The Relation between corporate governance and market value : mitigating endogeneity Problems

This paper investigates whether the adoption of good corporate governance practices influences the market value of Brazilian firms. For this purpose, we used an unbalanced panel over the period from 2002 to 2010, composed of 233 listed nonfinancial companies, for a total of 1,110 observations. Because of endogeneity problems such as omitted variables, simultaneity and the feedback effect, we employed dynamic multiple linear regression models, estimated by the system generalized method of moments (System GMM). The results indicate that firms listed in one of the three premium corporate governance segments of the BM&FBovespa (Level 1, Level 2 and Novo Mercado) are priced higher by the market compared to firms listed in the traditional trading segment. Therefore, corporate governance has a positive effect on the market value of Brazilian firms.


INTRODUCTION
ver the past three decades, the theme of corporate governance has gained increasing importance among academics and business leaders.Various factors have contributed to this trend.Becht, Bolton & Röell (2002) point out, among other phenomena, the corporate scandals in the United States between 2001 and 2003, which revealed the manipulation of financial statements, tax evasion and use of inside information by executives of large American companies.
More recently, the global financial crisis of 2008, triggered by problems in the American subprime mortgage market and excessively leveraged banks, corroborated the importance of improving corporate governance.According to Silveira (2010), besides excessively low interest rates and overly complex financial products, executive compensation systems linked to unsustainable results over the long run and failures of boards of directors also contributed to the crisis.
In Brazil, the liquidation by the Central Bank in the 1990s of large privately owned banks whose insolvency had been masked by fraudulent accounting heightened concern with improving corporate governance standards in the country (CARVALHO, 2002).This preoccupation was an important factor leading to the creation of three premium listing segments in terms of governance requirements by the São Paulo Stock Exchange (BOVESPA 1 ) in December 2000: Level I, Level 2 and Novo Mercado (New Market).Firms listed in these trading segments must have more transparent disclosure policies and provide greater protection for minority shareholders (BM&FBOVESPA, 2012).
Our main objective here is to investigate the relation between the adoption of better corporate governance practices and the market value of firms.Due to the presence of endogeneity problems, we employed dynamic multiple linear regression models with an unbalanced panel of firms listed on the BM&FBovespa between 2002 and 2010, for a total of 1,110 firm-year observations.We tested six specifications with the intention of answering the following question: Do Brazilian companies listed in premium corporate governance segments (L1, L2 and NM) have higher market value compared to firms listed in the traditional market segment?

THEORETICAL FRAMEWORK
Although there is no exact definition of corporate governance, it can be understood as the set of incentive and control mechanisms intended to minimize the agency costs arising from the conflict of interests between the suppliers of resources and the managers of companies (SILVEIRA, 2010).Agency theory is based on assumptions such as the inexistence of complete contracts and the opportunistic behavior of economic agents (JENSEN;MECKLING, 1976).These can result, among others, in certain practices by managers (or controlling shareholders) placing their own benefit over those of the shareholders (or the minority shareholders).Among these are decisions by managers to grant (SILVEIRA, 2010).Among these are omitted variables, simultaneity (or reverse causality) and the feedback effect, three possible sources of endogeneity (BÖRSCH-SUPAN; KÖKE, 2002;BARROS et al., 2010).
According to Silveira (2010), the omission of variables can result in spurious correlation between the variables of interest, because a relevant variable not included in the theoretical model can simultaneously influence the governance variables and the performance variables chosen.Two variables can be mutually correlated when neither is caused by the other, both of them instead being explained by a third variable.The use of control variables and fixed effects and random effects procedures, such as in the works of Bai et al. (2004) and Carvalhal-da-Silva & Leal (2005), can attenuate this problem.BARROS et al., 2010).The adoption of better governance practices can reduce the firm's cost of capital or increase its market value (SILVEIRA, 2010).However, the better performance of a firm can make more resources available for investment in its governance system, which in turn increases its efficiency.Silveira et al. (2009) built a panel composed of 823 observations of Brazilian firms between 1998 and 2004.When they applied regressions by ordinary least squares, Tobin's Q had a positive influence on the quality of the firms' governance practices.
The problem of reverse causality can be mitigated by using instrumental variables, as in the studies of Black, Jang & Kim (2006) and Silveira, Barros & Famá (2006).However, it is hard to find a set of valid instruments because while the first assumption, of a significant correlation between the instruments and the endogenous regressor, is easily verified, the second, of no correlation between the instruments and the model's error term, is not easy to check, since the error is not directly observable (BARROS et al., 2010).Black, Jang & Kim (2006) formulated an indicator to approximate the quality of governance practices of 525 South Korean companies, among other aspects covering stockholders' rights, the structure of the board of directors and the transparency in disclosing information.By employing regressions calculated by ordinary least squares (OLS), two-stage least squares (2SLS) and three-stage least squares (3SLS), the authors found evidence that corporate governance is able to explain the market value of the firms studied.
Silveira, Barros & Famá (2006) also applied multiple regressions estimated by OLS, 2SLS and 3SLS, to a sample of 154 Brazilian companies in 2000.As a proxy for the quality of governance practices they used an index calculated from a set of 20 objective binary questions.Their results indicated that governance quality has a positive influence on Tobin's Q and the price-to-book-value ratio and that rising market value can lead to the adoption of better governance standards by organizations (i.e., two-way causality between governance quality and valuation).
Finally, the feedback effect of the response variables to the regressors occurs when the past values of the dependent variable influence the present and/or future values of the independent variables (BARROS et al., 2010).According to Wintoki, Linck & Netter (2008), this problem is also known as "dynamic endogeneity" in corporate finance studies and can be mitigated by the inclusion of lags of the explained variable in the regression model, as done in the generalized method of moments (GMM).Wintoki, Linck & Netter (2008), in using fixed-effects analysis of a panel of 6,000 American companies, found a significant relationship between the structure of the board of directors and corporate performance.However, when they employed a system generalized method of moments (System GMM), size and level of board independence did not influence the return on sales or Tobin's Q.In turn, Ammann, Oesch & Schmid (2011) analyzed 2,300 companies in 22 developed countries over the period from 2003 to 2007.They found a positive influence of governance level on market value (Tobin's Q) when testing models estimated by fixed effects and GMM.

METHODS
The main objective of this article is to investigate the relation between the adoption of enhanced corporate governance practices and the market value of Brazilian companies.More specifically, we formulated the following question: Do Brazilian companies listed in premium corporate governance segments (L1, L2 and NM) have higher market value compared to firms listed in the traditional market segment?
For this purpose, we used a sample composed of nonfinancial companies with shares traded on the São Paulo Stock, Mercantile and Futures Exchange (BM&FBOVESPA) during the period from 2002 to 2010.The exclusion of financial institutions was due to the different accounting standards applicable to these firms, making it hard to compare the profitability and valuation metrics against those of companies in other sectors.We also excluded firms that reported negative equity in any year, since this would impair calculation of some indicators, like ROE and ROIC.
The final sample was composed of 233 firms for which all the data were available to calculate the variables in at least two years in the period analyzed.Therefore, we formed an unbalanced panel of 1,110 firm-year observations, although in the regression models the number of observations declined to 788 because of the estimation by the system generalized method of moments (System GMM).The variables were calculated from secondary data obtained from the Economática ® database.The accounting values refer to the financial statements for the fourth quarter of each year, and the stock quotations are the average annual price, obtained by averaging the closing price on the last trading day of each quarter.Since  The first, Tobin's Q, is hard to operationalize due to the inability to directly observe the market value of debts and replacement value of assets.Therefore, we chose the approximation proposed by Chung & Pruitt (1994: p. 72), according to which Tobin's Q can be calculated by Equation ( 1): QTOBIN = (1) Where: MVC: market value of the firm's common shares; MVP: market value of the firm's preferred shares; DEBT: book value of total debt (short-and long-term) minus current assets after exclusion of inventories.
The second proxy used was the enterprise value over total assets, obtained by Equation (2): EV/TA = (2) (*) Enterprise value = market value of total shares + short-and long-term bonds + short-and long-term loans + advances on foreign exchange contractscash and cash equivalents.

CONTROL VARIABLES
We also used the following control variables:  Profitability (operational performance).We used three profitability measures: (i) return on invested capital (ROIC); (ii) return on equity (ROE); and earnings before interest, taxes, depreciation and amortization over total assets (EBITDA/TA), calculated, respectively, by Equations ( 3), ( 4) and ( 5): EBITDA/TA = ( )×100% (5)  Leverage (LEV): measured by the sum of current and long-term liabilities divided by total assets at the end of the year.
 Volatility (VOLAT): measured by the monthly returns of the most liquid type of stock of each firm in each year, obtained by Equation ( 6): Where: : continuous monthly return of stock i; : average of the continuous monthly returns of stock i; and n = 12 (one year).
 Payout index (PAYOUT): defined as dividends paid per share over net earnings per share at the end of the year.
 Liquidity index (LIQ): Calculated by the Economática ® system for periods of 12 months, based on the trading volume of each stock, obtained by Equation ( 7): Where: p: number of days when there was at least one trade of the stock in the period analyzed; P: total number of days in the period analyzed; n: number of trades of the stock in the period analyzed; N: number of trades of all stocks in the period analyzed; v: monetary value of the trading of the stock in the period analyzed; V: monetary value of the trading of all stocks in the period analyzed.
 Capital expenditure rate (CAPEX): defined as capital expenditures over gross property, plant and equipment at the end of the year.
We also used annual dummy variables (YEAR), with value of one for the specific year and zero for other years, with the aim isolating the macroeconomic effects that affected the companies during the study period.They are coded in chronological order: D02 = 2002, D03 = 2003 and so on.

MODELS AND STATISTICAL TREATMENT
To ascertain whether the adoption of enhanced corporate governance practices influences the firms' market value, we employed dynamic multiple linear regression models, estimated by the system generalized method of moments (System GMM).For this purpose we used the Stata ® 11 statistical package, applying the xtabond2 tool of Roodman (2006Roodman ( , 2009)).
As a proxy for market value (MV), the dependent variable, we used Tobin's Q (QTOBIN) and enterprise value over total assets (EV/TA).Equation ( 8 In the formula, α is the intercept, which can be interpreted as the average specific effect of the set of firms, i and t represent, respectively, the firm and year, and + represents the decomposition of the random error term ( = + ).More specifically, is the error term of firm i in year t and is the unobserved heterogeneity of the firms in the sample, to capture any unobserved characteristics of firm i that do not vary over time.
Model (8) addresses the possible sources of endogeneity that can bias the parameters obtained by the regression technique.The control variables and the term have the objective of attenuating the omission of variables.The use of the first lag of the response variable  (2006,2009), with employment of the laglimits (3 3) function.

RESULTS
Tables 1 and 2 report the descriptive statistics of the variables.Regarding the participation of the companies in the premium corporate governance segments (CGQ), note that various firms migrated to or went public by listing directly in Level 1, Level 2 or Novo Mercado just after their creation by the Bovespa in December 2000, as reported by Gorga (2008).However, the number of observations of firms with shares listed in the traditional segment (654) represents more than half of the total (1,110).Besides this, the group of companies listed in one of the special governance segments presented higher means and medians of Tobin's Q and for enterprise value over total assets than the other set of firms.Regarding the total sample, the median value below 1 of Tobin's Q (as well as EV/TA) shows that the majority of observations refer to firms that if sold at that time for the price attributed by the market, would not have received enough to cover the replacement value of their assets.

MARKET VALUE AND ADOPTION OF ENHANCED GOVERNANCE PRACTICES
Table 3 shows the results of the dynamic multiple linear regression models, estimated by the System GMM, to explain the market value (Tobin's Q or enterprise value over total assets).In the six equations tested, the variables QTOBIN and EV/TA were positively influenced by their lagged values (L.QTOBIN and L.(EV/TA)), at 1% significance.These results indicate inertial behavior of the firms' market value.
With respect to the independent variable of interest, CGQ obtained positive and significant coefficients at 1% in five of the six specifications employed.These results suggest that the adoption of enhanced corporate governance practices has a positive effect on market value.More specifically, these results indicate that on average the firms listed in one of the premium governance segments (L1, L2 or NM) are more valued by the market than are firms listed in the traditional trading segment.
Hence, it can be inferred that the higher market value of the firms listed in the L1, L2 and NM segments results from their adherence to higher corporate governance standards, mainly in relation to protection of minority shareholders and transparency in the disclosure of information, meaning less pronounced information asymmetry, increasing demand for their shares.This hypothesis is consistent with the results of Carvalho & Pennacchi (2012), when analyzing the impact of the migration of 38 Brazilian firms to the premium segments between 2001 and 2006.The authors used the event study method and found a positive cumulative Bai et al. (2004) used control variables such as size, operating profit and indebtedness in their regression models.For a panel with 2,905 observations of Chinese corporations in the period from 1999 to 2001 they observed that firms are less valued by the market when: a) the government is the largest shareholder; and b) the CEO is also the chairman of the board.Carvalhal-da-Silva & Leal (2005) used as a proxy for the quality of corporate governance an index composed of 15 objective binary questions regarding the ownership structure, board composition, transparency of information disclosed and shareholders' rights.


L1: Maintenance of a free-float of at least 25% of capital; holding of public offerings to place shares through mechanisms that favor capital dispersion to a broader range shareholders; improved disclosure of quarterly information, including the obligation to report consolidated figures and special audit review; adherence to the disclosure rules for transactions involving assets issued by the company on the part of the controlling shareholders or company management; and disclosure of shareholder agreements and stock option programs.L2: To be classified at Level 2, in addition to the obligations of Level 1, the firm and its controlling shareholders must adopt a broader range of corporate governance practices and minority shareholder rights: a term of two years or less for the entire board of directors, without staggered elections; disclosure of the annual balance sheet in accordance with US GAAP or IAS; granting to all holders of common shares the same price obtained by the controlling shareholders on the transfer of control of the firm and 80% (100% as of 2011) of this price for preferred shareholders (tag along); voting rights granted to preferred shares in certain circumstances such as transformation, consolidation, spin-off or merger of the firm and approval of contracts between the firm and other companies of the same group; obligation to hold a tender offer by the economic value criterion should the firm be taken private or Level 2 registration be cancelled; and requirement to resolve corporate disputes by arbitration under the auspices of the Market Arbitration Chamber of the BM&FBovespa.BBR, Braz.Bus.Rev. (Engl.ed., Online),Vitória, v. 11, n. 1, Art. 5, p. 90 -110, jan.-feb.2014   www.bbronline.com.br NM: Participation in the Novo Mercado requires satisfying all the criteria of Level 2 plus the requirement only to issue common shares, so that all shareholders are assured of the right to vote.As suggested by Börsch-Supan & Köke (2002), we used two proxies for market value.
). 1 In 2006 the Bovespa merged with the Mercantile and Futures Exchange (BM&F) to form the BM&FBovespa.O BBR, Braz.Bus.Rev. (Engl.ed., Online), Vitória, v. 11, n. 1, Art. 5, p. 90 -110, jan.-feb.2014 www.bbronline.com.brA and b) how these mechanisms are related (SILVEIRA; BARROS; FAMÁ, 2006).Moreover, a significant portion of the articles have considered corporate governance with an exogenous variable, i.e., one not determined by its mechanisms or other attributes of the firm.For this reason, authors have not attempted to control for the possible sources of endogeneity, such as omitted variables, reverse causality and the feedback effect, which can invalidate the results Brazilian firms typically issue more than one type of stock (e.g., common and preferred), we selected the stock with highest liquidity in the respective year to calculate some market BBR, Braz.Bus.Rev.(Engl.ed., Online),Vitória, v. 11, n. 1, Art. 5,feb.2014 www.bbronline.com.br(L2) or Novo Mercado (NM) trading segments of the BMF&Bovespa in the respective year, and zero otherwise.The firms listed in these segments must satisfy stricter corporate governance standards than required by law, as described by Carvalho & Pennacchi (2012): ) can control for the feedback effect.Besides this, we believe the reverse causality was mitigated by the estimation of the System GMM by means of the xtabond2 tool of Roodman BBR, Braz.Bus.Rev.(Engl.ed., Online),Vitória, v. 11, n. 1, Art. 5, p. 90 -110, jan.-feb.2014www.bbronline.com.br (