The Differential Effects of Classified Boards on Firm Value

Transcription

1 The Differential Effects of Classified Boards on Firm Value SEOUNGPIL AHN, VIDHAN K. GOYAL and KESHAB SHRESTHA * March, 2009 ABSTRACT We find that classified boards actually increase the value of firms that have low monitoring costs and greater advising needs. Conversely, when firms are costly to monitor and have few advising needs, classified boards reduce value. Thus, the entrenchment effects of classified boards documented in much of the previous literature are found only for firms that are difficult for outside directors to monitor and that require smaller advisory roles by directors. We also show that board classification is associated with enhanced board stability. This enhanced board stability provided by classified boards has both negative and positive implications on firm value. Finally, we test the impact of board classification on pay-performance sensitivity and the investment efficiency. These results illustrate the channels through which board classification causes the observed differential effects on firm value among firms with various monitoring costs and advisory needs. JEL classification: G32, G34, K22 Keywords: classified boards, staggered boards, corporate governance, independent directors, board composition, board size We thank Michael Lemmon, Ronald Masulis, Karlyn Mitchell, Mark Walker, Bernard Yeung, and seminar participants at Midwest Finance Association meetings in Chicago 2009, National University of Singapore and Sogang University for helpful comments and suggestions. * Ahn is from the National University of Singapore (corresponding author, bizsa@nus.edu.sg, phone: ); Goyal is from Hong Kong University of Science and Technology; and Shrestha is from Nanyang Technological University. 1

2 1. Introduction A classified board is a governance structure that divides the board of directors into separate classes, with a single class of directors standing for re-election each year. Directors serving staggered terms are typically re-elected once every three years. Many commentators and shareholder activists claim that staggered terms insulate boards from removal by shareholders, thereby entrenching management. The consequent shirking, empire building, and extraction of private benefits by incumbents reduce firm value in firms that have classified boards. These claims have received some support in the literature. Several recent papers show that firms with classified boards have lower value. 1 Faleye (2007), for example, estimates that having a classified board reduces a typical firm s Tobin s q ratio by 13.15% (equal to a $795 million reduction in the typical firm s market value). Given these large economic magnitudes, it seems surprising that currently over 60% of large publicly traded U.S. firms maintain classified boards. This proportion has remained fairly stable over the last decade. In summary, the disadvantages of classified boards are well understood: mainly, they insulate management from discipline imposed by the market for corporate control and therefore increase agency conflicts. This lowers firm value. The advantages, however, are less well understood. Staggered terms provide directors with multi-year terms. This substantially enhances their incentives to bear the costs of investing in information required to actively monitor managerial performance and to provide advice and guidance to top managers on firm strategy and actions. 1 Recent research that documents a negative relation between classified board and Tobin s q includes papers by Gompers, Ishii, and Metrick (2003), Bebchuk, Cohen, and Ferrell (2004), Bebchuk and Cohen (2005), and Faleye (2007). 1

3 Classified boards also increase board stability and enhance director independence, thus insulating board members from outside pressures. 2 In the absence of board classification, directors will stand for election annually and there is a threat that a board member who refuses to conform to the dominant view during board meetings will not be renominated. Multi-year terms insulate directors from retribution (Wilcox, 2002). Finally, in firms with classified boards, directors have more discretion that can sometimes be value-enhancing. 3 In takeover situations, for example, board classification can improve the bargaining power of target managers over the transaction surplus on behalf of their shareholders. The tradeoffs associated with staggered director terms imply that firms will adopt classified boards if the benefits of classification exceed the costs of managerial entrenchment. We take the perspective that providing directors with multi-year terms is valuable for certain types of firms. The nature of a firm s assets and the scope and complexity of its operations will determine the effect of classified boards on firm value. When a firm s assets are information intensive, monitoring by outside directors is generally considered less effective. Raheja (2005) and Harris and Raviv (2008) argue that in firms with relatively opaque assets, an information advantage by insiders implies that boards will largely consist of insiders. Staggering terms of insiderdominated boards will further entrench managers. However, in firms with relatively transparent assets, board independence and stability, the increased incentives to bear the costs of investing in information, and greater discretion provided by staggered terms will increase the effectiveness of monitoring and hence result in higher firm value. 2 See Bradley et al. (2007). 3 For example, Robert C. Siegel, CEO and Chairman of Stride Rite Co., urged the company shareholders to vote against a shareholders proposal to remove board classification citing numerous benefits of board classification including board stability, effectiveness in long-term strategic planning, orderly succession of directors, and negotiation power in unsolicited takeover threats. He further pointed out that these beneficial aspects of classified boards enhance shareholders wealth (DEFA 14A, 1996). 2

4 Similarly, firms with greater advising needs, such as those with a greater scope and complexity of operations, will benefit from classified boards. By promoting board stability, board classification preserves directors incentives for making firm-specific investments required for their advisory role. Staggered terms also imply that a majority of directors serving at any given time has prior experience at the firm, likely to be more valuable for firms with greater advisory needs. Firms with complex operations often have greater advisory needs, and we expect classified boards in these firms to be value enhancing. We test these predictions on a sample of 6,410 firm-year observations over the 1998 to 2004 period. Consistent with our predictions, we find that classified boards increase firm value for firms that have low monitoring costs and high advising requirements. By contrast, board classification reduces value for firms with high monitoring costs and low advising needs. We also find that staggered terms increase board stability regardless of a firm s monitoring costs and advising needs. This suggests that the enhanced board stability provided by classified board structure has both negative and positive implications on firm value, and the net effect of classified boards on firm value through the enhanced board stability depends on a firm s information costs and advisory needs. Finally, we examine the channels through which board classification causes the observed differential effects on firm value. Specifically, we test the impact of board classification on payperformance sensitivity and the investment efficiency. The results show that board classification is associated with significantly lower CEO pay-performance sensitivity in firms with higher information costs and it is associated with significantly higher investment-q sensitivity in firms with more advisory needs. These results illustrate some of channels through which board 3

5 classification causes the observed differential effects on firm value among firms with various monitoring costs and advisory needs. Overall, the different pieces of evidence presented here paint a consistent picture of the effect of board classifications on firm value. Board classification is value enhancing for firms that are easy to monitor and have high advising needs. By contrast, classification is value destroying for firms that are difficult to monitor and have low advising needs. Our finding of a positive relation between classified boards and Tobin s q for certain classes of firms is new. The evidence in the paper adds to the growing literature that argues that the one size fits all approach to board structure is misguided. 4 The finding that classified boards increase firm value in certain cases stresses the importance of understanding the tradeoffs that firms face in adopting certain board structures. Our contribution is to explore these tradeoffs in the context of classified boards. The results imply that if all firms are compelled to conform to a single model of board structure, i.e., a single class with directors standing for election each year, value would be lost for firms in which classified boards are optimal. This paper is organized as follows. In Section 2, we discuss the related literature and develop our hypotheses. Section 3 describes our sample selection procedure and provides descriptive statistics of the sample firms. Section 4 presents the univariate results and examine the relation between Tobin s q and classified boards in multivariate tests. Section 5 presents results on the effect of classified boards on board stability. Section 6 examines the impacts of classified boards on pay-performance sensitivity and investment efficiency. Section 7 concludes the paper. 4 This recent literature includes Bhagat and Black (2001), McConnell (2002), Lehn, Patro, and Zhao (2007), Boone, Field, Karpoff, and Raheja (2007), Link, Netter, and Yang (2007), and Coles, Daniel, and Naveen (2008). 4

6 2. Information Costs, Firm Complexity, and Classified Boards Boards play two distinct and important roles they monitor managers and they provide advice to managers. 5 Board classification affects both the monitoring and advisory functions of boards. We begin with a discussion of how staggered boards affect the monitoring of managers, and then discuss the effect of staggered boards on the advisory function of boards. It is generally known that a board s ability to monitor managers depends, in part, on the nature of a firm s assets. Raheja (2005) and Harris and Raviv (2008) argue that monitoring by outside directors is less effective when information costs are high because outsiders are at an information disadvantage relative to insiders. In other words, outsider-controlled boards are less efficient in monitoring opaque firms (R&D intensive firms with high levels of intangible assets) compared with relatively transparent firms. Boards of such firms optimally consist of fewer outside directors. Staggering the terms of these insider-dominated boards will further entrench managers. Thus, in firms with largely opaque assets, classified boards will be associated with lower firm value. Alternatively, board classification may enhance managerial incentives to invest significantly more in firm-specific human capital. DeAngelo and Rice (1983) and Stein (1988) infer that antitakeover provisions including staggered boards dissuade opportunistic biddings in instances where the value of a firm s projects are opaque and thus can not be accurately conveyed to outside investors. This deterrence effect in turn enhances the incentives of insiders to invest their human capital in projects whose cash flows is less transparent to outside investors. In this case, we expect that in firms with high asset opacity, board classification may reduce managerial 5 See, for example, Williamson (1975), Fama (1980), Fama and Jensen (1983), and Hermalin and Weisbach (1998). 5

7 entrenchment and increase firm value. Whether classified boards reduce or exacerbate managerial entrenchment problem is, therefore, an empirical issue we test. The empirical tests rely on two primary measures of asset opacity the ratio of research and development expenditure (R&D) to assets and the ratio of intangible assets to total assets. Firms whose assets are largely intangible and have high R&D intensity are likely to be relatively opaque. These two measures of asset opacity are combined into a factor score, which is a linear combination of the transformed values of R&D intensity and asset intangibility. Factor scores are positively correlated to both R&D intensity and asset intangibility. Firms with above-median factor scores are termed opaque (OPAQUE=1) and those with below-median factor scores are termed transparent (OPAQUE=0). Alternative definitions of asset opacity that include other attributes such as firm age and variance of stock returns produced identical results. See Section D for more details. We now discuss the effect of board classification on the advisory role of boards. It is generally recognized that the advisory needs of firms vary. Klein (1998) and more recently Boone et al. (2007), and Coles, Daniel and Naveen (2008) argue that complex firms have greater advisory needs. Firm complexity varies with firm size, firm age, and diversity of operations. We take the perspective that the advisory role of the board requires that directors of complex firms invest significantly more in firm-specific human capital, which is not easily transferable. Board classification, which provides directors with a degree of permanence, aligns director incentives to invest in firm-specific human capital. Multi-year terms also provide boards with incentives to invest in a network of contacts that may be useful in resource gathering and obtaining new business. In addition, staggered terms are perhaps attractive to individuals with specific human capital to join a board. Finally, staggered terms imply that a majority of directors 6

8 serving at any given time has prior experience at the firm, which is likely to be valuable for firms with greater advisory needs. These arguments imply that classified boards will be value enhancing in firms with greater complexity, which results in higher advising needs. In firms with less advising needs (less complex or simple firms), the benefits of staggered terms are likely to be small. We measure firm complexity along multiple dimensions as in Coles, Daniel, and Naveen (2008). 6 First, it is commonly argued that the advisory needs of firms increase with firm size. Larger firms have more external contracting relationships and a greater dependence on the environment for resources. Thus, size increases firm complexity and the advisory needs of the firm. Second, firms operating in multiple segments tend to be more complex. CEOs managing diversified firms have greater advising needs. CEOs of firms with larger scopes of operations rely more on outside expertise from a greater number of industries. Finally, we expect older firms to have more complex contracting relationships. As firms grow older, they add more outside directors to their boards. The complexity of their operations and consequently their advising needs increase with firm age. As before, we compute a factor score based on firm size, firm age, and the number of segments. The score increases with greater firm complexity; it is positively related with firm size, firm age, and the number of business segments. Firms with above-median factor scores are referred to as complex firms (ADVICE=1) and those with below-median scores are referred to as simple firms (ADVICE=0). We test the robustness of the ADVICE variable by additionally including a firm s leverage. Coles, Daniel, and Naveen (2008) argue that higher leverage results 6 The term complexity may indicate the technological complexity of a firm s operations, which may make it difficult for outsiders to monitor the firm. Throughout this paper, we use the term firm complexity to indicate a larger asset base, diversity of operations, and greater interactions with the outside contracting environment. 7

9 in greater advising requirements because of their reliance on outside resources. 7 The results are robust to including leverage in the definition of complexity. In summary, we expect classified boards to increase firm value in firms that are relatively transparent (OPAQUE=0) and complex (ADVICE=1). We expect classified boards to reduce firm value in firms that are relatively opaque (OPAQUE=1) and simple (ADVICE=0). 3. Data and variables 3.1. Sample selection The initial sample consists of the universe of firms included on the RiskMetrics (formerly, Investor Responsibility Research Center, IRRC) directors and governance databases for the period from 1998 to The RiskMetrics governance database issued four volumes of data in years 1998, 2000, 2002, and Following Gompers, Ishii, and Metrick (2003) and many others, we assume that during the years between two consecutive publications, firms have the same governance provisions as in the previous publication year. This governance data is matched with RiskMetrics directors database excluding firms missing CEO information. This procedure yields 9,108 firm-year observations of the initial sample. From this initial sample firms, we exclude firms with sales revenues less than $20 million, firms in Real Estate Investment Trusts (SIC 6798), and those lacking the required financial data from Compustat annual files. After these selection procedures, our final sample consists of 6,410 firm-year observations by 1,555 firms. In addition, we collect the state incorporation data from COMPUSTAT quarterly files and the number of segments from Compustat Segment files, and firm age is estimated based on the listing dates in CRSP. 7 Booth and Deli (1999) and Guner, Malmendier, and Tate (2005) finds that bankers frequently serve on corporate boards to enhance access to outside financing. 8 The RiskMetrics Director data starts in But its coverage on board information, required in our study, is limited until the 1998 release. 8

10 3.2. Variables The definition of our key variables is summarized in Appendix A. To measure firm opacity, we define R&D intensity as the ratio of research and development expenses to assets, and asset intangibility as one minus the ratio of net property, plant and equipment to assets. To construct firm complexity measure, we define firm size as the natural logarithm of assets, firm age as the number of years since the firm first appeared on CRSP, and the scope of operations as the number of segments the firm is operating in Performance measure As a measure of performance, we approximate Tobin's q as the ratio of market value of assets to book value of assets, where market value is defined as the book value of assets minus the book value of equity and deferred taxes, plus the market value of equity. The analysis includes a number of explanatory variables to control for factors that are expected to affect Tobin's q directly. 9 Thus, we control for investment opportunities by including the ratio of capital expenditures to assets and the ratio of R&D expenses to assets. The q regressions also control for firm profitability measured by the return on assets (ROA). In addition, the literature shows that diversified firms have had low Tobin's q during the last several decades (Lang and Stulz, 1994; Berger and Ofek, 1996). We therefore include a variable that equals the number of business segments reported by the company in Compustat segment files. We also control for firm size and leverage following the previous literature. All of our specifications include year dummy variables and industry indicator variables (at the two-digit SIC level) to control for the year specific effects and industry effects, respectively. 9 Other studies that regress Tobin's q on a set of governance and financial explanatory variables include Morck, Shleifer, and Vishny (1988), Lang and Stulz (1994), Yermack (1996), Daines (2001), and La Porta et al. (2002). 9

11 Board characteristics and ownership In addition to firm characteristics, previous studies have identified several governance variables that are known to affect firm value. These include board size (Yermack (1996)), board composition (Rosenstein and Wyatt (1990)), leadership structure (Rechner and Dalton (1991)), insider ownership (Morck, Shleifer, and Vishny (1988); McConnell and Servaes (1990)), outside block ownership (Bethel, Liebeskind, and Opler (1988)), and independence of nominating committee (Callahan, Miller, and Schulman (2003)) Antitakeover measures Recently, Gompers, Ishii, and Metrick (2003) showed that q is negatively related with their index of 24 antitakeover provisions (hereafter, referred to as the G-index). The 23 remaining governance provisions (other than the classified board) are aggregated into a net G-index, computed as the G-index minus one for firms with a classified board and equal to the G-index for firms without classified boards Delaware incorporation There is a vigorous debate in the literature about the effect that corporate law has on firm value. With more than half of U.S. public firms incorporated in Delaware, this debate largely revolves around whether Delaware law matters. The question of whether Delaware law affects firm value remains unsettled. Daines (2001) finds that firms incorporated in Delaware have significantly higher firm value. He argues that the law in Delaware lowers acquisition costs and facilitates the sale of firms, thereby increasing firm value. These results have been challenged by Subramanian (2004) who argues that the Delaware effect has disappeared in the more recent period. To control for the potential effects of incorporation in Delaware on Tobin's q, we include an indicator variable, Delaware, that takes a value of 1 if a firm is incorporated in Delaware, and 10

12 zero otherwise Descriptive statistics Table 1 presents mean and median values of board structure variables, firm characteristics and other governance variables. Almost 61% of the firms have classified boards. In unreported tables, we find that the fraction of firms that have classified boards in the IRRC sample has remained stable over the entire sample period. Bebchuk and Cohen (2005) make a similar observation and report that over 60% of firms during the period from 1995 to 2002 had classified boards. The average Tobin's q is 1.98 (the median is 1.48). Consistent with the existing literature, Table 1 shows that firms with classified boards have significantly lower Tobin s q. Consistent with our predictions in Section I, we find that firms with classified boards have less informationintensive assets; they spend less on R&D and more of their assets are tangible. The median firm with a classified board is larger than the median firm without a classified board, although the averages suggest otherwise. Firms with classified boards are more diversified. They have 3.03 segments on average compared with 2.81 for firms without classified boards. Firms with classified boards are more likely to have multiple segments compared to firms without classified boards. Firms with classified boards are also more levered. The mean and median debt-to-asset ratio for firms with classified boards is significantly higher compared with that for firms without classified boards (0.31 versus 0.28 in average debt-to-asset ratio). Firms with classified boards have significantly larger boards, a larger fraction of independent directors, and fewer inside directors. Directors of firms with classified boards are older and they sit on more boards than do directors of firms without classified boards. Share ownerships of CEO and outside directors are, however, similar regardless of board classification. The CEO is more 11

13 likely to be the board chairman in firms with classified boards and the CEO more frequently sits on the nominating committee. Finally, we note that firms with classified boards score significantly higher on the net G-index. The average net G-index for firms with classified boards is 9.3 compared to 7.6 for firms without classified boards. 4. The relation between classified boards and firm value In this section, we provide evidence on whether the effects of classified boards on firm value are different among firms with various monitoring costs and advisory needs. We do so by first comparing Tobin s q among different types of firms and then testing the association between Tobin s q and classified boards in multivariate settings Univariate tests Table 2 presents mean and median Tobin s q for different types of firms with and without classified boards. Panel A, which stratifies the sample by asset opacity, shows that the lower Tobin s q of firms with classified boards is largely driven by firms that have high monitoring costs (OPAQUE=1). Opaque firms have higher q than transparent firms, but conditioning on these firms being opaque firms, Tobin s q of firms with classified boards is significant lower by (-0.24 in median) than q of firms without classified boards. Among transparent firms with low monitoring costs, the average and median Tobin s q are similar in firms with and without classified boards. Similarly, Panel B partitions firms on their advisory needs and board classification. The result shows that the lower Tobin s q for firms with classified boards is largely driven by simple firms with few advising needs. In complex firms with greater advising needs, the mean and median Tobin s q are similar in firms with and without classified boards. This indicates that the 12

14 valuation effect of classified boards is more positive in complex firms compared to it is in simple firms. The result in Pane B, however, is not sufficient to prove that complex firms with classified boards have higher q than those without classified boards, as we predict. Since our measures of OPAQUE and ADVICE dummy are not completely exclusive to each other, some firms partitioned as complex firms are opaque firms at the same time and therefore the positive effect of classified boards in complex firms could be tarnished by the negative effect of classified boards in opaque firms, resulting in the insignificant result in Panel B. 10 To explore this issue further, we examine the joint effect of asset opacity and firm complexity on the value of firms with classified boards and those without. We expect that the positive effect of classified boards will be most heightened in the combination of complextransparent firms, and the negative effect of classified boards will be most evident in opaquesimple firms. From Panel C of table 2, complex-transparent firms with greater advising needs and low monitoring costs (ADVICE = 1 and OPAQUE = 0), the median Tobin s q is indeed higher when firms have classified boards compared to when they do not. The average q is positive but insignificant By contrast, in opaque-simple firms with high monitoring costs and few advising needs (OPAQUE = 1 and ADVICE = 0), Tobin s q is significantly lower when firms have classified boards compared with when they do not. The difference of q is lowest (-0.28 in median) among these sub-groups. In other sub-groups of firms in-between these two extremes, board classification has moderate effect on q differences. This is consistent with our predictions that the negative effect of classified boards on firm value is largely concentrated among firms that are difficult to monitor (opaque firms) and it is less so among firms that are easier to monitor (transparent firms). The board classification benefits firms that have greater 10 The Pearson correlation coefficient between OPAQUE and ADVICE is from Table 3 correlation matrix. 13

15 advising needs (complex firms), but less beneficial for firms that have few advising needs (simple firms). The univariate test results presented in Table 2 do not consider the correlation structure of other variables which affect Tobin s q. The Pearson correlation matrix in Table 3 shows that some of these variables effects on Tobin s q. In addition to OPAQUE and ADVICE dummy variables and their corresponding firm characteristics variables, Net G-index, board size, and the proportion of inside and outside directors are significantly correlated with Tobin s q and classified boards. It is also noteworthy that a classified board is positively correlated with OPAQUE and it is negatively correlated with ADVICE. This is consistent with the discussion in Section 3 that complex firms have greater advising needs and they benefit more from staggering the terms of their directors. In opaque firms, classified boards are likely to exacerbate managerial entrenchment. Thus, opaque firms are less likely to adopt classified boards, while complex firm requiring greater advice are more likely to have classified boards. In the following sections, we examine the relation between classified boards and firm value in multivariate settings after controlling for the effects of other variables Multivariate results: Classified boards and firm value Baseline regression models We begin our tests on the relation between Tobin s q and classified boards with a regression specification similar to that used by Bebchuk and Cohen (2005) and Faleye (2007). The baseline specification is then augmented with interaction terms between classified boards and variables measuring the information intensity of assets and firm complexity. Column (1) of Table 4 reports results from the baseline regression of Tobin s q on the classified board indicator variable, log of firm size, log of number of segments, log of firm age, 14

16 leverage, ROA, and capital expenditures-to-assets ratio. All regression specifications include year and industry (two-digit standard industrial classification, SIC) indicator variables. The t- statistics are based on robust standard errors. Consistent with Bebchuk and Cohen (2005) and Faleye (2007), the results in column (1) show that classified boards are negatively associated with firm value. The coefficient of on the classified board variable is similar to that reported in the previous research. In column (2), we control for other corporate governance provisions by including the net G- index. Consistent with prior literature, Tobin s q is declines as the net G-index increases (see Gompers, Ishii, and Metrick, 2003; Bebchuk and Cohen, 2005). However, even with other governance provisions in the regression, the effect of classified boards on Tobin s q continues to remain negative and statistically significant. In column (3), we include additional variables measuring the information intensity of assets, such as the high R&D indicator variable and the intangibility ratio. Both of these variables are positively associated with Tobin s q but the negative effect of classified boards on Tobin s q persists in the results. Overall, these results match those in the existing literature, which makes sense, because our sample and methodology resemble those in other papers. In columns (4) to (6), we report results from regressions with interaction terms between classified boards and firm attributes measuring the information intensity of assets and the complexity of operations. These interaction terms allow us to test if classified boards enhance firm value in certain types of firms. Column (4) includes an interaction term between classified board and OPAQUE to test if classified boards in firms with more costly monitoring are less valuable. Monitoring costs increase as the information intensity of assets increases, resulting in less monitoring by outside 15

17 directors. Consequently, classified boards may exacerbate managerial entrenchment in such firms without any countervailing benefits. As expected, we find that the coefficient on the interaction between classified boards and OPAQUE is negative and statistically significant at the 1% level. From the bottom of Table 4, the total effect of classified boards in opaque firms is on Tobin s q. At the same time, we find that the coefficient on classified boards is insignificantly negative for transparent firms (OPAQUE = 0). These results confirm our univariate findings that the negative effect of classified boards on firm value is largely due to opaque firms (i.e., firms that are R&D intensive and with mostly intangible assets). In column (5), we include an interaction between classified boards and ADVICE in the baseline regression specification. As predicted, the estimated coefficient on this interaction term is positive and statistically significant at the 1% level. The F-test reported at the bottom of the table shows the total effect of classified boards in complex firms (ADVICE = 1) is and it is statistically significant at the 5% level. This indicates that firms with classified boards have higher Tobin s q when these firms have greater advising needs. By contrast, the coefficient on the classified board indicator is negative, suggesting that firms with classified boards have lower firm value when advising needs are low (i.e., ADVICE=0). In column (6), we examine the joint effect of both OPAQUE and ADVICE by including both dummy variables in the same specification. Control variables that constitute OPAQUE and ADVICE are excluded to avoid multicollinearity. 11 Consistent with the results in previous columns, we find a significant negative coefficient on the interaction of OPAQUE with classified boards and a significant positive coefficient on the interaction of ADVICE with classified boards. The total effect of classified boards is for opaque-simple firms (OPAQUE=1 and 11 The inclusion of all other control variables does not affect the coefficients on the classified board and its interaction terms. However, the coefficients on the OPAQUE and ADVICE dummy variables become insignificant. 16

18 ADVICE=0) and it is for complex-transparent firms (ADVICE=1 and OPAQUE=1). The total effect of classified boards for the remaining sub-category firms are for (OPAQUE=0 and ADVICE=0) and (OPAQUE=1 and ADVICE=1), which are located in-between the total effects of the two extreme groups. Overall, the result from column (6) indicates that classified boards have a negative impact on the value of opaque firms and conversely, a positive impact on the value of complex firms. Therefore, the effect of classified boards on Tobin s q differs by firms monitoring costs and advisory needs. Our results in Table 4 is also broadly consistent with Faleye (2007), who reports stronger negative effect of classified boards on Tobin s q for R&D intensive firms and weak negative effect for large firms Additional Control Variables We perform the first set of robustness tests by including additional governance variables in the regression specifications reported in Table 4. The additional variables are: (i) the free cash flow-to-assets ratio, (ii) CEO ownership (and CEO ownership-squared), (iii) combined CEO- Chair position, (iv) CEO in nominating committee, (v) CEO tenure, (vi) outside director ownership, (vii) Delaware incorporation, (viii) board size, (ix) fraction of insiders, (x) average directorships, and (xi) average director age. Table 5 reports results from regressions that include these additional control variables. As shown in columns (1) to (3), the addition of these variables has no material effect on our conclusions regarding the effect of monitoring costs and firm complexity on the sensitivity of 12 From Table 5 in Faleye (2007), the coefficient on classified boards is (significant at the 1% level) for R&D intensive firms and it is (significant at the 10% level) for large firms, compared to the coefficient of classified boards of for the entire sample firms. Faleye (2007) conjectures that R&D intensive firms and large firms are firms that potentially benefits from classified board structure and conclude they are not. Our prediction has subtle difference with his in that R&D intensity is a proxy for monitoring costs and firm size is a proxy for advisory needs. We also include other measures of firm diversity and firm age to infer a firm s advisory needs. 17

19 Tobin s q to the presence of classified boards. The results are virtually identical to those presented in columns (4) to (6) of Table 4. The next two columns of Table 5 examine if our results are robust to explicit consideration of other board structure variables and their interactions with firm attributes found to be important in explaining the variation in Tobin s q. Coles, Daniel, and Naveen (2008) show that firms with greater advising needs have larger boards and that Tobin s q is positively related to board size in such firms. In addition, previous studies find that firms with high R&D intensity and intangible assets have smaller boards. To control for the effect of board size and board composition on Tobin s q and to make sure that our findings of differential effect of classified boards are not merely capturing the effects of boards, we explicitly include (i) the log of board size and its interaction with OPAQUE and ADVICE indicator variable in column (4) and (ii) the fraction of insiders and its interaction with OPAQUE and ADVICE indicator variable in column (5). Consistent with the findings in Coles, Daniel, and Naveen (2008), the coefficient estimate on the interaction between the log of board size and ADVICE variable is significantly positive. More importantly, these additional variables have no effect on our conclusions regarding the differential effects of classified boards on Tobin s q among firms grouped by the monitoring costs and the complexity of their operations. We continue to find a significant positive coefficient on the interaction of ADVICE with classified boards and significant negative coefficient on the interaction of OPAQUE and classified boards. At the bottom of column (4) in Table 5, the total effects of classified boards for opaque-simple firms and for complextransparent firms remain statistically significant and in the predicted signs. Similarly, in column (5), the inclusion of the fraction of inside directors and its interaction with OPAQUE or ADVICE variable do not materially affect the coefficient on the interaction 18

20 terms of classified boards with OPAQUE or ADVICE indicator variables. The interaction of OPAQUE and the fraction of inside directors is positive, though it is statistically insignificant. Bebchuk and Cohen (2005) and Bates et al (2008) argue that board classification itself does not constitute a substantial impediment to a hostile takeover unless it is an effective classified board. We define effective classified boards as such a classified board is established in corporate charters or it is established in corporate bylaws and shareholders are not allowed to (1) modify the charter or bylaws without board consent, (ii) call special meetings, or (iii) replace directors without cause. A total of 160 cases (or, 4% of original classified boards) are reassigned as no board classification. 13 We then reestimate the regression model in column (6). This does not have a material effect on our coefficients estimated. We also conduct Median regression to further eliminate the potential influence of outliers. In column (7) of Table 5, the total effects of classified boards become smaller in their magnitude compared to those from OLS regressions in column (3)-(6). Nonetheless, the effect of classified boards in opaque-simple firms is negative and it is positive in complex-transparent firms, and they remain statistically significant at the conventional levels. In sum, the evidence in Table 5 reinforces our findings that in firms with lower monitoring costs and greater advising needs, the benefits of board stability outweigh any managerial entrenchment effects that may arise when director terms are staggered The data on the charter-based and bylaw-based classified boards are from Bebchuk and Cohen (2005). The data series ends in year 2002 and we assume that the information on classified boards remain the same by year We also thank professor Lucian Bebchuk for providing this data for the paper. 14 We also estimate the models in Table 5 using the Fama-Macbeth method. In the method, t-statistics for the coefficients on classified boards and its interaction term are qualitatively the same as these reported in Table 5. Cochrane (2001) also mentions that the Fama-Macbeth method does not make a difference to OLS regression results when regressors are time-invariant. In our case, the classified board variable is extremely persistent. Only about 5% of firms changed board classification during the sample period. 19

21 Addressing Endogeniety Concerns The results presented so far raise two endogeneity concerns. One concern, noted in much of the existing literature, is that managers of poorly performing firms may adopt antitakeover strategies to insulate themselves from the market for corporate control. These reverse causality issues are less important in the case of classified boards since these structures were predominantly adopted in the 1980s (Bebchuk and Cohen (2005); Masulis, Wang, and Xie (2007); Bates, Becher, and Lemmon (2008)). In the 1990s and thereafter, few publicly traded firms have been able to stagger the terms of their boards. Most adoptions of classified boards since 1990 have occurred at the time of IPOs. Excluding IPOs in the 1990s yields results that are similar to those reported earlier. The reverse causality also does not explain the positive impact of classified boards on Tobin s q in complex-transparent firms. The second concern is that an omitted variable that affects Tobin s q may also be related to a firm s choice of its board structure. To address this concern, we use a self-selection model (see Heckman, 1979, and Li and Prabhala, 2007, for details). In the first step, we estimate the probability of adopting classified boards using an instrument that affects the probability of classified board adoption but does not directly affect Tobin s q. The instrument we use is incorporation in Massachusetts. In 1990, the state of Massachusetts enacted the Massachusetts Classified Board Law which explicitly requires all public firms incorporated in Massachusetts to have classified boards by default. 15 Using state incorporation data in Compustat quarterly files, we construct a dummy variable, Massachusetts, which takes a value of one if the firm is incorporated in Massachusetts prior to year 1990, and zero otherwise. This yields 163 firm-year observations that incorporated in the state of Massachusetts. Incorporation in Massachusetts 15 Firms are allowed to opt out of the Massachusetts classified board law if there is an action of the board or shareholders approval (see Faleye, 2007). Nonetheless, the law causes majority of firms incorporated in Massachusetts to adopt classified boards. 20

22 increases the likelihood of adopting classified boards. At the same time, the introduction of the law itself is exogenous to any particular firm s decision and therefore it is unlikely to be related to the unobservable variables that affect Tobin s q. Table 6 presents results from probit models that explain the likelihood of observing classified boards in firms. As predicted, the coefficient on the Massachusetts indicator variable is positive and highly significant in all models. From column (6), the marginal effects based on coefficient estimated in column (5) suggest that firms incorporated in Massachusetts have a 16.2% higher likelihood of adopting classified boards. The results in columns (1)-(4) of Table 6 are consistent with our predictions. Opaque firms are less likely to adopt classified boards. In the presence of other antitakeover provisions, the sign on ADVICE variable switches to negative from the positive correlation reported in Table 3. Signs of coefficients are largely negative on variables representing disciplines on CEO s incentives. Disciplinary role of leverage and monitoring incentives from outside directors measured by ownerships of outside directors deter the adoption of classified boards. Variables related to the magnitude of private benefits such as free cash flows and net G-index increase the probability of classified board adoption. Reflecting the high correlation between the net G-index and the classified board indicator variable as reported in Table 3, the net G-Index increase the probability to adopt classified boards. Firms with CEO present at the nominating committee are more likely to adopt classified boards. At the low levels of CEO ownership in which CEO is less likely to be entrenched, Board size and composition have material impacts on the probability of adopting classified boards. The coefficient on board size is positive and significant in column (5). This indicates that firms with large boards may adopt classified boards to avoid nuisance to elect a large numbers of 21

23 directors annually. The proportion of insiders is negatively related to the probability of adopting classified boards. Table 7 presents results from the second-stage Tobin s q regressions corresponding to each column in Table 6 as the first-stage estimation. In all columns of Table 7, the estimated coefficients on λ (the inverse Mills ratio) are negative, but either insignificant or marginally significant at the 10% level. This suggests weak evidence that firm characteristics that affect a firm s decision to adopt classified boards are also associated with lower Tobin s q. In column (1), the coefficient on classified boards is and it is insignificant. This indicates that after correcting for the selection bias, the effect of classified boards on q is, on average, indifferent from zero for the entire sample firms. This contradicts our findings in Table 5 and previous literature, which documents significantly negative impact of classified boards on firm value. This insignificant effect of classified boards, however, is consistent with evidence from target returns reported in Bates et al. (2008). Our main focus is on the specification reported in columns (2)-(5), which includes interaction terms between classified boards and OPAQUE, ADVICE variables. Consistent with the findings in Table 5, we continue to find the differential effects of classified boards on firm value. In columns (2)-(5), the coefficient on the interaction of classified boards and OPAQUE is negative and significant, while the coefficient on the interaction of classified boards and ADVICE is positive and significant. At the bottom of Table 7, the total effects of classified boards for opaque-simple firms appear less negative than those reported in OLS regressions, and the total effects of classified boards on firm value for complex-transparent firms are positive at bigger magnitude compared to those reported in Table 5. 22

24 In sum, the evidence from self-selection models is consistent with our predictions that the effect of classified boards on q is positive in firms with low information costs and greater operational complexity, and it is negative in firms with high information costs and less operational complexity Additional Robustness Tests In this section, we discuss additional robustness tests that (a) employ alternative definitions of the OPAQUE and ADVICE dummies, (b) run separate regressions by firm classes, (c) control for the influence of outside blockholdings, (d) test whether our findings are affected by the inclusion of newly listed firms, (e) test whether the combination of poison pills with classified boards affect our results, and, finally, (f) test whether our findings are robust to different time periods, in particular the period surrounding the passage of the Sarbanes-Oxley act. These results are not reported in tables to save space Alternative Definitions of OPAQUE and ADVICE Variables Our main tests are based on two firm attributes, which measure the information intensity of a firm s assets (OPAQUE) and the complexity of its operations (ADVICE). The OPAQUE variable is a principal component factor extracted from R&D expense ratios and asset intangibility. The ADVICE variable is a principal component factor extracted from firm size, the number of segments, and firm age. We consider alternative definitions of OPAQUE that include additional proxies for monitoring costs including industry median q, stock return volatility and firm age. Similarly, we consider alternative definitions of ADVICE that include additional proxies of complexity including leverage and CEO ownership. To test if our results are sensitive to the construction of OPAQUE and ADVICE, we reestimate the regression specifications reported in Table 5 with these alternative definitions. Our 23

25 main results are robust to using these alternative constructions of OPAQUE and ADVICE. The coefficient on the interaction terms between classified boards and OPAQUE and that between classified boards and ADVICE have similar signs and significance. One notable difference is that the coefficient on classified boards itself is mostly insignificant in these specifications. Another concern with our proxy variables for monitoring costs and firm complexity is that they may not be mutually exclusive. For example, firm age may be related to both monitoring costs and firm complexity. To address this concern, we construct factors based on all proxy variables and then retain multiple factors from principal component analysis with eigenvalues in excess of 1. This exercise yields two factors. The first factor is positively correlated with R&D expenditure and asset intangibility and negatively correlated with firm age, firm size and the number of segments. The opposite correlations are observed with the second factor. Employing these two factors in place of OPAQUE and ADVICE, respectively, generates similar inferences. The interaction of the first factor with classified boards is significantly negative and the interaction of the second factor with classified boards is significantly positive Classified Boards and Tobin s q: Estimation by Firm Classes So far, we have relied on the interaction terms to capture the effect of complexity and opaqueness on the relation between classified boards and Tobin's q. To address concerns that the coefficients on control variables in the q regressions may differ systematically for different classes of firms, we re-estimate the effect of classified boards on Tobin s q for various firm classes based on the information intensity of assets and firm complexity. This does not have a material effect on our conclusions. 24