The Pennsylvania State University. The Graduate School. The Mary Jean and Frank P. Smeal College of Business Administration

Transcription

1 The Pennsylvania State University The Graduate School The Mary Jean and Frank P. Smeal College of Business Administration ESSAYS IN CORPORATE FINANCE: THE IMPACT OF CORPORATE GOVERNANCE DURING ECONOMIC STRESS AND TRACKING STOCK IN THE UNITED STATES A Thesis in Business Administration by Linda L. Miles 2005 Linda L. Miles Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy December 2005

2 The thesis of Linda L. Miles was reviewed and approved* by the following: Dennis P. Sheehan The Virginia and Louis Benzak Professor of Finance Thesis Adviser Chair of the Committee Laura C. Field Associate Professor of Finance Chris J. Muscarella L.W. Roy and Mary Louis Clark Teaching Fellow Professor of Finance Xiaoe-Jenny Li Associate Professor of Economics and Mathematics William A. Kracaw David Sykes Professor of Finance Chair of the Department of Finance *Signatures are on file in the Graduate School. ii

3 ABSTRACT Chapter I Does Corporate Governance Make a Difference in Hard Times? Using data from proxy statements, I investigate the impact of various firm characteristics and corporate governance measures on firm performance during a period of an exogenous economic shock. The first sample investigates the steel industry and its shock in The second sample is the hotel industry after the terrorist attacks of September 11, In the steel industry, the level of firm performance after the shock is strongly related to firm size and negatively related to the percentage of votes held by those affiliated with the firm. To assess the firm s ability to mitigate losses when the industry is in a severe downturn, I calculate each firm s change in performance from prior to the shock to after the shock as compared to the median industry decrease. Successful performance is positively related to the percentage of insiders on the board of directors. In the hotel industry, the level of firm performance after the shock is positively related to boards of directors of fewer than nine members and negatively related to the percentage of inside directors. When measuring the change in performance from before to after the shock, the magnitude of change is positively related to the percentage of inside directors and the existence of staggered terms for directors. The results suggest that there is a relationship between corporate governance and firm performance in hard times, though in a period of economic shock, it appears that boards of directors with significant insider participation are more successful in adapting to the changing market. These findings are consistent with prior research by Dallas (2003), suggesting the value of inside directors in enhancing the ability of the board of directors to perform its relational and strategic management roles. Chapter II Tracking Stock from Beginning to (almost) End Tracking stock is an equity structure whereby a firm issues stock tied to the performance of a portion of the corporation rather than the entire corporate entity. Tracking stock was first used in 1985, became more popular in the 1990 s and all but disappeared by This study extends prior research by lengthening the time horizon, investigating the market microstructure effects of the abnormal returns, and examining the effects of announcement of the elimination of the tracking stock structure. The issuance of tracking stock has a positive wealth effect, with a magnitude for announcement of approximately 3% and significant over the three-day event window. The wealth effects appear to be independent of market microstructure frictions. The returns are significantly different among the types of tracking stock, which supports increased corporate focus and the reduction in asymmetric information as explanations for the abnormal return to shareholders. The execution window effect is approximately 2% and significant for reorganization tracking stock. Execution window returns are insignificant for merger tracking stock. The announcement of the elimination of tracking stock structure by returning to a single common stock for the entire corporation is not a significant wealth event for the parent firms, but evidences significant excess returns of greater than 10% for the tracking stocks over the threeday event window. iii

4 TABLE OF CONTENTS List of Figures List of Tables Acknowledgements v vi vii Chapter I. DOES CORPORATE GOVERNANCE MAKE A DIFFERENCE IN HARD TIMES? Introduction 1 Review of Selected Literature and Motivation 3 Data 8 Empirical Results 11 Steel Industry 12 Hotel Industry 17 Conclusion 21 References 23 Appendix 25 Figures 27 Tables 31 Chapter II. TRACKING STOCKS: FROM BEGINNING TO (ALMOST) END Introduction 45 Emergence, development and dissolution of tracking stock 47 Prior tracking stock research 51 Motivation 53 Data 56 Methodology and empirical results 57 Wealth Effects and Robustness Analysis of Issuance of Tracking Stock 57 Intra-Day Analysis of Announcement CARs and Robustness Analysis 60 Cross-Sectional Analysis 63 Wealth Effects of Elimination of Tracking Stock 66 Conclusion 67 References 69 Tables 71 Figures 90 iv

5 LIST OF FIGURES 1.1 Steel Industry Production Seasonally Adjusted 1998 through CRSP Quarterly Percentage Returns 1989 through U.S. Hotel Industry Descriptive Measures 2000 through CRSP Monthly Percentage Returns 2000 through Announcement Date Daily Volume (-60, +10) Announcement Date Daily Volume (-10, +10) Order Flow Ratio and Spread (-3, +3) Announcement Date Returns (-3, +3) Execution Date Daily Volume (-60, +10) Execution Date Daily Volume (-10, +10) 95 v

6 LIST OF TABLES 1.1 Descriptive Statistics for Steel Industry and Sample 1989 to Steel Industry 1990 Proxy Data Statistics Coefficient Estimates from Univariate OLS Regressions, Steel Industry ROA Coefficient Estimates from Multivariate OLS Regressions, Steel Industry ROA Univariate OLS Regressions, Steel Industry Change in ROA Multivariate OLS Regressions, Steel Industry Change in ROA Univariate OLS Regressions of Steel Industry Performance on Governance Index Descriptive Statistics for Hotel Industry and Sample 2000 to Hotel Industry 2001 Proxy Data Statistics Coefficient Estimates from Univariate OLS Regressions, Hotel Industry ROA Coefficient Estimates from Multivariate OLS Regressions, Hotel Industry ROA Univariate OLS Regressions, Hotel Industry Change in ROA Multivariate OLS Regressions, Hotel Industry Change in ROA Univariate OLS Regressions of Hotel Industry Performance on Governance Index Firms Issuing Tracking Stock Firms Eliminating Tracking Stock Prior Empirical Research in Tracking Stock Cross-Sectional Analysis of Announcement Effect Cross-Sectional Analysis of Execution Date Effect Tracking Stock Pending Interval Wealth Effects of Tracking Stock at Announcement Date (-1, +1) Wealth Effects of Tracking Stock at Execution Date (-1, +1) Combined Announcement and Execution Date Wealth Effects of Tracking Stock Robustness Analysis Wealth Effects at Announcement Date Announcement Stock Returns on Intra-Day Intervals (-1, +1) Robustness Analysis Wealth Effects at Execution Date Ex-Date Stock Returns on Intra-Day Intervals (-1, +1) Cross-Sectional Analysis B T-tests for Differences of Mean for Cross-Sectional Analysis Wealth Effects on Announcement Date (-1, +1) of Reunification of Tracking Stocks Remuneration to Holders of Tracking Stocks Upon Reunification with Parent Firm Comparison of Cross-Sectional Analysis of Announcement Effect to Predictions Comparison of Cross-Sectional Analysis of Execution Date Effect to Predictions 89 vi

7 ACKNOWLEDGEMENTS I gratefully express my deep appreciation to my committee members, Dennis Sheehan, Laura Field, Chris Muscarella and Jenny Li for their discussions, support and guidance. I am indebted to Lynn Fisher for her friendship, insight and assistance. My colleagues at Illinois State University, particularly Joe Trefzger and Bill Scott, have encouraged me and provided valuable suggestions. In addition, thanks are extended to graduate assistant extraordinaire Archana Mandala. A loving thank you to my husband, Richard Vaughn, for all he did, and continues to do, in support of my academic career. And to my children, Jonathan, Jeremy, Jadrian and Devin, thank you for your understanding and love you are amazing individuals; I am so proud of you. All remaining errors are mine alone. vii

8 I. Does Corporate Governance Make a Difference in Hard Times? A. Introduction Corporate governance, in a broad view, can be defined as the characteristics of a firm and its board of directors that establish the framework for decision-making. The owners (shareholders) of the corporation typically rely on others (agents) to operate the firm. The owners elect the board of directors which is responsible for choosing and monitoring the chief executive officer (CEO) of the organization. The CEO, in turn, is responsible for management of the firm. As residual claimants, the wealth of the shareholders is negatively impacted by actions that reduce firm value. Jensen (1993) notes that the board of directors has the final responsibility for the functioning of the firm. Researchers studying corporate governance have long grappled with measuring the effectiveness of the board of directors. Jensen (1993) suggests that independent boards are more effective, with smaller boards more effective than larger boards. Empirical corporate governance studies on the impact of board composition on the wealth of the firm have produced supporting evidence that the market values independent directors. 1 Support for the effectiveness of smaller boards is provided by Yermack (1996). Empirical studies of the impact of board composition on firm performance are more ambiguous. Hermalin and Weisbach (1991) and Bhagat and Black (1999) report no relationship between board composition and firm performance. Denis and Sarin (1999) document significant changes in board composition and ownership structure after fundamental changes in the business environment, but find no impact of board structure on stock performance. 1 For example, Rosenstein and Wyatt (1990), Brickley, Coles and Terry (1994), and Shivdasani and Yermack (1999). 1

9 Dallas (2003) suggests that insider dominated boards are more effective for certain tasks; while outsider dominated boards are more efficient for other purposes. Her paper provides an explanation for the ambiguous, and sometimes conflicting, results in many governance studies. Hermalin and Weisbach (2003) assert that [b]oard independence does, however, matter for certain board actions, particularly those that occur infrequently or only in a crisis situation. I study the effect of variation in corporate governance characteristics upon the performance of firms within an industry when that industry is subjected to an exogenous economic shock. The effectiveness of various governance structures should be magnified during these times of fundamental changes in the business environment. The choice of single industry analysis eliminates any potential differences in optimal governance structure among industries. Choosing a point in time when all firms are impacted by an outside event allows examination of firms with different corporate governance structures; the superiority of a specific governance structure is expected to be reflected in better firm performance. My study investigates the effectiveness of firms in dealing with an exogenous economic shock in the steel industry and the hotel industry, at separate points in time. My data contains extensive documentation on firm, board, chief executive officer and director characteristics. In both industries, on a univariate basis, the level of firm performance after the shock is strongly and positively related to firm size. For the steel industry, in a multivariate analysis, firm size and firms with classes of stock that possess super voting rights are positively related to performance, while the percentage of votes held by those affiliated with the firm 2 and boards of fewer than nine directors are negatively related. In the hotel industry, boards of fewer than nine members is 2 Percentage affiliated votes is defined as the percentage of votes outstanding held by those affiliated with the firm, including affiliated directors, blockholders, officers and directors. 2

10 positively related, while the percentage of insiders on the board is negatively related to firm performance. To assess the firm s ability to mitigate losses when the industry is in a severe downturn, I calculate each firm s change in performance from prior to the shock to after the shock as compared to the industry median decrease. In both industries, whether measured on a univariate or multivariate basis, successful performance is positively and significantly related to the percentage of insiders on the board of directors, though the coefficients are small. For the steel industry, the percentage of officer and director holdings are negatively related in both the univariate and multivariate analysis. In the hotel industry, firms with staggered boards are positively related to firm performance. The results suggest that corporate governance does matter in the hard times, though the stronger performance by firms with a greater percentage of inside directors is contrary to the current trend towards boards primarily composed of outside directors. For the steel industry after the economic shock in 1990 and the hotel industry after the economic shock in 2001, firms with fewer inside directors were more strongly impacted by the economic downturn. The remainder of the paper is as follows. Section B reviews related literature and discusses the motivation for the study, Section C describes the data for the two samples, Section D discusses the empirical results, and Section E concludes. B. Review of Selected Literature and Motivation Empirical corporate governance studies on the impact of board composition on the wealth of the firm have produced evidence that the market values independent directors. Rosenstein and Wyatt (1990) report a positive announcement effect for firms announcing in The Wall Street 3

11 Journal the appointment of an outside director. While statistically significant, the economic effect is slight, at 0.2%. Brickley, Coles and Terry (1994) observe that stock market reaction is positive to the announcement of the adoption of poison pill provisions when the board of directors is composed primarily of outside directors and negative when it is not. Further investigation of the market s positive reaction to corporate governance changes relating to director independence is offered by Shivdasani and Yermack (1999). Their data is comprised of firms in the Fortune 500 in 1995, and records market reaction to director appointments during the years 1994 through They report that stock price reaction to the appointment of outside directors is lower when the CEO of the firm is a member of the nominating committee or there is no nominating committee. This lends support to the supposition that independent directors are a wealth enhancement to the firm. The size of the board of directors has also been considered. Jensen (1993) suggests that smaller boards are more effective than larger boards. Empirical evidence is provided by Yermack (1996) in a study of 452 large firms in the United States during the period from 1984 through He documents an inverse relationship between board size and Tobin s Q among these firms. Additionally, when there is a significant change in the size of the board of directors (a difference of at least 4 directors), the market reaction is positive for a decrease in size and negative for an increase, with an average magnitude of 2.5%. These findings appear to be unrelated to the proportion of outsiders on the board, as some of the boards that enlarged also increased their proportion of independent directors. Empirical corporate governance studies of the impact of board composition on firm performance are more ambiguous. Hermalin and Weisbach (1991) use a sample of 134 NYSE firms with information gathered for the years 1971, 1974, 1977, 1980 and They find no 4

12 relationship between board composition and performance. Bhagat and Black (1999) report no relationship between board composition and firm performance in their sample of 957 firms listed in the Institutional Shareholder Services database. Their study covers the period between 1985 and 1995, and includes information regarding board composition, stock price performance and financial performance. Gompers, Ishii and Metrick (2003) present the analogy of a corporation as a republic, likening the variation in methods of corporate governance to the differences between a democracy and a dictatorship. In a dictatorship, shareholders cede some of their rights of control to managers, ostensibly to gain greater returns. Gompers et al develop a governance scale based on 24 corporate governance provisions reported by the Investor Responsibility Research Center. In the study of 1500 large firms in the 1990 s, they find that firms with better corporate governance (greater shareholder rights, or democracies ) are strongly correlated with stock returns. Denis and Sarin (1999) study panel data on 583 publicly traded firms during the period 1983 through They document significant changes in board composition and ownership structure after fundamental changes in the business environment, but find no impact of board structure on return on equity stock performance. Hermalin and Weisbach (2003) discuss the joint endogeneity problem in empirical corporate governance studies of the relationship between firm performance and governance characteristics. While accounting measures of firm performance may be dependent upon the composition of the prior board, the future board composition is related to both the firm s performance and the prior board. Additionally, since board structure typically changes only as 5

13 the result of elections at the annual meeting, the question arises as to whether we are studying a system out of equilibrium. Dallas (2003) labels the three major roles of the board of directors as monitoring of management, relational, and strategic management, with the relational role described as the sharing of information and contacts among stakeholders of the corporation. Monitoring of management is most effectively performed by outsider dominated boards. Insiders and those closely associated with the corporation are more effective at the relational and strategic management roles. In this paradigm, the effectiveness of a particular board composition would depend upon the role(s) the board was required to perform for the corporation. A general definition of good corporate governance is the framework within a firm that successfully protects the financial rights of shareholders. In large corporations, the owners (shareholders) have managers run the firm. The board of directors is responsible for choosing and monitoring the chief executive officer (CEO) of the organization. The CEO, in turn, is responsible for the firm management. As residual claimants, the wealth of the shareholders is subject to diminishment by actions that reduce firm value. Jensen (1993) notes that the board of directors has the final responsibility for the functioning of the firm. Thus the board of directors is an important component of corporate governance. In the United States, boards of directors vary in size from very small to over two dozen members. Some firms elect the entire board annually, others have staggered terms. Boards may be comprised entirely of individuals with close ties to the firm, while other boards have a majority of unaffiliated members. Whether these structural differences lead to varying degrees of effectiveness remains an open question. 6

14 Jensen (1993) lists several attributes of effective boards, including independence from the CEO, the CEO and chairman of the board (COB) are different individuals, directors have substantial equity holdings in the corporation, and small size (eight or fewer members). 3 The Sarbanes-Oxley Act of 2002 (SOX) has mandated changes to corporate governance requirements; the New York Stock Exchange (NYSE) and NASDAQ have rules for their listed firms to comply with these provisions. The NYSE corporate governance rules approved by the United States Securities and Exchange Commission (SEC) in 2003 and updated in 2004, require that the board of directors of a listed firm have a majority of independent directors. 4 The NYSE noted in its original filing that there is public concern with the super-voting rights in some corporations and that guidelines to address these concerns may be established in the future. While the changes in regulations intuitively appear to improve shareholder protection, empirical evidence that a board dominated by outside directors correlates with increased shareholder wealth is lacking. Philippon (2003) creates a governance model in which his simulation results indicate that badly governed firms respond more to aggregate shocks than do well governed firms. If a difference in corporate governance structure is related to firm performance, these differences should be more readily observable during a period of economic stress. When a market is expanding, most firms are able to profit. However, in economic downturns, with firms competing for shares in a shrinking market, any competitive advantage becomes more pronounced. 3 Equity holdings are substantial with respect to the total wealth of the individual director, which does not necessarily mean that they comprise a significant proportion of the equity of the corporation. 4 The 2004 amendment strengthened the definition of independent by requiring that the label of independence only be applied to directors who are determined by the board to have no material relationship with the firm, and the basis for the determination must be stated in the proxy statement. Thus the label of independent can no longer be assigned merely by an individual director asserting his or her independence. 7

15 Gertner and Kaplan (1996) argue that optimal corporate governance may differ across industries. Similarly, Adams and Mehran (2002) state that performance results within a single industry may be less impacted by measurement errors than those across multiple industries. Kole and Lehn (1997) present an industry study of the changes in corporate governance after the exogenous shock of airline deregulation. They note that smaller boards are more effective, citing as reasons a decrease in free-riding and more rapid decision-making. Governance characteristics changed, albeit slowly, and all components did not change at the same rate. My study focuses on an industry in a time of negative economic stress, examining first the effect of variation in corporate governance upon the performance of firms in the steel industry when that industry is subjected to the exogenous economic shock of the recession. Next the effect of corporate governance upon the performance of firms in the hotel industry after the impact of the terrorist attacks of September 11, 2001 is examined. The effectiveness of various governance structures may be magnified during times of fundamental changes in the business environment, and any variations therein become more pronounced. The choice of a single industry analysis eliminates the variances in effectiveness among industries. Choosing a point in time when all firms are impacted by an outside event allows examination of different firm and board compositions managing the situation. Two industries, subjected to negative economic stress at different times, are studied to ascertain any commonalities in a relationship between governance structures and firm performance. C. Data The sample data set for the steel industry was compiled by obtaining all firms listed in Standard Industrial Classification (SIC) 331, Steel Works, from Center for Research in Security 8

16 Prices (CRSP), Standard & Poor s Compustat dataset (Compustat) and Value Line Investment Survey. All available proxy statements were obtained and data recorded for the period 1990 through Using each firm s proxy statement, the demographics of the firm and the seated board of directors in July 1990, the time of the economic shock to the steel industry, are used to examine the relationship between the governance structure and firm performance. The hotel industry sample was obtained from CRSP and Compustat listings of SIC 710, Hotels and Motels and North American Industry Classification System (NAICS) 7211 Traveler Accommodation. Proxy data was obtained for firms in the hotel industry to determine firm characteristics and board composition as of September 11, The classes of stock of the corporation and the voting rights for each class are recorded, as well as whether cumulative voting is permitted. The number of shares outstanding as of the record date is obtained, and the number of votes outstanding are calculated if there are different voting rights among the classes of stock. The existence of a tiered board (certain board seats must be filled by a pre-specified group, such as representatives of a labor union or preferred shareholders) and a staggered board (members terms are for more than one year and a portion of the board seats are up for election each year) are recorded as zero/one variables. Firm characteristics are further described by the total number of officers and directors and the total holdings, in terms of shares and votes, held by that group. 5 Further information is gathered on the CEO and each director. Sections on biographical information, stock holdings, material relationships, and notes from the proxy statements were used to record information about each director and her/his relationship to the firm. In conformance with the typical classification method, board members 5 Most firms report share holdings of officers and directors as a group in the proxy statement. Some firms instead report holdings for executives and directors in which case those are the figures recorded. 9

17 are identified as insiders, outsiders, or affiliates (sometimes called grays ), with affiliates including those who do not fit clearly into either of the other categories. An insider is a current or former employee of the firm, or the relative of a current or former employee. Affiliates may bring additional industry-specific knowledge to the board, perhaps enhancing the board s effectiveness in performing the relational and strategic management roles typically performed by insiders. Outsiders are those non-employees with no discernable relationship to the firm other than the position of director. 6 In addition, it is noted whether the information provided in the proxy statement shows the outside director possesses specific industry-related experience. The impact of the relative power of the chief executive officer (CEO) is well documented in the literature. Information gathered with respect to the CEO includes age, tenure as CEO, tenure with the firm and whether the position of chairman of the board of directors (COB) is held by the same individual. These variables will be used to attempt to separate the influence of the CEO on performance from the impact of board characteristics. Information is gathered about blockholders of the firm identified in the proxy statement as holding five percent (5%) or more of the outstanding shares of the company. Holderness (2003) states that outside blockholders have different incentives than insiders. In the industry sample, blockholders are classified as affiliated or unaffiliated with the firm, based on information presented within the proxy. Affiliated blocks are when the shareholder has an identifiable link to the firm, including employees, directors or founders of the firm and their relatives, employee stock ownership plans, joint venture firms, etc. Unaffiliated blocks have no identified link to the firm. The total number of blockholders is recorded, as well as the total number of shares and votes held by affiliated and unaffiliated blockholders. 6 The definition of outsider varies in the literature. Some definitions merely mean the individual is not an employee. My definition seeks to classify non-employees based on their relationship to the firm. Recent literature often uses independent for directors with no evident relationship with the firm. 10

18 Data on firm performance was obtained from Compustat s Industrial Annual dataset. Calculated ratios are expressed as percents. Return on assets (ROA) equals operating income divided by total assets (data items 13 and 6). Market capitalization is calculated as the share price at the end of the year times the number of shares outstanding (data items 199 and 25). D. Empirical Results Prior research suggests that the impact of corporate governance on firm performance will be reflected in the relationship between measures of performance and governance variables. Smaller boards, independent directors, officers and directors holding significant shares of the firm, and having separate individuals serve as chairman of the board and chief executive officer have been associated with improved firm performance. Agency theory suggests that large blockholders may serve as substitutes for corporate governance in monitoring the actions of management. However, as noted by Hermalin and Weisbach (2003), there arises also the issue of endogeneity, particularly with variables relating to the CEO. One would predict that a CEO would have long tenure only if the firm were performing well, with firm performance explaining CEO tenure. A similar argument could exist, though perhaps not as strongly, for the demographics of the seated board of directors, whereby only those types of directors most beneficial to the firm s needs, be they insiders, outsiders or affiliates, would retain seats on the board. This study seeks to examine a wide range of variables and their impact on firm performance during a period of economic stress. The null hypothesis is the assertion of Demsetz and Lehn (1985) that corporate governance structure is a choice variable in equilibrium; as such, no correlation between performance and corporate governance measures is expected. 11

19 Using ordinary least squares (OLS) regression the corporate governance measures will be used to explain firm performance, where H 1 : Firm performance = f (governance characteristics) H 0 : Firm performance f (governance characteristics) D.1. Steel Industry The steel industry in the United States sustained an exogenous economic shock in the early 1990 s. The National Bureau of Economic Research (NBER) sets the beginning of the recession early in that decade as July 1990 and the ending as March 1991, though unemployment continued to rise through June of According to the Federal Reserve Bank of Boston (1999), manufacturing and construction firms bore the brunt of the downturn, shedding one million jobs between them. The production of cars and light trucks was hammered hard, dropping by 28%. The steel industry felt the impact of both the general economic downturn and the reduction in product demand from the manufacturing sector. The Federal Reserve Board s Seasonally Adjusted Steel Industry Production is charted in Figure Steel production dropped dramatically, falling over 23 percent between August 1990 and March Aggregate data from CRSP shows a sharp decrease in the value weighted market returns and industry sample returns from the second to the third quarter of 1990, as charted in Figure 1.2. The median returns for the industry sample are -22.2%; the return for the market is -15.5%. Using a sample of all firms listed in SIC 331, Steel Works, from CRSP, Compustat and Value Line Investment Survey, for which proxy data, Compustat data and CRSP data are all 7 The NBER defines a recession as beginning at the peak of the economic cycle and ending at the trough. 8 The Federal Reserve Board uses an X-12 ARIMA adjustment for seasonality. 12

20 available, I study the impact of corporate governance in the steel industry. Table 1.1 contains descriptive statistics for the industry and the sample for the years 1989 through Also included are descriptive statistics for the firms in the steel industry included in the Governance Index by Gompers, Ishii and Metrick. 9 The 24 steel industry firms listed in the Governance Index for 1990 are all in the research sample. The firms on the index are large, with total assets of 19 of the 24 greater than the median of the sample. Financial ratios are calculated from Compustat data, with mean and median values shown in the table. Market capitalization falls dramatically from 1989 to 1990, consistent with Figure 1.2. Total assets and net sales rise through the period, while return on assets falls. Operating margin falls throughout. Table 1.2 summarizes the proxy statement data statistics for the sample firms for The mean board size is nine members, with a maximum of fifteen and a minimum of four. 10 About half of the boards are elected annually; the remainder had staggered terms. Five of the firms had tiered boards in which certain board positions were pre-specified to represent specific groups, such as labor unions or preferred shareholders. Twelve of the firms allowed cumulative voting for directors, whereby a shareholder controlled votes equivalent to the number of shares held multiplied by the number of directors positions up for election. The shareholder could then choose to cast all votes for one director, or spread them among the open positions. Recognizing that firm performance is influenced by many factors, it is not expected that any single variable will exhibit strong explanatory power. An initial examination of the variables included ordinary least squares (OLS) regressions of firm performance on individual 9 Permission to use the Governance Index was granted by the authors and is gratefully acknowledged. For a complete discussion of the index, see Gompers, Ishii and Metrick, Corporate Governance and Equity Prices, The Quarterly Journal of Economics, February Board size in the industry decreased from 1990 through 1995 by an average of two-thirds of a person, though the median change was zero. The greatest changes were in large boards, such as ARMCO from 15 directors in 1990 to 8 by 1995, Bethlehem Steel from 12 directors in 1990 to 14 by Some large boards remained large; US Steel (later USX) had 15 directors in 1990 and 1995, with an increase to 16 in

21 firm and board characteristics. Table 1.3 shows the coefficient estimates of the regressions, with the dependent variable firm performance, measured as the 1990 return on assets (ROA) minus the industry median return on assets. 11 The 1990 ROA is calculated as operating income divided by total assets. The log of total assets and the log of market capitalization are used as independent variables to capture the impact of size differences among firms. On a univariate basis, the independent variable with the strongest ability to explain firm performance is size of the firm, as measured by the log of market capitalization. Market capitalization is significant and positive at the 1% level and results in an R 2 of Companies whose CEOs had six or more years of experience were positively related to firm performance at the 5% level (R 2 of 0.16), as might be expected given the potential endogeneity issue between CEO tenure and firm performance. Other variables of significance, though possessing small coefficients, are the total number of officers and directors (positive), the percentage of directors with specific industry experience (positive) and the percentage of affiliated votes (negative). Regression (1) of Table 1.4 contains variables prior research has suggested are important components of effective corporate governance, including the percentage of officer and director shares, whether the CEO is also the CoB, small boards (fewer than nine directors), the percentage of independent directors, and if the board has a majority of outside directors. Firm size and small boards were significant; the regression has an adjusted R 2 of Regression (2) shows the results of combining the independent variables that evidenced the most predictive ability of firm performance from the univariate analysis of Table 1.3. The variables of size, number of officers and directors, small boards, percentage of affiliated votes and the percentage of directors with industry experience retain their significance in multivariate regression (3), steel industry median ROA is Results from measuring performance using Return on Equity were similar. 14

22 resulting in an adjusted R 2 of Regression (4) shows a distillation of the results of a stepwise regression in SAS with requirement for retention in the regression set at The resulting significant variables are firm size, the existence of super voting rights, and the percentage of affiliated votes, with an adjusted R 2 of To attempt to quantify each firm s ability to successfully manage the economic shock and look beyond the influence of size evident in the regressions in both Tables 1.3 and 1.4, I next examine the change in firm performance from the year prior to the shock to the year ending after the shock. The industry median change in performance is subtracted from the firm s change in performance. 12 This variable more specifically identifies how the firm reacted to the change. ROA figures were obtained from Compustat for each firm for The ability to react successfully to an economic shock is related to different variables than those that relate to a point in time measurement of ROA. On a univariate basis, shown in Table 1.5, the variables with the greatest descriptive ability were the total number of blocks of 5% or more (negative), officer and director share holdings as a percentage of shares outstanding (negative), and the percentage of insiders on the board of directors (positive). The resultant values of R 2 ranged from 0.14 to Regression (1) in Table 1.6 contains variables previously identified in the literature as associated with effective corporate governance. The only significant variable was the percentage of officer and director holdings; the regression has an adjusted R 2 of Including the independent variables with the greatest predictive ability from Table 1.5 in a multivariate regression resulted in an R 2 of 0.28, with only the percentage of outside directors significant and negative. Further reduction in the number of variables resulted in a more parsimonious model of regression (3) with an adjusted R 2 of 0.24, with only the percentage of officer and director industry median change in ROA is

23 holdings significant and negative. Regression (4) is the distillation of the results of a SAS stepwise regression, resulting in an adjusted R 2 of 0.29, with the percentage of inside directors positive and the only significant variable. These results suggest that corporate governance does matter in hard times; though in the steel industry in 1990, it appears the presence of an outsider dominated board of directors, typically viewed as a stronger governance provision, decreases the firm s ability to react to change in the business environment. Table 1.7 shows the coefficient estimates from the ordinary least squares (OLS) regressions of firm performance on the Governance Index of Gompers, Ishii and Metrick. The first panel uses 1990 ROA as the dependent variable, similar to the regressions of Tables 1.3 and 1.4. The second panel uses the change in ROA between 1989 and 1990 as the dependent variable, similar to the regressions of Tables 1.5 and 1.6. The independent variable in regression (1) is the firm s numerical value (G) on the Governance Index. The independent variable in regression (2) employs the Gompers, Ishii and Metrick definition of Dictatorship as firms where the value of G is 14 or greater on a scale from 0 to 24, which comprised their highest decile in the index. These firms were classified as having the weakest shareholders rights. Neither method of employing the Governance Index was a good predictor of either measure of firm performance. To attempt to ascertain whether the unexpected impact of governance characteristics on firm performance in a period of economic stress relate beyond the early 1990 s and the steel industry, another shock in a different industry in a more recent period was examined. 16

24 D.2. Hotel Industry The dramatic impact of the terrorist attacks on September 11, 2001 caused difficulties throughout the economy, with the travel industry particularly hard hit. The airline industry was already facing the challenges of consolidation after deregulation. However, the hotel industry had been strong, weathering the beginning of the recession 13 with only a mild downturn. After September 11, 2001, many companies cancelled travel plans of executives, conventions were scaled back or eliminated, and leisure trips by the populace greatly diminished. All these factors impacted the hotel industry, with industry profitability falling. Figure 1.3 charts the change from the prior year for three descriptive measures typically used in the hotel industry: occupancy rate, average daily rate and revenue per available room. Occupancy rate is the number of rooms sold divided by the number of rooms available; average daily rate is the room revenue divided by the rooms sold; revenue per available room is the room revenue divided by the number of rooms available. 14 In September 2001, all three measures fell and did not begin to rise above that level (marked as the 0.0 line in the figure) until September Figure 1.4 charts the CRSP monthly value weighted market returns and the returns for the hotel industry sample. For September 2001, the market return was -9.2%; the hotel industry sample was -24.8%. Using a sample of all firms listed in SIC 710, Hotels and Motels and North American Industry Classification System (NAICS) 7211 Traveler Accommodation, for which proxy data, Compustat data and CRSP data are all available, I study the impact of corporate governance on performance in the hotel industry after the economic shock of September 11, Table 1.8 contains descriptive statistics for the industry and the sample for the years 2000 through NBER sets the beginning of the 2001 recession as March 2001 with an end of November The data was provided by Smith Travel Research and is gratefully acknowledged. 17

25 Also included are descriptive statistics for the firms in the hotel industry included in the Governance Index by Gompers, Ishii and Metrick. The 13 hotel industry firms listed in the Governance Index for 2000 and 2002 are in the research sample. The firms on the index are large, with median total assets in excess of ten times greater than the median of the sample. The industry sample contains many small firms, as indicated by the vast difference between mean and median firm size. All initial regressions will include firm size as an independent variable to attempt to mitigate any impact of firm size on the results. Financial ratios are calculated from Compustat data, with mean and median values shown in the table. Market capitalization falls from 2000 to 2001, rising in 2002 but still below the values of 2000; total assets follows the same path. Net sales and return on assets evidence a dramatic decline in 2001 with only a slight increase to Operating margin falls throughout. Table 1.9 summarizes the proxy statement data statistics for the sample firms for As with the steel industry sample, the mean board size is nine members, with a median of eight. The maximum board size is twenty and minimum is three. The percentage of inside directors is similar to that of the steel industry sample, as is the mean and median CEO age of 58 and 57. In both samples, the majority of CEOs also hold the position of chairman of the board (60% for steel, 69% for hotels.) The hotel industry has a greater variance in insider holdings and ownership blocks than does the steel sample. Additionally, one firm has 98.9% of its votes controlled by directors and officers. Table 1.10 shows the coefficient estimates from the ordinary least squares (OLS) regressions of firm performance on board characteristics. The dependent variable in the regression is firm performance, measured as the 2002 return on assets (ROA) minus the industry 18

26 median return on assets, 15 calculated as operating income divided by total assets. 16 The log of total assets and the log of market capitalization are used as independent variables to capture the impact of size differences among firms. On a univariate basis, the independent variables with the strongest ability to explain firm performance are firm size (positive), the percentage of inside directors (negative), and directors with industry specific experience (positive). The resulting values of R 2 ranged from 0.14 to Regression (1) of Table 1.11 shows the results of regressions of variables typically associated with strong corporate governance. Firm size is the only significant variable, resulting in a regression with an adjusted R 2 of Regression (2) combines the independent variables that evidenced the most predictive ability of firm performance from the univariate analysis of Table While regression (2) has an adjusted R 2 of 0.26, there are no significant variables. When the number of independent variables is reduced in regression (3), the adjusted R 2 increases to 0.41 and the percentage of directors with industry experience gains significance at the 10% level. Regression (4) shows the distillation of the results of a SAS stepwise regression, with small boards positive and significant at the 10% level and the percentage of insiders negative and significant at the 1% level. The resulting regression has an adjusted R 2 of I next examine the change in firm performance from the year prior to the shock to the year ending after the shock, with the industry median change in performance subtracted from the firm s change in performance. 17 This variable more specifically identifies how the firm reacted to the change. Tables 1.12 shows the coefficient estimates from the OLS regressions of change in ROA from 2001 to 2002 for each independent variable hotel industry median ROA is Results from measuring performance using Return on Equity were similar industry median change in ROA is

27 The ability to react successfully to an economic shock is very different than the variables predictive of firm performance in the univariate regressions shown in Table 1.10 that relate to a point in time measurement of ROA. When explaining the change in ROA, the percentage of inside directors is significant, as it was in explaining ROA. However, that variable now carries a positive coefficient, as opposed to the negative coefficient evidenced in the regression of ROA. No other variables possess significant predictive capabilities. Table 1.13 shows four multivariate regressions. Regression (1), with an adjusted R 2 of 0.10, has a negative and significant coefficient for the variable indicating that the CEO is also the CoB. In Regression (2), the percentage of inside directors is positive and significant, while the percentage of block votes is negative and significant, resulting in an adjusted R 2 of Reducing the number of independent variables results in regression (3), with an adjusted R 2 of Regression (4), a distillation of the SAS stepwise regression, has the independent variables staggered board and percentage of inside directors both significant at the 1% level, with a resultant adjusted R 2 of Table 1.14 shows the coefficient estimates from the ordinary least squares (OLS) regressions of firm performance on the Governance Index of Gompers, Ishii and Metrick. The first panel uses 2002 ROA as the dependent variable; the second panel uses the change in ROA between 2001 and 2002 as the dependent variable. The independent variable in regression (1) is the firm s numerical value (G) on the Governance Index. The independent variable in regression (2) employs the Gompers, Ishii and Metrick definition of Dictatorship as firms where the value of G is 14 or greater on a scale from 0 to 24, which comprised their highest decile in the index. These firms were classified as having the weakest shareholders rights. As was the case with the 20

28 steel industry, neither method of employing the Governance Index was a good predictor of either measure of firm performance. E. Conclusion Using a data set of firms in the steel industry, I relate firm and corporate governance characteristics and chief executive officer demographics to firm performance after the economic shock of the recession in Return on assets is significantly related to firm size and negatively related to the percentage of votes held by affiliates of the firm. To assess a firm s competence in adapting to the changing business environment, I measure the difference between firm ROA before and after the shock, and then subtract the steel industry median change in ROA. The change in ROA reveals those firms that were able to minimize the reduction in return on assets. In these regressions, the variable most significantly related to successful performance, and possessing a negative coefficient, is the percentage of officer and director holdings. The percentage of insiders on the board is positively related, and boards with a majority of outsiders is negatively related. The Governance Index of Gompers, Ishii and Metrick (2003), a valuable tool in large multi-industry studies, adds no additional insight into the relationship between governance characteristics and firm performance in the steel industry in In the hotel industry, ROA is positively related to boards of fewer than nine directors and negatively related to the percentage of inside directors. When measuring the change in ROA before and after the shock, the percentage of inside directors again holds a predictive role, though this time as a positive coefficient. The reason for the change in sign is unclear. Directors holding staggered terms is also a positive variable. As with the steel industry, the Gompers, Ishii and Metrick (2003) Governance Index results in no significant indicators of firm performance. 21