Corporate boards and the leverage and debt maturity choices

Transcription

1 Int. J. Corporate Governance, Vol. 1, No. 1, Corporate boards and the leverage and debt maturity choices Jarrad Harford Foster School of Business, University of Washington, Box , Seattle, WA , USA jarrad@u.washington.edu Kai Li* Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC V6T 1Z2, Canada kai.li@sauder.ubc.ca *Corresponding author Xinlei Zhao Department of Finance, Kent State University Kent, OH 44242, USA xzhao@kent.edu Abstract: Debt, and in particular, short-term debt have the potential to discipline managers. We examine the role of the board in making financing decisions that provide this discipline. Specifically, given a firm s characteristics, we predict that stronger boards will force the firm to hold more debt and more short-term debt. Employing a rich dataset of board characteristics and controlling for other aspects of a firm s corporate governance, we find support for these hypotheses. Our simple measure of director power is a robust and promising measure of internal governance. Keywords: leverage; debt maturity; board effectiveness; incentive alignment; director power; governance. Reference to this paper should be made as follows: Harford, J., Li, K. and Zhao, X. (2008) Corporate boards and the leverage and debt maturity choices, Int. J. Corporate Governance, Vol. 1, No. 1, pp Biographical notes: Jarrad Harford is the Marion Ingersoll Endowed Professor of Finance at the Foster School of Business, University of Washington. His main research interests are corporate finance, mergers and acquisitions, payout policy, corporate governance. Kai Li is the W.M. Young Professor of Finance at the Sauder School of Business, University of British Columbia in Vancouver, Canada. Her main interests include corporate finance and corporate governance. Copyright 2008 Inderscience Enterprises Ltd.

2 4 J. Harford et al. Xinlei Zhao is currently an Assistant Professor of Finance at Kent State University. Her primary teaching and research interests are in the fields of empirical corporate finance and investments. 1 Introduction The separation of ownership and control in modern corporations leads to conflicts of interest between managers and shareholders (Jensen and Meckling, 1976). Past research, including Grossman and Hart (1982), Jensen (1986), Stulz (1990), Hart and Moore (1995), Rajan and Winton (1995) and Stulz (2000), has suggested leverage and debt maturity structure as effective ways to mitigate this agency problem. The intuition is that leverage, and particularly short-term debt, can reduce discretionary funds and subject managers to the scrutiny of the financial market and the threat of default, effectively curbing self-serving behaviour by managers. However, both leverage and debt maturity structure choices are influenced by managers, and they are not expected to voluntarily make the shareholders preferred leverage and debt maturity choices to constrain themselves. 1 Out of self-interest, managers would prefer less debt and/or debt with longer maturity. Consistent with that hypothesis, Berger et al. (1997) provide evidence that entrenched managers prefer equity to debt. Datta et al. (2005) show that managers with fewer equity-based incentives tend to employ more long-term debt, and Benmelech (2006) find that entrenched managers are less likely to finance with short-term debt. Further, Denis and Mihov (2003) show that managers with low ownership prefer diffuse public debtholders instead of concentrated private lenders. Empirical evidence has demonstrated that the board has the power to set financing policy and that the board uses it. Both Klein (1998) and Güner et al. (2006) find evidence that the composition of the board or finance committee impacts the firm s financing choices. 2 While we do not expect the control of agency conflicts to be the primary consideration in setting the level and maturity of debt, we would expect that, on the margin, boards do not overlook this tool available to them. The hypothesis we test in this paper is that a strong board is associated with more debt and shorter debt maturity. Boards are multi-faceted. It is a challenging task to characterise boards potential for effectiveness in monitoring. The different aspects of a board may contribute to a firm s corporate governance environment differently, with some aspects providing more effective monitoring than others. Further, some board aspects may be complementary to leverage, while others may be substitutes. In order to have a thorough understanding of the effect of the board, as part of the firm s internal corporate governance, on capital structure and debt maturity choices, we first construct a broad index of board characteristics. We then further refine our aggregate index into four subindices to capture distinct aspects of board functionality: monitoring effectiveness, incentive alignment, director power, and shareholder power. 3 Using a sample of S&P 1500 firms over the period and the panel data fixed-effects estimation, we show that our board index is not significantly related to leverage but is negatively associated with debt maturity. Among the four subindices, greater director power is the only subindex that is positively related to leverage and negatively related to debt maturity. The positive relation between director power and

3 Corporate boards and the leverage and debt maturity choices 5 leverage is offset by the negative relation between incentive alignment and leverage, leading to a lack of relation between the board index and leverage. Although director incentive alignment is negatively related to leverage, suggesting a possible substitute effect, this effect is not robust because of a lack of relation between incentive alignment and debt maturity. These results are obtained after controlling for an extensive list of other governance mechanisms in a firm. This finding suggests that director power is the most important characteristic for a board to have in order to be effective in implementing capital structure decisions that discipline managers. This paper contributes to the existing literature along the following dimensions. First, to the best of our knowledge, this study is the first in the literature to examine the role of corporate boards in debt maturity choices. Further, we control for a rich set of corporate governance mechanisms, both internal and external, to isolate the effect of the board on capital structure decisions. We also make a methodological contribution by constructing a broad board index to explore the role of corporate boards in firms financing decisions. Prior studies of corporate boards focus on at most a few aspects of the board, without being able to consider all of the dimensions of the board and the channels through which it becomes effective. The new dataset used here allows us to simultaneously examine a broad list of board attributes and assess which of them has the largest impact on the board s ability to influence financing decisions. The plan of the paper is as follows. We review the literature and develop our hypotheses in the next section. Section 3 describes our sample and variables and provides descriptive statistics. Section 4 presents the empirical results on the role of corporate boards in leverage and debt maturity structure choices and offers interpretations. Section 5 implements some additional investigation and Section 6 concludes. 2 Literature review and hypothesis development 2.1 Literature review The recent corporate scandals and the subsequent changes in securities laws and exchange rules highlight the importance of corporate boards in the mindset of regulators and the general public. Nonetheless, empirical evidence on the effectiveness of boards in corporate decisions and performance has been mixed. On the one hand, Baysinger and Butler (1985) find a positive relationship between board composition and performance, but there is no relation between changes in composition and changes in performance. Yermack (1996) shows a significant negative association between board size and firm value. Bhagat et al. (1999) show that higher dollar value of equity holdings by directors is associated with better corporate performance and a higher likelihood of managerial replacement after poor results. Adams and Ferreira (2004) show that board attendance is higher in the presence of higher meeting fees and that firms paying mostly a flat retainer overpay their directors on average. Adams et al. (2005) show that firms with powerful CEOs experience more variability in performance. Their measures of CEO power include: whether the CEO is also a founder, or the only insider on board, or also the chairman of the board. On the other hand, Hermalin and Weisbach (1991) show that there is no strong relationship between the fraction of outsiders on the board and firm performance. Brickley et al. (1997) do not find evidence that separation of CEO and chairman is

4 6 J. Harford et al. associated with higher market or operational performance. Core et al. (1999) find that outside directors are not more effective than inside directors in setting CEO compensation and conclude that the emphasis on directors and institutional ownership has been misplaced. Bhagat and Black (2002) show that low-profitability firms increase the independence of their boards, but firms with more independent boards do not perform better than other firms. Fich and Shivdasani (2006) show that firms in which a majority of outside directors hold three or more board seats have significantly lower valuation than firms in which a majority of outside directors hold fewer than three board seats. In summary, the empirical literature examining whether board size, composition, leadership, and incentives have real effects on firm performance and managerial incentives shows mixed results, and these studies motivate our adoption of ten board characteristics to broadly measure the effectiveness of the board. Given that boards meet infrequently, the monitoring role of the board is most likely to be detectable in specific, discrete corporate decisions, rather than in the day-to-day operations that contribute to long-run stock and operating performance. Indeed, the existing literature does deliver a much clearer message when examining the effect of the boards on top management turnover and mergers and acquisitions. Starting with Weisbach (1988), many papers have found boards dominated by independent directors to be more likely to make decisions that are in the interest of shareholders. 4 Our paper is most closely related to the literature examining the relation between corporate governance and firm capital structure decisions. Berger et al. (1997) study associations between managerial entrenchment and firms capital structures, and their results suggest that entrenched CEOs seek to avoid debt. Garvey and Hanka (1999) find that firms protected by second generation state antitakeover laws substantially reduce their use of debt, and that unprotected firms do the reverse. Datta et al. (2005) conclude that managerial ownership is effective in forcing firms to have shorter debt maturity. Benmelech (2006) finds that entrenched managers prefer long-term debt. 2.2 Our hypotheses The literature has suggested the following channels through which debt exerts its disciplinary role on managers. Grossman and Hart (1982) point out that since higher levels of debt increase the threat of bankruptcy which put managers under pressure due to the potential loss of control of their firms, higher levels of debt induce managers to avoid value-decreasing corporate decisions. Jensen (1986) further suggests that the fixed payments associated with debt reduce a firm s free cash flow, and effectively limit management s ability to waste corporate resources to benefit themselves. Moreover, the issuance of external debt also results in monitoring by bondholders, other lenders, investment bankers, and bond rating agencies. Other work has focused on the maturity structure of debt in addition to the level of debt to discipline managers. Easterbrook (1984), while primarily writing about dividends, notes that when a firm has to frequently issue new debt, its managers are reevaluated by capital providers at each issuance, increasing monitoring. Rajan and Winton (1995) point out that short-term loans give the bank unlimited power to act, while long-term loans with covenants only allow the bank to act if a covenant has been violated (based on verifiable information). They conclude that loans that are payable on demand or have short fixed maturities give lending institutions greater flexibility and control. Benmelech (2006) develops a model that endogenises debt maturity choice by self-interested

5 Corporate boards and the leverage and debt maturity choices 7 managers and predicts that entrenched managers will finance with long-term debt to avoid liquidation. Finally, DeAngelo et al. (2002) provide a case study illustrating the importance of short-term debt in properly constraining managers. A priori, capital structure and board strength could be substitutes or complements in controlling agency costs. They would be substitutes if managers choose a restrictive capital structure (higher leverage and shorter maturity) to bond themselves and establish a reputation with the capital markets in the absence of strong internal governance. Conversely, if restrictive capital structures are the outcome of strong boards imposing constraints on managers, then capital structure and board strength would be found to be complements. Previous work on the amount of leverage, by Berger et al. (1997) casts doubt on the substitutes view. Further, more restrictive capital structure has the advantage of utilising scrutiny by the capital markets to help the board (with limited time and resources) monitor managers. Thus, we reason that strong boards force managers to choose a level and maturity structure of debt that facilitates more (frequent) monitoring. Hence, we test the following hypotheses: H1: There is a positive relation between strong boards and the level of leverage H2: There is an inverse relation between strong boards and debt maturity. Specifically, these hypotheses predict that, ceteris paribus, our empirical measures of the strength, incentives, and effectiveness of the board will be positively related to the level of debt and negatively related to the maturity of debt in a firm. We predict that strong boards will be overall complements of restrictive debt policy. However, we admit the possibility that some components of board strength could be substitutes with debt policy and others could be complements, such that when we examine the individual subindices, the relation could go either way. Whether a specific board component is a complement or substitute of debt policy is an empirical issue. In addition to the agency conflict between managers and shareholders (represented by the board), a conflict of interest exists between shareholders and bondholders. Bondholders would prefer managers to undertake low-risk projects. Managers, due to their overexposure to firm-specific risk, would also prefer a more conservative investment profile and capital structure than would shareholders. Thus, when shareholders take actions to better control managers, bringing the firm s risk profile into line with shareholders preferences, the conflict between shareholders and bondholders is greater than in the case where managers are unconstrained. Klock et al. (2005) and Cremers et al. (2007) provide a similar argument that weaker shareholder control of managers can actually be preferred by bondholders and provide evidence from bond yields consistent with this argument. Bondholders concern about risk-shifting is largest for long-term debt because it limits their ability to reevaluate management. Thus, they are more likely to offer short-term debt to firms whose managers are closely monitored by a strong board. This effect reinforces the prediction in H2. 3 Sample formation and variable construction Our empirical design is to relate measures of corporate boards to leverage and debt maturity structure choices. We obtain detailed board information for S&P 1500 firms (including S&P 500, S&P Midcap 400, and S&P Small Cap 600) from the Investor

6 8 J. Harford et al. Responsibility Research Centre (IRRC), now part of Institutional Shareholder Services. The ISS/IRRC board data, our initial sample, covers the S&P 1500 firms for the proxy season, and thus our final sample will more likely be large firms. Following the prior literature, we restrict our sample to industrial firms excluding utilities and financials. Our initial sample is an intersection of the ISS/IRRC board data, ExecuComp for managerial ownership, Compustat for accounting information, and CRSP for stock prices for the period This leads to an initial sample of 6,145 observations. In order to isolate the effect of the board, we need to control for other governance mechanisms in addition to managerial ownership. Gompers et al. (2003) construct an external governance measure Gindex that proxies for the level of shareholder rights. Litov (2005) shows that firms with weak shareholder rights actually use more debt financing and have higher leverage ratios, and Benmelech (2006) finds that more entrenched managers tend to employ long-term debt. To make sure that our results will not be affected by the extent of managerial entrenchment in the corporate control market, we control for a modification to the Gindex the Eindex in our main analysis. Bebchuk et al. (2005) show that only six of the 24 provisions in the Gindex are significantly associated with firm value. The Eindex is not available on a continuous basis. For years where the Eindex data is not available, we follow the common practice of filling in the missing years with the prior data (e.g., assuming that 1999 has 1998 values, 2001 has 2000 values, and 2003 has 2002 values). Another important corporate governance measure is the presence of institutional investors. We obtain institutional shareholding information from Thompson Financial s 13F filings. We merge our initial sample with the data on the Eindex and the institutional shareholding. We employ two leverage measures. The book leverage ratio is defined as the ratio of total debt to total assets, and the market leverage ratio is the ratio of total debt to the market value of total assets, obtained as total assets minus book equity plus market value of equity. Following Barclay and Smith (1995), we measure debt maturity as the proportion of total debt maturing in more than T years, where T ranges from 1 (debt1) to 5 (debt5). We discard any firm-year observation where the book leverage and debt maturity ratio variables are less than 0% or more than 100%. And our final sample has 5,825 firm-year observations representing 1,130 unique firms with no missing data on all variables used in our multivariate analysis. Panel A of Table 1 provides descriptive statistics on leverage and debt maturity structure for our sample of firms. We note a large variation in leverage among our sample firms, with average at 17% (23%) and corresponding standard deviation at 15% (16%) for the market (book) leverage measure. Our market leverage measure is of similar magnitude as the one in Datta et al. (2005), and our book leverage measure is quite close to the leverage number in Berger et al. (1997) even though our sample and sample period do not overlap with Berger et al. (1997) and only overlap with the Datta et al. sample by a few years. This suggests that among large US firms, leverage ratios on average have been quite stable over the past 20 years, consistent with the conclusions of Lemmon et al. (2007). 5 Over 90% of the debt of a median firm in our sample matures in more than one year. The median proportion of debt maturing in more than three and five years is 70% and 47%, respectively. These numbers are much higher than similar numbers reported in Barclay and Smith (1995), but not much higher than those in Datta et al. (2005), suggesting that firms have lengthened their debt maturity in the late 1990s.

7 Corporate boards and the leverage and debt maturity choices 9 Table 1 Sample characteristics Panel A: Leverage and debt maturity Mean StdDev 25th percentile Median 75th percentile Market leverage Book leverage Debt Debt Debt Debt Debt Panel B: Firm characteristics Mean StdDev 25th percentile Median 75th percentile Firm size Tangibility M/B ratio Profitability Asset maturity Term structure Asset return StdDev Proportion of firms paying dividend 60.01% Proportion of firms with rating 54.06% Proportion of firms with investment grade rating 34.99% This table presents summary statistics of leverage, debt maturity, and firm characteristics for our sample firms. Our sample is an intersection of the ISS/IRRC board data, Gindex data, ExecuComp, Compustat, and CRSP for the period Our sample has 5,825 firm-year observations representing 1,130 unique firms. Market leverage is the ratio of total debt (data34 + data9) to market value of total assets which is computed as the sum of market value of equity (data199 data54) and book value of debt minus book value of equity. Book leverage is the ratio of total debt to total assets (data6). Debt1-Debt5 are the proportion of total debt maturing in one to five years, respectively. We drop observations where the leverage ratio and debt maturity ratio variables are outside the (0, 1) range. Firm size is sales (data12) in millions of 2004 dollars. Asset tangibility is the ratio of net property, plant, and equipment (data8) to total assets. M/B ratio is the ratio of market value of total assets to book value of total assets. Profitability is operating income before depreciation (data13) normalised by the lagged value of total assets. Asset maturity is the weighted average of current assets (data4) divided by the cost of good sold (data41), and net property, plant, and equipment divided by depreciation and amortisation (data14), following Datta et al. (2005). Term structure is the yield difference between 10-year T-bond and 3-month T-bill. Asset return StdDev is stock return standard deviation times the equity-to-firm value ratio following Barclay and Smith (1995). Panel A presents summary statistics of capital structure choices, and Panel B presents firm characteristics.

8 10 J. Harford et al. Panel B of Table 1 documents sample firm characteristics. Our sample firms are quite large, with median sales (in 2004 $) at 1.6 billion. A median firm in our sample has 26% of their assets in tangible assets, and an M/B ratio of 1.6. We see substantial variations across our sample firms in terms of operating income, asset maturity, and asset return standard deviation. Sixty percent of our sample firms pay dividends, and over half of them have bond ratings, with most of them obtaining an investment grade rating. Dividend payers have greater representation in our sample than as reported for all public firms in Fama and French (2001), due to the fact that our sample is skewed toward large firms. 3.1 The board index In this paper, we ask whether and how boards affect corporate financing decisions. Departing from existing work, we first construct an aggregate board index that captures different facets of the board, and then study the empirical relationship between this index and firms financing choices. The ISS/IRRC board data covers director identity and affiliation, director compensation and equity ownership, director tenure, committee membership, whether the CEO is the chairperson of the board, whether cumulative voting is allowed, and whether there is a provision that staggers the terms and elections of directors. The unique director identifier within the dataset allows us to compute board size, director/board independence, and how many board seats are held by a specific director. Based on the prior theoretical and empirical work summarised in the previous section, we single out ten board characteristics to be included in our board index construction (see detailed definitions in the Appendix). They are small board, not busy (board), board independence, no interlock, (board) equity ownership, (board) meeting fee, CEO not Chairperson of the Board (COB), blockholder on board, cumulative voting, and no classified board. Among the ISS/IRRC sample of firms, the median board size is nine, the median director equity ownership is $ 0.54 millions, the median board meeting fee is $ 1134, and the median number of board meetings per year is seven. Our board index is constructed to proxy for the balance of power between the board of directors and managers. We calculate our index for each firm-year in a straightforward manner: we add one point for every board measure that increases the likelihood that the board will be an effective monitor of managers in shareholders interests. Thus, the board index ( Bindex ) is just the sum of the points awarded for the existence of each board characteristic. 6 We also construct subindices for characteristics in each of the following four categories: Board Effectiveness (BEindex), Incentive Alignment (IAindex), Director Power (DPindex), and Shareholder Power (SPindex). By construction, the four subindices sum to the board index (see our Appendix for board characteristics included in the subindices). The maximum possible number for the Bindex is 10, and the maximum possible values for the BEindex, IAindex, DPindex, and SPindex are 2, 4, 2, and 2, respectively. The higher the values for the Bindex and the four subindices, the better is the quality of a firm s corporate governance.

9 Corporate boards and the leverage and debt maturity choices 11 In Table 2, Panel A presents the summary statistics of our Bindex and the four subindices, and Panel B presents the descriptive statistics for each of the ten board characteristics included in the Bindex. The median firm in our sample scores 5 out of 10 in the Bindex, with the mean score at 5.2. One focus of governance reform in the last decade has been the independence and incentives of the directors. Our sample firms score quite well on the dimension of incentive alignment of the board, with the majority of firms in our sample scoring 3 out of a possible maximum value of 4 in the IAindex. In particular, as shown in Panel B, 76% of the sample firms have a board with the majority of the directors being independent, and over 90% of our sample firms have a board free of directors with interlocking relationships. By contrast, most of our sample firms do not score well in terms of board effectiveness, director power, and shareholder power in particular. For instance, the majority of sample firms only score 1 out of 2 in the DPindex. The data as shown in Panel B confirms that in most firms, CEOs are also COBs. Only about half of the sample firms have blockholders sitting on the board. Table 2 Corporate governance characteristics Panel A: Our board indices Mean StdDev 25th percentile Median 75th percentile Bindex BEindex IAindex DPindex SPindex Panel B: Board characteristics Percent of firms with that feature Board effectiveness (BEindex) Small board 38.1 Not busy 91.2 Incentive alignment (IAindex) Board independence 76.3 No interlock 91.4 Equity ownership 49.1 Meeting fee 48.2 Director power (DPindex) CEO not COB 27.7 Blockholder on board 53.4 Shareholder power (SPindex) Cumulative voting 10.4 No classified board 38.6

10 12 J. Harford et al. Table 2 Corporate governance characteristics (continued) Panel C: Other governance mechanisms Mean StdDev 25th percentile Median 75th percentile Eindex Public pension holding Executive ownership This table presents summary statistics of corporate governance characteristics for our sample firms. Please refer to our Appendix for the definitions of the Bindex, BEindex, IAindex, DPindex, SPindex, and their corresponding constituents. The Eindex is defined based on Bebchuk et al. (2005) covering staggered boards, limits to shareholder bylaw amendments, supermajority requirements for mergers, supermajority requirement for charter amendments, poison pills, and golden parachutes. Public pension holding is defined as aggregate holdings by active public pensions as identified in Cremers and Nair (2005), normalised by the number of shares outstanding. Executive ownership is dollar value of common stock and restricted shares held by the top five executives in millions of 2004 dollars. Panel A presents summary statistics of our board indices, Panel B presents summary statistics of our ten board characteristics, and Panel C presents summary statistics of other governance mechanisms. 3.2 Other governance mechanisms A variety of firm-level mechanisms are associated with the governance of the public corporations. Other than the board measures we have constructed earlier, we also control for other governance mechanisms in order to isolate the effect of boards on corporate financing decisions. These firm-level mechanisms can be broadly grouped into two categories external and internal governance mechanisms. Takeovers and large shareholders are often seen as the primary external governance mechanisms, while incentive pay and the board of directors are the main internal mechanisms. To measure managerial entrenchment from an external governance point of view, we use the Eindex recently proposed by Bebchuk et al. (2005), which is based on a count of six provisions that reduce shareholder rights. They are staggered boards, limits to shareholder bylaw amendments, supermajority requirements for mergers, supermajority requirement for charter amendments, poison pills, and golden parachutes. We expect that firms with strong shareholder rights (lower value of the Eindex) are more likely to employ debt and ceteris paribus, more short-term debt. 7 Shleifer and Vishny (1986) suggest that large shareholders have incentive to monitor the management. Research by Del Guercio and Hawkins (1999) and Cremers and Nair (2005) show that public pension funds generally have fewer conflicts of interest and less corporate pressure than other institutional shareholders do. To measure this aspect of the governance of a firm, we construct the aggregate holdings by active public pensions as identified in Cremers and Nair (2005), normalised by the number of shares outstanding, Since Jensen and Meckling (1976), many studies have shown that managers will have more powerful incentives to make value-maximising corporate decisions when their stock ownership is high. Berger et al. (1997) show that firms whose CEOs have greater equity ownership are associated with the use of higher leverage. Denis and Mihov (2003) find that higher managerial ownership is associated with the use of more scrutinising private debt, and Datta et al. (2005) conclude that managerial ownership is negatively related to

11 Corporate boards and the leverage and debt maturity choices 13 debt maturity. We use the dollar value of stock ownership for the top five executives (inferences are unchanged if we focus on the CEO) as a measure of their cumulative equity-based compensation to capture the degree of incentive alignment between managers and shareholders. Note that our dollar value measure of managerial equity incentives is different from past work and is motivated by the following considerations. As Hall and Liebman (1998) argue, for large and diversely-held companies, a top executive might only hold a very small fraction of the firm s shares while those shares, in terms of dollar values, could be a significant portion of the executive s personal wealth, and hence leading to an alignment of interests between the executive and the shareholders. Measuring fractional ownership is best-suited for capturing incentives to over-consume on perquisites because the manager only bares a fraction of the costs. Using a dollar value measure captures the amount of wealth she has at risk in the firm and is an appropriate measure of incentive alignment on major corporate decisions such as financing policies. 8 Our definition of managerial stock ownership includes common stock and restricted shares and is scaled by the total number of shares outstanding. And our dollar value equity incentive is the product of managerial equity ownership and market capitalisation, measured in millions of 2004 dollars. Panel C of Table 2 presents the summary statistics of other corporate governance measures. The median sample firm scores 2 out of 6 in the Eindex. The median level of public pension fund holding in our sample firms is about 3%, similar to the number reported in Cremers and Nair (2005). The median (mean) dollar value of executive ownership of company shares is about $5 ($24) millions in 2004 $, while the median fractional executive ownership is about 1% among our sample firms (unreported), similar to the number reported in Datta et al. (2005) even though our sample periods only overlap by a few years. The low equity ownership by top executives in our sample is due to the fact that our sample includes the largest companies in the USA. Table 3 presents measures of the time-series variation in the main board index, its four subindices, and its ten board characteristic constituents. We first compute the average of year-to-year changes in the specific board index or characteristic for each firm. We then calculate the summary statistics for the cross-sectional sample of average annual changes for that specific measure. Panel A shows that there are minimal changes in the BEindex and SPindex, with the majority of firms experiencing no changes from year to year. There is relatively more temporal variation in the IAindex and DPindex. As a result, there is moderate temporal variation in the Bindex. Even though for each individual characteristic, there is a low probability of change in any particular year, that does not mean that the board characteristics are static. We construct another measure, change frequency, capturing the frequency with which any one or more of the board characteristics changes from year to year. For example, in each of the seven years, a different board characteristic could change. This is clearly a board in flux, but each individual characteristic would show a median change of zero from year to year. Our new measure would capture it as a median change of one, normalising the seven individual yearly changes by the maximum number of possible changes (7). The last row in Panel B presents the summary statistics of this new measure. It shows that on average, about half of the boards experience at least one change in their characteristics per year. Our evidence is consistent with Denis and Sarin (1999) who show that 65% of their sample firms exhibit large changes in ownership and board structure in any given year over the ten-year period they examined.

12 14 J. Harford et al. Table 3 Temporal variation in board indices and board characteristics Panel A: Time variation in our board indices Mean StdDev 25th percentile Median 75th percentile Bindex BEindex IAindex DPindex SPindex Panel B: Time variation in board characteristics Mean StdDev 25th percentile Median 75th percentile Board effectiveness (BEindex) Small board Not busy Incentive alignment (IAindex) Board independence No interlock Equity ownership Meeting fee Director power (DPindex) CEO not COB Blockholder on board Shareholder power (SPindex) Cumulative voting No classified board Change frequency This table presents the temporal variations in board indices and board characteristics. Please refer to our Appendix for the definitions of the Bindex, BEindex, IAindex, DPindex, SPindex, and their corresponding constituents. We first compute the average of year-to-year change in board indices and board characteristics for each firm. We then calculate the summary statistics for the cross-sectional sample of average annual changes. Change frequency refers to the frequency with which any board characteristic changes from year to year. This captures the probability that in any given year, some aspect of the board will change. Panel A presents the summary statistics of year-to-year changes in our board indices, and Panel B presents the summary statistics of year-to-year changes in board characteristics. Table 4, Panel A presents the correlation between leverage, debt maturity, and all our governance measures. Contrary to our hypothesis for debt levels and to prior literature, the results indicate significantly negative correlations between measures of leverage and strong governance measured by our board index. Nonetheless, the correlations between measures of debt maturity and our board index are significantly negative as predicted. However, omitted variable bias in univariate correlations can mask the true relationships between the variables. As a result, we later employ multiple regressions in order to detect causal relationships between leverage, maturity structure, and board measures after controlling for an extensive list of potential determinants of leverage and maturity structure choices.

13 Corporate boards and the leverage and debt maturity choices 15 Table 4 Simple correlation between corporate governance measures

14 16 J. Harford et al. Table 4 Simple correlation between corporate governance measures (continued)

15 Corporate boards and the leverage and debt maturity choices 17 Panel A also shows that correlations among the governance measures are mixed. The board subindices are, in general, positively, but weakly correlated. Our Bindex and subindex SPindex are strongly and negatively correlated with the Eindex. While both the BEindex and DPindex are significantly negatively correlated with Eindex, the correlations are 10% or lower. Executive ownership is negatively correlated with the Eindex and public pension holding. Table 4, Panel B presents the correlation between the ten board characteristics that we examine. Many of the correlations are sensible. A small board is more likely to have directors that do not sit on multiple boards, and less likely to be an independent board. Non-busy directors are more likely to have high equity ownership and are positively associated with the presence of blockholders on the board. In contrast, independent boards are associated with low equity ownership and the absence of blockholders on the board. The median director equity ownership is strongly and positively correlated with the presence of blockholders on the board. 4 Results on board monitoring 4.1 Model specification To test our hypotheses that there is a positive relation between our board constructs and the level of leverage, and a negative relation between our board constructs and debt maturity in a multivariate framework, we use two-stage least squares regression analysis as in Barclay et al. (2003) and Datta et al. (2005). We model total leverage in the first stage and then model the maturity structure in the second stage including our fitted leverage estimate as an explanatory variable. The first-stage regression model is Leverageit = α0 + β1board Measureit + β2gindexit + β3public Pension Holdingit + β4executive Ownershipit + β5firm Sizeit + β6asset Tangibilityit + β M / B Ratio + β Profitability + β Dividend Paying Dummy + e, 7 it 8 it 9 it it where the market leverage ratio is the dependent variable, and Board Measure is one or more of our board indices. Motivated by prior findings in Litov (2005), Cremers and Nair (2005) and Datta et al. (2005), we also include alternative governance mechanisms in the above regression. Firm characteristics used to determine leverage in the first stage are firm size, asset tangibility, M/B asset ratio, profitability, and the dividend-paying dummy. These variables are defined in Table 1 and are motivated by the empirical findings in Berger et al. (1997), Barclay et al. (2003) and Frank and Goyal (2007), among others. We employ two different sets of board measures when reporting our estimation results. In Model 1, we use the overall board index, namely, Bindex. In Model 2, we replace the Bindex with our four board subindices. In the second-stage regression, we control for leverage using the predicted leverage from the first-stage regression. 9 To properly identify the system of equations, in the second-stage regression, we exclude asset tangibility, profitability, and dividend-paying dummy, 10 and add asset maturity, a term structure measure, asset return volatility, a rating dummy, and an investment grade dummy following Barclay and Smith (1995), (1)

16 18 J. Harford et al. Guedes and Opler (1996) and Stohs and Mauer (1996). Specifically, we estimate the following model for debt maturity in the second stage: Debt Maturity = α + β Board Measure + β Gindex + β Public Pension Holding + β4executive Ownershipit + β5fitted Leverageit + β6firm Sizeit + β7m / B Ratioit + β8asset Maturityit + β9term Structuret + β10asset Return StdDevit + β11bond Rating Dummyit + β Investment Grade Dummy + e. it 0 1 it 2 it 3 it 12 Again, we employ two different sets of board measures when reporting our estimation results. Given that the data is an unbalanced panel, we employ the panel data estimator with fixed effects. The fixed-effects model relies on time series variation to identify the relation between board measures and financing choices. Hence it captures the dynamic and causal effect of boards on capital structure. In contrast, the commonly used cross-sectional regression is only able to identify (static) correlation but fails to establish (inter-temporal) causality. Moreover, the fixed-effects approach is robust to the presence of omitted time-invariant firm-specific variables (including the industry effect) that would lead to biased estimates in an ordinary least squares framework. Finally, Lemmon et al. (2007) present evidence of the persistence in capital structures and recommend the use of fixed-effects estimation in capital structure studies. The reported P-values are based on White heteroscedasticity-consistent standard errors adjusted to account for possible correlation within a (firm) cluster. We report results from the first-stage leverage regression in the first two columns of Table 5. Under Model 1, we find that the Bindex is not significantly related to the level of leverage. Under Model 2, we find that the DPindex is significantly and positively related to the level of leverage. The economic implication of this result is that an increase in the DPindex from the 25th percentile to the 75th percentile is expected to increase leverage by 0.6%. On the other hand, the IAindex is significantly and negatively related to leverage, suggesting a possible substitute effect between incentive alignment and leverage. 11 In addition, there are no significant associations between leverage and the other two subindices: the BEindex and SPindex. Columns 3 and 4 report the second-stage maturity regression results. The dependent variable here is the proportion of total debt maturing in more than three years (debt3). 12 Under Model 1, we find that the overall board index, Bindex, is significantly negatively related to debt maturity. The economic implication of this result is that an increase in the Bindex from the 25th percentile to the 75th percentile is expected to reduce the percentage of total debt maturing in three years or more by 1.6%, relative to an unconditional average of 63%. This finding lends support to our second hypothesis that a strong board is associated with the use of more short-term debt. Under Model 2, among our four board subindices, only the DPindex is found to be significantly negatively associated with debt maturity. The economic implication of this result is that an increase in the DPindex from the 25th percentile to the 75th percentile is expected to reduce the percentage of total debt maturing in three years or more by 1.9%. The IAindex is not significantly related to debt maturity, inconsistent with the substitute argument, even though it is significantly negatively associated with leverage. Therefore, we conclude that there is no strong evidence in support of the substitute effect. it it (2)

17 Corporate boards and the leverage and debt maturity choices 19 In other words, the DPindex is the only board subindex that shows a consistent relation with leverage and debt maturity, suggesting that director power relative to the CEO is the most important characteristic of the board in exerting their monitoring influence. Table 5 Leverage and maturity regressions Leverage regressions Debt maturity regressions Model 1 Model 2 Model 1 Model 2 Bindex * (0.304) (0.089) BEindex (0.482) (0.344) IAindex 0.007*** (0.000) (0.799) DPindex 0.006** 0.019** (0.015) (0.040) SPindex (0.419) (0.286) Eindex (0.434) (0.486) (0.989) (0.925) Public pension holding (0.137) (0.131) (0.153) (0.154) Executive ownership 0.008*** 0.008*** 0.014** 0.015*** (0.000) (0.000) (0.011) (0.010) Fitted leverage (0.394) (0.434) Firm size 0.034*** 0.036*** (0.000) (0.000) (0.451) (0.385) Asset tangibility (0.378) (0.365) M/B ratio 0.017*** 0.017*** (0.262) (0.302) (0.000) (0.000) Profitability 0.338*** 0.343*** (0.000) (0.000) Dividend-paying dummy 0.019*** 0.018*** (0.001) (0.002) Asset maturity 0.003* 0.003* (0.075) (0.072)

18 20 J. Harford et al. Table 5 Leverage and maturity regressions (continued) Leverage regressions Debt maturity regressions Model 1 Model 2 Model 1 Model 2 Term structure 0.013*** 0.013*** [0.001] [0.001] Asset return StdDev (0.110) (0.123) Bond rating dummy 0.109*** 0.110*** (0.003) (0.003) Investment grade dummy (0.589) (0.598) Intercept *** 0.818*** (0.884) (0.721) (0.000) (0.000) Number of observations R-squared Within Between Overall *, **,*** denote significance at the 10%, 5% and 1% levels, respectively. This table presents fixed-effects regression results of leverage and Debt3, which is the proportion of total debt maturing in more than three years. We use the market leverage as the dependent variable and it is the ratio of total debt (data34 + data9) to market value of total assets which is computed as the sum of market value of equity (data199 data54) and book value of debt minus book value of equity. Please refer to our Appendix for the definitions of the Bindex, BEindex, IAindex, DPindex, and SPindex. The Eindex is defined based on Bebchuk et al. (2005). Public pension holding is defined as aggregate holdings by active public pensions as identified in Cremers and Nair (2005), normalised by the number of shares outstanding. Executive ownership is dollar value of common stock and restricted shares held by the top five executives in millions of 2004 dollars. Fitted leverage is the fitted market total leverage from the first-stage leverage regression in the table. Firm size is the natural logarithm of sales (data 12) in millions of 2004 dollars. Asset tangibility is the ratio of net property, plant, and equipment (data8) to total assets. M/B ratio is the ratio of market value of total assets to book value of total assets. Profitability is operating income before depreciation (data13) normalised by the lagged value of total assets. Dividend-paying dummy equals one for firms paying dividends, and zero otherwise. Asset maturity is the weighted average of current assets (data4) divided by the cost of good sold (data41), and net property, plant, and equipment divided by depreciation and amortisation (data14), following Datta et al. (2005). Term structure is the yield difference between 10-year T-bond and 3-month T-bill. Asset return StdDev is stock return standard deviation times the equity-to-firm value ratio following Barclay and Smith (1995). Bond rating dummy equals one for firms with long-term bond rating available, and zero otherwise. Investment grade dummy equals one for firms with investment grade bond rating, and zero otherwise. Table 5 also presents the estimation results on the relationship between other corporate governance measures and leverage and debt maturity choices. We find that the Eindex is not significantly related to leverage once we account for the role of corporate boards in firm financing decisions. 13 Public pension holding is not significantly positively

19 Corporate boards and the leverage and debt maturity choices 21 associated with leverage, contradicting the prevailing view that these institutional investors are active monitors (Cremers and Nair, 2005). On the other hand, the negative relation between pension fund presence and leverage seems to suggest that due to their fiduciary duties, public pension funds tend to invest in firms with lower financial risk. Executive ownership is positively associated with the level of leverage, consistent with the findings in Berger et al. (1997); however, it is significantly and positively associated with debt maturity, in contrast to Datta et al. s (2005) findings without including board characteristics. 14 Further, neither the Eindex nor public pension holding is significantly associated with debt maturity. Finally, Table 5 shows the relation between other firm characteristics and leverage or debt maturity choices. There results are quite consistent with those from earlier studies. In particular, large less profitable non-dividend paying firms with lower growth opportunities are more likely to have higher leverage. Further, firms with shorter asset maturity are more likely to use short-term debt. The term structure is significantly positively related to the use of long-term debt, consistent with the tax hypothesis put forward by Brick and Ravid (1985): when the yield curve is upward sloping, issuing long-term debt reduces firms expected tax liability and consequently increases their current market value. Firms with a bond rating use more short-term debt, suggesting that firms with lower liquidation risk tend to shorten their debt maturity. This finding is different from Barclay and Smith (1995), but is consistent with the nonmonotonic relation between debt maturity and bond rating predicted by Diamond (1991) and confirmed by Guedes and Opler (1996), and Stohs and Mauer (1996). One caveat to our estimation results regarding the firm specific control variables is that several of these firm characteristics, such as the presence of bond rating, or asset maturity, do not change much from year to year, while our estimation technique the panel data fixed-effects estimator relies heavily on the time series variations in these control variables. 5 Additional investigation We discuss additional investigation we have done in this section. 5.1 Alternative samples So far, our sample excludes financial firms and utilities as is standard in the literature. We also conduct our main analysis using two other different samples. Our second sample follows: Barclay and Smith (1995) and Datta et al. (2005) to focus on industrial firms only (SIC ), leaving 6,445 firm-year observations. Our final sample does not impose any industry restrictions, yielding 9,086 firm-year observations. We find that these samples give very similar results as our main sample. We conclude that the effect of boards on capital structure choice is robust among firms of both unregulated and regulated industries. 5.2 CEO age It is possible that older CEOs will have risk preferences that lead them to choose more conservative capital structures. We add the CEO age variable to our main specifications and find that CEO age is only weakly associated with short debt maturity, and has no