Large Shareholder Turnover and CEO Compensation

Transcription

1 Large Shareholder Turnover and CEO Compensation Kyonghee Kim University of Pittsburgh, Pittsburgh PA February 2005 Abstract The corporate governance literature generally assumes that shareholders incentives to monitor management depend on how much of the firm the shareholders own. Large shareholders have a strong incentive to monitor management while shareholders who own a small number of shares have proportionally little incentive to monitor. This study proposes that another determinant of monitoring incentive is how long large shareholders (blockholders of at least 5% of ownership stake) hold their shares, which can be measured by the turnover rates of their shares. For blockholders who hold their shares for only a short period (i.e., those with a high turnover rate), direct monitoring of management is unlikely to be cost-effective. Instead, I expect that such blockholders will rely on using incentive contracts to motivate managers to maximize shareholder value. In contrast, blockholders with lower turnover rates will find it worthwhile to invest significant time and effort in monitoring management. Therefore, they can impose less risk on managers, reducing the required risk premium they must pay the CEO. This study tests hypotheses based on the preceding expectations and finds that when blockholder turnover is higher, CEO compensation imposes more risk (i.e., higher pay-performance sensitivity) on managers, and CEO expected compensation is higher. The increased risk is mainly from equity-based incentives in total CEO compensation but not from earnings-based cash incentives. Further, I find that when a firm has blockholders who are active in monitoring management, their direct monitoring reduces the effect of blockholder turnover on CEO incentive compensation. Keywords: Ownership structure, blockholder turnover, investment horizons, information asymmetry, incentive contracts I am indebted to my chair, Harry Evans and to other members of my dissertation committee (Shijun Cheng, Kenneth Lehn, Nandu Nagarajan and Dhinu Srinivasan). I also thank Jacob Birnberg, Mei Feng, Vicky Hoffman, Donald Moser, Sukesh Patro, Sally Widener, Michael Williamson and the workshop participants at Katz Graduate School of Business for their valuable comments and suggestions. I gratefully acknowledge the generosity of Jennifer Dlugosz, Rudiger Fahlenbrach, Paul Gompers and Andrew Metrick for making their ownership data available.

2 1. Introduction Large shareholders play a significant role in corporate governance. Prior research (e.g., Grossman and Hart [1980]; Morck, Shleifer and Vishny [1988]) relates shareholders incentives to monitor management to the size of their stake in the firm. Specifically, the larger the fraction of ownership held by a shareholder, the stronger the incentive to monitor. Based on this relation, researchers examine how ownership concentration and large shareholder characteristics affect CEO compensation (Mehran [1995]; Core, Holthausen and Larcker [1999]; Hartzell and Starks [2003]) or firm performance (McConnell and Servaes [1990]; Himmelberg, Hubbard and Palia [1999]). This study introduces a measure of large shareholder 1 turnover as an additional determinant of shareholders monitoring incentives. Specifically, large shareholder turnover, defined as the proportion of outside block ownership that changes hands annually, reflects the investment horizons of the firm s outside blockholders. I examine how outside blockholder turnover helps to explain the level and structure of CEO compensation after controlling for the effect of ownership concentration and types of block ownership. I also demonstrate how outside blockholder turnover interacts with the remaining dimensions of ownership structure to increase or decrease CEO payperformance sensitivity and the level of CEO compensation. To effectively monitor top management, blockholders expend time and resources in information collection. Blockholders with higher rates of turnover; i.e., shorter investment horizons, have less opportunity to monitor and learn about the firm and its management. As a result, there is greater information asymmetry between blockholders 1 Following the SEC filing criterion, I define a large shareholder as one who owns at least 5% of a firm s outstanding voting stock. I use the terms large shareholders and blockholders interchangeably. 1

3 with shorter investment horizons (higher turnover) and management than for blockholders with longer investment horizons (lower turnover). Consequently, such high turnover blockholders are likely to delegate more discretion to management and then rely on outcome-based compensation to encourage the management to use this discretion to maximize firm value. To compensate for the additional risk associated with greater reliance on outcome-based payment, expected management compensation will also be higher when blockholder turnover is higher. This reasoning suggests that firms with higher blockholder turnover are likely to have greater pay-performance sensitivity and larger total CEO compensation. On the other hand, the presence of active blockholders who can mitigate this agency problem will reduce these firms reliance on outcome-based compensation and their needs to pay larger total CEO compensation. The following example illustrates how blockholder turnover can vary crosssectionally and influence CEO compensation. Table A compares Alexander & Baldwin Inc., a transportation firm, with Teradyne Inc., a business equipment manufacturer. The two firms are comparable in asset size and have roughly the same magnitude of block ownership (average block ownership is 27.83% for Alexander & Baldwin versus 27.72% for Teradyne). However, the two firms experience very different blockholder turnover. Alexander & Baldwin (A&B) demonstrates little change in blockholders from year to year, as reflected in the blockholder turnover measure of 2.2% 2. In contrast, the pattern for Teradyne reflects much more dynamic block ownership with most blockholders in any given year leaving by the following year, resulting in blockholder turnover of 78.32%. Previous ownership studies (e.g., Mehran [1995]; Hartzell and Starks [2003]) 2 For details of the blockholder turnover calculation, see Section 4.3 Measurement of Blockholder Turnover. 2

4 Table A An Example of Block Ownership, Blockholder Turnover and CEO Compensation Company name: Alexander & Baldwin Inc. Company name: Teradyne Inc. Year Blockholder Name Holding % Year Blockholder Name Holding % First Hawaian Bank Wenberg Foundation Southeastern Asset Mgmnt First Hawaian Bank Wenberg Foundation Southeastern Asset Mgmnt Capital Group Companies First Hawaian Bank Wenberg Foundation Southeastern Asset Mgmnt Capital Group Companies Alexander Waterhouse First Hawaian Bank Wenberg Foundation Southeastern Asset Mgmnt Capital Group Companies Alexander Waterhouse Capital Group Cos Inc FMR Corp Metropolitan Life Insurance Pioneering Management State Street Research & Mgmnt Capital Group Cos Inc Pioneering Management Scudder Kemper Investment Neuberger and Berman LLC FMR Corp Neuberger and Berman LLC Capital Group Cos Inc Marsh & McLennan Cos Inc Four year average statistics Industry: Transportation Average block-ownership: 27.83% Average blockholder turnover: 2.20% Firm size (Total assets): $1,723 million Total CEO compensation: $1,627,495 CEO salary/total compensation: 39.92% CEO options/total compensation: Four year average statistics Industry: Business equipment Average block-ownership: 27.72% Average blockholder turnover:: 78.32% Firm size (Total assets): $1,307 million Total CEO compensation: $2,186,685 CEO salary/total compensation: 20.47% CEO options/total compensation: 53.25% treat such firms as A&B and Teradyne as having similar ownership structures because the magnitude of block ownership is similar, and the blockholders are primarily outside investors. In contrast, my premise is that the two firms will potentially demonstrate significant differences in governance characteristics because they experience very different levels of blockholder turnover. In fact, the two firms are quite different in terms of CEO compensation level and structure (see the bottom of Table A). The level of Teradyne s CEO compensation is greater, and the CEO compensation is more sensitivity to performance than at A&B. Specifically, total CEO compensation is $2,186,685 at Teradyne versus $1,627,495 at 3

5 A&B. Regarding the structure of CEO compensation, Teradyne CEO s compensation is more in the form of stock options (53.25%) than fixed salary (20.47%) whereas A&B CEO compensation is more in the form of fixed salary (39.92%) than stock options (25.49%). This study examines whether the pattern of blockholder turnover and CEO compensation exhibited in this example holds for a large sample of firms. My results support the hypothesis that firms design CEO compensation contracts in the context of their monitoring environment to minimize agency costs. Specifically, firms with shorter-horizon outside blockholders are more likely to use incentives than direct monitoring to induce desirable actions of their CEOs, and pay larger total CEO compensation for the risk associated with the incentives. Further, when providing incentives (i.e., increasing pay-performance sensitivity), these firms tie CEO compensation to market performance via stock option grants rather than relying on cash compensation based on earnings which is more subject to management discretion. This interpretation is further supported by my finding that the effect of outside blockholder turnover on CEO compensation is greater (smaller) in the absence (presence) of active blockholder monitors. This paper contributes to the corporate governance literature in three ways. First, I consider the simultaneous influence of three dimensions of ownership ownership concentration, ownership type and owners investment horizons on CEO compensation contracts. My analysis addresses the call by Demsetz and Villalonga (2001) for studies that provide a richer representation of ownership structure in order to provide a more accurate picture of the governance role of ownership. Specifically, existing literature relates higher outside ownership to greater CEO pay-performance sensitivity with the 4

6 assumption that concentration of outside ownership leads to higher monitoring by the outside shareholders (Mehran [1995]; Core, Holthausen and Larcker [1999]; Talmor and Wallace [2000]; Hartzell and Starks [2003]). This study examines outside block ownership both in terms of its level of concentration and rate of turnover and demonstrates that it is the rate of turnover but not the level of outside block ownership that influences CEO compensation. This result is consistent with Parrino, Sias and Starks (2003) that institutional investors (outside investors) influence the boards by voting with their feet rather than by exercising direct monitoring. Second, prior studies (Potter [1992]; Bushee [1998]; Gaspar, Massa and Matos [2004]) suggest that investors short horizons can increase agency costs by reducing the effectiveness of shareholders monitoring of management. In contrast, my study suggests that investors horizons are chosen simultaneously with the presence or absence of active blockholder monitors and the design of CEO compensation. As a result, firms with shorthorizon shareholders do not necessarily suffer higher agency costs. Instead, my results suggest that compensation contract design and the presence of active blockholders mitigate the potential agency costs associated with weaker monitoring by short-horizon blockholders. Third, recent studies on the effects of investors horizons on corporate governance have proxied shareholder horizons in a given firm by the size of institutional ownership in the firm and the institutional investors average portfolio turnover rate (Bushee[1998]; Gaspar, Massa and Matos [2004]); Dikolli, Kulp and Sedatole [2004]; Shin[2005]). Because institutional investors horizons can vary across the firms in which they invest, I directly measure the turnover rate of all blockholders in each of my sample firms. I then 5

7 examine the effect of blockholder turnover on CEO compensation in a simultaneous estimation framework to account for the potential endogeneity of blockholder turnover. Further, my measure of turnover focuses exclusively on large shareholders (blockholders) whose trading strategies are most likely to influence the firms in which they invest. Therefore, my measure should be more likely to detect the effect of shareholder turnover on CEO compensation. The rest of this paper is organized as follow. Section 2 reviews prior research, and Section 3 develops my hypothesis. Section 4 describes the data sources and sample, while Sections 5 and 6 presents the descriptive statistics and the results of empirical tests. Section 7 summarizes and concludes the results. 2. Blockholder Turnover and Investment Horizons As early as Berle and Means (1932), researchers have emphasized that separation of ownership and control creates potential agency problems (Jensen and Meckling [1976]). Shareholders mitigate this problem by monitoring management and designing incentive compensation contracts. Because direct monitoring 3 is costly, however, only shareholders with large investments will find direct monitoring cost-effective (Grossman and Hart [1980]; Shleifer and Vishny [1997]). Accordingly, prior research on the relation between corporate ownership and CEO compensation has focused primarily on the size and distribution of ownership stakes across various classes of investors, such as outside block ownership, managerial ownership and institutional ownership (Holderness and Sheehan [1988]; Mehran [1995]; Core, Holthausen and Larcker [1999]; Hartzell and Starks [2003]). This study contributes to the literature by introducing blockholder 3 I define direct monitoring as shareholder activities like sitting on the board of directors, and gathering information about firms and their management to evaluate managerial decisions and firm performance, and making shareholder proposals to influence the management. 6

8 turnover as an additional dimension of ownership structure which is related to shareholder monitoring. Recent research has established the importance of shareholder turnover (investor horizons) to corporate governance and firm performance. Froot, Perold and Stein (1992) use total share turnover based on all shareholders to measure investors investment horizons, but they emphasize that the turnover of large shareholders will have the most important monitoring implications. My sample firms have an average of 2.32 blockholders who together own an average of 23.23% of the firm s stock. This level of ownership is consistent with blockholder turnover exerting potentially significant effects on the sample firms governance mechanisms. Besides total share turnover, portfolio strategies of institutional investors are also used as a proxy for shareholder horizons of any specific firm in the investors portfolios (Bushee [1998]; Gaspar, Massa, and Matos [2004]). Specifically, Bushee (1998) classifies institutional investors into transient (short-horizon), quasi-indexer and dedicated (long-horizon) based on the investors investment portfolio strategies (portfolio diversification, portfolio turnover and sensitivity to current earnings news). He then uses a firm s shares held by each type of the institutional investors as a proxy for the firm s shareholder horizons. Similarly, Gaspar, Massa, and Matos (GMM hereafter) (2004) measure portfolio turnover of each institutional investor (called churn rate ). They use churn rate to construct a proxy for shareholder horizons of a firm in which the institutional investor invests. Specifically, they measure the firm s shareholder horizon by summing each institutional investor s churn rate weighted based on the investor s 7

9 shareholdings in the firm. The underlying assumption of these measures is that an institutional investor s horizon is constant across all firms in his portfolio. The general conclusion of these studies is that firms with shorter-horizon institutional investors have weaker monitoring of management and suffer higher agency costs. For example, Bushee (1998) finds that firms with shorter-horizon institutional investors are more likely to cut back their R&D expenses when their operating performance is below market expectations. He interprets this result as managers attempting to increase short-term profit at the expense of long-term firm value in response to the pressure from the shorter-horizon investors and the apparent weakness of shareholder monitoring in those firms. GMM (2004) hypothesize that weaker monitoring by short-horizon shareholders (high turnover) leads managers to pursue their self-interest at the expense of shareholders interest in takeover bargaining. Consistent with their hypothesis, they find that target firms with short-horizon shareholders receive lower takeover premiums. Bushee (1998) and GMM (2004) identify the potential importance of shareholder horizons to corporate governance. However, their analysis also raises the question of why we observe short-horizon investors playing such a significant role in corporate ownership if their short investment horizons increase agency costs and reduce long-term firm value. A possible explanation is that firms rely on other governance mechanisms to counter balance the effect of shareholders short-horizons. In this vein, this study examines how blockholder turnover is related to CEO compensation contracts. Two recent studies suggest that institutional investors horizons influence CEO compensation (Dikolli, Kulp and Sedatole [2004]; Shin [2005]). Using Bushee s (1998) 8

10 classification of institutional investors, Dikolli, Kulp and Sedatole (DKS hereafter) document that the greater the ownership by short-horizon (transient) institutional investors, the weaker is the pay-performance sensitivity of CEO cash compensation. Because they examine only CEO cash compensation, however, their study does not provide a complete story as to how institutional investors short-horizons influence design of CEO compensation contracts. For example, it is not clear from their study whether lower pay-performance sensitivity of CEO cash compensation (lower risk) is accompanied with lower total CEO compensation (lower risk premium) or off set by an increase in pay-performance sensitivity of CEO equity compensation (higher risk). In a similar study, Shin (2005) shows that institutional ownership is positively associated with CEO equity compensation regardless of the institutional investors horizons. However, it is difficult to assess whether his finding is applicable to other types of large shareholders, such as individual investor groups or corporations because he examines only institutional investors. In summary, the literature review suggests that while shareholder horizon is an important attribute of corporate ownership, ownership and shareholder horizons have been studied separately without much reference to each other. This study attempts to fill the gap by examining shareholder horizons in the context of existing ownership characteristics and other governance mechanisms, specifically CEO incentive contracts. 3. Hypotheses An owner can influence managers behavior by using a variety of mechanisms including direct monitoring and incentive contracting. Both methods are costly for the owner. For example, an owner must expend resources on gathering firm-specific 9

11 information to execute the monitoring process (Noe [2002]). In addition, Chidambaran and John [1999] argue that an owner bears direct monitoring cost in the form of reduced liquidity because he must hold enough shares and maintain the investment for a sufficiently long period of time to provide effective monitoring. On the other hand, imposing financial risk on a risk-averse manager through incentive compensation is also costly because the owner needs to pay the manager risk premium for the financial risk that the manager bears. Therefore, the owner will determine an optimal mix of monitoring and incentive compensation based on the marginal cost-benefit trade-off of each mechanism (Demsetz and Lehn [1985]; Core, Guay and Larcker [2003]; Hartzell and Starks [2003]). Besides the trade-off between monitoring costs and the risk premium associated with incentive compensation, Lafontaine and Bhattacharyya (1995) and Prendergast (2002) demonstrate how agents private information can influence optimal compensation structures. They argue that when an agent has private information about productivity but there is no effective mechanism for revealing this information, the principal can delegate more discretion to the agent and impose an output-based contract on the agent. In summary, theory suggests that firms with higher monitoring cost and a greater amount of CEOs private information due to less direct monitoring are more likely to rely on incentive compensation to induce desirable actions from managers. Existing literature suggests that blockholders with shorter investment horizons will invest less in monitoring (Bushee [1998]; Gaspar, Massa, and Matos [2004]). Specifically, short-horizon blockholders have less time to learn about the firm to provide effective monitoring. Further, these blockholders improve their portfolio value by trading 10

12 on their superior information about firm value than by providing direct monitoring (Bushee [1998]). Finally, because of limited direct monitoring by short-horizon blockholders, firms with such blockholders will delegate more to managers (Brickley, Smith and Zimmerman [1997]; Prendergast [2002]). These attributes of short-horizon blockholders suggest that firms with short-horizon blockholders are more likely to rely on incentive compensation than on direct monitoring to mitigate managerial agency problems. As the firms impose more compensation risk on managers, they also increase the amount of risk premium for the managers (Holmstrom [1979]). Therefore, I hypothesize the following: H1 Holding all else constant, firms with higher blockholder turnover are likely to have greater CEO pay-performance sensitivity and a higher level of CEO compensation. Most firms base CEO compensation primarily on some combination of market returns and earnings. Given that earnings are subject to management discretion and manipulation, earnings information in firms with a weaker monitoring environment is less informative about managerial effort and less useful for contracting. For example, Lambert and Larcker (1987) show that firms use greater stock-based compensation when accounting measures are noisy. Yermack (1995) also report that CEO compensation is more sensitivity to stock value in companies with noisy accounting data. Similarly, DKS (2004) argue that concentration of ownership by institutional investors with short-horizon (transient) tends to produce incentives for managers to make myopic decisions that adversely affect long-term firm value. Therefore, these firms rely less on earnings for CEO compensation contracts. Such concerns may lead firms with short- horizon 11

13 blockholders to tying CEO compensation to market returns in order to limit the extent of management discretion in earnings measurement. H2 Holding all else constant, firms with higher blockholder turnover are more likely to increase the pay-performance sensitivity of CEO compensation by tying CEO compensation to market returns than to earnings. Agency theory suggests that direct monitoring can be a cost-effective substitute for outcome-based incentive compensation in some circumstances (Holmstrom [1979]; Prendergast [2002]). In the present context, the presence of active blockholders who provide effective monitoring may reduce a firm s reliance on CEO incentive compensation. For example, Schleifer and Vishny (1986) demonstrate that large outside blockholders find positive benefits from direct monitoring of management. Consistent with this argument, Mehran (1995) finds that firms with large outside blockholders use less equity-based compensation than those with few outside blockholders, suggesting that direct monitoring reduces the needs for pay-performance sensitivity. Ke, Petroni and Safieddine (1999) also report that pay-performance sensitivity of CEO compensation is much stronger in publicly-held insurance companies with diffused ownership and less active monitors than in closely-held private insurance companies with concentrated ownership and better monitoring. They interpret this result as direct monitoring substituting for pay-performance sensitivity in the closely-held firms. In this study, I classify outside director blockholders and fiduciaries of ESOP blocks as active monitors. Bethel, Liebeskind and Opler (1998) report that activist blockholders are more likely to be on the board. Therefore, I expect outside blockholders who are also directors of the firm to be active in monitoring management. ESOP fiduciaries represent employee-shareholders and are expected to guard the employees 12

14 interests from any value-destroying acts by management. Faleye, Mehrotra and Morck (2003) argue that most ESOP, 401(K) plans, broad based stock option plans and stock purchase plans usually give employees full voting rights, and therefore, these stakes give employees and their fiduciaries a substantial voice in corporate governance. Talmor and Wallace (2002) support this argument by reporting that firms with ESOP blockholders tend to have lower CEO compensation. They interpret this result as ESOP blockholders providing a disciplining role in CEO compensation 4. Therefore, I expect that the presence of outside director or ESOP blockholders leads to less reliance on CEO incentive compensation. Likewise, as the risk premium associated with incentive compensation declines, so should the level of CEO compensation. H3 Holding all else constant, the presence of blockholders who are active in monitoring management reduces the effect of blockholder turnover on CEO pay-performance sensitivity and the level of CEO compensation. 4. Data and Research Method 4.1 Data sources and sample formation My sample consists of 931 publicly listed U.S. firms drawn from the Investor Responsibility Research Center s (IRRC) surveys in 1995, 1998 and 2000 (Rosenbaum 1995, 1998 and 2000). Block ownership data 5 are from Dlugosz, Fahlenbrach, Gompers 4 ESOPs are also viewed as management friendly (e.g., Borokhovich, Brunarski and Parrino [2000]) or being used by managers as a means to seal themselves from the market discipline (Bethel, et al. [1998]). Therefore, I re-estimate the empirical tests (Tables 4.1 & 4.2) only with director blocks in the monitor category (MONTR). The tests yield results that are qualitatively similar to those with ESOP blocks in MONTR. 5 The SEC requires that all beneficial owners of more than 5% of a company s voting stock be listed in the proxy statement, and therefore, jointly held shares are listed multiple times for each blockholder in the block. For example, Warren Buffett and Berkshire Hathaway jointly own 8.1% of Coca Cola Co. Warren Buffett is a director of Coca Cola. The Coca Cola proxy statement, therefore, lists the 8.1% held by Warren Buffett under Beneficial Ownership by Directors and Executive Officers, and also reports the 8.1% block held by Berkshire Hathaway under Principal Share Owners. Compact Disclosure frequently ignores the footnote details of these jointly held blocks. Instead, they list each blockholder as it appears in the summary table, resulting in multiple counts of the jointly held block ownership. DFGM segregate the shared block 13

15 and Metrick (2004) (DFGM hereafter) based on Compact Disclosure and proxy statements filed with the SEC. Financial, stock price and CEO compensation data come from COMPUSTAT, CRSP and EXECUCOMP, respectively. From DFGM s sample of 1,913 unique firms during the period from 1996 to 2001, I removed 767 firms that do not have at least four consecutive years of ownership data. I further dropped 30 firms due to errors in the ownership data or lack of proxy statements, and 185 firms due to missing data in COMPUSTAT, CRSP or EXECUCOMP. The final sample contains 931 firms. 4.2 Reclassification of Block Types DFGM classify blockholders into five categories: officers, directors, affiliated entities, Employee Stock Ownership Plans and outside blockholders. DFGM categorize individual blockholders who are directors of the firm as director blockholders, but categorize blockholders (often corporations or financial institutions) whose representatives sit on the board of the firm as outside blockholders rather than director blockholders. This classification may understate director block ownership. Therefore, I redefine director blockholders to include all non-manager blockholders who are directors of the firm or whose representatives sit on the board of the firm. Using the information in the proxy statements and existing information in DFGM s ownership data, I then reclassify the blockholder ownership into Manager blocks (MGRBLK), Director blocks (DIRBLK), Outsider blocks (OUTBLK), ESOP blocks (ESOPBLK) and Gray blocks (GRAYBLK). The five classes of block ownership are mutually exclusive, and the sum of the categories always yields the total block ownership of the firm. ownership to correct the errors and create variables that identify the blockholders with shared ownership. This process results in some blockholders with less than 5% ownership. Therefore, using the identifiers as suggested by DFGM, I verify the segregated block ownership against the proxy statements and sum them back to obtain the total corrected block ownership. 14

16 4.3 Measurement of Blockholder Turnover I define blockholder turnover in year t+1 as the proportion of total block ownership that changed hands during year t+1. The change in block ownership from the end of year t to the end of year t+1 reflects a combination of the exit of existing blockholders and the entry of new blockholders. For example, in Table A in the introduction, Teradyne s aggregate block ownership in 1998 and 1999 is 40.8% 6 and 19.6% 7, respectively. During the period from 1998 to 1999, three blockholders (Capital Group, Pioneering Management, Scudder Kemper Investment) owning 33.3% of Teradyne s stock exited the firm, and one new blockholder (FMR Corp) owning 13.6% entered. Therefore, my measure of Teradyne s blockholder turnover in 1999 is ( )/( ) = 77.65%. This measure is a modified version of the mutation ratio used by Franks, Mayer and Rossi (2004) (See Appendices 1A and 1B for more details). Because my study focuses on outside blockholder turnover, I break the total turnover into turnover of outside block ownership and turnover of all other types of block ownership. 4.4 Using Four-Year Averages Corporate governance including ownership structure changes very slowly, resulting in a serial correlation problem in a pooled time-series analysis (Demsetz and Lehn [1985]; Gompers, Ishii, Metrick [2003]). An examination of correlations among the annual ownership data indeed demonstrates that the sample firms experienced very little change in the level of block ownership (total and each blockholder type) as well as 6 Teradyne s total block ownership in 1998 consists of 14.1% for Capital Group, 5.5% for Pioneering management, 13.7% by Scudder Kemper and 7.5% by Neuberger and Berman (Refer to Table A in page 3). 7 Teradyne s total block ownership in 1999 includes 13.6% for FMR Corp and 6.0% for Neuberger and Berman (Refer to Table A in page 3). 15

17 outside blockholder turnover during the four-year sample period 8. Therefore, I use fouryear averages of the block ownership and financial data for each firm in the sample 9. This approach also minimizes the effect of temporary fluctuations in key variables. Additionally, the high correlations suggest that for this sample period, any systematic associations between ownership characteristics and CEO compensation are more likely coming from cross-sectional variation than from within firm variation, further supporting the use of four-year averages. 5. Results This section consists of two parts. The first part presents descriptive statistics, and the second part reports results of hypothesis tests. 5.1 Descriptive Statistics The sample firms are generally large with average assets and market value of equity of $7,042 million and $7,485 million, respectively (Table 1). The mean (median) annual stock return is 20.3% (14.7%). The mean (median) total CEO compensation is $4.64 million ($2.51 million), which is comparable to the $4.21 million ($1.76 million) for the period 1993 to 2000 reported by Core, Guay and Verrecchia (2003). OPTION is the average change in the value of stock options granted in the current year per $1,000 in shareholder wealth, and measures pay-performance sensitivity of stock option compensation (Yermack [1995]; Hartzell and Starks [2003]; Core, Guay and Larcker [2003]). The average OPTION is $1.92 per $1,000 of shareholder wealth with a median 8 Four year average correlations of the annual block ownership of the sample firms are 0.79 for total block; 0.94 for manager block; 0.89 for director block; 0.68 for outside block; 0.91 for ESOP block and 0.71 for affiliated (gray) block. The correlation for annual outside blockholder turnover is For the sample firms with more than four years of block ownership data, I randomly select the starting year to include only four years. Using four-year averages is consistent with Demsetz & Lehn (1985) and Demsetz & Villalonga (2001) who use five-year averages of block ownership in their study of ownership structure. 16

18 of $1.01. These values are larger than Hartzell and Starks corresponding mean and median values of $0.98 and $0.17 for the period 1992 to1997 and Yermack s $0.59 and $0.07 for the period 1984 to 1991, consistent with the increasing use of options in CEO compensation throughout the 1990s. The mean (median) block ownership of the sample firms is 23.23% (22.00%), and the mean (median) number of blockholders is 2.32 (2.25). Outsiders are the largest blockholder group with an average (median) ownership of 15.07% (13.57%), followed by manager blockholders (average 3.77%) and director blockholders (average 2.91%). However, the distributions of manager block ownership and director block ownership are highly skewed because 75% to 80% of all sample firms have no manager or director blockholders. The mean (median) turnover of total block ownership (TURNRATE) is 28.97% (25.42%). The total blockholder turnover is driven primarily by outside blockholders (OUTTURN) with a mean (median) turnover of 27% (22.82%) as compared with 1.85% (0.00%) for all other blockholders (manager block, director block, ESOP block and gray block). 5.2 The Effect of Blockholder Turnover on CEO Compensation Compensation Measures I use two measures of compensation level (total CEO compensation and CEO cash compensation) and two measures of pay-performance sensitivity (ratio of stock options granted to total CEO compensation and OPTION). Total CEO compensation (TCOMP) is the total payments made to a CEO for the fiscal year ($000), including salary, bonus, stock options, restricted stock and all other compensation. Cash compensation (CASH) is the sum of CEO salary and cash bonus 17

19 ($000) for the year. Stock options granted is the aggregate value of stock options granted to a CEO during the fiscal year valued using S&P s Black-Scholes method. OPTNRATE is the ratio of stock options granted to total CEO compensation and measures the extent to which a CEO s total compensation is tied to stock performance. OPTION is the change in the value of options granted during the fiscal year per $1,000 in shareholder wealth (Yermack [1995]; Hartzell and Starks [2003]). In addition to the two pay-performance sensitivity measures for stock option compensation (OPTNRATE, OPTION), I use two interaction terms to measure payperformance sensitivity of total CEO compensation and cash compensation. HITURN is an indicator variable for the firms with outside blockholder turnover above the 50 th percentile. In the total compensation (cash compensation) regression, [ADJRET* HITURN] is the interaction between market adjusted stock return (ADJRET) and the high turnover dummy (HITURN), and [ROA*HITURN] is the interaction between return on assets (ROA) and HITURN. Positive coefficients for the interaction terms would be consistent with total CEO compensation (cash compensation) being more sensitive to the performance measures when outside blockholder turnover is higher Hypotheses 1 & 2 Empirical Model: Hypothesis 1 predicts that firms with higher outside blockholder turnover will have higher pay-performance sensitivity and a greater level of CEO compensation. Hypothesis 2 further predicts that the increase in pay-performance sensitivity is primarily due to tying CEO compensation to market returns. To test these 18

20 hypotheses, I first estimate the CEO Compensation (Equation (1)) using OLS 10. Second, because outside blockholder turnover (OUTTURN) is potentially endogenous 11 to various firm characteristics that are related to monitoring cost-benefit trade-off, I reestimate the effect of OUTTURN on CEO compensation using the following system of Equations (1) and (2). Given the skewed distribution of several key variables, such as block ownership, firm size and stock option compensation, log transformations are used when appropriate. CEO Compensation 5 log (COMP) = α 0 + β 1 log(outturn) + β 2 (INTURN)+ Σ δ k log (BLOCK) + β 3 CEOTURN + k=1 β 4 log(asset) + β 5 [log(asset)] 2 + β 6 log (LEV) + β 7 log (DIV) + β 8 log (BETA ) + β 9 log (IDIORISK) + β 10 ROA + β 11 ADJRET + β 12 log (MKTBV) + β 13 (ROA*HITURN) + 11 β 14 (ADJRET*HITURN) + Σ γ m Industry dummy + µ. (1) m = 1 (Where k denotes blockholder types 1 through 5 and m denotes Fama-French industry portfolio dummies 1 through 11) Outside Blockholder Turnover log (OUTTURN) = ϕ 0 + λ 1 log (COMP) + λ 2 SNPDUMY i + λ 3 log (ASSET) + λ 4 [log(asset)] 2 + λ 5 log(lev)+ λ 6 log(div)+ λ 7 log (BETA) + λ 8 log (IDIORISK) + λ 9 ROA + λ 10 log(q) + λ 11 log(shrturn) + λ 12 log(astplc) +ν (2) When examining the relation between outside blockholder turnover and CEO compensation, accounting for endogeneity of outside blockholder turnover is important for two reasons. First, the observed relation between outside blockholder turnover and 10 Because of high correlations between OUTTURN and the block ownership variables, I tested for multicollinearity in the OLS regression. Except for firm size variables, none of the other variables in the OLS regression has VIF greater than 3, suggesting no significant multicollinearity problems in the model. 11 Hausman test results also reject the null hypothesis of the exogeneity of independent variables at p <

21 CEO compensation is potentially spurious due to omitted variables in Equation (1), resulting in non-zero correlation between outside blockholder turnover and the error term of Equation (1). The second issue is that firms could potentially be adopting specific compensation policies to attract certain type of investors, and, if so, CEO compensation could bring outside blockholder turnover rather than vice-versa. To account for these issues, I estimate the CEO compensation (COMP) and outside blockholder turnover (OUTTURN) simultaneously using 2SLS 12. In Equation (1), COMP is the natural log of CEO compensation measures. To capture the effect of outside blockholder turnover on CEO compensation, I include outside blockholder turnover (OUTTURN) as the key variable of interest. I also include turnover by all other types of blockholders (INTURN). To control for differential monitoring incentives and skills among blockholder types, I include the level of block ownership held by the five classes of blockholders (manager, director, outsider, ESOP and gray). Beside ownership characteristics, I use a large number of control variables because the existing literature on compensation suggests that the level of CEO compensation and pay-performance sensitivity are cross-sectionally related to various firm characteristics. For example, CEO turnover (CEOTURN) often results in significant fluctuations in CEO compensation because of severance payment to the exiting CEO or a hiring bonus paid to the incoming CEO. Therefore, I include the number of CEO turnovers over the four year period (CEOTURN) to control for the effect of CEO turnover. 12 To account for heteroscedasticity, I also used Heteroscedastic 2SLS Regression (SAS Proc Model procedures). The test yields results that are qualitatively similar to those of the traditional 2SLS simultaneous estimation. 20

22 Ceteris paribus, the larger the size of the company, the greater is the management s discretion to influence the shareholder wealth (Demsetz and Lehn [1985]; Core, Holthausen and Larcker [1999]).Therefore, firm size (ASSET) is expected to be positively associated with both CEO pay-performance sensitivity and the level of CEO compensation. To the extent that leverage (LEV) reflects agency costs between shareholders and debt-holders (Jansen and Meckling [1976]), firms with higher leverage will encourage their CEOs to engage in risky projects and pay greater CEO compensation in the form of stock options. Therefore, I expect a positive association between leverage and pay-performance sensitivity of CEO option compensation (OPTNRATE and OPTION). Dividend (DIV) paying firms tend to be mature firms with greater cash flows, and thus are expected to be negatively associated with equity-based CEO compensation (Smith and Watts [1992]). Idiosyncratic firm risk (IDIORISK) reflects volatility of a firm s operating environment. The ratio of market-to-book value of equity (MKTBV) is often used as a proxy for investment opportunities (Smith and Watts [1992]). To the extent that a volatile operating environment and investment opportunities are associated with greater complexity of managerial tasks and larger managerial discretion, IDIORISK and MKTBV will be positively associated with CEO pay-performance sensitivity and the level of CEO compensation (Talmor and Wallace [2002]). Agency theory suggests that the level of CEO compensation is positively associated with firm performance such as return on assets (ROA) or market adjusted stock returns (ADJRET). Finally, I include industry dummy variables because the use of option plans varies across industries (Core, Guay and Larcker [2003]). 21

23 To estimate outside blockholder turnover in Equation (2), I use a set of firm characteristics which are known to be related to ownership structure. They are firm size measured as total assets (ASSET), firm risk measured as market beta (BETA) and idiosyncratic firm risk (DIORSK), capital structure measured as leverage (LEV), dividend yield (DIV), operating performance measured as return on assets (ROA), share liquidity measured as share turnover rate (SHRTURN), asset characteristics measured as assets-in-place (ASTPLC) and S&P index dummy (SNPDUMY). Full descriptions of all the variables used in the models are in Appendix 2. Results for Hypotheses 1 & 2: Hypothesis 1 predicts that greater blockholder turnover will be associated with higher pay-performance sensitivity. As firms impose more compensation risk on risk-averse CEOs through higher pay-performance sensitivity, the level of CEO compensation should offset the greater risk. Therefore, Hypothesis 1 further predicts that outside blockholder turnover is positively associated with the level of CEO compensation. OPTNRATE, the ratio of stock options granted to total CEO compensation, and OPTION, the change in option value per $1000 shareholder wealth, measure payperformance sensitivity of CEOs option compensation. Tables 3.1 and 3.2 report the regression results of the two compensation measures. The OLS regression results in Table 3.1 Panel A show that outside blockholder turnover (OUTTURN) is highly positively associated with the ratio of stock options granted to total CEO compensation (OPTNRATE) (t = 7.59).This result holds after controlling for the level of five classes of block ownership and various economic determinants of CEO compensation. The simultaneous estimation results in Table 3.1 Panel B are consistent with the OLS 22

24 regression results. The positive and significant coefficient for OUTTURN (t = 4.23) suggests that firms with higher outside blockholder turnover use stock options more extensively for CEO compensation, increasing CEO pay-performance sensitivity. This result is further supported by the results for OPTION in Table 3.2. In all model specifications and all estimation methods of OPTION, the coefficients on outside blockholder turnover (OUTTURN) are positive and statistically significant at p < The results in Tables 3.3 and 3.4 show that the coefficient for outside blockholder turnover (OUTTURN) is positive and significant in total CEO compensation (Table 3.3, Panel B) (t = 2.33), but is insignificant in CEO cash compensation (Table 3.4, Panel B) (t = -0.90). These results suggest that in response to greater risk associated higher payperformance sensitivity, firms with higher outside blockholder turnover increases the level of total CEO compensation through other forms of compensation, such as long-term incentive payments, restricted stock or stock options rather than cash payments. This interpretation is consistent with Hypotheses 1 and 2. As a further examination of Hypotheses 1 and 2 for total CEO compensation (CEO cash compensation), I include two interaction terms in the compensation model: [ADJRET*HITURN] and [ROA*HITURN]. Positive and statistically significant coefficients for the two interaction terms would indicate that sensitivity of total CEO compensation (CEO cash compensation) to firm performance is greater when blockholder turnover is higher, supporting Hypothesis 1. Hypothesis 2 predicts that firms with higher blockholder turnover are more likely to use market adjusted returns (ADJRET) than earnings (ROA) to increase pay-performance sensitivity of total CEO compensation 23

25 (CEO cash compensation). Therefore, a coefficient for [ADJRET*HITURN] that is larger than the coefficient for [ROA*HITURN] would support Hypothesis 2. Table 3.3 reports the results for pay-performance sensitivity of total CEO compensation. The simultaneous estimation results in Panel B demonstrate that firms with low blockholder turnover tie CEO total compensation to earnings (ROA) (t=2.99) but not to market adjusted returns (ADJRET) (t = -0.38). However, when outside blockholder turnover is high, the firms reduce their use of earnings (ROA*HITURN) (t = -2.16) and rely more on market adjusted returns (ADJRET*HITURN) (t = 3.20) to determine total CEO compensation. Hypotheses 1 and 2 jointly predict a positive and significant coefficient for [ADJRET*HITURN]. Therefore, the results for total CEO compensation support both hypotheses. More interestingly, these results provide empirical evidence that firms choose a performance measure for CEO compensation based upon their monitoring environment. The results in Table 3.4 show that after accounting for the endogeneity of outside blockholder turnover (OUTTURN), coefficients of the two interaction terms are insignificantly different from zero in CEO cash compensation. This result suggests that outside blockholder turnover (OUTTURN) does not have significant effect on sensitivity of CEO cash compensation to either earnings (ROA) or market adjusted returns (ADJRET). Together with the positive effect of outside blockholder turnover on payperformance sensitivity of CEO option compensation, these results for CEO cash compensation indicate that firms with higher blockholder turnover increases CEO payperformance sensitivity by granting stock options rather than relying on earnings-based cash incentive payments. This interpretation also explains the insignificant association 24

26 between outside blockholder turnover (OUTTURN) and the level of CEO cash compensation. Overall, the results strongly support Hypotheses 1 and 2. First, firms with higher blockholder turnover tend to have greater CEO pay-performance sensitivity and larger total CEO compensation (H1). Second, to increase CEO pay-performance sensitivity, firms rely more on equity-based incentives than on cash incentives (H2). This interpretation is reinforced by the finding that high blockholder turnover firms rely more on market adjusted returns than on earnings for total CEO compensation. A further result to note is for the level of outside block ownership. Although Hartzell and Starks (2003) find that institutional ownership increases pay-performance sensitivity of CEO option compensation and decreases total CEO compensation, I find in Panel B of Tables 3.2 and 3.3 that after controlling for outside blockholder turnover, the level of outside block ownership has an insignificant association with both CEO payperformance sensitivity and the level of total CEO compensation. This result is consistent with outside blockholders exercising their influence on the firm by voting with their feet (Parrino, Sias and Starks [2003]) rather than by other monitoring activities Hypothesis 3 Hypothesis 1 implies that direct monitoring substitutes for outcome-based CEO incentives. Accordingly, Hypothesis 3 predicts that the presence of blockholders who are active in monitoring managers will reduce the effect of outside blockholder turnover on CEO pay-performance sensitivity as well as the level of CEO compensation. To test Hypothesis 3, I classify blockholders into those whom I expect to be more monitors of management (directors and ESOP fiduciaries) versus those who are more management 25