

small (250x250 max)
medium (500x500 max)
Large
Extra Large
large ( > 500x500)
Full Resolution


EXAMINATION OF EXECUTIVE COMPENSATION DETERMINANTS IN THE HOSPITALITY INDUSTRY: A QUANTILE REGRESSION APPROACH By SANG HYUCK KIM Bachelor of Science in Business Administration DongGuk University Seoul, Korea 2000 Master of Business Administration Western Illinois University Macomb, Illinois 2002 Master of Science in Accountancy University of Illinois UrbanaChampaign, Illinois 2003 Submitted to the Faculty of the Graduate College of the Oklahoma State University in partial fulfillment of the requirements for the Degree of DOCTOR OF PHILOSOPHY July, 2008 ii EXAMINATION OF EXECUTIVE COMPENSATION DETERMINANTS IN THE HOSPITALITY INDUSTRY: A QUANTILE REGRESSION APPROACH Dissertation Approved: Dr. Jerrold K. Leong Dissertation Adviser Dr. Woo Gon (Woody) Kim Dr. Murat Hancer Dr. William D. Warde Dr. A. Gordon Emslie Dean of the Graduate College iii DEDICATION This dissertation is dedicated to my parents , Jong Ho Kim, and Myung Hee Park. iv ACKNOWLEDGMENTS There are several people without whose support and encouragement I could not have completed the doctoral degree. I offer my sincere appreciation to my advisor, Dr. Woody (Woo Gon) Kim, for his guidance, encouragement, understanding, and friendship throughout my studies. He made himself readily available for consultation and offered constructive criticism, and his steady support and encouragement have been critical during my studies. I also thank my committee chair, Dr. Jerrold K. Leong, for his guidance, patience, understanding, and support. I sincerely appreciate my other committee members, Dr. Murat Hancer and Dr. William Warde, who greatly assisted me by providing the feedback necessary to bring this study together. I am very grateful to my parents in Korea, who have given me their love and support during my graduate studies in the United States. None of my work would have been possible without their devotion and encouragement. Finally, I thank my wonderful wife, Soo Lyun Cho, for her love, support, and encouragement throughout my academic career. v TABLE OF CONTENTS Chapter Page I. INTRODUCTION......................................................................................................1 1. Background of study............................................................................................1 2. Research Motive and Problem Statement ............................................................5 Research Motive ................................................................................................5 Problem Statement .............................................................................................6 3. Significance of the Study.....................................................................................7 4. Purpose of the Study ............................................................................................8 5. Organization of the Study ....................................................................................9 II. REVIEW OF LITERATURE..................................................................................10 1. The Agency Theory ...........................................................................................11 2. Financial Determinants of Executive Compensation.........................................13 3. Managerial Power Determinants of Executive Compensation ..........................23 4. OLS regression and Quantile regression............................................................28 5. Development of Hypotheses ..............................................................................32 III. METHODLOGY ...................................................................................................36 1. Data Collection and Sampling Procedures ........................................................36 2. Variable selection...............................................................................................41 Selection of dependent variable .......................................................................41 Selection of independent variables ..................................................................42 3. Data Analysis and Model...................................................................................46 vi IV. FINDINGS.............................................................................................................49 1. Description of Sample........................................................................................49 2. Findings of Ordinary Least Square (OLS) Regression Method.........................57 Results of OLS regression method for the Full sample ...................................58 Results of OLS regression method for the Hotel & Casino subsample..........60 Results of OLS regression method for the Restaurant subsample..................61 3. Findings of the Quantile Regression..................................................................63 Results of the quantile regression method for the Full sample........................64 Results of the quantile regression method for Hotel & Casino subsample ....69 Results of the quantile regression method for Restaurant subsample ............74 V. CONCLUSION......................................................................................................79 1. Summary of the study ........................................................................................79 Summary of the Full Sample: The Hospitality Industry..................................80 Summary of the Hotel & Casino Subsample..................................................83 Summary of the Restaurant Subsample..........................................................85 2. Implication of study ...........................................................................................88 3. Limitations and Suggestions for Future Research .............................................92 REFERENCES ............................................................................................................94 APPENDICES ...........................................................................................................100 APPENDIX A. EXECUTIVES IN HOTEL INDUSTRY (2006) .......................101 APPENDIX B. EXECUTIVES IN CASINO INDUSTRY (2006) .....................103 APPENDIX C. EXECUTIVES IN RESTAURANT INDUSTRY (2006)..........106 vii LIST OF TABLES Table Page 21. The classification of financial variables from the previous studies...................22 22. The variables of managerial power approach used in previous studies ............28 31. Classification of study sample...........................................................................41 41. Descriptive Statistics of Executive Total Cash Compensation .........................50 42. OLS regression summary for the Full sample (all hospitality companies) .......59 43. OLS regression summary for the Hotel & Casino subsample .........................60 44. OLS regression summary for the Restaurant subsample .................................62 45. Quantile regression summary for the Full sample (all hospitality companies) .67 46. Quantile regression summary for the Hotel & Casino subsample ...................72 47. Quantile regression summary for the Restaurant subsample ...........................77 viii LIST OF FIGURES Figure Page 41. Scatter plot for full sample (All hospitality industry)........................................51 42. Scatter plot for subsample (Hotel & Casino industy) ......................................52 43. Scatter plot for subsample (Restaurant industry) .............................................52 44. Histogram and Normal PP Plot for full sample (All hospitality industry).......54 45. Histogram and Normal PP Plot for subsample (Hotel & Casino industry) ....55 46. Histogram and Normal PP Plot for subsample (Restaurant industry) ............56 47. The coefficient graphs of all hospitality industry by Quantile regression.........68 48. The coefficient graphs of Hotel & Casino subsample by Quantile regression 73 49. The coefficient graphs of Restaurant subsample by Quantile regression ........78 1 CHAPTER I INTRODUCTION 1. Background of study In the last few decades, the topic of executive compensation has received a great deal of attention from both academic empirical researchers and practitioners of business management, especially those from the finance and accounting fields (Andjelkovic, Boyle, & McNoe, 2002; GrabkeRundell & GomezMejia, 2002; Gray & Cannella, 1997). The dominant topic of executive compensation studies has focused on examining the relationship between the executive’s compensation and the firm’s performance (Mishra, McConaughy, & Gobeli, 2000; Perry & Zenner, 2001). That is, executive compensation studies have been conducted on the basis of the payforperformance rule in the agency theory (GrabkeRundell & GomezMejia, 2002). According to the agency theory, proposed by Jensen and Meckling (1976), compensation packages should balance compensation value and the executive’s managerial performance by implementing and utilizing an appropriate payforperformance rule aligned to motivate the agent (the executive in this study), to attract and retain management talent, and to increase management performance in order to maximize shareholder wealth (Gu & Choi, 2004; 2 Kim & Gu, 2005; Perlik, 2002). In other words, the agent’s compensation contract should lead the executives of the firms to try to increase the firm’s performance, thereby achieving the goal of maximization of shareholder’s wealth through an increase in the firm’s stock price and a stable flow of dividends (Lippert & Porter, 1997; Grabke Rundell & GomezMejia, 2002). The payforperformance rule, then, supports the idea that the level of an executive’s compensation should be closely and positively linked to the firm’s performance (Hallock, 1998; Jensen & Murphy, 1990; Kato & Kubo, 2006). Because of theoretical confidence in the payforperformance rule, it has become an increasingly popular measure in agency theory research (Lippert & Porter, 1997; Perry & Zenner, 2001). Even so, the payforperformance rule has not always been supported by the empirical results of executive compensation studies (Andjelkovic et al., 2002; Gray & Cannella, 1997). As the numbers of studies that have found other affects on executive compensation, such as executive’s demographic characteristics and the structure of corporate governance, have increased, the support for the payforperformance rule has decreased (GomezMejia, Tosi, & Hinkin, 1987; Hebner & Kato, 1997; Nelson, 2005). In addition, the increasing attention on payforperformance among the public stimulated the development of regulations in the United States (Perry & Zenner, 2001). The United States Securities and Exchange Commission (SEC) announced a new compensation disclosure rule, beginning with the fiscal year of 1992, which required publicly held companies to include top executives’ compensation disclosures in annual proxy statements to the SEC (Vafeas & Afxentiou, 1998). Congress also established tax legislation, Section 162(m) of the Internal Revenue Code, to limit executive’s 3 compensation’s deduction for nonperformancerelated executive compensation to US$1 million in the publicly traded companies (Perry & Zenner, 2001). Both the SEC regulation and the tax legislation were expected to help determine clearer and more appropriate levels of executive compensation in U.S. publicly traded companies by encouraging companies to relate compensation to company performance (Perry & Zenner, 2001; Vafeas & Afxentiou, 1998). Today, the compensation packages of executives in publicly traded companies still have been spent huge amounts of money and have continually increased in value in order to attract and retain executives. For example, in 2006, Goldman Sachs’ CEO, Lloyd Blankfein, received compensation totaling $55 million in cash, stock options and restricted stock, a 76% increase in pretax compensation from 2005. John Mack, CEO of Morgan Stanley, increased his compensation to $41 million in 2006, a 43% increase from 2005 (Hahn, 2007). In the retail industry, George L. Jones, the president and CEO of book retailer Borders Group, Inc., received $3.37 million in compensation during fiscal year 2006 (Financial Times Information, 2007b). Contrary to above examples of increase in top executive compensation, some top executives have voluntarily reduced their annual salaries, sometimes drastically. For example, Roger Enrico, CEO of PepsiCo, dropped his $900,000 salary to $1 in 1998, 1999 and 2000 and donated his previous salary to scholarships for employees’ kids. Steve Miller, CEO of Delphi, dropped his salary from $1.5 million to a $1 after the company filed for bankruptcy protection. Rick Wagoner, GM’s CEO, cut his salary almost 50% in 2005 and volunteered for another 50% cut in his $2.2 million salary in 2006 (Kempner, 2007). Although some examples show that top executive’s compensation is decreased by several reasons, it is true that most industries 4 still pay huge amounts of money to acquire and keep talented—and sometimes notsotalented— executives. While the value of executives’ compensation packages can vary in response to such factors as firm performance, the structure of other companies’ compensation packages, and voluntary cuts by the executive himself or herself, questions about efficient and appropriate executive compensation packages in publicly traded firms have increased as executive compensation has increased (GrabkeRundell & Gomez Mejia, 2002). Since the agency theory was proposed by Jensen and Meckling (1976), numerous studies have been undertaken to find determinants of executive compensation. At the initial stage of executive compensation research, most studies were concerned with determining how executive compensation relates to financial performance standards (Carr, 1977; Core, Holthausen, & Larcker, 1999; Firth, Tam, & Tang, 1999). As mentioned earlier, the payforperformance rule has not always been supported by the empirical results of executive compensation studies. Thus, some researchers made efforts to extend executive compensation study by adding other factors, especially factors from managerial power approach (Yermack, 1995; Core et al., 1999; Hallock, 1997; Bebchuk & Fried, 2003; Grinstein & Hribar, 2004). Grinstein and Hribar (2004) stated the “managerial power approach,” which presents that compensation based on the payforperformance rule did not work optimally to enforce agents to maximize shareholder wealth if the agent had powers or influence over board decisions, including compensation decisions (Grinstein & Hribar, 2004). Based on the concept of the managerial power approach, researchers who questioned the payforperformance rule found that the characteristics of ownership structure and corporate governance also affected executive 5 compensation. As a result, research that investigates determinants of executive compensation should include variables such as ownership structure, number of board members, and whether the executive is on the company’s board of directors, so that both the payforperformance rule and the managerial power approach are considered in finding the determinants of executive compensation. 2. Research Motive and Problem Statement Research Motive The hospitality industry is not much different from other industries when it comes to compensation for top executives. For example, in 2004, Starwood Hotels & Resorts appointed Maven Steven Heyer, former president and COO of the CocaCola Co., as its new CEO, with a $1 million annual base salary for a fouryear initial term (Parets, 2004). The total compensation of David Brandon, CEO of Domino’s Pizza, Inc., increased from $1.81 million in 2004 to $21.9 million in 2006 (Snavely, 2006). The CEO of McDonalds, Jim Skinner, received $8.8 million in bonuses from 2004 to 2006 (Financial Times Information, 2007a). As in other industries, not all top executives in the hospitality industry have received huge compensation. The CEO of Planet Hollywood International, Robert Earl, was paid half of his $600,000 annual salary in 2001 because of bankruptcy (Schneider, 2007). Tim Taft was appointed as the new CEO of Pizza Inn with a firstyear salary of $1, although he received stock options (RobinsonJacobs, 2005). These are examples of 6 the hospitality industry’s following the payforperformance rule, although there are many exceptions. For example, when Denny’s restaurant faced a loss of $88.5 million in company earnings before interest and taxes in 2002, the CEO received a $1.3 million bonus (Perlik, 2002). On the other hand, Joseph P. Martori, CEO of ILX Resorts Incorporated, named the numbertwo topperforming CEO in HVS International’s 2002 Survey and beating out the CEOs of the Four Seasons, Marriott International, Starwood, Hilton and others, was one of the lowestpaid CEOs in the hotel industry, ranking 45th among 51 hospitality industry CEOs (Business Wire, 2003). Sometimes, then, the payfor performance rule does not explain the determinants of top executives’ compensation in hospitality industry well. Clear understanding of the determinants of executive compensation is necessary for stockholders or potential investors in the hospitality industry to judge whether the appropriate compensation is awarded. Although the hospitality companies have spent large amounts on executives’ compensation packages, little research has been done to investigate how that compensation is determined in the industry. Previous literature related to executive compensation in the hospitality and tourism field has examined the determinants of CEO’s compensation only with regard to either financial variables from the firm’s performance (Gu & Choi, 2004; Kim & Gu, 2005) or to gender difference (Skalpe, 2007). Problem Statement Although previous studies have expanded our knowledge of what determines executive compensation in the hospitality industry, it remains uncertain whether 7 hospitality companies follow only the payforperformance rule or whether other factors have an influence on determining executive compensation in the hospitality industry. 3. Significance of the Study Most literature related to executive compensation in the hospitality field has focused on financial determinants from the payforperformance rule (Gu & Choi, 2004; Kim & Gu, 2005; Skalpe, 2007). While the ownership structure and/or corporate governance variables from the managerial power approach have also been considered to be among the determinants of executive’s compensation for academic fields and other industries, to the best of my knowledge, there is no study that has considered whether the managerial power approach is a determinant of executive compensation in the hospitality industry. Therefore, this study combines the payforperformance rule and the managerial power approach, using both the financial variables from the payforperformance rule and the ownership and corporate governance variables from the managerial power approach, to investigate the determinants of executive compensation in the hospitality industry. In addition, Dyl (1988) found that the different types of industry influence on determining management compensation level. Other researchers also adopted a type of industry as dummy variable in their studies to examine whether the different type of industry influences on the level of the executive compensation (Dyl, 1988; Hallock, 1997; Yermack, 1995). Thus, this study also attempts to examine whether there is a difference 8 between different sectors (i.e., the hotel & casino vs. restaurant) in the hospitality industry regarding determinants of executive compensation. In terms of methodology, most research on executive compensation has used traditional multiple regression, such as Ordinary Least Square (OLS) regression and Weighted Least Square (WLS) regression analysis, to investigate the relationship between total cash compensation and financial variables from the payforperformance rule. The current study adopted Quantile regression analysis, which was developed by Koenker and Basset (1978), to allow examination of whether different levels of total cash compensation are related differently to each independent variable from the payforperformance rule and the managerial power approach. Unlike traditional multiple regression analysis, Quantile regression analysis is operated by a conditional quantile function that estimates the relationship between each independent variable and each segment (quantile) of the dependent variables. For this study, then, Quantile regression will allow us to investigate how each independent variable is related to each different segments of the executive’s total cash compensation. 4. Purpose of the Study The primary objective of this study is to examine whether the financial performance variables from the payforperformance rule and a company’s ownership and corporate governance structure from the managerial power approach is related to 9 executive compensation in the hospitality industry. More specifically, the purpose of this study is to: 1) Identify the determinants for executive compensation in the hospitality industry in terms of both the payfor performance rule and the managerial power approach; 2) Examine whether there is a difference between different sectors (i.e., the hotel & casino and restaurant) in the hospitality industry regarding determinants of executive compensation; and 3) Investigate whether different levels of executive compensation are differently related to selected variables from both the payforperformance rule and the managerial power approach in the hospitality industry. 5. Organization of the Study The composition of this study is as follows. Chapter I, the introduction section, presents the background, research motives, significance, and the purpose of the study. Chapter II, the literature review section, reviews previous literature dealing with agency theory, executive compensation with the payforperformance rule and managerial power approach, comparisons between OLS regression and Quantile regression analysis, and development of hypotheses for this study. Chapter III explains the research methodology, including data collection, sampling procedures, and data analysis and models. Chapter IV addresses the empirical results of the study and, finally, Chapter V concludes and discusses the study’s implications, contributions, and limitations. 10 CHAPTER II REVIEW OF LITERATURE Numerous studies have tried to verify what factor(s) determine executive compensation. Two main streams of research concerning executive compensation in the finance and accounting fields have emerged over the last 70 years. The fundamental difference between the two streams of research lies in the theoretical foundations of executive compensation: the payforperformance rule from agency theory, which focuses on the relationship between executive compensation and firm performance; and the ownership structure and corporate governance from managerial power approach, which emphasizes that, because the payforperformance rule does not always work to determine executive’s compensation, other factors, such as whether the executive is involved in company ownership, board size, and whether the executive is a board member, also influence the executive’s compensation. While neither approach is perfect in explaining what determines executive compensation, they each have their advantages and disadvantages. A current trend in the literature is to combine both approaches. The goal of this literature review is to address the previous studies regarding the effect of financial determinants and managerial power on executive compensation and to identify relevant variables and methodologies used in previous studies. This chapter has 11 five main sections. The first section summarizes the agency theory, proposed by Jensen and Meckling (1976). The next two sections present the prior studies of executive compensation determinants from the payforperformance and the managerial power approaches, respectively. The fourth section compares OLS regression with Quantile regression for this study. The last section proposes the hypotheses for the study. 1. The Agency Theory In traditional financial theory, the primary goal of business management is maximization of stockholder wealth in terms of maximization of the firm’s market value. Because of this, those who invest money in the company expect executives not only to improve their business processes but to increase the company’s value (Brigham, Gapenski, & Ehrhardt, 1999), so spending money for executive compensation is one of the investments shareholders make to maximize their wealth. In contrast to the traditional financial theory, Jensen and Meckling (1976) proposed “the agency problem” within the agency theory, which is that there is a conflict between the agent’s interests and the shareholders’ interests because of the separation of management from ownership (Jensen & Meckling, 1976). In other words, the agent (the executives, in this case) is more likely to pursue personal interests or goals than to maximize shareholders’ wealth (Dyl, 1988; Jensen & Meckling, 1976; Traichal, Gallinger, & Johnson, 1999). This conflict between agent and shareholder evokes several types of costs for shareholders. This “agent cost” is composed primarily of three types of 12 costs: monitoring cost, bonding cost, and residual loss. Monitoring cost is the cost for the principal (the shareholder in this case) to limit the discretionary behavior of the agent. Bonding cost refers to the costs for the agent (the executive, in this case) to guarantee his or her discretionary behavior. Residual loss is loss from conflicts between principals and agents (Dyl, 1988; Jensen & Meckling, 1976). Several remedies have been proposed to solve the agency problem, including monitoring the agent’s discretionary behavior and controlling the agent’s compensation packages (Dyl, 1988; Jensen & Meckling, 1976; Traichal et al., 1999) by providing sufficient agent compensation to motivate the executive to work toward the best interests of shareholders, i.e., achieving maximization of shareholder wealth (Kim & Gu, 2005). Jensen and Meckling (1976) also suggested that executive compensation could be determined by means of the payforperformance rule, which would establish an optimal compensation contract between agent and principal. According to the payforperformance rule in agency theory, the agent’s compensation should be determined by practical and reliable measures of firm performance, that is, the level of the agent’s compensation would be commensurate with his or her contribution to the firm’s value. In this context, compensation should be based on observable measures, such as market returns or profitability ratios, which maximize the value of a firm (Grinstein & Hribar, 2004). The payforperformance rule is frequently utilized as a standard by which to determine executive compensation by using firm performance (Gu & Choi, 2004). Thus, many extant studies have investigated the relationship between executive compensation and firm performance using several key financial variables, including firm size and 13 several types of firm performance measures (Anderson, Becher, & Campbell, 2004; Grinstein & Hribar, 2004; Gu & Choi, 2004, Kato & Kubo, 2006; Kim & Gu, 2005). 2. Financial Determinants of Executive Compensation Most executive compensation studies have adopted the firm’s performance in terms of the firm’s financial data, as estimators of each executive’s performance because of the difficulty of collecting relevant or sufficient data regarding executives’ direct contribution on firm performance. In other words, it is difficult to estimate each executive’s direct performance on the firm’s performance with financial and mathematical figures. The agency theory also suggests that executives’ managerial performance leads to improvement in the firm’s performance, which, in turn, links to increasing shareholder wealth (Gu & Choi, 2004). At the initial stage of executive compensation research, especially after Jensen and Meckling proposed the agency theory in 1976, the financial measures from the payfor performance rule were probably the most common measure utilized in research on determinants of executive compensation (Bebchuk & Fried, 2003; GomezMejia &Wiseman, 1997). Even though some critics decry the financial measures from the payfor performance rule, numerous studies have utilized the financial measures for firm performance. Thus, following these prior studies provides the theoretical justification for the current study to utilize relevant variables for measuring firm performance. 14 To measure the firm’s performance, researchers have adopted several types of financial figures as relevant variables. The three dominant streams of measurement in prior empirical executive compensation studies in the finance and accounting fields are marketbased measurements, accountingbased measurements, and growthbased measurements. First, the company’s market return in terms of stock returns is an indirect measure of a firm’s marketbased performance, because it is an important indicator of its business performance and shareholder wealth (Jensen & Murphy, 1990; Leone, Wu, & Zimmerman, 2006). Financial figures such as return on assets, earnings per share, and return on equity, are accountingbased measures of firm performance. The accountingbased ratio analysis is one of the tools used by financial managers and financial analysts to evaluate the financial position or performance of a firm (Chatfield & Dalbor, 2005). Finally, many studies have utilized the firm’s sales growth as a growthbased determinant of executive compensation (Firth et al., 1999; Kato & Kubo, 2006). Many prior studies of executive compensation based on the payforperformance rule have also used firm size as a control variable (Andjelkovic et al., 2002). Studies that followed often adopted these four financial measurements of firm performance in executive compensation research. Jensen and Murphy (1990) examined the association between top management’s pay and performance, adopting shareholder wealth in terms of stock returns as an estimator of managerial performance. The study found that top management compensation is highly sensitive to the stock returns of the company. One other example of a study regarding the sensitivity of stock returns on executive compensation was conducted by Leone et al. (2006). The authors examined the sensitivity of CEO cash compensation to stock returns with 9,858 CEOs in the ExecuComp database from 1993 to 15 2003 and found that CEO cash compensation is twice as sensitive to negative stock returns as it is to positive stock returns. The results supports that company’s stock return positively influence on determining CEO cash compensation. In addition, the reducing amount of CEO cash compensation in company with negative stock return is bigger than the increasing amount of CEO cash compensation in company with positive stock return. Gray and Cannella (1997) examined the role of firm’s risk in executive compensation, using several financial figures to identify the relationship between executive compensation and firm risks, return on sales, and firm size. The results of the study provided that firm risks have a significantly negative relationship with executive total compensation and firm size, and Jensen’s alpha has a significantly positive association with executive total compensation. The findings from this study supports that the executive compensation is determined by firm’s performance. Furthermore, the executive compensation is reduced when firm’s risk increases, as well as the executive compensation is increased when firm’s size and firm’s performance increase. Duru and Iyengar (1999) conducted a crosssectional research analysis of 225 firms in the electric utility industry (SIC code 4931) from 1992 to 1995 to examine the association between firm performance and CEO compensation components. The authors adopted the change in CEO compensation as the dependent variable and the changes in the firm’s financial figures as multiple independent variables to examine the sensitivity of CEO compensation to changes in firm performance. They used market returns, return on assets, earning per share, operating cash flow per share, and growth in sales to measure financial performance and showed a positive relationship between changes in compensation and changes in firm performance. More specifically, executive bonuses 16 were sensitive to changes in market return, and executive stock options were sensitive to changes in sales growth. Several studies have examined the relationship between executive compensation and firm performance during special events, like mergers and acquisitions (M&A). Anderson et al. (2003) investigated bank CEOs’ managerial incentives for bank mergers as they related to financial variables such as firm size and stock returns, and found that CEO compensation was in line with an increase in bank size, regardless of whether a merger or acquisition created value. Grinstein and Hribar (2004) also used financial variables, including firm size and ROA, stock return, and sales growth, to examine the determinants of CEOs’ bonuses for 327 large ($1 billion or more) M&A deals in publicly traded U.S. companies between 1993 and 1999. They found the firm size, ROA, stock return, and the acquisition dummy to be positively correlated with CEOs’ bonuses for the M&A deal. Some studies of executive compensation determinants have been performed in countries outside the U.S., such as Japan (Kato & Kubo, 2006), England (Eichholtz, Kok, & Otten, 2008), and China (Firth et al., 1999). Kato and Kubo (2006) examined the relationship between executive compensation and firm performance for Japanese firms from 1986 to 1995 using marketbased firm performance (stock returns), accountingbased firm performance (return on asset), growthbased firm performance (sales growth), and firm size. The results of this study supported that Japanese CEOs’ cash compensation was sensitive to firm performance (especially accountingbased performance) and that the bonus system made CEO compensation more responsive to firm performance. Eichholtz et al. (2008) used samples from 39 companies in the UK property industry from 1998 to 17 2003 and variables from both firm performance and corporate governance—total stock performance, Jensen alpha, earnings per share, dividend yield, and discount—to investigate the association between executive compensation and firm performance. They found that stock performance, Jensen alpha, earnings per share, and discount were not significantly related to executive cash compensation but that dividend yield was significantly negatively related to executive cash compensation. Thus, the study found a weak association between executive cash compensation and the payforperformance rule. Firth et al. (1999) used a sample of companies in Hong Kong and several variables from both the payforperformance rule and the managerial power approach— annual stock return, firm size, return on shareholder equity, and annual compound sales growth—and found that the companies in Hong Kong followed payforperformance rule, by showing that company size and accounting profitability are significantly related with executive compensation. Thus, executive compensation studies of three different countries cautiously supported an association between executive compensation and firm performance. Most research has measured firm performance with marketbased, accountingbased (mostly profitability measures), and growthbased measures, as well as a control variable for firm size. However, other types of accounting based financial ratios have been used to represent for firm performance in the accounting literature. The basic accountingbased financial ratios are generally divided into four categories: liquidity, activity, profitability, and coverage. Liquidity ratios are used to measure the company’s shortrun ability to pay its maturing obligations, activity ratios measure how effectively 18 and efficiently a company uses its assets, profitability ratios are measures of the degree of success or failure of company for a given period of time, and coverage ratios measure the protection of longterm creditors and investors (Brigham et al., 1999; Chatfield & Dalbor, 2005; Gallagher & Andrew, 1997; Kieso, Weygandt, & Warfield, 2001). Although most studies have adopted profitability ratios for firm performance from among the four classified accounting based ratio analyses, other ratios, like liquidity, activity and coverage ratios, might also be considered as measures of firm performance for the current study. OrtizMolina (2007) stated that executive compensation may not only depend on shareholder opinion, because the bondholders (debtors) but also have the power to influence executive compensation. Thus, OrtizMolina found that executive compensation was significantly sensitive to the types of debt in a company. Traichal et al. (1999) affirmed the importance of liquidity ratios and coverage ratios in executive compensation and adopted a modified liquidity ratio (ratio of shortterm debt divided by total assets) and coverage ratio (ratio of longterm debt divided by total assets). Since those two liquidity and coverage ratios are measurements that show the ability of a firm to pay back both shortterm and longterm debt, those two ratios may also be considered measurements of firm performance. In addition, some studies of executive compensation in the hospitality field have included all four types of accountingbased financial measurements (liquidity, efficiency, profitability, and coverage) as independent variables for their executive compensation studies (Gu & Choi, 2004; Kim & Gu, 2005). In a related study in the hospitality field, Cauvin (1979a) investigated the relationship between executive total compensation and company size, represented by sales, with 33 19 lodging companies in the U.S. Since the study was conducted before the SEC’s regulation requiring disclosure of top executive compensation disclosure was announced, the data for this study were collected from surveys based on the 1978 directory of Hotel/Motel Systems, published by the American Hotel and Motel Association. The author found that the relationship between executive total compensation and company sales was not similar, unlike the results from the studies in other industries. The results indicated that the executive total compensation in small hotel companies was equal to or more than the executive total compensation in large hotel companies. Cauvin investigated the relationship between executive compensation and company sales again in 1979, this time conducting nine simple regressions to examine the relationship between executive compensation in each of nine executive positions. The results showed that there was less relationship between executive compensation and company sales in the lodging industry than in other fields (Cauvin, 1979b). More recently, Gu and Choi (2004) researched the determinants of CEO compensation in the casino industry, using several types of financial measurements for firm performance: marketbased firm performance (annual change of stock price), accountingbased firm performance (return on assets for firm profitability, asset turnover ratio for firm efficiency, longterm debt ratio for firm debt leverage), and firm size (total assets). The results supported that profitability, firm size, debt leverage, and stock options were positively related to CEO cash compensation, while revenue efficiency (i.e., asset turnover) was negatively correlated. Kim and Gu (2005) also studied the determinants of CEO cash compensation in the restaurant industry based on the payforperformance rule using firm size, sales growth, ROI, and stock returns. The authors found that CEOs’ cash 20 compensation was positively associated with firm size and operating efficiency, while growth, debt leverage, profitability, and stock performance were not related. In addition, Namasivayam, Miao, and Zhao (2007) investigated the relationship between compensation and firm performance for 1,223 hotel companies in the U.S. using data gathered from the Hospitality Compensation and Benefit Survey of Smith Travel Research in 2001 to 2003. Unlike other studies of executive compensation in the hospitality industry or other industries, the authors adopted RevPar (Revenue per available room) as the hotels’ performance measurement. The results showed that both individual salary and benefits were significantly positively related to hotel performance for both management and nonmanagement employees. For the tourism industry, Skalpe (2007) examined the differences in CEO pay between Norway’s tourism and manufacturing industries with regard to the CEOs’ gender and age, as well as financial variables that included firm size and firm performance. The study found that there was a difference in CEO pay between genders in both industries, although the smaller companies showed a greater difference. A difference in salary by gender is particularly significant for the tourism industry because more female CEOs work in the tourism industry than in the manufacturing industry. As a result of extensive literature reviews of studies in the accounting and finance literature on executive compensation based on payforperformance rule, four major categories for measuring firm’s performance can be identified: marketbased firm performance, accountingbased firm performance, growthbased performance, and firm size. Table 21 shows a summary of the financial variables used in prior studies of executive compensation determinants. These financial variables can be utilized as the 21 basis by which select relevant variables of financial determinants for executive compensation in the current study. 22 Table 21. The classification of financial variables from the previous studies Type Variable Studies Total Asset (TA) Anderson et al. (2003); Firth et al. (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Kim & Gu (2005);Traichal et al. (1999) Firm size Sales Volume (SV) Cauvin (1979); Gray & Cannella (1997); Leone et al (2006); Skalpe (2007); Stock Return (SR) Anderson et al. (2003); Andjelkovic et al (2002); Duru & Iyengar (1999); Eichholtz et al (2008); Firth et al. (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Jensen & Murphy (1990); Kato & Kubo (2006); Leone et al (2006); Traichal et al(1999) Return on Asset (ROA) Andjelkovic et al (2002); Duru & Iyengar (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Kato & Kubo (2006);Leone et al (2006); Skalpe (2007); Return on Investment (ROI) GomezMejia et al (1987); Kim & Gu (2005); Return on Sales (ROS) Gray & Cannella, Jr (1997); Return on Equity (ROE) Andjelkovic et al (2002); Firth et al. (1999); GomezMejia et al (1987); Traichal et al(1999) Earnings per Share (EPS) Duru & Iyengar (1999); Eichholtz et al (2008); GomezMejia et al (1987); Perry & Zenner (2001) Firm Profitability Sales Growth (GS) Duru & Iyengar (1999); Firth et al. (1999); GomezMejia et al (1987); Grinstein & Hribar (2004); Kato & Kubo (2006); Kim & Gu (2005) Firm Liquidity Fixed Assets Turnover (FAT) Kim & Gu (2005) Firm Activity Asset Turnover (AT) Gu & Choi (2004); Kim & Gu (2005) Debt ratio (DT) Kim & Gu (2005) Firm Coverage Long Term Debt (LTD) Gu & Choi (2004); Traichal et al(1999) 23 3. Managerial Power Determinants of Executive Compensation Recent academic research has found it difficult to explain the determinants of executive compensation using only firm performance because numerous empirical results have supported that the payforperformance rule does not always work in determining executive compensation (Conyon, 1997; GomezMejia et al., 1987). Thus, researchers have turned their sights to finding other factors that might influence executive compensation, such as demographic characteristics (e.g., age, gender, and education level), compensation structure (e.g., the composition of compensation with stock options and cash compensation), and the variables from the managerial power approach (e.g., stock ownership, board size, compensation committee size) (Coles, McWilliams, & Sen, 2001; Nelson, 2005). Among those attempts to address other determinants of executive compensation, the dominant theoretical foundation is the managerial power approach proposed by Ouch and Maguire in 1975 (GomezMejia & Wiseman, 1997). According to traditional financial theory, especially agency theory, the board of a company can control an agent (executive) with compensation packages. However, the managerial power approach suggests that the executive would not consider shareholder wealth if he or she had the power to influence the board’s decisionmaking; that is, if the executive has enough governance power to affect the board’s decision process in establishing the executive’s compensation contract, the traditional financial view based on the payforperformance rule may not be an appropriate approach to finding the determinants of executive compensation (Bebchuk & Fried, 2003; Core et al., 1999; Grinstein & Hribar, 2004; Hallock, 1998; Yermack, 1995). Thus, the managerial power 24 approach has been combined with the payforperformance rule in recent empirical studies on executive compensation. There are two main components of the managerial power approach: stock ownership structure and board independence. For the stock ownership structure’s impact on executive compensation, the CEO who possesses a higher portion of the company’s outstanding stocks could have more power in the company and be more likely to use corporate resources for his or her own benefit (Khan, Dharwadkar, & Brandes, 2005). Thus, the executive with high stock ownership would extract greater overall levels of compensation (Ozkan, 2007). In other words, the level of executive compensation would be higher when the executive has higher stock ownership (Toyne, Millar, & Dixon, 2000). In addition, higher executive possession of company’s outstanding stocks would influence the composition of board members, because the voting rights to select directors are distributed according to the amount of company stock held. Thus, an executive who owns a great deal of stock may have enough power to affect his or her own compensation level by selecting sympathetic board members (GrabkeRundell & GomezMejia, 2002). Since the board of directors decides the level of executive compensation, the independence of the board has been regarded as one of the key factors in determining executive compensation. However, it is not always easy to keep the board independent of top executives in the company. For example, outside members of the board are less likely to conflict with the CEO when the CEO appoints the outside members. Furthermore, the board of directors tends to follow the opinion of compensation consultants who are hired by the CEO (Core et al., 1999). As a result, executive compensation might not be 25 determined with the company’s best interests in mind (Core et al., 1999). Therefore, the independence of the board should be considered in an executive compensation study. Numerous studies have been conducted to examine the association between executive compensation and selected variables from the managerial power approach. GomezMejia et al. (1987) studied the effect of ownership structure on CEO compensation by classifying sample companies into two categories using the 5 percent ownership convention (referring to whether one individual or organization holds more than five percent of the company’s outstanding stock and may, therefore, be able to affect decisions): managementcontrolled companies and ownercontrolled companies. The firm’s performance and size measures were also included in the study, which found that ownership structure significantly influenced the level of CEO compensation such that CEO in externally controlled firms receive more compensation on the basis of firm performance than do CEOs in internally controlled firms. Thus, executive compensation would be more likely to follow the payforperformance rule in externally controlled firms, and executive stock ownership is a key factor in determining executive compensation. Core et al. (1999) also researched the effect of corporate governance on executive compensation with 495 CEOs in 205 publicly traded U.S. firms. The authors utilized several relevant variables from both financial performance and the managerial power approach to identify determinants of executive compensation. Among the managerial power variables were board size, composition of board membership, whether the CEO served as chairperson of the board, and the CEO’s percentage of stock ownership. The authors found that there is a significantly negative relationship between CEO 26 compensation and board and ownership structure and concluded that CEOs received greater compensation when governance structures were less effective. Likewise, Yermack (1996) found that companies with small boards provided stronger CEO performance incentives from compensation. Other studies that investigated the relationship between CEO compensation and corporate governance from managerial power approach included that of Bebchuk and Fried (2003), which found that the CEO could influence board decisions by controlling the information about the company to board members and controlling the meeting time and agenda. Several studies have examined the relationship between executive compensation and relevant variables from the managerial power approach in different countries, such as China (Firth, Fung, & Rui, 2007), the United Kingdom (Ozkan, 2007), and Israel (Cohen & Lauterbach, 2008). Firth et al. (2007) examined how ownership structure and corporate governance influenced CEOs’ compensation in Chinese companies. They adopted variables based on the theoretical concepts from both the payforperformance and managerial power approaches. Board size, proportion of nonexecutive directors on the board, and a dummy variable (whether the CEO and the chairman of the board are the same person) were used to examine the effect of managerial power on CEO compensation. The study revealed that type of ownership and board size affected CEO compensation, as independent boards or nonexecutive directors of boards were more likely to implement performancerelated pay. Using board size, the composition of nonexecutive directors on board, and CEO stock ownership, Ozkan (2007) investigated how corporate governance influenced CEO compensation in 414 U.K. companies and found that larger board size and a higher 27 proportion of nonexecutive directors on the boards resulted in higher CEO compensation; thus, less corporate governance of executives led to higher executive compensation. Finally, Cohen and Lauterbach (2008) researched CEO compensation, as it related to company ownership, with 124 publicly traded firms in Israel from 1994 to 2001. The authors included independent variables of firm performance, firm size, board composition, demographics (education level and age), and company ownership and found that CEOs in CEOowned companies received significantly higher compensation than did CEOs who did not own part of the company. In addition, the payforperformance sensitivity was lower in CEOowned companies than in nonCEO owned companies, even though the difference was not statistically significant. These results also indicated that the CEOs who owned more company stock received higher compensation than did CEOs who owned less company stock, regardless of firm performance. Thus, executive compensation was related to corporate governance in a variety of different countries. The extant literature demonstrates that the importance of the managerial power approach has increased and that combinations of variables from the payforperformance and the managerial power approaches have come to the fore in executive compensation studies. Thus, the managerial power approach should be considered for the current study in order to derive more concise and meaningful information, so both the payforperformance and the managerial power approaches shall be included in this study. Table 22 summarizes selected variables from the managerial power approach used in prior studies which will form the basis of variables of managerial power determinants in this study. 28 Table 22. The variables of managerial power approach used in previous studies Type Variables Studies Executive shares Cohen & Lauterbach (2008); Core et al. (1999); Coles et al. (2001); GomezMejia et al. (1987); Khan et al. (2005); Ozkan (2007) Ownership Ownership Composition Core et al. (1999); Coles et al. (2001); Firth et al. (2007); Khan et al. (2005); Ozkan (2006); Toyne et al. (2000) Board size (Number of Board Director) Core et al. (1999); Firth et al. (2007); Grinstein & Hribar (2004); Hallock(1997); Ozkan (2007); Yermack (1995) Board Structure Cohen & Lauterbach (2008); Core et al. (1999); Coles et al. (2001); Firth et al. (2007); Grinstein & Hribar (2004); Board Independence Executive as Director (CEO as Board of chair) Conyon (1997); Core et al. (1999); Firth et al. (2007); Grinstein & Hribar (2004); 4. OLS regression and Quantile regression Several types of multiple regression analyses have been utilized in prior empirical studies to examine the relationship between executive compensation and selected variables based on payforperformance, managerial power, and demographic characteristics. These have included multivariate logistic regression (Gray & Cannella, 1997; Jensen & Murphy, 1990; Nelson, 2005), weighted leastsquares (WLS) regression 29 (Gu & Choi, 2004; Kim & Gu, 2005), and ordinary least squares (OLS) regression (Anderson et al., 2004; Core et al., 1999; Dyl, 1988; Firth et al., 2007; Firth et al., 1999; Grinstein & Hribar, 2004; GomezMejia et al., 1987; Hebner & Kato, 1997; Kato & Kubo, 2006; Traichal et al., 1999). Clearly, most researchers have used OLS regression analysis in these efforts. OLS regression achieves the parameter estimates of the model (model fit) and illuminates the relationship between the dependent variable and independent variables through the conditional mean function (Kutner, Nachtsheim, Neter, & Li, 2005; Pedhazur, 1997). The conditional mean function uses the conditional mean response to examine the relationship between the dependent variable and independent variable(s) (Hao & Naiman, 2007). One of the crucial factors in conducting OLS regression is reducing a heteroscedasticity problem by minimizing the sums of squared residuals in the regression equation. However, OLS regression analysis has been criticized for failing to minimize the sums of squared residuals in the regression equation (Koenker, 2005) because it is difficult to follow the equal variance of variables for social phenomena in the real world (Hao & Naiman, 2007). Because of this criticism, Koenker and Basset (1978) developed a new mechanism of regression analysis, called quantile regression analysis, which uses a conditional quantile function instead of a conditional mean. Quantile regression analysis can be used to examine the relationship between the dependent variable and independent variable(s) by estimating each quantile of response variables based on the conditional quantile function (Koenker & Hallock, 2001); thus, it can achieve flexibility by estimating a change in the entire range of the dependent variable through a change in each independent variable (Hao & Naiman, 2007). As a result, quantile regression 30 analysis has gradually emerged as the way to estimate the responses of various levels of a population from each independent variable (Koenker & Machado, 1999). More specifically, the conditional mean function in the OLS regression enables to estimate the coefficient of each independent variable by taking the value of parameters that minimize the sum of squared residuals. In other words, OLS regression minimizes the sum of squared vertical distances between data points (Xi, Yi) and the fitted line 0 1 β ˆ + β ˆ = Y ˆ (Hao & Naiman, 2007; Koenker, 2005; Koenker & Basset, 1978). The model for estimating the coefficient of OLS regression is shown as follows: Min 2 0 1 (Υ  (β + β )) i i i Σ x However, the conditional quantile function enables to estimate the coefficients of independent variables that minimize the sum of absolute residuals. In other words, quantile regression minimizes the sum of absolute vertical distances between observed value to its fitted value. The model for estimating the coefficient of medianregression line is follows: Min Υ  (β + β ) i 0 1 i i Σ x The median regression line should pass through a pair of sample, by half of data should be in the above median regression line, as well as the last half of data should be in the below median regression line (Hao & Naiman, 2007). By modifying above median regression function, the estimation of coefficients for pth quantile regression is derived as shown below: Min  (β + β ) + (1 P) Υ  (β +β ) ( ) 1 ( ) 0 <β +β (p) 1 ( ) 0 β +β Σ ( ) 1 ( ) 0 ( ) 1 ( ) 0 i P P i Yi x i p i Yi x P Y x x i p p i p p Σ ≥ 31 As shown above pth quantile regression model, pth quantile regression enables to estimate the coefficients ( ) 0 β ˆ p and ( ) 1 β ˆ p by using the weighted sum of distances between fitted values from ( ) 1 ( ) 0 β ˆ + β ˆ ˆ = p p i Y and the observed value of Yi, where 0 < P < 1 (Hao & Naiman, 2007; Koenker, 2005; Koenker & Bassett, 1978). In addition, Koenker and Hallock (2001) used one example of a CEO compensation topic to illustrate the need for quantile regression analysis for executive compensation study. They derived 1999 data from the EXECUCOMP database for CEO annual compensation in 1,660 firms and showed that executive compensation tends to increase with firm size. However, the low and high levels of CEO annual compensation were more highly related to firm size than were the middle range of CEO annual compensation, indicating that different levels of CEO compensation were differently related to firm size. The authors insisted that those kinds of results would be more frequent and would create more difficulty in explaining the relationship between executive compensation and covariates with OLS regression analysis when the sample size is larger and more covariates are included in the study. Thus, they suggested that quantile regression analysis would be a more effective method than the OLS method for executive compensation studies. In the current study, quantile regression analysis also enables examination of whether different levels of executive total cash compensation are related differently to each independent variable. More specific and concise results are expected from quantile regression analysis than would be expected from OLS regression analysis. 32 5. Development of Hypotheses Two main hypotheses for this study are proposed for examining the determinants of executive cash compensation in the hospitality industry using OLS regression and quantile regression with selected variables from both the payforperformance and the managerial power approach. The two main hypotheses were tested for three classes of samples: for Ha, all hospitality industry (H1), hotel and casino industry (H2), and restaurant industry (H3); and for Hb, and all hospitality industry (H4), hotel and casino industry (H5), and restaurant industry (H6). Hypotheses A Ha: The selected variables from both the payforperformance rule and the managerial power approach are significantly correlated with executive cash compensation in the hospitality industry. Ha1: The firm’s current ratio (CR) is significantly correlated with executive cash compensation in the hospitality industry. Ha2: The firm’s asset turnover (AT) is significantly correlated with executive cash compensation in the hospitality industry. Ha3: The firm’s debttoasset ratio (DT) is significantly correlated with executive cash compensation in the hospitality industry. Ha4: Firm size (FS) is significantly correlated with executive cash compensation in the hospitality industry. 33 Ha5: The firm’s Earnings per Share (EPS) is significantly correlated with executive cash compensation in the hospitality industry. Ha6: The firm’s sales growth (GS) is significantly correlated with executive cash compensation in the hospitality industry. Ha7: The type of executive (whether the executive is a director or not: PDIR) is significantly correlated with executive cash compensation in the hospitality industry. Ha8: The board size (the number of directors on the board: NDIR) is significantly correlated with executive cash compensation in the hospitality industry. Ha9: The compensation committee size (the number of directors on the compensation committee: NCCMT) is significantly correlated with executive cash compensation in the hospitality industry. Ha10: The number of the executive’s equity shares (Dummy variable: whether the executive has more than 5% of outstanding common stocks of company: SO) is significantly correlated with the executive cash compensation in the hospitality industry. One additional set of hypotheses was proposed for the quantile regression method. Each selected variable from both the payforperformance rule and the managerial power approach was tested by different levels of executive total cash compensation, leading to the following hypothesis: 34 Hypotheses B Hb: The selected variables from both the payforperformance rule and the managerial power approach are differently correlated with different levels of executive cash compensation in the hospitality industry. Hb1: The firm’s current ratio (CR) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb2: The firm’s asset turnover (AT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb3: The firm’s debttoasset ratio (DT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb4: Firm size (FS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb5: The firm’s Earnings per Share (EPS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb6: The firm’s sales (GS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb7: The type of executive (whether the executive is a director: PDIR) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb8: Board size (the number of directors on the board: NDIR) is differently correlated with different levels of executive cash compensation in the hospitality industry. 35 Hb9: Compensation committee size (the number of directors on the compensation committee: NCCMT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb10: The number of the executive’s equity shares (Dummy variable: whether the executive has more than 5% of outstanding company common stocks: SO) is differently correlated with different levels of executive cash compensation in the hospitality industry. 36 CHAPTER III METHODOLOGY 1. Data Collection and Sampling Procedures The main objective of this study is to examine which elements from the two approaches, financial performance and managerial power, are linked to executive cash compensation in the hospitality industry. Sample data were gathered from secondary databases, Standard & Poor’s COMPUSTAT database and proxy statements (DEF 14A) from SEC filings. The data collection procedure for this study was divided into two main processes: gathering firms’ financial data from the COMPUSTAT database to calculate financial measurements and collecting executive compensation data and data related to the managerial power approach from the proxy statements from SEC filings. If a company’s data was not available for one of following procedures, the observation was eliminated from the sample. The sample companies were limited to the companies that were on the list of COMPUSTAT database. Among the several subsidiaries of the hospitality industry were three major sectors: hotels, casinos, and restaurants. 37 1. Financial data for the sample companies were retrieved for each of the three main sectors of the hospitality industry from the COMPUSTAT database using Standard Industrial Classification (SIC) codes. A total of 150 hospitality company samples were collected. 1) Hotel Industry The initial sample consisted of all hotel companies (SIC code 7011). After excluding companies that did not have financial data for either 2005 or 2006, 15 hotel companies remained in the sample. 2) Casino Industry The initial sample consisted of all casino companies (SIC code 7990). After excluding companies that did not have financial data for either 2005 or 2006, 49 casino companies remained in the sample. 3) Restaurant Industry The initial sample consisted of all restaurant companies (SIC code 5812). After excluding companies that did not have financial data for either 2005 or 2006, 86 restaurant companies remained in the sample. 2. The 150 hospitality companies in the sample were matched to the SEC filing list to find executive compensation data and data related to the managerial power approach. After the matching process, 83 hospitality companies remained in the sample. 1) Hotel Industry Seven hotel companies were eliminated from the sample either because they were not listed in the SEC filings or because they didn’t have proxy 38 statements (DEF 14A) for 2005 and 2006. After excluding the 7 companies, 8 hotel companies remained in the sample. 2) Casino Industry Twentyeight casino companies were eliminated from the sample either because they were not listed in the SEC filing lists or because they didn’t have proxy statements (DEF 14A) for 2005 and 2006. After excluding the 28 companies, 21 casino companies remained in the sample 3) Restaurant Industry Thirtytwo restaurant companies were eliminated from the sample either because they were not listed in the SEC filing lists or because they didn’t have proxy statements (DEF 14A) for 2005 and 2006. After excluding the 32 companies, 54 casino companies remained in the sample. 3. Executive compensation and data related to the managerial power approach were retained from the proxy statements of the 83 hospitality companies for 2005 and 2006. Initially, data for 388 executives were gathered; after filtering, 331 executives’ data remained. 1) Filtering Executives’ data from the Hotel Industry The initial executive sample included 44 executives in 8 hotel companies. Nine executives were missing compensation data for either 2005 or 2006 and were eliminated, leaving 35 executives in the sample. 2) Filtering Executives’ data from the Casino Industry 39 The initial executive sample included 104 executives in 21 hotel companies. Seventeen executives were missing compensation data for either 2005 or 2006 and were eliminated, leaving 87 executives in the sample. 3) Filtering Executives’ data from the Restaurant Industry The initial executive sample included 240 executives in 54 hotel companies. Thirtyone executives were missing compensation data from either 2005 or 2006, leaving 209 executives in the sample. 4. With a total of 331 executives’ data remaining, the data was filtered again to remove executives who had a greater than 100% change in total cash compensation from 2005 to 2006 because such an unusual change in executive total cash compensation could skew results. 1) In the Hotel Industry One executive was removed because the executive had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 34 executives in the sample. 2) In the Casino Industry One executive was removed because the executive had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 86 executives in the sample. 3) In the Restaurant Industry Fourteen executives were removed because they had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 195 executives in the sample. 40 5. After transforming the actual cash compensation of executives to their natural logarithms, two outliers were detected in the restaurant sample as having a very low log value. The real dollar amounts of two executives’ annual total cash compensations (outliers) were same as $25,000. The $25,000 for each executive’s total cash compensation was too small, compared with other executives in the sample. Thus, the two outliers were deleted from the restaurant sample to achieve more efficient results. As a result of removing the two outliers from the restaurant sample, the number of executives in the restaurant sample decreased from 195 to 193, and the number of executives in the full sample decreased from 315 to 313. 6. Finally, the sample was divided into two subsamples: the hotel and casino industry made up one subsample and the restaurant industry made up the other. The hotel and casino companies were combined as one subsample because the number of executives in those industries was too small to conduct statistical analysis, especially regression analysis. Combining them made sense since hotel and casino companies are not always easily distinguished because some hotel companies also have casino facilities, and vice versa. As shown in Table 31, these procedures led to a total of 313 executives from 83 hospitality companies: 120 executives from the 29 hotel and casino companies and 193 executives from the 50 restaurant companies. 41 Table 31. Classification of study sample Category Type of Industry Number of companies Number of executives SubSample Hotel & Casino Industry 29 120 SubSample Restaurant Industry 50 193 Full Sample Hospitality Industry 79 313 2. Variable selection Based on the extensive literature review, eleven variables were selected for the study—one dependent variable, six variables from the payforperformance rule, and four variables from the managerial power approach. The dependent variable (executive total cash compensation) and the six financial variables were transformed by natural logarithm or calculated by formula to conduct the multiple regression analyses for this study. This section explains why the dependent and independent variables were selected for the purposes of this study, how the dependent variable and one independent variable (firm size) were transformed, and how the other financial measures to be utilized for this study were calculated. Selection of dependent variable Executive compensation consists of three main types of executive compensation: cashbased compensation (e.g., salary and bonus), deferred compensation (e.g., stock options), and benefitbased compensation (e.g. insurance and pensions) (Brigham & Houston, 2001). As has been the case with many prior studies, the current study used only the cashbased compensation (in this case, salary and bonus for 2006) as the 42 dependent variable (Gray & Cannelaa, Jr., 1997; Gu & Choi, 2004; Jensen & Murphy, 1990; Kim & Gu, 2005; Lippert & Porter, 1997). Other types of compensation were not included because they are difficult to translate into comparable (cash) amounts. In addition, total cash compensation was transformed by natural logarithm in order to avoid the statistical problem of heteroscedasticity that can result from the not equal variances of variables from raw data when conducting the regression analyses (Dyl, 1988; Ott & Longnecker, 2001). By adopting the base of natural logarithms for each executive’s total cash compensation, the dependent variable was transformed from the original values of executive total cash compensation to the log of executive total cash compensation. Selection of independent variables Firm performance variables Several types of financial measures for firm performance have been utilized in prior empirical executive compensation studies, primarily marketbased performance measures, accountingbased performance measures, and growthbased performance measures. The current study adopted both accountingbased and growthbased performance measures. Marketbased performance measures (e.g., stock returns) were not chosen for this study because they can be easily biased by “noise” that is not controlled by management (GomezMejia & Wiseman, 1997). Firm size was also utilized as a financial variable. The financial measures for firm performance were adopted and modified from Gu and Choi’s study (2004) and Kim and Gu’s study (2005) in the hospitality field. 43 As was mentioned in the literature review, accountingbased financial performance measures are generally divided into four categories: liquidity, activity, profitability, and coverage. Current ratio and quick ratio are common examples of liquidity ratios, which estimate a firm’s ability to pay back its shortterm debts. Current ratio (CR) was selected for this study, rather than quick ratio, because CR is most commonly used as a basic ratio for liquidity and because quick ratio excludes more liquid assets, like inventory, even though inventory is one of the most important assets in the hospitality industry (Chatfield & Dalbor, 2005). Activity ratios measure management’s effectiveness in employing its resources to the firm’s business and include mainly receivable turnover, inventory turnover, and asset turnover (Kieso et al., 2001). As in Gu and Choi (2004) and Kim and Gu (2005), asset turnover was selected for this study. Profitability ratios include return on assets, profit margin on sales, and earnings per share (EPS). Return on assets and EPS have often been used in executive compensation studies as an estimator of firm’s profitability (Duru & Iyengar, 1999; Eichholtz et al., 2008; GomezMejia et al., 1987; Perry & Zenner, 2001). EPS was selected as an estimator of firm’s profitability ratios for the current study, rather than return on assets, because EPS facilitates checking the firm’s profitability based on the amount of outstanding common stock, so EPS is an indicator of shareholder profits from the firm’s business activities in the fiscal year (Gallagher & Andrew, 1997). Finally, as has been the case in prior studies in the field, debt ratio (DT) (Kim & Gu, 2005; OrtizMolina, 2007) was selected as an estimator of the firm’s coverage ratios, which measure the firm’s ability to protect itself from its total debt (Brigham & Houston, 2001). 44 Most previous executive compensation studies have also added firm size and sales growth as financial determinants of executive compensation. Firm size has been used as a control variable in prior executive compensation studies because it is highly correlated with the level of executive compensation (GomezMejia et al., 1987). Thus, as in other studies, total assets (TA) was selected to estimate firm size for this study (Gu & Choi, 2004; Kim & Gu, 2005). Sales growth has also been viewed as an indicator of growthbased performance and was adopted from Kim and Gu’s study as a estimator of growthbased performance for the current study. Thus, a total of six financial variables from the payforperformance rule were utilized for this study. Accountingbased financial ratios were used to transform and calculate six financial variables from the payforperformance rule into independent variables. Firm size and sales growth rate were also calculated using formulas; firm size transformed the dollar amount of the firm’s total assets by natural logarithms to avoid the bias of heteroscedasticity. The following formulas were used to calculate the financial variables for this study: 1) Current Liabilities Current Assets Current Ratio (CR) = 2) Total Assets Net Sales Asset Turnover (AT) = 3) Common stock outstanding Net Income  Preferred Stock Dividends Paid Earinings per Share (EPS) = 4) Total Assets Total Debt Debt Ratio (DT) = 45 5) 2005 2006 2005 Sales Sales  Sales Sales Growth (SG) = 6) Firm size (FS) = Log(TotalAsset) Managerial power variables Several types of variables have been utilized in prior studies to investigate the effects of the stock ownership structure and board independence on executive compensation. For this study, four variables from the managerial power approach were selected: Number of board directors (NDIR), Number of compensation committee members (NCCMT), Executive as current director (PDIR), and the executive’s stock ownership (SO). Several studies have adopted NDIR and NCCMT to estimate the board’s independence (Core et al., 1999; Firth et al., 2007; Grinstein & Hribar, 2004; Hallock, 1997; Ozkan, 2007; Yermack, 1995). Real numbers for both variables were collected from the companies’ proxy statements (DEF 14A) in the SEC filing lists and recorded in the dataset. PDIR represents whether the executive is a current member of the board of directors and was a dummy variable, coded 0 if the executive was not a current board director or 1 otherwise. Finally, the executive’s stock ownership was included to examine the effect of ownership structure on executive compensation and was also a dummy variable, coded 0 if the executive has less than 5% of company’s common stocks or 1 otherwise. The classification rule for this variable was based on whether the executive held more than 5% of the company’s outstanding common stocks. Since 1960s, numerous researchers have used the cutoff point of 5% stock ownership convention in many empirical research, because 5% of stock ownership for publicly traded company has been considered as enough amounts of stocks to influence on the firm’s decision making 46 (GrabkeRundell & GomezMejia, 2002; GomezMejia et al., 1987). Thus, four variables from the managerial power approach were adopted and modified to examine the effects of corporate governance and stock ownership on executive compensation in the hospitality industry. 3. Data Analysis and Model This research is designed as a crosssectional data analysis to examine how each financial performance and managerial power variable is linked to executive cash compensation in the hospitality industry. The data analysis of this study consisted of a descriptive analysis, a reliability test, an Ordinary Least Squares (OLS) regression analysis, and a quantile regression analysis. A descriptive analysis summarized the sample’s financial characteristics (e.g., firm size, EPS, and sales) and corporate governance characteristics (e.g., number of board members, executive’s stock ownership, board characteristics). Several types of reliability tests were conducted to check the data before doing the OLS regression and quantile regression analyses. Scatter plots allowed outliers to be removed from the sample, and a histogram and normal probability plot tested the normality and linearity in order to check the assumptions of the multiple regression analysis. The OLS regression analysis and the quantile regression analysis were used to investigate the determinants of executive cash compensation, with total cash compensation as the dependent variable (Y) and all variables from both the payfor 47 performance rule and the managerial power approach as the independent variables (X). Quantile regression analysis allowed examination of whether different levels of total cash compensation are related differently to each independent variable from the payforperformance rule and the managerial power approach. To test the hypotheses proposed in literature review chapter, the multiple regression models for each industry were proposed as follows: Predicted Executive total cash compensation = β0+ β1 Current ratio(CR) it + β2 Asset turnover(AT) it + β3 Debt ratio(DT) it + β4 Firm size(FS) it + β5 Earnings per Share (EPS) it + β6 Sales growth(SG) it + β7 Executive as board directors(PDIR) it + β8 Number of board directors(NDIR) it + β9 Number of compensation committee members(NCCMT) it + β10 Executive’s stock shares (SO) it + ε it , Where, β0 = the intercept; β1,2…,10 = the beta coefficient or slope; and εit = the random error term or the residual portion; Total cash compensation it = the sum of executive’s annual cash salary and cash bonus for firm i in year t; Current ratio it = Current asset/Current liabilities for firm i in year t; Asset turnover it = Total sale (revenue)/ Average of asset for firm i in year t; Debt ratio it =Total liabilities / Total assets for firm i in year t; Firm size it = Log of the book value of total assets of firm i in year t ; Earnings per Share it = (Net income – preferred common stock dividend paid) / common stock outstanding for firm i in year t; Sales growth it = the percentage growth in sales for firm i from year t1 to year t; Executive as board director it = Whether the executive is also a member of the board for 48 firm i in year t (not current member of board = 0, current member of board = 1); Number of board directors it = Total number of board directors for firm i in year t; Number of compensation committee members it = Number of compensation committee members for firm i in year t; and Executive’s stock shares it = whether the portion of executive’s equity shares for firm i in year t is more than 5% of the firm’s outstanding common stock (less than 5% = 0, more than 5% = 1). 49 CHAPTER IV FINDINGS 1. Description of Sample Table 41 shows a frequency analysis for the characteristics of this study’s sample. Executive total cash compensation in the hospitality industry averages $559,484, range from $109,490 to $3,035,000. The average executive total cash compensation in hotel and casino companies is higher than that in restaurant companies, at $711,395 and $465,031, respectively; the median in hotel and casino companies is also larger than the median in restaurant companies. Furthermore, the mean of the percent change of executive compensation from 2005 to 2006 was negative at 8.95%, but the average percent change and the median percent change of executive total cash compensation in hotel and casino companies was more negative than was that for restaurants (14.98% and 5.20% average change, respectively; and 12.44% and 2.6% median change, respectively). 50 Table 41. Descriptive Statistics of Executive Total Cash Compensation (N=313) Sample Category Mean Std. Dev. Median Minimum Maximum Total Cash Compensation (2006) $559,484 $483,339 $388,600 $109,490 $3,035,000 All Hospitality Companies (N=313) % Change of Total Cash Compensation (2005  2006) 8.95 32.08 5.74 85.08 96.04 Total Cash Compensation (2006) $711,395 $587,548 $564,879 $109,490 $2,825,000 Hotel & Casino Companies (N=120) % Change of Total Cash Compensation (2005  2006) 14.98 35.09 12.44 85.08 50.95 Total Cash Compensation (2006) $465,031 $398,081 $339,984 $112,452 $3,035,000 Restaurant Companies (N=193) % Change of Total Cash Compensation (2005  2006) 5.20 29.53 2.60 71.27 96.04 In terms of corporate governance characteristics, the average number of board members is 8 for both the full sample (all hospitality industry) and the subsamples (Hotel & Casino industry and Restaurant industry), and the number of board members 51 ranges from 3 to 14. More than 30% of executives in the samples are board members and more than 30% of the executives in the samples also hold more than 5% of outstanding stock, which indicates that many have enough power to influence board decisions. Prior to performing OLS regression and quantile regression analysis, several tests for outliers, normality, and linearity were performed to check assumptions of the multiple regression method. The outliers were checked by developing scatter plots of samples; there were no outliers among dependent variables (Log TCC) in the full sample or the subsamples (Figures 41, 42 and 43). Figure 41. Scatter plot for full sample (All hospitality industry) 52 Figure 42. Scatter plot for subsample (Hotel & Casino industry) Figure 43. Scatter plot for subsample (Restaurant industry) 53 The normality and linearity of samples were also tested using histograms and normal probability plots of standardized residuals for dependent variables (Log of total cash compensation). As shown in Figure 44, standardized residuals of dependent variables in the full sample were normally distributed and had linearity. Figures 45 and 46 also show that nonnormality and nonlinearity were not detected in the subsamples of either the Hotel & Casino subsample or the Restaurant subsample. Thus, it was confirmed that data sets of both the full sample and the subsamples were appropriate to conduct multiple regression methods to examine the relationship between executive total cash compensation and independent variables selected for this study. 54 Figure 44. Histogram and Normal PP Plot for full sample (All hospitality industry) 55 Figure 45. Histogram and Normal PP Plot for subsample (Hotel & Casino industry) 56 Figure 46. Histogram and Normal PP Plot for subsample (Restaurant industry) 57 2. Findings of Ordinary Least Square (OLS) Regression Tables 42, 43, and 44 report the results of the OLS regression with the full sample and the two subsamples. Multicollinearity for the three multiple regression models had to be checked since high correlations among the variables would cause deviation or and/or misleading results in the multiple regression statistics by changing input variable in the regression model as variables were added in or deleted from the model (Pedhazur, 1997). The variance inflation factor (VIF) was used to check the impact of multicollinearity between each independent variable in the regression models. The higher the VIF number, the greater the impact of collinearity on the accuracy of the model (Ott & Longneker, 2001). The VIF values shown in Table 42, for the full sample, lie in the range between 1.091 and 3.361. This does not indicate a serious multicollinearity problem because the VIF is well below the problematic level of 10 (Ott & Longneker, 2001). The range of VIF values for the Hotel & Casino subsample are between 1.088 and 5.141 (Table 43), and Table 44 shows that the VIF values of the Restaurant subsample are between 1.275 and 2.906. Thus, there are no serious multicollinearity problems for the subsamples either. After testing multicollinearity using VIF values, OLS regression analyses were conducted to investigate the relationship between the dependent variable and 10 independent variables to examine the three main hypotheses (H1, H2, and H3). The dependent variable for the OLS regression models is the log of executive total cash compensation and the ten independent variables consisted of six financial variables and 58 four managerial power variables. The results of the OLS regression analyses are presented in Tables 42, 43, and 44. Results of OLS regression method for the Full sample Table 42 summarizes the results of the OLS regression for the full sample with six financial variables and four managerial power variables. Both the Rsquare (=0.646) and the adjusted Rsquare (=0.634) for this model were the appropriate level of goodness of fit for empirical study in social science fields. The pvalues of three of the financial variables (DT, FS, EPS) were less than 0.01 with positive coefficients, and the pvalues of the other three financial variables (CR, AT, GS) were larger than 0.05, so only DT, FS, and EPS were positively related to the dependent variable at a statistically significant level of 0.01. The pvalues of both PDIR and SO were less than 0.01, and PDIR and SO were positively associated with executive total cash compensation at a pvalue of 0.01. 59 Table 42. OLS regression summary for the Full sample (all hospitality companies) Variable T Value Significance Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.672 56.480 0.000 CR 0.021 1.239 0.216 0.713 1.403 AT 0.022 1.125 0.262 0.505 1.982 DT 0.115 3.419 0.001*** 0.788 1.269 FS 0.245 10.186 0.000*** 0.298 3.361 EPS 0.038 4.170 0.000*** 0.615 1.626 GS 0.079 1.739 0.083* 0.917 1.091 PDIR 0.182 6.749 0.000*** 0.658 1.520 NDIR 0.010 1.583 0.114 0.514 1.946 NCCMT 0.004 0.494 0.622 0.844 1.185 SO 0.092 3.196 0.002*** 0.578 1.730 N RSquare Adjusted R 313 0.646 0.634 Note: * P< 0.10, ** P< 0.05, ***P<0.01 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.672 + 0.115 Debt to asset ratio(DT) + 0.245 Firm size(FS) + 0.038 Earnings per share(EPS) + 0.182 Type of board directors(PDIR) + 0.092 Executive’s stock shares(SO). Thus, hypotheses H13, H14, H15, H17, and H110 were accepted at the 0.01 level, but hypotheses H11, H12, H16, H18, and H19 were not. 60 Results of OLS regression method for the Hotel & Casino subsample Table 43 summarizes the results of the OLS regression method for the Hotel & Casino subsample with six financial variables and four managerial power variables. Both the Rsquare (=0.708) and the adjusted Rsquare (=0.682) for this model were the appropriate level of goodness of fit. Like the OLS regression for the full sample, six financial variables were used for the OLS regression for this subsample with the result that the pvalues for four variables (DT, FS, EPS, and GS) were less than 0.01, and the pvalues of CR and AT were larger than 0.05. Thus, DT, FS, EPS, and GS were positively associated with the dependent variable with statistical significance at the 0.01 level. The pvalues of only two managerial power variables, PDIR and SO, were less than 0.05, so PDIR and SO were positively related with the executive total cash compensation at a pvalue of 0.05. Table 43. OLS regression summary for the Hotel & Casino subsample Variable T Value Significance Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.507 28.246 0.000 CR 0.051 1.373 0.173 0.532 1.881 AT 0.034 0.954 0.342 0.819 1.222 DT 0.176 2.797 0.006*** 0.646 1.547 FS 0.307 5.780 0.000*** 0.195 5.141 EPS 0.041 3.250 0.002*** 0.479 2.088 GS 0.141 2.662 0.009*** 0.919 1.088 PDIR 0.207 4.813 0.000*** 0.731 1.369 NDIR 0.002 0.153 0.878 0.388 2.576 NCCMT 0.008 0.413 0.680 0.630 1.587 SO 0.106 2.264 0.026** 0.638 1.568 N RSquare Adjusted R 120 0.708 0.682 Note: * P< 0.10, ** P< 0.05, *** P<0.01 61 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.507 + 0.176 Debt to asset ratio(DT) + 0.307 Firm size(FS) + 0.041 Earnings per share(EPS) + 0.141 Sales growth(GS) + 0.207 Type of board directors(PDIR) + 0.106 Executive’s stock shares(SO). Thus, hypotheses H23, H24, H25, H26, H27, and H210 were accepted at 0.05 level, while hypotheses H21, H22, H28, and H29 were not. Results of OLS regression for the Restaurant subsample Table 44 summarizes the results of the OLS regression for the Restaurant subsample. Both the Rsquare (=0.591) and the adjusted Rsquare (=0.598) for this model had the appropriate level of goodness of fit, even though both were less than those for the full sample or the other subsample. Contrary to the results of the OLS regression analyses for the full sample and the Hotel & Casino subsample, the results of the OLS regression for the Restaurant subsample had only two variables (DT and FS) in the financial variables with pvalues less than 0.05 and positive coefficients, indicating that DT and FS were positively related to the dependent variable at a statistically significant level of 0.05 and 0.01, respectively. The results of the Restaurant subsample were similar to those of the full sample and the Hotel & Casino subsample in terms of the managerial power variables, as the pvalues of 62 both PDIR and SO were less than 0.05. Thus, PDIR and SO were positively associated with executive total cash compensation at a pvalue at the 0.05 level. Table 44. OLS regression summary for the Restaurant subsample Variable T Value Significance. Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.715 42.041 0.000 CR 0.028 1.433 0.154 0.715 1.398 AT 0.023 0.850 0.396 0.649 1.542 DT 0.093 2.058 0.041** 0.784 1.275 FS 0.255 8.279 0.000*** 0.344 2.906 EPS 0.011 0.594 0.553 0.536 1.865 GS 0.127 1.183 0.238 0.781 1.280 PDIR 0.161 4.676 0.000*** 0.594 1.683 NDIR 0.012 1.421 0.157 0.471 2.124 NCCMT 0.010 0.927 0.355 0.764 1.310 SO 0.090 2.490 0.014** 0.526 1.900 N RSquare Adjusted R 193 0.591 0.568 Note: * P< .10, ** P< .05, P<0.01 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.715 + 0.093 Debt ratio(DT) + 0.255 Firm size (FS) + 0.161 Type of board directors(PDIR) + 0.090 Executive’s stock shares(SO). Thus, only hypotheses H33, H34, H37, and H310 were accepted at 0.05 level, but hypotheses H31, H32, H35, H36, H38, and H39 were not accepted. 63 3. Findings of the Quantile Regression After conducting the OLS regression analyses, the quantile regression analyses were conducted to test the three proposed hypotheses under the second main hypotheses (H3, H4, and H5) to determine whether the selected independent variables are differently related to different levels of executive compensation. The variables for the quantile regression were the same as the variables in the OLS regression analyses. There are two usual ways of interpreting the results of quantile regression: checking the statistical significance of the coefficients of each independent variable toward dependent variable, and checking the pattern of coefficients of each independent variable toward each quantile of the dependent variables. Tables 45, 46, and 47 show the results of the quantile regression analysis for the coefficient estimates of the model for the full sample and the two subsamples, and Figures 47, 48, and 49 show the pattern of the coefficients of each independent variable from the payforperformance rule and the managerial power approaches toward each quantile of dependent variable (the level of executive total cash compensation) for the three samples. The Xaxis for each graph shows the quantile of executive total cash compensation, and the Yaxis shows the coefficients of the independent variables from both the payforperformance rule and the managerial power approach. Red lines show the coefficients for the independent variables from the OLS regression analysis, and the black line represents the coefficients of the independent variable from the quantile regression analysis. The black shadow areas show the 95% confidence interval of coefficients for the independent variables from the results of the quantile regression analysis. 64 Results of the quantile regression method for the Full sample Table 45 shows the results of the quantile regression analysis for the coefficient estimates of the model with the full sample of the hospitality industry. Generally speaking, it looks similar to the OLS regression results for the full sample, even though the quantile regression provides more specific results than the OLS regression does. For example, for the financial variables, neither CR nor AT were correlated with executive total cash compensation at an alpha level of 0.05 in the OLS regression, but the quantile regression showed that both CR and AT were significantly related to executive total cash compensation at the 0.05 level for the low quantiles of compensation, the 0.1 0.2 and the 0.10.3 quantiles, respectively. Thus, executives who received lower cash compensation were influenced by CR and AT, while others were not. In addition, the coefficient graphs for both CR and AT (Figure 47) show that the coefficient values for CR and AT decreased as the level of executive compensation increased, indicating that executives at a lower level of compensation were more sensitive to both CR and AT than were the executives in the middle and upper level of compensation. For the DT and FS variables, the quantile regression analysis provided results similar to those of the OLS regression analysis (i.e., both DT and FS were significantly related to executive compensation in the full sample with statistical significance at an alpha level of 0.05). However, the coefficient graphs for DT show moderate volatility of coefficients from the lower quantile to the upper quantile of executive compensation. This indicates that DT was not differently related to the level of executive compensation with statistical significance. In contrast to the DT graph, the pattern of coefficients of the FS variable decreased from the lower quantile of executive compensation to the middle quantile, then 65 increased as it approached the upper quantile. Thus, the low and high levels of executive compensation were more related to firm size than was the middle range of executive compensation. The results of the quantile regression also show that EPS was not significantly correlated with executive total cash compensation for executives in the lower level of compensation, the 0.10.3 quantile, at an alpha level of 0.05, even though EPS was significantly correlated with total cash compensation in the OLS regression results. This suggests that EPS affects only the executives in the mid and upper levels of total cash compensation. The coefficient graph for the EPS variable in Figures 47 shows an increasing pattern for the coefficient value of EPS from the lower to the upper quantiles of executive compensation, so executives with lower compensation were less sensitive to EPS than were executives in the middle and upper levels of compensation. The result from the quantile regression also shows that GS was significantly correlated with executive total cash compensation for executives in the mid and upper levels of total cash compensation (0.50.8 quantile) at an alpha level of 0.05, even though GS was not significantly related with executive’s total cash compensation in the OLS regression results. The coefficient graph for the GS variable (Figures 47) shows the coefficient value for GS increasing as the level of executive compensation increases, so executives with lower compensation were less sensitive to GS than were executives in the middle and upper levels of compensation. The result of the quantile regression analysis of the four managerial power variables was not much different from that of the OLS regression. Both PDIR and SO were significantly correlated with executive total cash compensation at the 0.05 level, and 66 NDIR and NCCMT were not. Figure 47shows that the only pattern of the SO coefficient of SO was an increasing pattern from the low quantile to the high quantile of executive total cash compensation. By contrast, the pattern of PDIR coefficient had moderate variation. Thus, the effect of SO on executive compensation increased as executive compensation increased. From the results of the quantile regression analysis, hypotheses: H41, H42, H4 4, H45, H46 and H410 were accepted at 0.05 level, while hypotheses: H43, H47, H4 8, and H49 were not accepted. 67 Table 45. Quantile regression summary for the Full sample (all hospitality industry) Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 4.328 4.481 4.580 4.687 4.711 4.858 4.789 (Intercept) 4.787 4.813 Coefficient T value 40.872*** 44.184*** 42.525*** 40.364*** 38.154*** 38.339*** 33.963*** 28.451*** 25.382*** 0.066 0.044 0.026 0.015 0.012 0.009 0.004 0.002 0.016 CR Coefficient T value 3.316*** 2.317** 1.333 0.686 0.529 0.396 0.169 0.071 0.417 0.063 0.053 0.050 0.033 0.040 0.000 0.001 0.017 0.006 AT Coefficient T value 2.495** 2.121** 1.964** 1.248 1.449 0.014 0.020 0.520 0.159 0.159 0.099 0.094 0.092 0.085 0.117 0.100 0.080 0.070 DT Coefficient T value 4.039*** 2.776*** 2.607*** 2.426** 2.249** 3.238*** 2.583*** 2.085** 1.538 0.260 0.276 0.251 0.232 0.231 0.192 0.233 0.274 0.277 FS Coefficient T value 7.527*** 8.569*** 7.489*** 6.759*** 6.518*** 5.696*** 6.164*** 6.197*** 5.491*** 0.021 0.017 0.018 0.036 0.038 0.065 0.053 0.054 0.047 EPS Coefficient T value 1.409 1.297 1.311 2.324** 2.349** 4.268*** 3.571*** 3.667*** 2.937*** 0.010 0.012 0.065 0.137 0.168 0.183 0.170 0.140 0.071 GS Coefficient T value 0.137 0.169 0.876 1.755* 2.119** 2.561** 2.356** 1.943** 1.059 0.210 0.175 0.192 0.165 0.180 0.188 0.166 0.177 0.213 PDIR Coefficient T value 5.090*** 4.901*** 5.139*** 4.207*** 4.497*** 4.922*** 4.298*** 4.348*** 4.490*** 0.002 0.001 0.005 0.006 0.006 0.016 0.019 0.002 0.014 NDIR Coefficient T value 0.172 0.163 0.615 0.659 0.635 1.758* 2.111** 0.246 1.185 0.019 0.006 0.001 0.002 0.001 0.012 0.018 0.001 0.011 NCCMT Coefficient T value 1.770* 0.522 0.082 0.195 0.075 1.038 1.494 0.120 0.864 0.017 0.105 0.083 0.090 0.066 0.078 0.106 0.140 0.142 SO Coefficient T value 0.351 2.870*** 2.182** 2.276** 1.631 1.981** 2.610*** 3.309*** 3.085*** Sample Size N=313 N=313 N=313 N=313 N=313 N=313 N=313 N=313 N=313 Note: * P< .10, ** P< .05, P<0.01 68 Figure 47. The coefficient graphs of all hospitality industry by Quantile regression 69 Results of quantile regression method for Hotel & Casino subsample Table 46 shows the result of the quantile regression analysis for the coefficient estimates of the model with the Hotel & Casino subsample. Generally speaking, the results of the quantile regression analysis were similar to those of the OLS regression results for the subsample, but the quantile regression provides more specific results than the OLS regression. For example, neither CR nor AT in the quantile regression results were correlated with executive total cash compensation in the hotel and restaurant industry at an alpha level of 0.05, which is the same as the results from the OLS regression. However, the patterns of the coefficients of both the CR and AT variables (Figure 48) provided meaningful results, even though the CR and AT were not significantly related with the level of executive compensation. The patterns of the coefficient value for both CR and AT decreased as the level of executive compensation increased, indicating that the executives at lower levels of compensation were more sensitive toward both CR and AT than were the executives at middle and upper levels of compensation. In addition, DT was not significantly related to all quantile of executive compensation in the result from the quantile regression, while the OLS regression showed that DT is significantly related to executive compensation. The quantile regression provided, however, that DT was significantly related to compensation for the low quantile, 0.10.2 quantile, so only those executives at the low level of total cash compensation were influenced by DT. In addition, the coefficient graphs for the DT variable (Figure 48) show that the pattern of the coefficient value for DT decreased slightly as the level of executive compensation increased, suggesting that executives at 70 lower levels of compensation were slightly more sensitive to DT than were the executives in the middle and upper levels of compensation. The quantile regression analysis provided similar results for the FS variable as the OLS regression results for this subsample that FS was statistically significantly related to executive compensation at an alpha level of 0.05. However, the pattern of coefficients of the FS variable decreased from the lower quantile of executive compensation to the middle quantile, then increased to the upper quantile. This suggests that the low and high levels of executive compensation were more related to firm size than was the middle range of executive compensation. The results of the quantile regression show that EPS was not significantly correlated with executive total cash compensation for the low quantile (0.1 – 03 quantile) of executive compensation, even though EPS was significantly correlated with executive total cash compensation in the OLS regression results. It implies that EPS significantly affects only the executive in mid and upper level of total cash compensation. Furthermore, the coefficient graphs for the EPS variable (Figure 48) show that the coefficient value for EPS increased from the lower quantile of executive compensation to the upper quantile, so executives at the lower level of compensation were less sensitive to EPS than were executives in middle and upper levels of compensation. The quantile regression also shows that GS was significantly correlated with executive total cash compensation, but only for the upper level of compensation (0.8 0.9 quantile), even though GS was significantly related to compensation in the OLS regression results. Thus, GS affected only the executives at the upper level of total cash compensation. In addition, the coefficient graphs for the GS variable (Figure 48) show that the coefficient value for 71 GS increased as the level of executive compensation increased, which also supports the conclusion that those at the lower level of compensation were less sensitive toward GS than were those at the middle and upper levels. The result of the quantile regression analysis was not much different for the four managerial power variables than the results of the OLS regression. PDIR was significantly correlated with executive total cash compensation at the 0.05 level, whereas NDIR and NCCMT were not. Most notable were the results from the quantile regression for the SO variable, which showed that SO was significantly correlated with executive total cash compensation for executives only at the upper level of total cash compensation (0.70.8 quantile) at an alpha level of 0.05, even though SO was significantly related to executive total cash compensation in the OLS regression results. Of the two statistically significant variables (PDIR and SO) shown in Figure 48, only SO had a pattern of coefficients that increased from the low quantile to the high quantile of executive total cash compensation; the pattern of coefficients for PDIR had moderate volatility. It indicates that executives at the lower level of compensation were less sensitive to SO than were executives in middle and upper levels of compensation. After the quantile regression analysis, only hypotheses H53, H54, H55, H56, and H510 were accepted at 0.05 level, while hypotheses H51, H52, H57, H58, and H59 were not. 72 Table 46. Quantile regression summary for the Hotel & Casino subsample Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 Coefficient 4.060 4.100 4.296 4.616 4.776 4.794 4.803 4.772 4.957 (Intercept) T value 12.676*** 12.883*** 15.616*** 18.646*** 17.778*** 19.584*** 20.668*** 18.487*** 17.351*** Coefficient 0.117 0.125 0.070 0.013 0.012 0.014 0.029 0.031 0.024 CR T value 1.651 1.692* 1.077 0.208 0.188 0.254 0.571 0.588 0.417 Coefficient 0.043 0.053 0.014 0.055 0.055 0.067 0.051 0.025 0.059 AT T value 0.560 0.924 0.258 0.988 0.985 1.382 1.069 0.488 1.137 Coefficient 0.211 0.210 0.140 0.115 0.101 0.105 0.048 0.061 0.315 DT T value 2.264** 2.119** 1.538 1.219 1.030 1.236 0.618 0.726 1.351 Coefficient 0.384 0.326 0.294 0.242 0.259 0.280 0.306 0.335 0.288 FS T value 3.148*** 3.092*** 3.308*** 3.027*** 3.337*** 4.240*** 4.845*** 5.164*** 4.280*** Coefficient 0.000 0.013 0.023 0.045 0.057 0.073 0.066 0.061 0.054 EPS T value 0.016 0.598 1.173 2.154** 2.569*** 3.728*** 3.575*** 3.221*** 2.679*** Coefficient 0.004 0.074 0.154 0.105 0.169 0.162 0.138 0.227 0.279 GS T value 0.038 0.627 1.378 0.986 1.617 1.666* 1.524 2.236** 2.651*** Coefficient 0.200 0.223 0.180 0.184 0.206 0.177 0.179 0.182 0.210 PDIR T value 2.782*** 3.085*** 2.522** 2.454*** 2.733*** 2.773*** 2.827*** 2.766*** 3.599*** Coefficient 0.013 0.008 0.014 0.015 0.004 0.002 0.005 0.006 0.007 NDIR T value 0.596 0.359 0.662 0.746 0.218 0.143 0.329 0.385 0.376 Coefficient 0.057 0.009 0.002 0.002 0.020 0.028 0.036 0.027 0.043 NCCMT T value 1.159 0.276 0.058 0.092 0.721 1.219 1.716* 1.022 1.251 Coefficient 0.023 0.038 0.070 0.123 0.120 0.120 0.130 0.168 0.061 SO T value 0.254 0.444 0.852 1.544 1.566 1.900* 2.045** 2.461** 1.001 Sample Size N=120 N=120 N=120 N=120 N=120 N=120 N=120 N=120 N=120 Note: * P< .10, ** P< .05, P<0.01 73 Figure 48. The coefficient graphs of the Hotel & Casino subsample by quantile regression 74 Results of quantile regression for the Restaurant subsample Table 47 shows the results of the quantile regression analysis for the coefficient estimates of the model with the Restaurant subsample. The quantile regression provided more specific results than the OLS regression for restaurant subsample, even though the result of the quantile regression analysis for the restaurant subsample was similar to that of the OLS regression results in this subsample. In the quantile regression, unlike the results from the OLS regression, both CR and AT were correlated with executive total cash compensation at the lower quantile (0.1  0.3 quantile) of executive compensation at the alpha level of 0.05. In addition, the coefficient graphs for both CR and AT (Figure 49) show that their coefficient values decreased as the level of executive compensation increased. This suggests that executives at lower levels of compensation were more sensitive to both CR and AT than were executives at the middle and upper levels of compensation. The result of the quantile regression also showed that DT was not significantly related to executive compensation, while the result from the OLS regression showed the opposite. In addition, the coefficient graphs for the DT variable (Figure 49) show that the pattern of the coefficient value for DT was one of moderate volatility from the lower quantile to the upper quantile of executive compensation. It indicates that there is no different impact of DT on different level of executive compensation in Restaurant industry. The quantile regression analysis provided similar results for the FS variable as that of the OLS regression that FS was statistically significantly related to executive compensation at an alpha level of 0.05. However, the pattern of coefficients of the FS 75 variable decreased from the lower quantile of executive compensation to the middle quantile, then increased in the upper quantile. Thus, the low and high levels of executive compensation were more related with firm size than was the middle range. The results of the quantile regression also showed that EPS was not significantly correlated with executive total cash compensation for any quantile of executive compensation at an alpha level of 0.05, which was the same as the result from the OLS regression. However, the coefficient graphs for the EPS variable in Figure 49 show that the coefficient value for EPS increased from the lower quantile of executive compensation to the upper quantile, indicating that the executives at the lower level of compensation were less sensitive to EPS than were the executives at the middle and upper levels. Like the OLS regression result, the quantile regression also shows that the GS variable was not significantly correlated with executive total cash compensation for any level of executive total cash compensation at an alpha level of 0.05. In addition, the coefficient graphs for the GS variable (Figure 49) show moderate volatility of the coefficient from the lower quantile to the upper quantile of executive compensation. Thus, GS was not related to the level of executive compensation. The results of the quantile regression analysis for the four managerial power variables were not much different from those of the OLS regression. PDIR was significantly correlated with executive total cash compensation at the 0.05 level, while NDIR and NCCMT were not. However, the quantile regression for SO shows that SO was significantly correlated with executive total cash compensation only for executives at the upper level of compensation (0.70.9 quantile) at an 0.05 alpha level, while the OLS 76 regression shows that it was significantly related to executive total cash compensation in general. The coefficient pattern of SO (Figure 49) increased from the low quantile to the high quantile of compensation, suggesting that the effect of SO on executive compensation increase when the level of compensation increases. However, the pattern of coefficients for PDIR had moderate volatility, suggesting that there is no different effect of PDIR on different level of executive compensation. Furthermore, the pattern of the NDIR coefficient increased from the lower quantile to the upper quantile, which indicates that the executives at a lower level of compensation were less sensitive toward NDIR than were the executives at the middle and upper levels of compensation. However, the pattern of the NCCMT coefficient decreased from the lower quantile to the upper quantile, indicating that the executives at a lower level of compensation were more sensitive to NCCMT than were the executives at the middle and upper levels of compensation. After the quantile regression analysis, hypotheses H61, H62, H64, and H67 were accepted at 0.05 level, but H63, H65, H6 6, H68, H69, and H610 were not accepted at the 0.05 level. 77 Table 47. Quantile regression summary for the Restaurant subsample Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 Coefficient 4.465 4.504 4.488 4.614 4.662 4.846 4.841 4.884 4.999 (Intercept) T value 26.656*** 25.263*** 23.571*** 22.672*** 22.499*** 21.782*** 21.022*** 23.965*** 36.805*** Coefficient 0.061 0.050 0.053 0.025 0.017 0.017 0.024 0.009 0.029 CR T value 2.662*** 2.012** 2.004** 0.901 0.585 0.611 0.874 0.307 1.017 Coefficient 0.097 0.093 0.097 0.060 0.053 0.001 0.001 0.008 0.072 AT T value 2.262** 2.109** 2.106** 1.299 1.155 0.015 0.033 0.199 2.282*** Coefficient 0.030 0.037 0.059 0.058 0.066 0.070 0.053 0.077 0.094 DT T value 0.594 0.678 1.084 1.069 1.230 1.410 1.116 1.548 1.921 Coefficient 0.275 0.280 0.289 0.272 0.273 0.234 0.255 0.248 0.203 FS T value 5.345*** 5.336*** 5.107*** 4.744*** 4.744*** 4.080*** 4.364*** 4.720*** 5.258** Coefficient 0.016 0.017 0.007 0.011 0.016 0.016 0.019 0.043 0.058 EPS T value 0.605 0.578 0.242 0.368 0.522 0.527 0.676 1.536 2.295** Coefficient 0.108 0.124 0.040 0.037 0.046 0.172 0.248 0.306 0.220 GS T value 1.023 1.032 0.315 0.262 0.301 1.000 1.508 1.819* 1.133 Coefficient 0.127 0.160 0.167 0.159 0.190 0.181 0.153 0.220 0.162 PDIR T value 2.827*** 3.510*** 3.441*** 3.080*** 3.731*** 3.371*** 2.888*** 4.101*** 3.052*** Coefficient 0.004 0.008 0.002 0.005 0.003 0.012 0.012 0.015 0.042 NDIR T value 0.292 0.689 0.170 0.356 0.232 0.882 0.905 1.087 3.033*** Coefficient 0.002 0.006 0.005 0.007 0.007 0.009 0.012 0.017 0.020 NCCMT T value 0.144 0.383 0.265 0.420 0.402 0.586 0.766 1.197 1.556 Coefficient 0.055 0.089 0.077 0.080 0.048 0.067 0.098 0.113 0.158 SO T value 1.169 1.916* 1.568 1.541 0.941 1.335 1.983** 2.221** 2.922** Sample Size N=193 N=193 N=193 N=193 N=193 N=193 N=193 N=193 N=193 Note: * P< .10, ** P< .05, P<0.01 78 Figure 49. The coefficient graphs of the Restaurant subsample by Quantile regression 79 CHAPTER V CONCLUSION 1. Summary of the study This study has investigated the determinants of executive compensation in the hospitality industry with selected variables from both the payforperformance rule and the managerial power approach, using two multiple regression analysis methods: OLS regression and quantile regression. The study provides an empirical illustration of determinants of executive compensation in the hospitality industry as a whole, and in two subcategories, the Hotel & Casino category and the Restaurant category. OLS regression analysis was performed first to identify the determinants of executive compensation in the hospitality industry on the basis of both the payforperformance rule and the managerial power approach. At the second stage of analysis, quantile regression analysis was adopted to examine whether the independent variables were differently related to different levels of executive compensation in the hospitality industry. A summary and discussion of the empirical findings of this study are presented in the following sections. 80 Summary of the Full Sample: The Hospitality Industry OLS regression analysis The results of OLS regression analysis for the hospitality industry revealed that three financial variables, DT, FS, and EPS, and two managerial power variables, PDIR and SO, were positively related to executive compensation with statistical significance. Thus, payforperformance rules and managerial power variables both influenced executive compensation in the hospitality industry. While the financial variables suggested that firm size (FS) and firm profitability (EPS) positively affected executive compensation, the study found a different result from prior studies of executive compensation in the hospitality industry by showing that coverage ratio was positively related to executive compensation and that CR, AT, and GS were not. The results of the coverage ratio analysis revealed that executives in the hospitality industry were paid highly despite increasing risk to the firm. In general, a company with a higher debt ratio has a riskier financial status because high debt means a heavy interest burden and the need to repay principal (Chatfield & Dalbor, 2005) However, Jensen and Meckling (1976) suggested that agents (executives) in companies with high debt capital structures receive more compensation because of the incentive effects associated with debt: The agent is paid more for being willing to take on the challenge of activities which offer the possibility of very high payoffs, even when they have a very low probability of success. Such activities invoke an agency problem because the shareholders prefer that the company does not remain in a risky environment, but executives may prefer to invest in more risky projects in order to receive higher 81 compensation from big successes with risky projects. This result implies that there may be an agency problem in the hospitality industry. Three other financial variables—CR, AT, and GS—were not significantly related to executive compensation, so the hospitality industry only partially follows the payforperformance rule in determining executive compensation. Regarding the managerial power approach, the PDIR and SO variables were identified as determinants of executive compensation in the hospitality industry, while NDIR and NCCMT were not. These results support the idea that executives who serve on the board of directors receive more compensation than those who do not, regardless of the number of board members or the number of compensation committee members. Stock ownership by the executive also positively influenced executive compensation, so, in addition to payforperformance rules, how the executive was related to corporate governance influenced executive compensation. Quantile regression analysis The results of the quantile regression analysis provided more specific and concise results for the hospitality industry by examining the effects of each independent variable on different levels of executive compensation. The quantile regression analysis showed that the level of executive compensation was differently related to each independent payfor performance and managerial power variable. The variables were related to three different group of compensation—the lower level (0.10.3 quantile), the middle level (0.40.6 quantile), and the upper level (0.70.9 quantile). 82 Among the financial variables, the lower level of executive compensation was significantly related to CR and AT, while the middle and upper levels were significantly related to EPS and GS. That is, firm liquidity and efficiency were determinants of lower level compensation, while profitability in terms of EPS and GS determined middle and upper levels of executive compensation. In addition, the FS variable was significantly related to all levels of executive compensation, although the sensitivity of the lower and upper levels was greater than that of the middle level. Meanwhile, the coverage ratio (DT) was significantly related to the full range of executive compensation, although it was moderate. Thus, firm coverage was not differently related depending on the level of compensation. Among the four managerial power variables, only the stock ownership variable was differently related to levels of executive compensation in that the upper level of executive compensation was more sensitive than either the lower or middle levels. PDIR was significantly but moderately related to all levels of compensation, so it was not differently related to different levels of compensation. The results of the quantile regression analysis of the financial and managerial power variables also suggested that executive stock ownership and board independence (from the managerial power approach) influenced executive compensation in the hospitality industry, and that the hospitality industry also partially follows the payforperformance rule to determine executive compensation. It concluded that the hospitality industry weakly follows payforperformance rule to determine executive compensation and higher executive’s stock ownership and board nonboard independent from top executive may also influence on determining the level of executive compensation in 83 hospitality industry. In addition, it supports that the determinants of executive compensation differ between different groups of executive compensation level in the hospitality industry. Summary of Hotel & Casino Subsample OLS regression analysis The results of the OLS regression analysis provided a result similar to that of the full sample (the positive correlation of FS, EPS, DT, PDIR, and SO) for both financial variables and managerial power variables, except that GS was significantly related to compensation in the hotel and casino industry. Four financial variables (FS, EPS, GS and DT) were positively related to executive compensation, and the positive relationship between coverage ratio and executive compensation suggested that there may be a serious agency problem in the hotel and casino industry. In short, the results showed that the hotel and casino industry partially follows the payforperformance rule in determining executive compensation. The results of the managerial power approach were the same as that for the full sample in that the PDIR and SO variables were identified as determinants of executive compensation. Thus, the involvement of the executive in corporate governance influences 84 Quantile regression analysis The results of the quantile regression analysis were similar to those for the hospitality industry as a whole, supporting the idea that different levels of executive compensation are differently related to the independent variables. Among the financial variables for the payforperformance rule, the lower level of executive compensation was significantly related to only FS and DT. The FS variable was significantly related to all levels of executive compensation, although the sensitivity of FS on the lower and upper levels of compensation was more than it was for the middle level. However, coverage ratio (DT) was significantly related only to the lower level of executive compensation with a positive coefficient, but the sensitivity of DT on the level of executive compensation was moderate. Thus, firm coverage was not differently related to different levels of executive compensation. In addition, the coefficient graph of CR and AT showed that the lower level of compensation was more sensitive than were the middle and upper levels, which had no statistical significance. A cautious interpretation of this finding is that the firm’s liquidity and efficiency could have more influence on the lower level of executive compensation than on the middle and upper levels. In contrast to the lower level of compensation, the middle and upper levels were significantly related to EPS, and firm profitability was a determinant of the middle and upper levels of compensation. In addition, sales growth had different affects on the different levels of executive compensation, based on the coefficient graph of GS, which showed that the lower level of executive compensation was less sensitive to GS than were the middle and upper levels. 85 Among the four managerial power variables, the result for the Hotel & Casino subsample was the same as that of the hospitality industry as a whole, except that the stock ownership variable was differently related with levels of executive compensation, while the other three variables—PDIR, NDIR, and NCCMT—were not. The results of the quantile regression analysis of the financial and managerial power variables also suggested that executive stock ownership and board independence influenced executive compensation in the Hotel & Casino industry, and that the industry also partially follows the payforperformance rule to determine executive compensation. While the influence of the payforperformance rule is weak, higher executive’s stock owner
Click tabs to swap between content that is broken into logical sections.
Rating  
Title  Examination of Executive Compensation Determinants in the Hospitality Industry: a Quantile Regression Approach 
Date  20080701 
Author  Kim, Sang Hyuck 
Keywords  Hospitality Administration 
Department  Hospitality Administration 
Document Type  
Full Text Type  Open Access 
Abstract  This study has investigated the determinants of executive compensation in the hospitality industry with selected variables from both the payforperformance rule and the managerial power approach, using two multiple regression analysis methods: OLS regression and quantile regression. OLS regression analysis was performed first to identify the determinants of executive compensation in the hospitality industry on the basis of both the payforperformance rule and the managerial power approach. At the second stage of analysis, quantile regression analysis was adopted to examine whether the independent variables were differently related to different levels of executive compensation in the hospitality industry. Both the OLS regression analysis and the quantile regression analysis supported that the payforperformance rule and the managerial power approach influence executive compensation in the hospitality industry. The results of OLS regression analysis for the hospitality industry revealed that three financial variables, firm debt ratio, firm size, and EPS, and two managerial power variables, PDIR (current director or not) and executive stock ownership, were positively related to executive compensation with statistical significance. In addition, the quantile regression analysis showed that the level of executive compensation was differently related to each independent payforperformance and managerial power variable. In other words, the different measures could be utilized to determine different levels of executive compensation. 
Note  Dissertation 
Rights  © Oklahoma Agricultural and Mechanical Board of Regents 
Transcript  EXAMINATION OF EXECUTIVE COMPENSATION DETERMINANTS IN THE HOSPITALITY INDUSTRY: A QUANTILE REGRESSION APPROACH By SANG HYUCK KIM Bachelor of Science in Business Administration DongGuk University Seoul, Korea 2000 Master of Business Administration Western Illinois University Macomb, Illinois 2002 Master of Science in Accountancy University of Illinois UrbanaChampaign, Illinois 2003 Submitted to the Faculty of the Graduate College of the Oklahoma State University in partial fulfillment of the requirements for the Degree of DOCTOR OF PHILOSOPHY July, 2008 ii EXAMINATION OF EXECUTIVE COMPENSATION DETERMINANTS IN THE HOSPITALITY INDUSTRY: A QUANTILE REGRESSION APPROACH Dissertation Approved: Dr. Jerrold K. Leong Dissertation Adviser Dr. Woo Gon (Woody) Kim Dr. Murat Hancer Dr. William D. Warde Dr. A. Gordon Emslie Dean of the Graduate College iii DEDICATION This dissertation is dedicated to my parents , Jong Ho Kim, and Myung Hee Park. iv ACKNOWLEDGMENTS There are several people without whose support and encouragement I could not have completed the doctoral degree. I offer my sincere appreciation to my advisor, Dr. Woody (Woo Gon) Kim, for his guidance, encouragement, understanding, and friendship throughout my studies. He made himself readily available for consultation and offered constructive criticism, and his steady support and encouragement have been critical during my studies. I also thank my committee chair, Dr. Jerrold K. Leong, for his guidance, patience, understanding, and support. I sincerely appreciate my other committee members, Dr. Murat Hancer and Dr. William Warde, who greatly assisted me by providing the feedback necessary to bring this study together. I am very grateful to my parents in Korea, who have given me their love and support during my graduate studies in the United States. None of my work would have been possible without their devotion and encouragement. Finally, I thank my wonderful wife, Soo Lyun Cho, for her love, support, and encouragement throughout my academic career. v TABLE OF CONTENTS Chapter Page I. INTRODUCTION......................................................................................................1 1. Background of study............................................................................................1 2. Research Motive and Problem Statement ............................................................5 Research Motive ................................................................................................5 Problem Statement .............................................................................................6 3. Significance of the Study.....................................................................................7 4. Purpose of the Study ............................................................................................8 5. Organization of the Study ....................................................................................9 II. REVIEW OF LITERATURE..................................................................................10 1. The Agency Theory ...........................................................................................11 2. Financial Determinants of Executive Compensation.........................................13 3. Managerial Power Determinants of Executive Compensation ..........................23 4. OLS regression and Quantile regression............................................................28 5. Development of Hypotheses ..............................................................................32 III. METHODLOGY ...................................................................................................36 1. Data Collection and Sampling Procedures ........................................................36 2. Variable selection...............................................................................................41 Selection of dependent variable .......................................................................41 Selection of independent variables ..................................................................42 3. Data Analysis and Model...................................................................................46 vi IV. FINDINGS.............................................................................................................49 1. Description of Sample........................................................................................49 2. Findings of Ordinary Least Square (OLS) Regression Method.........................57 Results of OLS regression method for the Full sample ...................................58 Results of OLS regression method for the Hotel & Casino subsample..........60 Results of OLS regression method for the Restaurant subsample..................61 3. Findings of the Quantile Regression..................................................................63 Results of the quantile regression method for the Full sample........................64 Results of the quantile regression method for Hotel & Casino subsample ....69 Results of the quantile regression method for Restaurant subsample ............74 V. CONCLUSION......................................................................................................79 1. Summary of the study ........................................................................................79 Summary of the Full Sample: The Hospitality Industry..................................80 Summary of the Hotel & Casino Subsample..................................................83 Summary of the Restaurant Subsample..........................................................85 2. Implication of study ...........................................................................................88 3. Limitations and Suggestions for Future Research .............................................92 REFERENCES ............................................................................................................94 APPENDICES ...........................................................................................................100 APPENDIX A. EXECUTIVES IN HOTEL INDUSTRY (2006) .......................101 APPENDIX B. EXECUTIVES IN CASINO INDUSTRY (2006) .....................103 APPENDIX C. EXECUTIVES IN RESTAURANT INDUSTRY (2006)..........106 vii LIST OF TABLES Table Page 21. The classification of financial variables from the previous studies...................22 22. The variables of managerial power approach used in previous studies ............28 31. Classification of study sample...........................................................................41 41. Descriptive Statistics of Executive Total Cash Compensation .........................50 42. OLS regression summary for the Full sample (all hospitality companies) .......59 43. OLS regression summary for the Hotel & Casino subsample .........................60 44. OLS regression summary for the Restaurant subsample .................................62 45. Quantile regression summary for the Full sample (all hospitality companies) .67 46. Quantile regression summary for the Hotel & Casino subsample ...................72 47. Quantile regression summary for the Restaurant subsample ...........................77 viii LIST OF FIGURES Figure Page 41. Scatter plot for full sample (All hospitality industry)........................................51 42. Scatter plot for subsample (Hotel & Casino industy) ......................................52 43. Scatter plot for subsample (Restaurant industry) .............................................52 44. Histogram and Normal PP Plot for full sample (All hospitality industry).......54 45. Histogram and Normal PP Plot for subsample (Hotel & Casino industry) ....55 46. Histogram and Normal PP Plot for subsample (Restaurant industry) ............56 47. The coefficient graphs of all hospitality industry by Quantile regression.........68 48. The coefficient graphs of Hotel & Casino subsample by Quantile regression 73 49. The coefficient graphs of Restaurant subsample by Quantile regression ........78 1 CHAPTER I INTRODUCTION 1. Background of study In the last few decades, the topic of executive compensation has received a great deal of attention from both academic empirical researchers and practitioners of business management, especially those from the finance and accounting fields (Andjelkovic, Boyle, & McNoe, 2002; GrabkeRundell & GomezMejia, 2002; Gray & Cannella, 1997). The dominant topic of executive compensation studies has focused on examining the relationship between the executive’s compensation and the firm’s performance (Mishra, McConaughy, & Gobeli, 2000; Perry & Zenner, 2001). That is, executive compensation studies have been conducted on the basis of the payforperformance rule in the agency theory (GrabkeRundell & GomezMejia, 2002). According to the agency theory, proposed by Jensen and Meckling (1976), compensation packages should balance compensation value and the executive’s managerial performance by implementing and utilizing an appropriate payforperformance rule aligned to motivate the agent (the executive in this study), to attract and retain management talent, and to increase management performance in order to maximize shareholder wealth (Gu & Choi, 2004; 2 Kim & Gu, 2005; Perlik, 2002). In other words, the agent’s compensation contract should lead the executives of the firms to try to increase the firm’s performance, thereby achieving the goal of maximization of shareholder’s wealth through an increase in the firm’s stock price and a stable flow of dividends (Lippert & Porter, 1997; Grabke Rundell & GomezMejia, 2002). The payforperformance rule, then, supports the idea that the level of an executive’s compensation should be closely and positively linked to the firm’s performance (Hallock, 1998; Jensen & Murphy, 1990; Kato & Kubo, 2006). Because of theoretical confidence in the payforperformance rule, it has become an increasingly popular measure in agency theory research (Lippert & Porter, 1997; Perry & Zenner, 2001). Even so, the payforperformance rule has not always been supported by the empirical results of executive compensation studies (Andjelkovic et al., 2002; Gray & Cannella, 1997). As the numbers of studies that have found other affects on executive compensation, such as executive’s demographic characteristics and the structure of corporate governance, have increased, the support for the payforperformance rule has decreased (GomezMejia, Tosi, & Hinkin, 1987; Hebner & Kato, 1997; Nelson, 2005). In addition, the increasing attention on payforperformance among the public stimulated the development of regulations in the United States (Perry & Zenner, 2001). The United States Securities and Exchange Commission (SEC) announced a new compensation disclosure rule, beginning with the fiscal year of 1992, which required publicly held companies to include top executives’ compensation disclosures in annual proxy statements to the SEC (Vafeas & Afxentiou, 1998). Congress also established tax legislation, Section 162(m) of the Internal Revenue Code, to limit executive’s 3 compensation’s deduction for nonperformancerelated executive compensation to US$1 million in the publicly traded companies (Perry & Zenner, 2001). Both the SEC regulation and the tax legislation were expected to help determine clearer and more appropriate levels of executive compensation in U.S. publicly traded companies by encouraging companies to relate compensation to company performance (Perry & Zenner, 2001; Vafeas & Afxentiou, 1998). Today, the compensation packages of executives in publicly traded companies still have been spent huge amounts of money and have continually increased in value in order to attract and retain executives. For example, in 2006, Goldman Sachs’ CEO, Lloyd Blankfein, received compensation totaling $55 million in cash, stock options and restricted stock, a 76% increase in pretax compensation from 2005. John Mack, CEO of Morgan Stanley, increased his compensation to $41 million in 2006, a 43% increase from 2005 (Hahn, 2007). In the retail industry, George L. Jones, the president and CEO of book retailer Borders Group, Inc., received $3.37 million in compensation during fiscal year 2006 (Financial Times Information, 2007b). Contrary to above examples of increase in top executive compensation, some top executives have voluntarily reduced their annual salaries, sometimes drastically. For example, Roger Enrico, CEO of PepsiCo, dropped his $900,000 salary to $1 in 1998, 1999 and 2000 and donated his previous salary to scholarships for employees’ kids. Steve Miller, CEO of Delphi, dropped his salary from $1.5 million to a $1 after the company filed for bankruptcy protection. Rick Wagoner, GM’s CEO, cut his salary almost 50% in 2005 and volunteered for another 50% cut in his $2.2 million salary in 2006 (Kempner, 2007). Although some examples show that top executive’s compensation is decreased by several reasons, it is true that most industries 4 still pay huge amounts of money to acquire and keep talented—and sometimes notsotalented— executives. While the value of executives’ compensation packages can vary in response to such factors as firm performance, the structure of other companies’ compensation packages, and voluntary cuts by the executive himself or herself, questions about efficient and appropriate executive compensation packages in publicly traded firms have increased as executive compensation has increased (GrabkeRundell & Gomez Mejia, 2002). Since the agency theory was proposed by Jensen and Meckling (1976), numerous studies have been undertaken to find determinants of executive compensation. At the initial stage of executive compensation research, most studies were concerned with determining how executive compensation relates to financial performance standards (Carr, 1977; Core, Holthausen, & Larcker, 1999; Firth, Tam, & Tang, 1999). As mentioned earlier, the payforperformance rule has not always been supported by the empirical results of executive compensation studies. Thus, some researchers made efforts to extend executive compensation study by adding other factors, especially factors from managerial power approach (Yermack, 1995; Core et al., 1999; Hallock, 1997; Bebchuk & Fried, 2003; Grinstein & Hribar, 2004). Grinstein and Hribar (2004) stated the “managerial power approach,” which presents that compensation based on the payforperformance rule did not work optimally to enforce agents to maximize shareholder wealth if the agent had powers or influence over board decisions, including compensation decisions (Grinstein & Hribar, 2004). Based on the concept of the managerial power approach, researchers who questioned the payforperformance rule found that the characteristics of ownership structure and corporate governance also affected executive 5 compensation. As a result, research that investigates determinants of executive compensation should include variables such as ownership structure, number of board members, and whether the executive is on the company’s board of directors, so that both the payforperformance rule and the managerial power approach are considered in finding the determinants of executive compensation. 2. Research Motive and Problem Statement Research Motive The hospitality industry is not much different from other industries when it comes to compensation for top executives. For example, in 2004, Starwood Hotels & Resorts appointed Maven Steven Heyer, former president and COO of the CocaCola Co., as its new CEO, with a $1 million annual base salary for a fouryear initial term (Parets, 2004). The total compensation of David Brandon, CEO of Domino’s Pizza, Inc., increased from $1.81 million in 2004 to $21.9 million in 2006 (Snavely, 2006). The CEO of McDonalds, Jim Skinner, received $8.8 million in bonuses from 2004 to 2006 (Financial Times Information, 2007a). As in other industries, not all top executives in the hospitality industry have received huge compensation. The CEO of Planet Hollywood International, Robert Earl, was paid half of his $600,000 annual salary in 2001 because of bankruptcy (Schneider, 2007). Tim Taft was appointed as the new CEO of Pizza Inn with a firstyear salary of $1, although he received stock options (RobinsonJacobs, 2005). These are examples of 6 the hospitality industry’s following the payforperformance rule, although there are many exceptions. For example, when Denny’s restaurant faced a loss of $88.5 million in company earnings before interest and taxes in 2002, the CEO received a $1.3 million bonus (Perlik, 2002). On the other hand, Joseph P. Martori, CEO of ILX Resorts Incorporated, named the numbertwo topperforming CEO in HVS International’s 2002 Survey and beating out the CEOs of the Four Seasons, Marriott International, Starwood, Hilton and others, was one of the lowestpaid CEOs in the hotel industry, ranking 45th among 51 hospitality industry CEOs (Business Wire, 2003). Sometimes, then, the payfor performance rule does not explain the determinants of top executives’ compensation in hospitality industry well. Clear understanding of the determinants of executive compensation is necessary for stockholders or potential investors in the hospitality industry to judge whether the appropriate compensation is awarded. Although the hospitality companies have spent large amounts on executives’ compensation packages, little research has been done to investigate how that compensation is determined in the industry. Previous literature related to executive compensation in the hospitality and tourism field has examined the determinants of CEO’s compensation only with regard to either financial variables from the firm’s performance (Gu & Choi, 2004; Kim & Gu, 2005) or to gender difference (Skalpe, 2007). Problem Statement Although previous studies have expanded our knowledge of what determines executive compensation in the hospitality industry, it remains uncertain whether 7 hospitality companies follow only the payforperformance rule or whether other factors have an influence on determining executive compensation in the hospitality industry. 3. Significance of the Study Most literature related to executive compensation in the hospitality field has focused on financial determinants from the payforperformance rule (Gu & Choi, 2004; Kim & Gu, 2005; Skalpe, 2007). While the ownership structure and/or corporate governance variables from the managerial power approach have also been considered to be among the determinants of executive’s compensation for academic fields and other industries, to the best of my knowledge, there is no study that has considered whether the managerial power approach is a determinant of executive compensation in the hospitality industry. Therefore, this study combines the payforperformance rule and the managerial power approach, using both the financial variables from the payforperformance rule and the ownership and corporate governance variables from the managerial power approach, to investigate the determinants of executive compensation in the hospitality industry. In addition, Dyl (1988) found that the different types of industry influence on determining management compensation level. Other researchers also adopted a type of industry as dummy variable in their studies to examine whether the different type of industry influences on the level of the executive compensation (Dyl, 1988; Hallock, 1997; Yermack, 1995). Thus, this study also attempts to examine whether there is a difference 8 between different sectors (i.e., the hotel & casino vs. restaurant) in the hospitality industry regarding determinants of executive compensation. In terms of methodology, most research on executive compensation has used traditional multiple regression, such as Ordinary Least Square (OLS) regression and Weighted Least Square (WLS) regression analysis, to investigate the relationship between total cash compensation and financial variables from the payforperformance rule. The current study adopted Quantile regression analysis, which was developed by Koenker and Basset (1978), to allow examination of whether different levels of total cash compensation are related differently to each independent variable from the payforperformance rule and the managerial power approach. Unlike traditional multiple regression analysis, Quantile regression analysis is operated by a conditional quantile function that estimates the relationship between each independent variable and each segment (quantile) of the dependent variables. For this study, then, Quantile regression will allow us to investigate how each independent variable is related to each different segments of the executive’s total cash compensation. 4. Purpose of the Study The primary objective of this study is to examine whether the financial performance variables from the payforperformance rule and a company’s ownership and corporate governance structure from the managerial power approach is related to 9 executive compensation in the hospitality industry. More specifically, the purpose of this study is to: 1) Identify the determinants for executive compensation in the hospitality industry in terms of both the payfor performance rule and the managerial power approach; 2) Examine whether there is a difference between different sectors (i.e., the hotel & casino and restaurant) in the hospitality industry regarding determinants of executive compensation; and 3) Investigate whether different levels of executive compensation are differently related to selected variables from both the payforperformance rule and the managerial power approach in the hospitality industry. 5. Organization of the Study The composition of this study is as follows. Chapter I, the introduction section, presents the background, research motives, significance, and the purpose of the study. Chapter II, the literature review section, reviews previous literature dealing with agency theory, executive compensation with the payforperformance rule and managerial power approach, comparisons between OLS regression and Quantile regression analysis, and development of hypotheses for this study. Chapter III explains the research methodology, including data collection, sampling procedures, and data analysis and models. Chapter IV addresses the empirical results of the study and, finally, Chapter V concludes and discusses the study’s implications, contributions, and limitations. 10 CHAPTER II REVIEW OF LITERATURE Numerous studies have tried to verify what factor(s) determine executive compensation. Two main streams of research concerning executive compensation in the finance and accounting fields have emerged over the last 70 years. The fundamental difference between the two streams of research lies in the theoretical foundations of executive compensation: the payforperformance rule from agency theory, which focuses on the relationship between executive compensation and firm performance; and the ownership structure and corporate governance from managerial power approach, which emphasizes that, because the payforperformance rule does not always work to determine executive’s compensation, other factors, such as whether the executive is involved in company ownership, board size, and whether the executive is a board member, also influence the executive’s compensation. While neither approach is perfect in explaining what determines executive compensation, they each have their advantages and disadvantages. A current trend in the literature is to combine both approaches. The goal of this literature review is to address the previous studies regarding the effect of financial determinants and managerial power on executive compensation and to identify relevant variables and methodologies used in previous studies. This chapter has 11 five main sections. The first section summarizes the agency theory, proposed by Jensen and Meckling (1976). The next two sections present the prior studies of executive compensation determinants from the payforperformance and the managerial power approaches, respectively. The fourth section compares OLS regression with Quantile regression for this study. The last section proposes the hypotheses for the study. 1. The Agency Theory In traditional financial theory, the primary goal of business management is maximization of stockholder wealth in terms of maximization of the firm’s market value. Because of this, those who invest money in the company expect executives not only to improve their business processes but to increase the company’s value (Brigham, Gapenski, & Ehrhardt, 1999), so spending money for executive compensation is one of the investments shareholders make to maximize their wealth. In contrast to the traditional financial theory, Jensen and Meckling (1976) proposed “the agency problem” within the agency theory, which is that there is a conflict between the agent’s interests and the shareholders’ interests because of the separation of management from ownership (Jensen & Meckling, 1976). In other words, the agent (the executives, in this case) is more likely to pursue personal interests or goals than to maximize shareholders’ wealth (Dyl, 1988; Jensen & Meckling, 1976; Traichal, Gallinger, & Johnson, 1999). This conflict between agent and shareholder evokes several types of costs for shareholders. This “agent cost” is composed primarily of three types of 12 costs: monitoring cost, bonding cost, and residual loss. Monitoring cost is the cost for the principal (the shareholder in this case) to limit the discretionary behavior of the agent. Bonding cost refers to the costs for the agent (the executive, in this case) to guarantee his or her discretionary behavior. Residual loss is loss from conflicts between principals and agents (Dyl, 1988; Jensen & Meckling, 1976). Several remedies have been proposed to solve the agency problem, including monitoring the agent’s discretionary behavior and controlling the agent’s compensation packages (Dyl, 1988; Jensen & Meckling, 1976; Traichal et al., 1999) by providing sufficient agent compensation to motivate the executive to work toward the best interests of shareholders, i.e., achieving maximization of shareholder wealth (Kim & Gu, 2005). Jensen and Meckling (1976) also suggested that executive compensation could be determined by means of the payforperformance rule, which would establish an optimal compensation contract between agent and principal. According to the payforperformance rule in agency theory, the agent’s compensation should be determined by practical and reliable measures of firm performance, that is, the level of the agent’s compensation would be commensurate with his or her contribution to the firm’s value. In this context, compensation should be based on observable measures, such as market returns or profitability ratios, which maximize the value of a firm (Grinstein & Hribar, 2004). The payforperformance rule is frequently utilized as a standard by which to determine executive compensation by using firm performance (Gu & Choi, 2004). Thus, many extant studies have investigated the relationship between executive compensation and firm performance using several key financial variables, including firm size and 13 several types of firm performance measures (Anderson, Becher, & Campbell, 2004; Grinstein & Hribar, 2004; Gu & Choi, 2004, Kato & Kubo, 2006; Kim & Gu, 2005). 2. Financial Determinants of Executive Compensation Most executive compensation studies have adopted the firm’s performance in terms of the firm’s financial data, as estimators of each executive’s performance because of the difficulty of collecting relevant or sufficient data regarding executives’ direct contribution on firm performance. In other words, it is difficult to estimate each executive’s direct performance on the firm’s performance with financial and mathematical figures. The agency theory also suggests that executives’ managerial performance leads to improvement in the firm’s performance, which, in turn, links to increasing shareholder wealth (Gu & Choi, 2004). At the initial stage of executive compensation research, especially after Jensen and Meckling proposed the agency theory in 1976, the financial measures from the payfor performance rule were probably the most common measure utilized in research on determinants of executive compensation (Bebchuk & Fried, 2003; GomezMejia &Wiseman, 1997). Even though some critics decry the financial measures from the payfor performance rule, numerous studies have utilized the financial measures for firm performance. Thus, following these prior studies provides the theoretical justification for the current study to utilize relevant variables for measuring firm performance. 14 To measure the firm’s performance, researchers have adopted several types of financial figures as relevant variables. The three dominant streams of measurement in prior empirical executive compensation studies in the finance and accounting fields are marketbased measurements, accountingbased measurements, and growthbased measurements. First, the company’s market return in terms of stock returns is an indirect measure of a firm’s marketbased performance, because it is an important indicator of its business performance and shareholder wealth (Jensen & Murphy, 1990; Leone, Wu, & Zimmerman, 2006). Financial figures such as return on assets, earnings per share, and return on equity, are accountingbased measures of firm performance. The accountingbased ratio analysis is one of the tools used by financial managers and financial analysts to evaluate the financial position or performance of a firm (Chatfield & Dalbor, 2005). Finally, many studies have utilized the firm’s sales growth as a growthbased determinant of executive compensation (Firth et al., 1999; Kato & Kubo, 2006). Many prior studies of executive compensation based on the payforperformance rule have also used firm size as a control variable (Andjelkovic et al., 2002). Studies that followed often adopted these four financial measurements of firm performance in executive compensation research. Jensen and Murphy (1990) examined the association between top management’s pay and performance, adopting shareholder wealth in terms of stock returns as an estimator of managerial performance. The study found that top management compensation is highly sensitive to the stock returns of the company. One other example of a study regarding the sensitivity of stock returns on executive compensation was conducted by Leone et al. (2006). The authors examined the sensitivity of CEO cash compensation to stock returns with 9,858 CEOs in the ExecuComp database from 1993 to 15 2003 and found that CEO cash compensation is twice as sensitive to negative stock returns as it is to positive stock returns. The results supports that company’s stock return positively influence on determining CEO cash compensation. In addition, the reducing amount of CEO cash compensation in company with negative stock return is bigger than the increasing amount of CEO cash compensation in company with positive stock return. Gray and Cannella (1997) examined the role of firm’s risk in executive compensation, using several financial figures to identify the relationship between executive compensation and firm risks, return on sales, and firm size. The results of the study provided that firm risks have a significantly negative relationship with executive total compensation and firm size, and Jensen’s alpha has a significantly positive association with executive total compensation. The findings from this study supports that the executive compensation is determined by firm’s performance. Furthermore, the executive compensation is reduced when firm’s risk increases, as well as the executive compensation is increased when firm’s size and firm’s performance increase. Duru and Iyengar (1999) conducted a crosssectional research analysis of 225 firms in the electric utility industry (SIC code 4931) from 1992 to 1995 to examine the association between firm performance and CEO compensation components. The authors adopted the change in CEO compensation as the dependent variable and the changes in the firm’s financial figures as multiple independent variables to examine the sensitivity of CEO compensation to changes in firm performance. They used market returns, return on assets, earning per share, operating cash flow per share, and growth in sales to measure financial performance and showed a positive relationship between changes in compensation and changes in firm performance. More specifically, executive bonuses 16 were sensitive to changes in market return, and executive stock options were sensitive to changes in sales growth. Several studies have examined the relationship between executive compensation and firm performance during special events, like mergers and acquisitions (M&A). Anderson et al. (2003) investigated bank CEOs’ managerial incentives for bank mergers as they related to financial variables such as firm size and stock returns, and found that CEO compensation was in line with an increase in bank size, regardless of whether a merger or acquisition created value. Grinstein and Hribar (2004) also used financial variables, including firm size and ROA, stock return, and sales growth, to examine the determinants of CEOs’ bonuses for 327 large ($1 billion or more) M&A deals in publicly traded U.S. companies between 1993 and 1999. They found the firm size, ROA, stock return, and the acquisition dummy to be positively correlated with CEOs’ bonuses for the M&A deal. Some studies of executive compensation determinants have been performed in countries outside the U.S., such as Japan (Kato & Kubo, 2006), England (Eichholtz, Kok, & Otten, 2008), and China (Firth et al., 1999). Kato and Kubo (2006) examined the relationship between executive compensation and firm performance for Japanese firms from 1986 to 1995 using marketbased firm performance (stock returns), accountingbased firm performance (return on asset), growthbased firm performance (sales growth), and firm size. The results of this study supported that Japanese CEOs’ cash compensation was sensitive to firm performance (especially accountingbased performance) and that the bonus system made CEO compensation more responsive to firm performance. Eichholtz et al. (2008) used samples from 39 companies in the UK property industry from 1998 to 17 2003 and variables from both firm performance and corporate governance—total stock performance, Jensen alpha, earnings per share, dividend yield, and discount—to investigate the association between executive compensation and firm performance. They found that stock performance, Jensen alpha, earnings per share, and discount were not significantly related to executive cash compensation but that dividend yield was significantly negatively related to executive cash compensation. Thus, the study found a weak association between executive cash compensation and the payforperformance rule. Firth et al. (1999) used a sample of companies in Hong Kong and several variables from both the payforperformance rule and the managerial power approach— annual stock return, firm size, return on shareholder equity, and annual compound sales growth—and found that the companies in Hong Kong followed payforperformance rule, by showing that company size and accounting profitability are significantly related with executive compensation. Thus, executive compensation studies of three different countries cautiously supported an association between executive compensation and firm performance. Most research has measured firm performance with marketbased, accountingbased (mostly profitability measures), and growthbased measures, as well as a control variable for firm size. However, other types of accounting based financial ratios have been used to represent for firm performance in the accounting literature. The basic accountingbased financial ratios are generally divided into four categories: liquidity, activity, profitability, and coverage. Liquidity ratios are used to measure the company’s shortrun ability to pay its maturing obligations, activity ratios measure how effectively 18 and efficiently a company uses its assets, profitability ratios are measures of the degree of success or failure of company for a given period of time, and coverage ratios measure the protection of longterm creditors and investors (Brigham et al., 1999; Chatfield & Dalbor, 2005; Gallagher & Andrew, 1997; Kieso, Weygandt, & Warfield, 2001). Although most studies have adopted profitability ratios for firm performance from among the four classified accounting based ratio analyses, other ratios, like liquidity, activity and coverage ratios, might also be considered as measures of firm performance for the current study. OrtizMolina (2007) stated that executive compensation may not only depend on shareholder opinion, because the bondholders (debtors) but also have the power to influence executive compensation. Thus, OrtizMolina found that executive compensation was significantly sensitive to the types of debt in a company. Traichal et al. (1999) affirmed the importance of liquidity ratios and coverage ratios in executive compensation and adopted a modified liquidity ratio (ratio of shortterm debt divided by total assets) and coverage ratio (ratio of longterm debt divided by total assets). Since those two liquidity and coverage ratios are measurements that show the ability of a firm to pay back both shortterm and longterm debt, those two ratios may also be considered measurements of firm performance. In addition, some studies of executive compensation in the hospitality field have included all four types of accountingbased financial measurements (liquidity, efficiency, profitability, and coverage) as independent variables for their executive compensation studies (Gu & Choi, 2004; Kim & Gu, 2005). In a related study in the hospitality field, Cauvin (1979a) investigated the relationship between executive total compensation and company size, represented by sales, with 33 19 lodging companies in the U.S. Since the study was conducted before the SEC’s regulation requiring disclosure of top executive compensation disclosure was announced, the data for this study were collected from surveys based on the 1978 directory of Hotel/Motel Systems, published by the American Hotel and Motel Association. The author found that the relationship between executive total compensation and company sales was not similar, unlike the results from the studies in other industries. The results indicated that the executive total compensation in small hotel companies was equal to or more than the executive total compensation in large hotel companies. Cauvin investigated the relationship between executive compensation and company sales again in 1979, this time conducting nine simple regressions to examine the relationship between executive compensation in each of nine executive positions. The results showed that there was less relationship between executive compensation and company sales in the lodging industry than in other fields (Cauvin, 1979b). More recently, Gu and Choi (2004) researched the determinants of CEO compensation in the casino industry, using several types of financial measurements for firm performance: marketbased firm performance (annual change of stock price), accountingbased firm performance (return on assets for firm profitability, asset turnover ratio for firm efficiency, longterm debt ratio for firm debt leverage), and firm size (total assets). The results supported that profitability, firm size, debt leverage, and stock options were positively related to CEO cash compensation, while revenue efficiency (i.e., asset turnover) was negatively correlated. Kim and Gu (2005) also studied the determinants of CEO cash compensation in the restaurant industry based on the payforperformance rule using firm size, sales growth, ROI, and stock returns. The authors found that CEOs’ cash 20 compensation was positively associated with firm size and operating efficiency, while growth, debt leverage, profitability, and stock performance were not related. In addition, Namasivayam, Miao, and Zhao (2007) investigated the relationship between compensation and firm performance for 1,223 hotel companies in the U.S. using data gathered from the Hospitality Compensation and Benefit Survey of Smith Travel Research in 2001 to 2003. Unlike other studies of executive compensation in the hospitality industry or other industries, the authors adopted RevPar (Revenue per available room) as the hotels’ performance measurement. The results showed that both individual salary and benefits were significantly positively related to hotel performance for both management and nonmanagement employees. For the tourism industry, Skalpe (2007) examined the differences in CEO pay between Norway’s tourism and manufacturing industries with regard to the CEOs’ gender and age, as well as financial variables that included firm size and firm performance. The study found that there was a difference in CEO pay between genders in both industries, although the smaller companies showed a greater difference. A difference in salary by gender is particularly significant for the tourism industry because more female CEOs work in the tourism industry than in the manufacturing industry. As a result of extensive literature reviews of studies in the accounting and finance literature on executive compensation based on payforperformance rule, four major categories for measuring firm’s performance can be identified: marketbased firm performance, accountingbased firm performance, growthbased performance, and firm size. Table 21 shows a summary of the financial variables used in prior studies of executive compensation determinants. These financial variables can be utilized as the 21 basis by which select relevant variables of financial determinants for executive compensation in the current study. 22 Table 21. The classification of financial variables from the previous studies Type Variable Studies Total Asset (TA) Anderson et al. (2003); Firth et al. (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Kim & Gu (2005);Traichal et al. (1999) Firm size Sales Volume (SV) Cauvin (1979); Gray & Cannella (1997); Leone et al (2006); Skalpe (2007); Stock Return (SR) Anderson et al. (2003); Andjelkovic et al (2002); Duru & Iyengar (1999); Eichholtz et al (2008); Firth et al. (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Jensen & Murphy (1990); Kato & Kubo (2006); Leone et al (2006); Traichal et al(1999) Return on Asset (ROA) Andjelkovic et al (2002); Duru & Iyengar (1999); Grinstein & Hribar (2004); Gu & Choi (2004); Kato & Kubo (2006);Leone et al (2006); Skalpe (2007); Return on Investment (ROI) GomezMejia et al (1987); Kim & Gu (2005); Return on Sales (ROS) Gray & Cannella, Jr (1997); Return on Equity (ROE) Andjelkovic et al (2002); Firth et al. (1999); GomezMejia et al (1987); Traichal et al(1999) Earnings per Share (EPS) Duru & Iyengar (1999); Eichholtz et al (2008); GomezMejia et al (1987); Perry & Zenner (2001) Firm Profitability Sales Growth (GS) Duru & Iyengar (1999); Firth et al. (1999); GomezMejia et al (1987); Grinstein & Hribar (2004); Kato & Kubo (2006); Kim & Gu (2005) Firm Liquidity Fixed Assets Turnover (FAT) Kim & Gu (2005) Firm Activity Asset Turnover (AT) Gu & Choi (2004); Kim & Gu (2005) Debt ratio (DT) Kim & Gu (2005) Firm Coverage Long Term Debt (LTD) Gu & Choi (2004); Traichal et al(1999) 23 3. Managerial Power Determinants of Executive Compensation Recent academic research has found it difficult to explain the determinants of executive compensation using only firm performance because numerous empirical results have supported that the payforperformance rule does not always work in determining executive compensation (Conyon, 1997; GomezMejia et al., 1987). Thus, researchers have turned their sights to finding other factors that might influence executive compensation, such as demographic characteristics (e.g., age, gender, and education level), compensation structure (e.g., the composition of compensation with stock options and cash compensation), and the variables from the managerial power approach (e.g., stock ownership, board size, compensation committee size) (Coles, McWilliams, & Sen, 2001; Nelson, 2005). Among those attempts to address other determinants of executive compensation, the dominant theoretical foundation is the managerial power approach proposed by Ouch and Maguire in 1975 (GomezMejia & Wiseman, 1997). According to traditional financial theory, especially agency theory, the board of a company can control an agent (executive) with compensation packages. However, the managerial power approach suggests that the executive would not consider shareholder wealth if he or she had the power to influence the board’s decisionmaking; that is, if the executive has enough governance power to affect the board’s decision process in establishing the executive’s compensation contract, the traditional financial view based on the payforperformance rule may not be an appropriate approach to finding the determinants of executive compensation (Bebchuk & Fried, 2003; Core et al., 1999; Grinstein & Hribar, 2004; Hallock, 1998; Yermack, 1995). Thus, the managerial power 24 approach has been combined with the payforperformance rule in recent empirical studies on executive compensation. There are two main components of the managerial power approach: stock ownership structure and board independence. For the stock ownership structure’s impact on executive compensation, the CEO who possesses a higher portion of the company’s outstanding stocks could have more power in the company and be more likely to use corporate resources for his or her own benefit (Khan, Dharwadkar, & Brandes, 2005). Thus, the executive with high stock ownership would extract greater overall levels of compensation (Ozkan, 2007). In other words, the level of executive compensation would be higher when the executive has higher stock ownership (Toyne, Millar, & Dixon, 2000). In addition, higher executive possession of company’s outstanding stocks would influence the composition of board members, because the voting rights to select directors are distributed according to the amount of company stock held. Thus, an executive who owns a great deal of stock may have enough power to affect his or her own compensation level by selecting sympathetic board members (GrabkeRundell & GomezMejia, 2002). Since the board of directors decides the level of executive compensation, the independence of the board has been regarded as one of the key factors in determining executive compensation. However, it is not always easy to keep the board independent of top executives in the company. For example, outside members of the board are less likely to conflict with the CEO when the CEO appoints the outside members. Furthermore, the board of directors tends to follow the opinion of compensation consultants who are hired by the CEO (Core et al., 1999). As a result, executive compensation might not be 25 determined with the company’s best interests in mind (Core et al., 1999). Therefore, the independence of the board should be considered in an executive compensation study. Numerous studies have been conducted to examine the association between executive compensation and selected variables from the managerial power approach. GomezMejia et al. (1987) studied the effect of ownership structure on CEO compensation by classifying sample companies into two categories using the 5 percent ownership convention (referring to whether one individual or organization holds more than five percent of the company’s outstanding stock and may, therefore, be able to affect decisions): managementcontrolled companies and ownercontrolled companies. The firm’s performance and size measures were also included in the study, which found that ownership structure significantly influenced the level of CEO compensation such that CEO in externally controlled firms receive more compensation on the basis of firm performance than do CEOs in internally controlled firms. Thus, executive compensation would be more likely to follow the payforperformance rule in externally controlled firms, and executive stock ownership is a key factor in determining executive compensation. Core et al. (1999) also researched the effect of corporate governance on executive compensation with 495 CEOs in 205 publicly traded U.S. firms. The authors utilized several relevant variables from both financial performance and the managerial power approach to identify determinants of executive compensation. Among the managerial power variables were board size, composition of board membership, whether the CEO served as chairperson of the board, and the CEO’s percentage of stock ownership. The authors found that there is a significantly negative relationship between CEO 26 compensation and board and ownership structure and concluded that CEOs received greater compensation when governance structures were less effective. Likewise, Yermack (1996) found that companies with small boards provided stronger CEO performance incentives from compensation. Other studies that investigated the relationship between CEO compensation and corporate governance from managerial power approach included that of Bebchuk and Fried (2003), which found that the CEO could influence board decisions by controlling the information about the company to board members and controlling the meeting time and agenda. Several studies have examined the relationship between executive compensation and relevant variables from the managerial power approach in different countries, such as China (Firth, Fung, & Rui, 2007), the United Kingdom (Ozkan, 2007), and Israel (Cohen & Lauterbach, 2008). Firth et al. (2007) examined how ownership structure and corporate governance influenced CEOs’ compensation in Chinese companies. They adopted variables based on the theoretical concepts from both the payforperformance and managerial power approaches. Board size, proportion of nonexecutive directors on the board, and a dummy variable (whether the CEO and the chairman of the board are the same person) were used to examine the effect of managerial power on CEO compensation. The study revealed that type of ownership and board size affected CEO compensation, as independent boards or nonexecutive directors of boards were more likely to implement performancerelated pay. Using board size, the composition of nonexecutive directors on board, and CEO stock ownership, Ozkan (2007) investigated how corporate governance influenced CEO compensation in 414 U.K. companies and found that larger board size and a higher 27 proportion of nonexecutive directors on the boards resulted in higher CEO compensation; thus, less corporate governance of executives led to higher executive compensation. Finally, Cohen and Lauterbach (2008) researched CEO compensation, as it related to company ownership, with 124 publicly traded firms in Israel from 1994 to 2001. The authors included independent variables of firm performance, firm size, board composition, demographics (education level and age), and company ownership and found that CEOs in CEOowned companies received significantly higher compensation than did CEOs who did not own part of the company. In addition, the payforperformance sensitivity was lower in CEOowned companies than in nonCEO owned companies, even though the difference was not statistically significant. These results also indicated that the CEOs who owned more company stock received higher compensation than did CEOs who owned less company stock, regardless of firm performance. Thus, executive compensation was related to corporate governance in a variety of different countries. The extant literature demonstrates that the importance of the managerial power approach has increased and that combinations of variables from the payforperformance and the managerial power approaches have come to the fore in executive compensation studies. Thus, the managerial power approach should be considered for the current study in order to derive more concise and meaningful information, so both the payforperformance and the managerial power approaches shall be included in this study. Table 22 summarizes selected variables from the managerial power approach used in prior studies which will form the basis of variables of managerial power determinants in this study. 28 Table 22. The variables of managerial power approach used in previous studies Type Variables Studies Executive shares Cohen & Lauterbach (2008); Core et al. (1999); Coles et al. (2001); GomezMejia et al. (1987); Khan et al. (2005); Ozkan (2007) Ownership Ownership Composition Core et al. (1999); Coles et al. (2001); Firth et al. (2007); Khan et al. (2005); Ozkan (2006); Toyne et al. (2000) Board size (Number of Board Director) Core et al. (1999); Firth et al. (2007); Grinstein & Hribar (2004); Hallock(1997); Ozkan (2007); Yermack (1995) Board Structure Cohen & Lauterbach (2008); Core et al. (1999); Coles et al. (2001); Firth et al. (2007); Grinstein & Hribar (2004); Board Independence Executive as Director (CEO as Board of chair) Conyon (1997); Core et al. (1999); Firth et al. (2007); Grinstein & Hribar (2004); 4. OLS regression and Quantile regression Several types of multiple regression analyses have been utilized in prior empirical studies to examine the relationship between executive compensation and selected variables based on payforperformance, managerial power, and demographic characteristics. These have included multivariate logistic regression (Gray & Cannella, 1997; Jensen & Murphy, 1990; Nelson, 2005), weighted leastsquares (WLS) regression 29 (Gu & Choi, 2004; Kim & Gu, 2005), and ordinary least squares (OLS) regression (Anderson et al., 2004; Core et al., 1999; Dyl, 1988; Firth et al., 2007; Firth et al., 1999; Grinstein & Hribar, 2004; GomezMejia et al., 1987; Hebner & Kato, 1997; Kato & Kubo, 2006; Traichal et al., 1999). Clearly, most researchers have used OLS regression analysis in these efforts. OLS regression achieves the parameter estimates of the model (model fit) and illuminates the relationship between the dependent variable and independent variables through the conditional mean function (Kutner, Nachtsheim, Neter, & Li, 2005; Pedhazur, 1997). The conditional mean function uses the conditional mean response to examine the relationship between the dependent variable and independent variable(s) (Hao & Naiman, 2007). One of the crucial factors in conducting OLS regression is reducing a heteroscedasticity problem by minimizing the sums of squared residuals in the regression equation. However, OLS regression analysis has been criticized for failing to minimize the sums of squared residuals in the regression equation (Koenker, 2005) because it is difficult to follow the equal variance of variables for social phenomena in the real world (Hao & Naiman, 2007). Because of this criticism, Koenker and Basset (1978) developed a new mechanism of regression analysis, called quantile regression analysis, which uses a conditional quantile function instead of a conditional mean. Quantile regression analysis can be used to examine the relationship between the dependent variable and independent variable(s) by estimating each quantile of response variables based on the conditional quantile function (Koenker & Hallock, 2001); thus, it can achieve flexibility by estimating a change in the entire range of the dependent variable through a change in each independent variable (Hao & Naiman, 2007). As a result, quantile regression 30 analysis has gradually emerged as the way to estimate the responses of various levels of a population from each independent variable (Koenker & Machado, 1999). More specifically, the conditional mean function in the OLS regression enables to estimate the coefficient of each independent variable by taking the value of parameters that minimize the sum of squared residuals. In other words, OLS regression minimizes the sum of squared vertical distances between data points (Xi, Yi) and the fitted line 0 1 β ˆ + β ˆ = Y ˆ (Hao & Naiman, 2007; Koenker, 2005; Koenker & Basset, 1978). The model for estimating the coefficient of OLS regression is shown as follows: Min 2 0 1 (Υ  (β + β )) i i i Σ x However, the conditional quantile function enables to estimate the coefficients of independent variables that minimize the sum of absolute residuals. In other words, quantile regression minimizes the sum of absolute vertical distances between observed value to its fitted value. The model for estimating the coefficient of medianregression line is follows: Min Υ  (β + β ) i 0 1 i i Σ x The median regression line should pass through a pair of sample, by half of data should be in the above median regression line, as well as the last half of data should be in the below median regression line (Hao & Naiman, 2007). By modifying above median regression function, the estimation of coefficients for pth quantile regression is derived as shown below: Min  (β + β ) + (1 P) Υ  (β +β ) ( ) 1 ( ) 0 <β +β (p) 1 ( ) 0 β +β Σ ( ) 1 ( ) 0 ( ) 1 ( ) 0 i P P i Yi x i p i Yi x P Y x x i p p i p p Σ ≥ 31 As shown above pth quantile regression model, pth quantile regression enables to estimate the coefficients ( ) 0 β ˆ p and ( ) 1 β ˆ p by using the weighted sum of distances between fitted values from ( ) 1 ( ) 0 β ˆ + β ˆ ˆ = p p i Y and the observed value of Yi, where 0 < P < 1 (Hao & Naiman, 2007; Koenker, 2005; Koenker & Bassett, 1978). In addition, Koenker and Hallock (2001) used one example of a CEO compensation topic to illustrate the need for quantile regression analysis for executive compensation study. They derived 1999 data from the EXECUCOMP database for CEO annual compensation in 1,660 firms and showed that executive compensation tends to increase with firm size. However, the low and high levels of CEO annual compensation were more highly related to firm size than were the middle range of CEO annual compensation, indicating that different levels of CEO compensation were differently related to firm size. The authors insisted that those kinds of results would be more frequent and would create more difficulty in explaining the relationship between executive compensation and covariates with OLS regression analysis when the sample size is larger and more covariates are included in the study. Thus, they suggested that quantile regression analysis would be a more effective method than the OLS method for executive compensation studies. In the current study, quantile regression analysis also enables examination of whether different levels of executive total cash compensation are related differently to each independent variable. More specific and concise results are expected from quantile regression analysis than would be expected from OLS regression analysis. 32 5. Development of Hypotheses Two main hypotheses for this study are proposed for examining the determinants of executive cash compensation in the hospitality industry using OLS regression and quantile regression with selected variables from both the payforperformance and the managerial power approach. The two main hypotheses were tested for three classes of samples: for Ha, all hospitality industry (H1), hotel and casino industry (H2), and restaurant industry (H3); and for Hb, and all hospitality industry (H4), hotel and casino industry (H5), and restaurant industry (H6). Hypotheses A Ha: The selected variables from both the payforperformance rule and the managerial power approach are significantly correlated with executive cash compensation in the hospitality industry. Ha1: The firm’s current ratio (CR) is significantly correlated with executive cash compensation in the hospitality industry. Ha2: The firm’s asset turnover (AT) is significantly correlated with executive cash compensation in the hospitality industry. Ha3: The firm’s debttoasset ratio (DT) is significantly correlated with executive cash compensation in the hospitality industry. Ha4: Firm size (FS) is significantly correlated with executive cash compensation in the hospitality industry. 33 Ha5: The firm’s Earnings per Share (EPS) is significantly correlated with executive cash compensation in the hospitality industry. Ha6: The firm’s sales growth (GS) is significantly correlated with executive cash compensation in the hospitality industry. Ha7: The type of executive (whether the executive is a director or not: PDIR) is significantly correlated with executive cash compensation in the hospitality industry. Ha8: The board size (the number of directors on the board: NDIR) is significantly correlated with executive cash compensation in the hospitality industry. Ha9: The compensation committee size (the number of directors on the compensation committee: NCCMT) is significantly correlated with executive cash compensation in the hospitality industry. Ha10: The number of the executive’s equity shares (Dummy variable: whether the executive has more than 5% of outstanding common stocks of company: SO) is significantly correlated with the executive cash compensation in the hospitality industry. One additional set of hypotheses was proposed for the quantile regression method. Each selected variable from both the payforperformance rule and the managerial power approach was tested by different levels of executive total cash compensation, leading to the following hypothesis: 34 Hypotheses B Hb: The selected variables from both the payforperformance rule and the managerial power approach are differently correlated with different levels of executive cash compensation in the hospitality industry. Hb1: The firm’s current ratio (CR) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb2: The firm’s asset turnover (AT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb3: The firm’s debttoasset ratio (DT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb4: Firm size (FS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb5: The firm’s Earnings per Share (EPS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb6: The firm’s sales (GS) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb7: The type of executive (whether the executive is a director: PDIR) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb8: Board size (the number of directors on the board: NDIR) is differently correlated with different levels of executive cash compensation in the hospitality industry. 35 Hb9: Compensation committee size (the number of directors on the compensation committee: NCCMT) is differently correlated with different levels of executive cash compensation in the hospitality industry. Hb10: The number of the executive’s equity shares (Dummy variable: whether the executive has more than 5% of outstanding company common stocks: SO) is differently correlated with different levels of executive cash compensation in the hospitality industry. 36 CHAPTER III METHODOLOGY 1. Data Collection and Sampling Procedures The main objective of this study is to examine which elements from the two approaches, financial performance and managerial power, are linked to executive cash compensation in the hospitality industry. Sample data were gathered from secondary databases, Standard & Poor’s COMPUSTAT database and proxy statements (DEF 14A) from SEC filings. The data collection procedure for this study was divided into two main processes: gathering firms’ financial data from the COMPUSTAT database to calculate financial measurements and collecting executive compensation data and data related to the managerial power approach from the proxy statements from SEC filings. If a company’s data was not available for one of following procedures, the observation was eliminated from the sample. The sample companies were limited to the companies that were on the list of COMPUSTAT database. Among the several subsidiaries of the hospitality industry were three major sectors: hotels, casinos, and restaurants. 37 1. Financial data for the sample companies were retrieved for each of the three main sectors of the hospitality industry from the COMPUSTAT database using Standard Industrial Classification (SIC) codes. A total of 150 hospitality company samples were collected. 1) Hotel Industry The initial sample consisted of all hotel companies (SIC code 7011). After excluding companies that did not have financial data for either 2005 or 2006, 15 hotel companies remained in the sample. 2) Casino Industry The initial sample consisted of all casino companies (SIC code 7990). After excluding companies that did not have financial data for either 2005 or 2006, 49 casino companies remained in the sample. 3) Restaurant Industry The initial sample consisted of all restaurant companies (SIC code 5812). After excluding companies that did not have financial data for either 2005 or 2006, 86 restaurant companies remained in the sample. 2. The 150 hospitality companies in the sample were matched to the SEC filing list to find executive compensation data and data related to the managerial power approach. After the matching process, 83 hospitality companies remained in the sample. 1) Hotel Industry Seven hotel companies were eliminated from the sample either because they were not listed in the SEC filings or because they didn’t have proxy 38 statements (DEF 14A) for 2005 and 2006. After excluding the 7 companies, 8 hotel companies remained in the sample. 2) Casino Industry Twentyeight casino companies were eliminated from the sample either because they were not listed in the SEC filing lists or because they didn’t have proxy statements (DEF 14A) for 2005 and 2006. After excluding the 28 companies, 21 casino companies remained in the sample 3) Restaurant Industry Thirtytwo restaurant companies were eliminated from the sample either because they were not listed in the SEC filing lists or because they didn’t have proxy statements (DEF 14A) for 2005 and 2006. After excluding the 32 companies, 54 casino companies remained in the sample. 3. Executive compensation and data related to the managerial power approach were retained from the proxy statements of the 83 hospitality companies for 2005 and 2006. Initially, data for 388 executives were gathered; after filtering, 331 executives’ data remained. 1) Filtering Executives’ data from the Hotel Industry The initial executive sample included 44 executives in 8 hotel companies. Nine executives were missing compensation data for either 2005 or 2006 and were eliminated, leaving 35 executives in the sample. 2) Filtering Executives’ data from the Casino Industry 39 The initial executive sample included 104 executives in 21 hotel companies. Seventeen executives were missing compensation data for either 2005 or 2006 and were eliminated, leaving 87 executives in the sample. 3) Filtering Executives’ data from the Restaurant Industry The initial executive sample included 240 executives in 54 hotel companies. Thirtyone executives were missing compensation data from either 2005 or 2006, leaving 209 executives in the sample. 4. With a total of 331 executives’ data remaining, the data was filtered again to remove executives who had a greater than 100% change in total cash compensation from 2005 to 2006 because such an unusual change in executive total cash compensation could skew results. 1) In the Hotel Industry One executive was removed because the executive had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 34 executives in the sample. 2) In the Casino Industry One executive was removed because the executive had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 86 executives in the sample. 3) In the Restaurant Industry Fourteen executives were removed because they had a greater than 100% change in total cash compensation from 2005 to 2006, leaving 195 executives in the sample. 40 5. After transforming the actual cash compensation of executives to their natural logarithms, two outliers were detected in the restaurant sample as having a very low log value. The real dollar amounts of two executives’ annual total cash compensations (outliers) were same as $25,000. The $25,000 for each executive’s total cash compensation was too small, compared with other executives in the sample. Thus, the two outliers were deleted from the restaurant sample to achieve more efficient results. As a result of removing the two outliers from the restaurant sample, the number of executives in the restaurant sample decreased from 195 to 193, and the number of executives in the full sample decreased from 315 to 313. 6. Finally, the sample was divided into two subsamples: the hotel and casino industry made up one subsample and the restaurant industry made up the other. The hotel and casino companies were combined as one subsample because the number of executives in those industries was too small to conduct statistical analysis, especially regression analysis. Combining them made sense since hotel and casino companies are not always easily distinguished because some hotel companies also have casino facilities, and vice versa. As shown in Table 31, these procedures led to a total of 313 executives from 83 hospitality companies: 120 executives from the 29 hotel and casino companies and 193 executives from the 50 restaurant companies. 41 Table 31. Classification of study sample Category Type of Industry Number of companies Number of executives SubSample Hotel & Casino Industry 29 120 SubSample Restaurant Industry 50 193 Full Sample Hospitality Industry 79 313 2. Variable selection Based on the extensive literature review, eleven variables were selected for the study—one dependent variable, six variables from the payforperformance rule, and four variables from the managerial power approach. The dependent variable (executive total cash compensation) and the six financial variables were transformed by natural logarithm or calculated by formula to conduct the multiple regression analyses for this study. This section explains why the dependent and independent variables were selected for the purposes of this study, how the dependent variable and one independent variable (firm size) were transformed, and how the other financial measures to be utilized for this study were calculated. Selection of dependent variable Executive compensation consists of three main types of executive compensation: cashbased compensation (e.g., salary and bonus), deferred compensation (e.g., stock options), and benefitbased compensation (e.g. insurance and pensions) (Brigham & Houston, 2001). As has been the case with many prior studies, the current study used only the cashbased compensation (in this case, salary and bonus for 2006) as the 42 dependent variable (Gray & Cannelaa, Jr., 1997; Gu & Choi, 2004; Jensen & Murphy, 1990; Kim & Gu, 2005; Lippert & Porter, 1997). Other types of compensation were not included because they are difficult to translate into comparable (cash) amounts. In addition, total cash compensation was transformed by natural logarithm in order to avoid the statistical problem of heteroscedasticity that can result from the not equal variances of variables from raw data when conducting the regression analyses (Dyl, 1988; Ott & Longnecker, 2001). By adopting the base of natural logarithms for each executive’s total cash compensation, the dependent variable was transformed from the original values of executive total cash compensation to the log of executive total cash compensation. Selection of independent variables Firm performance variables Several types of financial measures for firm performance have been utilized in prior empirical executive compensation studies, primarily marketbased performance measures, accountingbased performance measures, and growthbased performance measures. The current study adopted both accountingbased and growthbased performance measures. Marketbased performance measures (e.g., stock returns) were not chosen for this study because they can be easily biased by “noise” that is not controlled by management (GomezMejia & Wiseman, 1997). Firm size was also utilized as a financial variable. The financial measures for firm performance were adopted and modified from Gu and Choi’s study (2004) and Kim and Gu’s study (2005) in the hospitality field. 43 As was mentioned in the literature review, accountingbased financial performance measures are generally divided into four categories: liquidity, activity, profitability, and coverage. Current ratio and quick ratio are common examples of liquidity ratios, which estimate a firm’s ability to pay back its shortterm debts. Current ratio (CR) was selected for this study, rather than quick ratio, because CR is most commonly used as a basic ratio for liquidity and because quick ratio excludes more liquid assets, like inventory, even though inventory is one of the most important assets in the hospitality industry (Chatfield & Dalbor, 2005). Activity ratios measure management’s effectiveness in employing its resources to the firm’s business and include mainly receivable turnover, inventory turnover, and asset turnover (Kieso et al., 2001). As in Gu and Choi (2004) and Kim and Gu (2005), asset turnover was selected for this study. Profitability ratios include return on assets, profit margin on sales, and earnings per share (EPS). Return on assets and EPS have often been used in executive compensation studies as an estimator of firm’s profitability (Duru & Iyengar, 1999; Eichholtz et al., 2008; GomezMejia et al., 1987; Perry & Zenner, 2001). EPS was selected as an estimator of firm’s profitability ratios for the current study, rather than return on assets, because EPS facilitates checking the firm’s profitability based on the amount of outstanding common stock, so EPS is an indicator of shareholder profits from the firm’s business activities in the fiscal year (Gallagher & Andrew, 1997). Finally, as has been the case in prior studies in the field, debt ratio (DT) (Kim & Gu, 2005; OrtizMolina, 2007) was selected as an estimator of the firm’s coverage ratios, which measure the firm’s ability to protect itself from its total debt (Brigham & Houston, 2001). 44 Most previous executive compensation studies have also added firm size and sales growth as financial determinants of executive compensation. Firm size has been used as a control variable in prior executive compensation studies because it is highly correlated with the level of executive compensation (GomezMejia et al., 1987). Thus, as in other studies, total assets (TA) was selected to estimate firm size for this study (Gu & Choi, 2004; Kim & Gu, 2005). Sales growth has also been viewed as an indicator of growthbased performance and was adopted from Kim and Gu’s study as a estimator of growthbased performance for the current study. Thus, a total of six financial variables from the payforperformance rule were utilized for this study. Accountingbased financial ratios were used to transform and calculate six financial variables from the payforperformance rule into independent variables. Firm size and sales growth rate were also calculated using formulas; firm size transformed the dollar amount of the firm’s total assets by natural logarithms to avoid the bias of heteroscedasticity. The following formulas were used to calculate the financial variables for this study: 1) Current Liabilities Current Assets Current Ratio (CR) = 2) Total Assets Net Sales Asset Turnover (AT) = 3) Common stock outstanding Net Income  Preferred Stock Dividends Paid Earinings per Share (EPS) = 4) Total Assets Total Debt Debt Ratio (DT) = 45 5) 2005 2006 2005 Sales Sales  Sales Sales Growth (SG) = 6) Firm size (FS) = Log(TotalAsset) Managerial power variables Several types of variables have been utilized in prior studies to investigate the effects of the stock ownership structure and board independence on executive compensation. For this study, four variables from the managerial power approach were selected: Number of board directors (NDIR), Number of compensation committee members (NCCMT), Executive as current director (PDIR), and the executive’s stock ownership (SO). Several studies have adopted NDIR and NCCMT to estimate the board’s independence (Core et al., 1999; Firth et al., 2007; Grinstein & Hribar, 2004; Hallock, 1997; Ozkan, 2007; Yermack, 1995). Real numbers for both variables were collected from the companies’ proxy statements (DEF 14A) in the SEC filing lists and recorded in the dataset. PDIR represents whether the executive is a current member of the board of directors and was a dummy variable, coded 0 if the executive was not a current board director or 1 otherwise. Finally, the executive’s stock ownership was included to examine the effect of ownership structure on executive compensation and was also a dummy variable, coded 0 if the executive has less than 5% of company’s common stocks or 1 otherwise. The classification rule for this variable was based on whether the executive held more than 5% of the company’s outstanding common stocks. Since 1960s, numerous researchers have used the cutoff point of 5% stock ownership convention in many empirical research, because 5% of stock ownership for publicly traded company has been considered as enough amounts of stocks to influence on the firm’s decision making 46 (GrabkeRundell & GomezMejia, 2002; GomezMejia et al., 1987). Thus, four variables from the managerial power approach were adopted and modified to examine the effects of corporate governance and stock ownership on executive compensation in the hospitality industry. 3. Data Analysis and Model This research is designed as a crosssectional data analysis to examine how each financial performance and managerial power variable is linked to executive cash compensation in the hospitality industry. The data analysis of this study consisted of a descriptive analysis, a reliability test, an Ordinary Least Squares (OLS) regression analysis, and a quantile regression analysis. A descriptive analysis summarized the sample’s financial characteristics (e.g., firm size, EPS, and sales) and corporate governance characteristics (e.g., number of board members, executive’s stock ownership, board characteristics). Several types of reliability tests were conducted to check the data before doing the OLS regression and quantile regression analyses. Scatter plots allowed outliers to be removed from the sample, and a histogram and normal probability plot tested the normality and linearity in order to check the assumptions of the multiple regression analysis. The OLS regression analysis and the quantile regression analysis were used to investigate the determinants of executive cash compensation, with total cash compensation as the dependent variable (Y) and all variables from both the payfor 47 performance rule and the managerial power approach as the independent variables (X). Quantile regression analysis allowed examination of whether different levels of total cash compensation are related differently to each independent variable from the payforperformance rule and the managerial power approach. To test the hypotheses proposed in literature review chapter, the multiple regression models for each industry were proposed as follows: Predicted Executive total cash compensation = β0+ β1 Current ratio(CR) it + β2 Asset turnover(AT) it + β3 Debt ratio(DT) it + β4 Firm size(FS) it + β5 Earnings per Share (EPS) it + β6 Sales growth(SG) it + β7 Executive as board directors(PDIR) it + β8 Number of board directors(NDIR) it + β9 Number of compensation committee members(NCCMT) it + β10 Executive’s stock shares (SO) it + ε it , Where, β0 = the intercept; β1,2…,10 = the beta coefficient or slope; and εit = the random error term or the residual portion; Total cash compensation it = the sum of executive’s annual cash salary and cash bonus for firm i in year t; Current ratio it = Current asset/Current liabilities for firm i in year t; Asset turnover it = Total sale (revenue)/ Average of asset for firm i in year t; Debt ratio it =Total liabilities / Total assets for firm i in year t; Firm size it = Log of the book value of total assets of firm i in year t ; Earnings per Share it = (Net income – preferred common stock dividend paid) / common stock outstanding for firm i in year t; Sales growth it = the percentage growth in sales for firm i from year t1 to year t; Executive as board director it = Whether the executive is also a member of the board for 48 firm i in year t (not current member of board = 0, current member of board = 1); Number of board directors it = Total number of board directors for firm i in year t; Number of compensation committee members it = Number of compensation committee members for firm i in year t; and Executive’s stock shares it = whether the portion of executive’s equity shares for firm i in year t is more than 5% of the firm’s outstanding common stock (less than 5% = 0, more than 5% = 1). 49 CHAPTER IV FINDINGS 1. Description of Sample Table 41 shows a frequency analysis for the characteristics of this study’s sample. Executive total cash compensation in the hospitality industry averages $559,484, range from $109,490 to $3,035,000. The average executive total cash compensation in hotel and casino companies is higher than that in restaurant companies, at $711,395 and $465,031, respectively; the median in hotel and casino companies is also larger than the median in restaurant companies. Furthermore, the mean of the percent change of executive compensation from 2005 to 2006 was negative at 8.95%, but the average percent change and the median percent change of executive total cash compensation in hotel and casino companies was more negative than was that for restaurants (14.98% and 5.20% average change, respectively; and 12.44% and 2.6% median change, respectively). 50 Table 41. Descriptive Statistics of Executive Total Cash Compensation (N=313) Sample Category Mean Std. Dev. Median Minimum Maximum Total Cash Compensation (2006) $559,484 $483,339 $388,600 $109,490 $3,035,000 All Hospitality Companies (N=313) % Change of Total Cash Compensation (2005  2006) 8.95 32.08 5.74 85.08 96.04 Total Cash Compensation (2006) $711,395 $587,548 $564,879 $109,490 $2,825,000 Hotel & Casino Companies (N=120) % Change of Total Cash Compensation (2005  2006) 14.98 35.09 12.44 85.08 50.95 Total Cash Compensation (2006) $465,031 $398,081 $339,984 $112,452 $3,035,000 Restaurant Companies (N=193) % Change of Total Cash Compensation (2005  2006) 5.20 29.53 2.60 71.27 96.04 In terms of corporate governance characteristics, the average number of board members is 8 for both the full sample (all hospitality industry) and the subsamples (Hotel & Casino industry and Restaurant industry), and the number of board members 51 ranges from 3 to 14. More than 30% of executives in the samples are board members and more than 30% of the executives in the samples also hold more than 5% of outstanding stock, which indicates that many have enough power to influence board decisions. Prior to performing OLS regression and quantile regression analysis, several tests for outliers, normality, and linearity were performed to check assumptions of the multiple regression method. The outliers were checked by developing scatter plots of samples; there were no outliers among dependent variables (Log TCC) in the full sample or the subsamples (Figures 41, 42 and 43). Figure 41. Scatter plot for full sample (All hospitality industry) 52 Figure 42. Scatter plot for subsample (Hotel & Casino industry) Figure 43. Scatter plot for subsample (Restaurant industry) 53 The normality and linearity of samples were also tested using histograms and normal probability plots of standardized residuals for dependent variables (Log of total cash compensation). As shown in Figure 44, standardized residuals of dependent variables in the full sample were normally distributed and had linearity. Figures 45 and 46 also show that nonnormality and nonlinearity were not detected in the subsamples of either the Hotel & Casino subsample or the Restaurant subsample. Thus, it was confirmed that data sets of both the full sample and the subsamples were appropriate to conduct multiple regression methods to examine the relationship between executive total cash compensation and independent variables selected for this study. 54 Figure 44. Histogram and Normal PP Plot for full sample (All hospitality industry) 55 Figure 45. Histogram and Normal PP Plot for subsample (Hotel & Casino industry) 56 Figure 46. Histogram and Normal PP Plot for subsample (Restaurant industry) 57 2. Findings of Ordinary Least Square (OLS) Regression Tables 42, 43, and 44 report the results of the OLS regression with the full sample and the two subsamples. Multicollinearity for the three multiple regression models had to be checked since high correlations among the variables would cause deviation or and/or misleading results in the multiple regression statistics by changing input variable in the regression model as variables were added in or deleted from the model (Pedhazur, 1997). The variance inflation factor (VIF) was used to check the impact of multicollinearity between each independent variable in the regression models. The higher the VIF number, the greater the impact of collinearity on the accuracy of the model (Ott & Longneker, 2001). The VIF values shown in Table 42, for the full sample, lie in the range between 1.091 and 3.361. This does not indicate a serious multicollinearity problem because the VIF is well below the problematic level of 10 (Ott & Longneker, 2001). The range of VIF values for the Hotel & Casino subsample are between 1.088 and 5.141 (Table 43), and Table 44 shows that the VIF values of the Restaurant subsample are between 1.275 and 2.906. Thus, there are no serious multicollinearity problems for the subsamples either. After testing multicollinearity using VIF values, OLS regression analyses were conducted to investigate the relationship between the dependent variable and 10 independent variables to examine the three main hypotheses (H1, H2, and H3). The dependent variable for the OLS regression models is the log of executive total cash compensation and the ten independent variables consisted of six financial variables and 58 four managerial power variables. The results of the OLS regression analyses are presented in Tables 42, 43, and 44. Results of OLS regression method for the Full sample Table 42 summarizes the results of the OLS regression for the full sample with six financial variables and four managerial power variables. Both the Rsquare (=0.646) and the adjusted Rsquare (=0.634) for this model were the appropriate level of goodness of fit for empirical study in social science fields. The pvalues of three of the financial variables (DT, FS, EPS) were less than 0.01 with positive coefficients, and the pvalues of the other three financial variables (CR, AT, GS) were larger than 0.05, so only DT, FS, and EPS were positively related to the dependent variable at a statistically significant level of 0.01. The pvalues of both PDIR and SO were less than 0.01, and PDIR and SO were positively associated with executive total cash compensation at a pvalue of 0.01. 59 Table 42. OLS regression summary for the Full sample (all hospitality companies) Variable T Value Significance Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.672 56.480 0.000 CR 0.021 1.239 0.216 0.713 1.403 AT 0.022 1.125 0.262 0.505 1.982 DT 0.115 3.419 0.001*** 0.788 1.269 FS 0.245 10.186 0.000*** 0.298 3.361 EPS 0.038 4.170 0.000*** 0.615 1.626 GS 0.079 1.739 0.083* 0.917 1.091 PDIR 0.182 6.749 0.000*** 0.658 1.520 NDIR 0.010 1.583 0.114 0.514 1.946 NCCMT 0.004 0.494 0.622 0.844 1.185 SO 0.092 3.196 0.002*** 0.578 1.730 N RSquare Adjusted R 313 0.646 0.634 Note: * P< 0.10, ** P< 0.05, ***P<0.01 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.672 + 0.115 Debt to asset ratio(DT) + 0.245 Firm size(FS) + 0.038 Earnings per share(EPS) + 0.182 Type of board directors(PDIR) + 0.092 Executive’s stock shares(SO). Thus, hypotheses H13, H14, H15, H17, and H110 were accepted at the 0.01 level, but hypotheses H11, H12, H16, H18, and H19 were not. 60 Results of OLS regression method for the Hotel & Casino subsample Table 43 summarizes the results of the OLS regression method for the Hotel & Casino subsample with six financial variables and four managerial power variables. Both the Rsquare (=0.708) and the adjusted Rsquare (=0.682) for this model were the appropriate level of goodness of fit. Like the OLS regression for the full sample, six financial variables were used for the OLS regression for this subsample with the result that the pvalues for four variables (DT, FS, EPS, and GS) were less than 0.01, and the pvalues of CR and AT were larger than 0.05. Thus, DT, FS, EPS, and GS were positively associated with the dependent variable with statistical significance at the 0.01 level. The pvalues of only two managerial power variables, PDIR and SO, were less than 0.05, so PDIR and SO were positively related with the executive total cash compensation at a pvalue of 0.05. Table 43. OLS regression summary for the Hotel & Casino subsample Variable T Value Significance Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.507 28.246 0.000 CR 0.051 1.373 0.173 0.532 1.881 AT 0.034 0.954 0.342 0.819 1.222 DT 0.176 2.797 0.006*** 0.646 1.547 FS 0.307 5.780 0.000*** 0.195 5.141 EPS 0.041 3.250 0.002*** 0.479 2.088 GS 0.141 2.662 0.009*** 0.919 1.088 PDIR 0.207 4.813 0.000*** 0.731 1.369 NDIR 0.002 0.153 0.878 0.388 2.576 NCCMT 0.008 0.413 0.680 0.630 1.587 SO 0.106 2.264 0.026** 0.638 1.568 N RSquare Adjusted R 120 0.708 0.682 Note: * P< 0.10, ** P< 0.05, *** P<0.01 61 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.507 + 0.176 Debt to asset ratio(DT) + 0.307 Firm size(FS) + 0.041 Earnings per share(EPS) + 0.141 Sales growth(GS) + 0.207 Type of board directors(PDIR) + 0.106 Executive’s stock shares(SO). Thus, hypotheses H23, H24, H25, H26, H27, and H210 were accepted at 0.05 level, while hypotheses H21, H22, H28, and H29 were not. Results of OLS regression for the Restaurant subsample Table 44 summarizes the results of the OLS regression for the Restaurant subsample. Both the Rsquare (=0.591) and the adjusted Rsquare (=0.598) for this model had the appropriate level of goodness of fit, even though both were less than those for the full sample or the other subsample. Contrary to the results of the OLS regression analyses for the full sample and the Hotel & Casino subsample, the results of the OLS regression for the Restaurant subsample had only two variables (DT and FS) in the financial variables with pvalues less than 0.05 and positive coefficients, indicating that DT and FS were positively related to the dependent variable at a statistically significant level of 0.05 and 0.01, respectively. The results of the Restaurant subsample were similar to those of the full sample and the Hotel & Casino subsample in terms of the managerial power variables, as the pvalues of 62 both PDIR and SO were less than 0.05. Thus, PDIR and SO were positively associated with executive total cash compensation at a pvalue at the 0.05 level. Table 44. OLS regression summary for the Restaurant subsample Variable T Value Significance. Collinearity Statistics Regression Coefficients Tolerance VIF Intercept 4.715 42.041 0.000 CR 0.028 1.433 0.154 0.715 1.398 AT 0.023 0.850 0.396 0.649 1.542 DT 0.093 2.058 0.041** 0.784 1.275 FS 0.255 8.279 0.000*** 0.344 2.906 EPS 0.011 0.594 0.553 0.536 1.865 GS 0.127 1.183 0.238 0.781 1.280 PDIR 0.161 4.676 0.000*** 0.594 1.683 NDIR 0.012 1.421 0.157 0.471 2.124 NCCMT 0.010 0.927 0.355 0.764 1.310 SO 0.090 2.490 0.014** 0.526 1.900 N RSquare Adjusted R 193 0.591 0.568 Note: * P< .10, ** P< .05, P<0.01 After the OLS regression analysis, the following model was accepted with statistical significance: Predicted Executive total cash compensation = 4.715 + 0.093 Debt ratio(DT) + 0.255 Firm size (FS) + 0.161 Type of board directors(PDIR) + 0.090 Executive’s stock shares(SO). Thus, only hypotheses H33, H34, H37, and H310 were accepted at 0.05 level, but hypotheses H31, H32, H35, H36, H38, and H39 were not accepted. 63 3. Findings of the Quantile Regression After conducting the OLS regression analyses, the quantile regression analyses were conducted to test the three proposed hypotheses under the second main hypotheses (H3, H4, and H5) to determine whether the selected independent variables are differently related to different levels of executive compensation. The variables for the quantile regression were the same as the variables in the OLS regression analyses. There are two usual ways of interpreting the results of quantile regression: checking the statistical significance of the coefficients of each independent variable toward dependent variable, and checking the pattern of coefficients of each independent variable toward each quantile of the dependent variables. Tables 45, 46, and 47 show the results of the quantile regression analysis for the coefficient estimates of the model for the full sample and the two subsamples, and Figures 47, 48, and 49 show the pattern of the coefficients of each independent variable from the payforperformance rule and the managerial power approaches toward each quantile of dependent variable (the level of executive total cash compensation) for the three samples. The Xaxis for each graph shows the quantile of executive total cash compensation, and the Yaxis shows the coefficients of the independent variables from both the payforperformance rule and the managerial power approach. Red lines show the coefficients for the independent variables from the OLS regression analysis, and the black line represents the coefficients of the independent variable from the quantile regression analysis. The black shadow areas show the 95% confidence interval of coefficients for the independent variables from the results of the quantile regression analysis. 64 Results of the quantile regression method for the Full sample Table 45 shows the results of the quantile regression analysis for the coefficient estimates of the model with the full sample of the hospitality industry. Generally speaking, it looks similar to the OLS regression results for the full sample, even though the quantile regression provides more specific results than the OLS regression does. For example, for the financial variables, neither CR nor AT were correlated with executive total cash compensation at an alpha level of 0.05 in the OLS regression, but the quantile regression showed that both CR and AT were significantly related to executive total cash compensation at the 0.05 level for the low quantiles of compensation, the 0.1 0.2 and the 0.10.3 quantiles, respectively. Thus, executives who received lower cash compensation were influenced by CR and AT, while others were not. In addition, the coefficient graphs for both CR and AT (Figure 47) show that the coefficient values for CR and AT decreased as the level of executive compensation increased, indicating that executives at a lower level of compensation were more sensitive to both CR and AT than were the executives in the middle and upper level of compensation. For the DT and FS variables, the quantile regression analysis provided results similar to those of the OLS regression analysis (i.e., both DT and FS were significantly related to executive compensation in the full sample with statistical significance at an alpha level of 0.05). However, the coefficient graphs for DT show moderate volatility of coefficients from the lower quantile to the upper quantile of executive compensation. This indicates that DT was not differently related to the level of executive compensation with statistical significance. In contrast to the DT graph, the pattern of coefficients of the FS variable decreased from the lower quantile of executive compensation to the middle quantile, then 65 increased as it approached the upper quantile. Thus, the low and high levels of executive compensation were more related to firm size than was the middle range of executive compensation. The results of the quantile regression also show that EPS was not significantly correlated with executive total cash compensation for executives in the lower level of compensation, the 0.10.3 quantile, at an alpha level of 0.05, even though EPS was significantly correlated with total cash compensation in the OLS regression results. This suggests that EPS affects only the executives in the mid and upper levels of total cash compensation. The coefficient graph for the EPS variable in Figures 47 shows an increasing pattern for the coefficient value of EPS from the lower to the upper quantiles of executive compensation, so executives with lower compensation were less sensitive to EPS than were executives in the middle and upper levels of compensation. The result from the quantile regression also shows that GS was significantly correlated with executive total cash compensation for executives in the mid and upper levels of total cash compensation (0.50.8 quantile) at an alpha level of 0.05, even though GS was not significantly related with executive’s total cash compensation in the OLS regression results. The coefficient graph for the GS variable (Figures 47) shows the coefficient value for GS increasing as the level of executive compensation increases, so executives with lower compensation were less sensitive to GS than were executives in the middle and upper levels of compensation. The result of the quantile regression analysis of the four managerial power variables was not much different from that of the OLS regression. Both PDIR and SO were significantly correlated with executive total cash compensation at the 0.05 level, and 66 NDIR and NCCMT were not. Figure 47shows that the only pattern of the SO coefficient of SO was an increasing pattern from the low quantile to the high quantile of executive total cash compensation. By contrast, the pattern of PDIR coefficient had moderate variation. Thus, the effect of SO on executive compensation increased as executive compensation increased. From the results of the quantile regression analysis, hypotheses: H41, H42, H4 4, H45, H46 and H410 were accepted at 0.05 level, while hypotheses: H43, H47, H4 8, and H49 were not accepted. 67 Table 45. Quantile regression summary for the Full sample (all hospitality industry) Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 4.328 4.481 4.580 4.687 4.711 4.858 4.789 (Intercept) 4.787 4.813 Coefficient T value 40.872*** 44.184*** 42.525*** 40.364*** 38.154*** 38.339*** 33.963*** 28.451*** 25.382*** 0.066 0.044 0.026 0.015 0.012 0.009 0.004 0.002 0.016 CR Coefficient T value 3.316*** 2.317** 1.333 0.686 0.529 0.396 0.169 0.071 0.417 0.063 0.053 0.050 0.033 0.040 0.000 0.001 0.017 0.006 AT Coefficient T value 2.495** 2.121** 1.964** 1.248 1.449 0.014 0.020 0.520 0.159 0.159 0.099 0.094 0.092 0.085 0.117 0.100 0.080 0.070 DT Coefficient T value 4.039*** 2.776*** 2.607*** 2.426** 2.249** 3.238*** 2.583*** 2.085** 1.538 0.260 0.276 0.251 0.232 0.231 0.192 0.233 0.274 0.277 FS Coefficient T value 7.527*** 8.569*** 7.489*** 6.759*** 6.518*** 5.696*** 6.164*** 6.197*** 5.491*** 0.021 0.017 0.018 0.036 0.038 0.065 0.053 0.054 0.047 EPS Coefficient T value 1.409 1.297 1.311 2.324** 2.349** 4.268*** 3.571*** 3.667*** 2.937*** 0.010 0.012 0.065 0.137 0.168 0.183 0.170 0.140 0.071 GS Coefficient T value 0.137 0.169 0.876 1.755* 2.119** 2.561** 2.356** 1.943** 1.059 0.210 0.175 0.192 0.165 0.180 0.188 0.166 0.177 0.213 PDIR Coefficient T value 5.090*** 4.901*** 5.139*** 4.207*** 4.497*** 4.922*** 4.298*** 4.348*** 4.490*** 0.002 0.001 0.005 0.006 0.006 0.016 0.019 0.002 0.014 NDIR Coefficient T value 0.172 0.163 0.615 0.659 0.635 1.758* 2.111** 0.246 1.185 0.019 0.006 0.001 0.002 0.001 0.012 0.018 0.001 0.011 NCCMT Coefficient T value 1.770* 0.522 0.082 0.195 0.075 1.038 1.494 0.120 0.864 0.017 0.105 0.083 0.090 0.066 0.078 0.106 0.140 0.142 SO Coefficient T value 0.351 2.870*** 2.182** 2.276** 1.631 1.981** 2.610*** 3.309*** 3.085*** Sample Size N=313 N=313 N=313 N=313 N=313 N=313 N=313 N=313 N=313 Note: * P< .10, ** P< .05, P<0.01 68 Figure 47. The coefficient graphs of all hospitality industry by Quantile regression 69 Results of quantile regression method for Hotel & Casino subsample Table 46 shows the result of the quantile regression analysis for the coefficient estimates of the model with the Hotel & Casino subsample. Generally speaking, the results of the quantile regression analysis were similar to those of the OLS regression results for the subsample, but the quantile regression provides more specific results than the OLS regression. For example, neither CR nor AT in the quantile regression results were correlated with executive total cash compensation in the hotel and restaurant industry at an alpha level of 0.05, which is the same as the results from the OLS regression. However, the patterns of the coefficients of both the CR and AT variables (Figure 48) provided meaningful results, even though the CR and AT were not significantly related with the level of executive compensation. The patterns of the coefficient value for both CR and AT decreased as the level of executive compensation increased, indicating that the executives at lower levels of compensation were more sensitive toward both CR and AT than were the executives at middle and upper levels of compensation. In addition, DT was not significantly related to all quantile of executive compensation in the result from the quantile regression, while the OLS regression showed that DT is significantly related to executive compensation. The quantile regression provided, however, that DT was significantly related to compensation for the low quantile, 0.10.2 quantile, so only those executives at the low level of total cash compensation were influenced by DT. In addition, the coefficient graphs for the DT variable (Figure 48) show that the pattern of the coefficient value for DT decreased slightly as the level of executive compensation increased, suggesting that executives at 70 lower levels of compensation were slightly more sensitive to DT than were the executives in the middle and upper levels of compensation. The quantile regression analysis provided similar results for the FS variable as the OLS regression results for this subsample that FS was statistically significantly related to executive compensation at an alpha level of 0.05. However, the pattern of coefficients of the FS variable decreased from the lower quantile of executive compensation to the middle quantile, then increased to the upper quantile. This suggests that the low and high levels of executive compensation were more related to firm size than was the middle range of executive compensation. The results of the quantile regression show that EPS was not significantly correlated with executive total cash compensation for the low quantile (0.1 – 03 quantile) of executive compensation, even though EPS was significantly correlated with executive total cash compensation in the OLS regression results. It implies that EPS significantly affects only the executive in mid and upper level of total cash compensation. Furthermore, the coefficient graphs for the EPS variable (Figure 48) show that the coefficient value for EPS increased from the lower quantile of executive compensation to the upper quantile, so executives at the lower level of compensation were less sensitive to EPS than were executives in middle and upper levels of compensation. The quantile regression also shows that GS was significantly correlated with executive total cash compensation, but only for the upper level of compensation (0.8 0.9 quantile), even though GS was significantly related to compensation in the OLS regression results. Thus, GS affected only the executives at the upper level of total cash compensation. In addition, the coefficient graphs for the GS variable (Figure 48) show that the coefficient value for 71 GS increased as the level of executive compensation increased, which also supports the conclusion that those at the lower level of compensation were less sensitive toward GS than were those at the middle and upper levels. The result of the quantile regression analysis was not much different for the four managerial power variables than the results of the OLS regression. PDIR was significantly correlated with executive total cash compensation at the 0.05 level, whereas NDIR and NCCMT were not. Most notable were the results from the quantile regression for the SO variable, which showed that SO was significantly correlated with executive total cash compensation for executives only at the upper level of total cash compensation (0.70.8 quantile) at an alpha level of 0.05, even though SO was significantly related to executive total cash compensation in the OLS regression results. Of the two statistically significant variables (PDIR and SO) shown in Figure 48, only SO had a pattern of coefficients that increased from the low quantile to the high quantile of executive total cash compensation; the pattern of coefficients for PDIR had moderate volatility. It indicates that executives at the lower level of compensation were less sensitive to SO than were executives in middle and upper levels of compensation. After the quantile regression analysis, only hypotheses H53, H54, H55, H56, and H510 were accepted at 0.05 level, while hypotheses H51, H52, H57, H58, and H59 were not. 72 Table 46. Quantile regression summary for the Hotel & Casino subsample Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 Coefficient 4.060 4.100 4.296 4.616 4.776 4.794 4.803 4.772 4.957 (Intercept) T value 12.676*** 12.883*** 15.616*** 18.646*** 17.778*** 19.584*** 20.668*** 18.487*** 17.351*** Coefficient 0.117 0.125 0.070 0.013 0.012 0.014 0.029 0.031 0.024 CR T value 1.651 1.692* 1.077 0.208 0.188 0.254 0.571 0.588 0.417 Coefficient 0.043 0.053 0.014 0.055 0.055 0.067 0.051 0.025 0.059 AT T value 0.560 0.924 0.258 0.988 0.985 1.382 1.069 0.488 1.137 Coefficient 0.211 0.210 0.140 0.115 0.101 0.105 0.048 0.061 0.315 DT T value 2.264** 2.119** 1.538 1.219 1.030 1.236 0.618 0.726 1.351 Coefficient 0.384 0.326 0.294 0.242 0.259 0.280 0.306 0.335 0.288 FS T value 3.148*** 3.092*** 3.308*** 3.027*** 3.337*** 4.240*** 4.845*** 5.164*** 4.280*** Coefficient 0.000 0.013 0.023 0.045 0.057 0.073 0.066 0.061 0.054 EPS T value 0.016 0.598 1.173 2.154** 2.569*** 3.728*** 3.575*** 3.221*** 2.679*** Coefficient 0.004 0.074 0.154 0.105 0.169 0.162 0.138 0.227 0.279 GS T value 0.038 0.627 1.378 0.986 1.617 1.666* 1.524 2.236** 2.651*** Coefficient 0.200 0.223 0.180 0.184 0.206 0.177 0.179 0.182 0.210 PDIR T value 2.782*** 3.085*** 2.522** 2.454*** 2.733*** 2.773*** 2.827*** 2.766*** 3.599*** Coefficient 0.013 0.008 0.014 0.015 0.004 0.002 0.005 0.006 0.007 NDIR T value 0.596 0.359 0.662 0.746 0.218 0.143 0.329 0.385 0.376 Coefficient 0.057 0.009 0.002 0.002 0.020 0.028 0.036 0.027 0.043 NCCMT T value 1.159 0.276 0.058 0.092 0.721 1.219 1.716* 1.022 1.251 Coefficient 0.023 0.038 0.070 0.123 0.120 0.120 0.130 0.168 0.061 SO T value 0.254 0.444 0.852 1.544 1.566 1.900* 2.045** 2.461** 1.001 Sample Size N=120 N=120 N=120 N=120 N=120 N=120 N=120 N=120 N=120 Note: * P< .10, ** P< .05, P<0.01 73 Figure 48. The coefficient graphs of the Hotel & Casino subsample by quantile regression 74 Results of quantile regression for the Restaurant subsample Table 47 shows the results of the quantile regression analysis for the coefficient estimates of the model with the Restaurant subsample. The quantile regression provided more specific results than the OLS regression for restaurant subsample, even though the result of the quantile regression analysis for the restaurant subsample was similar to that of the OLS regression results in this subsample. In the quantile regression, unlike the results from the OLS regression, both CR and AT were correlated with executive total cash compensation at the lower quantile (0.1  0.3 quantile) of executive compensation at the alpha level of 0.05. In addition, the coefficient graphs for both CR and AT (Figure 49) show that their coefficient values decreased as the level of executive compensation increased. This suggests that executives at lower levels of compensation were more sensitive to both CR and AT than were executives at the middle and upper levels of compensation. The result of the quantile regression also showed that DT was not significantly related to executive compensation, while the result from the OLS regression showed the opposite. In addition, the coefficient graphs for the DT variable (Figure 49) show that the pattern of the coefficient value for DT was one of moderate volatility from the lower quantile to the upper quantile of executive compensation. It indicates that there is no different impact of DT on different level of executive compensation in Restaurant industry. The quantile regression analysis provided similar results for the FS variable as that of the OLS regression that FS was statistically significantly related to executive compensation at an alpha level of 0.05. However, the pattern of coefficients of the FS 75 variable decreased from the lower quantile of executive compensation to the middle quantile, then increased in the upper quantile. Thus, the low and high levels of executive compensation were more related with firm size than was the middle range. The results of the quantile regression also showed that EPS was not significantly correlated with executive total cash compensation for any quantile of executive compensation at an alpha level of 0.05, which was the same as the result from the OLS regression. However, the coefficient graphs for the EPS variable in Figure 49 show that the coefficient value for EPS increased from the lower quantile of executive compensation to the upper quantile, indicating that the executives at the lower level of compensation were less sensitive to EPS than were the executives at the middle and upper levels. Like the OLS regression result, the quantile regression also shows that the GS variable was not significantly correlated with executive total cash compensation for any level of executive total cash compensation at an alpha level of 0.05. In addition, the coefficient graphs for the GS variable (Figure 49) show moderate volatility of the coefficient from the lower quantile to the upper quantile of executive compensation. Thus, GS was not related to the level of executive compensation. The results of the quantile regression analysis for the four managerial power variables were not much different from those of the OLS regression. PDIR was significantly correlated with executive total cash compensation at the 0.05 level, while NDIR and NCCMT were not. However, the quantile regression for SO shows that SO was significantly correlated with executive total cash compensation only for executives at the upper level of compensation (0.70.9 quantile) at an 0.05 alpha level, while the OLS 76 regression shows that it was significantly related to executive total cash compensation in general. The coefficient pattern of SO (Figure 49) increased from the low quantile to the high quantile of compensation, suggesting that the effect of SO on executive compensation increase when the level of compensation increases. However, the pattern of coefficients for PDIR had moderate volatility, suggesting that there is no different effect of PDIR on different level of executive compensation. Furthermore, the pattern of the NDIR coefficient increased from the lower quantile to the upper quantile, which indicates that the executives at a lower level of compensation were less sensitive toward NDIR than were the executives at the middle and upper levels of compensation. However, the pattern of the NCCMT coefficient decreased from the lower quantile to the upper quantile, indicating that the executives at a lower level of compensation were more sensitive to NCCMT than were the executives at the middle and upper levels of compensation. After the quantile regression analysis, hypotheses H61, H62, H64, and H67 were accepted at 0.05 level, but H63, H65, H6 6, H68, H69, and H610 were not accepted at the 0.05 level. 77 Table 47. Quantile regression summary for the Restaurant subsample Quantile Regression(%) Variables 10 20 30 40 50 60 70 80 90 Coefficient 4.465 4.504 4.488 4.614 4.662 4.846 4.841 4.884 4.999 (Intercept) T value 26.656*** 25.263*** 23.571*** 22.672*** 22.499*** 21.782*** 21.022*** 23.965*** 36.805*** Coefficient 0.061 0.050 0.053 0.025 0.017 0.017 0.024 0.009 0.029 CR T value 2.662*** 2.012** 2.004** 0.901 0.585 0.611 0.874 0.307 1.017 Coefficient 0.097 0.093 0.097 0.060 0.053 0.001 0.001 0.008 0.072 AT T value 2.262** 2.109** 2.106** 1.299 1.155 0.015 0.033 0.199 2.282*** Coefficient 0.030 0.037 0.059 0.058 0.066 0.070 0.053 0.077 0.094 DT T value 0.594 0.678 1.084 1.069 1.230 1.410 1.116 1.548 1.921 Coefficient 0.275 0.280 0.289 0.272 0.273 0.234 0.255 0.248 0.203 FS T value 5.345*** 5.336*** 5.107*** 4.744*** 4.744*** 4.080*** 4.364*** 4.720*** 5.258** Coefficient 0.016 0.017 0.007 0.011 0.016 0.016 0.019 0.043 0.058 EPS T value 0.605 0.578 0.242 0.368 0.522 0.527 0.676 1.536 2.295** Coefficient 0.108 0.124 0.040 0.037 0.046 0.172 0.248 0.306 0.220 GS T value 1.023 1.032 0.315 0.262 0.301 1.000 1.508 1.819* 1.133 Coefficient 0.127 0.160 0.167 0.159 0.190 0.181 0.153 0.220 0.162 PDIR T value 2.827*** 3.510*** 3.441*** 3.080*** 3.731*** 3.371*** 2.888*** 4.101*** 3.052*** Coefficient 0.004 0.008 0.002 0.005 0.003 0.012 0.012 0.015 0.042 NDIR T value 0.292 0.689 0.170 0.356 0.232 0.882 0.905 1.087 3.033*** Coefficient 0.002 0.006 0.005 0.007 0.007 0.009 0.012 0.017 0.020 NCCMT T value 0.144 0.383 0.265 0.420 0.402 0.586 0.766 1.197 1.556 Coefficient 0.055 0.089 0.077 0.080 0.048 0.067 0.098 0.113 0.158 SO T value 1.169 1.916* 1.568 1.541 0.941 1.335 1.983** 2.221** 2.922** Sample Size N=193 N=193 N=193 N=193 N=193 N=193 N=193 N=193 N=193 Note: * P< .10, ** P< .05, P<0.01 78 Figure 49. The coefficient graphs of the Restaurant subsample by Quantile regression 79 CHAPTER V CONCLUSION 1. Summary of the study This study has investigated the determinants of executive compensation in the hospitality industry with selected variables from both the payforperformance rule and the managerial power approach, using two multiple regression analysis methods: OLS regression and quantile regression. The study provides an empirical illustration of determinants of executive compensation in the hospitality industry as a whole, and in two subcategories, the Hotel & Casino category and the Restaurant category. OLS regression analysis was performed first to identify the determinants of executive compensation in the hospitality industry on the basis of both the payforperformance rule and the managerial power approach. At the second stage of analysis, quantile regression analysis was adopted to examine whether the independent variables were differently related to different levels of executive compensation in the hospitality industry. A summary and discussion of the empirical findings of this study are presented in the following sections. 80 Summary of the Full Sample: The Hospitality Industry OLS regression analysis The results of OLS regression analysis for the hospitality industry revealed that three financial variables, DT, FS, and EPS, and two managerial power variables, PDIR and SO, were positively related to executive compensation with statistical significance. Thus, payforperformance rules and managerial power variables both influenced executive compensation in the hospitality industry. While the financial variables suggested that firm size (FS) and firm profitability (EPS) positively affected executive compensation, the study found a different result from prior studies of executive compensation in the hospitality industry by showing that coverage ratio was positively related to executive compensation and that CR, AT, and GS were not. The results of the coverage ratio analysis revealed that executives in the hospitality industry were paid highly despite increasing risk to the firm. In general, a company with a higher debt ratio has a riskier financial status because high debt means a heavy interest burden and the need to repay principal (Chatfield & Dalbor, 2005) However, Jensen and Meckling (1976) suggested that agents (executives) in companies with high debt capital structures receive more compensation because of the incentive effects associated with debt: The agent is paid more for being willing to take on the challenge of activities which offer the possibility of very high payoffs, even when they have a very low probability of success. Such activities invoke an agency problem because the shareholders prefer that the company does not remain in a risky environment, but executives may prefer to invest in more risky projects in order to receive higher 81 compensation from big successes with risky projects. This result implies that there may be an agency problem in the hospitality industry. Three other financial variables—CR, AT, and GS—were not significantly related to executive compensation, so the hospitality industry only partially follows the payforperformance rule in determining executive compensation. Regarding the managerial power approach, the PDIR and SO variables were identified as determinants of executive compensation in the hospitality industry, while NDIR and NCCMT were not. These results support the idea that executives who serve on the board of directors receive more compensation than those who do not, regardless of the number of board members or the number of compensation committee members. Stock ownership by the executive also positively influenced executive compensation, so, in addition to payforperformance rules, how the executive was related to corporate governance influenced executive compensation. Quantile regression analysis The results of the quantile regression analysis provided more specific and concise results for the hospitality industry by examining the effects of each independent variable on different levels of executive compensation. The quantile regression analysis showed that the level of executive compensation was differently related to each independent payfor performance and managerial power variable. The variables were related to three different group of compensation—the lower level (0.10.3 quantile), the middle level (0.40.6 quantile), and the upper level (0.70.9 quantile). 82 Among the financial variables, the lower level of executive compensation was significantly related to CR and AT, while the middle and upper levels were significantly related to EPS and GS. That is, firm liquidity and efficiency were determinants of lower level compensation, while profitability in terms of EPS and GS determined middle and upper levels of executive compensation. In addition, the FS variable was significantly related to all levels of executive compensation, although the sensitivity of the lower and upper levels was greater than that of the middle level. Meanwhile, the coverage ratio (DT) was significantly related to the full range of executive compensation, although it was moderate. Thus, firm coverage was not differently related depending on the level of compensation. Among the four managerial power variables, only the stock ownership variable was differently related to levels of executive compensation in that the upper level of executive compensation was more sensitive than either the lower or middle levels. PDIR was significantly but moderately related to all levels of compensation, so it was not differently related to different levels of compensation. The results of the quantile regression analysis of the financial and managerial power variables also suggested that executive stock ownership and board independence (from the managerial power approach) influenced executive compensation in the hospitality industry, and that the hospitality industry also partially follows the payforperformance rule to determine executive compensation. It concluded that the hospitality industry weakly follows payforperformance rule to determine executive compensation and higher executive’s stock ownership and board nonboard independent from top executive may also influence on determining the level of executive compensation in 83 hospitality industry. In addition, it supports that the determinants of executive compensation differ between different groups of executive compensation level in the hospitality industry. Summary of Hotel & Casino Subsample OLS regression analysis The results of the OLS regression analysis provided a result similar to that of the full sample (the positive correlation of FS, EPS, DT, PDIR, and SO) for both financial variables and managerial power variables, except that GS was significantly related to compensation in the hotel and casino industry. Four financial variables (FS, EPS, GS and DT) were positively related to executive compensation, and the positive relationship between coverage ratio and executive compensation suggested that there may be a serious agency problem in the hotel and casino industry. In short, the results showed that the hotel and casino industry partially follows the payforperformance rule in determining executive compensation. The results of the managerial power approach were the same as that for the full sample in that the PDIR and SO variables were identified as determinants of executive compensation. Thus, the involvement of the executive in corporate governance influences 84 Quantile regression analysis The results of the quantile regression analysis were similar to those for the hospitality industry as a whole, supporting the idea that different levels of executive compensation are differently related to the independent variables. Among the financial variables for the payforperformance rule, the lower level of executive compensation was significantly related to only FS and DT. The FS variable was significantly related to all levels of executive compensation, although the sensitivity of FS on the lower and upper levels of compensation was more than it was for the middle level. However, coverage ratio (DT) was significantly related only to the lower level of executive compensation with a positive coefficient, but the sensitivity of DT on the level of executive compensation was moderate. Thus, firm coverage was not differently related to different levels of executive compensation. In addition, the coefficient graph of CR and AT showed that the lower level of compensation was more sensitive than were the middle and upper levels, which had no statistical significance. A cautious interpretation of this finding is that the firm’s liquidity and efficiency could have more influence on the lower level of executive compensation than on the middle and upper levels. In contrast to the lower level of compensation, the middle and upper levels were significantly related to EPS, and firm profitability was a determinant of the middle and upper levels of compensation. In addition, sales growth had different affects on the different levels of executive compensation, based on the coefficient graph of GS, which showed that the lower level of executive compensation was less sensitive to GS than were the middle and upper levels. 85 Among the four managerial power variables, the result for the Hotel & Casino subsample was the same as that of the hospitality industry as a whole, except that the stock ownership variable was differently related with levels of executive compensation, while the other three variables—PDIR, NDIR, and NCCMT—were not. The results of the quantile regression analysis of the financial and managerial power variables also suggested that executive stock ownership and board independence influenced executive compensation in the Hotel & Casino industry, and that the industry also partially follows the payforperformance rule to determine executive compensation. While the influence of the payforperformance rule is weak, higher executive’s stock owner 



A 

B 

C 

D 

E 

F 

I 

J 

K 

L 

O 

P 

R 

S 

T 

U 

V 

W 


