The factors affecting the differences in CEO salary from 1989 to 1990

In this paper, given the data set “ceosal1”, we will investigate the three factors, which are sales, return on equity roe and return on firm’s stock ros that can be used to explain the d

Trang 1

HANOI UNIVERSITY FACULTY OF BUSINESS ADMINISTRATION & TOURISMS

-oOo -ECONOMETRICS PROJECT The factors affecting the differences in CEO salary from 1989 to 1990

Trang 2

TABLE OF CONTENT

TABLE OF APPENDICES ii

LIST OF TABLES AND FIGURES iii

I Introduction 1

II Literature Review 1

1 Theoretical Foundation 1

1.1 Agency Problem 1

1.2 Executive remuneration 2

2 Empirical Findings: 2

III Methodology 3

IV Data Analysis and Results 4

1 Model testing 4

1.1 Testing on functional form 4

1.2 Testing the overall significance on all coefficients 6

1.3 Testing the model when dropping one variable 7

1.4 Testing on dummy variables 7

2 Checking the three errors 9

2.1 Multicollinearity 9

2.2 Heteroskedasticity 10

2.3 Autocorrelation 12

V Summary and Conclusion 13

REFERENCES 14

Trang 3

TIEU LUAN MOI download : skknchat@gmail.com

Trang 4

TABLE OF APPENDICES Appendix A: List of formulas

Trang 5

Trang 6

LIST OF TABLES AND FIGURES

Figure 1: lin – lin model 5

Figure 2: log – lin model 5

Figure 3: lin – log model 5

Figure 4: log – log model 6

Figure 5: log – log model after dropping “ros” 7

Figure 6: Regression with dummy variables 8

Figure 7: Variance-inflation factor method 10

Figure 8: Park Heteroscedasticity Test on Log[(log(sales)] 11

Figure 9: White Heteroskedasticity Test 12

Trang 7

I Introduction

CEO compensation structure is a key component in the corporate governance structures of firms and the process to obtain the information requires a lot of research and analysis for human resource department It is realized that the CEO salaries vary widely between different firms in the same area Different criteria are set to determine the appropriate salary level for CEOs and among that the firm performance can be regarded as an important indicator The question is how they affect the CEO compensation and related to each other In this paper, given the data set “ceosal1”, we will investigate the three factors, which are sales, return on equity (roe) and return on firm’s stock (ros) that can be used to explain the differences in CEO salaries and discuss the relationships between CEO compensation and these variables from the point of practical relevance, this study will bring a great contribution to regulators and companies in having a better understanding of how the CEO salary or the CEO compensation in general related to the company's performance, thus giving insights in how the paying structure of a firm based on pay-for-performance method can act as a means to tackle agency problems If there is a positive relationship between three variables and positive influence on salaries, the firm can have a more detailed picture of salary variation and consider making appropriate investment decisions for each scenario The purpose of our report is trying to solve three main questions: Firstly, what are the relationships between CEO salaries and sales, return on equity as well as return on firm’s stock? Secondly, are there any connections among these factors and how do they affect the salaries? And lastly, in the model, are there any errors detected and how to deal with this problem?

II.Literature Review

1 Theoretical Foundation

This paper studies whether the firm performance can impact on the variation of CEOsalary CEO salary is one type of financial incentives that is included in the Executivecompensation (Anon, 2018) Several literatures regarding this topic will be reviewed asfollowings:

According to Jensen and Meckling (1986), the agency relationship is "a contract which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision-making authority to the agent." This relationship can be the relationship between CEOs (the principal) and shareholders (the agent), in which the CEO is given some authority to act in the way that maximizing best interest of shareholders However, each party always focus on maximizing

1

Trang 8

their own utilities and it may arise a conflict (also known as "Agency problem") of interestsbetween the needs of the principal and the needs of the agent As the agency problem mayresult in conflicts and misunderstanding between two parties, different mechanisms havebeen found to help solve it (Jensen and Meckling, 1976) For instance, it can be minimized

by ownership structure (Jensen and Meckling, 1976; Fama and Jensen, 1983), the executiveremuneration contracts (Jensen and Murphy, 1990; Lambert et al., 1991) and financialstructure of the firm (Easterbrook, 1983; Jensen, 1986)

1.2 Executive remuneration

According to Shavell (1979), various forms of compensation provide different incentiveeffects on CEOs The relationship between agency problem and CEO compensation can beexplained by two different theories: optimal contract theory and managerial power oftheory (Bebchuk and Fried, 2003) Under the optimal contract theory (OCT), CEOcompensation can be seen as a remedy to the agency problem OCT suggests that boards,working in shareholders' interest, try to provide efficient incentives to managers throughtheir designed compensation schemes that eventually will help them to maximizeshareholder value Under the managerial power of theory, managers have more power thanthe boards, and thus, they have greater ability to extract higher compensation

2 Empirical Findings:

The relationship between executive compensation and agency theory can be found in several empirical research such as (Lambert et al.,1991; Gray and Cannella, 1997) Empirical evidence argued that there is a reduction of the agency problem when the compensation of the manager is tied

to the stock price (Lambert et al.,1991) In addition, Gray and Cannella (1997) found that incentive compensation can be considered as a tool that aligns the interest of both the principal and agent Besides, empirical studies which have examined the link between firm performance and CEO pay stated that the structure of CEO compensation differs from company to company (Murphy, 1999; Arya and Huey-Lian, 2004) Murphy (1999) documented that if the CEO meets the performance target, the CEO will receive the bonus which is usually determined as a given percentage of his salary There are conflicting results of CEO compensation and firm performance Tosi et al (2000) site the findings of two different set of studies One study by Finkelstein and Boyd (1998) report a significant correlation between Return on Equity and cash compensation of only 0.13 and supporting this finding, Johnson (1982) reported the significant correlation of 0.003 between the two variables To sum up all, in line with existing theory, findings of empirical studies suggested that the compensation contract is a useful mechanism for resolving the agency problem and

Trang 9

compensation policy is one of the most important factors in an organizations success(Fama, 1980).

In terms of model specification, among many factors leading to the salary volatility, thisresearch emphasizes on two typical categories: Internal factors: sales, ROE (Return onEquity) and ROS (Return on Sale); External factors (Dummy variables): Industrial firms,Financial Firms, Consumer Product Firms or Transport or Utilities Originally, theequation of our model expressed Salary with Internal factors under linear relationship

Salary = β 1 + β 2 * Sales + β 3 * ROE + β 4 * ROS + u

In which variables are:

a Dependent variable: Salary (Y)

We obtained Salary as Y (thousand $) of 209 observations in 1990 It is apparentlydifficult to identify one specific salary figures for each career, they vary dramatically based

on numerous factors All of them will have certain effects on Salary, therefore, byanalyzing the factors, employees can have a general view about salary fluctuation to makesuitable expenses or investment decisions

b Independent variable: Sales

Which measures revenue (million $) gained from firm’s volume of sales in 1990.The more effective and efficient the performance is, the higher the sale figure

Expectation: a positive relationship between Salary and Sales

c Independent variable: ROE (return on equity)

ROE = Net income (earnings)/Equity of shareholders

ROE is the ratio of net profit to equity, which reflects the ability to use the capital of the business for profit “This index is an accurate measure of how much a return is made and how much accrued interest it generates and it is often analyzed by investors for comparison with other peers in the market when they decide to buy shares of any company” Dr Quynh said The higher the ROE

is, the more effectively the company uses equity capital So why can the ROE affect to salary? When the ROE increases significantly, it means that the

3

Trang 10

company and its employees are operating efficiently This can influence positively to

staffs’ wages, maybe, the managers will raise their salary following a multiplier.1

Expectations: We expect a positive relationship between salary and ROE.

d Independent variable: ROS (return on sale)

ROS = Net profit after Tax / Sales

ROS indicates how much profit accounted for in revenue This ratio means that thecompany is profitable, the larger the ROS is, the greater the profit is So why should weconsider ROS as a variable that impacts in salary? Because when ROS increases, the profit

a company receives after exchanging also goes up It is cheerful news for a corporation andperhaps, they will have some bonuses for their employees adding to salary

Expectations: We expect a positive relationship between salary and ROS.

In terms of research design, the research follows the non-experimental researchdesign It identifies the relationship among salary and other factors The hypothesis testingand other analysis also revolve around these relationships

There are various tests used in this study in order to make the concrete conclusion

of the report In multiple regressions, we conducted 4 tests including testing the overallsignificance of all coefficients, the functional form of regression model and the dummyvariable We also check the errors including multicollinearity, heteroscedasticity, andautocorrelation

IV Data Analysis and Results

1 Model testing

In this part, we will run the OLS on three types of functional form that is lin-linmodel, semi-log model (lin-log & log-lin model), and log-log model to figure out the bestlinear unbiased estimator (BLUE) regression by using Eview10 However, with the

“ceosal1” data set given, we cannot run the log(ros) because of non-positive number in ros

figures in the excel file Therefore, it is obligated that we only obtain roe in lin-log model

and log-log model Through running four models, we will obtain R-squared (R2) andcovariance (CV) of each model and then choose the optimal model with highest R-squaredand lowest CV as followings

a Lin-lin model: Salary= β 1 + β 2 *sales+ βsales+ β 3 *sales+ βroe+ β 4 *sales+ βros + u

Trang 11

Dependent Variable: SALARY

Method: Least Squares

R-squared 0.031914 Mean dependent var 1281.120

Adjusted R-squared 0.017747 S.D dependent var 1372.345

S.E of regression 1360.113 Akaike info criterion 17.2874

Sum squared resid 3.79E+08 Schwarz criterion 17.35144

Log likelihood -1802.541 Hannan-Quinn criter 17.31334

F-statistic 2.252699 Durbin-Watson stat 2.118133

Prob(F-statistic) 0.083369

Figure 1: lin – lin model

b Log-lin model: Log(salary)= β1+ β2*sales+ βsales+ β3*sales+ βroe+ β4*sales+ βros + u

Dependent Variable: LOG(SALARY)

Sum squared resid 57.40118 Schwarz criterion 1.647853

c. Lin-log model: Salary= β1+ β2*sales+ βlog(sales)+ β3*sales+ βlog(roe)+ β4*sales+ βros +u

Dependent Variable: SALARY

Sum squared resid 3.71E+08 Schwarz criterion 17.32867

Figure 3: lin – log model

Prob(F-statistic) 0.009994

d. Log-log model: Log(salary)= β1+ β2*sales+ βlog(sales)+ β3*sales+ βlog(roe)+ β4*sales+ βros + u

Trang 12

Trang 13

Figure 4: log – log model

To conclude, with the highest R2 means that the highest percentage change in

salary are explained by sales, roe, ros jointly, and lowest CV means the most preferred

model, we find out log-log model is the optimal one The equation is:

Log(salary)= β 1 + β 2 *sales+ βlog(sales)+ β 3 *sales+ βlog(roe)+ β 4 *sales+ βros + u

1.2 Testing the overall significance on all coefficients

We have the equation of log-log model:

Ln(salary)= β 1 + β 2 *sales+ βlog(sales)+ β 3 *sales+ βlog(roe)+ β 4 *sales+ βros + u

Trang 14

1.3 Testing the model when dropping one variable

Dependent Variable: LOG(SALARY)

Prob(F-statistic) 0.000000 Figure 5: log – log model after dropping “ros”

Log(salary)=β 1 + β 2 *sales+ βlog(sales) + β 3 *sales+ β log(roe)+ β 4 *sales+ β ros + u (UR)

Dropping “ros”, we obtain a new equation:

Log(salary)= β 1 ’ + β 2 ’ *sales+ β log(sales) + β 3 ’ *sales+ β log(roe) + u (R)

Conclusion: Β4 is not statistically significant different from zero at α = 10%

1.4 Testing on dummy variables

We denote the variables as followings:

D1: industry D2: finance D3: consprod D4: utility

Trang 15

F-statistic 17.97162 Durbin-Watson stat

Figure 6: Regression with dummy variables

2.225653 Prob(F-statistic) 0.000000

D2 = 1, we have:

Log(salary) = β 1 + β 1 ’ * D2+ β 2 * log(sales) + β 3 * log(roe) + β 4 * ros + u

We perform hypothesis testing to check whether β 1 ’ is significant or not?

- Stating the hypothesis:

H0: β1’ = 0

H1: β1’ ≠ 0

- Compute: t-sta = 1.703 > Tc

0.05,205= 1.645 Reject H0

- Conclusion: β1’ is statistically significant different from zero at α = 10% D2 = 1,

we have the equation:

log(salary)= 4.5958 + 0.2549log(sales)+ 0.1285log(roe)+ 5.05- 05rosD3=1, we have:

Log(salary)= β 1 + β 1 ’’ * D3+ β 2 * log(sales)+β 3 * log(roe)+ β 4 * ros + u We perform hypothesis testing to check whether β 1 ’’ is significant or not?

- Stating the hypothesis:

H0: β1’’=0

H1: β1’’≠0

- Compute: t-sta=2.3295 > Tc0.05,205 = 1.645 Reject H0

- Conclusion: β1’’ is statistically significant different from zero at α = 10% D3 = 1,

we have the equation:

log(salary)= 4.6486+ 0.2549log(sales)+ 0.1285log(roe)+ 05ros D4=1, we have:

5.05E-Log(salary)= β 1 + β 1 ’’’ * D4+ β 2 * log(sales)+β 3 * log(roe)+ β 4 * ros + u We perform hypothesis testing to check whether β 1 ’’’ is significant or not?

8

Định dạng
Số trang	25
Dung lượng	213,52 KB