms. Stewards tend to value non-financial firm performance and feel the duty to act social responsible. Receiving performance based compensation is, according to stewardship theory, an ineffective method and might cause overinvestment in objectives covered by rewards and underinvestment in other objectives.
10 negatively influence the non-financial performance of firms, especially if they receive performance compensation. The same logic applies to the financial performance of firms. Following the above discussion, the third hypothesis can be formulated as:
Hypothesis 2a: There is a moderating relation between the format of compensation (fixed or performance based) and the relation with financial performance
Hypothesis 2b: There is a moderating relation between the format of compensation (fixed or performance based) and the relation with non-financial performance
3. RESEARCH DESIGN
3.1. Data
Most studies use managers or principals like CEOs and other top managers as data sources. With questionnaires and interviews they gather the data on CEO autonomy, goal alignment and financial and non-financial performance. Only a few papers use secondary data to investigate if the autonomy of CEOs will improve firm performance as expected by stewardship theory. Primary data gives more detailed information about, for example, the autonomy of the CEO but is very time consuming. The number of observations when using secondary data is much higher and there is no possibility of perception bias. Taking this into consideration, I decide to use secondary data.
11 the tenure of the executive cannot be correctly calculated when CEOs leave and later on return as CEO1.
The next step is to search in Bureau van Dijk’s database Orbis to acquire ISIN numbers; firm identification numbers needed to acquire financial information from Thomson Reuters Datastream. Companies were matched in Orbis on name, country of origin and ticker symbol resulting in 1460 identified firms.
The ISIN numbers are used to retain information from Thomson Reuters DataStream in which financial and other company information can be found. DataStream also gives access to Thomson Reuters Asset4, an environmental, social and governance information database. Due to the limited license that is granted to the University of Groningen a lot of variables cannot be retrieved. Other variables have a lot of missing observations. Two indicator variables were found having relatively many observations: if the company has a CSR committee and if the company report on belonging to a specific sustainability index. Both variables contain do not contain information about 2015. Therefore, the year 2015 is removed from the sample, decreasing the period to 2010-2014. For both variables 2860 observations were found.
3.1. Variables
3.1.1. Financial firm performance
To measure the financial performance of firms, I use three financial performance measures consistent with the literature. There are two kinds of financial performance measures: market-based and accounting-market-based. I use one market market-based measure (Tobin’s Q) and two accounting based measures (ROA and ROE).
The first variable is known as Tobin’s Q, a market based performance measure often used in stewardship and agency studies (Veprauskaite & Adams, 2013; Mamatzakis & Bermpei, 2015; Klein et al., 2005; Adams, Almeida & Ferreira, 2005). It is a theoretical construct widely used in the financial markets and it reflects the market performance of firms (Lopez Bernardo, Stockhammer & Lopez Martinez, 2016). It is a helpful measure for investors because it helps them to take investment decisions based on past and future performance. It is calculated as “the book value of total assets minus the book value of common equity plus the market value of common equity, divided by the book value of total assets” (Veprauskaite & Adams, 2013; Frijns et al., 2016). A Q ratio between 0 and 1 means that the cost to replace
1
12 the assets of the firm is greater than the value of its stock; the firm is undervalued. Above 1 is the other way around.
Return on assets (ROA) ratio is the second measure often used by studies (Graves & Shan, 2013; Veprauskaite, 2013; Lin, 2005; Klein et al, 2005; Mamatzakis & Bermpei, 2015). It shows how profitable a company is relative to the tangible assets it uses to generate cash flows (Veprauskaite & Adams, 2013). This operating performance ratio may foster a better view of the fundamentals of business because executives need to think about the assets that are best managed considering the capabilities of the firm (Hagel, Brown & Davison, 2010). It is calculated as the net income divided by the average total assets.
The third and final financial measure used by various papers (Veprauskaite & Adams, 2013; Donaldson & Davis, 1991; Mamatzakis & Bermpei, 2015; Lin, 2005) is the return on equity (ROE) ratio of firms. This accounting based measure is a profitability ratio like the ROA variable and indicates how effectively the money of stockholders is employed by the company (Veprauskaite & Adams, 2013). It is calculated by dividing the net income of the recent tax year by previous year’s book value of common equity (Veprauskaite & Adams, 2013).
It is expected that all three ratios are positively related to CEO decision-making autonomy. However, differences between Tobin’s Q, ROA or ROE can occur due to for example over or under valuation in the market or increase or decrease in assets or equity. Especially differences between the market based performance measure (Tobin’s Q) and the two accounting based performance measures (ROA and ROE) may exist because the market measure depends on the expectations and as a result the valuations of the investors. The operating measures do not depend on investor expectations. Because firms need to follow national and international principles while recording their financial statements, the accounting based measures are often considered more reliable.
3.1.2. Non-financial firm performance
13 The first indicator variable for non-financial performance is the presence of a corporate social responsibility committee (CSR committee). Firms that decide to create a new CSR committee signal their concern for social and environmental issues (Eberhardt-Toth, 2017). As a consequence, firms that established a CSR committee tend to value CSR more and therefore have higher non-financial firm performance. This dummy variable is coded as 1 if the firm has a corporate social responsibility committee (CSR) and 0 otherwise.
The second proxy is a CSR sustainability index (CSR Index) and indicates if the company belongs to a specific sustainability index. CSR indexes like the FTSE4Good Index or the Dow Jones Sustainability Index can be used to identify environmentally and socially sustainable companies. Firms that demonstrate strong environmental, social and governance practices are reported in this index and indicate to value CSR more than firms who do not belong to a certain CSR index. Consequently, firms reported on a CSR index have higher non-financial performance. The dummy variable is coded as 1 if the firm belongs to a CSR sustainability index and 0 otherwise.
3.1.3. Decision-making autonomy
The true decision-making autonomy of a CEO cannot be perfectly expressed in numbers and is hard to observe, therefore, I need to use proxies to estimate the decision-making autonomy of CEOs. The autonomy measures described in this study are power as autonomy measures. There are several proxies used by papers but none of the measures will capture every possible dimension of the decision-making autonomy (Veprauskaite & Adams, 2013). I use two different CEO-power related variables that will represent the degree of decision-making autonomy held by the CEO. These two sources of power are most researched and best supported in the literature (Veprauskaite & Adams, 2013; Combs et al., 2007; Mamatzakis & Bermpei, 2015).
14 the best proxies for decision-making autonomy. The annual title, obtained from Execucomp, is used to create a CEO-board duality indicator variable. When the CEO has the annual title of CEO and chairman, the dummy is coded 1, otherwise it is coded 0.
Tenure is the second variable and points out the number of years the CEO is in function. It is a key factor in the process of building decision-making autonomy. Decision-making power is positively related with the length of tenure because the CEO can influence the selection of board members (Combs et al., 2005; Veprauskaite & Adams, 2013). Furthermore, CEOs become more powerful because status and influence are gained when tenure increases (Perrone, Zaheer & McEvily, 2003). Tenure can improve performance because the CEO can apply the firm specific knowledge it gained in the previous years to influence board decisions (Brookman & Thistle, 2009). The tenure variable will describe the number of years the CEO is in post.
3.1.4. Moderator variable
In order to estimate the third equation, a moderator variable will be added. The moderator variable that will be added to the regression analyses is variable performance contract
(Pcontract). The dummy variable is coded 1 when the CEO receives compensation based on
performance and 0 otherwise. There will be performance based compensation when the CEO receives one or multiple bonuses in a year.
3.1.5. Control variables
Besides the variables that I use to construct the power index in order to measure the autonomy of the CEO, there are other variables that may influence the decision-making autonomy. Board size and the frequency of board interactions are variables used to control for potential influential factors of the board on CEO decision-making power. Board size (Size) indicates the total directors on the board and the frequency of board interactions (Meeting) is the total number of board meetings held during the fiscal year. Larger boards have increased monitoring capabilities and an additional pool of expertise, reducing the decision-making autonomy of the CEO (Guest, 2008).
15 With respect to firm performance, I need to control for other variables that have an influence on firm performance. In line with prior studies, I include firm characteristic variables like firm size (LnSales) and leverage (Leverage) and control for year effects (Lin, 2005; Veprauskaite & Adams, Klein et al., 2005; Combs et al., 2007; Graves & Shan, 2014)
Firm size controls for potential advantages of economies of scale and scope and market power (Klein et al., 2005). Size is measured in line with Veprauskaite & Adams (2013) as the natural logarithm of annual sales. This is a reasonable size measure because larger firms tend to have higher annual sales volumes than smaller firms.
The leverage of a firm is known to affect the performance of the firm (Lu & Beamish, 2004) and controls for risk characteristics of firms (Klein et al., 2005). Leverage is defined as the proportion of long-term debt to total assets.
3.2. Methodology
In order to answer the stated hypotheses using the data previously discussed, I will use three different models. With the use of the first model I will try to find evidence that the decision-making power of a CEO will positively influence the financial performance of firms (Hypothesis 1A). Second, I will use a similar model like the first one to find evidence that decision-making power has a positive effect on non-financial firm performance (Hypothesis
1B). Finally, I will add a moderator variable to estimate the moderating effect of performance
contracts on the financial and non-financial performance of firms. 3.2.1. Methodology Hypothesis 1A
I hypothesized the positive relation between the decision-making autonomy of a CEO and the financial firm performance. This can be characterized by the following expression:
Financial performanceit
= β1 + Dualityit + Tenureit + Sizeit + Meetingit + LnSalesit + Leverageit +
Industryeffectit + Yeareffectit + εit
(1)
where subscript i denotes ith firm (i = 1, …, 1441), subscript t denotes the tth year (t = 2010, …, 2014). Financial performance is one of the three dependent variables – ROA, ROE and
Tobin’s Q (as defined in the variable section). Duality and Tenure are the power variables that
16 3.2.2. Methodology Hypothesis 1B
The fourth and fifth models are rather similar to the first three but use different dependent variables to estimate the second hypothesis. In the fourth and fifth models, I try to estimate the influence of decision-making power of CEOs on the non-financial performance of firms. Because both dependent variables are dummy variables, a binary model needs to be used to find evidence that observations with particular characteristics falls in of the two categories: yes (y = 1) or no (y = 0). The expression can be formulated as:
Non-financial performanceit
= β1 + Dualityit + Tenureit + Sizeit + Meetingit + LnSalesit + Leverageit +
Industryeffectit + Yeareffecti + εit
(2)
Non-financial performance is one of the two dependent indicator variables – CSR Committee and CSR Index. The definition of all variables can be found in the variable section above and in Appendix 1. The model can be interpreted as when the autonomy increases, it is expected that the presence of a CSR committee or CSR Index increases. The results for the two non-financial firm performance variables can be found in models 4 and 5.
3.2.3. Methodology Hypothesis 2A and 2B
To find evidence for hypothesis 2A and 2B, an interaction variable is added to the models to estimate the moderation effect of a performance contract on the relationship between the power index and the performance of the firm. The model can be expressed as:
Performanceit
= β1 + Dualityit + Tenureit + (Pcontractit × (Dualityit + Tenureit)) + Sizeit +
Meetingit + LnSalesit + Leverageit + Industryeffectit + Yeareffectit + εit
(3)
17 3.3. Diagnostic tests and econometric issues
There are several factors that may be present and can influence the results of the models. I need to address possible issues in order to present reliable estimates of the financial and non-financial firm performance models as described in the methodology section.
In order to have reliable estimates there are several assumptions that must be satisfied. I will test if the assumptions are violated by employing several specifications tests.
3.3.1. Multicollinearity
One of the assumptions is the absence of multicollinearity. To test for multicollinearity, the correlation coefficients are checked and the Variance Inflation Factor (VIF) is used. The correlation coefficients give no indication of multicollinearity as the highest value reported is 0.557 between Tobin’s Q and ROA. The mean VIF is 1.31, with highest values 1.64 for the 2013 and 2014 year dummies, and therefore near the lower bound of 1. As a result, the correlation coefficients and the VIFs suggest there are no concerns for the possibility of multicollinearity in the data.
3.3.2. Heteroskedasticity
Furthermore, the models potentially exhibit heteroskedasticity. I test for the presence of heteroskedasticity by performing the Breusch-Pagan for heteroskedasticity test and found for four of the five models a p-value smaller than 0.01, indicating the possibility of heteroskedasticity at the 1% significance level. Only the CSR committee model had a p-value of 0.321 and is therefore not significant. To account for the presence of heteroskedasticity, the robust standard error options implemented in Stata12 will be used for the Tobin’s Q, the ROE, ROA and the CSR index models. The results for the Breusch-Pagan test for heteroskedasticity can be found in Appendix 2.
3.3.3. Fixed and Random effects model
18 financial firm performance can be found in Table 3 and will be discussed in the results section.
3.3.4. Probit and Logit model
In order to estimate the second model, the decision must be made which model to use because the two dependent non-financial performance variables are both indicator variables. Two models that can be used to estimate the non-financial dummy models are the probit model and the logit model. The probit model assumes a normal probability distribution while the logit model assumes logarithmic distribution. Differences between both models tend to be minimal when the estimated dataset is large. The differences between the estimations of the probit and logit model are as expected minimal. I decide to use the probit model because I assume a normal probability distribution of the presence of a CSR committee or if the firm belongs to a certain CSR Index. Firms may decide to create or to not create a CSR committee or take the effort to belong to a CSR Index or not. Furthermore, probit models can be generalized to account for heteroskedasticity that may be present in the CSR sustainability index data. The results for hypothesis 1B can be found in Table 4 and will be discussed in the results section.
3.3.5. Endogeneity and GMM
I need to be careful about drawing conclusions from the results of table 3 because there may be endogeneity issues. Hence, I need to account for the possibility of endogeneity in order to infer trustworthy conclusions from my results.
The econometric model that I employ to examine the empirical linkage between CEOs and their influence on the financial performance of firms is the generalized method of moments estimation (GMM) (Arrelano & Bond, 1991; Veprauskaite & Adams, 2013). I use in line with Mamatzakis & Bermpei (2015) the GMM estimation described by Roodman (2009) and known as the “xtabond2” specification in Stata. All explanatory variables are lagged by one year to mitigate endogeneity concerns.
19
4. EMPERICAL RESULTS
In this section, the descriptive statistics, the correlation coefficients, the results of the robust fixed effect models, the (robust) probit models and the GMM models are presented and discussed. Furthermore, answers will be provided to the stated hypotheses.
4.1. Descriptive statistics
Table 1 shows the descriptive statistics of all the dependent, explanatory and control variables used in this study. The sample contains 1441 firms located in the United States in the period 2010-2014, a period of 5 years, resulting in a total of 7205 firm observations.
Because some of the variables contain extreme values, I employed the winsorisation technique to remove these outliers. I removed the lowest 5 and the highest 5 percent of the sample to ensure that these results do not bias our estimates. As a result, standard deviations, skewness and kurtosis decreased for the variables to levels who are more in line with a normal distribution.
The statistics that are reported in Table 1 are relatively similar to statistics found in previous studies. The mean ROA is 6.2% is below the rate Veprauskaite & Adams (2013) reported but above the rate of Francoeur et al. (2017). The ROE of this sample is again lower than Veprauskaite & Adams (2013) who report a ROE of 15.1% and above the ROE (7%) of Lin (2005). The reported Tobin’s Q (1.346) is below the Tobin’s Q of Frijns et al. (2016) who reported 2.01 and the 1.56 reported in both Veprauskaite &Adams (2017) and Klein et al. (2005). However, the standard deviation in this study is 0.843, which is lower than that of prior studies.
20 4.2. Correlation coefficients
The correlation diagram in Table 2 shows the significant (p<0.01, two-tail) positive correlations between the three firm performance variables. The correlation of ROE and Tobin’s Q is lowest (0.355) and the correlation between ROA and Tobin’s Q the highest (0.557). All three correlations are relatively low, indicating that these variables capture different aspects of firm performance.
The same holds for the two non-financial performance indicators. The correlation between the two is significant (p<0.01, two-tail) and relatively low (0.409), indicating that both variables capture different non-financial performance aspects.
The correlation between the dependent (ROA, ROE, Tobin’s Q, CSR committee and CSR index) and the independent variables (CEO duality, CEO tenure, performance contract) showed both significant positive and negative correlation. CEO duality shows a significant positive relation with ROE while Tobin’s Q is negative related. The opposite holds for CEO tenure where ROE is negatively correlated and Tobin’s Q positive. Duality is both positively correlated with CSR committee and CSR index while the former has a negative correlation with tenure. The correlation between both decision-making autonomy variables is low (0.340) and significant at the 1% level. This indicates that both variables are not related and capture different aspects of CEO decision-making autonomy. Finally, all the dependent variables are negatively correlated with the presence of a performance contract, as expected. However, only ROE, CSR committee and CSR index correlation coefficients are significant.
4.1. Results Robust Fixed Effects regressions 4.1.1. Hypothesis 1A
As discussed in section 3.4.3, the Hausman test indicated that the fixed effects model is preferred above the random effects model in estimating the effect of CEO power on financial firm performance. The results for hypothesis 1 can be found in the “A” models of Table 3. In all three A-models, the power variables Duality and Tenure have positive coefficients. However, Duality is not significant in the ROA and ROE model. The Duality coefficient in the Tobin’s Q model is significant, at the 1% significance level. This indicates that the increase in the decision-making power of a CEO, proxied by CEO Chair-Duality increases the Tobin’s Q of the firm.
21 Table 1
Summary descriptive statistics
Variables Mean St. dev. Min Median Max Skewness Kurtosis Observations
Financial performance variables
ROA 0.062 0.059 -0.072 0.056 0.201 0.316 3.001 7007 ROE 0.104 0.142 -0.503 0.104 0.399 -0.770 6.024 6877 Tobin's Q 1.346 0.843 0.209 1.131 4.252 1.191 4.167 6511 Non-Financial performance variables CSR Committee 0.476 0.499 0 0 1 0.098 1.010 2860 CSR Sustainability Index 0.154 0.361 0 0 1 1.919 4.682 2860
CEO power variables
CEO-Chair Duality 0.522 0.500 0 1 1 -0.089 1.008 7205 CEO-Tenure 9.813 6.413 1 8 27 0.991 3.281 7205 CEO compensation Performance Contract 0.202 0.401 0 0 1 1.487 3.211 7205 Control variables Board structure Size 10.418 2.183 5 10 17 0.299 3.068 3511 Meeting 8.087 3.365 4 7 24 1.467 5.765 3486 Firm Size LnSales 14.360 1.698 4.605 14.255 20.001 0.067 3.998 7088 Leverage 0.720 0.837 0 0.458 3.695 1.665 5.338 7094 Industry
Basic industries & resources 0.110 0.313 0 0 1 2.487 7.187 795
Industrial goods & services 0.178 0.383 0 0 1 1.681 3.825 1085
Consumer goods 0.185 0.388 0 0 1 1.623 3.636 1125
Services 0.196 0.397 0 0 1 1.529 3.336 1195
Financials 0.184 0.388 0 0 1 1.631 3.659 1120
23 5% significance level and in the Tobin’s Q model at the 10% level. This indicates that also the second decision-making power proxy is positively related to firm performance.
This confirms our first hypothesis that CEO decision-making power has a positive influence on firm performance. The result is in line with the prediction of stewardship theory, which holds that a high degree of decision-making autonomy allows CEOs to effectively guide and control the firm, and therefore, has a potential positive effect on firms’ financial performance. The Board control variables are also in line with the predictions of stewardship theory. In all three models, the coefficients for both Size and Meeting are negative and significant at the 1% level, except in the ROE model where the Board Size is significant at the 10% level. The increase in the size and the number of meetings of the board are both indications of lower CEO decision-making autonomy because coordination and control from the board increases. The negative effects of both variables on the financial firm performance are therefore as expected. Hence, the decrease in decision-making autonomy lowers the CEOs ability to effectively guide and control the firm, potentially affecting the financial performance of the firm.
All tough the effects of CEO decision-making autonomy are not very strong, I can conclude that the results indicate a weak positive relationship between decision-making autonomy and financial firm performance. This is in line with stewardship theory and the null hypothesis of 1A, that there is no relation, can be rejected.
4.1.1. Hypothesis 2A
The B models of Table 3 include the interaction terms (Pcontract × Duality) and (Pcontract × Tenure). The intuition of the interaction variables is to show how much the effect of the power variables on the financial firm performance changes when CEOs are compensated based on their performance or not. The coefficients in models 1B and 3B are non-significant and show a positive Pcontract × Duality coefficient. The ROE coefficient is positive as well and significant at the 10% level. The Pcontract × Tenure coefficient is for all three models negative and only significant (p<0.1) in the ROE model.
24 Table 3
Financial Firm Performance Robust Fixed Effects Regression
(1A) (1B) (2A) (2B) (3A) (3B)
Variables ROA ROA ROE ROE Tobin’s Q Tobin’s Q
Duality 0.00205 0.00259 0.00206 0.00119 0.106*** 0.124*** (0.00193) (0.00207) (0.00494) (0.00541) (0.0298) (0.0323) Tenure 0.000337** 0.000365** 0.000236 0.000497 0.00485* 0.00583** (0.000162) (0.000172) (0.000403) (0.000436) (0.00258) (0.00277) Pcontract × Duality 0.00313 0.0180* 0.105 (0.00450) (0.0108) (0.0671) Pcontract × Tenure -0.000148 -0.00125* -0.00531 (0.000309) (0.000724) (0.00478) Board Structure Size -0.00270*** -0.00269*** -0.00221* -0.00214* -0.0707*** -0.0704*** (0.000476) (0.000476) (0.00117) (0.00116) (0.00735) (0.00736) Meeting -0.00318*** -0.00318*** -0.00597*** -0.00595*** -0.0435*** -0.0435*** (0.000288) (0.000290) (0.000730) (0.000730) (0.00372) (0.00374) Other Controls LnSales 0.00373*** 0.00373*** 0.0200*** 0.0199*** -0.0604*** -0.0607*** (0.000798) (0.000797) (0.00193) (0.00193) (0.0120) (0.0120) Leverage -0.0137*** -0.0137*** -0.00265 -0.00279 -0.126*** -0.126*** (0.00116) (0.00116) (0.00367) (0.00367) (0.0178) (0.0178) Constant 0.0772*** 0.0772*** -0.157*** -0.156*** 3.417*** 3.417*** (0.0113) (0.0112) (0.0282) (0.0282) (0.168) (0.168)
Industry Fixed Effects YES YES YES YES YES YES
Year Fixed Effects YES YES YES YES YES YES
Observations 3,447 3,447 3,380 3,380 3,290 3,290
R-squared 0.113 0.113 0.088 0.088 0.155 0.156
Number of firms 713 713 704 704 678 678
This table reports the financial firm performance regression results for the period 2010-2014. Duality and Tenure are the CEO power variables. Pcontract × Duality and Pcontract × Tenure are the interaction variables. All the remaining variables are defined in Appendix 1.All six models are fixed effects models using robust standard errors to account for the presence of heteroskedasticity. The standard errors are reported in parentheses.
25 Table 4
Non-Financial Firm Performance Probit Regression
(4A) (4B) (5A) (5B)
Variables CSR Committee CSR Committee CSR Index CSR Index
Duality 0.173*** 0.206*** 0.0124 0.0580 (0.0551) (0.0592) (0.0676) (0.0711) Tenure -0.0176*** -0.0168*** -0.0167*** -0.0151** (0.00474) (0.00505) (0.00586) (0.00619) Pcontract × Duality -0.248* -0.453** (0.132) (0.184) Pcontract × Tenure -0.00126 -0.00938 (0.00891) (0.0125) Board Structure Size 0.0913*** 0.0919*** 0.0834*** 0.0871*** (0.0140) (0.0140) (0.0161) (0.0163) Meeting 0.0127 0.0138* 0.000686 0.00350 (0.00802) (0.00806) (0.00971) (0.00978) Other Controls LnSales 0.464*** 0.462*** 0.367*** 0.373*** (0.0254) (0.0254) (0.0250) (0.0252) Leverage 0.147*** 0.146*** -0.0693* -0.0717* (0.0317) (0.0318) (0.0394) (0.0396) Constant -8.254*** -8.244*** -7.502*** -7.646*** (0.381) (0.382) (0.379) (0.386)
Industry Fixed Effects YES YES YES YES
Year Fixed Effects YES YES YES YES
Observations 2,812 2,812 2,812 2,812
26 Table 5
Financial Firm Performance GMM Panel Regression
(6A) (6B) (7A) (7B) (8A) (8B)
Variables ROA ROA ROE ROE Tobin’s Q Tobin’s Q
Dep. Var. (t-1) 0.625*** 0.596*** 0.738*** 0.627*** 0.924*** 0.922*** (0.0810) (0.0751) (0.133) (0.112) (0.0584) (0.0534) Duality 0.00566 0.00607 0.00417 0.00905 0.239** 0.188* (0.00930) (0.00887) (0.0268) (0.0251) (0.0999) (0.101) Tenure 0.000363 2.30e-05 0.00395* 0.00331* 0.0147** 0.0104 (0.000852) (0.000794) (0.00207) (0.00200) (0.00703) (0.00711) Pcontract × Duality 0.00260 -0.0517 0.0412 (0.0253) (0.0625) (0.246) Pcontract × Tenure 0.000184 0.00356 -0.00703 (0.00111) (0.00335) (0.0114) Board Structure Size -0.00110 -0.00226 -0.00116 0.00103 -0.0197 -0.0207 (0.00319) (0.00303) (0.0102) (0.00941) (0.0310) (0.0305) Meeting 6.48e-05 -0.000729 -0.0114** -0.0109** -0.0436** -0.0527*** (0.00174) (0.00163) (0.00477) (0.00455) (0.0207) (0.0203) Other Controls LnSales -0.00637 -0.000870 0.00828 0.00469 0.0144 0.0113 (0.00748) (0.00639) (0.0205) (0.0186) (0.0714) (0.0609) Leverage 0.00266 0.000993 -0.0277* -0.0301** -0.0470 -0.0499 (0.00479) (0.00432) (0.0164) (0.0149) (0.0513) (0.0468)
Industry Fixed Effects YES YES YES YES YES YES
Year Fixed Effects YES YES YES YES YES YES
Constant 0.135 0 0.0129 0 0.772 0.921 (0.101) (0) (0.261) (0) (1.107) (0.956) Hansen 0.148 0.298 0.545 0.593 0.000 0.000 Diff-Hansen 0.256 0.224 0.134 0.009 0.000 0.000 AR(1) 0.000 0.000 0.001 0.000 0.000 0.000 AR(2) 0.367 0.398 0.205 0.245 0.020 0.018 Instruments 46 58 46 58 46 58 Observations 2,684 2,684 2,739 2,739 2,620 2,620 Number of firm 699 699 712 712 675 675
This table reports the dynamic panel regression results for the period 2010-2014. Dep. Var. (t-1) is the lagged dependent variable (ROE, ROA, and Tobin’s Q). Duality and Tenure are the CEO power variables. Pcontract × Duality and Pcontract × Tenure are the interaction variables. All the remaining variables are defined in Appendix 1. AR(1) and AR(2) are the Arellano-Bond tests for the first and second-order autocorrelated
27 Because the variables have different signs and only two of the six are significant at the 10% level, I cannot draw any conclusions from the results. Hypothesis 1B cannot be confirmed nor rejected.
To ensure the validity of the A and B models in table 3, I also run the regression models excluding firms categorized in the financial sector. The results can be found in Appendix 5 and are consistent with the results of table 3.
4.2. Results Probit regressions 4.2.1. Hypothesis 1B
Important points can be discerned from Table 4 that contains the results of the probit regression analyses.
The Duality coefficients in both the CSR committee and the CSR Index models are positive and only the former is statistically significant (p<0.01). Tenure is for all four models significant (p<0.05) and negatively related to the non-financial performance indicators. This is not in line with the hypothesized sign and indicates that the Tenure proxy for our CEO decision-making power has a negative relation with CSR, the proxy for non-financial performance. This suggests that the longer the CEO stays in power, the lower the probability of a CSR committee or being listed on a CSR Index. This is not in line with stewardship theory that predicts a higher probability of CSR committees or being listed on an Index when the tenure of the CEO increases.
On the other hand, the dual position as CEO and Chair of the board has a positive effect on non-financial performance. The engagement in social responsible activities increases when the CEO is also Chair of the board; exactly what stewardship theory predicts.
It is hard to infer conclusions from the results presented in the A-models of Table 4 due to the different outcomes. On the one hand, when the CEO is also Chair of the board, it improves the non-financial performance. On the other hand, non-financial performance will be lower when the tenure of the executive officer increases. Due to the result differences, I can neither confirm nor deny hypothesis 1B.
4.2.2. Hypothesis 2B
28 influence the probability of the presence of a CSR committee and if its listed on a specific sustainability index. This is in line with the predictions of stewardship theory.
The explanation behind this relation is that performance based compensation is an extrinsic motivational factor and therefore a waste of resources because CEOs are intrinsically motivated to achieve firm goals. There is no need to align the interests of the principals and the CEO; the goals are already aligned. Moreover, financial incentives might distract the CEO from its initial objectives and as a result decrease performance.
To conclude, Stewardship theory states that CEOs receiving performance based compensation will potentially decrease the presence of CSR committees and CSR reporting. The absence of CSR committees or firms not mentioned on CSR indexes indicates lower non-financial performance. CEOs receiving performance based compensation lower the probability of CSR committees and CSR Indexes reporting and lower as a result the non-financial performance. I can reject the null hypothesis of hypothesis 2B that there is no moderating relation between the format of compensation (performance based compensation or not) and the relation with non-financial performance. The results indicate that performance based compensation lowers the non-financial firm performance. I can therefore reject the null hypothesis that there is no moderating effect of performance based compensation on non-financial firm performance. To ensure the validity of the A and B probit models in table 4, I also run the regression models excluding firms categorized in the financial sector. The results can be found in Appendix 4 and are consistent with the results of table 4.
4.3. Results Generalized methods of moments regressions
Due to the possibility of endogeneity within the financial firm performance models, GMM regressions are used. Table 5 shows the results.
The coefficients of the lagged dependent variables, ROE, ROA and Tobin’s Q are all statistically significant at the 1% level. The results therefore not reject the dynamic nature of CEO power and the financial performance models.
The Duality coefficient is in all three models positive but only significant at the 5% level in the Tobin’s Q model (8A). In the other two models is Duality not significant.
29 When adding the interaction variables in the B-models, the coefficients for Duality and Tenure maintain positive in all three models but are mostly insignificant. Only Duality in the Tobin’s Q model (8B) and Tenure in the ROE model (7B) are significant (p<0.1).
The interaction terms are in none of the models significant but almost all have the expected negative sign, predicted by stewardship theory. Only the Duality interaction is positive in the Tobin’s Q model (8B).
The xtabond2 specification that is used to perform the GMM analyses in Stata automatically includes the Hansen and the difference-in-Hansen test for over identifying restrictions and the validity of instruments, respectively. This is because GMM assumes instruments to be exogenous in order to be valid. Furthermore, the Arellano-Bond test for testing autocorrelation is reported.
In both ROE and ROA model, the Hansen test is not significant, indicating that there is no evidence to reject the null hypothesis that the model is over identified. The difference-in-Hansen statistic is also not significant, except for the ROE model at the 10% level, indicating that there is no evidence to reject the null hypothesis that the instruments or not valid. I conclude there is no evidence of exogeneity in the ROE and ROA models. In the Tobin’s Q model exogeneity is expected due to the significant Hansen and Difference-in-Hansen test. However, the Hansen test is prone to weaknesses and should not be relied upon to faithfully (Roodman, 2009).
As a result, the generalized methods of moments regressions give no helpful insights in the relation between the decision-making autonomy of CEOs and the financial firm performance. This is due to the insignificance of the coefficients in the ROA and ROE models and the possibility of exogeneity in the Tobin’s Q model. The same holds for the results of the added interaction variables.
5. LIMITATIONS AND FUTURE RESEARCH
30 arrange a license which grants access to a wider variety of databases. This improves the amount of data gathered and the number of variables that can be used.
Secondly, it can be interpreted that information between stewards and principals is symmetric (Madison et al., 2016). However, a limitation of this study is the unobserved information symmetry between steward and principal. I was not able to gather information about future goals of principles were known by the steward. Actions of stewards depend on this
information and may impact firm performance. Future research may investigate the goal and information alignment between CEO and principal and if CEOs and principals with aligned interest have relatively higher firm performance.
Thirdly, non-financial performance can cover many different aspects. CSR committee or CSR index are therefore variables that cannot cover all non-financial aspects of firms. The
assumption that both variables cover all non-financial firm performance is therefore a
limitation but was a decision that needed to be taken due to the scarce availability of data. For the future, it would be interesting to investigate for example employment turnover and
productivity. Fourthly, the sample only covers firms located in the US. Laws, regulations, business climate and many more influential factors differ from country to country. The generalizability of the results is therefore limited. To improve the generalizability of future research, it would be interesting to investigate country or regional differences.
6. CONCLUSION
31 The results reveal that even when controlling for board and firm characteristics, two CEO decision-making autonomy variables, Duality and Tenure, are positive related to financial firm performance. This is in line with stewardship expectations and confirms Hypothesis 1A. Furthermore, the results indicate a moderating effect of performance based compensation on the relation between CEO decision-making autonomy and non-financial firm performance (Hypothesis 2B). Autonomous CEOs receiving performance based compensation will lower the non-financial performance of the firm. Results for the moderating effect on financial performance were contradictory (Hypothesis 2A).
The results of the relation between autonomy and the non-financial performance were contradictory as well. Duality, one of the two autonomy variables, is positively related with non-financial performance while the other autonomy variable, Tenure, is negatively related to non-financial performance. Hypothesis 1B and 2A can therefore be neither confirmed nor denied.
32
7. APPENDIX
Appendix 2. Heteroskedasticity test results
Model Chi-square Degrees of freedom Probability value
ROA 21.37 1 0.000
ROE 102.06 1 0.000
Tobin’s Q 164.05 1 0.000
CSR Committee 0.99 1 0.321
CSR Sustainability Index 665.18 1 0.000
Appendix 1. List of Variables
Variable Source Definition
ROA Datastream Earnings before interest expenses and taxes divided by the average of two most recent year of book value of total assets
ROE Datastream Earnings after tax, minority interest and preferred
dividends for the most recent tax year divided by previous year's book value of common equity
Tobin’s Q Datastream Book value of total assets minus the book value of common equity plus the market value of common equity divided by the book value of total assets
CSR Committee Datastream A dummy variable that is coded as 1 if the firm has a corporate social responsibility committee(CSR) and 0 otherwise
CSR Index Datastream A dummy variable that is coded as 1 if the the firm report on belonging to a specific sustainability index and 0 otherwise
CEO Duality WRDS A dummy variable that is coded as 1 if the CEO also serves as board Chair and 0 otherwise
CEO Tenure WRDS The number of years a CEO has been as CEO
Perf Contract WRDS A dummy variable that is coded as 1 if the CEO receives performance based compensation and 0 otherwise Board Size Datastream The total number of inside and outside directors on the
board
Board meetings Datastream The total number of board meeting held during the year LnSales Datastream Natural logarithm of the value of Sales
33 The null hypothesis of the Breusch-Pagan test for heteroskedasticity is that the variance of the error terms is constant. I reject the null hypothesis that there is no heteroskedasticity in four of the five models. Only the CSR committee model is not significant.
Appendix 3. Hausman test results for fixed and random effects model
Model Chi-square Degrees of freedom Probability value
ROA 41.27 10 0.000
ROE 39.74 10 0.000
Tobin’s Q 96.99 10 0.000
Under the null hypothesis of the Hausman test, the random effects (RE) model is preferred while under the alternative hypothesis the fixed effects (FE) model is preferred. Because the null hypothesis is rejected, the fixed effects model is used.
Appendix 4. Non-Financial Firm Performance Probit Regression (excluding financials)
(4A) (4B) (5A) (5B)
Variables CSR Committee CSR Committee CSR Index CSR Index
Duality 0.0946** 0.113** -0.0370 -0.000774 (0.0612) (0.0655) (0.0768) (0.0807) Tenure -0.0138*** -0.0106* -0.0248*** -0.0212*** (0.00524) (0.00562) (0.00657) (0.00682) Pcontract × Duality -0.179 -0.442* (0.151) (0.230) Pcontract × Tenure -0.0156 -0.0276 (0.0100) (0.0179) Board Structure Size 0.108*** 0.111*** 0.108*** 0.115*** (0.0158) (0.0157) (0.0181) (0.0185) Meeting 0.00950 0.0110 0.00513 0.00902 (0.00909) (0.00915) (0.0111) (0.0112) Other Controls LnSales 0.433*** 0.429*** 0.318*** 0.323*** (0.0283) (0.0285) (0.0283) (0.0286) Leverage 0.110*** 0.106*** -0.0801* -0.0864** (0.0348) (0.0349) (0.0433) (0.0437) Constant -7.984*** -7.971*** -7.009*** -7.190*** (0.421) (0.425) (0.416) (0.425)
Industry Fixed Effects YES YES YES YES
Year Fixed Effects YES YES YES YES
Observations 2,229 2,229 2,229 2,229
This table reports the non-financial firm performance results for the period 2010-2014. Duality and Tenure are the CEO power variables. Pcontract × Duality and Pcontract × Tenure are the
34 Appendix 5. Financial Firm Performance Robust Fixed Effects Regression (excluding financials)
(1A) (1B) (2A) (2B) (3A) (3B)
Variables ROA ROA ROE ROE Tobin’s Q Tobin’s Q
Duality 0.00540* 0.00633** 0.0124 0.0164* 0.0358 0.0239 (0.00288) (0.00312) (0.00867) (0.00930) (0.0346) (0.0345) Tenure 0.000384 0.000286 0.00234** 0.00220** 0.0151*** 0.0133*** (0.000330) (0.000333) (0.000943) (0.00104) (0.00500) (0.00425) Pcontract × Duality 0.00488 0.0189 0.0644 (0.00552) (0.0121) (0.0665) Pcontract × Tenure -0.000433 -0.000387 -0.00800 (0.000492) (0.00111) (0.00588) Board Structure Size -0.000710 -0.000724 -0.00518** -0.00517** -0.0258*** -0.0251*** (0.000820) (0.000817) (0.00236) (0.00236) (0.00827) (0.00829) Meeting -0.00108*** -0.00109*** -0.00106 -0.00110 -0.00312 -0.00275 (0.000344) (0.000343) (0.000936) (0.000932) (0.00352) (0.00347) Other Controls LnSales 0.00793 0.00791 0.0705*** 0.0701*** 0.244*** 0.241*** (0.00535) (0.00538) (0.0157) (0.0157) (0.0652) (0.0646) Leverage -0.00941*** -0.00937*** -0.0253*** -0.0250*** 0.0426* 0.0415 (0.00236) (0.00236) (0.00951) (0.00952) (0.0256) (0.0258) Constant -0.0355 -0.0346 -0.862*** -0.859*** -2.014** -1.975** (0.0821) (0.0826) (0.241) (0.241) (0.996) (0.989)
Industry Fixed Effects YES YES YES YES YES YES
Year Fixed Effects YES YES YES YES YES YES
Observations 2,698 2,698 2,647 2,647 2,592 2,592
R-squared 0.029 0.030 0.040 0.041 0.044 0.046
Number of firms 557 557 550 550 530 530
This table reports the financial firm performance regression results for the period 2010-2014 excluding financial firms. Duality and Tenure are the CEO power variables. Pcontract × Duality and Pcontract × Tenure are the interaction variables. All the remaining variables are defined in Appendix 1.All six models are fixed effects models using robust standard errors to account for the presence of heteroskedasticity. The standard
35
8. REFERENCES
Adams, R. B., Almeida, H., & Ferreira, D. 2005. Powerful CEOs and their impact on corporate performance. Review of Financial Studies, 18(4): 1403–1432.
Anderson, D. W., Melanson, S. J., & Maly, J. 2007. The evolution of corporate governance: Power redistribution brings boards to life. Corporate Governance, 15(5): 780–797. Andreoni, J. 1989. Giving with Impure Altruism: Applications to Charity and Ricardian
Equivalence. The Journal of Political Economy, 97(6): 1447–1458.
Arellano, M., & Bond, S. 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies, 58(2): 277.
Argyris, C. 1973. Some Limits of Rational Man Organizational Theory. Public Administration Review, 33(3): 253–267.
Bhagat, S., & Bolton, B. 2008. Corporate Governance and Firm Performance. Journal of Corporate Finance, 14: 257–273.
Block, Peter 1996, Stewardship: Choosing Service Over Self-Interest. San Francisco: Berrett-Koehler Publishers.
Bosse, D. A., & Phillips, R. A. 2016. Agency theory and bounded self-interest. Academy of Management Review, 41(2): 276–297.
Brookman, J., & Thistle, P. D. 2009. CEO tenure, the risk of termination and firm value. Journal of Corporate Finance, 15: 331–344.
Combs, J. G., Ketchen, D. J., Perryman, A. A., & Donahue, M. S. 2007. The moderating effect of CEO power on the board composition-firm performance relationship. Journal of Management Studies, 44(8): 1299–1323.
Conyon, M. J. 2014. Executive compensation and board governance in US firms. Economic Journal, 124(574): 60–90.
Davis, J. H., Schoorman, F. D., & Donaldson, L. 1997. Toward a stewardship theory of management. Academy of Management Review, 22(1): 20–47.
Dedman, E. 2016. CEO succession in the UK: An analysis of the effect of censuring the CEO-to-chair move in the Combined Code on Corporate Governance 2003. British Accounting Review, 48(3): 359–378.
36 Driscoll, C., & Starik, M. 2011. For the Natural Environment Primordial Stakeholder: the
Conceptual Advancing Consideration of Stakeholder Status for the Natural Environment The Journal of Business, 49(1): 55–73.
Eberhardt-Toth, E. 2017. Who should be on a board corporate social responsibility committee? Journal of Cleaner Production, 140: 1926–1935.
Francoeur, C., Melis, A., Gaia, S., & Aresu, S. 2017. Green or Greed? An Alternative Look at CEO Compensation and Corporate Environmental Commitment. Journal of Business Ethics, 140(3): 439–453
Frijns, B., Dodd, O., & Cimerova, H. 2016. The impact of cultural diversity in corporate boards on firm performance. Journal of Corporate Finance, 41: 521–541.
Garner, J. L., & Harrison, T. D. 2013. Boards, executive excess compensation, and shared power: Evidence from non-profit firms. Financial Review, 48(4): 617–643.
Godos-Díez, J.-L., Fernández-Gago, R., & Martínez-Campillo, A. 2011. How Important are CEOs to CSR Practices? An Analyses of the Mediating Effect of the Perceived Role of Ethics and Social responsibility. Journal of Business Ethics, 98(4): 531–548.
Guest, P. M. 2008. The determinants of board size and composition: Evidence from the UK. Journal of Corporate Finance, 14(1): 51–72.
Hagel, J., Brown, J.S. & Davison, L. 2010. The Best Way to Measure Company Performance. Harvard Business Review, https://hbr.org/2010/03/the-best-way-to-measure-compan, March 4 2010.
He, L., Cordeiro, J. J., & Shaw, T. S. 2015. CEO power, equity ownership and underwriter reputation as determinants of lockup period length. Management Research Review, 38(5): 438–458.
Hendijani, R., Bischak, D. P., Arvai, J., & Dugar, S. 2016. Intrinsic motivation, external reward, and their effect on overall motivation and performance. Human Performance, 29(4): 251–274.
Jian, M., & Lee, K.-W. 2015. CEO compensation and corporate social responsibility. Journal of Multinational Financial Management, 29: 46–65.
Klein, P., Shapiro, D., & Young, J. 2005. Corporate governance, family ownership and firm value: the Canadian evidence. Corporate Governance: An International Review, 13(6): 769–784.
Lammers, J., Stoker, J. I., Rink, F., & Galinsky, A. D. 2016. To have control over or to be free from others? The desire for power reflects a need for autonomy. Personality and Social Psychology Bulletin, 42(4): 498–512.
