. Corporate Governance and Firm Value: The Case of Venezuela

Transcription

1 194 CORPORATE GOVERNANCE. Corporate Governance and Firm Value: The Case of Venezuela Urbi Garay and Maximiliano González* ABSTRACT Manuscript Type: Empirical Research Question/Issue: We examine the relationship between corporate governance and firm value, and evaluate the relatively understudied governance practices in Venezuela. Research Findings/Results: We construct a corporate governance index (CGI) for publicly-listed firms that is free of self-selection and self-reported bias and find that its mean value is below the emerging market average in general, and below the Latin American average in particular. This weak investor protection environment makes Venezuela a good setting to study how corporate governance practices affect firm value. We show that an increase of 1 per cent in the CGI results in an average increase of 11.3 per cent in dividend payouts, 9.9 per cent in price-to-book, and 2.7 per cent in Tobin s Q. These findings are robust after considering the potential endogeneity of our regression variables. Theoretical Implications: Results contrast to those reported in the US due to the higher interfirm variations in CGI. Our findings are consistent with the theoretical models that relate good corporate governance practices to higher investor confidence, and with the agency model of dividend payout. Furthermore, we conjecture that our results are generalizable mainly to other countries where investor protection is low. Practical Implications: Two direct insights to policy makers and practitioners follow from our analysis: first, managers in weak investor protection environments could differentiate their firms adopting corporate policies to improve their governance structure; and second, our measure of governance practices gives investors a quantitative tool to better assess Venezuelan firms. Keywords: Corporate governance rating/index, corporate performance, South America INTRODUCTION More companies in a growing number of countries are increasingly attempting to adopt better corporate governance practices. In the case of Latin America, the Andean Development Corporation (Corporación Andina de Fomento CAF) recently presented an outline for a corporate governance Andean Code (CAF, 2005). Furthermore, the larger companies of the region, especially those that belong to the financial sector, are in the process of adopting other international codes of best corporate governance practices, such as the Sarbanes-Oxley Act and the Principles of Corporate Governance developed by the Organization for Economic Co-operation and Development (OECD, 1999). It is not difficult to predict that the success or failure of these * Address for correspondence: Suite 11629, 6910 N.W. 50 Street, Miami, FL/ Tel: (ext. 3369); mgf@adm.uniandes.edu.co initiatives will depend on the real impact that they may have on the financial performance and market valuation of the companies that adopt them. La Porta, López-de-Silanes, Shleifer and Vishny (1997, 1998, 2000a) show that the legal framework that firms and investors face differs significantly around the world, in part, because of differences in legal origin. They argue that investors are less protected in French Civil Law countries, compared with countries from the Common Law origin. All countries in Latin America have the same legal origin, which is French Civil Law. They also find that Latin American countries perform even worse than the average French Civil Law countries in terms of investor rights, and argue that this helps explain the low level of financial development and the small size of stock exchanges of these countries. Chong and López-de-Silanes (2007) confirm these findings for a more recent period. Furthermore, according to Djankov, La Porta, López-de-Silanes and doi: /j x

2 CORPORATE GOVERNANCE AND FIRM VALUE 195 Shleifer (2008) Venezuela exhibits one of the worst scores in terms of investor protection. The weak investor protection inherent in many Latin American countries offers an opportunity for firms to differentiate themselves from the rest and to send strong and credible signals to attract investors by self-adopting good corporate governance practices and policies, thus partially compensating investors for the weak legal environment in which these firms operate. Klapper and Love (2004) and Durnev and Kim (2005) show that corporate governance provisions matter more in countries with weak legal protection. We know relatively little about the potential impact that the adoption of corporate governance practices may have on company value in Latin America (see Chong and López-de- Silanes, 2007, for a recent review of this evidence). Measuring this effect is important for the region because the success or failure of implementing good corporate governance practices may be greater if the market rewards those companies that adopt them. In the case of the US, the empirical evidence shows either no effect or an economically small effect. 1 Black (2001) argues that perhaps these weak results in the US arise because the variation in firm governance is small given that the minimum quality of corporate governance, which is set by law and by norms, is very high in that country. On the other hand, interfirm governance variation is found to be much larger in Venezuela. This should not come as a surprise, as a country with weaker laws and norms offers a wider range for governance differences between firms and, therefore, the potential for stronger results on the effects of governance on firm value. Furthermore, even though Venezuela is the fourth largest economy in Latin America (after Brazil, Mexico, and Argentina), relatively little is known about corporate governance practices in this country. In sum, Venezuela represents a very strong case study. We evaluate the current state of corporate governance practices in Venezuela by constructing a corporate governance index (CGI) for all firms listed in the Caracas Stock Exchange (CSE) as of the end of 2004 and comparing the results to other emerging and Latin American countries. We then evaluate whether firm dividend payout policies, price-to-book multiple, and Tobin s Q (TQ) are related to our CGI. By undertaking a single country-study approach, we attempt to perform a straightforward empirical test that has the advantage of avoiding some of the potential econometric problems involved in cross-country studies such as the omitted variable bias and the usually high across-firm heterogeneity. In general, we find a positive and strong relation between our index of corporate governance and the payout ratio, price-to-book multiple, and TQ for firms in Venezuela. From the composition of the index, we find that the subindexes on ethics and conflicts of interest, composition and performance of the board of directors, and shareholders rights explain much of the cross-sectional difference in payout ratio; on the other hand, the subindex regarding ethics and conflicts of interest can explain much of the results when price-to-book and TQ are used as dependent variables. These results add to the growing literature that supports the idea that in countries with relatively low investor protection, good corporate governance practices and policies could be used as an efficient mechanism for firms that want to distinguish themselves to attract investors. Although our results are tentative given the small size of the CSE, they passed a series of robustness checks that attempted to tackle, among other potential problems, the issue of endogeneity, a common concern found in this literature. Our paper is similar to Black (2001) and Judge, Naoumova and Koutzevol (2003) who tested the relation between corporate governance and firm value in Russia, a transition economy characterized by weak investor protection. Both papers have a small sample and Russia, like Venezuela, is also a country that scores low in terms of investor protection and exhibits a high interfirm variation in corporate governance practices. Our paper is also related to recent country studies done in Latin America 2 and especially with Garay and González (2005), who also studied the case of Venezuela. The evidence reported in this paper is important not only for Venezuela but also for other emerging markets in the process of attempting to improve their corporate governance practices. The evidence we show here adds to the growing literature worldwide that indicates that firms can differentiate themselves by adopting better corporate governance practices and policies. That is, even in a weak investor protection environment, firms can increase their market value by adopting good corporate governance measures. The rest of the paper is organized as follows: first, we review the growing literature on corporate governance and market valuation, concentrating on recent papers that are based on Latin America. Second, we construct a CGI for Venezuela and compare it with other emerging economies and, more importantly, to other Latin American countries. Third, we present the data and conduct our econometric analysis testing the relation between a firm s dividend payout ratio, price-to-book, and TQ, and our CGI. Fourth, we perform a number of robustness checks to our main findings. In the last section we present the conclusions and policy recommendations, as well as its potential practical applications and suggestions for future studies. LITERATURE REVIEW Many definitions of corporate governance stress the potential conflicts of interest between insiders (managers, boards of directors, and majority shareholders) and outsiders (minority shareholders and creditors) of the company. The set of internal and external mechanisms to balance these conflicts of interest is what it is usually known as corporate governance. The effect that a set of good corporate governance practices may have on firm s value is, however, an empirical question. Recently, different studies, trying to measure quantitatively the quality of corporate governance, have created indexes based on legal, accounting, and firm-level financial information. Gompers, Ishii and Metrick (2003) construct a CGI based on 24 governance rules for 1,500 large US firms, and show that firms with higher corporate governance scores had higher firm value. La Porta et al. (1997) study a sample of 49 countries and conclude that countries with legal systems based on Civil

3 196 CORPORATE GOVERNANCE Law, especially the French legal system, provide less protection to investors and have less developed capital markets, particularly when compared with countries from the Common Law origin. These authors also conclude that dividend policy constitutes an essential tool to reduce agency conflicts to minority investors. 3 These findings are consistent with the theoretical model presented in La Porta, López-de-Silanes, Shleifer and Vishny, (2002), where the positive effects of good corporate governance practices on firm valuation are explained by higher investor confidence. This situation lowers the cost of capital and, ultimately, increases firm value. Also, these results are consistent with the agency model of dividend payout in the corporate governance framework developed in La Porta, López-de-Silanes, Shleifer and Vishny (2000b). Since the seminal empirical papers of La Porta et al. (1997, 1998, 2000a) showing that laws that protect investors differ significantly across countries, in part because of differences in legal origin, the academic focus has shifted to study corporate governance in the international setting. 4 Klapper and Love (2004) was among the first and more comprehensive papers focusing on corporate governance in emerging markets. Using firm-level evidence on corporate governance practices for 495 companies from 25 emerging markets, they show that better corporate governance is highly correlated with better operating performance and market valuation. Many country-studies have used a methodology that is very similar to that of Klapper and Love (2004). For example, Black, Jang and Kim (2006a) constructed a CGI for South Korea; and Black (2001) and Black, Love and Rachinsky (2006b) both studied how their CGI affects firm value in Russia. The empirical evidence for Latin America has also grown rapidly in recent years. Leal and Carvalhal-da-Silva (2005) studied Brazil, Chong and López-de-Silanes (2006) studied Mexico, Lefort and Walker (2005) studied Chile, and Garay and González (2005) studied Venezuela. All these papers show that, on average, a good set of corporate governance practices and policies is positively related to firm value. These findings in Latin America are especially important because the weak investor protection inherent in this region offers an opportunity for firms to differentiate themselves to attract investors by self-adopting good corporate governance practices. Easterbrook and Fischer (1991) argue that firms themselves, when it is optimal to do so, could offer private contracts with better terms than can be offered by the rigid legal system. In the same manner, Diamond (1989, 1991) presents a theoretical discussion of the effects of a firm s reputation on its access to external financing, and Coffee (1999) argues for a global convergence in corporate governance that is independent of the local legal environment. Empirically, Klapper and Love (2004) and Durnev and Kim (2005) find that corporate governance practices play a more important role in countries where legal protection is weak. That is, firm-level improvements in corporate governance could, in some way, bypass the obstacles and inefficiencies of a country s legal system. That makes Venezuela a good setting to corroborate the effect good corporate governance practices have on firm valuation, given the overall low scores this country exhibits in terms of investors protection and the high interfirm variation in corporate governance practices observed in this country. This suggests the following hypothesis: Hypothesis 1: Better corporate governance practices will be positively related to firm valuation in Venezuela. This paper is similar to Garay and González (2005) because both papers use firm-level data for Venezuelan listed firms. However, the two papers differ in three important aspects. First, we present a more detailed analysis of each of the questions in our CGI and exclude all questions that are not directly applicable to the Venezuelan market. In contrast, Garay and González (2005) used a standard and more general questionnaire that was very similar to the one used by Klapper and Love (2004). Second, we answered the questions directly and therefore our paper is less likely to suffer from self-selection and self-reported bias. Third, here we have directly addressed the endogeneity issue, a typical concern in this type of empirical analysis. Moreover, the focus in Garay and González (2005) was not to test whether corporate governance affects market valuation but if financial performance somehow affects CEO turnover. Corporate Governance Index (CGI) Most studies on firm-level evidence on corporate governance practices gather their information using questionnaires filled by the companies themselves. This methodology presents various potential problems, among others: a low response rate, especially from those companies whose corporate governance practices are poor (self-selection bias); and, for the firms that do respond to the questionnaire, there is a tendency to present themselves not as they are at the moment when the questionnaire is being completed, but as they want to see themselves in the future (self-report bias). In our paper we follow a different route to construct our CGI. In the same spirit of Leal and Carvalhal-da-Silva (2005), we answer the questions ourselves using publicly available information. From Leal and Carvalhal-da-Silva (2005) s 24 questions we ended up with 17 questions that are applicable to the Venezuelan setting. 5 Each one of these 17 questions was answered using publicly available information. We then grouped the questions into four subindexes, namely: information disclosure (five questions), composition and performance of the board of directors (five questions), ethics and conflicts of interest (three questions), and shareholders rights (four questions). We report our results for each subindex in Table 1 for the 46 companies listed in the CSE in the year The disclosure subindex shows that only 19.6 per cent of the firms disclose penalties against management in case of deviating from the corporate governance policy; 82.6 per cent report their audited financial statements on time; only 17.4 per cent use international accounting standards; 84.8 per cent hire internationally recognized auditors; and 50 per cent disclose information on managerial compensation. The arithmetic mean for this subindex is 50.9 per cent. According to the composition and performance of the board of directors subindex, for 60.9 per cent of the firms in the sample, the chairman of the board is also the CEO or general manager; 56.5 per cent have monitoring committees;

4 CORPORATE GOVERNANCE AND FIRM VALUE 197 TABLE 1 Corporate Governance Index (CGI) These questions were answered by the authors for each of the 46 Venezuelan firms that were listed in the Caracas Stock Exchange (BVC) in 2004 to determine for each firm its CGI. The answer to each question is either Yes or No. If the answer is Yes, we add 1, and if the answer is No, we add 0. All answers are based on publicly available information. The primary sources of information are firms financial statements, bylaws, minutes of meetings, and annual reports available at the CNV. At the end of each question, there are remarks in italics on whether what is stated in the question is stipulated in the Venezuelan Code of Commerce. Arithmetic Affirmative N Questions mean answers SUBINDEX DISCLOSURE 50.9% 1 Does the company indicate in its charter, annual reports, or in any other manner, the 19.6% 9/46 penalties against the management in case of breach of its desired corporate governance practices? Required by Generally Accepted Auditing Standards. 2 Does the company present reports of its audited financial statements on time? Required 82.6% 38/46 by the CNV. 3 Does the company use international accounting standards? Required by Generally 17.4% 8/46 Accepted Auditing Standards. 4 Does the company use any recognized auditing firm? Required by the CNV and by 84.8% 39/46 Generally Accepted Auditing Standards. 5 Does the company disclose, in any form whatsoever, the compensation of the general 50.0% 23/46 manager and of the board of directors? Required by the CNV. SUBINDEX COMPOSITION AND PERFORMANCE OF THE BOARD OF DIRECTORS 54.4% 6 Are the chairman of the board of directors and the general manager two different 60.9% 28/46 people? Not required by any legal instrument. 7 Does the company have monitoring committees, such as appointment or compensation 56.5% 26/46 or auditing committees, or all of these? The auditing committee is established in the Venezuelan Code of Commerce. 8 Is the board of directors clearly comprised of external directors and possibly independent 32.6% 15/46 ones? Stipulated in the Code of Commerce, but not limited to the fact that they be independent. 9 Is the board of directors comprised of five to nine members, as per recommendation 73.9% 34/46 of good international corporate governance practices? Not required by any legal instrument or regulatory entity. 10 Is there a permanent auditing committee? Stipulated in the Code of Commerce. 47.8% 22/46 SUBINDEX EHTICS AND CONFLICTS OF INTEREST 39.9% 11 Is the company free of any penalty or fine for breach of good corporate governance 82.6% 38/46 practices or of any rules of the CNV during the last year? CNV rules. 12 Taking into account the agreements among shareholders, are the controlling 30.4% 14/46 shareholders owners of less than 50% of the voting shares? Not established in any legal instrument or by any regulatory entity. 13 Is the capital/voting rights ratio of controlling shareholders higher than 1? Not 6.5% 3/46 established in any legal instrument or by any regulatory entity. SUBINDEX SHAREHOLDERS RIGHTS 16.3% 14 Does the company charter or any other verifiable means facilitate the voting process of 28.3% 13/46 the shareholders beyond that established by law? Stipulated in the Code of Commerce. 15 Does the company charter guarantee additional voting rights to that established by law? 13.0% 6/46 Stipulated in the Code of Commerce. 16 Are there pyramidal structures that reduce concentration of control? Not established in 15.2% 7/46 any legal instrument or by any regulatory entity. 17 Are there agreements among shareholders that reduce concentration of control? Not 8.7% 4/46 established in any legal instrument or by any regulatory entity. AVERAGE CGI (equally weighting the four subindexes) 40.3% Source: Comisión Nacional de Valores (CNV), Código de Comercio, The questionnaire is adapted from Leal and Carvalhal-da-Silva (2005) to the Venezuelan setting.

5 198 CORPORATE GOVERNANCE 32.6 per cent have external directors 7 ; 73.9 per cent have a board composed of between 5 to 9 members; and 47.8 per cent have a permanent audit committee. The arithmetic mean for this subindex is 54.4 per cent. The ethics and conflicts of interest s subindex shows that 82.6 per cent of the companies are free from penalties or fines on the part of the regulatory agency (the Comisión Nacional de Valores); there exists a shareholder that controls less than 50 per cent of the firm s shares in 30.4 per cent of the firms in the sample; and in 6.5 per cent of the firms, the capital to voting rights ratio of majority shareholders is higher than 1. The arithmetic mean for this subindex is 39.9 per cent. Finally, the shareholders rights subindex shows that only 28.3 per cent of the firms in the sample facilitate the voting process beyond what is required by law; only 13.0 per cent have voting rights beyond that required by law; only 15.2 per cent do not exhibit a pyramidal structure that reduces the concentration of control 8 ; and 8.7 per cent report special agreements among shareholders that reduce the concentration of control. The arithmetic mean for this subindex is a very low 16.3 per cent. Taking together these averages, we can conclude that only around half of the firms in our sample comply with the requirements of the disclosure of the composition and performance of the board of directors and more work needs to be done in terms of ethics and conflicts of interest, and, especially, in terms of shareholders rights. At the firm level the highest overall CGI was 71.7 per cent and the lowest was 16.7 per cent. We found a much larger variation in Venezuelan firms corporate governance practices when compared with the US (results are not reported here). The average CGI in the sample is a low 40.3 per cent. In Table 2 Panel A we compare our CGI with the results reported by Klapper and Love (2004) who analyzed 495 firms in 25 emerging countries, 9 Lefort and Walker (2005) who studied 181 firms in Chile, and Leal and Carvalhal-da- Silva (2005) who studied 214 firms in Brazil. Table 2 shows that Venezuela is 14 percentage points below the emerging market average and 19 percentage points below Chile, which is the leading country in Latin America in terms of financial development and investor protection (Chong and López-de- Silanes, 2007). The Venezuelan average is closer to the one reported for Brazil. In Panel B we summarize the results obtained for each subindex and compare them with the results presented in Garay and González (2005) and in Lefort and Walker (2005) for Venezuela and Chile, respectively. Overall, the CGI we obtained produces a score 14 percentage points below the CGI reported by Garay and González (2005). As mentioned before, this difference could represent an overestimation on that paper due to the self-selection and self-reported bias generated when firms executives completed the questionnaires. Only in the composition and performance of the board of directors (Board) subindex do we find similar results. We also include in this panel the score reported by Lefort and Walker (2005) for Chile. The CGI for Chile is close to 20 percentage points higher than the CGI for Venezuela. Only in the subindex of ethics and conflicts of interest (Ethics) are the scores relatively close. Finally, in Table 2 Panel C we show the correlation matrix among the subindexes. As expected, all subindexes are positively and significantly related to the overall CGI. Chong and López-de-Silanes (2006) report a similar finding for Mexico, even though their corporate governance components are not exactly comparable to ours, and Leal and Carvalhal-da-Silva (2005) do not provide a correlation matrix for Brazil. On the other hand, each of our subindexes shows little correlation with the other subindexes (none of the correlation coefficients are statistically different from zero). Interestingly, each subindex seems to be taking into account a different dimension of the overall governance of the firm. Overall, these results confirm that Venezuela represents a good case study to test whether firms can somehow bypass a poor investor protection environment by voluntarily adopting good corporate governance practices. A relatively high CGI is an indicator that firms can use to attract investors. We want to verify whether investors in Venezuela recognize this signal by assigning a higher market valuation to such firms. DATA Having shown that Venezuela is a strong case study to test whether corporate governance is related to firm valuation and dividend payout, in this section we present the dependent, independent, and control variables used to formally test our hypothesis. Dependent Variables We use three alternative dependent variables to test our hypothesis. First, we use the dividend payout ratio (DPR), which is measured as the quotient between cash dividends and net earnings. La Porta et al. (2000b) show that firms in countries where investors are better protected exhibit higher dividend payouts than firms in countries where investors are poorly protected. On the other hand, Black et al. (2006a) and Leal and Carvalhal-da-Silva (2005) do not find support for this hypothesis in the cases of South Korea and Brazil, respectively. The second dependent variable is the price-to-book ratio (price-to-book value or PBV), measured as the quotient between per share market price and book value. The priceto-book is a valuation measure that has been used in corporate governance studies by authors such as Leal and Carvalhal-da-Silva (2005) for Brazil. Finally, we use the TQ as the third of our dependent variables. This variable was computed as the market value of the firm s assets (book value of assets - book value of equity + market value of equity) divided by the book value of assets. TQ can be considered the classic valuation measure and has been used extensively in the corporate governance literature (see, for instance, Morck, Shleifer and Vishny, 1988; La Porta et al., 2002; Gompers et al., 2003). Information regarding each one of these variables was obtained from the CSE Anuario (2004 yearbook) and corresponds to year-end values. Economatica s database was also used in some cases to confirm the validity of stock market prices data. Independent Variables As we mentioned in the previous section, the CGI was constructed based on 17 questions pertaining to different

6 CORPORATE GOVERNANCE AND FIRM VALUE 199 TABLE 2 Comparative Analysis In this table we compare our corporate governance index (CGI) to similar studies done in other emerging markets. Panel A presents basic statistics comparing 25 different emerging markets (Klapper and Love, 2004) together with the CGI calculated for Chile (Lefort and Walker, 2005) and Brazil (Leal and Carvalhal-da-Silva, 2005). Panel B divides the CGI into its four subindexes and compares the values with a similar study for Venezuela (Garay and González, 2005) and Chile (Lefort and Walker, 2005). Panel C shows the correlation matrix of each of the subindexes (p-values are reported below each correlation coefficient). Panel A: Comparative statistics for the Venezuelan CGI versus other emerging market studies Description This paper Klapper and Love (2004) Lefort and Walker (2005) Leal and Carvalhal-da-Silva (2005) Mean Median NR Standard deviation NR 8.33 Minimum NR Maximum NR Country Venezuela 25 EM Chile Brazil Observations Source: The above-mentioned papers. All numbers (except the number of observations) are expressed in percentages. EM = Emerging Markets; NR = not reported. Panel B: Comparative subindex for the Venezuelan CGI versus other studies in Venezuela and in Chile This paper (46 firms) Garay and González (2005) Lefort and Walker (2005) Subindex Questions Score (%) Questions Score (%) Questions Score (%) Ethics Board Shareholders Disclosure Overall CGI Panel C: Subindex correlation matrix CGI Disclosure Board Ethics Shareholders CGI 1 Disclosure Board of directors Ethics and conflicts of interest Shareholders rights corporate governance practices. We answered these questions for each of the 46 Venezuelan firms that were listed in the CSE in 2004 to determine for each firm its CGI. The answer to each question is either Yes or No. If the answer is Yes, we add 1 and if the answer is No, we add 0. All answers are based on publicly available information. These 17 questions were answered after reviewing each firm s financial statements, bylaws, minutes of the boards of directors and shareholders meetings, and annual reports available at the Comisión Nacional de Valores library.

7 200 CORPORATE GOVERNANCE We then grouped the questions in four subindexes: information disclosure (DIS, five questions), composition and performance of the board of directors (BOA, five questions), ethics and conflicts of interest (ETH, three questions), and shareholders rights (SHA, four questions). Control Variables We use the following three variables as controls: company size (CS), measured as the natural logarithm of the book value of assets, return on assets (ROA), measured as operating earnings (EBIT) divided by total assets, and leverage (LEV), measured as the quotient between total debt and total assets. Information regarding each one of these variables was obtained from the CSE Anuario (2004 yearbook) and corresponds to year-end values. ECONOMETRIC ANALYSIS In order to perform the statistical tests and the multivariate regressions, a preliminary analysis of the information available was carried out following a procedure similar to that used by Black et al. (2006a), in order to exclude from the sample all those companies with missing information or whose standardized errors exceeded the standard deviations in +/ in each of the variables used. Also, companies without any market transaction during the year were deleted from the sample. From this analysis a total of 13 companies were excluded, therefore the sample was reduced to 33 companies. 10 Of these 33 companies, 12 were banks, 2 were bank-related financial institutions, and 19 were firms that belonged to the industrial and service sectors. In Table 3 we report the descriptive statistics for the variables used in the analysis that follows. Panel A includes statistics for the complete sample of 33 firms, while panel B shows the descriptive statistics for the reduced sample of 19 nonfinancial institutions. In our complete sample the average firm pays 20 per cent of its net income in dividends, has a price-to-book multiple equal to 1.07 and a TQ equal to When we restrict our sample to only nonfinancial firms, the mean values of these three variables decline slightly to 16 per cent, 0.85 per cent, and 0.90 per cent, respectively. In terms of our CGI, the reduced sample of 33 firms shows an average value equal to 8.30 over a maximum of 17 points (one point for each question answered as yes ), or 49 per cent in percentage terms. 11 For the nonfinancial sample, the CGI declines to 7.95; this is consistent with the fact that financial institutions are more regulated and are subject to more scrutiny in terms of information disclosure and other legal requirements. The nonfinancial sample tends to be more profitable in terms of ROA. The firms included in the complete sample tend to be larger and, as expected, more leveraged. In Table 4 we report a pair-wise correlation matrix for the variables used in this study. It shows that the CGI is positively correlated to the three alternative dependent variables previously defined (dividend payout, price-to-book, and TQ), not only in the complete sample (Panel A), but also in the sample restricted to nonfinancial institutions (Panel B). In the first case, the correlation is always statistically significant, and in the second case, it is statistically significant for the dividend payout. With respect to the firm s size (CS), consistent with the findings obtained by Leal and Carvalhal-da-Silva (2005), we obtain a positive and significant coefficient for both samples. That is, larger firms tend to exhibit better corporate governance practices. In Panel C we also report the correlation coefficients between the performance measures and each of the corporate governance subindexes. The Board, the Ethics and Conflicts of Interest, and the Shareholders rights subindexes are positive and significantly related to dividend payout. For the other two performance measures, only the Ethics and Conflicts of Interest subindex shows a significant correlation coefficient. The Disclosure subindex did not show significant coefficients in any of the three performance measures. Dividend Payout Ratio In Table 5 we show the results of Ordinary Least Squares (OLS) regressions for the dividend payout ratio on the CGI, and the control variables for Model 1 includes the CGI as the sole explanatory variable. 12 Here, an increase of one point in the CGI causes an increase of 11.3 per cent in the dividend payout ratio. This result is statistically significant (t = 3.69, p <.01) and almost triples the 4.32 per cent increase found by Garay and González (2005), also for Venezuela. However, this result differs from that of Leal and Carvalhalda-Silva (2005) for Brazil, who did not find a significant relation between these parameters. Models 2, 3, and 4 include one control variable in the estimation. In each case the positive sign and the statistical significance of CGI is preserved (t = 3.50, p <.01; t = 2.93, p <.01; and t = 3.99, p <.01, respectively). Model 5, which includes all control variables considered together, also shows a positive and statistically significant sign for CGI (t = 3.69, p <.05). 13 Table 5 also shows that results are maintained and that the economic impact is stronger when financial institutions are excluded from the sample (Panel B). Also, in both samples we reject the hypothesis that dividend payout and CGI are independent variables using a nonparametric test (Spearman). Finally, we regressed dividend payout for the whole sample with each of the subindexes (Panel C) and also considering the four subindexes together (we show in Table 2 Panel C that there was very little correlation among the subindexes). These results confirm that three of the subindexes (board of directors, ethics and conflicts of interest, and shareholders rights) independently affect, in a positive and statistically significant way, the firm s dividend payout (t = 1.91, p <.05; t = 2.19, p <.05 and t = 2.33, p <.05, respectively). On the other hand, the disclosure subindex does not affect in any significant way the dividend payout. These results serve as a preliminary evidence to conclude that the factor driving CGI are the subindexes on board of directors, ethics and conflicts of interest, and shareholders rights. Price-to-Book Value In Panel A of Table 6 we present our OLS estimation of price-to-book on the CGI, and the control variables for 2004.

8 CORPORATE GOVERNANCE AND FIRM VALUE 201 TABLE 3 Descriptive Statistics The variables are identified as follows: dividend payout ratio (DPR), price-to-book value (PBV), Tobin s Q (TQ), corporate governance index (CGI), return on assets (ROA), company size (CS), and leverage (LEV). The total sample is composed of 33 companies (Panel A) and it is reduced to 19 companies (Panel B) when financial firms are excluded. All numbers represent 2004 values. Panel A: Complete sample Variable Observation Mean Standard deviation Minimum Maximum Dependent variables DPR PBV TQ Independent variable CGI CGI Control variables ROA CS LEV Panel B: Sample excluding financial institutions Variable Observation Mean Standard deviation Minimum Maximum Dependent variables DPR PBV TQ Independent variable CGI CGI Control variables ROA CS LEV Model 1 includes the CGI as the sole explanatory variable. An increase of one point in the CGI causes an average increase of 9.9 per cent in the PBV. This result is statistically significant (t = 2.37, p <.05), and its magnitude more than doubles the 4.2 per cent increase calculated by Garay and González (2005) also for Venezuela. Leal and Carvalhal-da-Silva (2005) did not find a significant relationship between these two variables in Brazil. Models 2, 3, and 4 include one control variable. In each of these models, CGI maintains the sign and its statistical significance in Model 2 (t = 2.17, p <.05) and Model 4 (t = 2.19, p <.05), and it is marginally statistically significant in Model 3 (t = 1.70, p <.10). Model 5, which includes all control variables together, also shows a statistically significant sign for CGI (t = 2.08, p <.05). Using a nonparametric test (Spearman) we reject, at the 5 per cent confidence level, the hypothesis that PBV and CGI are independent. Results using the reduced sample of nonfinancial institutions were not included because the regression, as a whole, was not statistically significant in any of the models. The small size of the CSE may help explain the lack of statistical power. We also regressed PBV on each of the subindexes and considering the four subindexes together (results available from the authors). Although all subindexes were positive, only ethics and conflicts of interest was statistically significant. In the case of Brazil, although Leal and Carvalhal-da-Silva (2005) also found that the coefficients of each of the subindexes were positive, these authors also found that none of the components of the CGI was statistically significant explaining PBV. Similar results were obtained by Chong and López-de-Silanes (2006) for the case of Mexico, although we must advise that their CGI components classification is different (more disaggregated) than

9 202 CORPORATE GOVERNANCE TABLE 4 Correlation Matrix This table presents the pair-wise correlation matrix for the variables used in this study. The variables are identified as follows: Dividend payout ratio (DPR), price-to-book value (PBV), Tobin s Q (TQ), corporate governance index (CGI), return on assets (ROA), return on equity (ROE), company size (CS), and leverage (LEV). Panel A presents the results using all firms in the sample; Panel B presents the results excluding financial firms; and Panel C presents the correlation coefficients of each performance measure with each of the subindexes. p-values are reported below each correlation coefficient. Panel A: Complete sample DPR PBV TQ CGI ROA ROE CS LEV DPR 1 PBV TQ CGI ROA ROE CS LEV Panel B: Sample excluding financial institutions DPR PBV TQ CGI ROA ROE CS LEV DPR 1 PBV TQ CGI ROA ROE CS LEV Panel C: Performance variables correlation with subindexes DPR PBV TQ Disclosure Board of directors Ethics and conflicts of interest Shareholders rights

10 CORPORATE GOVERNANCE AND FIRM VALUE 203 TABLE 5 Dividend Payout Regressions Ordinary Least Squares regressions results of the dividends payout ratio (DPR) on the corporate governance index (CGI) for five different model specifications. The variables are identified as follows: return on assets (ROA), company size (CS), leverage (LEV), disclosure subindex (DIS), board of directors subindex (BOA), ethics and conflicts of interest subindex (ETH), and shareholders rights subindex (SHA). Panel A presents the results using all firms in the sample; Panel B presents the results excluding financial firms; and Panel C presents the results using as independent variable each subindex coefficients. p <.10, *p <.05, **p <.01. The t-values are reported below each estimated coefficient. Panel A: Complete sample Variables Model 1 Model 2 Model 3 Model 4 Model 5 CGI 0.11** 0.11** 0.11** 0.12** 0.09* (3.69) (3.50) (2.93) (3.99) (2.58) ROA (-0.91) (-.91) CS (0.24) (1.12) LEV * (-1.69) (-2.25) Cons ** -0.66* * (-2.82) (2.38) (-1.17) (2.24) (-1.56) Obs F 13.60** 7.18** 6.62** 8.61** 5.24** R Adj. R Spearman s rho 0.58** Panel B: Sample excluding financial institutions Variables Model 1 Model 2 Model 3 Model 4 Model 5 CGI 0.18** 0.18** 0.18** 0.18** 0.18* (4.51) (4.35) (3.64) (4.24) (2.91) ROA (-0.68) (-0.58) CS (0.01) (0.02) LEV (-0.50) (-0.42) Cons ** -1.22** ** (-3.92) (-3.48) (-1.32) (-3.03) (-0.92) Obs F 20.38** 10.11** 9.59** 9.87** 4.55* R Adj. R Spearman s rho 0.56** Panel C: Subindexes Variables Model 1 Model 2 Model 3 Model 4 Model 5 DIS (0.07) (0.18) BOA 0.10* 0.11* (1.91) (2.31) ETH 0.23* 0.18 (2.19) (1.92) SHA 0.15* 0.16** (2.33) (2.82) Cons (0.71) (-0.51) (-0.45) (0.97) (-1.61) Obs F * 4.78* 5.45* 4.51** R Adj. R Spearman s rho **

11 204 CORPORATE GOVERNANCE TABLE 6 Price-to-Book Value (PBV) and Tobin s Q (TQ) Regressions Ordinary Least Squares regressions results of PBV and TQ on the corporate governance index (CGI) for five model specifications. The variables are identified as follows: return on assets (ROA), company size (CS), and leverage (LEV). p <.10, *p <.05, **p <.01. The t-values are reported in the parenthesis behind each estimated coefficient. Panel A: Dependent variable PBV Model 1 Model 2 Model 3 Model 4 Model 5 CGI 0.10* (2.37) 0.09* (2.17) 0.09 (1.7) 0.09* (2.19) 0.11* (2.08) ROA (-1.27) (-1.17) CS 0.02 (0.50) (-0.71) LEV 0.37 (1.35) 0.44 (1.30) Cons (0.65) 0.39 (1.03) (-0.25) 0.08 (0.21) 1.12 (0.84) Obs F 5.59* 3.66* * 2.23 R Adj. R Spearman s rho 0.36* Panel B: Dependent variable TQ Model 1 Model 2 Model 3 Model 4 Model 5 CGI 0.03 (1.85) 0.03 (1.67) 0.03 (1.53) 0.03 (1.68) 0.04* (2.00) ROA (-1.12) (-1.25) CS 0.00 (0.02) (-1.26) LEV 0.12 (1.29) 0.19 (1.59) Cons. 0.73** (5.84) 0.77** (5.92) 0.72 (1.93) 0.68** (5.21) 1.27** (2.79) Obs F R Adj. R Spearman s rho 0.33 ours, making comparisons between the two works more difficult to interpret. Tobin s Q Ratio Panel B of Table 6 shows the results of the TQ regressions on the CGI and the control variables. In Model 1 we report, using the CGI as the sole explanatory variable, that an increase of one point on the CGI causes an average increase of 2.7 per cent in the TQ. This result is marginally statistically significant (t = 1.85, p <.10), being consistent with the results obtained by Garay and González (2005) for Venezuela and by Leal and Carvalhal-da-Silva (2005) for Brazil, who found an increase in TQ of 2.24 per cent and 3.1 per cent, respectively. Klapper and Love (2004) also found a positive and significant relation between the TQ and the CGI for their sample of emerging market firms. Models 2, 3, and 4 include one control variable in the estimation and Model 5 includes all the control variables considered together. The CGI maintains the sign in each of the models but the coefficient is statistically significant only in Model 5 (t = 2.00, p <.05). Also, using a nonparametric test (Spearman), we marginally reject the hypothesis that the TQ and CGI are independent at the 10 per cent confidence level. We do not show the results of the sample excluding the financial institutions because they were not statistically significant in any of the models. Once again, the small size of the CSE may help explain the lack of statistical power. We also regressed the TQ to each of the subindexes and also considering the four subindexes together (results available from the authors). Although all subindexes were posi-

12 CORPORATE GOVERNANCE AND FIRM VALUE 205 tive, and as it was the case when PBV was used as the dependent variable, only the ethics and conflicts of interest subindex was statistically significant. There exists a vast literature on the potential effects of conflicts of interest between controlling and outside shareholders on firm value and profitability. For instance, Morck et al. (1988) find, for a sample of US firms, that profitability first rises as ownership concentration increases (a finding that is consistent with the incentive hypothesis), and then falls after a certain point. These authors contend that this fall in profitability is due to an excessive voting power concentration or entrenchment, which leads to a fall in corporate value as the likelihood of expropriation increases. Lins (2003) analyzes 18 emerging markets and finds that the evidence in favor of entrenchment is stronger than it is for incentives. This author also argues that, in countries were legal protection is weak, the existence of large nonmanagerial block holders helps mitigate the potentially negative effect of control concentration on firm value. In the case of Brazil, although Leal and Carvalhal-da-Silva (2005) also found that the coefficients of each of the subindexes were positive, they reported that only the disclosure component of the CGI was statistically significant explaining TQ. Similar results were obtained by Chong and López-de- Silanes (2006) for Mexico. Overall, we find a positive and significant relation between dividend payout ratio and firm valuation (PBV and TQ) and our CGI. Firms with a better CGI tend to pay more dividends and are more valuable for investors in terms of their price-to-book multiple and their TQ ratio. Results also suggest that in a weak investor protection environment such as Venezuela s, firms are able to send strong signals to the market by voluntarily improving their corporate governance practices, something that allows them to differentiate from the rest. ROBUSTNESS CHECKS In this section we perform several robustness checks to validate our previous results. Huber/White/Sandwich Estimator of Variance The first robustness check we perform consists in estimating once again all the regression coefficients but this time using the Huber/White/Sandwich estimator of variance. This procedure generates larger standard errors and, therefore, the estimated t-values are much smaller than those obtained by the traditional OLS procedure. The signs and the statistical significance of the main results remain (results are not shown but are available upon request). Endogeneity Many empirical studies on corporate governance are subject to criticism, given the likely endogeneity of the CGI that is present in this type of studies. 14 For example, firms that need to finance their growth could be tempted to improve their corporate governance practices in order to reduce their cost of capital. This expected growth should therefore be priced in the firm s valuation, creating a positive correlation between the CGI and value ratios such as TQ. If this occurs then valuable firms will choose to adopt better governance practices and not the other way around. In the previous section we attempted to mitigate this problem by adding several control variables that could proxy for growth opportunities such as size, operating performance, and leverage to our regression model and showed that our results were not spuriously caused (at least not by these omitted variables). However, including control variables in the regressions is not enough to dissipate the likely endogeneity between a firm s value and its CGI. The first step we followed to tackle the possible existence of endogeneity in our model consisted in estimating all the regressions but this time using financial information corresponding to the year In this approach, we regressed the three alternative dependent variables (dividend payout, price-to-book, and TQ) for 2005 against the 2004 s CGI. Signs for all coefficients were preserved but statistical significance was lost, something that can be understandable given the small sample size. 15 The second step was to construct the CGI for the year 2006, even though we were only able to compute a preliminary index for that year given that not all the necessary information was either still available or was final. Following the approach of Chong and López-de-Silanes (2006), we averaged the CGI for 2004 and 2006 and run each model again. In this new set of results we obtained smaller coefficients. For example, when estimating dividend payout we obtained a statistically significant lower coefficient, (t = 2.50, p <.05), versus the original shown in Table 5; when estimating PBV we obtained a marginally statistically significant lower coefficient, (t = 1.85, p <.10), versus the original shown in Table 6; and, when estimating TQ, we obtained also a marginally statistically significant lower coefficient, (t = 1.78, p <.10), versus the original shown also in Table 6. Although the coefficients obtained under this approach were somehow smaller, they were all positive and kept their statistical significance. Although this set of results somehow reduced our endogeneity concerns, they were not able to eliminate them given the low statistical power of the model. However, the fact that both the positive signs of our CGI were preserved and that similar statistical significance was verified when we used the average CGI provides a base to be optimistic regarding the direction of the causality of our test. The next step we followed to tackle the endogeneity problem was to find instruments or a group of exogenous variables which are related to CGI but that are not necessarily related to any of the three alternative dependent variables. More specifically, we used the following three measures: first, a dummy variable called ADRUSA that takes the value of 1 if the company had American Depositary Receipts (ADRs) outstanding in the period and 0 otherwise. The second measure is a variable called CHAIND, which is the percentage change in board independence from the year 2000 to the year We proxy board independence, following Garay and González (2005), as the difference between the fraction of outside directors minus the fraction of inside directors in the board. A director is