BACHELOR THESIS
Gender diversity in the boardroom and firm
performance: Evidence from US S&P500 firms
Luyuan Shao, Economics and business, University of Amsterdam
2018/6/20
Advisor: Patrick Stastra
1
1
I would like to sincerely thank my advisor Patrick for his helpful comments and patience.
Abstract: Using the sample consisting of 138 S&P 500 US firms between 2010 and 2012, this paper
tested whether gender diversity on board influence company performance or not. We selected the
presence of female directors and proportion of female directors in the boardroom as our
explanatory variables and return on asset as well as cash flow from operations to total asset were
chosen as dependent variables. By employing Ordinary Least Squares and Random Effect Model,
this study indicates that gender diverse issue in the boardroom may have no significant influence
on performance of companies.
2
Statement of Originality
This document is written by Student [Luyuan Shao] who declares to take full responsibility for the
contents of this document.
I declare that the text and the work presented in this document are 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.
3
Contents
1. Introduction ·······4
2. Literature Review and the Development of Hypothesis········6
2.1 Empirical Research ···6
2.2 Theoretical Perspective ···7
2.2.1 Agency Theory···7
2.2.2 Social Psychological Theory ···8
2.3 Development of Hypothesis ···9 3. Research Method····9 3.1 Sample Collection ···9 3.2 Dependent Variables ···9 3.3 Independent Variables ···10 3.4 Control Variables ···10 3.5 Descriptive Statistics ···13 3.6 Regression Models ···14
4. Regression Results and Findings ······15
4.1 OLS Regression Results ···15
4.2 OLS Regression Results after Adding More Control Variables and Random Effect Model Results ·16 4.3 Robustness Check ···21
5. Conclusion ·····21
6. Reference List ·····24
4
1. Introduction
The outbreak of financial crisis in 2008 has revealed severe problems in corporations, which results a rising
attention among the policy makers toward how to reach an effective corporate governance structure. Lacks of
monitoring in the company is a major cause for corporate failure according to the agency theory (CĂRĂUȘU,
2015). This has led many regulators as well as scholars to focus on the composition of the board regarding to
gender, ethnicity, percentage of independent board members and the background of directors, etc. Gender
diversity, as one of the widely discussed topic by the general public, has been included in guidelines and
legislations in many countries. David et al. (2010) state regardless of how much a vital role that corporate
governance plays in a company, many companies fail to manage it and are not alert to its importance.
Consequently, the US government introduced a massive piece of laws named Sarbanes-Oxley Act of 2002 to
regulate the companies. Other countries like Norway amended the Public Limited Companies Act in 2005,
stated that if there are more than nine directors in the boardroom, the proportion of female should be no less
than 40%. Similarly, In the Law of Equality passed by the Spanish Parliament in 2007, it requires 40% of all
board seats should be occupied by females for listed firms. In 2011, Lord Davies set a number of targets for the top performance companies in the UK to recognize the beneficial result of gender diversity in boardrooms.
Nevertheless, in some developing countries, like China and Thailand, currently there is no requirement for
gender quota on company board. This may be caused by the lack of attention on feminist consciousness,
which results in an automatic impression that women are expected to stay at home and take responsibility of
the housework duties rather than going out for searching a job.
These regulations and policies give rise to at least two arguments for gender diversity in companies. One
argument is corresponding to the agency theory, which states that enforcing a certain female board director
quota through enhanced corporate monitoring would have a positive impact on the performance of
organizations. According to Singh and Vinicombe (2004), Females in general are more serious when making
decisions than males. Additionally, gender diversity of directors might result in more diverse perspectives
toward business issues, less information bias when making decisions as well as more creativity within the firm.
This argument has also been documented by a large amount of empirical studies in different countries (Daniel
et al. 2015; Gul et al. 2011; Vafaei et al. 2015; Ararat et al. 2015; Carter et al. 2003). On the other hand,
5
Ferreria (2009), and Smith, Smith and Verner (2006) are consistent with the social psychology theory. They
state that more women directors might give rise to more conflict in the boardroom, thus leading to the
decision-making process less efficient.
Consequently, empirical findings from prior studies are still inconclusive. It remains a question on whether it
is a wise choice for countries to demand more gender quotas for the company boardrooms. However, any
relationship should be set based on existing evidences. If the quota policy intervention improves the
performance of companies, then it serves as an efficient monitoring device for firms and proves
gender-diverse firms perform better than homogenous firms. Whereas, if there is no such effects, then these
quota put forward by policy makers may simply help social fairness.
The main purpose of this research is to examine whether a gender diverse board enhances organization
performance or not by using a sample consists of 414 observations (138 firms) from the US S&P 500 index
during 2010-2012. Although, there are other studies also explore the US S&P 500 firms, the improvement of
this study is that we use data from a more recent year period, and employ different measurements of
performance (ROA and CFO/TA). We implement the OLS regression both with and without control of the
industry dummies and year dummies. Besides, we use the Random Effect methodology as well, which was
rarely in previous studies. For the purpose of reliability, we also do robustness checks for both OLS
methodology and REM methodology. We failed to identify any significant relation between gender diversity
and the performance of companies. This result does not support the requirements for female quota set by
many countries that were aiming at improve the performance of the firms.
This research is organized as follows: section 2 provides some existing literatures and theories related to
gender diverse and performance of company topic as well as explanation of the hypothesis development based
on empirical findings. Section 3 illustrates the research method including sample collection, variables used in
the analysis, descriptive statistics and regression models. Furthermore, section 4 discusses the results of the
6
2. Literature Review and the Development of Hypothesis
2.1 Empirical Research
The notice of the importance of board monitoring has attracted public attentions on board diversity. The board
composition has been considered vital in approaching the interest of stakeholders in organizations, providing
information to managers as well as monitoring to ensure making decisions efficiently (Marinova, Plantenga
and Remery, 2010). Gender diversity, average age of directors, board size as well as independent board
members are main compositions of board diversity, which were used frequently in previous literatures by
scholars to discover how the boardroom composition affects the decision-making process and, hence,
influencing the company performance. Gender diversity is one of the most crucial variables of board
composition. Studies on this topic have been prompted by the fact that female is underrepresented in
organization boards (Yap, Chan and Zainudin, 2017).Although there are already numerous studies in this area,
Marinova et al. (2010) conclude that the empirical result regarding to whether diversity in the boardroom
affect organization performance in an optimistic or passive way is still inconclusive.
Research on gender diversity in the boardroom and organization performance generally has reached two
conclusions. One line of studies state that more females on the corporation’s board leading to a good business
performance. Daniel et al. (2015), using firms from four different Asian countries, and employing Return on
Equity (ROE) as measurement of performance, reported a positive link between gender diversity and ROE.
Another research, Gul et al. (2011) measured firm performance by stock price informativeness (public
information disclosure), concluded that gender diversity on corporate board has a significant and positive
influence on organization’s stock price informativeness by enhancing information disclosure for large firms
and providing incentives for gathering private information in small firms. In Australia, Vafaei et al. (2015)
documented a significant and positive connection between board diversity with respect to gender and business
performance using top 500 listed Australian companies. Moreover, Ararat et al. (2015) examined the emerging
markets with data collected from the BIST-100 index, they observed a non-linear but positive link between
multiple diversity and firm’s performance measured by ROE and MTB (Market-to-book ratio). Earlier studies,
Carter et al. (2003) reported that the percentage of female directors is positively connected to organization’s
7
Another line of studies state that there is no impact or adverse impact of gender diversity on business
performance. David et al. (2010) examined the effect of the diversity pertaining to gender and ethnicity on
organization performance of US S&P500 index companies for the period of 1998 to 2002. They observed that
more women directors in the boardroom do not contribute to a good performance of organizations. This is
consistent with the study conducted by Siantar (2016), their work examined listed India organizations on the
National Stock Exchange (NSE) using data from 2006 to 2015. Similarly, although three different measures of
gender diversity were used by Yap, Chan and Zainudin (2017), they did not find any relationships between the
diverse of the gender in the boardroom and Tobin’s Q listed on the top 100 FTSE Bursa Malaysia between
2009 and 2013. In addition, Rose (2007) did not conclude that board gender diversity has a significant link
with Tobin’s Q by using samples from the Danish companies. In addition, using the data collected from US
firms, Adam and Ferreria (2009) reported that the percentage of female in the boardroom does not contribute
to the increase in Tobin’s Q. On the contrary, Smith, Smith and Verner (2006) employed gross value added
measure as the first quartile turnover rate of firm relative to the turnover rate of the industry, a new method to
measure business performance which has not been employed by other scholars before. They documented
gender diversity has negative impacts on the gross value added for Danish companies.
The implication is difficult to deduce based on these limited 11 previous studies, since they investigated in
various areas, countries, used different variables and models etc. In order to get a better insight of the causal
relation between the diverse boards and organization performance, except for just focusing on the empirical
studies, two well-established theories related to this topic are also taken into account in this paper.
2.2 Theoretical Perspective 2.2.1 Agency Theory
Agency theory is a commonly used theory in explaining how to reach an effective corporate governance
structure. It provides a fundamental concept that the functions of corporate board are monitoring and
controlling (David et al., 2010). According to Gallego et al. (2009) a more diverse boardroom may improve
monitoring of managers so that the independence of board can be enhanced. For example, board with female
directors might provide a fresher view on complex managerial decisions than the board only contains male
8
Vinicombe (2004) highlighted that females perform more seriously in their roles than males, which enhances
the governance of the corporation. However, not all evidence shows that female directors have a positive
effect on board monitoring, the results are still mixed. When analyzing the commercial banks of ASEAN-5,
Ramly et al. (2015) observed that independent directors serve as a monitoring device for the banks and thus
increase bank efficiency. Nonetheless, this positive impact weakens when the directors are female.
Concluding, research on agency theory lists some evidences that encourage firms to include more female
directors in the corporations, but it does not rule out the probability that a diverse board might have an adverse
effect.
2.2.2 Social Psychology Theory
Social psychology theory analyzes social behaviors of individuals as well as groups (Cremer & Murnighan,
2011). It is applied widely by organizations to manage its employees. Social psychology theory considers how
individual might behave when he/she has a majority status (e.g., male representatives) versus when he/she
might have a minority status (e.g., female representatives) (Mogbogu, 2016). While agency theory may prefer
gender diversity on the board, social psychology theory presumes that gender diverse in the boardrooms may
adversely affect the performance of the firm. In the hiring decision process within an organization, stereotype
toward female persist although some studies show that female get slightly higher scores than male on job
performance estimate (Robbins and Judge, 2016). Even when males and females have an equally level of
abilities, men is more likely to be chosen as a leader in an organization than women (Robbins and Judge,
2016). Besides the stereotype toward female, another reason most organizations may prefer a less diverse
board is that gender diversity may generates more conflicts among employees. Campbell and Minguez-Vera
(2008) argue that including more females on board induces various ideas on a single issue, causing the
decision-making procedure more complex and potentially more time-consuming. It is also argued that social
psychological procedure plays a crucial role in the cooperativeness of the boardroom. Members tend to be
more homogeneous in a group are often more productive with less emotional conflicts (Williams and O’Reilly,
1998). However, there are still arguments that support diversity of gender in organizations. For instance,
Siantar (2016) indicated that firm performance can be enhanced by heterogeneous boards with more creative
9
research on the social psychology theory provides results that are generally opposite from the agency theory
on the board gender diversity issue.
2.3 Development of Hypothesis
In conclusion, the outcomes of both empirical research and theoretical evidence are mixed. Five empirical
studies provided a positive impact of diversify of gender in the boardroom on organization performance. Six
empirical studies indicated neither causal relationship nor negative relations were found between the gender
diversity on the board and organization performance. Even though the agency theory support for board gender
diversity to some extent, the social psychology theory does not prefer a more diverse board. Therefore, since
the limited evidence does not show a clear path towards the connection between board gender diversity and
organization performance, the fundamental hypothesis of this research is stated as follow:
Board gender diversity does not affect organization performance.
3. Research method
3.1 Sample Collection
The sample for this thesis consists of a balanced panel of companies listed in the S&P500 index over the
three-year period of 2010-2012. The identities of board directors were directly obtained from the Institutional
shareholder services (ISS), which is one of the vendors of Wharton Research Data Service (WRDS).
Financial data to compute the cash flow from operation to total assets ratio, return on asset and leverage ratio
were extracted from the Compustat-Capital IQ database, a vendor of WRDS as well. Due to the missing
corporate governance data and firm financial data, the balanced panel sample contains 414 firm-year
observations.
3.2 Dependent Variables
A recent research by David et al. (2010) on gender and ethnic diversity and organization performance in US
S&P500 firms used Tobin’s Q and Return on Asset (ROA) as measurement of financial performance. They
found that the relationship is mixed when they employing different response variables. In order to enhance the
10
(ROA) and cash flow from operation to total asset (CFO/TA) respectively. According to Vafaei et al. (2015),
ROA is accounting-based measurement, which is a useful tool to assess how well the company has performed
in the previous years. The ROA of a firm, measured by dividing firm’s net income by total assets, is a good
indicator of how firm’s total assets have been generalized to make income for its stakeholders (Yap, Chan and
Zainudin, 2017). The higher the ratio, the more effective the company is. This is because the firm is utilizing
more earnings with less investment (Christopher, Siantar and Mahajan, 2016). On the other hand, CFO/TA is
an economic-based measurement, which indicates how efficiency the firm is in using assets (Smallbusiness,
2018). Vafaei et al. (2015) calculated CFO/TA as cash flow from operation activities divided by the firm’s
total assets.
3.3 Independent variables
This paper employed two proxies as independent variable of interest to measure gender diversity. Yap, Chan,
and Zainudin (2017) suggested that the mixed measures attributed to more comprehensive examinations of the
representations of female directors in the boardrooms. Therefore, gender diversity is measured by a dummy
variable (fem) indicating the presence of women directors in the boardroom and also measured by the
percentage of female board directors (propfem), which is calculated by dividing total amount of women
directors on the board by the total amount of board directors.
3.4 Control Variables
Omitted-variable bias occurs if one or more variables are left out. But in fact may have a relation with either
dependent variables or independent variables. In line with previous researches, six control variables are
included in this research to reduce the omitted variable bias. These are Board size (bsize), proportion of
independent board directors (propib), average age of board directors (aveage), leverage of the firm (leverage),
industry dummy (indusdum) and year dummy (yeardum) respectively.
Adams and Ferreira (2009) together with Carter et al. (2010) and Christopher, Siantar and Mahajan (2016)
suggest that board size need to be controlled in the regression model. However, there are mixed conclusions
between the board size and company performance based on previous researches. Jackling and Johl (2009)
11
(2015) stated a strong relationship between board size and ROA &ROE, while they did not find any
relationships using Tobin’s Q and CFO/TA as measurements. Moreover, based on the agency theory, Yermak
(1996) argued that board size has a negative impact on financial performance. Hence, in order to reach a more
accurate result, board size were controlled in the regression model.
Independent board directors are defined by the Institutional shareholder services (ISS) as directors of a board
who have no direct links with the company. The causality between independent board directors and
organization performance has already been investigated by several studies, reaching different conclusions. In
their study, Elgaied and Rachdi (2008) presented that independent board directors play a significant role in the
analysis of organization performance measured by ROA. Min and Verhoeven (2013) suggest that independent
board members enhance firm performance through Tobin’s Q and firm productivity. However, Prabowo and
Simpson (2011) found no significant connection when investigating the share of independent board directors
and organization performance. Consistent with Carter et al. (2010), who also examined the US S&P500 firms,
the independent board directors is treated as a control variable in this paper.
In addition to the independent board directors, the third control variable is the average age of board directors,
which is defined by the sum of ages of all board directors divided by the sum of the number of total directors.
According to Vroom and Pahl (1971), older managers are more likely to avoid risk-taking investment. While
younger managers are more willing to seek for risky project with higher return, therefore firms with younger
managers will result in a higher growth in performance than firms with relatively older managers (Hambrik
and Mason, 1984). In Germany, in their study using 149 German firms during the period of 2009-2011,
Eulerich et al. (2014) pointed out that board age is expected to have a negative influence on organization
performance. Furthermore, Nakano and Nguyen (2011) demonstrated that board age has a significantly
negative influence on company performance in Japan. Both of the results suggest that older board may
decrease firm’s performance. In this research, the sample of the average age of board directors for the 138 US
S&P500 firms is roughly sixty-three years old. Therefore board average age is expected to have a negative
relationship with firm’s performance based on the previous literatures discussed above.
The impact from the firm’s financial leverage is another major area of interest in numerous studies. Ilyukhin
12
showed a passive effect of financial leverage on ROA, ROE and operating margin. Besides, Ruland and Zhou
(2005) and Robb and Robinson (2009) also pointed out that financial leverage indeed can help to improve
firm performance. Additionally, Ibhagui (2018) figured out a significant negative influence of using debt on
financial performance when a sample is consisting of small-sized companies. However this effect diminishes
as the size of the firm increases. Finally, they induced a positive relation between leverage and Tobin’s Q for
Nigeria listed companies. Consequently, leverage of the firm (leverage) is included as a control variable, and
it is defined by dividing long-term debt by market value to equity.
Industry dummies are included as control variables. Since the data indicate industry sector for each firm in
WRDS is presented as a code number, it cannot be directly verified which industry each firm belongs to.
Hence, information for industry is hand-collected from the Wikipedia website corresponding with companies
using the Global Industry Classification Standards (GICS). This is an industry classification developed by
MSCI Inc. and Standard& Poor’s (S&P) in 1999 and is used as a benchmark for S&P and MSCI. Therefore,
following GICS, the sample consists of 10 main industry sectors shown in chart 1 below.
Chart 1 Sample Composition by Industry
Sample composition by industry
Industrials Health care Energy Consumer discretionary Consumer staples Materials Information technology Financial Telecommunication services Real estate
13 3.5 Descriptive Statistics
The statistical characteristics regarding to the distribution of female directors on the board between 2010 and
2012 are presented in table I in the Appendix. Since the data for this paper is balanced, the number of
companies for the sample years is constant. The distribution of women board directors is stable, from 8.6014%
in 2010 to 8.6631% in 2012. Although more and more countries introduced regulations for improving the
female representatives on the board in the recent years, male directors still take up most positions in the
boardroom.
Furthermore, when looking at the distribution of women directors in 10 industry sectors (see table II in the
Appendix). Real estate industry has the highest proportion of female directors, with approximately 10.34% of
women directors sitting on the board. In contrast, the energy industry sector is the lowest with only 6.40% of
female directors sitting on the board. This gap for the proportion of women directors in each industry sector
indicates that we should control for industries in the regression model. Otherwise the coefficient of our main
explanatory variable may be biased.
Table 1 below provides key information on variables for board and financial performance of the sample. The
second column illustrates that within the three-year period, there are 414 firm-year observations in the sample.
The mean of the two main performance variables ROA and CFO/TA are 0.2258 and 0.4036 respectively,
which indicates every dollar that firm invested in asset generated approximately 23 cents in income and 40
cents in operation cash flow. The standard deviation of ROA and CFO/TA are 0.2016 and 0.3237, which
means there are no significant outliers among 414 firm-year observations. The percentage of women board
directors ranges from 0% to 55.56%, with an average of 15.94%. The proportion of independent directors
ranges from 44.44% to 100%, with an average of 83.08%. Out of 100 directors, 83 directors are independent.
This number is not surprising in the US. According to Baum (2017) nearly 85% board director members were
independent in 2013 by analyzing the US public firms. Finally, almost every firm has long-term debt to
finance capital. The minimum and maximum leverage ratio is 0 and 2.8908, indicates that liabilities are 0%
14 Table 1 Description of firm characteristics (2010-2012)
Variable N Mean Median Standard deviation min max
ROA 414 .2258 .2027 .2016 -.9142 1.9677 CFO/TA 414 .4036 .3255 .3237 -.0581 2.2036 fem 414 .8937 1 .3086 0 1 propfem 414 .1594 .1429 .0988 0 .5556 bsize 414 10.3164 10 2.5348 6 32 propib 414 .8308 .8571 .0944 .4444 1 aveage 414 63.0390 63 3.1000 54 74.1429 leverage 414 .2832 .1946 .3315 .0000 2.8908
Table III (see Appendix) reports the correlation coefficients of the various variables. Severe multicollinearity
will influence the regression result when there is an obvious relationship between two explanatory variables.
According to (Studenmund, 2011), it is problematic if the value of the correlation coefficient is larger than
0.80. From the data shown below, it seems that the severe multicollinearity does not occur since all the values
of coefficients are very low.
3.6 Regression Models
In order to well-examine the causality of gender diversity and organization performance, two proxies of board
gender diversity (fem and propfem) are used as independent variables, two measures of firm performance
(ROA and CFO/TA) are used as dependent variables. The fundamental model is:
=α + + + + + + 2+ 3+
=α + + + + + + +
=α + + + + + + + +
=α + + + + + + + +
222 Notes: indusdum variable includes Information Technology dummy (ITdum), Consumer Staples (CSdum), Energy dummy (Energydum), Materials
dummy (Materialsdum), Telecommunication Services (TSdum), Consumer Discretionary dummy (CDdum), Industrials dummy (Industrailsdum), Financial dummy (Financialdum) and Real Estate dummy (REdum). Industry dummy equals 1 if the observation is belongs to a specific industry, equals 0 otherwise. To be specifically, we have 10 different industries in our sample, which means we use 9 industry dummies, and 1 reference industry. The results of these 9 industry dummies are all comparing with the reference industry. In our sample, we use the Health Care industry as the reference industry.
3
15
Based on our hypothesis in the literature research, the following hypotheses are tested:
H0: = 0 H1: 0
Where is the coefficient of gender diversity (fem and propfem)
Firstly, we employ the ordinary least squares (OLS) regression model, and then employ the Hausman test to
decide whether to apply the Fixed Effects Regression Model (FEM) or the Random Effects Regression Model
(REM). This is for checking if the results are consistent with OLS. Next, we run robustness check for both
OLS methodology and REM methodology. The list of variables and their definitions are shown in the table IV
in the Appendix.
4. Regression Results and Findings
4.1. OLS Regression Results
The preliminary OLS regression output without being controlled for the industry dummy variable and year
dummy variable is summarized in Table 2. We can notice that all of our regression models present a negative
relationship between our main explanatory variables and dependent variables. Model 1, model 3 and model 4
all have a significantly negative coefficient, meaning that gender diversity result in a significant negative
impact on firm’s performance at 90%, 99% and 95% level respectively. Moreover, for the control variables,
propib is significantly as well as positively associated with the ROA for model 1 and model 2 respectively.
However, this significant relationship disappeared when using CFO/TA as the performance measure. Leverage
however, seems to have a negative impact on the performance measure, with the significant coefficients of
-0.2190(0.0280) and -0.2200(0.0282) for model 1 and model 2. The relationship is negative since debt has the
potential to decrease the revenues as more cash is used to service that debt. For other control variables, there is
16
Table2. OLS Regression for Gender Diversity and Firm’s Performance without Including Industry Dummies and Year Dummies.
Notes: In parentheses: standard error. *, **, *** denotes p < 0.1, p < 0.05, p <0.01 respectively.
4.2 OLS Regression Results after Adding More Control Variables and Random Effect Model Results
For the purpose of enhancing reliability and accuracy of the results, we apply two methods to check our
results. The correlation between independent variables and unobserved variables may disconcert both
explanatory variables and dependent variables. This will cause omitted variable bias of our output. As
illustrated in section 3, the percentage of women directors varies a lot across different industries. Hence, we
add industry dummies and year dummies in the regression model as our control variables as well as use
Before robust After robust
Model 1 ROA Model 2 ROA Model 3 CFO/TA Model 4 CFO/TA Model 1 ROA Model 2 ROA Model 3 CFO/TA Model 4 CFO/TA fem -.701** (.0318) - -.2355*** (.0539) - -.0701 ( .0525) - -.2355*** (.0857) - propfem - -.0090 (.1010) - -.3188* (.1736) - -.0090 (.1167) - -.3188 (.2078) bsize .0021 (.0037) .0005 (.0037) .0009 (.0063) -.0036 (.0064) .0021 (.0033) .0005 (.0034) .0009 (.0052) -.0036 (.0051) propib .2691*** (.1013) .2185** (.1030) .1153 (.1720) .0265 (.1771) .2691*** (.0938) .2185** (.0904) .1153 (.1626) .0265 (.1551) aveAGE .0014 (.0031) .0027 (.0032) -.0027 (.0053) -.0008 (.0055) .0014 (.0032) .0027 (.0035) -.0027 (.0049) -.0008 (.0046) leverage -.2190*** (.0280) -.2200*** (.0282) -.0159 (.0475) -.0241 (.0485) -.2190*** (.0232) -.2200*** (.0237) -.0159 (.0346) -.0241 (.0359) constant .0153 (.2194) -.0654 (.2209) .6830* (.3723) .5253 (.3797) .0153 (.2172) -.0654 (.2313) .6830* (.3586) .5253 (.3441) Industry dummy NO NO NO NO NO NO NO NO Year dummy NO NO NO NO NO NO NO NO N 414 414 414 414 414 414 414 414 R-square .1465 .1363 .0466 .0102 .1465 .1363 .0466 .0102 Adjusted R-square .1360 .1257 .0349 -.0019 - - - -
17 Random Effect Regression to diminish omitted variable bias.
Table 3 below depicts that including industry dummies and year dummies weakens the significance of the
relationship between the fem and ROA, propfem and CFO/TA. The coefficient of propfem turns into
insignificant while the coefficient for fem still remains significant compared to the result in table 2 when there
is no industry dummies and year dummies included. Concerning the control variables, there is no major
change except for the coefficient of leverage. The relationship for leverage and CFO/TA convert to significant
compared to table2, which indicates that a high leverage ratio will lead to a decrease in company performance.
As for the dummy variables, all of the industry dummies listed in table 3 are comparing with a randomly
selected reference industry (health care industry). For instance, as illustrated in table 3, all the coefficients for
consumer staples industry dummy, telecommunication services industry dummy and real estate industry
dummy are positive and significant, which reveals that firms in those industries performed better than firms in
health care industry. Regarding to the year dummies, there is no significant evidence showing that firm
18
Table3. OLS Regression for Gender Diversity and Firm’s ROA and CFO/TA Including Industry Dummies and Year Dummies.
Notes: In parentheses: standard error. *, **, *** denotes p < 0.1, p < 0.05, p <0.01 respectively.
Before robust After robust
Model 1 ROA Model 2 ROA Model 3 CFO/TA Model 4 CFO/TA Model 1 ROA Model 2 ROA Model 3 CFO/TA Model 4 CFO/TA fem -.0611** ( .0316) - -.1579*** (.0478) - -.0611 ( .0521) - -.1579** (.0718) - propfem - -.0489 (.1029) - -.1627 (.1572) - -.0489 (.1075) - -.1627 (.1662) bsize -.0031 ( .0036) -.0044 (.0036) -.0026 (.0055) -.0058 (.0054) -.0031 ( .0042) -.0044 (.0046) -.0026 (.0047) -.0058 (.0045) propib .2586*** ( .10) .2344** (.1023) .0692 (.1514) .0164 (.1562) .2586*** ( .0902) .2344*** (.0927) .0692 (.1350) .0164 (.1409) aveAGE .0034 ( .0031) .0039 (.0031) -.0076 (.0046) -.0064 (.0047) .0034 ( .0036) .0039 (.0038) -.0076* (.0040) -.0064* (.0038) leverage -.2375*** ( .0276) -.2389*** (.0278) -.1425*** (.0418) -.1466*** (.0424) -.2375*** ( .0272) -.2389*** (.0272) -.1425*** (.0480) -.1466*** (.0492) ITdum -.0113 ( .0389) -.0048 (.0390) -.0447 (.0590) -.0288 (.0595) -.0113 ( .0255) -.0048 (.0247) -.0447 (.0403) -.0288 (.0391) CSdum .1463*** ( .0350) .1484*** (.0357) .0934* (.0530) .1012* (.0546) .1463*** ( .0275) .1484*** (.0290) .0934** (.0378) .1012*** (.0387) Energydum .0469 ( .0343) .0601* (.034) .3460*** (.0519) .3777*** (.0520) .0469 ( .0380) .0601* (.0366) .3460*** (.0579) .3777*** (.0633) Materialsdum -.0468 (.0353) -.0470 (.0356) -.0462 (.0535) -.0490 (.0543) -.0468** (.0196) -.0470** (.0198) -.0462 (.0287) -.0490* (.0281) TSdum .1468*** (.0521) .1456*** ( .0523) .7157*** (.0789) .7130*** (.0799) .1468* (.0767) .1456** ( .0769) .7157*** (.1518) .7130*** (.1528) CDdum -.0014 (.033) -.0006 (.0332) -.0044 (.0501) -.0021 (.0507) -.0014 (.0253) -.0006 (.0255) -.0044 (.0389) -.0021 (.0382) industrialsdum -.0366 (.0304) -.0367 (.0306) -.0454 (.0460) -.0462 (.0467) -.0366* (.0203) -.0367* (.0209) -.0454 (.0342) -.0462 (.0351) Financialdum .1728*** (.0453) .1780*** (.0459) .0562 (.0686) .0670 (.0700) .1728** (.0836) .1780** (.0870) .0562 (.0706) .0670 (.0600) REdum .1694** (.0757) .1626** (.0763) .4334*** (.1146) .4133*** (.1165) .1694** (.0691) .1626** (.0691) .4334*** (.0896) .4133*** (.0891) 2011dum .0051 (.0213) .0056 (.0213) .0179 (.0322) .0193 (.0326) .0051 (.0231) .0056 (.0231) .0179 (.0318) .0193 (.0325) 2012dum .0007 (.0213) -.0000 (.0214) .0260 (.0323) .0246 (.0327) .0007 (.0233) -.0000 (.0236) .0260 (.0320) .0246 (.0324) constant -.0759 (.2154) -.1239 (.2157) .9402*** (.3266) .8250** (.3295) -.0759 (.2154) -.1239 (.2337) .9402*** (.3266) .8250** (.2922) N 414 414 414 414 414 414 414 414 R-square 0.2675 0.2610 0.3477 0.3316 0.2675 0.2610 0.3477 0.3316 Adjusted R-square 0.2380 0.2312 0.3214 0.3047 - - - -
19
Random Effects Regression Model (REM) and Fixed Effects Regression Model (FEM) are both efficient
analytical tools in dealing with panel data as well as help to control the hidden heterogeneity. FEM is used
when the mean of groups are fixed. Besides, it allows missing information on the hidden individual effects to
correlate with the independent variables. Opposite from FEM, the mean of groups in REM are random. When
individual effects have no significant correlation with the parameters in the model, REM is more efficient
(Gallego et al., 2009).
We first apply the Hausman Test to check whether Random Effects Regression is more suitable than Fixed
Effects Regression. From the Hausman test, the results show that the p-values of our four regression models
are 0.3368, 0.3638, 0.8648 and 0.8917 respectively. Due to the p-values are all larger than 0.05 at the 95%
level (see Appendix V), REM is more appropriate than FEM for our sample.
Table 4 displays the output from REM, the result seems in conformity with the result in table 3. The data
suggest that the presence of women significantly lowers the performance of the company, while the
percentage of the women directors has no significant influence on performance of the company. Furthermore,
by applying REM, we get significantly higher R-squares for all models compared to the OLS with industry
20
Table4. Random Effects Regression Model for Gender Diversity and Firm’s ROA and CFO/TA
Notes: In parentheses: standard error. *, **, *** denotes p < 0.1, p < 0.05, p <0.01 respectively.
Before robust After robust
REM 1 ROA REM 2 ROA REM 3 CFO/TA REM 4 CFO/TA REM 1 ROA REM 2 ROA REM 3 CFO/TA REM 4 CFO/TA fem -.0645* ( .0340) - -.1228*** (.0385) - -.0645 ( .0334) - -.1228 (.0824) - propfem - -.0274 (.1186) - -.1978 (.1487) - -.0274 (.1285) - -.1978 (.1779) bsize -.0027 ( .0045) -.0040 (.0045) .0021 (.0065) .0005 (.0065) -.0027 ( .0053) -.0040 (.0052) .0021 (.0075) .0005 (.0070) propib .2365** ( .1134) .2164* (.1157) -.0018 (.1388) .0107 (.1410) .2365** ( .1053) .2164** (.1102) -.0018 (.1288) .0107 (.1323) aveAGE .0003 ( .0037) .0010 (.0037) -.0107** (.0051) -.0105* (.0052) .0003 ( .0029) .0010 (.0031) -.0107** (.0052) -.0105* (.0054) leverage -.2228*** ( .0341) -.2247*** (.0343) -.0893* (.0486) -.0949* (.0491) -.2228*** ( .0328) -.2247*** (.0327) -.0893** (.0448) -.0949** (.0459) ITdum -.0147 ( .0467) -.0059 (.0515) -.0447 (.0932) -.0299 (.0944) -.0147 ( .0339) -.0059 (.0335) -.0447 (.0576) -.0299 (.0564) CSdum .1418** ( .0467) .1429*** (.0474) .0822 (.0860) .0928 (.0875) .1418*** ( .0426) .1429*** (.0435) .0822 (.0599) .0928 (.0597) Energydum .0491 ( .0452) .0652 (.0451) .3532*** (.0819) .3758*** (.0828) .0491 ( .0458) .0652 (.0454) .3532*** (.0953) .3758*** (.0990) Materialsdum -.0497 (.0472) -.0490 (.0475) -.0553 (.0873) -.0584 (.0886) -.0497* (.0280) -.0490* (.0275) -.0553 (.0418) -.0584 (.0415) TSdum .1401** (.0694) .1390** ( .0697) .6897*** (.1275) .6894*** (.1291) .1401 (.0921) .1390 ( .0919) .6897*** (.2348) .6894*** (.2383) CDdum -.0040 (.0441) -.0029 (.0443) -.0130 (.0815) -.0092 (.0825) -.0040 (.0389) -.0029 (.0394) -.0130 (.0618) -.0092 (.0604) industrialsdum -.0344 (.0406) -.0338 (.0408) -.0421 (.0751) -.0423 (.0761) -.0344 (.0310) -.0338 (.0318) -.0421 (.0543) -.0423 (.0556) Financialdum .1705*** (.0604) .1778*** (.0610) .0448 (.1112) .0477 (.1130) .1705** (.0737) .1778** (.0782) .0448 (.1141) .0477 (.1091) REdum .1659* (.1011) .1609** (.1019) .4184** (.1873) .4002** (.1901) .1659** (.0694) .1609** (.0683) .4184*** (.1204) .4002*** (.1197) 2011dum .0049 (.0171) .0053 (.0172) .0165 (.0157) .0182 (.0159) .0049 (.0207) .0053 (.0207) .0165 (.0167) .0182 (.0178) 2012dum .0016 (.0172) -.0006 (.0173) .0247 (.0159) .0246 (.0161) .0016 (.0191) -.0006 (.0195) .0247 (.0165) .0246 (.0166) constant .1328 (.2583 -.0615 (.2595) 1.1044*** (.3445) 1.0166*** (.3528) .1328 (.2056) -.0615 (.2595) 1.1044*** (.3445) 1.0166*** (.3363) N 414 414 414 414 414 414 414 414 R-square 0.2653 0.3818 0.3784 0.3631 0.2653 0.2589 0.3416 0.3259 Wald chi2 84.00 79.80 87.00 76.61 97.96 100.43 51.00 49.95
21 4.3 Robustness Check
Non-robust estimator is calculated by using the sample mean, which is useful for symmetric distributions.
While robust estimator is calculated by the sample median and is helpful when there are extreme observations.
From table 1, we can see that the spreads of ROA and CFO/TA are extremely large, with the lowest value of
-91.42% and -5.81%, the highest value of 196.77%, 220.36% for ROA and CFO/TA respectively. Additionally,
the difference of the mean and median for ROA, CFO/TA and fem from table 1 also suggests that there might
be some outliers. This indicates that the answer from OLS and REM might be misleading in our sample. To
further ensure the reliability of the relationship concluded from OLS and REM, we also conduct robustness
checks (see the right four columns in table 2, table 3 and table 4) for each model. It is clear that after
concluding robustness checks, all the coefficients are presented insignificant in the random effect model and
only one estimator is shown to be significant in OLS model. On the whole, based on the robustness checks
together with the preliminary OLS model without industry dummies and year dummies, OLS model with
industry dummies and year dummies as well as random effect model, we fail to find any significant link for
gender diversity and organization performance.
5. Conclusion
The failures of corporate governance in recent decades have led politicians and scholars to focus on the
composition of boardrooms in organizations. Among different researches examining the corporate governance
topics, gender diversity in the boardroom stands out particularly due to the trend that more and more countries
implemented regulations requiring female director quotas in organizations. However, it still remains a
question that whether it can enhance organization performance or not through requiring female director quotas.
There are mainly two arguments from the existing literatures. Agency theory states that more gender diverse
board leads to a better firm performance through the function of board monitoring. Social psychology theory
claims that gender diversity can result in more conflicts between board members thus has a negative impact on
performance of companies. Moreover, results from empirical researches are inconclusive as well. Drawing on
the mixed evidence from both theoretical perspective and empirical perspective, the hypothesis of this study is
that requiring a more gender diverse board will not result in a better performance of companies.
22
results from OLS, we ran another model named REM. It showed that the result is consistent with the results
from OLS. We also did robustness checks for the outputs from OLS and REM. It gave us more precise results
than non-robust results since robustness checks using the median of the sample, which exclude large outliers.
Overall, from our regression methodology using the presence of female directors and proportion of female
directors as main independent variables, ROA and CFO/TA as dependent variables, we conclude that there do
not appear any changes in firm performance that are resulted from the perspective with and without gender
diversity of the board. A board with more gender diversity does not lead to a better performance of the
company.
Accordingly, the regulations and policies that requiring a certain quota for the amount of women directors in
organizations might not be directly associated with performance of corporations. Instead, the trend toward
female quotas in the boardrooms might be pushed by publics. In their study, Marinova, Plantenga and Remery
(2010) state that the balance representation of female in top management is connected with social justice
rather than merely a matter of firm performance. On the other hand, the imbalance between the labor force
and the low employment rate might be another reason that was considered by the governments on introducing
female quotas policies. In conclusion, when countries introduce policies to mandate a certain female quota for
companies, they should take into account the impact of the policy on the economic benefits of the company
seriously.
This paper contributes a better insight of the gender diversity topic in corporate governance. First of all, we
use a combination of OLS regression and REM regression as our methodologies, which is more rigorous than
previous papers with just one regression methodology. Moreover, we included 9 industry dummies to reduce
the omitted variables bias. In most of the previous studies, although 2SLS or OLS regression model were
conducted, the industry dummies were hardly employed as control variables, which might lead to a biased
result.
One limitation of this research is that we only considered the US S&P 500 companies. For different countries
with different cultural and policies, the result might vary. Another restriction is that the impact of gender
diversity on company performance might be lagged (She, Sainter and Mahajan, 2016). Which means the fact
23
research should use lagged firm’s performance variables to check for the result. Additionally, listed firms may
be more likely to hire more female board members. For future research, it seems more persuasive if non-listed
24
6. Reference List
Adams, R., & Ferreria, D. (2009). Women in the Boardroom and Their Impact of Governance and Performance.
Journal of Financial Economics, (94) , pp.291-309.
Ararat,M., Aksu, M., & Cetin,A. T. (2015). How Board Diversity Affects Firm Performance in Emerging Markets: Evidence on Channels in Controlled Firms. Corporate Governance: An International Review, 2015, 23(2): 83-103.
Baum, H. (2017). The Rise of the Independent Director: A Historical and Comparative Perspective. Max Planck
Private Law Research Paper, No.16/20.
Campbell, K. & Minguez-Vera, A. (2008). Gender Diversity in the Boardroom and Firm Financial Performance.
Journal of Business Ethics, 83: 435-451.
CĂRĂUȘU, D.N. (2015). Monitor and Control in Companies: An Agency Theory Approach. Journal of Public
Administration, Finance and Law, Special Issue 2/2015.
Daniel, C.M.L., Roberts, H., Rosalind H. W. (2015). Board Gender Diversity and Firm Performance: Empirical Evidence from Hong Kong, South Korea, Malaysia and Singapore. Pacific-Basin Finance Journal, 35(2015) 381-401.
De Cremer, D., Van Dick, R., & Murnighan, J. K. (2011). Social Psychology and Organizations. Taylor and
Francis. DOI: 10.4324/9780203846957.
Elgaied, M. & Rachdi, H. (2008). Composition, Structure of Board of Directors and Performance: The Case of American Firms.
Eulerich, M., Velte, P., & Uum, C. V. (2014). The impact of management board diversity on corporate performance - an empirical analysis for the German two-tier system Problems and Perspectives in Management, 12(1), 25-39.
Gul, F. A., Srinidhi, B., & Ng, A. C. (2011). Does Board Gender Diversity Improve the Informativeness of Stock Prices? Journal of Accounting and Economics, 51(2011) 314-338.
Hambrick, D. C., & Mason, P. A. (1984). Upper Echelons: The Organization as a Reflection of Its Top Managers.
Academy of Management Review, 9(2), 193-206.
Ibhagui, O.W., (2018). Leverage and Firm Performance: New Evidence on the Role of Firm Size. North American Journal of Economics and Finance (2018).
Ilyukhin, E. (2015). The Impact on Financial Leverage on Firm Performance: Evidence from Russian. Journal of
Corporate Finance Research, Vol. 9, No. 2, pp. 24-36, 2015.
Jackling, B. J. & Johl, S. (2009). Board Structure and Firm Performance: Evidence from India’s Top Companies.
Corporate Governance: An International Review, 17: 492-509.
Kang, H., Cheng, M., & Gray, S. J. (2007). Corporate Governance and Board Composition: Diversity and Independence of Australian Boards. Corporate Governance: An International Review, 15(2), 194–207.
Marinova, J., Plantenga, J. & Remery, C. (2010). Gender Diversity and Firm Performance: Evidence from Dutch and Danish Boardrooms. Tjalling C. Koopmans Research Institute, discussion paper series nr: 10-03.
Min, B-S. & Verhoeven, P. (2013). Outsider Board Activity, Ownership Stucture and Firm Vlue: Evidence from Korea. International Review of Finance, 13:2, 2013: pp. 187-214.
Mogbogu, O. (2016). Women on the Board of Directors and their Impact on the Financial Performance of a Firm: An Empirical Investigation of Female Directors in the United States Technology Sector. These and
Dissertation.625.
Nakano, M., & Nguyen, P. (2011). Do older boards affect firm performance? An empirical analysis based on Japanese firms.
25
Prabowo, M. & Simpson, J. (2011). Independent Directors and Firm Performance in Family Controlled Firms: Evidence from Indonesia. Asian Pacific Economic Literature, 25(1), 2011: pp. 121-132.
Ramly, Z., Chan, S-G., Mustapha, M. Z., & Sapiei, N. S. (2015). Gender Diversity, Board Monitoring and Bank Efficiency in ASEAN-5. South East Asia Journal of Contemporary Business, Economics and Law, Vol.7, Issue 1(Aug.) ISSN 2289-1560.
Rose, C. (2007). Does Female Board Representation influence Firm Performance? The Danish Evidence.
Corporate Governance: An International Review, 15: 404-413.
Smith, N., Smith, V., & Verner, M. (2006). Do Women in Top Management Affect Firm Performance? A Panel Study of 2500 Danish Firms. International Journal of Productivity and Performance Management, 55: 569-593. She, C., Sainter, D. & Mahajan, A. (2016). Effects of Board Gender Diversity on Firm Performance and Director
Compensation in India.
Vroom, V. H., & Pahl, B. (1971). Relationship between age and risk taking among managers. Journal of Applied
Psychology, 55(5), 399-405.
Williams, K. Y. & O’Reilly, C. A., III (1998). The Complexity of Diversity: A Review of Forty Years of Research. In E. Gruenfeld & M. Neale (Eds.), Research on Managing in Groups and Teams. Thousand Oaks, CA: Sage. Yap, I. L.-K, Chan S.-G.,& Zainudin,R. (2017). Gender Diversity and Firms’ Financial Performance in Malaysia.
Asian Academy of Management Journal of Accounting and Finance, 13(1), 41-62.
Yermack, D. (1996). Higher Market Valuation of Companies with a Small Board of Directors. Journal of Financial
26
7. Appendix
Table I Gender diversity in each of the sample years
Year Number of companies
Number of female board dire tors
Total amount of board directors
Proportion of female board directors
2010 138 123 1430 8.6014%
2011 138 122 1418 8.6037%
2012 138 125 1423 8.6631%
Table II Directors data by industry sector during the period of 2010-2012
Industry sector: Total amount of female board directors
Total amount of board directors
Proportion of female directors
Real estate 6 58 10.3448% Telecommunication services 15 152 9.8684% Health care 61 633 9.6367% Industrials 69 722 9.5568% Materials 40 436 9.1743% Consumer discretionary 51 565 9.0265% Consumer staples 45 505 8.9109% Information technology 28 368 7.6087% Financial 18 254 7.0886% Energy 37 578 6.4014%
Table III Correlation coefficients
ROA CFO/TA fem propfem bsize propib aveage leverage ROA 1.0000 CFO/TA .04469 1.0000 fem -.0851 -.2166 1.0000 propfem .0293 -.0940 .5573 1.0000 bsize .0251 -.0325 .1669 .0590 1.0000 propib .0683 -.0170 .2183 .2604 -.0938 1.0000 aveage .0406 .0165 -.1868 -.2433 .0024 -.0173 1.0000 leverage -.3531 -.0173 .0157 .0455 -.0804 .0875 -.0039 1.0000
27 Table IV: Summary of variables and definitions
Variable name Definitions
Dependent variables:
Return on asset (roa) Net income divided by total asset. Cash flow from operation to total
asset(cfo/ta)
Cash flow from operations divided by total asset.
Main explanatory variables:
Gender diversity dummy(fem) Equals 1 if at least one female in the firm’s boardroom, equals 0 otherwise. Proportion of female on
board(propfem)
Number of female directors on board divided by the total number of the firm’s board directors.
Control variables:
Board size (bsize) Total number of directors in the boardrooms. Proportion of independent
board(propib)
Number of independent board members divided by the total number of board directors.
Average age of board directors(aveage)
Sum of ages of all directors divided by the total number of board directors Leverage of the firm(leverage) Long-term debt divided by the market value of equity ratio
Information technology industry dummy (ITdum)
ITdum=1 indicating a firm is in the information technology industry sector, o otherwise.
Consumer staples industry dummy (CSdum)
CSdum=1 indicating a firm is in the consumer staples industry sector, 0 otherwise. Energy industry dummy
(Energydum)
Energydum=1 indicating a firm is in the energy industry sectors, 0 otherwise. Materials industry dummy
(Materialsdum)
Materialsdum=1 indicating a firm is in the materials industry sectors, 0 otherwise. Telecommunication services
dummy (TSdum)
TSdum=1 indicating a firm is in the telecommunication industry sectors, 0 otherwise.
Consumer discretionary industry dummy (CDdum)
CDdum=1 indicating a firm is in the consumer discretionary industry sector, 0 otherwise.
Industrials industry dummy (Industrialsdum)
Industrialsdum=1 indicating a firm is in the industrial industry sector, 0 otherwise. Financial industry dummy
(Financialdum)
Financialdum=1 indicating a firm is in the financial industry sector, 0 otherwise. Real estate industry dummy
(REdum)
REdum=1 indicating a firm is in the financial industry sector, 0 otherwise. Year 2011 dummy (2011dum) Equals 1 if the observation is in 2011, otherwise 0.
28
Table V: Hausman test between Random Effect Model and Fixed Effect Model
Hausman test (REM vs FEM) ROA Female dummy ROA Proportion of female CFO/TA Female dummy CFO/TA Proportion of female Prob> Chi2 0.3368 0.3638 0.8648 0.8917