Amsterdam Business School
The effect of gender diversity in higher management on firm
performance
Is there a difference between sectors?
Name: Micha van der Veen Student number: 10986766
Thesis supervisor: ir. drs. A.C.M. de Bakker Date: 23 June 2017
Word count: 10346
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Micha Nicolaas van der Veen 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.
Abstract
This study examines the relationship between female participation and financial performance for companies in North America. Figures show that women are underrepresented in higher management positions. Previous studies suggest that more female participation in top management positions positively influences firms’ financial performance. Or, for some studies, at least a positive trend is expected. This study specifically examines whether more women in higher management positions contributes to a higher financial performance within high technology & life science companies. These high technology & life science companies attempt to be progressive in the broadest sense of participation. Previous research and figures show that also these companies lagging behind if it concerns women in higher management positions. This research strives to contribute to awareness of more gender diversity within higher management positions at both high technology & life science and other companies.
The research is performed by means of three hypotheses. The hypotheses are converted into regressions, which are tested with the statistical program STATA. In STATA both fixed and random effects are used as the regression technique. Hypothesis 1 focusses on whether the percentage of women present in higher management is positively associated with firms’ financial performance. Results show that the first hypothesis cannot be rejected, meaning that the percentage of women in higher management has a positive effect on firms’ financial performance. Hypothesis 2 states that the percentage of female managers within high technology & life science companies is lower than in other companies. The outcome of the regression indicates that hypothesis 2 cannot be rejected. Hence, the percentage of female managers within high technology & life science companies is lower than for other companies. Hypothesis 3 is concentrated on whether the relationship between the percentage of women and financial performance is more positive in high technology & life science companies than in other companies. Regression results illustrate no significant effect for this relationship. Statistically it cannot be concluded that the financial performance is more positive in high technology & life science companies when the percentage of women in management is higher. Hence, hypotheses 3 can be rejected.
One can conclude from this research that more women in higher management positions will positively influence firms’ financial performance. This effect however is not stronger for high technology & life science companies.
Table of contents
1 Introduction ... 6
1.1 Motivation ... 6
1.2 Research question ... 7
2 Theoretical framework ... 9
2.1 Agency theory and economic model assumptions ... 9
2.2 Female participation... 11
2.3 Behavioral difference in management style and level of risk ... 13
2.4 Financial performance ... 14 2.5 Hypotheses development ... 15 3 Methodology ... 16 3.1 Conceptual model ... 16 3.2 Testing method ... 16 3.3 Empirical model ... 16 3.4 Regression models ... 17
3.4.1 Regression model hypothesis 1 ... 17
3.4.2 Regression model hypothesis 2 ... 17
3.4.3 Regression model hypothesis 3 ... 18
4 Data ... 20
4.1 General description of the panel database... 20
4.2 Description of actual data set used ... 20
4.2.1 Needed data ... 21
4.2.2 Specific calculations ... 21
4.2.3 Composing the dataset ... 22
4.3 Descriptive statistics ... 25
4.3.1 Core statistics ... 25
4.3.3 Multicollinearity analysis ... 26
5 Results... 27
5.1 Regression results ... 27
5.1.1 Evaluation of results hypothesis 1 ... 28
5.1.2 Evaluation of results hypothesis 2 ... 29
5.1.3 Evaluation of results hypothesis 3 ... 29
5.2 Normality of error term ... 30
6 Conclusion and recommendations ... 31
References ... 34
1 Introduction
This research investigates whether gender diversity on higher management level affects the financial performance of US companies. In particular the difference between ‘general’ and high technology & life sciences companies will be investigated. The introduction starts with a motivation that subsequently results in the research question.
1.1 Motivation
Despite many efforts by governments to encourage more gender diversity in leadership, there are still too few women in higher management of publicly traded companies. Most executives believe that gender diversity in leadership is correlated to better financial performance1. What is the reason
that fewer women than men are holding a higher management position within publicly traded companies?
Yet there is a rising trend at the number of women studying and the Knowledge Center ‘Catalyst’ learns us that women earn more degrees than men. The rising trend of women studying has been known for a longer period and statistics of the NCES (National Center for Education Statistics) show that the gap between women and men studying at universities is widening. Therefore one might expect that more higher management positions will be held by women or at least a positive trend will be anticipated.
The US population consists of 50.8 percent of women (2015, August) and therefore they make up the majority. Women earn around 60 percent of undergraduate degrees and around 60 percent of master’s degrees. Despite the fact that they hold almost 52 percent of the professional job levels, women in North America are essentially behind men when it comes to their representation in higher management positions. At S&P 500 companies, the labor force consists out of 45% women. Only 25% women hold executive- and senior management positions and only 19% is present in the board. At CEO-level it is even more dramatic because only 4.6% of women holds such a position2.
Bell and White (2014) investigated gender diversity in Silicon Valley. Most companies at Silicon Valley are high technology & life sciences companies. These companies are known to be progressive in the field of technology, attracting good and educated staff from all over the world,
1 Global Survey McKinsey 2010. This survey can be found on the website: http://www.mckinsey.com/business-functions/organization/our-insights/moving-women-to-the-top-mckinsey-global-survey-results
2 Center for American Progress. The data and information can be found on the website: https://www.americanprogress.org/issues/women/reports/2015/08/04/118743/the-womens-leadership-gap/
and focused on diversity in general (e.g. race, gender, origin). The latter might indicate that these advanced companies have a large proportion of women at higher management level. The researchers collected data from 100 Standard & Poor’s companies (S&P 100 companies) and Silicon Valley top 150 companies (SV 150 companies). S&P 100 companies have relative more female directors than SV 150 companies (S&P 100 average = 21.0% of directors and SV 150 average = 10.2% of directors in the 2014 proxy season). What can be seen in the data is that high technology & life science companies at Silicon Valley are doing even worse in terms of gender diversity3.
Following Westphal and Milton (2000) women often bring a different perspective on complex questions. They say that this can benefit proper informational biases in strategy formulation and problem solving. Hence, this supports the idea that more women functioning on a higher management level can enhance the business and therefore the firm performance.
This study contributes to the promotion of gender diversity in leadership by exploring the difference of value creation for shareholders by male managers and female managers. In addition, this study investigates the impact of more gender diversity in leadership at high technology & life science companies.
1.2 Research question
Firms’ financial performance is depending on many factors. The fact is that a firm’s financial performance is unknown in advance. Companies have influence on some of these factors such as the choice of managers and whether this manager is a man or a woman. An interesting factor is the view on shareholders versus personal gain of a manager.
Everyone, also managers, are looking for personal gain. Therefore it is not realistic to believe that managers (both male and female) only think about the shareholders of the company they work for or which they are leading. Interesting is to look at the differences between female and male management board in this occurrence. Hence, the effect of gender diversity in a company’s performance can be analyzed on different levels. The aim of this study is to examine the relationship between gender of the higher management layer and firm’s financial performance. In addition, a distinction in type of firms will be investigated. Therefore the main questions this research seeks to answer are:
3 Note that the top 15 of the SV 150 companies are closer to the S&P 100 companies concerning gender diversity in (higher) management.
• What is the effect of female participation on higher management level on the company’s financial performance?
• Is it depending on the type of business whether gender diversity contributes to the financial performance of a company?
This research contributes to the current literature as data from US (North America) Companies is used, and in addition, as the differences between high technology & life science companies and “other companies”4 are analyzed. Taking it into the broader scope of firms in the
US, instead of only high technology & life science companies in Silicon Valley.
The remainder of this thesis is organized as follows. In Section 2, the theoretical framework is given, where various results of similar studies and other relevant literature are presented. With the relevant literature, the research question is transformed to three hypotheses. The hypotheses development is described in Section 2.5. In Section 3, the research methodology is explained. In the 4th Section the data is presented and details are explained. In Section 5 the results will be given.
Finally, in Section 6 the conclusions and recommendations are discussed. The last sections provide a reference and appendices overview.
4 Other companies in this occurrence means, companies other than high technology & life science companies. Financial companies have also been taken out of the data. In section 4 more details about the data is given.
2 Theoretical framework
In this chapter underlying theories of the research will be presented. In addition definitions and general terms are described. The literature of agency theory and economic model assumptions is described in Section 2.1. In Section 2.2 the literature of female participation is represented. In Section 2.3 literature about behavioral difference in management style and level of risk is given. Section 2.4 describes the literature about financial performance. Finally the hypothesis development is represented in Section 2.5.
2.1 Agency theory and economic model assumptions
Agency theory is indispensable in the perspective of research on companies’ financial performance. Managers of publicly traded companies are the ones in charge to create value for the shareholders of those publicly traded companies. As stated in the Introduction (see Section 1.2), managers are looking for personal gain. This aspect can return into problems and the theory behind this tendency is called agency theory.
Eisenhardt (1989) describes agency problem as something that occurs when cooperating parties have different goals and division of labor. She states that agency theory specifically is directed at the pervasive agency relationship, where a party (the principal) delegates work to another party (the agent), who performs the work. In the agency theory a metaphor of a contract is used to try to describe this relationship (Jenssen & Meckling, 1976). In agency relationships, problems can occur and agency theory is about resolving two problems that can occur. The first problem that can occur is described by Eisenhardt (1989) as: ‘the desires or goals of the principal and agent conflict’. The second one she describes to occur is that: ‘it is difficult or expensive for the principal to verify what the agent is actually doing’. Therefore it cannot be checked if the agent behaved like he or she is designated to behave. When the principle and agent have different thoughts about risk, the problem of risk sharing can also occur. Eisenhardt (1989) suggests that the principle and agent have different thoughts about the actions to take concerning risk. Hence, the focus of the agency theory determines the best efficient contract possible for the principle-agent relationship. Table 1 provides an overview about agency theory (Eisenhardt, 1989).
Key idea Principle-agent relationships should reflect efficient organization of information and risk-bearing costs
Unit of analysis Contract between principal and agent
Human assumptions Self-interest
Bounded rationality Risk aversion
Organizational assumptions
Partial goal conflict among participants Efficiency as the effectiveness criterion
Information asymmetry between principal and agent
Information assumption Information as a purchasable commodity
Contracting problems Agency (moral hazard and adverse selection) Risk sharing
Problem domain Relationships in which the principal and agent have partly differing goals and risk preferences (e.g., compensation, regulation, leadership, impression management, whistle-blowing, vertical integration, transfer pricing)
Table 1: Agency Theory Overview (Eisenhardt, 1989)
The first reasons for agency problems, generally speaking, is the fact that people prefer more over less (greed) and that people avoid risk (risk aversion). It is therefore not illogical that an old saying on Wall Street is that markets are driven by just two emotions: fear and greed5. Another
assumption for agency problems refers to information asymmetry (Auronen, 2003). Information asymmetry can be divided in different asymmetric information models like screening, adverse selection, signaling and moral hazard. In this study, moral hazard is highlighted because it is most applicable in the relation of this research. Moral hazard refers to the misalignment of interests between principal (company) and agent (manager) because agent may prefer leisure activities over productive effort. In other words, as stated earlier, managers are looking for personal gain.
Most larger publicly held firms have separation of ownership and control. Based on the paper Jensen & Meckling (1976) separation of ownership and control introduces agency problems. With the delegation of decision rights from the shareholders in the direction of the higher management board, agency costs of equity arise. Agency costs of equity are distinct to be the sum of monitoring costs of the principal, bonding costs by the agent and residual loss. The reason is a divergence of interests between CEO and/or higher management board and outside investors (shareholders). Managers are not personally effected (or only a fraction if they own shares or equity) with costs that they take out to maximize their own utility by consuming perks. The more shares or equity the agent (management board) possess, the less he or she is motivated to consume these perks because a larger part of their wealth is tied to, for example, the performance measure stock price.
In summary of the above the following can be said; people, and thus managers, prefer more over less and are risk averse. Employees, and thus managers, have different interests compared to their investors (shareholders). But, what is the difference between male and female managers in this occurrence? Behavioral differences and management styles between men and women will be described in Section 2.3. Before looking into these differences, the literature about female participation is presented. Female participation is an important part of the formulation of the research question. More women than men are part of the American population, more women than men have (university) degrees and many more examples exist. Therefore it is unexpected that women are less part of higher management positions. Apart from the fact that this is unexpected, it is also interesting to get more information and feeling about the figures and other relevant information about female participation.
2.2 Female participation
The number of female managers rises. It is a logical effect of the growing number of women participating in the workforce. Despite the rise and in relation to the total workforce, there are still too few women in higher management. A study about gender diversity, carried out by McKinsey (a global consultancy firm), shows that most executives believes that gender diversity in leadership can result in a better financial performance for companies. Other studies, described in this section, support the idea that more gender diversity in leadership is linked to a better financial performance. Therefore, it is strange that the higher management board of companies still consists largely out of men.
Francoeur et al. (2008) investigated whether the participation of women in the firm’s board of directors positively influences the financial performance. They used data from the 500 largest
Canadian firms. Unlike other researchers they used the Fama and French (1992, 1993) model to account for the level of risk of the firm. The Fama and French (1992, 1993) is a valuation framework to take the level of risk into consideration when comparing firm performances. The outcome of the research is that a higher proportion of women officers (not board directors) generate positive and significant better returns for the firm. Hence, this research shows the support to current policies to encourage enhancement of women in business. Their results also indicate that firms operating in complex environments do generate positive and significant abnormal returns when they have a high proportion of women officers. High technology & life science companies do operate in complex environments and these companies, based on the results of Francoeur et al. (2008), should therefore hire more female officers.
Shrader et al. (1997) state in their study that women bring skills such as teambuilding and employee development to companies that are very much in tune with today’s competitive realities. In their study they assume that companies that hire more female managers do better in recruiting a capable management layer from the available talent pool. And therefore, they are in a better position to connect with customers, employees and other populations. Further they refer to a study that consisted of an interview on managers of North American firms, performed by Jelinek and Adler (1988). They found that women are very successful at developing interpersonal relations and cooperative alliances with their foreign counterparts. Eventually Shrader et al. (1997) do not find that more women in higher management is associated with a higher firm performance. The reasons they give are that women get assignments that have less instrumental impact on the firm and that women are not present enough in higher management positions to have much of an impact on the firm or its financial performance.
Campbell and Minquez-Vera (2008) performed the research of gender diversity and firm’s financial performance on data of Spanish companies. They used panel data analysis where the Tobin’s Q6 has been used as dependent variable and the measures for women’s participation in
the board as independent variable. The test results show that women on the board of directors have no direct effect on firm’s value, however gender diversity within companies shows to have a positive result on firm value. Despite the fact that Campell and Minquez-Vera (2008) did not find a significant effect for financial performance of gender diversity on board level, they did find an effect that triggers the importance of female participation within companies.
As stated in the introduction (see Section 1.1), Bell and White (2014) investigated gender diversity in Silicon Valley. They focus on high technology & life sciences companies. They found that several factors contribute to the under participation of women at high technology & life sciences companies. The first reason is because there is a relative lack of gender diversity at most senior levels of leadership in public companies. Second, companies often reflect conditions that existed and individual decisions that were made dozens of years ago. Therefore it is interesting to investigate whether high technology & life science companies have a relatively higher financial performance than non- high technology & life science companies.
To investigate this properly, several control variables must be taken into account. In section 3 an underpinning of the final control variables is given. In the following section the behavioral differences in management styles and risk taking between men and women is described.
2.3 Behavioral difference in management style and level of risk
In the perspective of female participation it is interesting to examine the behavioral differences in management styles and risk taking between men and women.
Company X takes more risk, or is in the opportunity to take more risk, than company Y. A company has the opportunity to choose its own management board. When considering a particular management board, companies will choose management that fits best with their strategy. But still, there is a difference between men and women in management style and thus in risk taking. Choosing a man or a woman will influence the way the company is managed.
Various management styles exist. Limbare (2012) uses eight management styles in his research. The management styles used are: deserter leadership style, missionary style, autocrat leadership style, compromiser style, bureaucrat style, developer leadership style, benevolent autocrat style and executive style.
Rotemberg and Saloner (1993) state that an autocratic leader might do well in pulling the firm out of a crisis and a more democratic leader will better stimulate creativity and intrapreneurship. In high technology and life science companies creativity is highly appreciated.
Before looking at the management styles that usually applies to women or men, it is valuable to mention that a review on 80 different studies by Eagly (1995) resulted in an outcome that male and female leaders were equally effective. Based on this result, one might expect that the choice between a male or female manager does not matter. The research performed by Eagly (1995) also revealed that women were more effective leaders in dominated or female-oriented settings. The same applies for men and male-dominated or male-female-oriented settings.
The latter indicates and points again that it is interesting to look at the difference of female participation on higher management in high technology & life science companies. This research seeks to answer whether the type of business and a diverse management contributes to a higher financial performance. Companies are less hierarchical nowadays. In the fast moving digital world, employees want to develop their personal skills and choose a thought-out career path. According to Eagly (1995) women seem to be better able to help employees in motivating them and be a “transformational” leader and role model. One can therefore assume that women are at least better people managers. But, does this eventually lead to a better financial performance? In the next section, financial performance and the variable to measure financial performance will be described.
2.4 Financial performance
It will be clear that financial performance is an important subject in the whole of this study. Measuring the financial performance is the way to find out if more women in higher management positions is appropriate for the company in terms of adding value. This section is devoted to this research because financial performance can be measured in many different ways. In this section, studies about the best measure for financial performance will be described, which will eventually lead to the best (chosen) measure for this research.
Both accounting-based and market-based measures are widely accepted as valid indicators of firm financial performance (Gentry & Shen, 2010). For accounting-based measures, researchers generally use return on assets (ROA), return on sales (ROS), and return on equity (ROE). For market-based measures, researchers use stock marked-based measures such as Tobin’s Q and market return (Hult et al., 2008). Another often used accounting-based measure is return on investment (ROI). ROI has been criticized in many researches (Fisher, 1984; Fisher and McGowan, 1983; Benston, 1985) as being an inadequate indicator of economic rate of return but research performed by Jacobson (1987) suggests that ROI is clearly warranted as input in the evaluation of business unit profitability.
There are quite some different views on defining financial performance, including accounting- and market-based financial performance. In this research ROA is used as the indicator for financial performance. ROA is an appropriate accounting-based measure that is well known in the accounting environment. Furthermore ROA is used as dependent variable in similar researches (e.g. Campbell and Minquez-Vera, 2008) on this topic. Hence regression results can be compared for these researches.
2.5 Hypotheses development
This section describes the translation of the research questions into hypotheses. The study will be performed using three hypotheses. Each of the hypothesis contain a brief introduction of their development.
As stated in the introduction (see 1.1), Westphal and Milton (2000) say that women often bring a different perspective on complex questions. According to a global survey performed by McKinsey (October 2010)7, a majority of executives believe that gender diversity in leadership is
linked to better financial performance. These statements support the idea that more women functioning on a higher management position can enhance firms’ financial performance. To test this idea the following hypothesis is used:
H1: The percentage of women in higher management has a positive association with firms’ financial
performance
The second hypothesis focusses on the difference in female managers representation between high technology & life sciences companies compared and non- high technology & life sciences companies. This is a control hypothesis for the expectation that within high technology & life science companies fewer women are present at higher management positions. Therefore hypothesis two is as follows:
H2: The percentage of female managers within high technology & life science companies is lower than the
percentage of female managers within other companies
As stated in Section 2.2 there are several factors that contribute to the under participation of women at high technology & life sciences companies. These factors correspond to those of companies in general. The main reason we discussed was the relative lack of gender diversity at most senior levels of leadership in public companies. Therefore it is interesting to investigate whether more gender diversity at high technology & life science companies have a higher influence on financial performance than non- high technology & life science (other) companies. This can be tested using the following hypothesis:
H3: The relationship between the percentage of women and financial performance is more positive in high
technology & life science companies than in other companies
The three hypotheses will be translated into regressions in Section 3.3.
7 Global Survey McKinsey 2010. This survey can be found on the website: http://www.mckinsey.com/business-functions/organization/our-insights/moving-women-to-the-top-mckinsey-global-survey-results
3 Methodology
In this section the research methodology is explained. This section starts with the conceptual model and method used to test the different hypotheses. Subsequently the empirical model is given. In Section 3.4 a regression per hypothesis is presented. The last section contains a clear overview of all the variables needed to perform this research.
3.1 Conceptual model
The conceptual model of this research is as follows:
Figure 1: Conceptual model 3.2 Testing method
The influence of gender diversity on financial performance will be tested. In addition the influence of gender diversity within different type of firms (with focus on high technology & life science companies) will be examined. The method used to test the three different hypothesis is ordinary least square (OLS). This is a method used to estimate the coefficient (β’s in Equation 1 in Section 3.3) in a linear regression model. OLS tries to minimize the sum of squares of the difference between the observed values in the data and the predicted values by the linear model which is based on explanatory variables.
3.3 Empirical model
Regression analysis will examine the influence of gender diversity on company financial performance.
Percentage of women in higher management
Financial performance of a company (ROA)
High technology & life science companies Control variables H1, H3 H1, H3 H2 H2 H3
Let the gender diversity of the firm be defined as the fraction of women working on a higher management level within the company. This research will specifically look at higher management, meaning an executive board function. As a measure for financial performance ROA is used based on previous literature (see Section 2.4).
Panel data will be used to test the different hypotheses. Note that panel data is used because the data (explained in Section 4) consists of both cross section data and a time dimension. The standard form of a panel data model is as follows:
𝑌𝑖𝑡 = 𝛽0+ 𝛽1𝑋1𝑖𝑡+ 𝛽2𝑋2𝑖𝑡+ ⋯ + 𝛽𝑗𝑋𝑗𝑖𝑡+ ⋯ + 𝜀𝑖𝑡 Equation 1: Standard form of a panel data model
As an example 𝑌 represents the firm’s performance (i.e. dependent variable) and the 𝑋𝑗′𝑠 the measure of women participation at higher management level within a firm and other control variables (i.e. independent variables). With 𝛽0 the constant, 𝛽𝑗 ′𝑠 the coefficient of variable j and ε the residual of the regression.
3.4 Regression models
In the following subsections the complete regression to be tested per hypothesis is given. Each regression defines its dependent, independent and controlling variables.
3.4.1 Regression model hypothesis 1
This research wants to test firms’ financial performance and whether gender diversity in higher management is positive associated with it. The dependent variable in regression one is return on assets. The independent variable in regression one is female participation (fraction). Based on Campbell and Minquez-Vera (2008) and Shrader, Blackburn and Iles (1997), leverage and firm size are used as control variables. Therefore, the first hypothesis can be translated into the following regression:
R1: 𝑅𝑂𝐴𝑖𝑡 = 𝛽0+ 𝛽1𝐹𝑃𝑖𝑡+ 𝛽2𝐿𝐸𝑉𝐸𝑅𝑖𝑡+ 𝛽3𝑆𝐼𝑍𝐸𝑖𝑡+ 𝜀𝑖𝑡
For description of the variables, see Table 2.
3.4.2 Regression model hypothesis 2
In this regression the focus is only on female participation difference between sectors. Hence the regression is straight forward in which only the dummy variable is used to indicate whether it
concerns the high technology & life science company or other company. Also the control variables from regression one are taken into account. The second hypothesis can be translated into the following regression:
R2: 𝐹𝑃𝑖𝑡 = 𝛽0+ 𝛽1𝐷𝐻𝑇𝑖𝑡+ 𝛽2𝐿𝐸𝑉𝐸𝑅𝑖𝑡 + 𝛽3𝑆𝐼𝑍𝐸𝑖𝑡 + 𝜀𝑖𝑡
For description of the variables, see Table 2.
3.4.3 Regression model hypothesis 3
This regression is similar to regression one, in that it focusses on the financial performance. However, in addition the difference in financial performance between sectors is analyzed. Therefore, the same control variables are used as in regression one with an addition of the dummy variable to indicate the sector. Also the interaction between female participation and sector is taken into account. The third hypothesis can be translated into the following regression:
R3: 𝑅𝑂𝐴𝑖𝑡 = 𝛽0+ 𝛽1𝐹𝑃𝑖𝑡+ 𝛽2𝐿𝐸𝑉𝐸𝑅𝑖𝑡+ 𝛽3𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽4𝐷𝐻𝑇𝑖𝑡+ 𝛽5𝐹𝑃𝑖𝑡𝐷𝐻𝑇𝑖𝑡+ 𝜀𝑖𝑡
For description of the variables, see Table 2.
In this research six variables are used of which two dependent variables (i.e. ROA and percentage of women at board level) and two independent variables. To control for other characteristics, there are two more explanatory variables added to the equation. In addition a cross variables is used. In order to perform the tests for the different hypotheses, data is required for the defined variables. More information about the data can be found in Section 4.
The table on the following page provides an overview of all the variables used in this research.
Variable Description Type of variable
ROA Return on Assets Dependent
FP Female Participation (fraction) Dependent (R2) and
independent (R1) DHT Dummy variable High technology & life
science companies. Equals 1 for high
technology & life science companies, equals 0 for other companies
Independent
LEVER Leverage (ratio: total debt to total equity) Control variable SIZE Size of the company (Log total assets) Control variable
Table 2: Variables used with description
The variables in Table 2 will be used in the different regression models for each hypotheses as developed in Section 2.5. For the specific calculation, if required, see Section 4.2.
4 Data
In this section the datasets used, the modifications performed on the data, and the statistical program needed for testing, are introduced. First a general description of the panel database will be presented. Second, the description of actual data set used is presented. Section 4.3 will describe the descriptive statistics.
4.1 General description of the panel database
In this section the data used to investigate the effect of diversity within a firm on financial performance is given. The data used for this study is retrieved from WRDS. Specifically, the data in Compustat & Execucomp is used which provides a wide range of information on listed companies from North America. For this study data is taken for the (fiscal year) period 2009-2016. The financial crisis ended halfway 2009 according to the US Nation Bureau of Economic Research8, meaning data starting from the year 2010 will be unbiased regarding the crisis. We
extracted one additional year (i.e. 2009) in order to calculate the ROA (uses average of total assets, see Section 4.2). The information in the WRDS Compustat and Execucomp databases is gathered by Standard & Poor’s.
4.2 Description of actual data set used
For this study the influence of gender diversity on financial performance will be investigated. Furthermore the influence of gender diversity within different type of firms (with focus on high technology & life science companies) will be examined. This research uses data from 1,975 US firms and the firms are all from different sectors (financial services companies are excluded). In Section 3.5 the variables needed are presented. In order to get to these variables, data from both Compustat and Execucomp databases is gathered. The data needed to get to the variables as presented in Section 3.5, are described in Section 4.2.1. Calculations made to get to certain variables are presented in Section 4.2.2. The modifications made to get the actual data set used are described in Section 4.2.3.
The in money expressed terms (such as financial performance) are presented in US dollars. It should be noted that the panel data represents an unbalanced panel data. This refers to the fact
8 The US National Bureau of Economic Research (NBER) is a private nonprofit research organizations that performs unbiased economic research in the US among public policymakers, business professionals, and the academic community.
that not all companies have data for all fiscal years. The panel data tests are conducted with the statistical program STATA.
4.2.1 Needed data
The following table provides an overview of the required data:
Variable Needed data Source
Return on assets (ROA) Earnings before interest and taxes & assets total
Compustat
Female participation (FP) Gender, Company ID Number, Fiscal Year
Execucomp
Dummy High technology & life science companies (DHT)
GIC Groups (Compustat), Industry Group, Industry Group Description
Execucomp
Leverage (LEVER) Total debt & total equity Compustat
Size of the company (SIZE) Total Assets Compustat
Other Compustat Global Company Key, Company Name, ISO Currency Code, Fiscal Year-end Month
Compustat
Table 3: Needed data per variable 4.2.2 Specific calculations
Variables ROA, FP and LEVER need more explanation because they consist of calculations. The calculations are as follows:
𝑅𝑂𝐴𝑡 = 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒9 (𝐴𝑠𝑠𝑒𝑡𝑠 𝑇𝑜𝑡𝑎𝑙𝑡− 𝐴𝑠𝑠𝑒𝑡𝑠 𝑇𝑜𝑡𝑎𝑙𝑡−1)/2 𝐹𝑃 =𝑇𝑜𝑡𝑎𝑙 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐹𝑒𝑚𝑎𝑙𝑒 𝐻𝑖𝑔ℎ𝑒𝑟 𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝑠 𝑇𝑜𝑡𝑎𝑙 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐵𝑜𝑎𝑟𝑑 𝑀𝑒𝑚𝑏𝑒𝑟𝑠10 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 = 𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑏𝑡 𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦
9 Where net income in ROA is defined as earnings before interest and taxes
10 Where total number of board members per company is calculated using Company ID Number and Fiscal Year from the Execucomp database in combination with Global Company Key and Data Year – Fiscal from Compustat
4.2.3 Composing the dataset
In this research, multiple exclusions to the datasets Compustat and Execucomp are performed. The following tables (Table 4 and Table 5, see next page) contain the exclusions performed per dataset. The first table contains the exclusions performed to the Compustat dataset. Before data was extracted, financial services companies are excluded and the currency is set to US dollars (Canadian dollars are excluded). Both active as inactive companies are extracted. Data starting from January 2009 until December 2016 is used. Because the dataset contains broken fiscal years, also 2008 figures are present in the database. Because only fiscal yearend figures for the year 2009 are needed, the 2008 figures are excluded. The original Compustat dataset contained 73,465 observations. After excluding the 2008 figures, the dataset contained 72,484 observations. Also financial data is missing for the variables needed in this research. All observations without earnings before interest & taxes, total debt (including current), total equity (common/ordinary), and total assets is excluded. After excluding these missing data, the dataset contained 47,614 observations. As also the specification of the sectors is needed, the data for which the sector is not defined has to be removed. In the Compustat dataset the GIC Groups is needed as identification of the sector. 266 items are missing, so the final Compustat dataset contains 47,348 observations. Before this dataset can be combined with the Execucomp dataset, all required calculations are performed (i.e. ROA and leverage). Two new columns are created which calculate ROA, leverage and natural logarithm of total assets as defined in Section 4.2.2. After these calculations the fiscal year 2009 data is deleted, which result in 41,166 observations. Note that 2009 data was only needed for the ROA calculations.
Description Remaining observations
Number of observations original dataset 73,465
Exclude 2008 figures 72,484
Exclude missing data 47,614
Exclude observations without defined sector 47,348
Exclude fiscal year 2009 after calculations 41,166
Final dataset before combining with Execucomp 41,166
Table 4: Exclusions dataset Compustat
In addition, data from the Execucomp dataset is required. The data range chosen for this dataset is from 2010 to 2016. The number of observations extracted from the Execucomp dataset are 72,679. There are no empty cells for the Company ID number to match with Compustat, and there are also no empty cells for the variable Gender. Hence, no modifications are required for this step. However, the sector code (Industry Group, is equal to GIC Groups in Compustat) contains empty cells and these observations will be omitted from the final dataset. The sector code in Execucomp also contains zeros, which identifies that no sector is applicable. For this reason these are also deleted from the dataset. The final number of observations are 72,600.
Description Remaining
observations
Number of observations original dataset 72,679
Exclude zeros in sector code 72,600
Final dataset before combining with Compustat 72,600
Table 5: Exclusions dataset Execucomp
The next step is to combine both datasets. This is done based on the company codes within both datasets (i.e. Compustat – Global Company Key, Execucomp – Company ID Number) and the fiscal year. The Compustat data contains the company level data over the fiscal years, whereas the Execucomp data contains the data in individual board member per company, per fiscal year. The data required from Execucomp is the fraction of female board members. Hence, the fraction
data per company and per fiscal year is matched to the company and fiscal year in the Compustat data.
It can be the case that not for every company available in Compustat there is data available in Execucomp. Meaning, that on the company board level no data exists. Hence, for all companies where no board level data is given, data has to be omitted from the final dataset. Compustat data contained so far 41,166, however after combining the data with Execucomp the data set is reduced to 11,843 observations. The final number of observations equals data of 1,975 companies.
The final step in the dataset is to put a label on each sector as a high technology & life science company or as another company. In Appendix Table 1 the different sector in the dataset are presented. The sectors with the Industry Group codes 2010, 2510, 3510, 3520, 4510, 4520, 4530 and 5010 are labeled as high technology & life science companies.
In the dataset a few outliers are encountered. The most significant outliers are found in the variable Leverage. The maximum is 3,569.4 and the minimum is -776.6. Based on the histogram (see Appendix Figure 1) Leverage values out of the range -500.0 to 500.0 are excluded. This results in in omitting 8 observations from the dataset (see Appendix Figure 2). Hence, the dataset consists of 11,835 observations. In Appendix Figures 3, 4 and 5 the histograms of the variables Size, ROA and FP are also presented.
The original dataset, before the exclusions, consisted of 13,066 companies (i.e. unique company names and codes for all 73,465 observations in Compustat). After the exclusions data of 1,975 companies is available. Therefore, the final dataset contains 15.1% of the companies of the original dataset. However, missing observations have been excluded in the original dataset which resulted in 8,865 companies for which data was available. Subsequently the final dataset contains 22.3% of the companies of the original dataset for which data was available in both Compustat and Execucomp. On average each company contains 5.99 observations (i.e. 11,835 observations divided by 1,975 companies), meaning each company on average has data available for 6 fiscal years. In this research, various unavoidable exclusions were made. The available data contains almost a quarter of the companies with on average 6 of the 7 fiscal years, and is therefore assumed to be representative.
4.3 Descriptive statistics
In this section the core statistics, the correlation matrix and the multicollinearity analysis is given.
4.3.1 Core statistics
To get more insight into the data, descriptive statistics and graphs are presented. Table 6 shows the observations, the mean, the standard deviation and the minimum and maximum of the variables used. Table 6 contains statistic data for the variables over the fiscal years 2010 until 2016 for 11,835 observations.
Variable Mean Standard
deviation
Minimum Maximum
ROA 0.0901383 0.1155776 -1.223365 1.270595
FP (fraction) 0.0759367 0.1230372 0 0.8
Leverage 0.8527123 10.77146 -290.5 384.8333
Size (log total assets) 7.785496 1.748536 1.272005 15.00578
DHT 0.434981 0.4957754 0 1
FP x DHT 0.0316362 0.0857469 0 0.6666667
Table 6: Descriptive statistics of the database
One can interpreted from Table 6 that Leverage has a minimum of -290.5, which may be treated as an outlier. As shown in Appendix Figure 2, the histogram contains multiple observations in the same area. Therefore, the minimum figure is not assumed to be an outlier.
4.3.2 Correlation matrix
Also a correlation matrix is presented. With the correlation matrix it can be identified whether there is an indication of multicollinearity (i.e. high correlation between two variables). The correlation ranges from minus 1 to 1 where the former indicates a perfect negative linear correlation and the latter a perfect positive linear correlation. Most of the variables have a close to zero correlation (which indicates no multicollinearity). The interaction term has a correlation close to a half with FP and DHT but this is as expected. The sector dummy has a negative correlation with Size (i.e. -0.1918) which indicates that high tech & life science companies on average have a lower company size. In the next section, an analysis on multicollinearity will be given.
ROA FP Leverage Size DHT FP x DHT ROA 1.000 FP 0.0723 1.000 Leverage -0.0082 -0.0083 1.000 Size 0.0411 -0.0202 0.0307 1.000 DHT 0.0010 -0.0229 -0.0115 -0.1918 1.000 FP x DHT 0.0240 0.5641 -0.0093 -0.0571 0.4205 1.000
Table 7: Correlation matrix (11,835 observations) 4.3.3 Multicollinearity analysis
In this section, a multicollinearity analysis will be given. To test for multicollinearity the Variance Influence Factor (VIF) test can be used to measure the linear coherence between the explanatory variables. The VIF is calculated as 1 / (1-R2). The VIF results are given in Table 8 where it can be
seen that the mean VIF is 1.44. The maximum VIF value is 2.03 and belongs to the interaction variable (i.e. FP x DHT). Assuming the average VIF, the variance of a coefficient is 44% larger than it would be if the predictive variables would be uncorrelated with all other predictive variables. A high VIF indicates instability (high variance) of the regression coefficient. There are no absolute criteria for permissible multicollinearity in the form of VIF. The main rule of thumb is that multicollinearity is high when the VIF value is larger than four or five. Therefore, it can be concluded that no multicollinearity is present in the variables.
Variable VIF FP 1.67 Leverage 1.00 Size 1.04 DHT 1.43 FP x DHT 2.03 Mean VIF 1.44
5 Results
In this chapter the results of the study will be presented. The results will be described per hypothesis. They will be linked to the literature and the research methodology. Table 9 provides output from STATA for the performed regressions.
5.1 Regression results
In this research both fixed effects and random effects is used. For panel data, fixed effects is the main technique to use. However, for regression 2 and 3 fixed effects cannot be used because collinearity between the sector dummy and the panel ID (which is the company indication) exists. Note that the panel data is unbalanced. The number of fiscal years is 8 and the total number of observations is 11,835 which includes company data over different fiscal years.
Before the coefficient results of regression 1 (see Section 3.4.1) are presented, a Hausman-test is performed on the fixed and random effect results. This Hausman-test is used to verify which regression technique to use. The Hausman test statistic for regression 1 gives a p-value lower than 0.05 which indicates that a fixed effect model should be used. Therefore, in regression 1 fixed effects is used. Using fixed effects gives the opportunity to discover the relation between the predictive variable and the outcome variable within, for example, companies. Each company has its individual features that may or may not influence the predictive variable. In this research being a male manager or a female manager could namely effect the financial performance of their company. Managers may influence or bias the predictive variable or the outcome variable and control is needed in this existence. The error term and the predictive variable correlations can be cleared using fixed effects which leads to evaluating the net result of the predictive variables on the outcome variable.
For regression 2 the Hausman-test indicates that random effects should be used (i.e. p-value test statistic =0.179 >0.05). The Hausman-test on regression 3 suggests at a 5% significance level, the fixed effect method. However, given that in the fixed effect model the sector dummy will be omitted due to collinearity, it is decided to use random effects for regression 3 as well. In this way, the dummy variable can be estimated and interpreted. With the random effect technique, the variation across entities is expected to be random and uncorrelated in relation to the predictive or independent variables of the model. Another reason to use random effects for regression 3 is because it is possible to include time invariant variables such as female participation, which would be absorbed by the intercept using the fixed effects technique.
In Table 9 the regression results are presented. The table contains the dependent variables for each regression in the header. The constant, the independent variables and the control variables are presented in the left block of the table. First the coefficient of the variable is given and then the standard deviation of the coefficient is given in brackets. In the lower block of the table the number of observations, the R2 and the p-value are displayed.
(1) ROA (2) FP (3) ROA _cons 0.0301916* (0.0173542) 0.1068318*** (0.0100717) 0.0488407*** (0.0096344) FP 0.0222813** (0.0112455) - 0.0445493*** (0.0129014) Leverage -3.43e-07 (0.0000664) -0.000146** (0.0000591) -7.75e-06 (0.0000657) Size 0.0074825*** (0.0022214) -0.0034135*** (0.0011884) 0.0045755*** (0.0011341) DHT - -0.0087242* (0.0050807) 0.0050451 (0.0047785) FP x DHT - - -0.0191968 (0.0198552) N 11,835 11,835 11,835 R2 (overall) 0.0031 0.0011 0.0059 p-value 0.0019*** 0.0011*** 0.0000*** * p < 0.10, ** p < 0.05, *** p < 0.01
Table 9: OLS regressions
5.1.1 Evaluation of results hypothesis 1
H1: The percentage of women in higher management has a positive association with firms’ financial
performance
In Table 9 it can be seen that the coefficient of FP is positive (i.e. 0.0222813) and significant at a 5% level. The coefficient of Leverage is negative and not significant. From an economic point of view the negative sign is unexpected. Furthermore the coefficient of Size is positive and significant
at a 1% level. The remaining coefficient of the constant is positive and significant at the 10% level. Overall the panel data regression is statistically significant at the 1% level. The overall R2 equals
0.0031 which indicates a very weak model fit of the data. Note that R2 indicates how much of the
variability within the data is explained in the model and the level range is between 0 and 1. To summarize, hypothesis 1 cannot be rejected. Hence, the percentage of women in higher management has a positive effect on firms’ financial performance.
5.1.2 Evaluation of results hypothesis 2
H2: The percentage of female managers within high technology & life science companies is lower than the
percentage of female managers within other companies
In the STATA output for regression 2 it can be seen that the coefficient of Leverage is negative and close to zero but statistically significant at a 5% level. The coefficient of Size is also negative but significant at a 1% level. The latter indicates that the bigger the company, the less women are relatively titled with a higher management position which is unpredicted. Furthermore, the constant in this regression is significant at a 1% level. DHT has a negative and significant effect (at the 10% level) on female participation. The overall model is also significant at a 1% level. Hence, hypothesis 2 cannot be rejected. This means that the percentage of female managers within high technology & life science companies is lower than the percentage of female managers within other companies.
5.1.3 Evaluation of results hypothesis 3
H3: The relationship between the percentage of women and financial performance is more positive in high
technology & life science companies than in other companies
Finally the STATA output for regression 3 is discussed. In the STATA output for regression 3 it can be seen that the coefficient of FP is positive and significant at the 1% level. As in regression 1, the coefficient of Leverage is negative and not significant. Furthermore, the coefficient of Size is positive and significant at the 1% level. The coefficient DHT is positive but not statistically significant. The interaction term is negative and also not significant. In this model the constant is positive and significant at the 1% level. In summary, the following can be said: the percentage of women and the size of the company has a positive effect on firms’ financial performance. Statistically it is not shown that female participation within high technology & life science companies contributes to those firms’ financial performances. Therefore hypothesis 3 can be rejected.
5.2 Normality of error term
In addition to the regression table (Table 9), a histogram is made for the residual terms (see Appendix Figures 6, 8 and 10). In order to test whether these residuals shown in the histograms are normally distributed, different techniques can be used. As a start a Skewness and Kurtosis test and a Kolmogorov-Smirnov test is performed to check for normality. Based on the normality tests in STATA, the distribution of the residuals can be distinguished from normally distributed data. Hence, the residuals are non-normal. For all three hypothesis the previously mentioned tests give a p-value of 0.0000. The latter indicates that the null hypothesis of normality is rejected. However, given the plots one could argue differently. In many literature it is recommended to use QQ-plots to indicate for normaly distributed residuals, instead of a normality test. This is because the normality test is quit sensitive such that with a large number of observations, the chance of rejecting is larger.
A QQ-plot of the residuals of regression 1 is plotted in Appendix Figure 7. The distribution of the residuals are only deviating in the tails. However, in the mid-range the quantiles are aligned. For the residuals of regression 3, the QQ-plot (Appendix Figure 11) has a similar shape and can be described equal to regression 1. For regression 2 the QQ-plot is quit deviating. Therefore it can be assumed that those error terms are not normally distributed.
In panel data regression it is assumed that the error term of the regression is normally distributed. Though, the normality tests performed on the error term of the regressions in this study show that the normality assumption is violated. This means that inconsistencies in the resulting regression significance values could arise. For this reason one should be careful with interpreting the final regression results.
6 Conclusion and recommendations
In this chapter the conclusion and recommendations are presented. In addition the limitations and possible follow-up research is discussed.
With the aid of panel data from Compustat and Execucomp it is investigated whether there is a positive relationship between female participation in higher management and firms’ financial performance. In addition the effect of the company sector is considered. Prior research in gender diversity on firms’ financial performance suggests that there exists a positive relationship. For this reason, it would be interesting to see whether the same result would apply for the chosen data in Compustat and Execucomp. In particular, data of companies in the north of America is investigated.
The results of this study show that hypotheses 1 and 2 cannot be rejected and hypothesis 3 can be rejected. The first hypothesis is focused on the percentage of women in higher management and whether it is positively associated with firms’ financial performance. The results show that the percentage of women in higher management has positive effects on the financial performance of the companies in North America which are part of the dataset. The second hypothesis focuses on the assumption that the percentage of female managers within high technology & life science companies is lower than in other companies. Regression results show that the percentage of female managers is lower in high technology & life science companies. Hypothesis 3 investigates whether the relationship between the percentage of female higher management and performance is more positive in high technology & life science companies than in other companies. The results of the regression give no indication for this effect.
Summarizing, this means that there is a significant positive effect of gender diversity on North American firms’ financial performance. Furthermore it can be concluded that discrimination of gender exists for high tech & life science companies. Female participation at management level is significantly lower in the high tech & life science companies than other companies. To conclude, the study does not show any significant results for positive relationship between females participation at higher management in high tech & life science companies than for other companies. It could be argued that the number of observations with female higher managers may be too small to appropriately estimate the parameters in this study. A follow up study, making use of a larger sample with active female higher management, will have to confirm this.
In this research the statistical program STATA is used for the statistical analysis. In which both random and fixed effects regressions are executed. From the correlation point of view it was already noted that the different variables suggested no distinctive relation. A pre-analysis could have been performed to specifically investigate the most suitable control variables. In this analysis the control variables were chosen based on previous literature.
Another important point; a more in-depth study could verify if other indications of financial performance give a better representation. Some variables in the data showed a wide-range of values, for example leverage. For this variable a few outliers were detected and omitted. However, the data remains to have many negative leverage values. Consequently this could have disrupted the estimation of the coefficient.
The outcome of the study is not in line with the study of Francoeur et al. (2008) who investigated the effect of female participation in the board of directors on firms’ financial performance within Canadian firms. They used the Fama and French (1992, 1993) model to control for the level of risk and also data from another country is used. Therefore, caution is required when comparing the studies. Interesting is that Francoeur et al. (2008) found a significant result for female participation within management positions (instead of board positions) on firms’ financial performance. The latter indicates the importance of gender diversity within firms. Another result that arises from the research of Francoeur et al. (2008) is that firms operating in complex environments do generate positive and significant abnormal returns when they have a high proportion of female officers. In this study it is assumed that high technology & life science companies do operate in complex environments as well. The result of Francoeur et al. (2008) is therefore in contrast with the result of this study, as no significant effect is found on financial performance for more female board members within high technology & life science companies.
In a study from Shrader et al. (1997) it is found that women are better in connecting with customers, employees and other constituencies. This again suggests the positive impact of gender diversity within companies. However, Shrader et al. (1997) did not find a positive relationship between higher percentage of female board members and firms’ financial performance. Note that the research is performed in 1997 and at that time fewer women were active at senior management level. In Shrader et al. (1997) research a safety margin is taken into account for evaluating the results. In particular, they discuss the impact that women are given in a firm and they have less the opportunity to show higher management capabilities. This is because in general they assume women get less challenging assignments and less possibility to perform at higher management levels.
Just like in this research, Campbell and Minquez-Vera (2008) used panel data analysis to investigate whether gender diversity in higher management is positively associated with firms’ financial performance. Their results show that female participation on the board of directors have no direct effect on firm’s value within Spanish companies. Results however do indicate a positive outcome on firm value and that shows the importance of female participation within companies. In general it is clearly observed that a positive relation is found, however not always significant.
In summary it can be concluded that gender diversity among companies has impact. This study has statistically shown that more women at board level positively influence firms’ financial performance. On the other hand, within high tech & life science companies this is not demonstrated. There is many more to investigate in this area. For instance, the use of more advanced regression methods or expanding to global levels to do a country comparison in this field. This is left for further research.
References
Auronen, Lauri. "Asymmetric information: theory and applications."Seminar of Strategy and International Business as Helsinki University of Technology (2003).
Bacidore, Jeffrey M., et al. "The search for the best financial performance measure."Financial Analysts Journal(1997): 11-20.
Bell, David A., and Shulamite Shen White. "Gender Diversity in Silicon Valley: A Comparison of Silicon Valley Public Companies and Large Public Companies. " Fenwick & West LLP (2014).
Benston, George J., "The Validity of Profits-Structure Studies with Particular Reference to the FTC’s Line of Business Data " American Economic Review (1985): 75, 37-67
Campbell, K., & Mínguez-Vera, A. (2008). "Gender diversity in the boardroom and firm financial performance." Journal of business ethics, 83(3), 435-451.
Chemla, Gilles, and Christopher A. Hennessy. "Skin in the game and moral hazard." The Journal
of Finance 69.4 (2014): 1597-1641.
Donaldson, Lex, and James H. Davis. "Stewardship theory or agency theory: CEO governance and shareholder returns." Australian Journal of management 16.1 (1991): 49-64.
Dwyer, Sean, Orlando C. Richard, and Ken Chadwick. "Gender diversity in management and firm performance: the influence of growth orientation and organizational culture." Journal of
Business Research 56.12 (2003): 1009-1019.
Eagly, Alice H., Steven J. Karau, and Mona G. Makhijani. "Gender and the effectiveness of leaders: a meta-analysis." (1995): 125.
Eisenhardt, Kathleen M. "Agency theory: An assessment and review." Academy of management
review 14.1 (1989): 57-74.
Fama, Eugene F., and Kenneth R. French. "Industry costs of equity."Journal of financial economics43.2 (1997): 153-193.
Fisher, Franklin M., "The Misues of Accounting Rates of Return: Reply. " American Economic
Review (1984): 74, 509-17
Fisher, Franklin M., and McGowan, John J., "On the Misuse of Accounting Rates of Return to Infer Monopoly Profits " American Economic Review (1983): 73, 82-97
Francoeur, C., Labelle, R., & Sinclair-Desgagné, B. (2008). "Gender diversity in corporate governance and top management." Journal of business ethics, 81(1), 83-95.
Gentry, Richard J., and Wei Shen. "The relationship between accounting and market measures of firm financial performance: How strong is it?"Journal of Managerial Issues(2010): 514-530. Haveman, Heather A., and Lauren S. Beresford. "If you’re so smart, why aren’t you the boss?
Explaining the persistent vertical gender gap in management."The ANNALS of the American Academy of Political and Social science639.1 (2012): 114-130.
Hult, G. Tomas M., et al. "An assessment of the measurement of performance in international business research."Journal of International Business Studies39.6 (2008): 1064-1080.
Jacobson, Robert. "The validity of ROI as a measure of business performance."The American Economic Review77.3 (1987): 470-478.
Jensen, Michael C., and William H. Meckling. "Theory of the firm: Managerial behavior, agency costs and ownership structure." Journal of financial economics 3.4 (1976): 305-360.
Limbare, Sameer. "Leadership styles & conflict management styles of executives."Indian Journal of Industrial Relations(2012): 172-180.
Kravitz, David A. "More women in the workplace: is there a payoff in firm performance?." The
Academy of Management Executive 17.3 (2003): 148-149.
Merchant, K. and Van der Stede, W. (2012) "Management control systems: Performance measurement, evaluation and incentives." 3rd ed. (Harlow: Pearson).
Rotemberg, Julio J., and Garth Saloner. "Leadership style and incentives." Management Science 39.11 (1993): 1299-1318.
Shrader, Charles B., Virginia B. Blackburn, and Paul Iles. "Women in management and firm financial performance: An exploratory study." Journal of managerial issues (1997): 355-372. Statham, Anne. "The gender model revisited: Differences in the management styles of men and
women." Sex Roles 16.7-8 (1987): 409-430.
Shrader, C. B., Blackburn, V. B., & Iles, P. (1997). "Women in management and firm financial performance: An exploratory study." Journal of managerial issues, 355-372.
Westphal, J. D., & Milton, L. P. (2000). "How experience and network ties affect the influence of demographic minorities on corporate boards." Administrative Science Quarterly, 45(2), 366-398.
