Concentration of Ownership in Brazilian Quoted Companies*

Concentration of Ownership in Brazilian Quoted Companies* TAGORE VILLARIM DE SIQUEIRA** Abstract This article analyzes the causes and consequences of concentration of ownership in quoted Brazilian companies, and is based on a sample of 278 companies. The tests undertaken are in accordance with previous studies by Demsetz and Lehn (1985) and Thomsen and Pedersen (1997), that attempted to answer questions regarding the factors determining the degree of concentration of share ownership within companies, as well as the impact of such factors on financial performance. The results of the study support the conclusion that the degree of concentration of ownership of Brazilian companies is affected by market regulation, by the size of the firm, and by its capital structure. However, only one of the models tested showed a statistically significant relationship between concentration of ownership and financial performance. While the test resulted in a low R 2 coefficient, results did confirm a negative impact of ownership concentration on corporate financial performance. This result indicates that companies with a greater dispersion of stockholder ownership tend to perform better. * Translation: BCBR Business Communications Brazil. ** Economist assigned to the BNDES Planning Department (tsiqueir@bndes.gov.br). 1

1. Introduction In recent years, the subject of corporate governance has attracted more attention from analysts, as a result of the impact of changes in corporate ownership. Such changes can, for example, affect financial performance, investment programs, the growth and even survival of companies. In the light of the above, a number of studies were carried out on European, U.S. and Asian companies starting in the 1980s, with the aim of examining the factors that determine the degree of concentration of share ownership, and their effects on corporate performance. These studies tested the hypotheses arising from the theory of the firm, that factors such as the degree of sector regulation, the size of the form and the volatility of markets in which companies were present, would have a significant impact on the degree of concentration of ownership. Demsetz and Lehn (1985), for example, carried out a pioneering study of a sample of 511 U.S. companies, that identified the following factors that determined concentration of ownership: the degree of regulation, the size of the company and the volatility of profitability. The authors, nevertheless, failed to identify a significant impact of concentration of ownership on corporate performance. Thomsen and Pedersen (1997) tested the model proposed by Demsetz and Lehn, introducing other variables and considering a sample of 388 companies in 12 European countries. The study confirmed the relevance of such variables as size, capital structure and volatility of profitability, and the tests carried out confirmed Demsetz and Lehn s conclusion that concentration of ownership does not have a significant impact on corporate performance. This study investigates the repercussions of changes in the ownership structure of Brazilian companies, considering a sample of 278 quoted companies. The main objective was to analyze the factors that determine concentration of ownership, and their effects on the financial performance of companies. The models mentioned above were tested, and new independent variables were introduced. On the basis of the tests undertaken, we observed that the degree of concentration of stockholder ownership of Brazilian companies is influenced by market regulation, the size of the firm, and capital structure. With regard to the impact of the degree of concentration of stockholder control on the financial performance of companies, a statistically significant negative effect on corporate performance was observed in one of the models considered. While the adjustment in this case generated a very low R 2 coefficient, the result indicates that companies with a greater dispersion of stockholder control tend to achieve a better financial performance. This article consists of four sections plus this introduction: Section 2 presents the points of reference for the theoretical debate from which the hypotheses for analyzing the determinants of the degree of ownership control were drawn; Section 3 characterizes the sample, and defines the variables used in the study; Section 4 presents an analysis of the impact of ownership concentration on performance. 2

2. The Theoretical Debate on the Factors that Determine Concentration of Corporate Ownership The hypotheses tested in this study are drawn from the existing debate within the theory of the firm on the modern corporation, the structure of ownership and the conflict of interests between controllers and managers, that was first discussed in the 1930s by Berle and Means in their classic study The modern corporation and private property (1932). Various studies have been based on the arguments of Berle and Means, with a view to deepening knowledge of the structure of corporate ownership, the impact of changes in this structure, and the relationships between stockholders and managers, as can be seen, for example, from Demsetz (1983), Demsetz and Lehn (1985), Holderness and Sheehan (1988), Shleifer and Vishy (1986), Williamson (1988), Zeckhouser and Pound (1990), Buzzacchi and Colombo (1996) and Thomsen and Pedersen (1997). A good deal of these studies concentrated their investigations on the causes and consequences of changes in ownership structure. Two of the main questions that they sought to answer were: What factors determine the degree of concentration of ownership? And what is the impact of concentration of ownership on the financial performance of companies? For Demsetz and Lehn (1985), the size of the firm, uncertainty and regulation are among the principal factors that influence the concentration of stockholder control of companies. The influence of the size of the firm on the degree of concentration of ownership arises from the assumption that the increase in value of a given portion of the firm would be compatible with a dispersion of ownership. In addition, large companies could also have high costs of capital and a high risk of maintaining the degree of concentration of stockholder control. Demsetz and Lehn, nevertheless, comment that large companies will tend, by virtue of their aversion to risk, to have a low degree of ownership concentration. The volatility of profitability will influence the concentration of ownership to the degree that there is a conflict of interest between managers and stockholders. In this way, this conflict will be of greater or lesser degree, depending on the volatility of markets. An increase in market volatility (associated with changes in prices, technology and market share) would cause an increase in the concentration of ownership, and could even reduce the freedom of action of the managers. The authors propose that this effect could be measured using some parameter of volatility of the financial performance of companies, such as an index of profitability. A high degree of volatility during a certain time period could lead to an escalation in the conflicts between managers and owners, and thus could cause a change in the structure of ownership. The authors argue that the degree of concentration of ownership will tend to be high in markets with a high degree of volatility, with controllers playing also an active role in company business. With regard to the effects of market regulation, Demsetz and Lehn argue that defined operating rules for companies can lead to a reduction in concentration of ownership. In the case of Brazil, the ownership structure of public service utility companies is affected not only by the regulatory structure of the sector, but also by the 3

large stakes held by the State as majority stockholder. It should, nevertheless, be remembered that this second factor is likely to become less relevant over the next few years, as a result of the privatization process for public sector companies. Demsetz and Lehn s study confirmed the relevance of factors such as regulation, company size and volatility of profitability in determining the degree of concentration of stockholder control of companies. Market regulation had a negative effect on the degree of concentration of ownership or, in other words, the definition of operating rules led to an increase in the dispersion of stockholder control, as a result of a reduction in uncertainty. This effect can even reduce the conflict of interest between managers and controllers, increasing the degree of autonomy of the former group. The size of the firm had also a negative effect on the degree of concentration of ownership, indicating that there is a tendency among U.S. companies towards the dispersion of stockholder control, as these increase in size. Volatility of profitability had also a positive effect, showing that there is a direct relationship between the increase in uncertainty with regard to profits and the degree of concentration of stockholder control. In other words, companies in highly volatile markets tend to be tightly controlled, with managers subject to reduced freedom of action. With regard to the effects of concentration of ownership on performance, Demsetz and Lehn failed to identify any significant effect. In addition to the variables mentioned above, Thomsen and Pedersen (1997) introduced a capital structure variable (net equity/total assets) into their model for determining concentration of ownership, with the aim of testing the effect exerted on it by specific investments in large-scale plants. The authors formulated the hypothesis that an increase in the above variable (net equity/total assets) would be accompanied by a reduction in the concentration of ownership. Thomsen and Pedersen tested the model proposed by Demsetz and Lehn, considering a sample of 350 companies from 12 European countries, and confirmed the relevance of variables such as regulation (negative effect), size (negative effect) and volatility of profitability (positive effect). In addition, the capital structure variable introduced by the authors in the model was shown to have a negative effect. The results of the tests thus carried out also confirmed Demsetz and Lehn s result that there was no statistically significant effect of concentration of ownership on performance. With regard to the effects of concentration of stockholder control on financial performance, Thomsen and Pedersen tested three models: a) for the first, in accordance with Demsetz and Lehn (in which performance is a function of regulation, company size, volatility of profitability, capital intensity, technological intensity and concentration of ownership), the degree of concentration of stockholder control did not have a statistically significant effect on performance; b) for the second, based on a study by Zeckhouser and Pound (in which performance is a function of technological intensity and concentration of ownership), concentration of ownership had a statistically significant negative effect on performance for companies with a low degree of technological intensity, but the authors came to the conclusion that this effect was not significant after introducing a variable that designated the country of origin of the company; and c) for the third model, proposed by the authors themselves (in which 4

performance is a function of company growth and concentration of ownership), it was possible to conclude that the relationship between concentration of ownership and performance is influenced by the rate of company growth. In other words, rapid growth is correlated with a higher degree of dispersion of stockholder control, while slower growth corresponds to a higher degree of concentration of ownership. 3. Characterization of the Sample The financial information for the companies considered in this study was taken from the Economática database, from the Information Center of the Brazilian Securities Commission (CVM), from the Stock Exchange Information Center, and from the publication Brazil Company Handbook. Table 1 presents the variables considered in this study, together with their respective descriptions. TABLE 1 Variables with Respective Descriptions VARIABLE DESCRIPTION CON1 Concentration of ownership, measured by the percentage of the total number of shares in the company held by the majority stockholder CON2 Concentration of ownership, measured by the percentage of the total number of voting shares in the company held by the majority stockholder UTIL Dummy variable in which public service utility companies = 1, and other companies = 0 INDUSTRY Dummy variable in which industrial companies = 1, and other companies = 0 REG Dummy variable in which aviation companies and banks = 1, and other companies = 0 DIMEN Proxy for firm size, measured by total assets (average 1994/96) INCERT Proxy for volatility of profitability, measured by the standard deviation of return on equity (1994/96) ESCAP Proxy for capital structure of companies, measured by ratio of net equity to total assets (average 1994/96) NATION Dummy variable in which Brazilian-controlled companies = 1, and subsidiaries of foreign companies = 0 PERF Return on equity (average 1994/96), cut off between +100 and 100, and weighted by net equity TECN Dummy variable in which companies whose technology expenditures are equal to or greater than 0.5 = 1, and less than 0,5 = 0 INTCAP Proxy for capital intensity, measured by ratio of total assets/net sales (average 1994/96) CPV Cost of products sold (average 1994/96) IMPORTS Dummy variable in which sectors suffering serious competition from imports = 1, and other sectors = 0 GROWTH Annual growth rate of net sales (average 1994/96) The sample considered in this study is composed of 278 quoted companies, which were classified on the basis of Thomsen and Pedersen s (1997) proposal into three categories according to the degree of concentration of ownership, namely: dispersed ownership, when the majority stockholder has less than 20% of control, dominant ownership, when the majority stockholder has between 20% and 50% of control, and majority ownership, when the majority stockholder has more than 50% of control. Considering that Brazilian companies have both voting and non-voting shares, the two forms of concentration of ownership were considered in all the tests undertaken. Tables 2 and 3 present the principal characteristics of the sample, according to the classification by concentration of ownership mentioned above. In Table 2, which considers concentration of ownership on the basis of the total number of shares, the sample shows a degree of equilibrium between the three categories of concentration of ownership, with 21% of companies under dispersed control, 48% under dominant 5

control, and 31% under majority control. In Table 3, on the other hand, which considers concentration of ownership on the basis of the total number of voting shares, it can be seen that the vast majority of companies still have highly concentrated stockholder control, with majority control accounting for 68% of the sample. Companies under dominant control account for some 25% of the sample, but companies under dispersed control for only 7%. It is worth pointing out that among the companies presented in Tables 2 and 3, those with a higher degree of dispersion of stockholder control reported superior financial performances during the period under consideration, in terms of all of net sales growth, return on equity, return on assets and net margin. An analysis of the influence of concentration of ownership on financial performance for the set of companies in the sample is presented in Section 5. TABLE 2 Principal Characteristics of the Sample by Concentration of Ownership (% of Total Number of Shares held by Majority Stockholder) VARIABLE ITEM CONCENTRATION OF OWNERSHIP Dispersed (Up to 20%) Dominant (Between 20-50%) Majority (Over 50%) Companies No. 59 134 85 % 21.22 48.20 30.58 Net Sales (US$ ) Average 1994/96 747,372,086 616,356,712 602,835,742 Net Equity (US$) Average 1994/96 426,809,267 662,983,788 692,786,233 Total Assets (US$) Average 1994/96 1,547,788,386 1,751,848,988 1,357,340,821 Cost of Sales (CPV) (US$) Average 1994/96 469,820,845 256,924,340 284,343,090 Net Profit (US$) Average 1994/96 25,443,552 19,248,744 24,508,146 TECN No. 24 58 21 % 40.68 43.28 24.71 UTIL No. 2 7 17 % 3.39 5.22 20.00 REG No. 3 10 6 % 5.08 7.46 7.06 NATION No. 56 123 68 % 94.92 91.79 80.00 INDUSTRY No. 46 99 48 % 77.97 73.88 56.47 IMPORTS No. 18 39 13 % 30.51 29.10 15.29 Net Equity/Total Assets (%) Average 1994/96 47.79 41.65 53.93 Total Assets/Net Sales (%) Average 1994/96 293.53 190.92 432.67 Std. Dev. of ROE Average 1994/96 42.43 142.47 43.20 GROWTH (%) Average 1994/96 28.92 34.03 23.67 ROE (%) Average 1994/96 10.02-9.63-1.78 ROA (Total Assets) (%) Average 1994/96 1.12-2.43-0.13 Net Margin (%) Average 1994/96 5.50 0.09-10.56 NB: ROE = Return on Equity, ROA = Return on Assets. Other variables as abbreviated in Table 1. 6

TABLE 3 Principal Characteristics of the Sample by Concentration of Ownership (% of Total Number of Voting Shares held by Majority Stockholder) VARIABLE ITEM CONCENTRATION OF OWNERSHIP Dispersed (Up to 20%) Dominant (Between 20-50%) Majority (Over 50%) Companies No. 21 69 188 % 7.55 24.82 67.63 Net Sales (US$) Average 1994/96 1,276,138,418 545,834,974 602,719,994 Net Equity (US$) Average 1994/96 760,109,432 454,102,664 667,889,515 Total Assets (US$) Average 1994/96 2,733,540,359 1,512,183,147 1,489,148,294 Cost of Sales (CPV) (US$) Average 1994/96 743,955,105 269,679,262 272,911,707 Net Profit (US$) Average 1994/96 43,060,381 21,178,758 20,888,172 TECN No. 9 30 64 % 42.86 43.48 34.04 UTIL No. 1 3 22 % 4.76 4.35 11.70 REG No. 1 5 13 % 4.76 7.25 6.91 NATION No. 19 65 163 % 90.48 94.20 86.70 INDUSTRY No. 15 55 123 % 71.43 79.71 65.43 IMPORTS No. 5 19 46 % 23.81 27.54 24.47 Net Equity/Total Assets (%) Average 1994/96 44.49 46.79 46.77 Total Assets/Net Sales (%) Average 1994/96 334.91 236.69 299.59 Std. Dev. of ROE Average 1994/96 34.50 48.74 112.65 GROWTH (%) Average 1994/96 26.46 43.09 25.43 ROE (%) Average 1994/96-2.43 0.42-4.42 ROA (Total Assets) (%) Average 1994/96 1.14-1.06-1.20 Net Margin (%) Average 1994/96 4.05 2.87-4.41 4. The Determinants of Concentration of Stockholder Control This section analyzes the principal causes of concentration of ownership, taking as a reference the models proposed by Demsetz and Lehn (1985), and Thomsen and Pedersen (1997). In other words, we consider that the degree of concentration of stockholder control is influenced by a series of variables that include market regulation, company size and volatility of profitability. In this study, tests were undertaken both with regard to concentration of ownership as measured by the holding of the principal stockholder as a percentage of the total number of shares (CON1), (See Models 1 and 2), and as measured by the holding of the principal stockholder as a percentage of the total number of voting shares (CON2), as is shown in Models 3 and 4. 7

Correlation Matrix 1 VARIABLES 1 2 3 4 5 6 7 1 CON1 1 2 UTIL 0.25 1 3 REG 0.01-0.09 1 4 DIMEN -0.05 0.18 0.42 1 5 INCERT -0.04-0.04-0.04-0.05 1 6 ESCAP 0.15 0.18-0.29-0.10-0.18 1 7 NATION -0.18 0.11 0.05 0.09 0.03-0.03 1 MODEL 1 MODEL 2 VARIABLE Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic CON1 Intersection 38.12 1.54 24.81 46.61 4.69 9.93 UTIL 22.23 4.82 4.62 22.58 4.78 4.72 REG 8.41 6.01 1.40 12.15 6.06 2.00 DIMEN 0.00 0.00-2.08 0.00 0.00-1.92 INCERT 0.00 0.00-0.48 0.00 0.00 0.03 ESCAP 0.09 0.05 1.91 NATION -15.14 4.24-3.57 No. 278 278 R² 0.08 0.13 F 5.68 6.79 Within Model 1, two variables were identified as statistically significant: UTIL (public service utility companies), and DIMEN (size of company), both of which had an extremely weak effect on concentration of ownership. These results are at variance with those presented by Demsetz and Lehn (1985), and Thomsen and Pedersen (1997), who identified a negative effect on concentration of stockholder control, both with regard to public service utility companies and to company size. In other words, an increase in the size of the firm leads to greater dispersion of stockholder control. As Demsetz and Lehn pointed out, aversion to risk is likely to cause owners to share control of companies as they grow in size. A possible explanation for the differences in results may lie in the forms of financing adopted by Brazilian companies, most of which tend to raise funds through bank loans or issues of debt securities (such as debentures and bonds) rather than by issuing shares. In this way, an increase in size of such companies would not necessarily be accompanied by a dispersion of stockholder control. The positive effect of public service utility companies (energy, telecommunications, water and sanitation) on concentration of ownership results from the dominant role of the State as majority stockholder in such companies. At the same time, the privatization of such companies will undoubtedly lead to a dispersion of stockholder control. In addition to the variables mentioned in the previous model, it was verified in Model 2 that the high degree of regulation (REG), the capital structure (ESCAP), and 8

the country of origin of the company (NATION), also play a significant role in determining the concentration of ownership in the observed companies. Against expectations, aviation and financial companies (REG), both of which operate in sectors with a high degree of regulation, were observed to have a positive impact on concentration of stockholder control. The result obtained on the basis of the sample of Brazilian companies can be partly explained by the high proportion of families among the controlling stockholders of the companies under consideration. The capital structure variable showed a positive effect on concentration of ownership, indicating that the higher the specific investment in large-scale plants, the higher the degree of concentration of stockholder control among the companies in the sample. This result was the opposite of that found by Thomsen and Pedersen, who identified a negative effect of capital structure on concentration of stockholder control, thereby confirming the hypothesis presented in their study, namely: the higher the specific investment in large-scale plants, the lower the degree of concentration of stockholder control. The introduction of the variable NATION allowed us to differentiate between companies under Brazilian ownership, and the subsidiaries of foreign companies. The negative effect of this variable indicates that an increase in the number of Brazilian companies in the sample is accompanied by a reduction in the concentration of company ownership. In other words, subsidiaries of foreign groups tend to present a higher degree of concentration of stockholder control than Brazilian companies. The volatility of profitability, measured by the standard deviation of return on equity, showed a statistically insignificant coefficient in both of the models considered. The principal conclusion drawn from this result is that for the observed companies, the uncertainty with regard to profitability or to volatility of profitability does not yet have a significant impact on the degree of concentration of ownership. On the basis of the sample of observed Brazilian companies, it can even be stated that the conflict of interest between stockholders and managers is unaffected by market volatility (whether this is due to prices, technology or market share). That is, it can be said that the degree of concentration of stockholder control, and in the last analysis, the freedom of action of managers, is not affected by market volatility. By contrast, the studies by Demsetz and Lehn (1985), and Thomsen and Pedersen (1997), identified a positive correlation between volatility of profitability and concentration of ownership. In other words, an increase in volatility of profitability led to an increase in concentration of stockholder control. Models 3 and 4 considered the concentration of ownership as measured by the holdings of voting shares by majority stockholders, CON2. The R 2 and F tests produced low coefficients, with results that were worse than those obtained in Models 1 and 2. 9

Correlation Matrix 2 VARIABLE 1 2 3 4 5 6 7 1 CON2 1 2 UTIL 0.10 1 3 REG -0.02-0.09 1 4 DIMEN -0.08 0.18 0.42 1 5 INCERT 0.07-0.04-0.04-0.05 1 6 ESCAP 0.02 0.18-0.29-0.10-0.18 1 7 NATION -0.09 0.11 0.05 0.09 0.03-0.03 1 MODEL 3 MODEL 4 VARIABLE Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic CON2 Intersection 59.24 1.71 34.72 65.99 5.34 12.35 UTIL 11.10 5.35 2.07 12.00 5.45 2.20 REG 3.77 6.67 0.57 4.27 6.90 0.62 DIMEN 0.00 0.00-1.68 0.00 0.00-1.61 INCERT 0.00 0.00 1.13 0.00 0.00 1.19 ESCAP 0.00 0.06 0.05 NATION -7.97 4.82-1.65 No. 278 278 R² 0.03 0.04 F 1.79 1.65 Table 4 presents the measures of convergence and dispersion of the variables considered in this section. It may be observed that the concentration of ownership is greater when the majority stockholdings of voting shares are considered. The sample included 26 public service utility companies and 19 aviation and financial companies. Of the companies in the sample, 247 were under Brazilian control. TABLE 4 Measures of Convergence and Dispersion VARIABLE AVERAGE STD. DEVIATION TOTAL MIN MAX CON1 (%) 39.39 23.34 3.22 99.84 CON2 (%) 59.66 25.23 9.62 100.00 UTIL (No.) 26 0 1 REG (No.) 19 0 1 DIMEN (US$ Million) 1,574.06 3,939.83 7.80 31,619.21 INCERT (%) 90.89 586 0.38 9,091.76 ESCAP (%) 46.71 28.74-188.77 98.12 NATION (No.) 247 0 1 NB: Sample composed of 278 companies. 10

5. Concentration of Ownership and Corporate Financial Performance In this section we analyze the effects of concentration of stockholder control on corporate financial performance. Tests were carried out on the basis of models proposed by Demsetz and Lehn (1985), Zeckhouser and Pound (1990) and Thomsen and Pedersen (1997), in which performance was assumed to be a function of regulation, company size, technological intensity, capital structure and growth. We also introduced variables not considered by the above authors, such as IMPORTS (a proxy for sectors which suffer a higher degree of competition from imported products) and CPV (the 1994/96 average of the cost of products sold). Model 5, which is presented below, and which results from tests carried out with all the variables considered in this study, was originally proposed by Demsetz and Lehn, and subsequently retested by Thomsen and Pedersen. The concentration of ownership considered in this model, CON1, is measured by the percentage of the total number of shares held by the largest stockholder. Correlation Matrix 3 VARIABLES 1 2 3 4 5 6 7 8 1 PERF 1 2 UTIL 0.25 1 3 INDUSTRY -0.08-0.43 1 4 TECN -0.08-0.18 0.30 1 5 INCERT -0.11-0.03 0.01-0.05 1 6 INTCAP 0.29 0.33-0.25-0.12-0.09 1 7 CPV 0.49 0.00-0.15-0.13-0.04 0.16 1 8 CON1 0.03 0.33-0.13-0.09-0.03 0.00-0.11 1 MODEL 5 VARIABLE Coefficient Std. Error t Statistic PERF Intersection -33.29 11.20-2.97 UTIL 72.56 18.70 3.88 INDUSTRY 21.13 9.09 2.33 TECN 1.07 6.61 0.16 INCERT -0.01 0.00-1.16 INTCAP 0.06 0.02 2.84 CPV 0.06 0.01 8.74 CON 1 0.04 0.14 0.28 No. 239 R² 0.34 F 17.17 In accordance with the model, we may conclude that concentration of ownership does not have a statistically significant effect on the financial performance of companies. This result coincides with the conclusions of Demsetz and Lehn, and Thomsen and Pedersen, with regard to a sample of U.S. and European companies, for which stockholder control did not affect corporate financial performance. 11

The variables that showed statistically significant results were UTIL, INDUSTRY, INTCAP and CPV, all of which had a positive effect on performance, even if for the last two variables, the effect was a weak one. Variables such as TECN (proportion of spending on technology), INCERT (volatility of profitability), DIMEN (company size) and IMPORTS (dummy variable, in which 1 = companies in sectors facing a higher degree of competition from imported products), did not produce significant results for the tests undertaken. Thomsen and Pedersen s study identified a positive effect of company size on performance. The impact of capital intensity on financial performance was not significant. The study by the above authors identified a negative effect with regard to public service utility companies, although the majority of such companies considered by them were private companies operating in markets with regulatory structures that favored competition, leading to greater transparency of financial performance. In Model 6, the concentration of ownership, CON2, was measured by the share of the total voting shares held by the majority stockholder. The principal results are presented below, and it may be verified that the concentration of ownership, in terms of the share of the total voting capital held by the majority stockholder, also failed to demonstrate a statistically significant effect on company performance, as was the case in Model 5. The variables UTIL, INDUSTRY, INTCAP and CPV nevertheless produced statistically significant coefficients. Correlation Matrix 4 VARIABLES 1 2 3 4 5 6 7 8 1 PERF 1 2 UTIL 0.25 1 3 INDUSTRY -0.08-0.43 1 4 TECN -0.08-0.18 0.30 1 5 INCERT -0.11-0.03 0.01-0.05 1 6 INTECP 0.29 0.33-0.25-0.12-0.09 1 7 CPV 0.49 0.00-0.15-0.13-0.04 0.16 1 8 CON2-0.09 0.15-0.01-0.01 0.07-0.06-0.20 1 MODEL 6 VARIABLE Coefficient Std. Error t Statistic PERF Intersection -29.16 12.21-2.39 UTIL 75.16 18.07 4.16 INDUSTRY 21.19 9.09 2.33 TECN 0.95 6.60 0.14 INCERT -0.01 0.00-1.15 INTCAP 0.06 0.02 2.79 CPV 0.00 0.00 8.57 CON2-0.04 0.12-0.32 No. 239 R² 0.34 F 17.17 12

Public service utility companies and industrial companies demonstrated positive effects on performance. Capital intensity was also shown to have a significant, albeit weak positive effect in performance, indicating that the greater the degree of capital intensity of companies, the superior their respective financial performance. Financial Performance, Technological Intensity and Concentration of Ownership Models 7 and 8 are based on the relationship between financial performance, technological intensity, and the degree of concentration of stockholder control, that was originally proposed by Zeckhouser and Pound (1990), and retested by Thomsen and Pedersen (1997). In both studies, companies with a low degree of capital intensity showed statistically significant relationships, although these did not coincide, with the first model showing a positive correlation involving an integrated effect by the two independent variables, while the second identified a negative effect. In contrast to the studies mentioned above, the results obtained in Models 7.1, 7.2 and 7.3, that are presented below, did not show any statistically significant relationships (concentration of ownership is measured by the proportion of the total number of shares held by the majority stockholder, with the table merely showing the signs of the coefficients). MODEL 7.1 MODEL 7.2 (TECN Low Intensity) MODEL 7.3 (TECN High Intensity) VARIABLE Coefficient Coefficient Coefficient PERF Intersection + + + TECN - CON1 - - - No. 244 141 103 NB: The results were not statistically significant. An analysis of Models 8.1, 8.2 and 8.3, which relate performance to the degree of technological intensity, and the proportion of the total voting shares held by the majority stockholder, also failed to produce statistically significant results. MODEL 8.1 MODEL 8.2 (TECN Low Intensity) MODEL 8.3 (TECN High Intensity) VARIABLE Coefficient Coefficient Coefficient Std. Error t Statistic PERF Intersection + + 29.28 13.62 2.15 TECN - CON2 - - -0.32 0.21-1.50 No. 239 136 103 R² 0.01 0.00 0.02 F 1.62 0.00 2.26 NB: The results were not statistically significant. 13

If it had been possible to find a significant relationship for Model 8.3, we could conclude that in technology-intensive sectors, the concentration of stockholder control is inversely correlated with financial performance. In other words, the lower the degree of concentration of control in technology-intensive companies, the better the financial performance, and vice-versa. In the case of Brazil, a possible inverse relationship between concentration of ownership and financial performance could be interpreted as the result of a method discovered by companies for increasing access to new technologies, which thereby overcame a restriction on growth. Companies in technology-intensive sectors can, for example, form technological partnerships with foreign companies, with a view to improving access to technology. These associations could take the form of the sale of a stake in the company s capital to the new partners. Thomsen and Pedersen found an inverse relationship between concentration of ownership and performance for companies with a low degree of technological intensity. Basing their conclusions on a sample of European companies, the authors demonstrated that companies in sectors with a low degree of technological intensity tend to achieve better results in so far as they show a greater dispersion of stockholder control. It should nevertheless be remembered that after considering the variable NATION, the same authors rejected these results. Financial Performance, Company Growth and Concentration of Ownership Models 9 and 10, which are presented below, relate financial performance to company growth and concentration of ownership. These models were first proposed by Thomsen and Pedersen (1997). Model 9, which considered the concentration of ownership as measured by the share of the total voting shares held by the majority stockholder, CON1, failed to demonstrate statistically significant results. Correlation Matrix 5 VARIABLE 1 2 3 4 5 1 PERF 1 2 GROWTH -0.17 1 3 CON1 0.00-0.07 1 4 INDUSTRY -0.03 0.08-0.14 1 4 NATION -0.10-0.05-0.19-0.11 1 14

MODEL 9 VARIABLE Coefficient Std. Error T Statistic PERF Intersection 51.50 18.03 2.86 GROWTH -0.20 0.07-2.85 CON1-0.11 0.17-0.64 INDUSTRY -5.68 9.28-0.61 NATION -25.07 13.11-1.91 No. 259 R² 0.04 F 2.90 The results of Model 9 with regard to the categories for concentration of stockholder control (dispersed, dominant and majority) also failed to produce statistically significant results. In contrast to Thomsen and Pedersen s (1997) model, growth did not have a significant effect on the performance of Brazilian companies in any of the regressions considered. The same authors concluded that an improvement in the performance of the European companies may be associated with a greater dispersion of stockholder control as such companies grow. VARIABLE PERF MODEL 9.1 (CON1 up to 20%) MODEL 9.2 (CON1 between 20% and 50%) MODEL 9.3 (CON1 above 50%) Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic Intersection 185.38 53.73 3.45 25.14 20.67 1.22 47.68 23.30 2.05 GROWTH -0.28 0.17-1.70-0.20 0.09-2.26-0.19 0.14-1.36 NATION -7.05 23.24-0.30 11.50 12.50 0.92-24.82 16.48-1.51 INDUSTRY -147.69 48.26-3.06-23.35 18.30-1.28-11.38 20.02-0.57 No. 56 125 78 R² 0.17 0.06 0.05 F 3.68 2.41 1.29 NB: The results were not statistically significant. Model 10, which relates performance to company growth and concentration of ownership (as measured by the proportion of the total number of voting shares held by the majority stockholder, CON2), produced statistically significant coefficients, although the joint explanatory power of the set of variables was low, with an R 2 coefficient of only 5%. Growth of net sales had a negative, albeit weak effect on company performance. This effect may be interpreted as a trade-off on the part of companies, between higher growth, and the maintenance of good financial performance. That is, as companies grow, they tend to become less demanding of good results. At the 5% significance level, it may be accepted that concentration of ownership influences performance, with the negative effect indicating that the reduction in the degree of stockholder control leads to an improvement in the performance of the observed companies. 15

Correlation Matrix 6 VARIABLE 1 2 3 4 5 1 PERF 1 2 GROWTH -0.17 1 3 CON2-0.09-0.05 1 4 INDUSTRY -0.03 0.08-0.01 1 5 NATION -0.10-0.05-0.08-0.11 1 MODEL 10 VARIABLE Coefficient Std. Error t Statistic PERF Intersection 63.16 17.89 3.53 GROWTH -0.21 0.07-2.93 CON2-0.27 0.15-1.77 INDUSTRY -4.95 9.11-0.54 NATION -25.25 12.80-1.97 No. 259 R² 0.05 F 3.61 When Model 10 is considered by category of concentration of ownership (dispersed, dominant and majority), it may be observed that the results are not statistically significant. VARIABLE MODEL 10.1 (CON2 up to 20%) MODEL 10.2 (CON2 between 20% and 50%) MODEL 10.3 (CON2 above 50%) Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic Coefficient Std. Error t Statistic PERF Intersection 203.17 88.68 2.29 15.09 37.43 0.40 86.80 47.10 1.84 GROWTH -0.43 0.40-1.09-0.32 0.12-2.66-0.55 0.25-2.22 INDUSTRY 4.19 47.41 0.09 27.73 20.46 1.36-55.73 30.35-1.84 NATION -166.21 73.72-2.25-15.61 31.90-0.49 3.98 41.69 0.10 No. 21 63 180 R² 0.26 0.13 0.05 F 1.98 2.82 3.05 NB: The results were not statistically significant. 6. Final Considerations Our study of concentration of ownership allows us to conclude that factors such as market regulation, company size and capital structure influence the degree of concentration of stockholder control of Brazilian companies. The high degree of control exerted by the State on public service utility companies, often as the majority stockholder, explains the positive impact of regulation 16

on concentration of ownership. With privatization, however, the degree of concentration of stockholder control of such companies will undoubtedly become more fragmented. Against expectations, other sectors with a high degree of regulation that were considered in the study, such as aviation and finance, were also shown to have positive effects on concentration of ownership. This result may be explained by the high proportion of families that exert stockholder control over the companies in these sectors. The positive effect of company size on the degree of concentration of ownership may be explained by the forms of funding chosen by Brazilian companies, which tend to prefer bank loans or debt issues to equity issues. In this way, unlike U.S. and European companies, the increase in size of Brazilian companies does not result in a dilution of stockholder control. The capital structure of companies, as measured by the ratio of net equity to total assets, was shown to have a positive effect on concentration of ownership. This indicates that larger specific investments in large-scale plants are accompanied by an increase in the degree of concentration of stockholder control. The introduction of the variable NATION into the tests led to the observation that the subsidiaries of foreign groups tend to show a greater concentration of stockholder control than is the case for Brazilian companies. With regard to market volatility, it was observed that the coefficient of volatility of profitability did not have statistically significant effects on the model. It can thus be stated that market volatility (due to changes in prices, technology or market share) does not have a significant impact on the degree of concentration of Brazilian companies, and may thus have no particular impact on the conflict of interest between stockholders and managers. The model proposed by Thomsen and Pedersen, that relates performance to growth, and to the degree of concentration of ownership, produced a test result, albeit with a very low R 2 coefficient, that identified a negative effect of concentration of ownership on performance. This result indicates that a greater degree of dispersion of stockholder control may be associated with an improvement in corporate financial performance. Finally, it may be said that the relationship between the degree of concentraion of stockholder control and the financial performance of companies can be better understood, when variables such as technological intensity or company growth are included in the analysis. For companies that require high rates of growth, it is thus possible to increase the degree of dispersion of stockholder control as a way of promoting access to new technologies, as well as of realizing new investments. In addition, it should be remembered that, in the final analysis, the performance of companies is influenced by the degree of competition within their respective markets. 17

Appendix 1. Causes of Concentration of Ownership Demsetz and Lehn s Models MEASUREMENT OF LA5 LA20 LAH LA5 LA20 LAH CONCENTRATION OF OWNERSHIP Intercept -2.10 (11.9) -1.01 (6.0) 3.83 (13.5) -1.92 (9.7) -0.68 (4.0) 4.21 (13.8) UTIL -1.20-1.44-2.07......... (10.0) (12.6) (10.8) FIN -0.47-0.49-0.78......... (4.3) (4.7) (4.4) MEDIA 0.70 (3.4) 0.75 (3.8) 0.94 (2.8) 0.68 (3.3) 0.72 (4.0) 0.91 (2.8) EQUITY -3.51 (2.7) -2.98 (2.4) -4.25 (2.0) -3.90 (2.9) -3.82 (3.3) -5.09 (2.5) SE 17.94 (5.9) 11.14 (3.9) 26.46 (5.4) 14.83 (4.4) 5.60 (1.9) 20.11 (3.9) SE² -39.38 (4.1) -25.26 (2.8) -59.70 (3.9) -31.98 (3.2) -12.14 (1.4) -44.49 (2.9) No. 511 511 511 406 406 406 R² 0.33 0.36 0.33 0.12 0.09 0.09 F 41.5 46.5 41.2 13.4 9.3 10.0 NB: LA5 = proportion of total shares held by 5 largest stockholders; LA20 = proportion of total shares held by 20 largest stockholders; LAH = Henfindahl index of concentration of ownership, calculated as the sum of the square of the percentage holdings controlled by each stockholder; UTIL = public service utility companies; FIN = financial sector companies; MEDIA = media sector companies; EQUITY = proxy for size; and SE = standard error of rate of return; t Statistic in brackets. Thomsen and Pedersen s Models VARIABLE MODEL 1 MODEL 2 (Ownership Share - CON) (Ownership Share - CON) Intercept 36.08 (13.60) 36.32 (2.45) UTIL -13.11 (-1.30) -11.09 (-1.11) MEDIA -1.62 (-0.16) -1.85 (-0.18) SIZE -0.51 (-3.65) -0.45 (-3.21) STD 0.55 (2.43) 0.40 (1.75) STD² -0.006 (-1.78) -0.005 (-1.65) E/A... -25.65 (-2.49) DUAL... 6.28 (1.99) No. 388 387 R² 0.06 0.09 F 4.72 5.08 NB: CON = proportion of total shares held by majority stockholder; UTIL = public service utility companies; MEDIA = media companies; SIZE = total assets as a proxy for size; and STD = standard deviation of return on equity; t Statistic in brackets. 18

Appendix 2. Consequences of Concentration of Ownership Demsetz and Lehn s Models VARIABLE MODEL 1 (Profit Rate) MODEL 2 (Profit Rate) MODEL 3 (Profit Rate) Intercept 0.24 (6.2) 0.27 (11.7) 0.35 (4.2) UTIL -0.13 (3.4) -0.10 (2.4) -0.13 (3.1) FIN -0.07 (3.6) -0.06 (3.3) -0.07 (3.5) CAP 0.04 (0.7) 0.05 (0.8) 0.04 (0.7) ADV 0.42 (1.9) 0.47 (2.3) 0.42 (1.9) RD -0.11 (0.4) -0.07 (0.3) -0.11 (0.4) ASSET 5.70 (0.8) 8.14 (1.2) 5.97 (0.9) SE -0.29 (1.1) -0.43 (2.0) -0.29 (1.1) LA5-0.02...... (0.9) LA20... -0.004... (0.2) LAH...... -0.02 (0.9) Nº 511 511 511 R² 0.10 0.10 0.10 F 7.2 7.2 7.1 Nota: CAP = capital spending as a proportion of total sales; ADV = advertising spending as a proportion of total sales; RD = R&D spending as a proportion of total sales; and ASSET = total assets; t Statistic in brackets. Thomsen and Pedersen s Models MODEL ESTIMATES 1 ROE = 35.31-13.61UTIL + 29.08MEDIA - 1.24CAP + 0.22RD + 0.13SIZE - 0.14STD - 0.007COM (-4.07) (2.33) (-0.93) (0.12) (3.05) (-2.23) (0.19) R² = 0.09 F= 5.30 N = 388 2 ROE = 17.29-1.18RD + 0.05CON + 0.14RD*CON R² = 0.04 F= 5.69 N = 427 (-0.47) (0.85) (2.18) Industries with low R&D/ROE = 18.17-0.09COM Industries with high R&D/ROE = 17.29 + 0.05CON (-3.66) (0.59) 3 ROE = -5.21 + 0.69GROWTH + OWNER + GROWTH*OWNER + INDUSTRY + NATION (9.91) (9.03) (9.19) (3.23) (6.24) R² = 0.48 F = 4.15 No. = 427 Dispersed Ownership: ROE = -5.21 + 0.69GROWTH + INDUSTRY + NATION Dominante Ownership: ROE = 3.54-0.002GROWTH + INDUSTRY + NATION (-4.26)* Majority Ownership: ROE = 0.87 + 0.13GROWTH + INDUSTRY + NATION (-2.18)** *(t for the difference based on the equation for dispersed ownership) **(t for the difference based on the equation for dominant ownership) NB: ROE = return on equity (average 1990/93); CON = holding of majority stockholder in 1990; CAP = total assets/sales (average 1990/93); UTIL = public service utility companies, industry dummy 1= public service utilities; MEDIA = industry dummy 1= media sector; RD = R&D intensity, industry dummy 1 = R&D spending as a proportion of sales >1%; SIZE = total assets (average 1990/93); STD = standard deviation of return on equity (1990/93); E/A = net equity/total assets; and NATION = a category for each of the 12 countries considered; t Statistic in brackets. 19

Bibliographical References BUZZACCHI, L., COLOMBO, M. Business groups and the determinants of corporate ownership. Cambridge Journal of Economics, Vol. 20, pp. 31-51, 1996. BRAZIL COMPANY HANDBOOK. Rio de Janeiro: IMF Editora Ltda., 1994 and 1997. DEMSETZ, H. The structure of ownership and the theory of the firm. Journal of Law and Economics, Vol. 26, pp. 375-390, 1983. DEMSETZ, H., LEHN, K. The structure of corporate ownership: causes and consequences. Journal of Political Economy, Vol. 93, No. 6, pp. 1,155-1,177, 1985. HOLDERNESS, C. G., SHEEHAN, D. P. The role of majority shareholders in publicly heldcorporations. Journal of Financial Economics, Vol. 20, pp. 317-346, 1988. SHLEIFER, A., VISHY, R. W. Large shareholders and corporate control. Journal of Political Economy, Vol. 94, No. 3, pp. 461-488, 1986. THOMSEN, S., PEDERSEN, T. European ownership concentration: causes and consequences. Institute of International Economics and Management/Copenhagen Business School, 1997. WILLIAMSON, O. E. Corporate finance and corporate governance. Journal of Finance, Vol. 43, pp. 567-591, 1988. ZECKHOUSER, R. J., POUND, J. Are large shareholders effective monitors? An investigation of share ownership and corporate performance. In: HUBBARD, G. R. (ed.). Asymmetric information, corporate finance and investment. University of Chicago Press, 1990. 20