DISCLOSURE OF FINANCIAL INSTRUMENTS AND CAPITAL COST OF BRAZILIAN COMPANIES

Transcription

1 ISSN: Vol. 1 No. 9 [16-27] DISCLOSURE OF FINANCIAL INSTRUMENTS AND CAPITAL COST OF BRAZILIAN COMPANIES Rodrigo Fernandes Malaquias Universidade Federal de Uberlândia Uberlândia, Minas Gerais, Brazil rodrigofmalaquias@yahoo.com.br Sirlei Lemes Universidade Federal de Uberlândia, Brazil Uberlândia, Minas Gerais, Brazil sirlemes@uol.com.br Luciano Ferreira Carvalho Universidade Federal de Uberlândia, Brazil Uberlândia, Minas Gerais, Brazil lucianofc@netsite.com.br Marcelo Tavares Universidade Federal de Uberlândia, Brazil Uberlândia, Minas Gerais, Brazil mtavares@ufu.br ABSTRACT This work was developed in order to investigate whether the cost of capital of Brazilian companies with shares listed in the U.S. capital market correlates with their level of disclosure of financial instruments. Based on previously literature, we established the hypothesis that the association between the two variables would be negative and statistically significant. The level of disclosure was measured in two different accounting reports: in Form_20F filed at SEC and the Standardized Financial Statements (accounting reports that follow accounting principles generally accepted in Brazil). As the period of analysis included five fiscal years and the sample comprised 24 non-financial companies, 240 different reports were collected resulting in 240 disclosure indexes. Based on the analysis of panel data, the results indicated that for the companies studied, regardless of the type of report analyzed, the cost of capital and the level of disclosure did not behave as two variables negatively related. Keywords: Disclosure, Cost of capital; Brazilian companies. 1. INTRODUCTION The information provided by companies in their financial reports presents itself as relevant to academic research. For Akhtaruddin (2005, p. 400), the effective functioning of capital markets [ ], significantly depends on the effective flow of information between the company and its stakeholders. This research fits in the line of research that seeks the association between the disclosure index and the capital cost associated with companies. Other studies (such as Kristandl and Bontis, 2007) support the hypothesis that this association exists and is negative, i.e., the higher the level of disclosure, the lower the cost of capital of companies. In the Brazilian context, there are also studies that have confirmed this hypothesis (such as Lima et al. 2007; Alencar, 2007). Society for Business Research Promotion 16

2 However, transactions with financial instruments have had serious financial losses in the international economic scene (for example, see: Hernandez, 2003; Aguiar and Hirano, 2003), a fact which shows that the level of information provided by companies in their accounting reports regarding such operations has the potential to have a material effect on user decisions. According to Botosan (1997), there are two different schools of thought that defend the negative association between cost of capital and level of disclosure. The first school argues that a greater level of disclosure increases the liquidity of the assets traded on the stock exchange by the companies, a fact that leads to a reduced capital cost by reducing transaction costs and by increasing demand for the shares of company. In the same line, Amihud and Mendelson (1991) consider that the level of disclosure reflects in the liquidity of company shares, recommending that they engage in disclosure practices. For Young and Guenther (2003, p. 554), "one potential barrier to international capital mobility is differences between domestic and foreign investors in the cost of becoming informed." The second school, still according to Botosan (1997), argues that a greater level of disclosure reduces the risk estimation in projections made by investors. According to Handa and Linn (1993, p. 82), "investors attribute more systematic risk to an asset with low information than to an asset with high information." Barry and Brown (1985, p. 407) also consider that "securities for which there is relatively little information are demonstrated to have relatively higher systematic risk when that risk is properly measured, all other things being equal." In this context, the research question that guided this study was following: do the cost of capital and the level of disclosure of financial instruments have a negative association for Brazilian companies? Based on previous literature and the arguments presented in the previous paragraphs, we have established the hypothesis that the association between the two aforementioned variables is negative and statistically significant, i.e., the higher the value of one variable, the lower the value of the other. Thus the null hypothesis established and tested in this study was the following: H0: The relationship between the level of disclosure and the cost of capital is negative. The alternative hypothesis was that the relationship between the aforementioned variables is not statistically significant. For the proposed analysis, the level of disclosure was measured in two types of reports disclosed by the listed companies: Form_20F and Standardized Financial Statements (DFP), taking into consideration that the informational content of these two kinds of reports is different (Darós and Borba, 2005). One of the contributions related to this work is the analysis of the importance that external users of accounting reports give to more detailed information on financial instruments. The evidence from the empirical results can serve as an argument for such operations to be better detailed in the accounting reports of Brazilian companies, a practice that, as already shown by previous studies (as Murcia and Santos, 2009), still leaves much to be desired. The following section describes the theoretical framework and previous research which has been developed on the theme. As will be shown, despite earlier studies having been developed on disclosure and cost of capital, the analysis of the specific impact of information on financial instruments in the cost of capital of the companies has not been studied by any of the researched works. Therefore, this work is distinguished by allowing a differentiated level of depth on the financial information relating to this subject. 2. THEORETICAL REVIEW Since the action of making evident, according to Aquinas and Santana (1992), means to disclose clearly and the delivering of complete and reliable information on the financial position and results of a company must be guaranteed by the issuance of financial statements, this being one of objectives of accounting (Iudícibus, Martins and Gelbcke, Society for Business Research Promotion 17

3 2006; Santos et al., 2006), it is understood that disclosure is a means for accounting to meet its purposes. In a sample of 122 companies, Botosan (1997) found that for firms that attract a low number of analyst following, the results indicate that a higher level of disclosure is associated with a lower cost of equity. As for companies with a high number of analyst following the results do not show such an association. Kristandl and Bontis (2007), investigating the association between the level of voluntary disclosure and cost of equity in 95 companies listed in the markets of Austria, Germany, Sweden and Denmark, found a negative relationship between the level of future-oriented information and cost of equity of companies, also found on the other hand, a positive relationship between the level of information given and the cost of equity. Analyzing the relationship between level of disclosure and volatility in stock prices in a sample with 600 banks in 31 countries over the period from 1993 to 2000, Baumann and Nier (2004) found that banks with higher levels of disclosure have lower volatility in their share price, consequently, the authors suggest that a lower stock price volatility can result in a lower cost of capital for companies. In the context of Brazilian companies, research has been developed by analyzing the relationship between disclosure and cost of capital. Lima et al. (2007), with a sample of companies listed in BM&FBOVESPA (Bolsa de Valores, Mercadorias e Futuros de São Paulo - São Paulo Stock Exchange) over the period 2000 to 2005, observed an inverse relationship between voluntary disclosure and cost of debt. Alencar (2007), taking into account the nature of information present in the voluntary and mandatory accounting reports, and particularly those that related to the projection of future cash flows, found that the cost of equity and the level of disclosure had a negative relationship. The sample also referred to the companies listed in BM&FBOVESPA, in the periods of 1998, 2000, 2002, 2004 and On the other hand, Mazer (2007), still with a sample of companies listed in BM&FBOVESPA for the period of 2005, found results that do not support the claim that a higher level of transparency implies in a significant reduction in the cost of equity of the companies studied. The disclosure measure used by the author involved content published on the website of the company and also the issues concerning the quality of the information in the financial statements. Nakamura et al. (2006), using different measures to calculate the cost of equity of companies, argue that the model chosen to calculate the cost of capital affect the results of empirical tests, and that we can see a significant impact on one model of disclosure and not significant in another. The differing findings present in some of the studies presented vary not only because of the instrument used to measure the level of disclosure but also in function of the method used to measure the cost of capital. The instrument adopted for this research relates to the disclosure of financial instruments, given their controversial discussion in academic circles (see, for example: Darós and Borba, 2005). In the world financial literature, serious financial losses with operations relating to financial instruments are observed, particularly derivatives, involving traditional cases such as Barings Bank, Metallgesechaift, Procter & Gamble, Daiwa Securities, Orange County, among others (Hernandez, 2003; Aguiar and Hirano, 2003). Believing that a measure of the level of disclosure based on the provision of information on financial instruments is a relevant variable, we will check its association with the cost of capital presented by Brazilian companies in this work. 3. METHOD As the study aims to analyze the association between the level of disclosure and the cost of capital of Brazilian companies in relation to the level of information presented in two different types of accounting reports, the sample consisted of Brazilian companies listed on Society for Business Research Promotion 18

4 NYSE (New York Stock Exchange) and BM&FBOVESPA simultaneously, during the period of analysis. 3.1 Selection of companies and obtaining reports Brazilian companies listed on Bovespa and the NYSE and are subject to the emission of two different accounting reports are those that, in this last Stock Exchange, issue ADRs levels II and III. In this case, such companies issue the reports following the local accounting standard (Brazilian GAAP) and file the Form_20F at SEC. The period of analysis refers to accounting reports for the years from 2002 to The choice of the period of 5 years was considered reasonable to be able to capture the effect of disclosure on the cost of capital. Additionally, the year 2006 was used as the target year for data collection because of the objective of the study to determine the relationship between disclosure and cost of capital for 'pure' financial reports, i.e., based on genuinely Brazilian accounting standards. After 2006, two regulatory frameworks began to change the Brazilian GAAP, in order to converge it to the IFRS (International Financial Reporting Standards) issued by the IASB (International Accounting Standards Board). The first was a review in 2007 of the Corporations Law (Law 6404/76) in which changes were introduced that were close in some points to the IFRS. The second was the creation of the CPC board (Comitê de Pronunciamentos Contábeis - Accounting Pronouncements Committee) which stated to issue standards similar to the IASB. In 2011, there were 41 approved standards already, besides the Framework for the Preparation and Presentation of Financial Statements of the CPC for SME and Interpretations. As such standards were approved, CVM (Comissão de Valores Mobiliários - Securities and Exchange Commission, the Brazilian SEC), made them mandatory for listed companies. In parallel, the CVM (and other regulators such as the Central Bank) determined that all consolidated financial statements were to be issued under IFRS in Thus, as of 2007, the accounting reports of Brazilian companies began to be issued in convergence with IFRS, issuing the reports in full IFRS at the end of After a research based on data from CVM (2008), we identified records of 102 companies that traded some sort of stock in the U.S. market. A filter in those records excluded financial companies, companies that did not negotiate ADRs level II or III and companies with accounting reports unavailable for the years of 2002 to These financial companies were excluded due to the fact that they were already using financial instruments in their operating context. Thus, we obtained a selection of 24 companies, as described in Appendix A. The Form_20F were collected from the records of the NYSE (2008), whereas the Standardized Financial Statements were obtained from the BM&FBOVESPA (2008) site. When more than one report was available for download, we chose one that was initially published by the company. These procedures resulted in the collection of 240 reports: 120 Standardized Financial Statements, and 120 Form_20F covering 5 years from 24 companies. 3.2 Disclosure index and processing of data In order to get the disclosure index, we proceeded to implement a data collection instrument in the accounting reports, which allowed the quantification of the level of disclosure in such reports. The instrument used is an adaptation of one previously used by Lopes and Rodrigues (2007), which aimed to measure the level of disclosure of financial instruments for Portuguese companies, in relation to the requirements placed by the IASB. The original instrument was based on IAS 32 and 39, comprising 54 questions. However, in 2008, these requirements were transferred to IFRS 7, which showed the need to update the data collection instrument. So, we decided to update the questionnaire according to the requirements of IFRS 7 (IASB, 2006c), which resulted in a new questionnaire, consisting of 45 questions, available in Appendix B of this research. This new questionnaire has the same characteristics as the one previously used by Lopes Society for Business Research Promotion 19

5 and Rodrigues (2007), and all variables are dichotomous, with equal weight and adjusted for not applicable items. The adoption of different weights could cause a bias and make the considerations to be obtained with this study unrealistic. Adjustment for non-applicable items was aimed at not hurting businesses that did not disclose some kind of information since they didn t perform that operation. The questionnaire was done through a content analysis (Bauer, 2002). After analysis of the content obtained from the data collection instrument, we built up the scores. These scores are, therefore, a quantitative assessment of the disclosure practices of a company. They are not a qualitative indicator of the value of that information (Khanna, Palepu and Srinivasan, 2004, p. 482). As all variables are dichotomous, when the report had met a certain item, the value 1was assigned to the variable, if not, then the value 0 was assigned. Equation (1) shows how the scores were measured for each company. Where: S: total score for each company, according to the report analyzed; i: starting number for the questions of the instrument to be verified; n: maximum number (45 for this case) for the variation of the sum; d i : the disclosure of each item of the instrument (1, if disclosed, 0 if not shown). S = n d i i = 1 (1) In relation to all relevant issues, this calculation is done according to Equation (2) below. Where: T: total number of questions applicable to each company; i: starting number for the questions of the instrument to be verified; n: maximum number (45 for this case) for the variation of the sum; q i : each question to be applied (1, if the variable is applicable, 0 if not). T = n q i i = 1 (2) Once the total number of items to be applied is obtained, this value should form the denominator of the equation that will return the disclosure index value per company, as shown by Equation (3). S Index = x 100 T Where: S: total score for each company (see Equation (1) of this work); T: total number of applicable questions (see Equation (2) of this work). The calculations for the disclosure indexes were made through the use of spreadsheets. The measure for the cost of capital of companies was obtained from the ECONOMATICA database, and represents the total cost of capital, i.e. the weighted average cost of capital that each of the companies presented in a given period. We established the following as control variables: Total Assets, measured by total assets reported by the company in the respective year of analysis as an indicator of company size; and time of issue of ADRs on the NYSE, as measured by time since the firm issues ADRs level II or III in the said stock, seeking to have a level of transparency indicator for the company. Once the quantitative variables were obtained, we performed statistical tests, the first of which was multiple regression analysis with panel data. The choice of the most appropriate model for testing hypotheses (stacked data, fixed effects or random effects) was made based on Tests Breusch-Pagan and Hausman. The exploratory analysis of variables was performed (3) Society for Business Research Promotion 20

6 initially by using descriptive statistics, followed by calculating the Pearson correlation coefficient. In addition, the Box-Cox transformation was applied to the variables that presented problems for normality using the Kolmogorov-Smirnov test, seeking better fits for the final model. For the quantitative analysis, it was considered that the reflex of the level of disclosure provided by companies in their financial report would be the cost of its capital in the year following that in which the report was released. For example, the level of disclosure of the accounting year 2002 was associated with the cost of capital for the year 2003, and so on. 4. RESULTS After obtaining the quantitative variables needed for analysis, we observed that there was a very high dispersion for the variable that represents the total cost of capital of companies. Thus, we decided, initially, to replace its extreme outliers by the variable median. For an exploratory analysis of the database obtained, we prepared descriptive statistics of the variables, as summarized in Table 1 below. [Insert Table 1 here] As indicated by results in Table 1, the variables do not exhibit normal or approximately normal distribution with a conservative level of significance, i.e., above 20%, and it is understood that this could affect the data analysis by regression model. Thus, we decided to apply the Box-Cox transform to all of them, so that there would be the same treatment pattern of the study variables. With the results obtained, we applied the Kolmogorov- Smirnov test again (nonparametric) for all transformed variables, and the normality indicators improved significantly. However, the variable Total Assets continued to have a low level of significance. We decided then to represent it by the value of its Napierian Logarithm, which did not allow the rejection of the hypothesis of approximately normal distribution at a 5% significance level. With the transformed variables, we decided to analyze the correlation between them, as shown in Table 2. [Insert Table 2 here] As indicated in Table 2, the statistically significant correlations observed occurred between the disclosure indicators and the size, this relationship between the two pairs of variables being positive. For the other variables, there is no evidence of statistically significant correlation, which can be changed taking into account the multivariate analysis, which is the next step of this work. Table 3 below, describes the six models tested. Different models were made considering the variables "disclosure" (Dis.NYSE and Dis.Bov) as explanatory due to the positive, strong and statistically significant correlation observed among them. This correlation could cause problems of multicollinearity when using both in the same model, which could also be solved by the creation of a factor by factor analysis. However, we decided perform the analysis in two different models, which may indicate if the disclosure effect in different reports is dissimilar in the dependent variable. [Insert Table 3 here] In order to compare the standard error of the betas generated in each of the models described, all six models were developed and tested in statistical software. The results of the betas and their standard errors for each of the models are summarized in Table 4, below. [Insert Table 4 here] We must say that the R-square presented by the models was, in all cases, less than 5%, indicating very low levels of capacity of explanation and predictability. Thus, even before checking which models would be best to test the hypothesis established in the work, we Society for Business Research Promotion 21

7 understand that the variables segregated to explain the total cost of capital do not seem to have a significant relationship with this variable. However, this comment can be confirmed (or rejected) based on the analysis which will be developed below. As shown in Table 4, for most dependent variables, the standard error of the generated model based on Fixed Effects is lower (in relation to the beta model, i.e., not absolute value), which could indicate the best settings. This comparison was formalized by the Breusch- Pagan and Hausman tests, as shown below. The Breusch-Pagan test showed that at the level of 1%, the null hypothesis that the stacked data model is better than the fixed effects model, for both types of analysis, was rejected. That is, by this test, the fixed effects model would provide better adjustments. After that, we performed a Hausman test to compare the settings of the fixed and random effects model. At the 5% level of significance, the null hypothesis that the fixed effects model provides the best fit was not rejected. Thus, there is evidence provided by both the Hausman test and the Breusch-Pagan test that the best model to carry out the tests of the coefficients found is the fixed effects model. The next step then consists in the analysis of the signs and significance levels for the betas of the regression model, with the main objective being to check whether the beta for the disclosure variables pointed to a negative and statistically significant relation to the total cost of capital variable. However, the two models generated based on fixed effects did not show a level of significance for the F test below 5%, which indicates that through the comparison of means by analysis of variance (ANOVA), the betas of the model are not statistically different from zero. Thus, it makes no sense to analyze its signal, and this leads to the conclusion that for the companies studied, the level of disclosure of financial instruments does not seem to affect the total cost of capital of such companies. The results we found have the potential to indicate that the market may not have realized the level of disclosure of financial instruments present in accounting reports. This fact would lead to two implications: the liquidity of the shares would remain the same and there would be no risk reduction. Consequently, there would be no negative correlation between the level of disclosure of financial instruments and the cost of equity, which is a component of the total cost of capital. Moreover, one factor that could explain these results is the limitation of this survey, related to the data collection instrument used. A measure for the level of disclosure that involves factors other than the disclosure of financial instruments could point to more consistent results with the existing theory on the subject. Another factor that could explain these results is also linked to the type of measurement provided by the data collection instrument adopted, because it quantifies the level of disclosure without any qualitative measure, insofar that if the information disclosed in reports on financial instruments was positive (gains, for example) or negative (losses, for example), this fact is not captured, according to the methodology described above. 5. CLOSING REMARKS This work was developed in order to investigate, for Brazilian companies with shares listed in the U.S. capital market, if the cost of capital correlates with their level of disclosure of financial instruments. Based on previous literature, we elaborated the hypothesis that the association between the two variables would be negative, i.e., the higher the level of disclosure promoted in financial reports, the lower the cost of capital, and vice versa. The level of disclosure was measured in two different accounting reports: in Form_20F filed with the SEC and the Standardized Financial Reports, these statements in Brazilian GAAP being the ones that companies listed in BOVESPA must provide to the Brazilian capital market. As the period of analysis included five accounting periods, and the sample comprised 24 non-financial companies, we analyzed 240 different accounting reports, resulting in 240 rates of disclosure. Society for Business Research Promotion 22

8 The measures for the total cost of capital variable were obtained from the ECONOMATICA database. The results are based on three models of panel data: stacked data, fixed effects and random effects. The Breusch-Pagan, and the Hausman test indicated that the fixed effects model was the most suitable for the analysis of the results. The hypothesis test indicated that, for the companies and the period studied, the total cost of capital and the level of disclosure did not show a negative relationship, even if the results were controlled by size (measured by total assets) and an indicator of transparency (issuance time of ADRs on the NYSE) of the companies analyzed. This result may have occurred because the level of disclosure of financial instruments, although a relevant item included in the annual financial reports of companies, is not among one of the main items taken into account by external users of accounting reports for risk level evaluation of companies.in other words, if higher rates of disclosure of financial instruments gave users more information and, consequently, lower levels of exposure to risk when making an investment or granting a loan to these businesses, a negative relationship between variables should be confirmed. We would also like to remark that the fact of not taking into consideration only companies with high liquidity belonging to the BOVESPA Index may have been a factor that affected the results obtained by the research, as Alencar (2007) and Lima et al. (2007) observed a negative relationship between the study variables for firms with high liquidity, that is, companies which would tend to be more sensitive to new information released in the capital markets, than firms with lower liquidity. However, Mazer (2007) also examined companies of that said index and, as in this work, did not observe a significant relationship between disclosure and cost of capital. The limits of this study are present in four aspects. The first is related to the measures for the cost of capital, where we considered the total cost of capital. The second limit is the use of a data collection instrument that only measures the level of disclosure of financial instruments, rather than the total level of disclosure. Similar to the second, the third limit is the use of a data collection instrument that measures the level of disclosure quantitatively, therefore failing to verify if the information that was disclosed in the financial reports was positive or negative. The fourth limit refers to the sample used, which is made up exclusively of Brazilian open capital companies listed on the NYSE. This sample may show some bias in that all the companies that comprise it may have similar characteristics to each other in relation to more homogeneous rates of representing the total cost of capital. We suggest for future research: i) increasing the research sample to include publicly traded Brazilian companies not listed on the NYSE, using the variable of participation in the said foreign exchange as a dummy control variable in the models tested, ii) elaborating a similar analysis, but using a data collection tool that quantifies the level of disclosure in relation to other matters, not only to the disclosure of financial instruments and iii) segregating the total cost of capital between cost of equity and cost of debt, checking the impact of segregation on the results to be obtained. REFERENCES Aguiar, A. B.; Hirano, A. (2003) Os Impactos do Fair Value como Base de Valor para Instrumentos Financeiros Derivativos na Atual Estrutura da Contabilidade Um Enfoque Normativo. In: XXVII ENANPAD, 2003, Atibaia, SP. Anais do XXIX ENANPAD. Akhtaruddin, M. (2005) Corporate Mandatory Disclosure Practices in Bangladesh. The International Journal of Accounting. v. 40. p Alencar, R. C. (2007) Nível de Disclosure e Custo de Captial Próprio no Mercado Brasileiro. PhD Thesis in Accounting. São Paulo: FEA/USP. Amihud, Y.; Mendelson, H. (1991) Liquidily, Asset Prices and Financial Policy. Financial Analysis Journal. v. 47. n. 6. p Aquino, W.; Santana, A. C. (1992) Evidenciação. Caderno de Estudos FIPECAFI. n. 5. Barry, C. B; Brown, S. J. (1985) Differential Information and Security Market Equilibrium. Journal of Financial and Quantitative Analysis. v. 20. n. 4. p Society for Business Research Promotion 23

9 Bauer, M. W. (2002) Análise de Conteúdo Clássica: uma revisão. In: Bauer, M. V.; Gaskell, G. (Orgs.). Pesquisa Qualitativa com Texto, Imagem e Som: um manual prático. Petrópolis: Vozes. Cap. 8. p Baumann, U.; Nier, E. (2004) Disclosure, Volatility, and Transparency: an empirical investigation into the value of bank disclosure. Economic Policy Review - Federal Reserve Bank of New York. v. 10. n. 2. p Botosan, C. A. (1997) Disclosure Level and The Cost of Equity Capital. The Accounting Review. v. 72. n. 3. p BM&FBOVESPA Bolsa de Valores de São Paulo. (2008). Available on Bussab, W. O.; Moretin, P. A. (2003) Estatística Básica. 5. ed. São Paulo: Saraiva. CVM Comissão de Valores Mobiliários (2008) Available on Darós, L. L.; Borba, J. A. (2005) Evidenciação de Instrumentos Financeiros Derivativos nas Demonstrações Contábeis: uma análise das empresas brasileiras. Revista Contabilidade & Finanças. n. 39. p Handa, P.; Linn, S. C. (1993) Arbitrage Pricing with Estimation Risk. Journal of Financial and Quantitative Analysis. v. 28. n. 1. p Hernandez, F. G. H. (2003) Derivatives and The FASB: visibility and transparency? Critical Perspectives on Accounting. vol. 14. p IASB International Accounting Standards Board. (2006a) IAS 32 Financial Insruments: presentation. IASB International Accounting Standards Board. (2006b) IAS 39 Financial Insruments: recognition and measurement. IASB International Accounting Standards Board. (2006c) IFRS 7 Financial Insruments: disclosure. Iudícibus, S.; Martins, E.; Gelbcke, E. R. (2006) Manual de contabilidade das sociedades por ações: aplicável às demais sociedades. 6. ed. rev. e atual. 9. reimp. São Paulo: Atlas. Khanna, T.; Palepu, K. G.; Srinivasan, S. (2004) Disclosure Practices of Foreign Companies Interacting With U.S. Markets. Journal of Accounting Research. v. 42. n. 2. Kristandl, G.; Bontis, N. (2007) The Impact of Voluntary Disclosure on Cost of Equity Capital Estimates in a Temporal Setting. Journal of Intellectual Capital. vol. 48. n. 4. p Lima, G. A. S. F.; Salotti, B. M.; Carvalho, L. N. G.; Yamamoto, M. M. (2006) Governança Corporativa. In: Lima, I. S.; Lima, G. A. S. F.; Pimentel, R. C. Curso de Mercado Financeiro: tópicos especiais. São Paulo: Atlas. Cap. 16. p Lopes, P. T.; Rodrigues, L. L. (2007) Accounting for Financial Instruments: an analysis of the determinants of disclosure in the Portuguese stock exchange. The Journal of Accounting. v. 42. p Mazer, L. P. (2007) O Impacto do Nível de Transparência no Custo de Capital Próprio das Empresas do Ibovespa. Master s Thesis in Accounting. Ribeirão Preto: FEARP/USP. Murcia, F. D; Santos, A. (2009) Regulação Contábil e Divulgação de Informações de Operações com Instrumentos Financeiros Derivativos: Análise do Impacto da CVM nº 566/08 da CVM nº 475/08 no Disclosure das Companhias Abertas no Brasil. In: XXXIII ENANPAD, 2009, São Paulo, SP. Anais do XXXIII ENANPAD. Nakamura, W. T.; Gomes, E. A.; Antunes, M. T. P.; Marçal, E. F. (2006) Estudo sobre os Níveis de Disclosure Adotados pelas Empresas Brasileiras e seu Impacto no Custo de Capital. In: XXX ENANPAD, 2006, Salvador BA. Anais do XXX ENANPAD. NYSE New York Stock Exchange. (2008) Available on Santos, J. L.; Schmidt, P.; Gomes, J. M. M.; Fernandes, L. A. (2006) Contabilidade Geral. 2. ed. São Paulo: Atlas. Young, D.; Guenther, D. A. (2003) Financial Reporting Enviroments and International Capital Mobility. Journal or Accounting Research. v. 41. n. 3. p Society for Business Research Promotion 24

10 Appendix A: Sample nº Name ADR Level nº Name ADR Level 1 Aracruz Celulose S.A. III 13 Perdigão S.A. III 2 Brasil Telecom Part. S.A. II 14 Petróleo Brasileiro S.A. II and III 3 Brasil Telecom S.A. II 15 SABESP III 4 Braskem S.A. II 16 Sadia S.A. II 5 Cia. Brasileira de Distribuição II 17 Tele Norte Celular Part. S.A. II 6 Cia. de Bebidas das Américas II 18 Tele Norte Leste Part. S.A. II 7 Cia. Energética de Minas Gerais II 19 Telecomunicações de São Paulo II 8 Cia. Paranaense de Energia III 20 Telemig Celular Part. S.A. II 9 Cia. Siderúrgica Nacional II 21 Tim Part. S.A. II 10 Cia. Vale do Rio Doce II and III 22 Ultrapar Part. S.A. III 11 Empresa Bras. de Aeronáutica III 23 Vivo Part. S.A. II 12 Gerdau S.A. II 24 Votorantim Celulose e Papel S.A. III Source: CVM (2008). Appendix B Data Collection Instrument (Questionnaire) Score (if disclosed and applicable) Financial Instruments - Accounting policies 1 Held for trading assets/liabilities 1 2 Held-to-maturity assets 1 3 Loans and receivables originated by the enterprise 1 4 Available-for-sale financial assets 1 5 Other financial liabilities 1 6 Trade date or settlement date 1 Financial Instruments - Fair Value 7 Fair value of assets and liabilities (grouped by classes) 1 8 Measurement method 1 9 Significant assumptions 1 Financial Instruments - Risks 10 Risk management policy 1 11 Monitoring and controlling policy 1 12 Segregation by risk types 1 13 Description of how those risks arises 1 14 Exposure to risk 1 15 Methods used to measure the risk 1 Derivatives - Accounting Policies 16 Objectives of holding or issuing derivatives 1 17 Accounting policies and methods adopted 1 Derivatives - Hedging 18 Hedging description 1 19 Financial instruments designated as hedging instruments 1 20 Fair values of those financial instruments 1 21 Nature of the risks being hedged 1 22 If hedge accounting is applicable 1 23 Type of hedge relationship adopted 1 Derivatives - Fair value 24 Measurement method 1 25 Significant assumptions 1 Interest Rate Risk Society for Business Research Promotion Sensitivity analysis 1 27 Methods and assumptions used in the sensitivity analysis 1 Currency Rate Risk 28 Sensitivity analysis 1

11 Derivatives - Hedging 18 Hedging description 1 19 Financial instruments designated as hedging instruments 1 20 Fair values of those financial instruments Nature of the risks being hedged 1 ISSN: If hedge accounting is applicable Vol. 1 No. 9 [00-00] 1 23 Type of hedge relationship adopted 1 Derivatives - Fair value 24 Measurement method 1 25 Significant assumptions 1 Interest Rate Risk 26 Sensitivity analysis 1 27 Methods and assumptions used in the sensitivity analysis 1 Currency Rate Risk 28 Sensitivity analysis 1 29 Methods and assumptions used in the sensitivity analysis 1 Other Prices Risk 30 Sensitivity analysis 1 31 Methods and assumptions used in the sensitivity analysis 1 Credit risk 32 Counterparties identification 1 33 Maximum amount of credit risk exposure (by class) 1 34 Analysis of the age of financial assets 1 35 Criteria used to determine allowance 1 36 Significant concentration of credit risk 1 37 Description of the collateral policies 1 38 Information about the credit quality 1 Collateral 39 Terms and conditions (relative to its pledge or associated with its use) 1 40 Carrying amount (if pledged) or fair value (if held) 1 Liquidity risk 41 A maturity analyze, showing the remaining contractual maturities (time bands) 1 42 Description of how manages the liquidity risk 1 Other 43 Impairment losses 1 44 Criteria to determine that there is objective evidence that an impairment loss has occurred 1 45 Total interest income and total interest expense (separately) 1 Source : Addopted from Lopes and Rodrigues (2007). Table 1: Descriptive Statistics of Variables Variables N Mean Standard Deviation Minimum Maximum Kolmog. Smirnov Signif. Cap. Cost ,684 7,748 0,000 34,700 1,121 0,162 DisBov ,952 8,802 17,778 53,333 1,192 0,117 DisNYSE ,444 7,414 29,545 70,455 1,131 0,155 Total Assets (R$ mil) ,183 0,000 TmpADR (years) 120 5,292 2,535 0,000 14,000 1,141 0,148 Notes: Cap Cost: total cost of capital for businesses; DisBov: level of disclosure provided to the Brazilian capital market; DisNYSE: level of disclosure provided to the U.S. capital market; total assets (R$ thousand): the total asset value of companies, in thousands of dollars; TmpADR (year) period, in years that the companies are listed on the NYSE; Kolmog. Sminov: Kolmogorov-Smirnov Statistics for the test of nonparametric normality; Signif.: Significance for the normality test. Society for Business Research Promotion 26

12 Table 2: Correlation between Variables Matrix Variables Cap. Cost DisBov DisNYSE Total Assets TmpADR Cap. Cost DisBov DisNYSE Total Assets TmpADR Pearson Correlation 1,000 0,062 0,103 0,007 0,110 Sig. (2-tailed) 0,503 0,263 0,935 0,233 Pearson Correlation 0,062 1,000 0,684 0,345-0,081 Sig. (2-tailed) 0,503 0,000 0,000 0,379 Pearson Correlation 0,103 0,684 1,000 0,290-0,011 Sig. (2-tailed) 0,263 0,000 0,001 0,905 Pearson Correlation 0,007 0,345 0,290 1,000 0,011 Sig. (2-tailed) 0,935 0,000 0,001 0,909 Pearson Correlation 0,110-0,081-0,011 0,011 1,000 Sig. (2-tailed) 0,233 0,379 0,905 0,909 Notes: Cap Cost: total cost of capital for businesses; DisBov: level of disclosure provided to the Brazilian capital market; DisNYSE: level of disclosure provided to the U.S. capital market; Total Assets, total asset value of companies, in thousands of dollars; TmpADR: period, in years that the companies are listed on the stock exchanges in New York; the calculations were done after the transformation of the variables. Table 3: Models tested in the study Model Dependent Variable Independent Variable 1 Independent Variable 2 Independent Variable 3 Pooled Data Cap.Cost Dis.NYSE Total Assets TmpADR Fixed Effects Cap.Cost Dis.NYSE Total Assets TmpADR Random Effects Cap.Cost Dis.NYSE Total Assets TmpADR Pooled Data Cap.Cost Dis.Bov Total Assets TmpADR Fixed Effects Cap.Cost Dis.Bov Total Assets TmpADR Random Effects Cap.Cost Dis.Bov Total Assets TmpADR Notes: Cap Cost: total cost of capital for businesses; Dis.Bov: level of disclosure provided to the Brazilian capital market; Dis.NYSE: level of disclosure provided to the U.S. capital market; Total Assets: total asset value of companies, in thousands of dollars; TempoADR: period, in years that the companies are listed on the stock exchanges in New York. Table 4: Statistics obtained for the six models tested Variables Pooled Data Fixed Effects Random Effects Variables Pooled Data Fixed Effects Random Effects Dis.Bov (beta) 0,0007 0,0009 0,0006 Dis.Bov (beta) 0,0030-0,0034-0,0015 standard error 0,0007 0,0013 0,0009 standard error 0,0076 0,0144 0,0095 Tot.Assets (beta) -0,0094 0,9838 0,0940 Tot.Assets (beta) 0,0043 0,9959 0,1307 standard error 0,1056 0,4396 0,1680 standard error 0,1084 0,4410 0,1719 TmpADR (beta) 0,1234-0,6737 0,0791 TmpADR (beta) 0,1242-0,3779 0,1593 standard error 0,2689 0,5449 0,3219 standard error 0,2723 0,5354 0,3140 Constant (beta) 7, ,0000 4,6109 Constant (beta) 7, ,0000 4,5435 standard error 3, ,2452 4,7390 standard error 3, ,2621 4,7752 R-Square < 5% < 5% < 5% R-Square < 5% < 5% < 5% Notes: Cap Cost: total cost of capital for businesses; Dis.NYSE: level of disclosure provided to the U.S. capital market; Dis.Bov: level of disclosure provided to the Brazilian capital market, Total Assets: total asset value of companies, in thousands of dollars; TempoADR: period, in years that the companies are listed on the stock exchanges in New York; the calculations were done after the transformation of variables. Society for Business Research Promotion 27