Comprehensiveness of Corporate Social Responsibility Reports in Brazilian Companies – An Analysis of Its Evolution and Determinants

This work aims to assess the degree of comprehensiveness of Corporate Social Responsibility () reports of Brazilian companies and its determinants. Literature suggests that a CSR report is considered comprehensive when it contains three types of information for each CSR item published: vision and objectives; management actions; and performance indicators. A content analysis of 272 CSR reports of Brazilian companies that follow GRI guidelines was conducted for the period 2010-2013. Results indicate that, despite the still low coverage of contents in CSR reports, there has been an increase in the degree of comprehensiveness over the period of study. Some firm attributes affect the comprehensiveness degree of CSR reports: ownership concentration in hands of the main shareholder, company presence in the ISE (Corporate Brazilian Sustainability Index), the environmental risk of firm industry, firm size and profitability.


Introduction
Corporate Social Responsibility (CSR) notions allude to the company's ability to respond to the expectations of a broad spectrum of stakeholders that present different demands. Organizations, however, have been promoting CSR actions aiming to respond to these demands to obtain competitive advantage (Archel, Husillos, Larrinaga, & Spence, 2009;Branco & Rodrigues, 2008;Donaldson & Preston, 1995;Robertson & Nicholson, 1996).
CSR information reports aim to transmit social, political and economic meaning, showing society the organization's concern with social issues, strengthening the relationship between the organization and its stakeholders, helping to mitigate conflicts, as well as to legitimate the organization's activities (Deegan, 2002;Neu, Warsame, & Pedwekk, 1998). KPMG's (2013) study shows that the number of CSR reports published has increased, as well as the volume of the information disclosed, internationally and in Brazil. Despite this growth, it must be assessed if the respective reports provide objective information that allow for a precise estimate of the company's social performance (Adams, 2004;Archel, Fernández, & Larrinaga, 2008;Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens, 2011;Hopwood, 2009;Tschopp & Huefner, 2015;Unerman, 2000). This reality has fostered research about the effectiveness of CSR reports in satisfying the demand for information, and the actual company's degree of responsibility (Adams, 2004). Results have shown that CSR reports seem to have more content disclosed about objectives and intentions than on effective social actions (Hopwood, 2009). To advance this matter, it was pointed comprehensiveness, and have a predominance of corporative rhetoric in detriment of disclosing concrete actions and performance indicators regarding CSR.
An international research by KPMG about corporative reports shows that about 93% of the 250 largest companies in the world disclose reports of this nature, and that there is also a tendency to conduct external auditing of these reports. In Brazil, the percentage of CSR reports submitted to external evaluation is 56%, while the global average is 38% (KPMG, 2013). In December 2011, the São Paulo Stock Exchange (BM&FBovespa) issued an External Statement recommending the companies listed to indicate in the Reference Form if they disclose Sustainability Reports or similar document, and where it is available. If not, they must explain why they do not do it (BM&FBOVESPA, 2011). This rule, baptized "Report or Explain", might have contributed to the increase from 45.31% to 66.29% in the accession of companies to the sustainability disclosure reports between May 2012 and June 2013 (KPMG, 2013). In the Brazilian market, studies about CSR reports found results in the same direction of increase in accession and in auditing of reports, as well as more compliance to quality standards (Corrêa, Souza, Ribeiro, & Ruiz, 2012;Crisóstomo, Prudêncio, & Forte, 2017).
The evolution in the quality and comprehensiveness of the information disclosed in CSR reports of Brazilian companies might not follow a linear path with the quantity of information disclosed. However, it is pertinent the proposition that the growth in the number of reports with good compliance to quality standards, such as the GRI guidelines and the external auditing of reports, are factors that strongly contribute for the quality and comprehensiveness of the disclosed information. This information makes for the proposition of a hypothesis about the evolution of the comprehensiveness degree of CSR reports in Brazilian companies.
Hypothesis 1: There has been an advance in the degree of comprehensiveness of CSR reports disclosed by Brazilian companies.
Companies with a higher degree of concern with CSR and sustainability actions tend to give more importance to reports that disclose information of this nature, as a way to be more transparent and pursue image and reputation gains, as well as legitimacy of its activities (Adams, 2008;Bebbington, Larrinaga-González, & Moneva-Abadía, 2008;. This broader importance must be reflected in the quality of the report, which includes its depth, completeness, comprehensibility, or comprehensiveness. Market indices have been proposed to evaluate the degree of firm concern with CSR and sustainability actions, such as the Dow Jones Sustainability Index (DJSI) of New York/USA, the FTSE4Good of London/United Kingdom, and the JSE Socially Responsible Investment (SRI) Index of Johannesburg/South Africa (Marcondes & Bacarji, 2010). In Brazil, there is the Corporate Sustainability Index (Índice de Sustentabilidade Empresarial -ISE), which has also been used as an indicator of the level of attention the company gives to social responsibility, and environmental and corporate sustainability issues.
The argumentation is that companies that compose these indices would have a higher CSR and sustainability disclosure standard because of the competitive process that they go through in order to be a member of the sustainability index in a certain year, and the higher degree of visibility they consequently have by composing the index. The suggestion is that the ISE participation amplifies the market knowledge about the company commitment with sustainable development, equity, transparency, and accountability (BM&FBOVESPA, 2016). Under this argument, the ISE presence would grant the company a gain in image among its many stakeholders, facilitating, thus, the legitimacy achievement for its activities. In this context, there are results in Brazil which indicate that the ISE participation influences the voluntary disclosure of socioenvironmental information (Braga, Oliveira, & Salotti, 2009;Machado, Macedo, Machado, & Siqueira, 2012;Murcia, Rover, Lima, Fávero, & Lima, 2008).
This line of thinking suggests that companies that integrate the ISE index manifest more concern with the valorization of their institutional image and, as such, disclose CSR information in a more complete and comprehensive fashion, as formulated in hypothesis 2: Hypothesis 2: The level of comprehensiveness of CSR information disclosed by companies that compose the ISE document is higher than that of other companies.
The literature has suggested that firm ownership structure might have effects on the conflicts of interests between the company's main stakeholders -shareholder, manager, and creditor -, and also on the company's strategic policies, under the argument that there can be prevalence of certain shareholders interests. For instance, there are results that indicate the influence of the ownership structure on the company's value and performance (Allen & Phillips, 2000;Villalonga & Amit, 2006), as well on the company's dividend policy in many ways (DeAngelo, DeAngelo, & Skinner, 2008;Harada & Nguyen, 2011;López-Iturriaga & Crisóstomo, 2010). There is also evidence that certain traits of the ownership structure affect company's investment policy and capital structure (Crisóstomo & Pinheiro, 2015;Goergen & Renneboog, 2001;Schiantarelli & Sembenelli, 2000). Considering that the ownership structure interferes in various company policies, it is plausible to suggest that it might also influence the CSR policy and its respective disclosure.
Besides the pressure from society for company social responsibility, owners and managers started to glimpse the possibility of CSR being an important tool of legitimacy, as well as improvement of the company image and reputation, which would be a motivating factor to carry out such projects that are proposed as able to create value (Chiu & Sharfman, 2011). Large controlling shareholders are very interested in firm legitimacy, reputation and image improvement for the company, given their superior interest in creating value for the company in the medium and long term, differently from minority shareholders that might prioritize a more short-term perspective. In fact, under this argument of value creation interest in the long term through firm image and reputation improvement due to consistent CSR actions, there are results that confirm the positive effect of ownership concentration on the company's CSR policy (Crisóstomo & Freire, 2015;Godos-Díez, Fernández-Gago, & Cabeza-García, 2012). Specifically in regard to disclosure, there are results that indicate a positive effect of ownership concentration in hands of the government on the CSR disclosure degree (Eng & Mak, 2003).
In Brazil, the strength of the main shareholder stands out a able to influence on company's policies. The interest in company legitimacy as well as improvement of reputation and image is very much associated with the identity of the main shareholder, and this situation can make them prioritize CSR and its respective disclosure. It is intuitive to suggest that more social and sustainability actions for the company motivated by a pursuit of legitimacy and reputation gain may conduct to a higher disclosure degree on these actions, which would be associated to a broader comprehensiveness of the CSR and sustainability reports, as proposed in the following hypothesis: Hypothesis 3: There is a positive relation between ownership concentration in hands of a controlling shareholder and the level of comprehensiveness on CSR reports.
Society's growing concern with environmental issues has put pressure on firm management so that management is aware of these issues, and such management concern may lead to also improve other aspects of social responsibility. Environmentally sensitive sectors of the economy are those that involve a higher risk of environmental impact and as such are more susceptible to criticism and penalties regarding their activities (Reverte, 2009). This higher attention on these companies can motivate a more intense social and environmental action, considering that they must be more aware to preventing environmental hazards and can also be seeking to conduct more social action in general in the search for legitimacy, and image improvement.
In Brazil, the Law 10.165/2000, which deals with the National Environmental Policy, in its annex VIII classified the economic activities according to industry propensity to environmental impact in three levels: low, medium, and high environmental impact potential. There are areas of the economic trade that are not classified in either of these levels. Previous studies noted that companies of more environmentally sensitive areas disclosed more social and/or environmental information, probably due to the higher visibility that these companies now have, and the need for adapting to more rigorous requirement standards concerning the environment (Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens, 2011;Brammer & Pavelin, 2004;Crisóstomo, Souza, & Parente, 2012;Reverte, 2009). From this discussion, it is plausible to propose that reports containing CSR information disclosed by companies from riskier industries would be more comprehensive than those of other companies with lesser potential for environmental impact, as expressed in the following hypothesis: Hypothesis 4: CSR Reports disclosed by firms from sector with higher environmental risk according to the National Environmental Policy present superior level of comprehensiveness compared to reports from other companies.
The Stakeholder theoretical framework proposes that there is a virtuous cycle between CSR and company performance, under the argument that CSR actions are able to create value to the company, since society has a positive sensibility for this type of corporative action (Baron, Harjoto, & Jo, 2011;Freeman, Wicks, & Parmar, 2004;Salazar, Husted, & Biehl, 2012;Waddock & Graves, 1997). The Slack Resource Theory, which proposes a better financial performance as generator of more resources availability that can be directed to CSR, is an additional argument in favor of the aforementioned virtuous cycle between profitability and CSR. In turn, more CSR action requires more disclosure, which naturally has a cost that is more easily handled by companies with more profitability. This way, more profitability might allow for more complete and comprehensive CSR reports, as hypothesized bellow: Hypothesis 5: Profitability positively contributes for the degree of comprehensiveness of CSR reports.
Firm size has been suggested as favorable to CSR policy due to the higher availability of resources, either infrastructure or money, for the execution of social policy. Furthermore, as the company grows, it gathers more visibility and interacts with a broader group of stakeholders, facing a greater demand for CSR, as well as for disclosure of these actions, which is of interest since it will contribute to improve its reputation and pursue of legitimacy (Andrade, Bressan, Iquiapaza, & Moreira, 2013;Artiach, Lee, Nelson, & Walker, 2010;Lourenço & Castelo Branco, 2013;Orlitzky, 2001;Ullman, 1985;Ziegler & Schröder, 2010). In this sense, it is proposed the hypothesis that larger companies will invest more in the completeness and comprehensiveness of their CSR reports as a way to better meet society's information demands for information, as synthesized in the hypothesis that follows: Hypothesis 6: The size of the company influences positively the comprehensiveness degree of its CSR reports.

Methodology
The data was analyzed through qualitative and quantitative approach. Initially, GRI reports were qualitatively analyzed. Then, a quantitative analysis was conducted, including a detailed description of the sample and the measurements used. Then, tests for the difference in means on the degree of comprehensiveness on CSR reports were run, and econometric models estimated to assess the drivers for the degree of comprehensiveness on CSR reports.

Content Analysis
To obtain an indication of CSR items disclosed by companies and the comprehensiveness of the information that accompanied each item disclosed in terms of vision and objectives, management approach, and performance indicators, it was used the content analysis table developed by Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011) to evaluate the comprehensiveness of the information about social responsibility disclosed in companies' Annual Report.
Content analysis technique requires a coding structure based in precise rules about information, representative of the content analysis characteristics, that is reliable and facilitates the interpretation of the data (Bardin, 2011). The coding structure of the content analysis was oriented by the GRI Sustainability Report Guidelines, version G3, with the intent of avoiding possible disparities between the items analyzed, since the analysis contemplates reports from the period comprising 2010-2013. Figure 1 presents the coding structure used in the content analysis in the shape of decision tree previously adopted from Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011). The codification tree consists of two dimensions: (i) the content, and (ii) the type of information. The dimension "content" refers to the information content and comprises two levels of analysis: areas and items. The "area" corresponds to the sustainability performance categories defined according to the GRI Guidelines. The "items" correspond to the specific indicators defined for each area. The dimension "type of information" is intended to verify the information comprehensiveness for each disclosed item considering the nature of the information it contains, examining if it contemplates (a) vision and objectives (VO); (b) management approach (MA); and (c) performance indicators (PI), which are considered essential aspects to the effectiveness of accountability (Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens, 2011;Robertson & Nicholson, 1996;Vuontisjärvi, 2006). Content analysis requires the selection of a "unit of analysis" that can be characterized as a content segment that can be allocated to a certain category (Guthrie, Cuganesan, & Ward, 2008). In line with previous studies, this work used sentences as units of analysis (Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens, 2011;Guthrie, Cuganesan, & Ward, 2008;Hackston & Milne, 1996). Sentences that served as script for coding decisions correspond to the indicators defined by the GRI Guidelines for sustainability reports, according to the details in Appendix A (Frame 1).
To verify the presence of an item and the type of information revealed by it in the report reading, the identification code was initially searched in the GRI content index (for instance, "EC1", "EN26", "SO1"). In those instances when the report did not present the GRI summary, phrase structures and key terms were used (for instance, "local suppliers," "health and safety," "organizational environment," "human rights") to locate items in the reports.
The verification that an item identified in the report contemplates VO (vision and objectives) type of information was conducted using key terms or words that denote intention, policies, values or objectives of the company in the CSR context, such as: "The correct residue management is a Level 2: Items Level 1: Areas commitment of [...]," "The development and qualification of our collaborators are amongst our main values". To verify MA (management approach) information, terms that indicate initiatives, actions, projects and programs effectively implemented by the company were used, such as: "In 2013, we offered more than 305 thousand hours of training [...]," "The focus of private social investment continues to be social inclusion with emphasis in education. Among the initiatives developed in 2013, we can mention [...]," "US$ 415 million in funding and credits were liberated via […], the local suppliers' development program of..." Finally, information about PI (performance indicators) was considered the data expressed in quantitative (absolute, relative, graphic) or qualitative values that indicate progresses or deficiencies on the company's CSR performance. As examples of PI information, we can mention: "During the year, no accidents with our own collaborators were registered [...]," "In 2013, we had our lowest turnover index of the past three years -7.8%, with 9% in 2012," "The supplier expense percentage of the surroundings in relation to the total supplier expenses was: 2011 -3.8%, 2012 -4.0%, 2013 -4.9%."

Index for the degree of comprehensiveness of CSR report
Information obtained in the content analysis served as the basis for the construction of two metrics to measure the degree of comprehensiveness of the reports (IDCR), which aim to reveal the extension to which a company discloses the types of information for the items reported. To evaluate the degree of comprehensiveness, or completeness, of CSR reports disclosed by Belgian companies, Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011) developed an index represented by the quotient between the number of items for which all three types of information (VO, MA and PI) are disclosed and the number of items disclosed by the company. By this formulation, only the items for which the three types of information (VO, MA and PI) are simultaneously presented are taken into account.
Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011) argue that items covered by only one of the disclosure types (VO, MA or PI) are vague and cannot capture the contextualization of CSR disclosure. In this sense, Wood (1991) observes that formal policies might not be reflected in firm behavior or programs that denote CSR performance, as well as performance indicators might be due to institutional factors such as laws, regulations, changes and adjustments in business activities, being not consequence of firm CSR decisions, which, according to McVea & Freeman (2005), consist in the sum of the company's obligations for a specific group of stakeholders. On the other hand, there can be social programs within the companies that have high social performance without support from any kind of formal policy (Wood, 1991). This way, the disclosure of policies, goals and/or objectives, accompanied by performance actions or indicators associated to them, as well as the disclosure of indicators associated to policies or programs and actions might provide some relevant information about the firm CSR. This argumentation suggests that the simultaneous disclosure of a minimum of two types of information might have the potential of meeting the stakeholders' informative demands concerning the firm CSR.
This work uses two metrics to measure the degree of comprehensiveness of CSR reports: the index proposed by Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011), and another proposed new index that also takes into account the item with the three types of information, and also those items covered by two types of information (VO, MA or PI).
The index proposed by Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011) is obtained by the quotient between the number of items disclosed for which there is exposition involving the three types of information (VO, MA or PI) and the total items about which the company disclosed information, according to equation (1).

IDCR1 =
Number of items for which all three types of information (VO, MA, PI) are disclosed (1)

Number of items disclosed by the company
The index for the degree of comprehensiveness of the report (IDCR1) defined by this metric reveals the measure in which a company discloses all three types of information for the items it reports. This way, if a company discloses only one CSR item, and about this one item all three types of information is disclosed, the degree of comprehensiveness of the report will be 1.0. The same way, if a company discloses the totality of items considered, and all of them contemplate the three types of information, the degree of comprehensiveness of the report will also be 1.0. On the other hand, a company might disclose information for a huge number of items without this reflecting in the degree of comprehensiveness of the report, if the disclosed items do not simultaneously contemplate the three types of information.
Considering that the simultaneous disclosure of two types of information is already able to transmit a certain degree of information (Wood, 1991), this work proposes an alternative metric to analyze the degree of comprehensiveness of the reports that takes into account items covered by a minimum of two types of information (VO, MA and PI). Following the strategy of previous studies that attributed weight to the types of information disclosed (Al-Tuwaijri, Christensen, & Hughes II, 2004;Hughes, Anderson, & Golden, 2001;Wiseman, 1982), the degree of comprehensiveness of CSR reports disclosed by companies was obtained by the ratio between the weighted sum of the items that takes into account exactly two types of information (VO and MA; VO and PI; or MA and PI) and those that incorporate exactly the three types of information (VO, MA and PI) with a higher weight, and the total number of reported items. (2)

Number of reported items
In equation (2), IDCR2 is the index for the degree of comprehensiveness of the report. VOMA is the number of items that catches exactly information of types VO and MA. VOPI is the number of items that gets exactly information of types VO and PI. MAPI is the number of items that catches exactly information of types MA and PI. VOMAPI is the number of items that disclose simultaneously the three types of information, VO, MA and PI.

Model
To contrast the research hypotheses, tests for the difference in means were executed for the degree of comprehensiveness of the reports. Mainly, econometric models were estimated, in which the degree of comprehensiveness of the reports (IDCR1 and IDCR2) is the dependent variable. The estimated models are based on equation (3): In equation (3) IDCR i,t is the degree of comprehensiveness of the report of company i during the period t. Table 3 sums up the operationalization of independent variables used in the model. D_ISE is a dummy that indicates the presence of company i in the ISE index during year t.
MAIN_SHARE is the proxy for the ownership concentration in the hands of the main shareholder. Two proxies are used for such concentration. Firstly, it is used a dummy indicating the presence of a major shareholder, i.e., a shareholder that holds more than 50% of the voting shares (D_MAJOR). Alternatively, it is also used the proportion of shares in the hands of the main shareholder (CONC1). D_ENV is a dummy variable that is set to 1 when the firm belongs to high risk industry, i.e., an industry with high potential to cause environment damage according to the Law 10.165/2000. The size of the company (SIZE) is proxied by the natural logarithm of the firm total assets. The profitability is proxied by the return on assets (ROA). Dummy variables of firm industry and year are also included in the model.

Sample
The sample used is compose for 265 annual observations from 98 companies listed in BM&FBovespa that published Social Responsibility Reports according to the GRI guidelines in the 2010-2013 period. Companies whose CSR reports were available for download in the GRI website until June 31 st , 2015 were included in the sample. The sample comprises an ample range of firm industries which is important in such studies (Table 1). The option for companies listed in the stock exchange was made because public companies have the ownership structure as more relevant. Besides, these companies tend to adopt broader and more comprehensive information disclosure policies in order to pursue reduction of informative asymmetry (Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens, 2011;Branco & Rodrigues, 2008). Additionally, financial data are timely available and subject to auditing processes. And the option for CSR reports that follow the GRI guidelines was made because of the worldwide recognition they have been reaching and the continuous increase of its use as a tool for communicating the companies' responsible behavior before the stakeholders, as well as the structured format according to guidelines, principles and indicators that allow for a more adequate analysis of the information about CSR performance in various dimensions (Brown, de Jong, & Levy, 2009;Nikolaeva & Bicho, 2011). Table 2 shows the number of reports that disclose at least one information item (vision and objectives, management approach, performance indicator) about CSR practices in each GRI sustainability dimension. Results show that CSR reports disclosed by Brazilian companies via GRI have a higher proportion of information disclosed than of non-disclosure for all sustainability dimensions. The less contemplated is the economic sustainability dimension, maybe due to the fact that companies might be prioritizing more specific social and environmental aspects of sustainability, which may be seen as the main objective of the report. It must be noted, in this sense, the highlighted number of reports with information about the dimension labor practices and decent work (99.2%), and environmental issues (97.7%).  Table 3 presents the frequency of reported items of all GRI sustainability dimensions, and the type of information disclosed about each one. Results indicate that there is a higher proportion of disclosure for the three types of information (vision and objectives, management approach, performance indicator) when compared to non-disclosure. The higher proportion of information on the social and environmental performance indicator (PI) is worth mentioning, reaching 75.7%. That means a possible trend for the disclosure of concrete information on CSR.  Table 4 contains the types of information disclosed for each sustainability dimension defined by the GRI. A positive result is that the largest proportion of the disclosed items (26.4%) present the three types of information (VO, MA and PI). 62.7% of the reports present information containing two or three types of information (VO and MA; VO and PI; MA and PI; and VO, MA and PI). Regarding disclosed information covered by only one type of information (VO or MA or PI), information of the PI (performance indicators) type stands out in comparison to the other two, in the opposite direction of Belgian companies, that prioritize information about specific actions according to Bouten, Everaert, Van Liedekerke, De Moor, & Christiaens (2011). The economic sustainability dimension presents a higher degree of comprehensiveness according to the proportion of items covered by information of two or three types, which is 71.7% (sum of VO and MA; VO and PI; MA and PI; and VO, MA and PI), followed by the sustainability dimensions Labor practices and decent work (66.6%), and Environment ( It can be noticed from Graphic 1 that the proportion of CSR items covered by information of three types (VO, MA and PI) had relevant and consistent growth during the period analyzed, in contrast to the decline, or maintenance, of items that present only one or two types of information. This denotes a trend of improvement in the comprehensiveness degree of CSR reports. This evaluation is deepened using the comprehensiveness indices (IDCR1 and IDCR2). Table 5 presents the descriptive statistics of the comprehensiveness index of CSR information per GRI sustainability dimension, measured by both metrics proposed (equations 1 and 2) The IDCR1 index considers comprehensive the item that incorporates the three types (VO, MA and PI) ( Table 5, Panel A), and IDCR2 takes into account items that contemplate information of two types (VO and MA; VO and PI; MA and PI) or of three types (VO, MA and PI) for each item disclosed (Table 5, Panel B).

Evolution of the degree of comprehensiveness of CSR reports
The higher proportion of items in some dimensions observed in Table 4 is observed in the index for the comprehensiveness of CSR reports exhibited in Table 5. In fact, the test for the difference in means (F) among the indexes in GRI sustainability dimensions shows that there are differences among the average index of each sustainability dimension. In reality, there are CSR sustainability dimensions more covered by the reports disclosed by Brazilian companies. As in the proportion of items disclosed, the Economic, Labor practices and decent work, and Environment dimensions seem to present superior degrees of comprehensiveness. Data exhibited in Table 6 indicates that the average degree of comprehensiveness of the information present in the Brazilian companies' CSR reports (IDCR1 and IDCR2) has increased during the period. By both metrics used (IDCR1 and IDCR2) it can be observed an average growth in the index for comprehensiveness of the reports along the period of study. The test F for the difference in means among the years (ANOVA) showed that indeed there is a difference in mean annual values of the index for comprehensiveness. The average degree of the index of comprehensiveness is actually superior in more recent periods. In addition to the annual comparison, the degree of comprehensiveness of the reports was also compared between periods of two years using t-test. The results confirm the superiority of the average index of comprehensiveness in more recent periods (Table 7). Results exhibited in Tables 6 and 7 support the proposition of hypothesis 1 that the degree of comprehensiveness of the CSR information disclosed by Brazilian companies has evolved throughout the years. This means that companies have evolved in the direction of improvement in the quality of the information disclosed in their CSR reports. More than a natural evolution of the CSR disclosure process, it is possible that the aforementioned BM&FBovespa bulletin "Report or Explain" has been an additional factor for the increase in the degree of comprehensiveness of the information disclosed in the latest two-year period.

Driving factors of the degree of comprehensiveness in CSR reports
According to the hypotheses proposed, it was suggested that some firm attributes might interfere in the degree of comprehensiveness of the CSR report: company presence in the ISE index, existence of a major shareholder, high risk firm industry as established by Law 10.165/2000, as well as firm profitability and size. Table 8 brings an initial analysis by comparing the average index for the degree of comprehensiveness in CSR reports, measured by both metrics proposed (IDCR1 and IDCR2), between firms listed in the ISE index and non-participant firms. From these results, it can already be noted that in fact the index for comprehensiveness in CSR reports of the ISE firms is higher for both metrics used. This is a strong indicator that belonging to the ISE might actually be a factor that stimulates a higher completeness of the CSR report, as suggested by hypothesis 2 (Table 8, Panels A and B). A comparative analysis of the average degree of comprehensiveness in CSR reports between companies with and without a major shareholder shows that this degree is higher for companies that have one major shareholder by both metrics used (Table 9, Panels A and B). This result is in the direction of what was proposed in hypothesis 3, signaling that a controlling shareholder tends to invest more in CSR and in the respective disclosure, in order to gain reputation and improve firm image.  Table 10 contains results of the comparison of comprehensiveness degree average in CSR reports between companies considered environmentally impactful according to the Law 10.165/2000. It is noted that statistically there is no significant difference between the average degree of comprehensiveness in CSR reports from companies from riskier industries when compared to those not from high risk industries, as proposed by hypothesis 4. The analysis of the specific index for the degree of comprehensiveness for the environmental issue, i.e., an index that considers the comprehensiveness of the report taking into account only the environment sustainability dimension, was done to deepen this analysis. Results shown in Table 11 denote a higher degree of comprehensiveness of CSR reports for companies from environment riskier industries, as suggested by hypothesis 4.  Table 12 shows the results of the estimates of models proposed in equation (3) that have the degree of comprehensiveness of CSR reports as dependent variable. The findings go in the same direction as those obtained by the tests for the difference in means between company groups. First of all, it can be observed that the presence in the ISE index is a company attribute that has a very positive effect on the degree of comprehensiveness of CSR reports, as hypothesized. In fact, being listed in the ISE (D_ISE) seems to be able to stimulate a broader disclosure of CSR and sustainability actions, because of the company's larger visibility and the search for gain image and reputation, as well as to seek legitimacy, as suggested in hypothesis 2.
In this same line of argumentation about the objective, the ownership concentration held by the main shareholder is another attribute of the company that contributes to increase the degree of comprehensiveness of CSR reports, as proposed by hypothesis 2. This positive effect is observed with the use of the two proxies proposed, which are: the presence of a major shareholder (D_MAJOR) ( Table 12, models 1 and 3), and the proportion of ownership concentrated in hands of the main shareholder (CONC1) ( Table 12, models 2 and 4). A controlling shareholder, besides tending to have a long-term perspective in the company ownership, has his/her name very much associated with the company, which is a strong motivating factor to pursue gaining in image and reputation for the company as proposed under hypothesis 3.
As foreseen by hypothesis 4, the fact that the company is considered potentially aggressive to the environment according to Brazilian environmental law (D_AMB) contributes positively for a higher degree of comprehensiveness in CSR reports. The fact that the firm is listed as a high environment risk company according to its sector actually seems to be a factor that stimulates -or even forces -the company to present CSR reports with a higher degree of comprehensiveness.
Also as theoretically foreseen, firm size and profitability of the company has a positive influence on the comprehensiveness degree of CSR reports.

Conclusions
This study made an analysis of the comprehensiveness of CSR information disclosed by Brazilian companies based on social responsibility reports that follow the GRI guidelines for sustainability reports. The work defined as comprehensive the contextualized disclosure that comprises information about vision and objectives, management actions, and performance indicators associated with the CSR items. This way, a more comprehensive report is the one that presents more information, i.e., consisting of a more complete or more comprehensible report. The reports' comprehensiveness was analyzed via a content analysis technique that resulted in the definition of an indicative index for the degree of comprehensiveness of the CSR report.
The research used 265 annual reports of corporative social responsibility, elaborated according to GRI guidelines, from 98 Brazilian companies listed in BM&FBovespa during the period 2010-2013. Results show that, in fact, there are firm attributes that interfere in the degree of comprehensiveness of CSR reports of Brazilian companies. Besides that, it was also noted a significant evolution in the degree of comprehensiveness of CSR reports. The evolution in the comprehensiveness degree of CSR reports of Brazilian companies was observed by verifying degree of comprehensiveness significantly higher in more recent periods.
Despite the progress in the degree of comprehensiveness of CSR reports, it is clear that there is still a low comprehensiveness degree in CSR reports in Brazil. The measurement of the degree of comprehensiveness takes into account vision and objectives (VO), management approach (MA) and performance indicators (PI). It is observed that companies tend to report more information regarding the execution of effective actions of social performance, as performance indicators and management approach, in detriment of vision and objectives. Thus, it can be said that Brazilian companies still have to advance in disclosing their behavior regarding CSR.
About the firm attributes that influence the degree of comprehensiveness of CSR reports, there are in fact company attributes that matter in this context. As suggested, there is evidence that companies more concerned with sustainability -here included environmental, social, and management aspects, brought together by the presence in the ISE -present a higher degree of comprehensiveness in CSR reports. Companies that compose the ISE index in fact fill many requirements associated to its social concern and sustainability. Companies with such traits tend to elaborate better CSR reports that will therefore be more comprehensive and complete. This attention to CSR reports might have a positive effect on the legitimacy of firm activities, as well on firm reputation and image.
Ownership concentration in hands of the main shareholder is also another company attribute that contributes to increase the degree of comprehensiveness of CSR reports. This positive effect of ownership concentration on the quality of CSR disclosure possibly comes from the controlling shareholder's influence on the company's policies, and his/her interest in legitimacy and improvement of the firm image that, in Brazil, is very much associated with the reputation of the shareholder himself/herself. It is possible to suggest that the firm social and sustainability action, motivated by the search for legitimacy and gain of reputation, leads to extensive sustainability disclosure of these actions in companies with highly concentrated ownership in hands of the main shareholder, resulting in a higher degree of comprehensiveness of CSR reports.
It is also important to report that companies from industries considered to have higher environmental impact, according to the Law 10.165/2000, present a higher level of comprehensiveness in the disclosure of information on the specific environmental sustainability dimension. Although the Law 10.165/2000 does not mention disclosing of social information, the fact that it lists a group of firm industries more potentially aggressive to the environment might be motivating these companies to have more effective environmental concern, as the literature has suggested. In this sense, this possible enforcement might also have effect in the quality of CSR reports' that, this way, tend to be more comprehensive, or complete, in the disclosure of information concerning environmental concerns.
Results obtained with this work amplify the knowledge about the voluntary disclosure of CSR information, since it assesses the disclosure of a group of companies that follow GRI guidelines, which nowadays constitute the main reference for disclosure in the field of corporative social responsibility in Brazil. The work also contributes for the criticism of CSR reports, adding empirical evidence for the quality of disclosed information and, still, for the advancement of studies related to the disclosure of CSR actions, by identifying types of information that might better describe the social operation practices and assessment of results of these actions, in view of the efficacy of communication with stakeholders. residues EN17 -Relevant indirect emissions of greenhouse gases, by weight. EN19 -Emissions of substances that damage the ozone layer, by weight. EN20 -NOx, SOx and other significant atmospheric emissions, by type and weight. EN21 -Total water discard, by quality and destination. EN22 -Total residue weight, by type and discard method. EN23 -Number and total volume of significant spillage. Products and services EN26 -Initiatives to mitigate the environmental impacts of products and services, and the extension of the reduction of these impacts. EN27 -Recuperation of products and its packages. Conformity EN28 -Significant fines and non-monetary sanctions resulting of nonconformity to environmental laws and regulations.

Human Rights
Investment and buying processes practices HR1 -Inclusion of contractual clauses and/or evaluation relative to human rights in investment contracts. HR2 -Evaluation of hired companies and critical suppliers regarding human rights and the actions taken. Non-discrimination HR4 -Cases of discrimination and the actions taken. Child labor HR6 -Activities identified as of significant risk-occurrence of child labor, and the actions taken to contribute to abolish child labor. Forced labor or analogous to slavery HR7 -Activities identified as of significant risk-occurrence of forced labor or analogous to slavery, and the actions taken to contribute to eradicate forced labor or analogous to slavery. Conformity Significant fines and non-monetary sanctions resulting of violation of Human Rights.
Labor practices and decent work Work LA1 -Situation of workers, by type of job, work contract and region.
LA2 -Employee turnover, by age group, gender and region. Relationship between workers and governance LA4 -Employees included in collective negotiation agreements. LA5 -Notification in advance regarding operational changes, including if this procedure is specified in collective negotiation agreements. Health and security at work LA7 -Rates of lesions, occupational diseases, days missed, absenteeism, and deaths related to work, by region. LA8 -Ongoing programs of education, training, counseling, risk prevention and control to assist employees, their families or members of the community in relation to severe illnesses. Training and education LA10 -Employee trainings, discriminated by job category. Diversity and equal opportunities LA13 -Composition of the groups responsible for corporative governance and employees discriminated by category, according to gender, age group, minorities, and other diversity indicators. LA14 -Proportion of base salary between men and women, by functional category.

Employee satisfaction
Policies or actions directed to the employees' quality of life, and results of satisfaction surveys with the employees.

Conformity
Fines and reparations from labor claims.

Product liability
Client health and security PR1 -Phases of the products and services cycle of life in which impacts in health and security are evaluated aiming improvement, and indication of products and services subject to these proceedings. Labeling of services and products PR3 -Type of information about products and services required by labeling proceedings, and the percentage of products and services subject to these demands. Client satisfaction PR5 -Information and results of satisfaction surveys with clients. Marketing communications PR6 -Accession programs to laws, norms and voluntary codes regarding marketing communications, including publicity, sales, and sponsorship. Conformity PR8 -Complaints regarding privacy violation or loss of client data. PR9 -Serious fines for non-conformity with laws and regulations regarding the supply and usage of products and services.

Society
Community SO1 -Nature, scope, and efficacy of any programs and practices to evaluate and manage the impacts of activities in communities, including entry, operation, and exit.