DOES THE INSTITUTIONAL ENVIRONMENT AFFECT CSR DISCLOSURE? THE ROLE OF GOVERNANCE

The aim of this article is to analyze whether the institutional environment has a direct effect on Corpo-rate Social Responsibility (CSR) reporting practices or if this effect is explained by the influence of the institutional environment on Corporate Governance (CG) mechanisms. To conduct our study, we focused on two countries that reflect different types of institutional environment: relation-based (Brazil) and rule-based (Spain). Based on our results, we can affirm that the institutional environment influences CG mechanisms (Board Size and Reference Shareholder) as well as companies’ CSR disclosure. Additionally, the CG mechanisms affected by the institutional environment also help to explain differences in CSR reporting practices. As relation-based societies evolve into rules-based environments, the information disclosed about CSR becomes more complex due to a strengthening of CG mechanisms


INTRODUCTION
Today, the importance of Corporate Social Responsibility (CSR) in business is undeniable (Lee, 2008) as well as the relevance of CSR communication strategy. Despite the diversity of options available to disclose about CSR (Ihlen, Bartlett, & May, 2011), most companies have chosen to do it by means of a CSR/ sustainability report.
The publication of sustainability reports has increased significantly in recent years (KPMG, 2011) even though CSR disclosure is not mandatory in most countries. Studies have pointed a few reasons behind this growth in CSR disclosure, particularly the responsibility of companies towards their stakeholders in the pursuit of legitimacy (Crowther, 2000), the rewards companies expect for their socially/ environmentally friendly behaviors and for decreasing information asymmetries (Bushman, Piotroski, & Smith, 2004), and the various pressures from their environment (Young & Marais, 2012).
It is theoretically consistent with Institutional Theory to affirm that organizations adapt their structures and policies to institutional norms (Dimaggio & Powell, 1983) in order to survive in the market.
However, research on the effects of the institutional environment can be complex due to the significant difficulty to isolate variables and distinguish those effects from the impact the institutional environment has on other internal-context variables (Adams, 2002).
These mechanisms are not limited to the Board of Directors (BD), comprehending also other features of the ownership structure of companies.
The considerations above and the evidence that sustainability disclosure policies emerge from BD (as supported by Agency and Stakeholder Theories -Freeman, 1984;Jensen & Meckling, 1976) motivate us to inquire whether the institutional environment has a direct effect on CSR reporting practices or if this effect is partially explained by the influence that the institutional environment has on companies' CG characteristics.
Consequently, the aim of this article is to analyze whether the institutional environment has a direct effect on CSR reporting practices of if this effect is explained by the influence of the institutional environment on CG mechanisms. According to suggestions by several researchers (Adams, 2002;Young & Marais, 2012) and considering the aim of our study, we focused on a comparison between two countries: Brazil and Spain. This is because these country's institutional environments are clearly different (Li & Filer, 2007) and because both countries have been considered pioneers or top-ranked in the reporting of CSR information (Gallego, Formigoni, & Antunes, 2014;Grecco, Filho, Segura, García, & Rodríguez, 2013;KPMG, 2013;. Additionally, according to previous empirical evidence, there are differences concerning CG mechanisms: Brazilian companies' BDs are weaker and less effective than Spanish ones (Andrés, Azofra, & Lopez, 2005;Black, Carvalho, & Gorga, 2010).
Thus, our sample consists of 125 listed companies from Brazil and Spain with CG and financial data available at the ASSET4 Database (an international database with information about listed companies' CG and CSR).
In our study, we used measures that reflect the complexity of companies' CSR reporting. We used an index intended to reflect CSR reporting practices; we considered also the actions taken to increase the credibility and quality of such disclosures (Amran, Lee, & Devi, 2014;Boiral, 2013); the extent to which the disclosed information followed the Global Reporting Initiatives (GRI) guidelines; whether such adequacy was subject to external checking; and, finally, whether the information had been assured. With regard to CG mechanisms, we focused on the ones indicated by the literature as having an influence on CSR reporting practices: Board Size, Board Independence, Board Activity, CSR Committee, and Reference Shareholder.
Considering our results, we can affirm that the institutional environment has a strong influence on the complexity of the CSR reporting practices as well as on two of the CG mechanisms analyzed: Board size and the presence of a Reference Shareholder. Moreover, these two CG characteristics are the ones that significantly affect CSR disclosure. When CG variables are introduced in the analysis, the institutional environment alone no longer has any significant effect on CSR reporting. Therefore, we can conclude that the effect of the institutional environment on CSR reporting practices is mediated by CG mechanisms, particularly Board Size and the existence of a Reference Shareholder.
With regard to the composition of our sample, it is worth highlighting that, in rule-based societies, the information disclosed about CSR presents a higher level of complexity than in relation-based societies.
The present article is organized as follows. In the next two sections, we focus on the theoretical framework adopted, a review of the literature, and the development of our hypotheses. In the following section, we look more closely into the sample and the

CSR DISCLOSURE
Communicating CSR has become a priority for companies in order to ensure the satisfaction of stakeholders' demands. Although there is a wide range of alternatives for how to report about CSR (Ihlen et al., 2011), most companies choose to include CSR-related information in their official written documents since markets, investors and stakeholders usually recognize this way better than other options (Gray, Kouhy, & Lavers, 1995).
Once a company has decided to report about CSR in its documents, the next step is to choose whether it is going to do that in its annual reports or through a more specific document, i.e., the sustainability or CSR report. This document aims to inform stakeholders about the social, environmental and economic impact (Triple Bottom Line approach -Elkington, 1998) of companies' activities during a given financial year (Rodríguez, Frías, & García, 2014).
In recent years, there has been an increase in the information disclosed about CSR (KPMG, 2013). However, the credibility and quality of CSR reports have been challenged (Boiral, 2013;Amran et al., 2014). Consequently, companies have started to implement some more complex CSR disclosure practices in order to make their reports more credible.
Thus, companies can decide to report according to international standards (GRI guidelines) in order to enable the comparison and harmonization of CSR information between companies (Adams, 2002;Boiral, 2013). Later, these reports can get a specific "application level", which reflects to what extent each company's report follows the GRI guidelines. Kolk and Perego (2010), García, Sierra, and Zorio (2013), and Castelo, Delgado, Ferreira, and Pereira (2014), among other scholars, argue that the adoption of the GRI guidelines is an important driver for enhancing the quality of CSR/sustainability reports. Moreover, this application level can be self-reported or externally checked (by a third-party or GRI). In this respect, the external checking of a CSR report's adequacy to GRI gives more credibility to the report (Fernández, Romero, & Ruiz, 2012;Prado et al., 2009).
Finally, the maximum level of information complexity is provided by the assurance of the CSR report. This is generally perceived as a mechanism to secure CSR reports' credibility (Fernández, Romero, & Ruiz, 2014;Simnett, Vanstraelen, & Chua, 2009) and quality Perego & Kolk, 2012).

CSR REPORTING, CORPORATE GOVERNANCE MECHANISMS AND INSTITUTIONAL ENVIRONMENT
Several researchers have inquired about what factors could explain the differences found in CSR disclosure practices. Empirical evidence suggests that, in addition to industry pressures and the institutional environment, other factors that can significantly affect CSR reporting practices are company size, profitability and CG mechanisms. Firstly, differences in the level of adoption of CSR disclosure practices have been detected between companies depending on their size (Clarkson, Overell, & Chapple, 2011;Fernández et al., 2012). Normally, company size is significantly and positively associated with the level of voluntary disclosure (Belkaoui & Karpik, 1989;Hackston & Milne, 1996). More recently, several researchers have confirmed that a firm's size can play a relevant role in its CSR disclosure (Michelon & Parbonetti, 2012;Michelon et al., 2015;Rodríguez et al., 2014).
In contrast, the relationship between profitability and voluntary disclosure is not completely clear. However, main disclosure theories, such as Agency and Political Cost theories, argue for a positive correlation (Reverte, 2009). Nevertheless, most of the empirical evidence has not found any significant results on CSR disclosure (Michelon & Parbonetti, 2012;Rodríguez et al., 2014;Said, Zainuddin, & Haron, 2009).
According to these theories, the BD seeks to safeguard the interests of all stakeholders by disclosing information that increases transparency, thus reducing information asymmetry and avoiding opportunistic behaviors.
There are several BD characteristics (BD size, level of activity of the BD, presence of independent directors in the BD), as well as other CG mechanisms (the existence of a CSR Committee in the company and the presence of a Reference Shareholder), that seem to show an effect on sustainability disclosure practices (Amran et al., 2014;Said et al., 2009).
Empirical evidence about the connection between board size and disclosure is not completely clear, although this connection seems to be related with board effectiveness.
While several studies argue that smaller boards are more advocate the convenience of larger boards as these could manage more effectively the complexity of today's business environments (Frías, Rodríguez, & García, 2013;Gandía, 2008;Kent & Stewart, 2008;Willekens, Bauwhede, Gaeremynck, & Gucht, 2005). This debate about effectiveness is also extended to the level of activity of BDs. In this respect, some researchers argue that more active boards would be more effective (Lipton & Lorsch, 1992), while others have suggested that an excessive number of meetings implies less effectiveness (Vafeas, 1999).
In both cases (Board size and level of activity), BD effectiveness is closely linked to the information disclosed by companies (García, Rodríguez, & Gallego, 2011). In other words, evidence suggests that more effective boards would disclose more information, which includes more information about CSR.
Another aspect of this governance mechanism that is commonly examined is board independence. The literature has shown board independence to be positively associated with disclosure practices since independent directors are less aligned with the management, and encourage firms to disclose more and increase information quality as an attempt to protect the interests of all stakeholders (Forker, 1992;Michelon & Parbonetti, 2012; among others). Nevertheless, some contributions show a negative relationship (Haniffa & Cooke, 2005;Prado & García, 2010), while others identify a non-significant relationship (Frías et al., 2013;García et al., 2011).
According to Ullman (1985), the establishment of a CSR Committee is an obvious indication of the company's CSR engagement level. Since one of the functions of this committee is to ensure the quality of the company's sustainability disclosure, the existence of a CSR Committee can be viewed as a sign of the company's commitment to stakeholders, in line with the evidence found by Michelon and Parbonetti (2012)  The last CG mechanism that the literature points as capable of affecting CSR disclosure is the existence of a Reference Shareholder (someone who has the majority of voting rights, veto power or golden share in a company). Ownership structure is identified as one of the key elements that explain international variations in the CG structure (Hoshi, Kashyap, & Scharfstein, 1990;Porta et al., 1999). The presence of a Reference Shareholder is one of the most relevant characteristics of ownership which will certainly affect the company's strategic decisions as he/she has enough power to appoint directors to the board (Boyd, 1994).
Considering that several characteristics of the BD can affect CSR reporting practices, as discussed earlier, one can see the potential influence that the existence of a Reference Shareholder would have on CSR disclosure practices.
Additionally, Agency Theory (Ross, 1973) also contribute in this line of thought as some authors argue that shareholders are more in favor of CSR than managers (Miras, Escobar, & Carrasco, 2014) since the former are concerned about long-term repercussions and aware of the negative consequences of not being socially responsible (Graves & Waddock, 1994;Margolis & Walsh, 2003). Therefore, the first hypothesis is as follows: H1: CG mechanisms have an influence on CSR reporting practices.
Furthermore, many researchers have stated the relevant role that industry plays on CSR disclosure (Fifka, 2013;Kolk & Perego, 2010;Simnett et al., 2009;Sweeney & Coughlan, 2008). In this respect, the main arguments affirm that the nature of activities determines the industry's attitude to social responsibility along with the need for legitimization in the markets or the need for improving firms' reputation due to the negative impact of their activities (greenwashing).
Finally, companies are also under another kind of pressure to disclose about CSR, i.e., pressure from their institutional environment (Adams, 2002;Kolk & Perego, 2010;Lattemann, Fetscherin, Alon, Li, & Schneider, 2009). This is theoretically supported by Institutional Theory, which states that organizations are continuously adapting their structures and policies to the institutional norms and cultural context in which they operate (Dimaggio & Powell, 1983;Scott & Meyer, 1994) as an attempt to survive in the markets. Apart from the influence of regulation in each institutional environment, companies will imitate the successful behavior or practices carried out by their peer organizations (Dimaggio & Powell, 1983). This is the argument followed by North (1991) and Williamson (2000) when they try to explain the concept of institutional environment.
The most relevant evidence in the literature that enhances the existence of this relationship was highlighted by Adams (2002): (1) the strong dependence between the issues that are worthy reporting and the institutional environment; (2) the divergences found in companies' main CSR concerns also depend on the institutional environment; and (3) stakeholders' demands for firms to be socially responsible vary according to differences in the institutional environment. Considering all these arguments, our second hypothesis affirms that: H2: CSR disclosure is different depending on the institutional environment in which the company operates.
Nevertheless, Adams (2002) pointed out the complexity of analyzing the institutional environment's influence since it is very difficult to isolate the contextual variables and to differ its influence from the indirect effects derived from the influence that the institutional environment has on the other internal context variables.
This brings us to consider whether the effects of the institutional environment on CSR reporting is not direct but could be explained instead by the mediation of CG mechanisms, a mediation that is strongly influenced by the institutional environment (Doidge et al., 2007;Klapper & Love, 2004;Li & Harrison, 2008a, 2008b) and, at the same time, significantly relevant to CSR disclosure practices (Haniffa & Cooke, 2005;Michelon & Parbonetti, 2012).
On the one hand, Charkham (1994) and Porta et al. (1999) remarked that one of the significant effects of the institutional environment on the internal variables of companies is the variation in corporate ownership structures. In this perspective, Klapper and Love (2004) and Doidge et al. (2007) highlighted two country-related characteristics that are determinant in the impact of institutional environment: investor protection and the costs of accessing capital markets. The empirical evidence in this respect (Li & Harrison, 2008a, 2008b

METHODOLOGY
This study is predominantly explanatory and uses a quantitative approach. In this section, the study's sample selection, variables and statistical method are explained.

Country selection and sample
Previous research about CSR disclosure has taken the institutional environment of the company into consideration in different ways. On the one hand, there are some studies that assume this influence by focusing only on one institutional environment (Gallego et al., 2014;Kuo, Yeh, & Yu, 2012;Michelon et al., 2015;Paul et al., 2006, Prado et al., 2009  The latter has allowed expanding the scope of research since it classifies each society either as a Rule-Based or a Relation-Based environment (Li, Park, & Li, 2004). To that end, GEI considers five relevant aspects of each environment: political rights, rule of law, free press, the quality of accounting standards, and general trust levels. In Rule-Based environments, regulation is transparent and efficient, and, consequently, the information reported by companies is credible. However, the control exercised by the government in Relation-Based environments can trigger a general mistrust in the information provided, and companies feel less pressure to disclose about their practices (Lattemann et al., 2009 (1) CSR information in a report Gray, Kouhy, & Lavers (1995) (2) GRI adoption level Adams (2002), Boiral (2013) (3) Externally checked Prado, Gallego, & García (2009) (1) there were not many hypotheses and variables included, and (2) the variables were observable and not composed of several items.
Ordinal regression minimizes the differences of the sum of the squares between the dependent variable and a weighted combination of the independent variables. The estimated coefficients reflect how changes in the factors affect the level of complexity of the CSR reporting practices analyzed.

Descriptive statistics
We present the sample's distribution of industries by country (Table 1), the sample's descriptive statistics for each country (Table 2), and the bivariate correlations between all the variables included in the study (Table 4).   As shown in Table 4, there is a significant, positive correlation between CSR reporting practices and three of the independent variables (Board Size, CSR Committee and Reference Shareholder). These results support that CSR disclosure decisions lie with BDs (Adams, 2002;Haniffa & Cooke, 2005;Michelon & Parbonetti, 2012  This evidence is logical and consistent with Boyd's (1994) argument that when there is a Reference Shareholder, several directors will be appointed by him/her.

Multivariate tests
All models in the study are statistically relevant according to their statistics (-2-log-likelihood), and their results are summarized in Tables 7 and 8. Firstly, it is worth highlighting that the institutional environment has a strongly significant influence on the complexity Once these first analyses were finished, we focused on the ordinal regressions of the models proposed (Table 6). Firstly, we found that including CG variables has a positive impact on the explanation of CSR report complexity (R 2 increases). Therefore, we conclude that CG characteristics affect CSR disclosure (i.e., Hypothesis 1 is accepted). These results are in line with previous evidence in this respect.  (Young & Marais, 2012). Previous research has pointed out that the presence of independent directors in boards has a positive effect on the level of CSR disclosure (Lattemann et al., 2009), although we did not confirm this in our findings. This disagreement could be explained by the low percentage of external directors in the companies included in the sample.
The empirical evidence suggests that the CG mechanisms with a strong influence on CSR reporting practices (Board Size and Reference Shareholder) are the ones that show significant differences according to the institutional environment (see Table 3).
When the impacts of CG mechanisms and institutional environment on CSR reporting practices were considered simultaneously, we found that the institutional environment was no longer significant. The goodness of fit of this model is quite close to that of the model which did not consider the institutional environment (only a 1% increase). Only one of the CG variables (the existence of a Reference Shareholder) continued showing a significant influence on CSR disclosure.
At this point, we were able to conclude that the influence of the institutional environment on CSR disclosure is mediated by both CG mechanisms (Board Size and Reference Shareholder).
With regard to the control variables (size and ROA), our results confirmed that company size and profitability have a positive impact on the complexity of CSR disclosure practices, although in some cases, the significance of that impact decreased when CG variables were included in the models. All regressions were also controlled by industry.

CONCLUSIONS
Considering that the aim of this paper is to analyze whether the effect of the institutional environment on CSR reporting practices is direct or if this effect can be explained by the influence that the institutional environment has on CG mechanisms, we can conclude that CG mechanisms mediate the influence that the institutional environment has on CSR reporting practices. Not all CG characteristics were found to be determinant, i.e., only Board Size and the existence of a Reference Shareholder were found to have a significant impact on the complexity of those reporting practices.
This research reveals that it is strongly relevant to know and study the regulation and recommendations concerning CG-related issues for each country (which are reflected on each country's Code of Good Governance) as well as their actual implementation in companies since CG features could be decisive for several strategic decisions, particularly policies of voluntary disclosure.
Specifically, our study focused on comparing two institutional environments (Rule-Based and Relation-based). Our results support that CSR disclosure complexity is higher in Rule-Based institutional environments.
The main contribution of this study is our finding that the higher complexity of the CSR information disclosed in Rule-Based societies comes from the effect of these societies' institutional environment on CG mechanisms. Therefore, as Relation-Based societies evolve into Rules-Based environments, the information disclosed about Corporate Social Responsibility becomes more complex due to a strengthening of Corporate Governance mechanisms.
Our results suggest that further investigations should be conducted including more countries in the sample. Moreover, future research on this topic might consider using other statistical techniques such as SEM.