افشا و هزینه سرمایه سهام در بازارهای نوظهور: مورد برزیل

In this paper, we conjecture that the weak association between disclosure and cost of equity capital found in the literature (Botosan, 1997) can be caused by the high-level corporate disclosure environment found in the United States. We hypothesize that in low-level corporate disclosure environments the variability in disclosure practices across firms will be larger than in the United States, and, consequently, the marginal effect of voluntary disclosure policies will be higher. Using a newly developed Brazilian Corporate Disclosure Index (BCDI), our results confirm this hypothesis. Disclosure is strongly associated with ex ante cost of equity capital for Brazilian firms. The results are more pronounced for firms with less analyst coverage and low ownership concentration, as expected.

There is an important strand of the financial accounting literature that investigates the relation between disclosure and cost of equity capital (Botosan, 1997, Botosan and Plumlee, 2002, Hail, 2002, Francis et al., 2005 and Chen et al., 2003). The basic idea is that higher levels of disclosure contribute to a reduction in information asymmetry between managers and investors and, consequently, cause a reduction in the idiosyncratic component of cost of equity capital (Verrechia, 2001 and Diamond and Verrechia, 1991). However, results of these investigations have not been conclusive (Botosan, 1997). Some authors (e.g., Hail, 2002) argue that the absence of statistical and economically significant associations between disclosure and cost of capital can be the result of measurement problems because both variables are not directly observed and proxies need to be used. In this paper, we investigate another possibility. We conjecture that the weak relation between firm-level disclosure measures and cost of equity capital is not significant in the United States because the overall disclosure level is already high and firm-level actions do not have a significant marginal impact. Consequently, the weak association between disclosure and cost of capital observed in the United States may result from low variation in disclosure levels because the mandatory disclosure threshold is already high. We believe that firms' actual disclosure policies depend on their incentives. However, incentives only play an important role if the firm's chosen level of disclosure is superior to the minimum level required by the market regulator. If the firm's policy put its disclosure level below the minimum required level, then it has to be increased by force of regulation. Thus, if the minimum disclosure level is already high, many firms will not adopt their optimal policies and instead will adopt the minimum level. Consequently, the disclosure variation will be reduced in the whole sample of firms. Conversely, in environments where the required minimum disclosure level is not so high, it is more likely that firms will present a higher cross-sectional variation in actual (adopted) disclosure policies. Based on prior research (Lopes & Walker, 2008), this high variation in disclosure levels is what we expect to see in Brazil. One could argue about the relevance of a single-country analysis. Recent cross-sectional studies (Francis et al., 2005) investigated the relation between disclosure indexes and cost of equity capital for a sample of firms extracted from 34 countries. We believe that more detailed within-country studies can complement the results of cross-country investigations. We believe there is considerable sample selection bias in the databases used in recent work. In Francis et al. (2005), for example, only 10 Brazilian firms are covered and there is no discussion about sample selection procedures and representativeness. Other studies that investigate corporate governance arrangements across a large number of firms have the same problem. Doidge et al. (2007) consider only 28 Brazilian firms of which 14 are cross-listed. Lang, Raedy, and Wilson (2006) only consider 1 Brazilian firm. We believe these small and biased samples can compromise the results. To investigate our hypothesis, we built a detailed disclosure index and applied it to a more representative sample of firms listed in Brazil. To proxy for disclosure, we built the Brazilian Disclosure Index (BCDI), which measures disclosure across 6 components and 47 specific attributes. The index is applied to the 50 most liquid shares traded on the São Paulo Stock Exchange (BOVESPA) for the years 1998, 2000, 2002, 2004, and 2005. Our sample is very close to the Bovespa Index, which is composed of the most liquid shares. The index is based on a set of questions used in previous research (Botosan, 1997 and Francis et al., 2005) and adaptations to reflect Brazilian regulations and accounting standards. Table 1 features the questionnaire and the percentage of positive answers. Our questionnaire was not sent to the firms or to analysts; rather, it is based on objective answers obtained from public sources of information — annual reports, websites, BOVESPA filings, and the files obtained from the Brazilian Securities and Exchange Commission (CVM). Answers are binary, with 1 indicating a positive answer and 0 indicating a negative answer. We believe that the BCDI provides a comprehensive picture of corporate disclosure policies in Brazil. We are not aware of a detailed investigation of disclosure policies done outside the United States. To measure cost of equity capital, we used the model proposed by Ohlson and Juetnner-Nauroth (2005), which has been used in previous papers (e.g., Francis et al., 2005). Table 1. Composition of BCDI and percentage of positive answers over time. 1998 (%) 2000 (%) 2002 (%) 2004 (%) 2005 (%) General information about the company — GI 1. Does the annual report provide a general description of the firm's business? 93 100 96 100 94 2. Does the annual report provide a description of the major corporate goals for the future? 33 38 57 65 74 3. Are these corporate goals expressed quantitatively? 13 13 28 27 54 4. Does the report describe the major markets where the firm operates? 82 91 100 100 92 5. Are these markets described quantitatively? 36 55 66 71 88 6. Does the annual report describe the major products or services that the firm is involved with? 82 89 98 98 98 7. Does the report provide quantitative information about the firm's major products or services? 44 60 72 71 88 8. Does the annual report describe the general business and economic environment where the firm is immersed? 80 81 91 94 92 9. Are the main corporate events, like the building of a new plant, internal restructurings, and acquisition of new equipments, described in detail? 76 79 85 94 94 10. Are these main corporate events described quantitatively? 36 36 49 55 82 Relationship with employees and managers — REM 11. Does the firm inform the number of employees? 56 74 85 96 90 12. Does the firm inform the mean compensation per employee? 24 34 47 55 64 13. Are the training and developing investments described quantitatively? 36 36 62 63 82 14. Does the firm provide quantitative information about employees like turnover, satisfaction, and value added per employee? 20 28 57 76 60 15. Does the firm provide information about management compensation? 33 55 70 82 62 Non-financial information about markets, sales and products — NFI 16. Does the report inform the firm's market share? 27 51 51 63 86 17. Are the revenues detailed by product, business segment, currency, and quantity? 60 68 77 84 88 18. Are the revenues informed in quantity as well as currency terms? 64 68 83 82 82 19. Is there a time series of revenues that allows for the calculation of growth rates (3 or more years)? 31 30 36 49 66 20. Are the prices of each product sold informed? 7 11 9 12 30 21. Is there information provided about new products or services that will be launched? 7 13 21 18 64 Forecasting — F 22. Did the firm inform sales forecasts for the next year? 0 2 2 6 16 23. Did the firm inform sales forecasts for more than one year ahead? 0 2 2 2 10 24. Did the firm inform earnings forecasts for next year? 0 0 0 2 12 25. Did the firm inform earnings forecasts for more than one year ahead? 0 0 0 0 4 26. Did the firm inform cash flow forecasts for the next year? 0 0 0 0 2 27. Did the firm inform cash flow forecasts for more than a year ahead? 0 0 0 0 2 28. Did the firm inform forecasted investments in research and development and other intangible assets? 0 2 9 8 26 29. Did the firm inform market share forecasts? 0 0 0 0 12 Discussion and analysis of financial data — DAF 30. Did the firm present a time series of revenues (3 or more years)? 18 28 47 65 56 31. Did the firm comment about and explain changes in revenues over the last years? 20 34 43 41 82 32. Did the firm present time series information about cost of goods sold and services provided? 2 6 17 18 30 33. Did the firm comment and explain about cost of goods and services sold and provided over the last years? 4 15 23 22 56 34. Did the firm present time series information about earnings (3 or more years)? 20 36 62 67 62 35. Did the firm comment and explain about changes in earnings over the last year? 24 38 51 45 84 36. Did the firm present time series information about general and administrative expenses (3 or more years)? 0 6 13 22 24 37. Did the firm comment and explain about changes in general and administrative expenses over the last year? 2 13 17 20 50 38. Did the firm present time series information about financial expenses and revenues (3 or more years)? 0 4 15 27 28 39. Did the firm comment and explain about changes in financial expenses and revenues over the last year? 4 13 32 39 58 40. Did the firm inform the amount invested in research and development and other intangible assets? 4 17 23 20 36 41. Did the firm comment and explain about changes in other items like receivables and non-operational expenses? 4 19 30 31 36 Other Information — OI 42. Did the firm present time series information about the Return on Assets (ROA) for 3 or more years? 0 2 11 12 8 43. Did the firm present time series information about the Return on Equity (ROE) for 3 or more years? 7 9 30 33 14 44. Did the firm present time series information about asset turnover for 3 or more years? 0 2 6 8 6 45. Did the firm explain how taxes are calculated? 36 66 81 78 66 46. Did the firm present the Social Balance or Statement of Value Added? 24 55 79 80 70 47. Did the firm present statement of cash flows? 13 32 60 65 66 Table options The results confirm our hypothesis. Initially we observe a lower mean and greater variation in disclosure scores in Brazil than previously reported in the United States (Botosan, 1997), Swiss (Hail, 2002) and for a sample of developed countries (Francis et al., 2005) corroborating the idea that in low-level disclosure environments there will be more cross-sectional variations in disclosure policies. It is very hard to perform international comparisons between our index (BCDI) and the indexes used in previous studies because they use different methodologies and consequently arrive at different results. However, we believe our findings are robust because this phenomenon (huge variation in disclosure policies) has already been reported in the literature, which our evidence corroborates. Francis et al. (2005), see Table 2) reports the disclosure levels in 34 countries according to the Center for International Financial Analysis Research (CIFAR) index. Brazil presents the higher standard deviation of the 34 countries investigated. Standard deviation for the subsample of Brazilian firms is 17.11, while the mean deviation number for all countries is 7.09. For the United States, the standard deviation of the CIFAR metric is 4.42. Thus, we believe there is compelling evidence (based on our study as well as others) that the variation of disclosure levels in Brazil is far superior to the levels in other developed countries and especially to the levels in America. Table 2. Descriptive statistics. Variables Mean Standard Deviation Max Min ks 0.24 0.13 0.66 0.01 BCDI 0.43 0.22 0.78 0.15 SIZE 13.58 4.51 23.8 1.05 MB 1.47 2.34 4.67 0.2 BETA 0.61 0.32 1.32 0.24 DE − 0.26 0.37 − 0.12 − 0.67 ROA 0.17 0.45 0.67 − 0.23 GROWTH 0.08 0.37 0.45 − 0.017 EXTFINANCE 0.22 0.36 0.76 0.00 Number of observations 276 firm-year observations for 1998, 2000, 2002, 2004, and 2005 This table presents descriptive statistics for a sample selected of the 50 most liquid shares (trading volume) traded on the São Paulo Stock Exchange (BOVESPA: where ks refers to cost of equity capital calculated according to the Easton PEG ration ( Easton, 2004; Hail & Leuz, 2005); BCDI refers to the score on the Brazilian Corporate Disclosure Index; SIZE is the natural logarithm of total assets; MB measures the market-to-book ratio; BETA is the stock market beta; DE is the natural logarithm of the debt-to-equity ratio. ROA is return over operational assets calculated as the ratio between annual operating profits and non-current assets; GROWTH is the change in revenues from year t to year t − 1; EXTFINANCE is the ratio of foreign debt outstanding divided by total long-term liabilities — this measure relates to amortized cost. Table options Our main analysis reveals that disclosure levels measured by the BCDI are negatively associated with cost of equity capital for our sample of Brazilian firms using a panel data specification and several control variables as suggested by Larcker and Rusticus, 2005 and Nikolaev and van Lent, 2005. This result shows a more significant association between voluntary disclosure actions and cost of capital for firms immersed in the low-disclosure Brazilian regime than previously found in the United States. Additionally, we investigate the relation between the six BCDI components and cost of capital. Interestingly, results show that the relation between disclosures of non-financial information is positively correlated with cost of capital. One possible explanation for this result is that firms can reveal valuable proprietary information about their business activities through voluntary disclosure of non-financial information (Verrechia, 2001). One alternative explanation for this result is that non-financial information is disclosed more frequently and, consequently, is more used by speculators which induce volatility in stock returns thus increasing cost of capital. Unfortunately, we do not have evidence to prove which of the following explanations is better suited for our sample. We also investigate whether the effect of disclosure on cost of equity is more pronounced for firms that receive less analyst coverage. We hypothesize that disclosure and analyst coverage act as substitutes, and the effect of disclosure on firms' cost of equity capital is less pronounced for firms that receive more attention from analysts. Our results confirm this hypothesis and show that the impact of disclosure on cost of equity is three times higher for firms with lower coverage. Our results are robust for controlling for cross-listing. An increase of 1 point in BCDI causes a decrease of 9 basis points in cost of equity for firms that receive more coverage from analysts and a decrease of 25 basis points for firms that receive less coverage. In addition, we examine the impact of ownership concentration on the relation between disclosure and cost of capital. We also expect disclosure to be less important for firms with higher ownership concentration. Controlling shareholders have direct access to insider information and do not depend on public disclosures. The results confirm our expectations: an increase of 1 point in BCDI causes a decrease of 27 basis points in cost of capital for firms with low ownership concentration. For firms with high ownership concentration, the relation is not statistically significant. We also investigate the likely determinants of BCDI scores and find, as expected, that growth opportunities and board composition are positively related to the association between disclosure and cost of capital. We perform several robustness checks and controls. Our results appear to be robust to changes analysis, Fama–Macbeth regressions, the use of instrumental and proxy variables, and other attempts to uncover endogeneity and other econometric problems. Our results continue to be statistically significant in these alternative specifications. Additionally, we also investigate the relation between BCDI and cost of debt capital and find a strong negative statistical association (as expected). Thus, we believe all results are robust and statistically reliable with important economic meanings. We contribute directly to the literature on the relation between disclosure and cost of capital (Botosan, 1997, Botosan and Plumlee, 2002, Hail, 2002 and Francis et al., 2005) by showing that high disclosure levels are strongly associated with lower cost of capital for firms immersed in countries with huge variation in disclosure policies. Previous research on this topic has been based on samples of firms from developed countries with low dispersion in disclosure levels. Our research also contributes to recent studies that investigate the determinants of the actual properties of accounting reports (Ball et al., 2000, Ball et al., 2003 and Ball and Shivakumar, 2005), which suggest that financial reporting practices depend on managers' incentives to provide informative numbers and not on standards and regulations. This literature, however, is silent about the effect of firm-level actions designed to improve the quality of financial reports. We show that financial reporting practices of firms with incentives to produce high-quality reports reduce significantly their cost of equity capital even under inimical circumstances. Our research does not directly address the policy implications related to disclosure levels. However, our results coupled with other previous works (Lopes & Walker; Ball et al., 2000 and Ball et al., 2003) seem to suggest that optimum disclosure levels (and the quality of accounting reports associated) are a function of firms' incentives and that an increase in these levels are related to a decrease in cost of equity capital for firms immersed in low-disclosure-level countries. Consequently, these results seem to suggest that rules intended to increase the disclosure threshold in countries with low institutional disclosure levels seem reasonable and could provide important reductions in cost of capital for firms. However, additional research must be conducted on this topic before any reliable conclusion is established. The rest of the paper is organized as follows. Section 2 explains the motivation behind the work and develops the hypothesis. Section 3 details the research design employed. Section 4 documents the data selection process, the construction of the BCDI, and the results of our econometric specifications. Section 5 offers additional analysis. Section 6 presents a conclusion.

We investigate the relation between disclosure and cost of capital for Brazilian listed firms. We select the most liquid shares on the São Paulo Stock Exchange (BOVESPA) in 1998, 2000, 2002, 2004, and 2005. To proxy for disclosure level, we built the BCDI, which measures disclosure over six dimensions. To proxy for cost of capital, we use the model proposed by Hail and Leuz (2005) based on Ohlson and Juetnner-Nauroth (2005) and Easton (2004). Using a panel data approach, we show that there is a significant negative association between disclosure and cost of equity capital for the firms in our sample. Additionally, we show that this relation is more pronounced for firms that receive less attention from analysts and have dispersed ownership structures. For these firms, an increase of 1 point in the BCDI results in a decrease of 26 (for firms with less coverage) and 27 (for firms with dispersed ownership) basis points in cost of capital. We also investigate the impact of BCDI scores on the relation between disclosure and cost of capital. While the results are different from those obtained elsewhere, this work has implication for a number of groups. To financial executives, it shows that increased levels of disclosure result in inferior cost of capital, especially for firms that receive less attention from analysts. For regulators, our results demonstrate that firms with superior growth opportunities will adopt better disclosure practices voluntarily even if they are immersed in weak governance and accounting regimes. We also contribute to a recent strand of the international accounting literature that investigates the determinants of the properties of actual reports. This literature (Ball et al., 2000, Ball et al., 2003 and Ball and Shivakumar, 2005) has shown that the properties of published accounting numbers depend more on the incentives managers face to provide informative numbers than on standards and regulations. Our results confirm the hypothesis that firm-level incentives also play a significant role on firms' financial reporting practices.