**Preface to "Bank Management, Finance and Sustainability"**

Sustainable banking incorporates environmental, social and governance (ESG) criteria into traditional banking activities and makes ESG benefits a key objective of the new banking business models. Banks'managerial and investment choices are made taking into account not only the aspects of risk and return, but also their social and environmental impacts.

Sustainability represents an opportunity for banks as it contributes to improvements in trust in the banking system. However, sustainable business models must be financially viable so that they can have a positive impact on banks'profitability, stimulating the long-term growth and resilience of the banking industry and overall financial stability.

Banks are widely acknowledged as playing a crucial role in achieving the Sustainable Development Goals (SDGs), as they can promote responsible investments and integrate environmental and social criteria into lending and investment strategies. Financial intermediaries can support projects and activities that create a measurable positive economic, social and environmental impact by providing easier access to capital. Furthermore, they can have an active role in improving the financial awareness, inclusion and resilience of the most vulnerable individuals in society.

The present volume collects the contributions selected for publication in the Special Issue entitled *Bank Management, Finance and Sustainability* of the journal *Sustainability*, for which we served as guest editors. This collection includes both empirical and theoretical studies, covering a wide range of themes related to sustainable banking and finance.

> **David Aristei, Manuela Gallo** *Editors*

### *Article* **Using Environmental, Social, Governance (ESG) and Financial Indicators to Measure Bank Cost Efficiency in Asia**

**Hai-Yen Chang , Lien-Wen Liang \* and Yu-Luan Liu**

Department of Banking & Finance, Chinese Culture University, Taipei City 11114, Taiwan; irischang1014@gmail.com (H.-Y.C.); 4334031allen@gmail.com (Y.-L.L.)

**\*** Correspondence: liang.lienwen@gmail.com; Tel.: +886-2-2861-0511 (ext. 36237)

**Abstract:** Environmental, social, and governance (ESG) practices have been used as non-financial indicators to measure bank performance worldwide in the last decade. The United Nations (UN) has specified 17 Sustainable Development Goals (SDGs) for the implementation of these ESG concepts. However, it remains unclear whether the costs of ESG have exceeded the benefits. The purpose of this study is to examine the impact of ESG on the cost efficiency of developed and developing Asian banks using a two-step approach comprising stochastic frontier analysis (SFA) and stochastic metafrontier analysis (SMF). The data sample from 2015 to 2018 is separated into two groups: 60 Asian developed economies and 85 developing economies. The results show that banks in the developed Asian economies become more cost-efficient through environmentally friendly activities. The banks in the developing Asian economies increase their cost efficiency by socially responsible activities and improved governance. Moreover, banks in the developed Asian economies outperformed those in the developing Asian economies in terms of technology gap ratio (TGR) and metafrontier cost efficiency (MCE). The results of this study benefit not only investors and bank managers but also the entire banking sector and the world economy.

**Keywords:** ESG; sustainable development goals (SDGs); bank efficiency; bank cost; stochastic frontier analysis; stochastic metafrontier analysis

#### **1. Introduction**

Financial institutions play an important role in national and international trade, and in the process of globalization. Banks serve as intermediary institutions for intermediation, channeling funds from savers to borrowers to enable business developments and investments [1]. Financial institutions are also crucial in the international markets because banks support companies in conducting international trade in which foreign exchange and letters of credit are often needed. Furthermore, banks facilitate the globalization process. Banks provide their customers with convenient and low-cost ways, such as an internet banking system, to pay and track funds [2]. In addition, banks assist multinational firms in achieving foreign direct investments and listing their stocks in overseas countries, thus helping these corporations to expand globally.

Karray and Chichti [3] claimed that banks must achieve optimal performance to support regional development and strengthen their function as intermediary institutions. Efficiency is an important indicator of bank performance which is measured mostly by financial data. Battese and Coelli [4] began to use bank efficiency to measure bank performance, followed by other researchers [5–7]. Efficiency is measured by comparing inputs such as cost of borrowed funds, cost of tangible assets and labor, against outputs such as loans, income-generating assets, and deposits. It is essential to measure bank efficiency because the failure of the bank or inadequate cash due to loan collection problems jeopardizes the economic lives of millions of individuals [8]. Miralles-Quiros et al. [9] expected banks to play a dual role concerning the sustainability of the financial sector with one involving

**Citation:** Chang, H.-Y.; Liang, L.-W.; Liu, Y.-L. Using Environmental, Social, Governance (ESG) and Financial Indicators to Measure Bank Cost Efficiency in Asia. *Sustainability* **2021**, *13*, 11139. https://doi.org/ 10.3390/su132011139

Academic Editors: David Aristei, Manuela Gallo and Olaf Weber

Received: 11 August 2021 Accepted: 5 October 2021 Published: 9 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

financial performance and the other corporate governance and social responsibilities at a strategic level [10,11]. Banks must not only focus on profitability but also on corporate governance [12,13].

In 2005, the United Nations (UN) proposed the Principles for Responsible Investment (PRI) which highlighted environmental, social, and governance (ESG) issues [14]. In 2015, the UN further announced the 17 Sustainable Development Goals (SDGs) of the 2030 agenda for sustainable development [15]. As intermediary institutions, banks primarily use the funds from their depositors to conduct banking business; therefore, banks must monitor the cost-benefit effect more prudently in order to safeguard their depositors' money. Banks' efficiency in using such funds is a crucial indicator of bank sustainability in the long run. Although it is important for banks to execute ESG practices, the literature examining whether banks generate more revenue as a result of implementing ESG programs to cover the associated expenditures is scant. Consequently, it remains unclear whether the implementation and disclosure of ESG activities increased or decreased bank cost efficiency. Moreover, prior research using regression models to analyze bank financial performance also mostly focused on the performance of banks in the developed countries, such as the U.S. and Europe. Few studies discussed the efficiency of the rapidly growing banking industry in Asia. This study fills such a research gap.

The purpose of this study is to examine the impact of ESG on the cost efficiency of developed and developing Asian banks using a two-step approach comprising stochastic frontier analysis (SFA) developed by Battese and Coelli [4] and stochastic metafrontier analysis (SMF) proposed by Huang et al. [6]. We divided the Asian banks into two groups based on the International Monetary Fund (IMF) definitions. One group included 60 banks in five developed Asian economies. The other group included 85 banks in 13 developing Asian economies. In the first step, we investigated the impact of ESG and bank-specific indicators on the efficiency of the two groups of banks. In the second step, we studied the impact of macroeconomic factors on bank efficiency and compared the technology gap ratio (TGR) and metafrontier cost efficiency (MCE) of the two groups of banks.

The results indicate that environment variables significantly increased bank cost efficiency in the developed economies, but not in the developing economies. The social and governance variables increased bank cost efficiency in the developing economies, but not in the developed economies. In addition, banks in the developed economies exhibited higher cost efficiency than their counterparts in the developing economies in Asia.

The study contributes to the literature in three ways. First, to the best of our knowledge, this is the first study to incorporate the 17 SDGs to examine bank cost efficiency. Second, this study compared the cost efficiency of banks in developed and developing economies in the fast-growing Asian region, which was rare in the literature. Third, this is the first study to apply SFA and SMF to analyze the impact of ESG on Asian bank cost efficiency. The results of this study benefit not only the bank managers and investors but also the entire banking sector.

#### **2. Literature Review**

Dahl et al. [16] elucidated that the western countries led the development of the modern banking industry around the world in terms of size, growth, business models, and innovation. However, in the last decade, Asian banks have expanded quickly and developed innovative products designed to satisfy the needs of a larger group of customers. This change reflects not only the increasingly important role of Asian banks in global trade and economic growth but also Asia's leadership in delivering new technologies and business models. From 2016 to 2021, the top five largest banks by asset size (Industrial & Commercial Bank of China, China Construction Bank, Agricultural Bank of China, Bank of China, Japanese Mitsubishi Bank) were in Asia [17].

In addition, more than 40 of the world's 100 largest banks based on asset size are Asian, accounting for approximately 50% of global market capitalization [16]. Moreover, Asia was the largest regional banking market in the world for the last decade. On the

whole, the Asian banks generated pretax profits exceeding \$700 billion, which accounted for 37% of the global banking profit. With the rise of the income level of the middle-class group in Asia, Dahl et al. [16] expected the financial assets held by households in Asia to reach \$69 trillion by 2025, representing approximately 75% of the global amount.

Despite the growth of the financial sector, banks are affected by systemic risks, which arise when a set of adverse events in the markets threatens to disrupt the bank functions of intermediation [2]. The systemic risk in the economy, such as a decline in the gross domestic product (GDP) and high unemployment rate, could lead to the instability of the banking system. For instance, a high unemployment rate is likely to aggravate the default rate of bank loans, which impedes further bank lending and tightens bank credit policy, leading to a recession and widespread failure of loan payment. Such results hamper the bank's role in facilitating economic growth.

#### *2.1. ESG*

The PRI announced by the UN in 2005 highlighted the influence of ESG issues on the performance of investment portfolios. Subsequently, guidelines for environmental stewardship, social responsibility, and corporate governance gradually directed the evaluation of the firms. Environmental stewardship refers to the firm's actions concerning the natural environment with a focus on the reduction in waste and pollution, greenhouse gas emissions, and climate change [14]. Social responsibility is similar to the concept of corporate social responsibility (CSR). Social responsibility means fair and beneficial business practices for labor, respect for human rights, the establishment of a safe environment, and service to the community [14]. Governance refers to the proper management of the company in addition to economic prosperity. Firms should formulate appropriate policies, especially related to business ethics, disclosure of information, and board composition to govern their business operations [18].

In 2015, the UN announced the 17 SDGs which can be divided into five categories: people (no poverty, zero hunger, good health and well-being, quality education, gender equality), planet (clean water and sanitation, affordable and clean energy, climate action, life below water, life on land), prosperity (decent work and economic growth, industry, innovation, and infrastructure, responsible consumption and production), peace (reduced inequality, sustainable cities and communities, peace, justice, and strong institutions), and partnership (entering into partnerships to reach the goals) [15].

In 2018, the UN released a report entitled "Integrating the SDGs into Corporate Reporting: A Practical Guide", to help corporations to set objectives and disclose their ESG activities. In 2018, approximately 40% of the world's 250 largest companies reported SDGs and included the global goals in their annual reports [19]. The achievements of the SDGs by all nations would create new opportunities and an increase in efficiency for an estimated \$12 trillion in four economic systems: food and agriculture, cities, energy and materials, and health and well-being [19].

In 2018, the UN established a special task force to analyze the relationship between ESG investing and returns. The UN also recommended aligning financial systems with sustainable development. Prior researchers began to examine the impact of the disclosure of ESG activities on bank performance.

#### *2.2. Environmental*

The stakeholder theory explains the dynamics of ESG and shareholder value [20]. Shareholders are the primary stakeholders in a firm; hence companies should perform business activities to maximize shareholder interests. Therefore, negative consumer attitudes toward a firm's products and services or non-compliance with government regulations and environmental practices may decrease shareholder value [21].

However, ESG may increase bank expenditures due to the additional investment requirements in environmental activities, such as reduction in carbon emissions, use of renewable energy, prevention of air and water pollution, planting trees, etc. Many banks

implemented environmental activities as a result of government requirements that need to be considered when evaluating the performance of listed firms [18]. The question of whether over-investment in environmental activities leads to a favorable financial position remains unanswered in the literature [22].

Prior studies indicated that the impact of environmental activities on bank performance varied. Some researchers found that environmentally friendly activities improved a bank's financial performance. In other words, banks that disclosed efforts of minimizing carbon emissions generated greater profits. Such disclosure also increased the bank's market value [14]. Buallay [23] studied the performance of 235 banks from 2007 to 2016 and ascertained that environmental disclosure positively affected the banks' return of assets (ROA) and market value as measured by Tobin's Q. Similarly, Miralles-Quirós et al. [24] studied 51 banks in the U.S. and Europe from 2002 to 2015. These authors claimed that environmental endeavors positively influenced the banks' market value and earnings per share (EPS). Crespi et al. [18] examined ESG activities and financial performance using data for 727 financial firms from 22 developed countries from 2006 to 2017. The results revealed that a higher environmental score led to increased profitability.

In contrast, other studies found that the disclosure of environmental activities had a negative impact on banks. For example, Forgione et al. [5] used a one-step SFA method to examine ESG and bank efficiency in primarily developed economies from 2013 to 2017. They found the disclosure of environmental activities reduced bank efficiency. Similarly, Dell'Atti et al. [25] investigated the impact of the banking industry during the 2008 subprime mortgage crisis by studying the correlation between bank reputation and economic performance. The results suggested that environmental activities had a negative but insignificant effect on reputation and bank performance. In a study by Di Tommaso and Thornton [26], the European banks that received high ESG scores by engaging in more carbon-emission-reduction activities became less willing to take a risk, thus diminishing bank value for the shareholders.

Following the practices of banks in the U.S. and Europe, banks in Asia also invested in environment-friendly activities. These environment-friendly policies may produce a positive influence on bank performance in the medium to long run.

Therefore, we developed the first hypothesis:

#### **Hypothesis 1 (H1).** *Environmental variables have a positive impact on bank cost efficiency.*

#### *2.3. Social*

CSR can be explained by the stakeholder theory [27]. The theory states that firms should service a multitude of stakeholders, including shareholders, customers, and employees, rather than shareholders only, so that firms may boost the popularity of products and financial performance [20].

These CSR activities include the production of high-quality products and services for customers, payment of fair salaries to employees, provision of health care and educational programs to the community, in addition to profit maximization for shareholders. However, previous studies found the relationship between social activities and firm performance to be mixed.

Some studies of developed countries such as the U.S., Canada, and other European countries revealed a negative relationship between the disclosure of social activities and bank performance in terms of earnings and ROA because the large costs of social welfare exceeded the benefits [9,11,23,26].

However, CSR may produce a positive influence on bank performance due to a better perception of the stakeholders of the firm's attitude toward social responsibility. Shakil et al. [22] argued that because stakeholders were more interested in the firms' disclosure of social activities, and the implementation of CSR programs may lead to an overall improvement of the firm performance. Dell'Atti et al. [25] studied the correlation between firm reputation and economic performance using 75 large international banks

during the 2008 subprime mortgage crisis. The results suggested that social welfare was positively correlated with firm reputation with some possibility of improving the firms' economic performance. Similarly, Forgione et al. [5] found that the disclosure of CSR activities had a positive impact on bank efficiency only in common law countries, such as the U.S., Australia, and countries with stakeholder protection. These studies confirmed the stakeholder theory that activities benefiting stakeholders increase their contributions to the firms and led to improved financial results.

Therefore, we assumed that CSR activities have a positive relationship with the performance of Asian banks. Thus, we developed the second hypothesis:

#### **Hypothesis 2 (H2).** *Social variables have a positive impact on bank cost efficiency*.

#### *2.4. Governance*

Corporate governance refers to the proper management of a company. For instance, firms should follow good business ethics, as well as disclosure and accountability practices [22]. Sustainable business policies cover the areas such as disclosure of financial and operational information to increase stakeholders' confidence in the company, gender equality, board diversity to allow various opinions on the firm operations, and so on [28].

The agency theory explains the reasons for the increasing importance of good corporate governance over the last decade. According to agency theory, a conflict between shareholders and managers occurs when management interests are not aligned with those of the shareholders [29]. Good corporate governance aims to align the interests of shareholders and managers so that the two groups of people cooperate to strengthen firm performance [5]. Hence, companies with strong corporate governance may reduce the conflict between shareholders and managers [30]. Companies with poor corporate governance are likely to face high agency problems and lower profitability [24].

Prior studies reported mixed results regarding the impact of corporate governance on bank performance [9,11,12,23,31,32]. Birindelli et al. [32] used a fixed-effects panel regression model to analyze the relationship between the composition of the board of directors and the ESG performance among 108 listed banks in the U.S. and Europe from 2011 to 2016. They used female directors, the board size, CSR committee as the governance variables. The empirical results presented an inverted U-shaped relationship between the female directors and firm performance. The evidence suggested that only a genderbalanced board had a positive impact on the bank's overall ESG performance. In addition, ESG programs produced a positive impact on the board size and the existence of the CSR committee. Miralles-Quirós et al. [9] investigated the relationship between ESG and bank performance using 51 banks in the U.S. and Europe from 2002 to 2015. The results indicated that governance had a positive influence on market value and EPS. In addition, Miralles-Quirós et al. [11] scrutinized ESG and bank financial performance in Europe and found that the governance factor produced a positive effect on bank market value.

However, other researchers found governance negatively affected bank performance in emerging countries and some European countries [12,23,31]. Azmi et al. [12] examined the relationship between the disclosure of ESG activities and bank value based on 251 banks from 2011 to 2017 from 40 emerging economies. The results revealed that governance had a negative impact on bank market value. El Khoury et al. [31] investigated the financial performance of 46 banks in the Middle East, North Africa, and Turkey (MENAT region) from 2007 to 2019. The empirical evidence showed that in the long run, bank costs exceeded the benefits of social and governance programs. Similarly, Buallay [23] found that governance disclosure negatively impacted the financial performance of European banks.

Based on the assumption that governance benefits bank performance, we developed the third hypothesis:

**Hypothesis 3 (H3).** *Governance has a positive impact on bank cost efficiency*.

#### *2.5. Financial Variable*

Prior studies used bank-specific (loan, deposits, interest, etc.) and macroeconomic indicators (unemployment rate, GDP, etc.) of the countries in which banks were headquartered to examine bank performance [2,7,10,23,31]. The combination of bank-specific and macroeconomic indicators provided for a comprehensive analysis of the banks and revealed the factors that contributed the most to bank performance.

#### *2.6. SFA and SMF*

Battese and Coelli [4] proposed SFA to examine the cost inefficiency of the panel data of firms. Subsequently, Huang et al. [6] proposed SMF to compare the efficiencies of different decision-making units (DMU) by computing their TGR and MCE. Banks in various countries used different knowledge and technologies to develop their products and services. The difference in technology, measured by TGR, contributed to the variations in bank performance [6,33].

Based on the literature review, this paper uses a two-step stochastic frontier analysis process composed of SFA and SMF to estimate the cost inefficiency of two groups of banks adopting distinct technologies [34]. In the first step of the analysis, the within-group variation in the firms' technical efficiencies, which is frequently associated with firm-specific exogenous variables, is calculated [4]. In the second step, the between-group variation in the technology gap ratios which commonly stems from group-specific environmental differences is computed [6]. The two-step stochastic frontier analysis is more powerful than the conventional regression models because the two-step analysis not only identifies the significant variables affecting bank cost efficiency but also compares the cost efficiency of two groups of banks.

#### **3. Method**

#### *3.1. Data Collection*

This study collected data for 145 banks located in Asian economies from 2015 to 2018, from the BankFocus database. The data were separated into two groups for a comparison of the bank's efficiency based on the bank classification by the IMF [35]. The IMF classifies countries/regions into advanced (known as "the developed economies") and emerging and developing economies (referred to as "the developing economies") by three main criteria: GDP per capita, export diversification (a country/region must export a wide array of commodities, not just a few commodities to be considered "developed economy"), and integration into the global financial system (including both an economy's volume of international trade and its adoption of and participation in international financial institutions). The IMF uses either the sums of the weighted average of data for individual countries/regions. This study adopts the IMF bank classification that is readily available but moves China to the developed economy group considering China has become the second-largest economy in the world by GDP since 2010 and has occupied nearly 20% of the top 100 banks in the world since 2015.

As the result of the bank division, one group of this study contained the data for 60 banks from five developed economies (China, Hong Kong, Japan, Korea, and Taiwan) with 240 observations. The other group included the data for 85 banks from 13 developing economies such as India and Pakistan, with 340 observations. Table 1 lists the banks in the two groups of developed and developing economies.


**Table 1.** Division of banks into the developed and developing economies.

#### *3.2. Variables and Definitions*

The intermediation approach proposed by Sealey and Lindley [36] defines the relationship of input and output used in the efficiency measurement. This approach focuses on bank activities performing the function of intermediation to distribute savers' deposits to borrowers in the form of loans [3,37]. In other words, the efficiency of the bank is measured by its ability to convert resources into income-generating financial assets.

Based on the literature review [7,8,36], the inputs of this study are deposits, labor, and fixed assets. However, we found the data on the number of employees either missing or unavailable for many Asian banks in the sample; we therefore used total assets to indicate labor based on the literature by Altunbas et al. [38], Altunbas et al. [39], Gaganis and Pasioura [40], Weill [41], Fries and Taci [42], Huang et al. [43]. In the previous studies, the price of labor was defined as the ratio of personnel expenses to total assets. Hence, the price of labor is significantly correlated with total assets. Therefore, this study used total assets as the proxy for labor input [38–43]. The outputs are loans, investment, and fee income. Table 2 lists the inputs, outputs, definitions, and descriptive statistics.

The outputs are loans, investment, and fee income. We obtained the means of input and output variables of the two bank groups from the T-test, which shows a significant difference between the two bank groups. Table 2 lists the inputs, outputs, definitions, and descriptive statistics.

Both non-financial and financial variables are used to measure bank efficiency in this study based on the literature. This study adopts both non-financial and financial variables to measure bank efficiency [7,8]. The non-financial variables are the 17 SDGs by the UN and its divisions into the three dimensions of ESG [14]. The UN provided the ESG score indicated by a color scheme for each economy. The green color indicates "good SDG achievement" with a score of 3. The yellow and orange colors (orange colors only available from 2017) mean "challenges remained" and "significant challenges" with a score of 2. The red color means "a major challenge" with a score of 1. The UN gave a higher score for better ESG performance of an economy.


**Table 2.** Definitions of Variables and Descriptive Statistics.

Notes: 1. Standard deviations are expressed in parentheses. 2. All the data were deflated using the consumer price index from the IMF with the year 2010 as the base year. 3. \*\*\* indicates significance levels of 1%.

> The financial variables are divided into bank-specific variables and macroeconomic variables for the economies in which the banks were headquartered [2,10,23,31]. The literature used the bank-specific financial indicators as variables to show the distinct characteristics of banks [43–46]. Pasiouras and Kosmidou [44] used cost-to-income ratio, liquidity ratio, equity ratio, and asset size. Liang, Chang, and Lin [45] adopted non-performing loans (NPL), loan-loss-reserve ratio, and non-interest expense ratio. The rationale for using the cost-to-income ratio is that it indicates the efficiency of cost management, measuring the degree to which banks generate revenues relative to expenses. Higher cost-to-income ratios imply less efficient cost/profitability management [44]. The major element of bank cost is employee salaries and benefits.

> The macroeconomic variables include unemployment rate, gross domestic product (GDP) per capita, and GDP growth rate. Table 3 shows the inefficiency variables and their definitions.



**Table 3.** *Cont.*


Note: 1. The Hong Kong Monetary Authority (HKMA) defines the net stable fund ratio as available stable funds divided by expected stable funds. However, this study computes the liquidity ratio by taking liquid assets divided by total deposits plus short-term funding, due to a lack of data on the net stable fund ratio. 2. non-interest expense refers to an operating expense separated from interest expense and loan loss reserves. Non-interest items include payroll, rent, utilities, information technology costs, etc.

#### *3.3. SFA*

In the first step of the analysis, we applied SFA as proposed by Battese and Coelli [5]. The stochastic cost frontier function is set as translog, based on Christensen et al. [47], which is homogeneous at the first degree [42,48–50]. In order to eliminate the heteroskedasticity problem, we also used the price of labor (*P*2*it*) to normalize total costs and input prices proposed by Allen and Rai [51], Berger and Mester [52], Kraft et al. [50]. The cost function is set as Equation (1):

ln( *TCit P*2*it* ) = *<sup>α</sup>*<sup>0</sup> <sup>+</sup> *<sup>α</sup>*<sup>1</sup> ln*Y*1*it* <sup>+</sup> *<sup>α</sup>*<sup>2</sup> ln*Y*2*it* <sup>+</sup> *<sup>α</sup>*<sup>3</sup> ln*Y*3*it* <sup>+</sup> *<sup>β</sup>*<sup>1</sup> ln *P*1*it <sup>P</sup>*2*it* + *<sup>β</sup>*<sup>2</sup> ln *P*3*it <sup>P</sup>*2*it* + <sup>1</sup> 2 *δ*11(ln*Y*1*it*) <sup>2</sup> + <sup>1</sup> 2 *δ*22(ln*Y*2*it*) 2 +<sup>1</sup> 2 *δ*33(ln*Y*3*it*) <sup>2</sup> + *<sup>δ</sup>*<sup>12</sup> ln*Y*1*it* ln*Y*2*it* + *<sup>δ</sup>*<sup>13</sup> ln*Y*1*it* ln*Y*3*it* + *<sup>δ</sup>*<sup>23</sup> ln*Y*2*it* ln*Y*3*it* + <sup>1</sup> 2 *<sup>γ</sup>*11<sup>h</sup> ln( *P*1*it P*2*it* ) i2 +<sup>1</sup> 2 *<sup>γ</sup>*33<sup>h</sup> ln( *P*3*it P*2*it* ) i2 <sup>+</sup> *<sup>γ</sup>*<sup>13</sup> ln *P*1*it <sup>P</sup>*2*it* ln *P*3*it <sup>P</sup>*2*it* <sup>+</sup> *<sup>ρ</sup>*<sup>11</sup> ln*Y*1*it* ln *P*1*it <sup>P</sup>*2*it* <sup>+</sup> *<sup>ρ</sup>*<sup>13</sup> ln*Y*1*it* ln *P*3*it <sup>P</sup>*2*it* <sup>+</sup>*ρ*<sup>21</sup> ln*Y*2*it* ln *P*1*it <sup>P</sup>*2*it* <sup>+</sup> *<sup>ρ</sup>*<sup>23</sup> ln*Y*2*it* ln *P*3*it <sup>P</sup>*2*it* <sup>+</sup> *<sup>ρ</sup>*<sup>31</sup> ln*Y*3*it* ln *P*1*it <sup>P</sup>*2*it* <sup>+</sup> *<sup>ρ</sup>*<sup>33</sup> ln*Y*3*it* ln *P*3*it <sup>P</sup>*2*it* <sup>+</sup>*τ*1*<sup>t</sup>* ln*Y*1*it* <sup>+</sup> *<sup>τ</sup>*2*<sup>t</sup>* ln*Y*2*it* <sup>+</sup> *<sup>τ</sup>*3*<sup>t</sup>* ln*Y*3*it* <sup>+</sup> *<sup>λ</sup>*1*<sup>t</sup>* ln *P*1*it <sup>P</sup>*2*it* + *<sup>λ</sup>*2*<sup>t</sup>* ln *P*3*it <sup>P</sup>*2*it* + *ω*1*t* + *ω*2*t* <sup>2</sup> + *uit* + *vit* (1)

where *i* denotes the *i-th* bank; *t* denotes the time period; *TCit* is the total cost of the *i*-th bank during period *t*; *Y*1*it* is the total loan amount of the *i-th* bank during period *t*; *Y*2*it* is the *i-th* bank's total investment during period *t*; *Y*3*it* is the total fee income of the *i-th* bank during period *t*; *P*1*it* is the funding price (interest) of the *i-th* bank during period *t*; *P*2*it* denotes the labor price of the *i-th* bank during period *t*; *P*3*it* denotes the capital price of the *i-th* bank during period *t*; *uit* denotes random error, *vit* ∼ *N*(0, *σ* 2 *v* ) means statistical noise. Nonnegative random errors *uit* represent cost inefficiency, which follows the truncated-normal distribution as *uit* ∼ *N*+(*mit* = *δ* ′*Zit*, *σ* 2 *u* ). *uit* and *vit* are independent of each other.

The inefficiency model used in this study is expressed in Equation (2):

$$m\_{\rm it} = \theta\_0 + \theta\_1 Z\_{1\rm it} + \theta\_2 Z\_{2\rm i} + \theta\_3 Z\_{3\rm i} + \theta\_4 Z\_{4\rm i} + \theta\_5 Z\_{5\rm i} + \theta\_6 Z\_{6\rm i} + \theta\_7 Z\_{7\rm i} + \theta\_8 Z\_{8\rm i} + \theta\_9 Z\_{9\rm i} \tag{2}$$

where *θ* denotes the estimated parameter; and *Zit* the inefficiency parameter. The inefficiency variables include ESG (environmental (*Z*1*it*), social (*Z*2*it*), governance (*Z*3*it*) with

a score of 1 to 3 assigned by the UN, NPL ratio (*Z*4*it*), BIS ratio (*Z*5*it*), cost-to-income ratio (*Z*6*it*), liquidity ratio (*Z*7*it*), loan loss reserve ratio (*Z*8*it*), and non-interest expense ratio (*Z*9*it*).

#### *3.4. SMF*

The second step of this analysis uses SMF as proposed by Huang et al. [7] to estimate the metafrontier cost function, and then measure the inefficiency of different DMUs. The SMF approach not only includes statistical inferences to replace the mathematical programming technique when estimating group frontiers but also considers error terms and group heterogeneity.

Prior literature discussed heterogeneous market structures and measured systemic risk based on the capital flows between groups of banks [53] or based on banks' market returns which are aggregated in bank groups [54].

We first applied SFA to estimate the group-specific frontier cost, then used SMF to estimate the metafrontier cost. Moreover, the SMF approach can directly estimate the technology gaps which are represented by the one-sided term. The technology gaps can be further specified as a function of bank-specific variables beyond the control of banks. The metafrontier cost is based on the concept that all DMUs in the various cost groups have potential access to an array of production technologies, but each may choose a particular technology depending on specific circumstances, such as regulation, the environments, risk (systemic risk or non-systemic risk).

The method used in this study is based on the two-step stochastic frontier approach for estimating the metafrontier proposed by Huang et al. [6]. In the first step, prior researchers [6] used the stochastic frontier regression method to estimate the group-specific frontier. In the second step, these researchers applied the stochastic metafrontier regression method to estimate the metafrontier that specifically takes into consideration the estimation error of ⌢ *f w* (*Xwit*) in estimating *f w* (*Xwit*).

*t t* The SMA regression method is used to obtain the frontier of cost efficiency (CE) of each bank group. Equation (3) explains that the cost efficiency of the *i* th DMU in group *w* th at *t* -th period is accounted for by the group-specific exogenous variables *Zwit*. The CE calculated using *CEit* = *e <sup>u</sup>it* has a value between one and infinity (1 < *CEit* < ∞). A lower group's *CEit* value means lower cost inefficiency (higher cost efficiency). On the contrary, a

higher group's *CEkt* value means higher cost inefficiency (lower cost efficiency). Thus, Cost Efficiency (CE) is expressed in Equation (3):

$$\mathcal{C}E\_{it}^{w} = \frac{\mathcal{C}\_{wit}}{f\_t^{w}(X\_{wit})e^{V\_{wit}}} = e^{U\_{wit}} \tag{3}$$

where *w* denotes a group.

The common underlying metafrontier cost function for all bank groups in the *t-th* period is defined as (*f M t* (*Xwit*)) [6]. The metafrontier (*f M t* (*Xwit*)) envelops all individual groups' frontiers (*f w t* (*Xwit*))*,* which is expressed in Equation (4):

$$f\_t^w(X\_{\rm wit}) = f\_t^M(X\_{\rm wit})e^{\mathbf{1}\_{\rm wit}^{\rm M}}, \; w = 1, \; \mathbf{2}, \; \dots, \; \mathbf{W}; \; \mathbf{i} = 1, \; \mathbf{2}, \; \dots, \; \mathbf{N}\_{\rm j}; \; t = 1, \; \mathbf{2}, \; \dots, \; \mathbf{T} \tag{4}$$

Because *U<sup>M</sup> wit* ≥ 1, the metafrontier cost for all groups must be smaller than or equal to the estimated group cost frontier *f M t* (*Xwit*) ≤ *f w t* (*Xwit*). TGR is the distance from the cost frontier of the *w* group to the metafrontier cost due to differences in the economic or non-economic factors. A higher TGR indicates a greater distance between the cost frontier of one particular group and the metafrontier cost. TGR is calculated using Equation (5).

$$TGR\_{it}^{w} = \frac{f\_t^{w}(X\_{wit})}{f\_t^{M}(X\_{wit})} = e^{U\_{wit}^{M}} \ge 1\tag{5}$$

The inefficiency of *DMU<sup>i</sup>* under *Xwit* produces random interference for output *Cwit* with the group inefficiency captured by *Vwit* and *Uwit*. Non-negative *U<sup>M</sup> wit* reflects the TGR between the group cost frontier and metafrontier cost. The result indicates that although *DMU<sup>i</sup>* has reached the highest cost efficiency within the group, it still has room for improvement when compared to the metafrontier cost.

Based on technology gap ratio expressed as *TGR<sup>w</sup> it* = *f w t* (*Xwit*) *f <sup>M</sup> t* (*Xwit*) , DMU cost efficiency expressed as *CE<sup>w</sup> it* = *f w t* (*Xwit*)*e Uwit f w t* (*Xwit*) <sup>=</sup> *<sup>e</sup> <sup>U</sup>wit* ; random errors expressed as *<sup>C</sup>wit f w t* (*Xwit*)*e Uwit* = *e Vwit* , we can obtain Equation (6):

$$\frac{\mathbb{C}\_{wit}}{f\_t^M(X\_{wit})} = TGR\_{it}^w \times \mathbb{C}E\_{it}^w \times e^{V\_{wit}} \tag{6}$$

The MCE (*MCEjit*) for the bank groups can be expressed using Equation (7):

$$\text{MCE}\_{\text{wit}} = \frac{\text{C}\_{\text{wit}}}{f\_t^M(\text{X}\_{\text{wit}})e^{V\_{\text{wit}}}} = TGR\_{\text{if}}^w \times \text{CE}\_{\text{if}}^w \tag{7}$$

#### **4. Results**

#### *4.1. Descriptive Statistics*

Table 4 exhibits the descriptive statistics for the inefficiency variables for the two groups of banks. Overall, the banks in the developed economies all had higher ESG means than the ones in the developing economies. We further examined each of the ESG variables. Regarding the environmental variable, the means of the banks in the developed and developing economies were 1.64 and 1.56, respectively. Regarding the social variable, the means of the banks in the developed and developing economies were 2.01 and 1.61, respectively. Regarding the governance variable, the means of the banks in the developed economies and developing economies were 1.78 and 1.38, respectively. Moreover, we tested for multicollinearity between the two groups of banks. The results showed the correlation coefficient of each variable between the developed and developing economies to be less than 0.6; therefore, the problem of collinearity did not exist. Table 4 provides the descriptive statistics of inefficiency variables.


**Table 4.** Descriptive Statistics of Inefficiency Variables.

#### *4.2. SFA*

In the first step of the analysis, we applied SFA. Before estimating the stochastic frontier cost functions for the two groups of banks, we performed the likelihood ratio (LR) test to verify whether the proposed inefficiency model was well developed, with Equation (8):

$$LR = -2\{\ln[L(H\_0)] - \ln[L(H\_1)]\}\tag{8}$$

where ln[*L*(*H*0)] is the log likelihood of the translog cost function without the inefficiency variables. ln[*L*(*H*1)] is the log likelihood of the translog cost function with inefficiency variables.

The null hypothesis (H0) means that no difference existed between the two groups of banks. The opposite hypothesis (H1) assumes the existence of a significant difference between the two groups of banks. The results show that the LR statistic for the banks in the developed Asian countries was 64.5084 and that for developing Asian countries was 87.2316. The statistics of both bank groups are above the Chi-square *X* 2 0.01,9 = 21.6660, thus significantly rejecting H0. The results indicate that the inefficiency variables should be included in the SFA method. Thus, the proposed inefficiency model was suitable for this research. Table 5 presents the estimation of stochastic frontier cost functions for the two groups of Asian banks.


**Table 5.** Stochastic Frontier Cost Functions of Two Groups of Asian Banks.

Note: \*\*\* denotes 1% significance level; \*\* denotes 5% significance level; \* denotes 10% significance level.

The study used the inefficiency model of the banks in the developed and developing economies in Asia to identify the impact of the ESG and financial variables on the cost inefficiency of the banks. Table 6 presents the empirical results for the two groups of Asian banks based on the cost inefficiency model.


**Table 6.** Results of Cost Inefficiency Model.

Note: \*\*\* denotes 1% significance level; \*\* denotes 5% significance level; \* denotes 10% significance level.

#### *4.3. ESG*

#### 4.3.1. Environmental (E)

The results in Table 6 show that in the developed Asian economies, the environmental variables are negatively related to bank cost inefficiency at the 5% significance level. However, in the developing Asian economies, the environmental variables are negatively correlated with bank cost inefficiency but insignificantly.

The outcome suggests that in developed Asian economies, environmentally friendly activities such as reducing water pollution, eliminating carbon dioxide emissions, and using renewable energy, not increased the intangible values of the banks, but also tangible values. The environmental activities disclosed in the banks' annual reports enhanced bank reputation and simultaneously diminished bank cost inefficiency. These results are consistent with previous studies that found that bank investments in environmentally friendly practices were able to save energy and fuel costs for banks considerably due to the bank-wide inclusion of the relatively larger scale of the energy-saving plans [12,18,23,26]. However, in the developing Asian economies, banks are unable to recover the costs incurred for implementing environmentally friendly activities. These findings correspond to the literature in that although banks spend money reducing environmental harm, these banks suffer from poor environmental regulations and government incentives. Therefore, banks in the developing economies increased expenditure on the necessary equipment and facilities to improve their environments but failed to reduce costs in the long run.

Therefore, H1 is accepted for developed Asian economies but rejected for developing Asian economies.

#### 4.3.2. Social (S)

The results in Table 6 indicate that in the developed Asian economies, social variables have a positive relationship with bank cost inefficiency at the 1% significance level. In the developed Asian economies, banks that implement social welfare programs for stakeholders such as fair employee salaries, safe work environment, and community services increased costs that cannot be compensated for by higher revenue. This outcome is consistent with the literature in that the large banks in the developed economies are expected to show altruistic behavior by caring for employees and serving the communities, hence these social activities did not create more business for these banks [9,11,23,26].

However, in the developing Asian economies, social variables have a negative relationship with bank cost inefficiency at the 1% significance level. This result suggests that socially responsible activities are considered strategic behavior for the banks in developing economies. Such a finding is consistent with the literature that care for employees and neighbors enhances bank reputation and consumer confidence, thus attracting more customers to interact with socially responsible banks. Consumers could even be willing to pay a higher price to purchase financial products and services from banks that frequently

announced new social programs. Furthermore, the higher revenue generated by these banks enables them to hire more qualified workers to enhance their cost efficiency [43].

Therefore, H2 is rejected for developed Asian economies but accepted for developing Asian economies.

#### 4.3.3. Governance (E)

The results in Table 6 reveal that in the developed economies, the governance variables have a positive relationship with bank cost inefficiency at the 1% significant level. In the developed economies, the design of innovative products and services, gender equality, and peace-building activities resulted in higher costs than profits. Such a phenomenon is consistent with prior studies [12,23,31] and occurs because the implementation of corporate governance is internal. The additional costs for such internal improvements cannot be directly turned into more business and higher revenue for banks. This outcome suggests that because the banks in the developed economies tend to be more globalized, they are anticipated to align with the UN guidelines and managed according to international standards. These banks are expected to invest in corporate governance to maintain their reputation and service quality. However, these efforts simply meet the customers' expectations and do not attract more customers. These efforts can be regarded as altruistic behavior [5].

In contrast, in the developing economies, the governance variables have a negative relationship with bank cost inefficiency at the 1% significance level. This outcome implies that these banks can attract more customers or increase customer willingness to purchase more financial products or services as a result of bank investments in governance activities. Such acts are considered strategic behavior for banks because their improvement in bank management is likely to change customer perception and differentiate themselves from the banks that do not pay attention to governance. Hence, banks in developing economies can attract customers and generate higher revenue due to improved governance which can be announced publicly.

Therefore, H3 is rejected for developed Asian economies but accepted for developing Asian economies.

#### *4.4. Bank Financial Ratios*

Regarding bank financial indicators, three variables are worth noting. First, NPL has a significantly positive relationship with bank cost inefficiency in both developed and developing economies. Second, BIS adequacy ratio has a significantly negative relationship with bank cost inefficiency in both developed and developing economies. This outcome suggests that banks should focus on reducing NPL and increasing bank capital to improve cost efficiency.

Third, the cost-to-income ratio has a significantly negative relationship with bank cost inefficiency in the developed economies but a significantly positive relationship with bank cost inefficiency in the developing economies. Such results mean that increasing costs in the developed economies increases bank cost inefficiency. However, increasing costs in the developing economies decreased bank cost inefficiency. The outcome implies that the banks in the developing economies can generate greater returns for their spending [46].

#### *4.5. SMF Cost Function*

In the second step of the analysis, we adopted SMF to estimate the metafrontier cost functions for the two groups of banks. Before estimating the metafrontier cost function, we conducted the LR test to determine whether a difference exists between the two groups of banks in terms of the stochastic frontier. The null hypothesis was set as *H*<sup>0</sup> : *β<sup>F</sup>* = *βN*, which means that no difference existed between the two bank groups. The opposite H<sup>1</sup> means a significant difference existed between the two bank groups. The LR test statistic was 218.56, which was higher than the threshold calculated by Chi-square *χ* 2 0.01,27= 46.96, thus rejecting H<sup>0</sup> at a 1% significance level. The results of the T-test and the LR test both

confirm a difference in the stochastic frontier between the two groups of banks. Therefore, it is feasible to estimate the metafrontier cost function.

To conduct SMF, we added a bank-specific variable, namely, asset size, and macroeconomic variables (unemployment rate, GDP per capita) to investigate their impact on bank cost efficiency. Table 7 contains the metafrontier cost function results estimated using SMF.


**Table 7.** SMF cost function results.

Note: \*\*\* denotes 1% significance level; \*\* denotes 5% significance level; \* denotes 10% significance level.

The results in Table 7 show that asset size has a negative relationship with cost inefficiency. This outcome means that larger banks can reduce cost inefficiency due to their economy of scale. The unemployment rate has a positive relationship with cost inefficiency. The GDP per capita has a negative relationship with bank cost inefficiency. The results using macroeconomic factors indicate that higher unemployment increases bank costs. In contrast, higher GDP per capita increases bank cost efficiency.

#### *4.6. TGR and MCE*

The SMF regression method was used to compute both the TGR and MCE of the banks in the developed and developing Asian economies. Figure 1 shows the TGR of the σஜ ଶ γ

> two groups of banks. The banks in the developed economies had average TGR values between 1.0065 and 1.0070. The banks in the developing economies had average TGR values between 1.0739 and 1.1134. The TGR values of the banks in the developed economies are closer to one (1) than those of the banks in the developing economies. The bank group cost of the developed Asian economies is closer to the metafrontier cost, which means they managed costs better. The TGR of banks in the developing economies rose from 2015 to 2018. The increase in the technology gaps each year from 2015 to 2018 suggests that banks in the developing Asian economies deteriorate in their abilities to control costs.

**Figure 1.** TGR of the banks in the developed and developing economies.

Figure 2 depicts the MCE of the two groups of banks in Asia. The banks in the developed economies had average MCE values between 1.3206 and 1.4242. The banks in the developing economies had average MCE values between 1.6638 and 1.6477. The results indicate that the MCE of the banks in the developed economies is superior to that of the banks in the developing economies because the values are closer to one (1). Overall, banks in the developed economies were more cost-efficient considering the ESG, bank-specific and macroeconomic variables altogether. However, the cost inefficiency of these banks continued to rise each year, notably from 2015 to 2016 and slightly from 2016 to 2018. This outcome suggests that the disclosure of ESG activities caused banks in the developed economies to be cost inefficient. It can be inferred that the ESG activities were considered altruistic behavior for these banks with costs not turned into more profits.

**Figure 2.** MCE of the banks in developed and developing economies.

#### **5. Robustness Test**

We conducted the robustness tests to confirm the impact of ESG on bank cost efficiency. The results show that the banks in the developed economies have significantly higher cost efficiency than the ones in the developing economies concerning CE, TRG, and MCE. Therefore, the results from the division of bank groups are robust.

#### **6. Conclusions**

The concept of ESG has gained increasing importance around the world in the last decade. Significantly, the UN specified 17 SDGs indicating actions in each of the ESG areas. An increasing number of companies are using ESGs as a measure of performance in addition to the widely used financial (bank-specific and macroeconomic) ratios. In the last ten years, Asian banks have grown rapidly and are expected to surpass the banks in other continents in terms of personal financial assets by 2025. Therefore, the sustainability of the Asian banks must be maintained to stabilize the economy. However, it remains unclear whether the additional costs incurred by banks due to ESG practices can be compensated for by higher revenue. This study applied a two-step analysis comprising SFA and SMF to examine the impact of ESG (17 SDGs) and financial indicators (bank-specific and macroeconomic) on the cost efficiency of the banks in developed and developing economies in Asia from 2015 to 2018.

In the first step of the analysis, we applied SFA. The results indicate that environmental variables increased the cost efficiency of the banks only in the developed Asian economies. Therefore H1 (Environmental variables have a positive impact on bank cost efficiency) is partially accepted. The social variables decreased the cost efficiency of the banks in the developed Asian economies but increased the cost efficiency of the banks in the developing Asian economies. Therefore, H1 (social variables have a positive impact on bank cost efficiency) is partially accepted. Governance decreased the cost efficiency of the banks in the developed Asian economies but increased the cost efficiency of the banks in the developing Asian economies. Therefore, H3 (Governance has a positive impact on bank cost efficiency) is partially accepted. Overall, only environmentally friendly activities helped banks in the developed economies to become cost efficient, possibly due to the large-scale energy saving schemes implemented bank-wide. Socially responsible activities and good governance aided banks in the developing economies to become more cost efficient, probably due to enhanced reputation and consumer confidence.

In addition, this study examined the impact of bank-specific factors on bank cost efficiency. The results from the cost-to-income ratio varied. Increasing costs in the developed economies increased bank cost inefficiency. However, increasing costs in the developing economies decreased bank cost inefficiency. This outcome suggests that banks in the developing economies were able to generate more revenues with more spending.

In the second step of the analysis, the SMF approach was utilized to compare the TGR and MCE for the two groups of banks. The results indicate that the banks in the developed economies had higher cost efficiency and a smaller technology gap than their counterparts in the developing economies. More importantly, banks in the developing Asian economies increased the technology gap each year, indicating that they deteriorated in their abilities to control costs effectively. The MCE of the banks in the developed economies is superior to that of the banks in the developing economies. Despite their higher cost efficiency, the banks in the developed economies continued to rise from 2015 to 2018, which implies that the overall ESG costs did not allow the banks in the developed economies to generate more revenues.

The current findings confirm that implementing environmentally-friendly practices such as using clean water, green energy, and recyclable products increased bank cost efficiency in the developed Asian economies. Therefore, these banks are recommended to focus on environmental activities. The empirical evidence also indicates that executing socially responsible activities, such as care for employees and community, and governance increased the bank cost efficiency in the developing Asian economies. Therefore, banks

in the developing Asian economies could engage more in socially responsible programs and governance. In addition, the results of this study may help investors in selecting banks with different ESG emphases.

Asian banks have gained increasing importance in the last decade with more than 30% of the top 100 banks in the world originating in Asia. Although the results of this study pertain to Asia, they could be generalized to banks in other developed and developing economies. The findings of the ways ESG impacts cost efficiency not only benefit investors and bank managers, but also the entire banking sector and the global economy seeking sustainable developments. In particular, the developing banks that emphasize ESG practices are likely to become more financially sustainable due to higher cost efficiency and a stronger connection with society. Moreover, banks that allocate more resources to ESG activities can better fulfill the needs of individuals and organizations, thus propelling global economic growth and sustainability.

Future research may aim at analyzing the direct impact of SDGs on value creation for banks, such as the actions which could be adopted to improve shareholder interests or customer satisfaction. The differences in the SDG approaches practiced by banks may be scrutinized to identify those which are most effective. In this way, we could deepen our understanding of the ways in which ESG impacts the financial industry to achieve sustainable development.

**Author Contributions:** Conceptualization, H.-Y.C.; methodology, L.-W.L.; software, Y.-L.L.; validation, H.-Y.C.; formal analysis, H.-Y.C. and L.-W.L.; investigation, L.-W.L.; resources, Y.-L.L.; data curation, Y.-L.L.; writing—original draft preparation, L.-W.L. and Y.-L.L.; writing—review and editing, H.-Y.C.; visualization, L.-W.L.; supervision, H.-Y.C.; project administration, L.-W.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank the anonymous referees for providing helpful comments and suggestions which lead to the improvement of the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Does Gender Diversity A**ff**ect the Environmental Performance of Banks?**

### **Clara Gallego-Sosa 1,\*, Yakira Fernández-Torres <sup>2</sup> and Milagros Gutiérrez-Fernández <sup>2</sup>**


Received: 4 November 2020; Accepted: 4 December 2020; Published: 5 December 2020

**Abstract:** Climate change is one of the greatest challenges facing humanity today. Therefore, all segments of society must act together to stop the deterioration of the planet and the depletion of its resources. The business sector must play an active role in acting responsibly toward the environment. Given the importance of this issue, major efforts have been made to analyze the environmental performance of the most polluting sectors. In contrast, other sectors that are also of great interest due to their contribution to sustainable development, such as the banking sector, have been overlooked. Notable factors conditioning performance include aspects of corporate governance such as gender diversity. However, the empirical evidence reveals a lack of consensus regarding the influence of women directors on corporate environmental performance. This background motivates the study of the commitment of the banking sector to reducing their environmental impact and the analysis the influence of board gender diversity on environmental performance. Data for the period 2009 to 2018 on 52 banks from the most polluting Western regions were studied using descriptive statistics and fixed effects econometric estimation to test the relationship between a selection of relevant variables. The key conclusions are that banks are committed to protecting the environment and that there are no significant differences between banks' commitment to the planet on the basis of board gender diversity.

**Keywords:** corporate social responsibility; environmental performance; climate change; gender diversity; board of directors; banking sector

#### **1. Introduction**

Climate change is one of the biggest challenges facing the planet. It is of vital importance, given its role as a cause of global warming. This phenomenon is having serious consequences throughout the entire planet, including rising sea levels, the flooding of low-lying coastal areas, extreme weather conditions, and severe difficulties for plants and animals to adapt to the new temperatures, potentially leading to the extinction of some species [1]. This irreversible damage is largely caused by human activity. Although some greenhouse gases (GHGs) are released naturally, human actions such as the burning of coal, oil, and gas are increasing the concentration of these gases [2]. Accordingly, in addition to appealing to governments to act following the adoption of the Sustainable Development Goals (SDGs), the United Nations has also called upon the private sector, civil society, and individuals. This call has been made under the premise that joint action is needed to achieve sustainable global economic development that respects the planet and its resources [3]. Otherwise, an environmental catastrophe is foreseen within 30 years [4].

Companies affect their surroundings through their economic activity. Therefore, it is essential that, while pursuing economic profit, they also ensure a positive social and environmental impact, as well as a close-knit relationship with their stakeholders [5]. For this reason, they should consider all elements of corporate social responsibility (CSR) [6]. As business strategies have shifted towards more environmentally responsible practices in an attempt to achieve sustainable development [7], the number of researchers in this area has likewise grown. These scholars have examined the relationships between environmental performance and other variables, including the characteristics of the board of directors [8].

Among the numerous board characteristics that can be used for analysis, gender diversity is considered an essential factor for responsible practices [9]. It is therefore to be expected that women directors act differently from men in response to climate change, given women's greater awareness of the threat it poses and their greater willingness to combat it [10]. The greater commitment of women to ethical standards helps them address social and environmental problems in a more sensible manner [11,12]. For example, Arayssi, Dah, and Jizi [13] have provided evidence that women managers increase the environmental performance of companies by disseminating information to stakeholders and participating in decision making on environmental undertakings. However, the results of previous studies offer mixed conclusions [14–16], highlighting the need for further research to clarify the direction and robustness of the relationship between gender diversity and environmental performance.

According to Pillai et al. [17], the role of the private sector is fundamental to increase awareness and corporate action in support of the 2030 SDGs. It is imperative that firms apply their creative and innovative capacities to resolve the challenges of sustainable development [18]. Most research on the environmental performance of the business community tends to focus on the sectors that are considered the worst polluters, such as the manufacturing industry [14,19]. Insufficient attention has been paid to the service sector, particularly the banking sector, given its central role in the economy and its contribution to sustainable development.

The growing role of the financial sector in the development of Western economies over the last 30 years must be addressed. The role of the banking sector has conditioned both long-term economic growth and the volatility of this growth [20]. Similarly, the importance of banks is supported by their mission, namely, to act as intermediaries tasked with efficiently allocating resources by channeling the savings of one group of individuals toward another group of individuals in need to funds. This second group then invests these funds, thereby creating development and social well-being. Moreover, the banking sector also has a relevant role in ensuring that the business community adopts the SDGs, given that substantial amounts of funding are needed to implement these SDGs [21].

According to Buchner et al. [22], large investments are needed to research alternatives to fight climate change, and a high level of financing is required to implement projects to develop these alternatives and ensure a sustainable planet. Therefore, although the activities of banks do not have a direct impact on the environment, they can exert a positive influence by financing projects that help mitigate harmful effects [23]. Hence, banks play a central role in environmental performance by providing financial resources to other sectors.

Consequently, the importance of this sector leads us to analyze its involvement in environmental action. Similarly, we aim to ascertain whether gender diversity in the managerial echelons of banks actually leads to a difference in their environmental performance. In the literature, the corporate governance of such entities has generally been linked to aspects such as economic or financial performance [24–26], with some recent studies linking it to environmental, social, and governance (ESG) performance [27,28]. To the best of our knowledge, however, only one study has examined the environmental performance of the banking sector [29], although this issue has been studied using multi-sector samples [19,30,31]. Likewise, it may be concluded that the possible relationship between board gender diversity and environmental performance in this sector has received scant attention.

Consequently, our research aims are justified by the importance of climate change, the possible influence of gender diversity on this phenomenon, the lack of consensus in the literature, the lack of studies of the banking sector's role in this area, and the status of North America and Europe as the most polluting Western regions [32]. Our first research aim is to analyze the environmental performance scores of the European and North American banking sector. This analysis can shed light on the level of involvement of banks in undertaking environmentally friendly, or at least non-harmful, actions to mitigate climate change. In addition, we also aim to ascertain whether gender diversity on the board of directors is a differentiating factor among banks with different environmental performance scores. The study is based on the SDGs pertaining to the "Planet" pillar, which is focused on the environment and the fight against climate change [3]. A sample of the largest European and North American banks by market capitalization was used to conduct descriptive statistical analysis and to estimate a fixed effects model.

To the best of our knowledge, this research differs from earlier studies in the following ways. These differences constitute the contribution of our study. First, this study offers the first characterization of the environmental involvement of the European and North American banking sector. This aspect is important for the literature for two reasons. First, for years, the banking sector has been viewed as a non-polluting sector given the nature of its activity [33]. Accordingly, interest in environmental concerns began to target the banking sector much later than the manufacturing sector [34]. Moreover, this sector represents a system that operates under centralized economic and monetary principles. Accordingly, the social and environmental costs associated with banking activity have traditionally been overlooked. Therefore, the banking sector still lacks these values in a context in which environmental damage has enormous scope and has accentuated the inequalities between the rich and the poor. Thus, it is fundamental to achieve social and environmental justice [35,36] and to strengthen social capital [37]. In addition, banking operations can have a powerful environmental impact through the intense use of energy required for the upkeep of buildings and electronic equipment, the generation of waste, and the distribution of financial resources for purposes that ultimately affect society and the environment [33]. This study aims to cover a gap in the research on the role of businesses in protecting the environment. The study achieves its aim by examining a sector that has received little attention (given the perception of the sector as a non-polluter) and that has taken a long time to become involved in protecting the environment, despite its undeniable strategic role in the value chain of the economic system.

Another differentiating factor of this study with respect to existing research is that it provides the first analysis of the role of board gender diversity in the environmental performance of the most polluting Western regions (North America and Europe). The study thus provides critical evidence to fill the current gaps in the literature. Specifically, the studies of this relationship reveal a lack of consensus. They have virtually ignored the banking sector, despite the aforementioned distinctive characteristics of this sector regarding its involvement in fighting climate change and its role as an economic agent. These factors indicate the need for special attention to be paid to this sector.

Finally, this study also offers the first use of a particular measure of gender diversity. The aim is to test the argument that it is necessary to achieve 30% representation of women board members to bring about change in the trend of environmental performance in the firms under analysis. An additional advantage is that the empirical analysis is based on a greater number of gender diversity measures than typically found in previous studies.

The article has five further sections following this introduction. Section 2 presents a review of the literature on the relationship between environmental performance and gender diversity. Section 3 explains the sample selection and provides justification for the method. Section 4 presents and discusses the results. Finally, Section 5 offers the conclusions, limitations, and proposals for future lines of research.

#### **2. Board Gender Diversity and Environmental Performance: Literature Review and Research Hypothesis**

Gender diversity is the subject of current debate in developed countries [38]. Numerous authors have studied the influence of women on corporate social responsibility (CSR) and, more specifically, the environmental dimension of CSR [16].

Several of these studies have concluded that companies with female representation on the board are more socially and environmentally responsible [19,39]. For example, they engage in more fundraising for social benefits [40], greater participation in decision making on environmental undertakings [13], and increased dissemination of non-financial information to stakeholders, notably with the Carbon Disclosure Project (CDP) [41]. Similarly, according to Cucchiella, D'Adamo, Gastaldi, Koh, and Rosa [42], women are more likely to adopt renewable energy systems as an alternative energy.

Li et al. [43] stress the importance of board gender diversity, specifically in the most polluting companies, because it encourages better development of environmental policies. The importance of gender diversity stems from the differences in the moral and social value systems of the two genders [44]. For instance, women are more aware of the importance of the stakeholders' well-being [45]. They are also more collaborative, which encourages the sharing of information [46]. Men, in contrast, are more competitive and ambitious [47]. However, the effectiveness of the role of women on the board may be weakened by increased conflict among members when there are at least three women directors [48].

Given these differentiating characteristics between men and women, the values and attributes of board members condition the board's decision making [49]. Therefore, the possible relationship between board gender diversity and CSR is based on three principal theories [50]: agency theory, resource dependence theory, and stakeholder theory.

The first of these three theories, agency theory [51], is based on the idea that greater board independence, due to, among other factors, the greater variety introduced by gender diversity, enabling greater control by reducing the costs derived from agency problems [52], thereby improving environmental action [15].

Similarly, the members of the board act as intermediaries between the company and the outside world. Therefore, resource dependence theory [53] would suggest that the inclusion of women on the board of directors enables greater access to resources and information channels by providing a wider network of contacts, which is particularly important for increasing the value of the business [54]. This situation can lead to better decision making [55] and the implementation of CSR policies such as those relating to the environmental dimension [15,56].

Third, the relationship studied in the present research can also be explained in terms of stakeholder theory [57], given that gender diversity can influence the implementation of environmental practices to meet the expectations of stakeholders [15]. The argument behind this idea is that women focus more on social well-being, given attributes such as emotional intelligence and the ability to understand and represent the needs of stakeholders [45].

In addition to the earlier arguments, an assessment of the consensus of previously reported empirical results reveals mixed findings. Some studies have confirmed a positive and significant relationship between the variables of interest. For example, Elmagrhi et al. [15] showed that the proportion of women on the board of directors positively affects environmental performance in terms of both putting in place environmental strategies and implementing and disseminating these strategies. Furthermore, Lu and Herremans [14] showed the existence of a positive relationship between gender diversity and environmental performance, emphasizing the significance of the results in relation to companies with a bigger environmental impact. Similarly, Liu [58] reported that companies with greater gender diversity are sued less often for environmental infringements.

By contrast, studies such as that of Walls, Berrone, and Phan [59] have shown that gender diversity does not influence environmental performance, a finding that has also been reported in relation to the banking sector [60]. This finding concurs with those of Prado-Lorenzo and García-Sánchez [61], who reported a positive but non-significant relationship between board gender diversity and the dissemination of information on GHG emissions. Alazzani et al. [16] found a positive influence of women on social performance but not on environmental performance, with this relationship being determined by the culture of the location where the company operates. This finding is supported by those of Fakoya and Nakeng [62], who reported that an increase in the number of women is not related

to greater energy use, focusing their analysis on responsible banks according to the Johannesburg Stock Exchange (JSE) Socially Responsible Investment (SRI) Index.

Despite the reported findings, the following hypothesis may be stated based on the earlier arguments that women are more sensitive to environmental issues and that their presence contributes to improving the effectiveness of the board of directors: *the presence of women on the board of directors contributes to better environmental performance of the European and North American banking sector.*

#### **3. Sample and Method**

#### *3.1. Sample*

Our study sample consists of the largest 52 banks in Europe (28 banks) and North America (24 banks) by market capitalization. The sample thereby covers the most polluting regions in the West [32] for the period 2010 to 2018. The selected banks had a market capitalization of more than \$10 billion on 3 December 2019, according to Thomson Reuters Eikon [63]. This database has been used as a data source in previous studies (e.g., [16,64]). The criterion of market capitalization was used because Li et al. [43] suggest that companies with greater market capitalization protect the environment to a greater degree, possibly because they have more resources to combat environmental pollution. There are also more data available on listed and large companies. The aforementioned source was used to gather the data for the dependent and independent variables used in this study. The independent variables consist of the variables of interest (gender measures) to address our second research aim, as well as the control variables.

### *3.2. Dependent Variable*

Environmental performance is usually measured by indicators composed of a weighting of environmental items. These scores of environmental performance are typically compiled by large companies, which have access to extensive information on firms' environmental performance. This study follows the approach adopted in previous studies [31,59,65]. The environmental score (EnvSc) is published by Thomson Reuters Eikon [63]. This score takes values between 0 and 100 and gives a score calculated as the weighted sum of the three categories that form this pillar: resource use (20 indicators), emissions (22 indicators), and environmental innovation (19 indicators).

#### *3.3. Independent Variables*

The board gender diversity variables chosen for this study are those that have been most widely used in previous environmental and corporate governance studies [29,66]. This choice of variables enabled verification of the robustness of the results given the use of multiple measures (Table 1).


Source: Compiled by the authors.

*Dum1* was included to control for differences between banks with no female directors on their board and those with at least one. Some studies have reported that women are more aware of environmental problems and lead to more egalitarian, social, and environmental organizations [64,68].

*Dum3* was included because several studies have shown that the presence of at least three women on the board of directors enhances the role of women [58,69]. The reason is that the presence of only one or two women on the board is insufficient to bring about change because their opinion is more likely to be ignored [70]. Similarly, Liu [58] reported that firms with more than three women are sued less often for environmental infringements. However, the impact of a critical mass of women directors on environmental sustainability has received little attention [29].

In view of the previous arguments, the *Nwom* variable was included. It has been observed that environmental performance increases when there are more women on the board of directors [71]. The variable *Pwom* was also included because, in addition to the number of women on the board of directors, the proportion of women on the board is also important. It has been argued that the percentage of women on the board is positively related to environmental performance because women have greater environmental awareness [15,31].

The variables *Dum30* and *Dum40* were also included to control for differences between banks that have a board with, respectively, at least 30% and at least 40% women directors and banks with a proportion of women directors below these thresholds. These variables were included because the threshold of around 30% in the proportion of female directors explains a shift in the trend of environmental performance in the banking sector [29]. Furthermore, the effect of gender diversity on the board of directors should lead to better performance if there is a balanced gender distribution on the board; that is, 40% to 60% of board members are women [72]. To the best of our knowledge, this is the first time that *Dum30* has been used in this stream of literature, and the use of *Dum40* is relatively new in studies of gender and corporate governance [66]. Finally, the *Blau* index was used to measure the gender diversity of the board [67]. Several studies have reported that this index offers a good measure of diversity [14,73].

To improve the specification of the model, six control variables were included. These variables have been linked to environmental performance in previous studies [41,64]. Five are governance variables (*Ndir*, *CEODual*, *CSRCom*, *EnvTra*, *DirBon*) and one is an economic indicator (*SBank*).

Specifically, we selected the size of the board of directors (*Ndir*: number of directors). According to Kaspereit, Lopatta, and Matolcsy [74], it has a positive influence on CSR, thereby confirming its relationship with environmental concern [59]. It was therefore expected that companies with larger boards would have better environmental performance [75]. It was also important to consider CEO duality, which occurs when the same person simultaneously holds the position of CEO and chair of the board of directors (*CEODual*: dummy that takes the value 1 if there is CEO duality, and 0 otherwise). Studies, such as that of Galbreath [76], have shown that companies with CEO non-duality make greater efforts to tackle climate change.

Likewise, we considered whether each bank had a CSR committee (*CSRCom*: dummy that takes the value 1 if such a committee exists, and 0 otherwise). The aim of such a committee is to increase the awareness of employees about the environmental aspects of their work and their responsibility for the reduction of negative impacts on the environment, positively influencing the development of carbon strategies [41]. Furthermore, Orazalin [77] reported that the adoption of CSR committees improves the effectiveness of CSR strategies, leading to improved environmental and social performance. *EnvTra* was also included in the study (dummy that takes the value 1 if there are environmental management training policies, and 0 otherwise) because employee training and development practices condition a company's environmental performance [78].

We considered the existence of bonus policies for responsible practices by board members (*DirBon*: dummy that takes the value 1 if there are bonus policies, and 0 otherwise). Previous studies, such as that of Williams [79], have also examined this variable because an increase in salary is related to meeting sustainability goals [59]. Finally, consistent with the approach of Haque [64], we included

26

a variable to capture the size of the company in terms of number of employees (*SBank*: annual average number of employees). This variable was log-transformed to reduce the distortions caused by outliers.

#### *3.4. Method*

This study has two aims. We describe the procedure in each case. First, to characterize environmental performance, we used descriptive statistics. Common statistics were obtained to arrange and analyze the properties of the data. Regarding the relationship between gender diversity and environmental performance, in line with previous studies [64,74,80], we used panel data to perform the econometric estimation of a linear static equation, which is shown below:

$$\begin{array}{ll}\text{EnvSc}\_{\text{it}} = \beta\_1 + \beta\_2 \text{ Gen}\_{\text{it}} + \beta\_3 \text{ Ndir}\_{\text{it}} + \beta\_4 \text{ Coo} \text{Dual}\_{\text{it}} + \beta\_5 \text{ CSRCom}\_{\text{it}} + \\ \beta\_6 \text{ EnvTr}\_{\text{it}} + \beta\_7 \text{ DirBon}\_{\text{it}} + \beta\_8 \text{ Skank}\_{\text{it}} + \eta\_{\text{i}} + \varepsilon\_{\text{it}} \end{array} \tag{1}$$

Here, *EnvSc* is the environmental indicator, *Gen* denotes each of the seven selected gender measures, *Ndir* refers to the number of directors on the board, *CEODual* is the measure of CEO duality, *CSRCom* indicates whether there is a CSR committee, *EnvTra* indicates whether there are environmental management training policies, *DirBon* indicates whether there are bonus policies for responsible practices, *SBank* is the average number of employees, η<sup>i</sup> is the unobservable individual effect, and εit is the random error term for company *i* in period *t*.

Two procedures can be used to estimate linear static equations with panel data: fixed effects models or random effects models. To determine which should be used, the assumption of absence of correlation between the unobservable individual effect and the explanatory variables must first be verified using the Hausman test. The results of the test are provided for each estimated equation in Table 4 under "*p* value (Hausman: FE/RE)." If this hypothesis of absence of correlation is rejected, then the only consistent estimator is the fixed effects estimator. However, if this hypothesis is not rejected, then both estimators are consistent, the difference being that the random effects estimator is the efficient estimator [81]. Therefore, the decision of which model to use for the analysis in this study was based on the results of the Hausman test. These results show the existence of correlation between the explanatory variables and the unobservable individual effect. Therefore, the results indicate that the fixed effects model should be used because it offers the only consistent estimator.

Crucially, for the fixed-effects estimator to be consistent, it requires the assumption of exogeneity of the explanatory variables to hold [81]. In view of the possible existence of endogeneity in the model due to the simultaneous causality between the dependent and independent variables [31,58], we tested the hypothesis of absence of correlation between the explanatory variables and the error term using the Hausman test. The result is given in Table 4 under "*p* value (Hausman: FEIV/FE)." The results show that the aforementioned assumption of exogeneity holds in all cases. The first lags of the explanatory variables were used as instruments [82].

However, robustness analysis was performed by repeating the estimation of the equation using a random effects model and the generalized least squares estimator. Given that it was also necessary to meet the assumption of exogeneity of the explanatory variables [81], we checked this assumption, providing the results in Table 5 under "*p* value (Hausman: REIV/RE)."

Finally, estimation was performed using a variances-covariance matrix of errors that were robust to heteroscedasticity between individuals and to serial correlation of the errors of the same individual. Time dummies were also included to control for any unobservable factors that could influence the behavior of the dependent variable over time.

#### **4. Results**

Tables 2 and 3 provide a general description of the variables for the sample and a comparison of the environmental performance scores at different levels of gender diversity. As Table 2 shows, *EnvSc* has a relatively high value, with an arithmetic mean of 75.43. The standard deviation indicates

low heterogeneity in the data, indicating the reliability of this mean value. The results for the 25th and 50th percentiles show that 75% of the observations of *EnvSc* have a score of more than 70, while 50% have a score of more than 80. Therefore, the vast majority have high scores, given that the maximum score is 100. The highest score is 97.84, and only 25% of the observations have scores above 90 (75th percentile). These data show that most of the analyzed banks have good environmental performance. Thus, in response to our first objective, we can conclude that the European and North American banking sector has a high level of involvement in actions to mitigate the effects of climate change. Similarly, according to Azarkamand et al. [83], companies are increasingly implementing measures to fight against them.


**Table 2.** Descriptive analysis.

Source: Compiled by the authors using Stata 16, StataCorp LLC, Badajoz, España.

**Table 3.** Descriptive statistics of *EnvSc* based on the value of observations of *Nwom* and *Pwom* with respect to the median values.


Source: Compiled by the authors using Stata 16, StataCorp LLC, Badajoz, España.

We now consider the gender variables. On average, 94.63% of observations in the sample indicate that there is at least one woman on the bank's board of directors (see *Dum1*). This result implies that there are still leading banks with no women on their boards, although this is not generally the case. The average value of *Dum1* is greater than that of *Dum3*. It can therefore be deduced that in some banks with female representation on the board, there are few than three female directors. However, in almost 67% of the observations, there are at least three female directors on the board. As reflected by the 50th percentile, in 50% of cases, the board has at least three women directors.

Moreover, although the average number of women directors is 3.57, the standard deviation reveals heterogeneity in the data, with the *Nwom* variable taking values between 0 and 10. In cases with a value of 0, there are no women on the board, while the maximum number of female directors is 10. The data show that in 75% of the observations, this number is less than 5 (75th percentile).

The low presence of women on the board is further reflected by the fact that the average proportion of women directors is slightly less than 25%. The proportion of women is less than 30% and 40% in most cases, as reflected by *Dum30* and *Dum40*. Therefore, the data reflect the under-representation of women on the boards of directors of the banks in the sample. This finding is also corroborated by the *Blau* index. Despite showing that there is at least gender parity in one company (see maximum value of *Blau*), the average value is 0.1336.

We now consider the other corporate governance variables and the company size indicator (number of employees). On average, the banks have approximately 14 board members. There is little dispersion of the observations around this mean value, with the largest boards comprising 28 members and the smallest consisting of five. The opposite is true of the *Sbank* variable. The values for *SBank* fall within a wide range (1250 to 33,012 employees). This high dispersion, together with the values for the percentiles, indicates the variation of the sampled banks in terms of size. Furthermore, just under 50% of the banks have a CEO who is also the chair of the board. Regarding CSR and the environmental training of the board members, the average values of *CSRCom* and *EnvTra* imply that many banks have a specific CSR committee as well as policies for the training of board members in environmental matters.

Table 3 shows the descriptive statistics for *EnvSc*. The data are shown separately for banks with fewer than 3.73 women directors and those with 3.73 women directors or more. The data are also shown separately for banks with less than 25% women directors and those 25% or more women directors. Here, 3.73 and 25% are the respective median values of *Nwom* and *Pwom* for the sample.

In observations for which the number of women directors or the proportion of women directors is greater than or equal to the respective median value, the environmental performance score is approximately 13 points higher. Specifically, in cases where there are at least 3.73 women directors, the mean environmental performance score is 82.46. If the opposite is true, the score is 68.66. The same occurs with the percentage of women directors. When at least 25% of the board members are women, the dependent variable has a mean value of 81.30. By contrast, when this proportion is lower, the mean value of *EnvSc* is 68.28. The dispersion of observations around the mean value of *EnvSc* is greater in the sub-samples covering the lowest 50% of scores for the diversity measures. Likewise, there is a notable difference (of around 30 points) between the values at the 25th percentile of *EnvSc* for the two subsamples under these two criteria. In each case, the value is much higher for the subsample where the number of women directors and the proportion of women directors is greater than or equal to the median (49.3 vs. 80.73 and 49.3 vs. 79.13, respectively).

Finally, Table 4 shows the fixed effects estimates of the proposed equation. An equation was estimated for each of the seven proposed gender measures. The models were statistically significant at the 99% confidence level in all cases, as reflected by the *p* value of the *F* test.

We now consider the results of the regressions shown in Table 4. Regarding the relationship between *EnvSc* and the gender variables, only two of the coefficients associated with these explanatory variables are statistically significant (*Nwom* and *Dum30*). Therefore, the results indicate that none of the following measures results in a better environmental score of the sampled banks: raising the proportion of female directors (*Pwom*), having greater gender parity among directors (*Blau*), increasing the number of women on the board from zero to at least one (*Dum1*), increasing the number of women on the board to at least three (*Dum3*), or having at least 40% female representation on the board (*Dum40*). Our results thus confirm the conclusions of previous studies [60–62].

As shown in Table 2, there is a major gender imbalance on the boards of directors in the sample due to a clear predominance of men. Therefore, the benefits of female representation in the top echelons of these organizations are not apparent because a gender-balanced board is necessary for the role of women to truly influence company policies and performance [72]. This point has already been made by Konrad et al. [70], who argued that the number of women on the board caused the difference between a notable and non-notable effect of the female presence on that board. Those authors based their argument on the fact that only if there are several women on the board will they be able to break down the predominant gender stereotypes and on the fact that there must be a critical mass of women to change the male-dominated communication dynamic.


**Table 4.** Dependent variable: *EnvSc*. Fixed effects estimator.

Source: Compiled by the authors using Stata 16, StataCorp LLC, Badajoz, España. \*\*\* significant at the 99% level, \*\* significant at the 95% level, \* significant at the 90% level. R<sup>2</sup> (Within): coefficient of determination of the transformed model (within group). *p* value (F): *p* value of the test of model significance. *p* value (Hausman: FE/RE): *p* value of the Hausman test under the null hypothesis of absence of correlation between the explanatory variables and the individual unobservable effect. *p* value (Hausman: FEIV/FE): *p* value of the Hausman test under the null hypothesis of absence of correlation between the explanatory variables and the error term. The time dummies are omitted for brevity and practicality. The estimation was performed with errors that are robust to heteroscedasticity and autocorrelation.

Regarding the statistically significant coefficients (*Nwom* and *Dum30*), the evidence reveals a negative relationship at a confidence level of 95% and 90%, respectively. These results imply that as the number of women on the board of directors increases and female representation reaches at least 30%, the environmental performance score for the studied banking sector worsens. These results do not necessarily imply that women are unaware of environmental risks and are therefore less sensitive to taking environmental action, as confirmed by studies that report a positive relationship between board gender diversity and environmental performance [14,15,43]. The results merely indicate that having women on the board does not positively influence the environmental performance score. This situation may be due to a possible increase in conflict between board members such that, instead of leading to environmentally responsible decision making, this conflict would decrease consensus and therefore lead to poorer performance, as indicated by Bernardi and Threadgill [48].

Finally, to test the robustness of these results, Table 5 shows the estimations using a random effects model. As observed, none of the coefficients associated with the gender explanatory variables is significant. Therefore, the evidence confirms that the female representation on the boards of directors of the analyzed firms does not contribute to explaining their environmental performance.

Consequently, the proposed hypothesis cannot be verified. To conclude, we should stress the positive and significant relationship between the environmental performance score and *EnvTra*, *SBank*, *Ndir*, and *CSRCom*, as reflected by all equations in Tables 4 and 5. These results imply that the environmental performance score improves when environmental management training policies are put in place, the number of employees and directors is increased, and a CSR committee is established.


**Table 5.** Dependent variable: *EnvSc*. Random effects model.

Source: Compiled by the authors using Stata 16, StataCorp LLC, Badajoz, España. \*\*\* significant at the 99% level, \*\* significant at the 95% level, \* significant at the 90% level. *p* value (Wald): *p* value of the test of model significance. *p* value (Hausman: REIV/RE): *p* value of the Hausman test under the null hypothesis of absence of correlation between the explanatory variables and the error term. The time dummies are omitted for brevity and practicality. The estimation was performed with errors that are robust to heteroscedasticity and autocorrelation.

#### **5. Conclusions**

This study aimed to identify the behavior of the European and North American banking sector in response to climate change. The goal was to determine whether board gender diversity is a differentiating factor among banks with different environmental behavior. This question is highly relevant, given the lack of studies on this topic. To achieve our aims, descriptive statistics and a fixed effects model were used to analyze a sample of the largest European and North American banks in terms of market capitalization.

First, the results show that the analyzed banks generally have high environmental performance scores. This finding reflects the importance with which the sector views this problem. Second, only two of the seven gender measures used in the estimations have statistically significant coefficients, both negative. Similarly, the robustness analysis shows that none of the gender variables has a significant coefficient. Therefore, the results support the negligible effect of a greater presence of women directors on environmental performance scores. However, the literature offers several arguments for the sensitivity of women toward caring for the environment and their greater concern for different stakeholders. Consequently, this finding can be explained by the gender imbalance on the boards of directors of the banks under study, which have a clear under-representation of women. This under-representation of women on the board would imply that the role of women does not influence company policies and performance.

Thus, although the literature suggests that female representation on boards of directors is important for corporate performance, there are still many organizations, such as those in the banking sector, where the presence of women directors is low. This situation is noteworthy given that it is important not only for there to be women on the board but also for women to have a decent level of representation among the directors. Otherwise, a male predominance can prevent exposure to different perspectives provided by women and the positive influence that they can have on the processes and decision making of the board. Therefore, one of the implications of this study is that it reveals the need for a gender balance on the board of directors to ensure that female talent is fully utilized in an organization's decision-making processes. Doing so creates a context that enables women to broaden the perspectives of the board of directors by considering issues that involve different stakeholders and the community, such as protection of the environment.

Finally, it is worth noting this study's limitations. These limitations fundamentally derive from the use of a composite indicator as a dependent variable. When this variable is created as an average of indicators, the sub-indicators with low values are masked and are offset by those with high values. Other indicators of environmental performance should be used in future studies to provide robustness analysis of the results. An example would be the those related to the SDGs. These indicators that measure commitment to the SDGs could show the level of involvement of banks in achieving these vital global goals, as well as providing evidence of the influence of gender diversity on reaching these goals.

**Author Contributions:** Conceptualization, C.G-S., Y.F.-T., and M.G.-F.; methodology, C.G.-S., Y.F.-T., and M.G.-F.; software C.G.-S., Y.F.-T., and M.G.-F.; validation, C.G.-S., Y.F.-T., and M.G.-F.; formal analysis, C.G.-S., Y.F.-T., and M.G.-F.; investigation, C.G.-S., Y.F.-T., and M.G.-F.; resources, C.G.-S., Y.F.-T., and M.G.-F.; data curation C.G.-S., Y.F.-T., and M.G.-F.; writing—original draft preparation, C.G.-S., Y.F.-T., and M.G.-F.; writing—review and editing, C.G.-S., Y.F.-T., and M.G.-F.; funding acquisition, C.G.-S., Y.F.-T., and M.G.-F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Junta of Extremadura and was co-financed by the European Regional Development Fund, grant number GR18124.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
