Culture, Board Composition and Corporate Social Reporting in the Banking Sector

: This paper contributes to the debate on the corporate governance of financial institutions, by studying the effect of different board characteristics on the level of corporate social responsibility (CSR) disclosures of banks. For that, we use a sample composed by 159 banks over the period 2004 – 2010. We found that independent directors and gender diversity favor the disclosure CSR information in baking sector. But, these results are moderated by the national cultural system; concretely, previous positive effects of independence and diversity of banks ’ boards on CSR reporting are reduced in countries with a weaker cultural system, that is, individualist, masculine and vertically stratified societies, that are little indulgent and short-term oriented and show high levels of uncertainty avoidance.


Introduction
Through Corporate Social Responsibility (CSR) disclosure, companies improve corporate transparency by reporting not only financial information but also their social and environmental performances to stakeholders and society. Most studies on CSR performance and CSR disclosures focus on non-financial sectors. Traditionally, authors considered that the banking sector impacts scarcely on the society and environment (Elkington 1994) but this view is shifting. According to Moyo and Rohan (2006), the banking sector is essential for the promotion of sustainable development, since financial intermediaries are crucial for stimulating economic activity, showing consequences for the society and environment (Thompson and Cowton 2004). The banking sector has particular characteristics; thus, financial companies are usually dropped from samples of analysis (Deegan et al. 2002;La Porta et al. 1998). Such sector is complex, with large information asymmetries among different stakeholders. Banks have a higher number of interest groups than non-financial organizations (Mehran et al. 2011) because there are agents who are specific to the banking sector, for example depositors, debt holders, deposit insurance authorities and so on. In this regard, corporate governance in banks is relatively more important than in other sectors, because internal conflicts may cause a loss of confidence of the market on the ability of financial institutions to manage investments, which may result finally in financial crisis (García-Marco and Robles-Fernandez 2008). Moreover, the analysis of the association between the professional background of board members and CSR is highly relevant in a world of continual governance scandals, failures, opacity and social and environmental excesses.
This study is focused on one of the most important corporate governance tool, namely board of directors. Its composition is essential to represent the views and interests of different stakeholders and prevent inappropriate uses of the resources provided by depositors, debt holders and shareholders. Moreover, it has been evidence that the board of directors may reduce information asymmetries and conflicts of interests among diverse stakeholders (De Andrés and Vallelado 2008).
Thus, this study joins two relevant lines of research, CSR and corporate governance, focusing on the banking sector. Concretely, we analyze whether the board independence and the diversity of their members affect the level of CSR disclosures in the banking sector. This question has been previously analyzed in non-financial sectors (e.g., Ayuso and Argandoña 2007;Chen and Jaggi 2000;García-Sánchez et al. 2014;Haniffa and Cooke 2002;Ibrahim and Angelidis 1994) but studies on financial institutions are scarce and focused on specific regions, for example Barako and Brown (2008) in Kenya, Khan (2010) in Bangladesh, Htay et al. (2012) in Malaysia and Sharif and Rashid (2014) in Pakistan.
Additionally, we expect the national cultural systems affect CSR disclosures. Previous literature on corporate governance has evidenced the role of regulatory and institution contexts in financial reporting but it has not been large developed for other kind of reporting, such as CSR (Jackson and Apostolakou 2010). Thus, using Hofstede's six dimensions, we consider culture as moderator in the link between the board composition of banks and their CSR disclosures. Therefore, this paper contributes towards a better understanding on how CSR and board relate to each other and also to previous literature examining the influence of Hofstede's dimensions in this association. We expect that banks in individualist and masculine societies, with high levels of uncertainty avoidance and which are vertically stratified, little indulgent and focused in the short term, tend to show lower levels of CSR disclosures.
This study is structured in six sections: after introduction, section two is dedicated to explain the expected effect of the board independence and diversity the CSR disclosures of financial institutions; the third explains the role of the national cultural system in the previous relationship; section four focuses on the methodology, that is, the sample, variables and models for testing the hypotheses; the results of the descriptive and exploratory analyses are presented in section five; and we finish with some concluding remarks.

Board Composition and CSR Disclosures: Research Hypotheses
Nowadays, the relevance of economic, social and environmental issues has increased and stakeholders' demands go beyond traditional financial information (Jamali et al. 2008). These three issues are encompassed in the concept known as CSR: the economic dimension refers to the responsibility of an organization to be effective and efficient, competitive and profitable (Schwartz and Carroll 2003); social issues are related to philanthropic behaviors and the promotion of the welfare and human rights; and, finally, the environmental feature represents the policies and systems implemented by the organization to take care of the impact of their activities on the environment (Hubbard 2009).
Information about CSR practices is generally demanded by a wide range of stakeholders, such as clients, suppliers, employees, public administrations and the public in general, because it facilitates the understanding of consequences of corporate activities beyond financial results. Normally, this kind of information is disclosed on the CSR report, which serves to show the behavior of the organization towards the society (Archel 2003).
Regarding the banking sector, the Basel Committee on Banking Supervision considers information transparency to be an essential element of a secure and effective banking system (Basel Committee on Banking Supervision 1998). It recommends the regular publication of information, which facilitates the decision making by stakeholders. However, studies on the information disclosures of financial institutions are scarce, especially ones that are focused on CSR information, since traditionally the literature has defended that financial institutions show low impacts on CSR issues (Elkington 1994). However, this conception has changed in the last decades, as some authors have shown; for example, Tsang (1998) showed banking sector in Singapore is more transparent than others as tourism and food sectors. Financial institutions tend to disclose information on human rights and the impacts of their economic activities on the society, as Abu-Banker and Naser (2000) and Zéghal and Ahmed (1990) showed for a sample of Canadian and Jordanian banks, respectively.
More recently, Halabi et al. (2006) also showed that Australian banks disclose information about human rights, human resources and environmental issues and this information is usually high quality. Douglas et al. (2004) found that financial entities tend to publish information about corporate governance, human resources and effects on the community. More recently, Thomson and Jain (2010) analyzed the CSR reporting of two large Australian banks, applying game theory. Due to information disclosures having benefits and costs, 1 the reporting strategy may be analyzed as a "prisoner's dilemma," being "disclosing" the best strategy for both.
One of the main advantages of disclosing information is the reduction of opportunistic behaviors derived from informative asymmetries between shareholders and managers (Hosmer 1994). In fact, monitoring such behavior is an essential role of the board of directors, because corporate information is used by external parties in their decision-making processes (Michelon and Parbonetti 2010). As the CSR reporting is a decision taken by the board (Hertz et al. 2012;Michelon and Parbonetti 2010), the personal characteristics of the board members (character, ideals, abilities, knowledge, etc.) may affect the decisions taken by directors. Among these characteristics, we can highlight independence and gender diversity, because independent and female directors may provide unique resources in terms of knowledge, experiences, opinions and so on.

Independence of the Board and CSR Reporting
Independent directors are professionals without any relationship with the management of the company, so it is improbable that they interfere in corporate decisions with their personal opinions (Wan-Hussin 2009). The presence of independent directors is a control mechanism because they make more objective decisions than managers and shareholders (Pincus et al. 1989;Rosenstein and Wyatt 1990). According to Mehran et al. (2011), they provide new points of view that are different from more traditional standpoints, which are more focused on financial issues. In fact, Ibrahim and Angelidis (1995) suggested that independent directors usually concern with diverse stakeholders, so it is less probable that a company retains useful information for their interest groups (Chau and Gray 2010).
Furthermore, prestige, reputation and job opportunities of these directors are intensely interconnected with the corporate reputation (Masulis and Mobbs 2013); therefore, directors would be interested in showing a good corporate image because it affects their self-reputation. In that respect, CSR is essential nowadays, since the current context lead organizations to exercise a responsible behavior for competing (Garrigues-Walker and Trullenque-San Juan 2008).
According to these arguments, we may expect that the degree of CSR disclosures is increased in organizations with larger share of independent directors on the board. This relationship has been analyzed previously for non-financial companies (e.g., García-Sánchez et al. 2014;Haniffa and Cooke 2005;Prado-Lorenzo et al. 2009). In the specific case of the banking sector, the Basel Committee suggested that the presence of independent directors is essential for the effectiveness of the board (Basel Committee on Banking Supervision 1998). However, the literature in this respect is scarce and focused on specific regions. Concretely, Barako and Brown (2008) suggested a positive link between the board independence and CSR reporting in the Kenyan banking sector. Htay et al. (2012) show similar findings in the case of Malaysian listed banks; and Khan (2010) and Sharif and Rashid (2014) did so for banks in Bangladesh and Pakistan, respectively. Sharma (2013) showed that 1 Some advantages of CSR disclosures are: capital cost reduction, uncertainty reduction, investors' and employees ' attraction, criticism reduction, etc. (Adams 2002;Filatotchev et al. 2007). Among the most important disadvantages are the disclosure of valuable information to competitors and the monetary cost of developing CSR practices, which impacts on the financial results (Thomson and Jain 2010). the independence of the board affects the web-based disclosures of listed commercial banks in Nepal.
In accordance with previous findings, we propose the first hypothesis as: Hypothesis 1 (H1). Banks with a higher proportion of independent directors on the board tend to show a higher level of CSR disclosures.

Diversity of the Board and CSR Reporting in the Banking Sector
Diversity among board members is considered to be necessary for the board's effectiveness (Van der Walt and Ingley 2003). It introduces heterogeneity into the decision-making process with differences among opinions and ideas. Thus, diversity will allow understanding the interests of different stakeholders (Ayuso and Argandoña 2007; Van der Walt and Ingley 2003).
Among the characteristics that give diversity to the board, gender diversity-understood as female participation in the board-is one of the most important (Carter et al. 2003;Williams 2003) and it is crucial to governance and other corporate issues in all modern organizations. The previous literature has characterized women as more empathic than men and having greater concern for others (Eagly and Karau 1991). They are also characterized as being more democratic, participatory, cooperative and accepting of others' ideas (Eagly and Johannesen-Schmidt 2001;Eagly et al. 2003). According to this female personal behavior, boards with larger proportion of women tend to take care of desires and interests different stakeholders beyond just shareholders (Konrad and Kramer 2006). So, probably they pay attention to social and environmental issues to a greater extent (Ibrahim and Angelidis 1994). Accordingly, it is usually expected that gender diversity in boards of directors promotes CSR reporting.
The link between the proportion of women on board and CSR disclosures has been previously analyzed in non-financial organizations (e.g., Ayuso and Argandoña 2007;Fernández-Feijoo et al. 2012;Ibrahim and Angelidis 1994;Rao et al. 2012;Liao et al. 2015). However, in the case of the banking sector, previous literature is scarce. As far as we know, Barako and Brown's (2008) study is unique in relation to this question; they suggest that Kenyan banks with more female directors show higher levels of CSR disclosures. Accordingly, we propose the following hypothesis that will be tested for an international sample of banks: Hypothesis 2 (H2). Banks with a higher proportion of female directors on the board tend to show a higher level of CSR disclosures.

Effect of the National Cultural System: Hofstede's Dimensions
A national culture is formed by different values, opinions and approaches that lead the individuals of a society to act in a certain way. It affects the people's everyday lives, defining their roles, rights and duties (Hofstede 1983). The cultural system of a society explains the differences with other societies, in terms of desires, demands and preferences of stakeholders (Tsakumis 2007). Thus, we expect that the affinity of the culture towards CSR issues may determine the demands of stakeholders for such information and therefore, it affects the level of CSR disclosures.
Here, we use Hofstede's model to represent cultural context (Hofstede et al. 2010); that model (or in its previous versions) has been widely used in the CSR literature (e.g., Fernández-Feijoo et al. 2012;Kim and Kim 2010;Maignan 2001;Ringov and Zollo 2007;Williams and Zinkin 2008). Currently, Hofstede's model consists of six dimensions to describe the national cultures over the world: (i) individualism/collectivism; (ii) masculinity/femininity; (iii) uncertainty avoidance or level of confidence; (iv) power distance or hierarchy; (v) long-/short-term orientation; and finally, (vi) indulgence/restraint.
The "individualism" (vs. collectivism) dimension refers to the relevance of "I" ahead of "we." Individualist societies hold self-actuations and individuals tend to take care of themselves or their immediate families. However, in collectivist societies, individuals think more of themselves as members of a group, having strong links to the society (Hofstede et al. 2010). Accordingly, it is expected that socially and environmentally responsible practices are more common in collectivist societies, in which all individuals take care of others and thus the demands of all stakeholders are important.
The "masculinity" (vs. femininity) feature represents the gender differences in the society. Masculine cultures are characterized by competition, financial profitability and rewards for success; while feminine societies favor cooperation and they are more stakeholder-oriented, beyond shareholders, taking a wider range of demands into account. In this regard, we expect that organizations in such societies tend to show higher levels of CSR disclosures.
The "uncertainty avoidance" dimension denotes the tolerance to unreliable situations and unexpected events. Societies that do not tolerate uncertainty, with fear of change, tend to enact strict rules on individuals' conducts and behaviors; while, societies that are more tolerant of uncertainty are more open to change, more flexible and accept new ideas and behaviors. Accordingly, we expect that organizations that operate in context with a lower level of uncertainty avoidance tend to show higher levels of CSR discloses than others operating in societies where CSR practices are motivated via legislation.
The component called "power distance" refers to the levels of hierarchy in a society. In other words, vertically stratified societies have a larger distance of power and greater differences between the social statuses (Hofstede et al. 2010). Power distance is negatively related to information transparency, so we expect that CSR disclosures are larger in contexts with low power distance, in which individuals require a justification on power inequality.
Other dimension is the "long-term" (vs. short-term) orientation, which may be understood as the links of society with the past and the future. Those individuals that have a long-term orientation believe that the truth is conditional to the context and the time and they take a pragmatic approach; however, those with a short-term orientation follow traditions and view changes with suspicion. Accordingly, we expect that societies characterized by a short-term orientation tend to demand traditional reports that are focused on financial issues, while those with a long-term orientation tend to be focused on other topics, such as CSR.
The "indulgence" (vs. restraint) dimension refers to the degree of control on personal desires and wishes. Individuals in indulgent societies realize their impulses and desires and they tend to be optimistic and have a positive attitude. Thus, we expect that organizations in such societies tend to be aware of CSR issues and report more information about those practices.
The Hofstede's model positions each country relative to other countries through a score on each of these six dimensions. It represents the cultural position of each country, from a strong to a weak position. Concretely, weak cultural systems are individualist, masculine, vertically stratified and short-term oriented and also little indulgent and little tolerant for ambiguity; and, strong cultures are collectivist and feminine societies, without clear hierarchy between social segments, tolerant with the ambiguity, more oriented to the long-term and more indulgent (Hofstede 2011).
This concept is similar to the conventional concept of "cultural distance" that is used in the international business literature to represent the cultural context. However, "cultural distance" and "cultural position" concepts are not totally equal; the former refers to the cultural differences between countries, while the second refers to the absolute cultural characteristics of each country, that is, it does not represent a difference but a level or a position regarding the cultural dimensions. Here, we use the "cultural position" approach to represent (strong or weak) cultural level of each country. Cultural distance indicators,-for example, Kogut and Singh's (1988) index or the Euclidean distance index of Drogendijk and Slangen (2006)-ignore the actual cultural characteristics or position on cultural dimensions (Drogendijk and Holm 2015). We are interested in representing position of the countries regarding the cultural characteristics, so it would not be appropriate to use the "cultural distance" concept.
The cultural context affects the personal characteristics (values, opinion, attitudes, etc.) of the population (Hofstede and McCrae 2004) and so the orientation of directors toward CSR. Previously, scholars have noted the effect of the cultural context on voluntary reporting, such as Williams (1999) in Asia-Pacific and Cooke (2002, 2005) in Malaysia. Buhr and Freedman (2001) showed that the highest levels of environmental disclosures in Canada can be explained by the collectivistic nature of the society in comparison with the US context. Similarly, Jackson and Apostolakou (2010) explained the superiority of Anglo-Saxon organizations in terms of CSR by the liberal orientation of such economies in comparison to other economies in continental Europe.
As the shared values and attitudes among the society influence the personal values of each individual (Hofstede and McCrae 2004), the orientation of independent and female directors toward CSR disclosures may be changed according to the cultural context. Then, we expect that the positive view of independent and female directors toward CSR disclosures may turn into negative in weak societies that are characterized by being individualist, masculine, vertically stratified, little tolerant of uncertainty, more oriented to the short-term and less indulgent. Such as we indicated previously, weak societies are less cooperative and leave aside the demands of different stakeholders; they tend to be more pessimistic and leery of innovations, being focused on traditional (financial) issues and CSR practices are motivated via legislation. Therefore, weak societies tend to give more importance to disclosures of financial information, so, the positive orientation of independent and female directors towards CSR disclosures may be limited by the weakness of the cultural system.
Accordingly, we expect that our Hypotheses H1 and H2 may be moderated by the cultural context, such as the following hypotheses propose:

Sample of Financial Institutions
To test the hypotheses, we use a sample of 159 banks from different countries: Canada, France, Germany, Italy, the Netherlands, Spain, Sweden, the UK and the USA. The sample period is 2004-2010, resulting in a panel with 877 observations. It is unbalanced because some banks leave the sample in some years due the availability of data on the different databases where we obtained the information.
Concretely, the level of CSR disclosures is obtained from two sources: (i) the indicators of utility provided by the Ethical Investment Research Service (EIRIS), which is a global provider of environmental, social and governance performance of organization all over the world and it provides ratings and indices that cover ethical and responsible aspects; and (ii) the CSR reports that have been published annually on each bank's website, in order to check whether they are adapted to the Global Reporting Initiative (GRI) guidelines. The information on board composition is provided by Spencer and Stuart, which is one of the world's leading consulting firms. It regularly publishes some board indexes that show the governance practices among leading public corporations around the world. Finally, data on corporate characteristics are obtained from the Thomson One Analytic database, which provides information on financial statements, ownership and capital structure, analyst rating, corporate governance and company filings, among other modules.
To our knowledge, our sample is the largest in literature on CSR disclosures in banking sector. Other previous papers focus on single countries and that obviously reduces the number of financial institutions involved in the study (e.g., Barako and Brown 2008;Htay et al. 2012;Khan 2010;Sharif and Rashid 2014;Sharma 2013 The level of CSR disclosures is difficult to be represented because it should refer, not only to the quantity but also to the quality of such information. Traditionally, the level of CSR disclosures has been represented by the number of words, sentences, sentences, or even pages (Samaha et al. 2012), which refer to the quantity of information. However, the quality should be taken into account, because organizations with bad CSR indicators tend to disclose descriptive and imprecise information to avoid comparison with those organizations that show the best CSR indicators (Clarkson et al. 2008). Because of that, here we refer to the extent of utility and comparability of CSR information.
Although there is no homogeneity in measuring CSR disclosures, this paper uses the approach of García-Sánchez and Martínez-Ferrero (2017, 2018, as a way to represent the utility and comparability of CSR information. On the one hand, utility dimension refers to the degree of adaptation to users' demands, such as the need for this information to be useful to decision making. For that, we use the utility indicator provided by the EIRIS database, which evaluates the usefulness of information related to four areas, namely employees and human rights, ethical practices, environmental issues and impacts on the community. The evaluation is based on three levels: low, medium and high. Operatively, as it is shown in Table 2, banks with a low level of utility obtain 5 points; with a medium level, 7.5 points; and with a high level, 10 points. As there are four items (areas), the sum of the scores obtained in each area constitutes the variable Utility that takes values between 0 and 40, from the minimum to the maximum level of utility.
On the other hand, comparability dimension denotes the similarity degree among the information reported by different organizations, or by the same organization in different periods of time. For that, we have revised the CSR reports published by each bank of the sample, checking whether the CSR disclosures are standardized to the Global Reporting Initiative (GRI) guidelines 2 , which have been widely used in the previous literature to represent the degree of comparability of CSR information (e.g., Frías-Aceituno et al. 2013;García-Sánchez et al. 2014;Legendre and Coderre 2012;Nikolaeva and Bicho 2011;Prado-Lorenzo et al. 2009). Concretely, CSR reports may be adapted to the GRI guidelines at three levels, C, B, or A (from lower to higher quality), regarding the extent of application of the guidelines. In addition, the symbol "plus" (+) is also available for each level (C+, B+, or A+) if an external assurance was used for the CSR report. According to these levels, we create the variable Comparability, following the scoring criteria that are showed in Table 2. To be specific, if the CSR report is adapted to level C, we assign 10 points to that bank and 10 additional points if it is assured (C+); if the report is adapted to level B, we assign 30 points and 40 in the case of B+; and, if it is adapted to level A, we assign 50 points, being 60 in the case of A+. As we can see, there are 6 levels and each of them is valuated with 10 points. Levels are exclusive, so we assign points to each by cumulating the points of previous levels (e.g., if a company has 40 points, its CSR report shows a 2 According to these guidelines, the information contained in the CSR reports should be global, comparable and harmonized, to ensure that it shows all relevant information, being represented in monetary or numerical terms and being comprehensible for all stakeholders. level B+, because obviously, it complies with requirements of level C (10 points), C+ (10 points), B (10 points) and B+ (10 points). The amount of points is used as the variable Comparability. Then, the higher the value, the higher the level of comparability.

Utility
Points Item 1 Whether bank discloses information about employee conditions and human rights. 5, 7.5, 10 Item 2 Whether bank discloses information about ethical practices. 5, 7.5, 10 Item 3 Whether bank discloses information about the community and society in general. 5, 7.5, 10 Item 4 Whether bank discloses information about environmental issues. 5, 7.5, 10 Comparability Points Item 5 Whether CSR report is adapted to the level C of the GRI guidelines.
10 Item 6 Whether CSR report is adapted to the level C+ of the GRI guidelines. 20 Item 7 Whether CSR report is adapted to the level B of the GRI guidelines.
30 Item 8 Whether CSR report is adapted to the level B+ of the GRI guidelines. 40 Item 9 Whether CSR report is adapted to the level A of the GRI guidelines. 50 Item 10 Whether CSR report is adapted to the level A+ of the GRI guidelines. 60 CSRdisclosure = Utility + Comparability 0-100

Notes:
Adaptation to the level C of the GRI guidelines means that CSR report includes the following indicators: • Profile disclosures: statements 1.1; from 2.1 to 2.10; from 3.1 to 3.8; from 3.10 to 3.12; from 4.1 to 4.4; 4.14 and 4.15.

•
Disclosures on the management approach: not required.

•
Performance indicators and sector supplement performance indicators: a minimum of any 10 performance indicators, including at least one from each of social, economic and environment. Performance indicators may be selected from any finalized sector supplement but 7 of 10 must be from the original GRI guidelines. Adaptation to the level B of the GRI guidelines means that CSR report includes the following indicators: • Profile disclosures: statements 1.1; 1.2; from 2.1 to 2.10; from 3.1 to 3.13; from 4.1 to 4.17.

•
Disclosures on the management approach: for each indicator category.

•
Performance indicators and sector supplement performance indicators: a minimum of any 20 performance indicators, including at least one from each of economic, environment, human rights, labor, society and product responsibility. Performance indicators may be selected from any finalized sector supplement but 14 of 20 must be from the original GRI guidelines. Adaptation to the level A of the GRI guidelines means that CSR report includes the following indicators: • Profile disclosures: statements 1.1; 1.2; from 2.1 to 2.10; from 3.1 to 3.13; from 4.1 to 4.17.

•
Disclosures on the management approach: for each indicator category.

•
Performance indicators and sector supplement performance indicators: each core and sector supplement indicator. The symbol "plus" (+) indicates that the CSR report is external assured. Source: The authors, based on García-Sánchez and Martínez-Ferrero (2017, 2018.
Finally, we create the variable CSRdisclosure as the sum of Utility and Comparability, taking values between 0 and 100. CSRdisclosure is the dependent variable in the econometric models used to test the hypotheses. Additionally, Utility and Comparability are used as dependent variables, with the aim of testing whether the results are robust for both characteristics.

Independence and Diversity of the Board: Explanatory Variables
On the one hand, the degree of the board independence is denoted by the proportion of independent directors on the board (called Independent), which is used for testing the Hypothesis H1. On the other, gender diversity is measured by the proportion of female directors on the board (called Women), which allows testing the Hypothesis H2.
Further, the Hypotheses H3 and H4 suggest the moderating effect of the national cultural system on previous hypotheses. To represent the cultural system, we use the extended "6D model of Hofstede" that comes up six dimensions that a society needs in order to organize itself. These six dimensions have been described in the Section 3: Individualism, Masculinity, Uncertainty Avoidance, Power distance, Long-term Orientation and Indulgence. Each of them is expressed on a scale from 0 to 100 on the Geert Hofstede website (www.hofstede-insights.com). They refer to each country, so all banks that belong to the same country show the same value for each dimension.
We have calculated the sum of the six dimensions for each country in order to create a global variable that represents the "cultural position" (from a strong to a weak position). As we indicated previously, strong societies are collectivist, feminine, without clear hierarchy between social segments, flexible to the uncertainty, more oriented to the long-term and more indulgent; while weak societies are more individualist, masculine, vertically stratified, little tolerant of uncertainty, more oriented to the short-term and less indulgent. Each dimension takes values from 0 to 100, being the higher the level of the most dimensions, the weaker the cultural context 3 , so the global indicator is called Cultural_weakness. Additionally, Cultural_weakness is interacted with Independent and Women variables, with the aim of testing the moderating role of the cultural context, such as it was proposed in Hypotheses H3 and H4.
Despite critics of Hofstede's dimensions (Brett and Okumura 1998;Schwartz 1994), 4 the validity of these dimensions has been supported by many other studies (Søndergaard 1994; Van Oudenhoven 2001;Drogendijk and Slangen 2006), being appropriate indicators to order countries according to their cultural systems. Table 3 shows the values obtained for each country. We can see that the USA is the most individualist country, while Spain is the most collectivist. Sweden highlights because of the role of women in society, while Italy is the most masculine cultural context. Spain and France show the highest level of uncertainty avoidance and Sweden shows the lowest one. Despite there are no important differences in terms of power distance, France is the most vertically stratified while Sweden shows the minimum level of hierarchy. Germany highlights because of the long-term orientation while the USA is the country with the most short-term view. Finally, Sweden is the most indulgent society, while Italy is the lowest one. In conclusion, Sweden is the country with a stronger cultural system and Italy is the weaker context. Finally, the models include also some control variables: Yeart are t dummy variables that take the value 1 in year t and 0 otherwise (t = 2004-2010); Countryj are j dummy variables that take the value 1 for country j and 0 otherwise (j = Canada, France, Germany, Italy, the Netherlands, Spain, Sweden, the UK and the USA). The empirical results are also controlled by some corporate characteristics, such as the banks size by the logarithm of the total assets at book value (Bank Size), economic profitability by the return-on-assets (ROA) and the ratio of loans to total assets at book value (Loans). 3 A higher level of the most dimensions (Individualism, Masculinity, Uncertainty Avoidance, Power distance) represents a weaker cultural system, except in the case of Long-term orientation and Indulgence dimensions. Thus, to be consistent, we use the opposite values in the case of the two last dimensions (100-original value) for creating the global indicator, namely Cultural_weakness. 4 The most of criticisms refer to the validity of data, since they are obtained from an IBM employees survey that was not designed to identify cultural dimensions, so it could be no representative of the general population characteristics (Schwartz 1994).

Models for the Analyses
To test the Hypotheses H1 and H2, we use the following model: The role of the national cultural system in the relationships suggested in Hypotheses H3 and H4 are tested using the following model: In the both models, sub-index i and t represent banks and years in the sample banks, respectively, being i = [1, 159] and t = [2004,2010]; β are estimated coefficients whose results will be commented in the following section; ηi represents the unobservable heterogeneity, which refer to characteristics of the sample individuals (banks) that are different among them but invariant over the time; and μit is the classical disturbance terms.
Another common problem in studies on finance and accounting is endogeneity (Pindado and Requejo 2014), which is defined as the correlation between explanatory variables included in the models and the error term (Wooldridge 2010). Endogeneity usually appear because of: (i) the omission of relevant variables; (ii) errors in measuring variables due to the use of proxies; and (iii) reverse causality between dependent and independent variables. In our models, endogeneity appear because of the three reasons, since other board characteristics (e.g., the board size, nationality of directors, duality, board meetings, etc.) and corporate variables (e.g., listing status, leverage or gearing, assets-in-place, shareholder market return, etc.) because they are strongly correlated with other independent and control variables and they may enter multicollinearity problems. Further, the level of CSR disclosures is represented by proxy variables, which may enter error in measuring and there is also reverse causality between CSR disclosures and some corporate characteristics, for instance size and profitability.
The coefficients of the two models could be estimated by using the fixed-or random-effects estimators, they require some initial conditions: errors must be conditionally homoscedastic and not serially correlated. These conditions are checked by using the Breush Pagan and the Wooldridge tests, respectively. The p-values for the both are lower than 0.05, which lead us to reject the null hypotheses of homoscedastic and not serially correlated errors, respectively, at 95% confidence level. In addition, since it is possible that endogeneity problems appear in our models, it is necessary to use instrumental variables. Due to the presence of heteroscedasticity, serial correlation and endogeneity problems, we use the two-step system estimator proposed by Arellano and Bover (1995). Table 4 shows the results of the descriptive statistics and correlations among variables entered into the models. It does not include the variable Cultural_weakness, because it has already been described in Table 3. The mean value of CSRdisclosure is 16.44, which is a very low value, since CSRdisclosure takes values between 0 and 100. This means that the banks of the sample tend to show a low level of CSR disclosures. More specifically, the levels of utility and comparability are very low (9.19 for Utility, which takes values between 0 and 40 and 7.23 for Comparability, which takes values between 0 and 60). In Table 5, we can see the distribution of observations by country, according to the level of utility and comparability. The financial institutions from Canada and the US disclose lower-quality CSR information, on average, or even do not disclose information about some issues. Among the countries that disclose good CSR information, we can highlight Spain and Germany. The UK shows a high level of useful CSR information but such information tends to be rarely comparable. In addition, the mean values of Independent and Women show that, in general, around 69% of directors are independent and around 9% are women. We can also see the mean values of company size, profitability and loans, as well as the bivariate correlations among the variables.

Results of the Descriptive and Exploratory Analyses
Regarding the bivariate correlations, CSRdisclosure variable has a high correlation with the other dependent variables, Utility and Comparability, because CSRdisclosure is the sum of the two. However, independent and control variables are weakly correlated (lower than 0.3).
The estimated coefficients of models 1 and 2 are shown in Table 6, which displays the effects of the board independence and diversity on CSR disclosures. In these models, the dependent variable is CSRdisclosure, which represents jointly the level of utility and the comparability of CSR information. In Model 1, the variable Independent impacts positively on CSRdisclosure and it is statistically relevant at the 99% confidence level (β1 = 0.2906, p < 0.001). This means that banks with a higher proportion of independent directors on their board tend to disclose a higher level of CSR information, which is in accordance with hypothesis H1. In addition, the variable Women show a positive coefficient, which is statistically relevant at 99% (β2 = 7.9603, p < 0.001), meaning that CSR disclosures tend to be more important (in terms of utility and comparability) in banks with more female directors on their board. This is in agreement with Hypothesis H2.
We find evidence to support Hypotheses H1 and H2, which suggest that the independence and gender diversity favor useful and comparable CSR disclosures in the banking sector. This is in accordance with the results obtained by other scholars in specific countries such as Barako and Brown (2008) in Kenya, Htay et al. (2012) in Malaysia, Khan (2010) in Bangladesh and Sharif and Rashid (2014) in Pakistan. Our findings add international evidence to such previous studies.
Further, Hypotheses H3 and H4 are tested by estimating model 2. Concretely, the cultural context is represented by the variable Cultural_weakness (the higher the value, the weaker the cultural system). The results for the variables Independent and Women are similar to those obtained in model 1: they impact positively on CSRdisclosure, being statistically relevant at the 99% confidence level (β1 = 16.4699, p < 0.001; β2 = 24.3276, p < 0.001). However, the interactions of these two variables with Cultural_weakness have negative coefficients (β4 = −4.3309, p < 0.05; β5 = −5.9244, p < 0.001). This means that the cultural system affects the link between the CSR disclosures and the proportion of independent and female directors, such as we proposed in Hypotheses H3 and H4. Specifically, the positive effect of Independent and Women on CSRdisclosure is lower in societies that are more focused on individuals' own personal satisfaction than that of the community, preferring traditions and fearing change and being few indulgent, having the most important roles in the society controlled by men, being vertically stratified and being more focused on the short term. Figures 1 and 2 visually depict the moderating effect of the cultural context on the link between CSRdisclosure and Independent and Women variables.  Statistically significance represented by: † p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001. Our findings indicate that the impact of the board composition on CSR disclosures is moderated by the national cultural system. More concretely, CSR practices are less important in countries characterized by individualism, rigid rules and high-power distance, dominated by males, focused on the short term and having little indulgence with the impulses of individuals, since corporations tend to pay more attention to financial results and therefore the demands of all stakeholders are not relevant. These characteristics are represented by the values and approaches of people (Hofstede 1983), including directors. Thus, the opinions of independent and female directors about CSR are moderated by such characteristics and it is possible that they reduce the level of CSR disclosures according to their cultural system.
These findings are very relevant because they show the essential role of independent and female directors for CSR disclosures. Their opinions, beliefs and values affect the decisions on CSR reporting. However, the national culture influences such opinions, beliefs and values, so CSR disclosures depend on the orientation of the society toward CSR, which is represented by the independent and female directors.  (2) is a serial correlation test of order 2 using residuals in first differences, asymptotically distributed as N(0,1) under the null hypothesis of no serial correlation. (iv) Hansen is a test of the over-identifying restrictions, asymptotically distributed as χ 2 under the null hypothesis of no correlation between the instruments and the error term, degrees of freedom in parentheses. (v) Statistically significance represented by: † p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001. (vi) CSRdisclosure represents the level of utility and comparability of CSR disclosures; Independent represents the percentage of independent outside directors in the board; Women is the proportion of female directors in the board; Cultural_weakness represents the weakness/strength of the national cultural system by using the six Hofstede's dimensions; Independent * Cultural_weakness, Women * Cultural_weakness are the interaction between Independent and Women and variable Cultural_weakness; Company Size is the logarithm of total assets at book value; Loans is the ratio of loans to total assets at book value; ROA represents the profitability by the return-on-assets; Countryj are dummy variables that takes the value 1 for each j country (Canada, France, Germany, Italy, Netherlands, Spain, Sweden, the UK and the USA); Yeark are dummy variables that takes the value 1 for each k years (2004)(2005)(2006)(2007)(2008)(2009)(2010).

Robustness Checking
In this section, we check the robustness of the previous results to other measures of CSR disclosures; we break down the CSRdisclosure variable into two components, namely Utility and Comparability. This allows us to test whether the board composition affects both the utility and the comparability of CSR information or whether there are differences between these two characteristics of information. Accordingly, new models are proposed, changing the dependent variables and the results are shown in Tables 7 and 8 Focusing on Utility (see Table 7), the results for model 3 are in accordance with those obtained for model 1: Independent and Women impact positively on Utility (α1 = 1.1336, p < 0.001; α2 = 12.1126, p < 0.001). This suggests that independence and gender diversity increase the disclosures of CSR information that is useful for stakeholders. The results obtained in model 4, in which the variable Cultural_weakness is entered, indicate that the positive effect of Independent and Women variables on Utility (α1 = 11.0831, p < 0.001; α2 = 22.8438, p < 0.001) is reduced in countries with weaker cultural systems, since the interaction terms are negative (α4 = −3.0678, p < 0.001; α5 = −6.2724, p < 0.001). Figures 3 and 4 represent the moderating effect of the cultural system on the link between Utility and Independent and Women variables, respectively.  (2) is a serial correlation test of order 2 using residuals in first differences, asymptotically distributed as N(0,1) under the null hypothesis of no serial correlation. (iv) Hansen is a test of the over-identifying restrictions, asymptotically distributed as χ 2 under the null hypothesis of no correlation between the instruments and the error term, degrees of freedom in parentheses. (v) Statistically significance represented by: † p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001. (vi) Utility represents the level of utility of CSR disclosures; Independent represents the percentage of independent outside directors in the board; Women is the proportion of female directors in the board; Cultural_weakness represents the weakness/strength of the national cultural system by using the six Hofstede's dimensions; Independent * Cultural_weakness, Women * Cultural_weakness are the interaction between Independent and Women and variable Cultural_weakness; Company Size is the logarithm of total assets at book value; Loans is the ratio of loans to total assets at book value; ROA represents the profitability by the return-on-assets; Countryj are dummy variables that takes the value 1 for each j country (Canada, France, Germany, Italy, Netherlands, Spain, Sweden, the UK and the USA); Yeark are dummy variables that takes the value 1 for each k years (2004)(2005)(2006)(2007)(2008)(2009)(2010).  Focusing on the level of comparability, we can see the estimated coefficients in Table 8. For Model 5, the results are similar to previous ones, that is, the level of comparability increases with the presence of independent and female directors (γ1 = 0.3021, p < 0.001; γ2 = 27.1152, p < 0.001). In the case of Model 6, in which the variable Cultural_weakness is entered, the positive effect of Independent and Women variables on Comparability (γ1 = 80.0811, p < 0.001; γ2 = 23.5038, p < 0.05), is moderated by the strength of the national cultural system, since the interaction terms have negative coefficients (γ4 = −2.192, p < 0.001; γ5 = −6.9289, p < 0.001). Figures 5 and 6 depict the moderating effect of Cultural_weakness variable on the link between Comparability and Independent and Women variables, respectively.  (2) is a serial correlation test of order 2 using residuals in first differences, asymptotically distributed as N(0,1) under the null hypothesis of no serial correlation. (iv) Hansen is a test of the over-identifying restrictions, asymptotically distributed as χ 2 under the null hypothesis of no correlation between the instruments and the error term, degrees of freedom in parentheses. (v) Statistically significance represented by: † p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001. (vi) Comparability refers to level of application of GRI guidelines; Independent represents the percentage of independent outside directors in the board; Women is the proportion of female directors in the board; Cultural_weakness represents the weakness/strength of the national cultural system by using the six Hofstede's dimensions; Independent * Cultural_weakness, Women * Cultural_weakness are the interaction between Independent and Women and variable Hofstede; Company Size is the logarithm of total assets at book value; Loans is the ratio of loans to total assets at book value; ROA represents the profitability by the return-on-assets; Countryj are dummy variables that takes the value 1 for each j country (Canada, France, Germany, Italy, Netherlands, Spain, Sweden, the UK and the USA); Yeark are dummy variables that takes the value 1 for each k years (2004)(2005)(2006)(2007)(2008)(2009)(2010).

Concluding Remarks
The banking sector is essential in the promotion of sustainable development, since its function of financial intermediation is essential to stimulate the economic development. Thus, their corporate decisions may affect not only their self-performance but also the society in general. Among corporate decisions, CSR strategies are essential nowadays to compete in the market. Accordingly, this paper is focused on CSR reporting of banking sector, checking whether some characteristics of the boards may stimulate CSR disclosures.
For that, we use an international sample of 159 financial institutions over 2004-2010. Our findings show that the board independence and gender diversity of directors in financial institutions favor the level of CSR disclosures, in terms of utility and comparability. However, such effect is affected by the cultural context, which is represented by Hofstede's dimensions (Hofstede 1983). Concretely, the positive effect of independent and female directors on the CSR disclosures has been found only in banks that operate in strong cultural systems. However, CSR disclosures are reduced in individualist societies, with fear of changes, vertical stratification, male domination, a short-term focus and little indulgence of the impulses and desires of individuals.
Our findings contribute to research on corporate governance of financial institutions, in which it has been observed that "the field would benefit from further international studies" (Basel Committee on Banking Supervision 1998). Further, this study also adds evidence to literature on CSR disclosures in banking sector, which is currently growing. These findings are very relevant due to the impact of financial institutions on the society and the environment (Moyo and Rohan 2006; Thompson and Cowton 2004).
In addition, to the best of our knowledge, this study is the first in showing that the role of the national cultural in banking sector. This suggests that the values, opinions and approaches of individuals regarding CSR affect stakeholders' demands and thus the decisions of the board. The board of banks may favor CSR disclosures in societies with strong culture but financial reporting is more relevant in individualist societies that are scarcely indulgent with the desires of individuals, dominated by males, focused on the short term, fear changes and have vertically stratified powers. Thus, independence and diversity are not enough to ensure high-quality CSR disclosures, because stakeholders are not interested in such practices.
Despite previous contributions, this paper is not free of limitations. Firstly, CSR disclosures are measured by proxy variables, since it is a complex and wide concept that is difficult to be represented. In this study, we use some proxies that are generally accepted and have been used previously by other scholars to represent CSR disclosures (García-Sánchez and Martínez-Ferrero 2017, 2018); but, future studies may check the robustness of our findings by using other measures of CSR disclosures. Secondly, a vast number of our sample banks belong to Anglo-Saxon context, mainly the USA and UK, thus, it could be interesting increasing the sample coverage, which will strengthen our findings on the cultural context. Finally, future research may control also the effect of the crisis, since financial institutions were very involved.

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