The original sample is compiled from the Investor Responsibility Research Center (IRRC), which collects data on the governance index. Because the IRRC collects data only periodically, our index is restricted to the years in which the IRRC has data on corporate governance, e.g., 1998, 2000, and 2002. We assume that the adoption of anti-takeover provisions for every specific firm is stable and tends to be constant in the short run. According to this assumption, for the year that IRRC do not have any publication, we use the index recorded by IRRC at prior period to proxy for the missing data. Data of other variables are obtained from Compustat.
Two industries are traditionally heavily regulated: financial and utility. The nature of financial firms, particularly depository institutions, is such that leverage cannot be interpreted the same way as in industrial firms. In addition, because regulators already provide a certain degree of monitoring, managers of regulated firms should be less able to reap private benefits at the expense of shareholders (Booth, Cornett, and Tehranian, 2002; Kole and Lehn, 1997).
Considering about the characteristics of these two industries, we exclude firms who’s SIC code fall between 6000 to 6999, or 4000 to 4999. To drive out the influences of extreme samples, we exclude samples for every calculated variable which follows in the top 1% or in the bottom 1%.
3.2 Model selection
Before going through our regression analysis, we calculate the variation inflation factors (VIF) of the control variables to test for the existing of the multicolinearity. The explanatory variables with higher VIF have more serious multicolinearity problem. To avoid further estimated bias, we have to ensure that the multicolinearity problem is under controlled. The calculated VIFs of our control variables are less than 3, while most of them are under 1.5. This means the multicolinearity problem of the explanatory variables is not serious.
The objective of our cross-sectional tests is to draw inferences about the relation among corporate governance, agency cost, capital structure, and firm value, while controlling for a number of other factors. As mentioned before, the task is complicated because it can be argued that corporate governance, agency cost, capital structure, and firm value are all jointly determined.
There exists a well-known problem of endogeneity bias in econometrics. Endogeneity bias happens in the situation that the relation between variables violates the major assumption of Original Least Squares (OLS) that explanatory variables are uncorrelated with the error terms.
There are properly two reasons causing the endogeneity problem. One is the existing of omitted
9
variables which is another variable correlated with both dependent and explanatory variables. So that after fitting the OLS model, there is still a relationship with this other variable and the residuals. The other most important source of endogeneity is reverse causality. To truly be able to make a causal claim, we need a truly exogenous variable called instrumental variable which is not related to any of the other variables in the system, unobserved and observed. To test the endogeneity problem, we employ the Hausman test for our regression model. The result is significant and rejected the hypothesis, consist with our prediction of existing of endogeneity problem.
Since we wouldn’t be able to use the OLS model to do our tests, the most common method for doing the actual estimation, three-stage least squares (3SLS), will help us clarify this issue.
3SLS is a statistical technique to analyze multivariate data. It combines two stage least squares (2SLS) with seemingly unrelated regression (SUR). Three stage least squares estimates are obtained by first estimating a set of nonlinear (or linear) equations with cross-equation constraints imposed, but with a diagonal covariance matrix of the disturbances across equations. This is the constrained two stage least squares estimator. The parameter estimates thus obtained are used to form a consistent estimate of the covariance matrix of the disturbances, which is then used as a weighting matrix when the model is reestimated to obtain new values of the parameters.
In addition, we employ the generalized method of moments (GMM) estimation which places no restrictions on either the unconditional or conditional variance matrix of the disturbance term.
Under the GMM framework we can obtain the efficient estimator which has the minimum asymptotical covariance matrix without making any additional assumptions. In this paper, we use panel data to construct our regression analysis. The advantage of GMM that allow conditional heteroskedastic on the disturbance term accords with what we need to get the most robust results.
We also use a J test to test for the overidentification while holding the GMM estimation. The results indicate that the instrument variables we use are efficient for dependent variables.
3.3 Regression analysis
Considering the potentially endogenous problem, we use 3SLS and GMM estimation models to test the relationship between corporate governance, firm value and leverage. The regression models are developed following:
t
10 refer to firm i. CGi,t isthe measure of corporate governance. Leveragei,t is measured as the ratio of total debt to total assets. Sizei,t is the scale of company measured by nature logarithm of sales.
GOi,t is refer to growth opportunity. Profitabilityi,t is measured by the ratio of EBIT to sales.
S&P500 is a dummy variable, if the firm observed is included in S&P500, it is recorded 1, or 0 otherwise. Competitioni,t is referred to product market competition, while Uniquei,t is a proxy for product uniqueness. FA ratioi,t is referred to fixed asset divided by total assets. R&Di,t is research and development expenditures scaled by sales. NDTXi,t means non-debt tax shields.
3.4 Variable construction 3.4.1 Firm value (Tobin’s Q)
Our firm valuation measure is Tobin’s Q, which has been used for this purpose in corporate-governance studies since the work of Demsetz and Lehn (1985) and Morck, Shleifer, and Vishny (1988). We follow Gompers, Ishii and Metrick (2003) method for the computation of Tobin’s Q [the ratio of a firm’s market value of assets over its book value of assets, where the market value of assets is computed as the book value of assets plus the market value of common stock less the sum of the book value of common stock and balance sheet deferred taxes].
Considering the industry effect, we also compute the median Q in each year in each of the 48 industries classified by Fama and French (1997) and subtract it from firms’ specific Tobin’s Q to obtain an industry-adjusted Tobin’s Q.
3.4.2 Corporate governance (CG)
Takeovers and takeover threats are the source of corporate governance considered in this paper. A great deal of theory and evidence suggests that takeovers address governance problems (see, e.g., Jensen (1988) and Scharfstein (1988)). Takeovers also typically increase the combined value of the target and the acquiring firm, indicating that firm performance is expected to improve posttakeover (Jensen and Ruback (1983)). Moreover, it is generally poorly performing firms that are targeted (Morck, Shleifer, and Vishny (1989)). However, a poorly performing firm can resist a takeover by adopting anti-takeover provisions (ATPs) in its charter. For our proxy of
11
corporate governance, the main interest is in measuring the extent to which a firm is protected against a takeover. This protection can take the form of direct anti-takeover provisions as well as other devices that provide managerial protection by restricting shareholder power to change charter provisions, to call for a shareholder meeting, or to overrule the management during a takeover attempt.
We incorporate the firm-specific defense mechanisms in place by using the index compiled by GIM from the Investor Responsibility Research Center (IRRC) publications. GIM (2003) introduce G-index which ranges from 0 to 24 as the proxy of corporate governance. They consider 24 different provisions in five categories—tactics for delaying hostile bidders, voting rights, director/officer protection, other takeover defenses, and state laws. G-index is formed by adding one point for every specific defensive provision adopted to restrict shareholder rights for each firm. As G-index increases, firms are expected to experience bad corporate governance and decreasing firm value. The G-index does not require judgments about the efficacy or wealth effects of any of these provisions. GIM only consider the impact on the balance of power.
We view this index as a measure of anti-takeover protection. Following Cremers and Nair (2005), we simply use a linear transformation of this index, CG = 24 − G-index, for ease in exposition. As a result, a larger value of CG signifies a higher vulnerability to takeovers, in turn, signifies a higher level of corporate governance quality.
In the robust test, to ensure that our results are not driven by any alternative interpretation of this index, we also adopt E-index, as mentioned before, constructed by Bebchuk, Cohen, and Ferrell (2005), to be taken as the measure of corporate governance. As Bebchuk, Cohen, and Ferrell (2005) emphasized, the E-index is expected to have stronger effect on firm value than G-index has. For ease in exposition, as the same reason, we employ a transferred index, CG’= 6 − E-index, to measure corporate governance.
3.4.3 Capital structure (Leverage)
Following Harvey, Lins, and Roper (2004), we use leverage (the ratio of total debt to total asset) as the measure of capital structure. Since Barclay et al. (2003) argue that book leverage is an instrument for the ratio of debt to a firm’s assets in place, we use leverage measured using book values throughout our analysis. We treat each year as a separate observation in order to allow for the possibility that leverage determinants like size and performance may change over time.
12
3.4.4 Instrument variables in equation (2)
One shortcoming of any instrumental variable technique is that it requires the identification of some number of exogenous variables that plausibly affect only value, or leverage, or ownership, but not all three. Here we want to find a instrument variable that affects only corporate governance.
Managers of firms operating in more competitive industries are less likely to shirk or put valuable corporate resources into inefficient uses, since the margin for error is thin in these industries and any missteps can be quickly exploited by competitors, which seriously jeopardizes firms’ prospects for survival and managers’ prospects for keeping their jobs.
The competitive environment can affect corporate governance structures in positive directions. If product market competition disciplines managers, then the marginal benefit of additional governance would be low, as competition would substitute for other mechanisms (Leibenstein, 1966; Hart, 1983). Alternatively, a competitive environment could raise the marginal cost of poor managerial decisions, resulting in a positive association between competition and internal governance strength.
Following Gillan, Hartzell, and Starks (2003), we try to capture the competitive structure of an industry with two different measures. The first is the Herfindahl index, calculated as the sum of squared market shares of all COMPUSTAT firms in each Fama-French (1997) industry. The second is each industry’s median ratio of selling expenses to sales, which Titmanand and Wessels (1988) argue acts as a proxy for product uniqueness.18 Industries with lower Herfindahl indices and industries where member firms have similar products have more competitive product markets.
For each year, we define an industry as competitive (unique) if the industry’s Herfindahl index (median ratio of selling expense to sales) is in the bottom (top) quartile of all 48 Fama-French industries. Both these two measures are used as the proxies for product market competition.
3.4.5 Control variables of Leverage
Numerous studies have argued that leverage may be positively affected by firm size.
Following Titman and Wessels (1988) and Johnson (1997), we use the natural logarithm of sales as a proxy for firm size. The composition of the firm’s assets has been found to affect capital structure decisions (Titman and Wessels, 1988 and Mehran, 1992). Hence, we include the fixed-asset ratio in the regression analysis. As in Johnson (1997), the fixed-asset ratio is property, plant, and equipment to total assets. Myers (1977) identifies growth opportunities as a significant determinant of capital structure. Similarly, Rozeff (1982) finds empirical support for growth
13
opportunities as a relevant variable. Profitability may be relevant to capital structure decisions.
Myers (1984) suggests that managers have a pecking order in which retained earnings represent the first choice, followed by debt financing, then equity. Thus, the pecking order hypothesis would imply a negative relationship between profitability and leverage. We employ the earnings before interest and taxes (EBIT) to sales ratioto control for profitability. DeAngelo and Masulis (1982) contend that non-debt tax deductions substitute for the tax shield benefits of debt. As a result, firms with greater non-debt tax shield would be expected to have lower levels of debt. We define non-debt tax shields as the ratio of the sum of depreciation and amortization to total assets.