The Effect of Refusing a Bonus by a Top
Executive: An Event Study
R. Vendelbos
University of Amsterdam 10406581
15 June 2015
Bachelor thesis Finance & Organization Supervisor: Msc M. Koudstaal
2
Verklaring eigen werk
Hierbij verklaar ik, Ramon Vendelbos, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op
me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de
referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de
3
Abstract
This paper analyses the effect of refusing a bonus by top executive on the companies’ share price. Agency theory and the efficient market hypothesis, predict no effect, whereas stewardship theory and theories on irrational behaviour of investors suggest a positive effect. To examine the effect of the refusal of a bonus by a top executive, an event study is performed. The event is the announcement of the refusal of a bonus by a top executive and an event window of 10 days prior to the refusal of the bonus announcement and 10 days after the announcement is used. I also perform a more rigourous cross-sectional analysis.
The results show more negative than positive abnormal returns with regard to a top executive refusing his bonus. This might explained by that the refusal of a bonus has a positive effect and that the negative observations are caused by negative company news which in some cases is related to the announcement. Overall, the findings in this study indicate that the announcement of the refusal of a bonus by a top executive has a positive but insignificant effect on the companies’ share price
4
Table of content
1. Introduction……….6
2
.
Literature Review………...72.1. Agency theory and Stewardship Theory………...7
2.1.1. Agency Theory………..7
2.1.2. Stewardship Theory………...…8
2.2. Market Reaction………8
2.2.1 Irrational behaviour of investors……….………...8
2.2.1. Efficient market hypothesis………..9
3.1 Data and Methodology……….9
3.1 Data………9
3.2 Event study…...…....………10
3.2.1 Event of interest………...………11
3.2.1 Event window and estimation period……….………..11
3.2.3 Selection criteria………...11
3.2.4 Calculating the abnormal return………...12
3.2.5 Calculating the CAR………13
3.2.6 T-test.………13
3.3 Cross-sectional analysis………...14
3.4 Descriptive statistics…...…..………..……….15
4. Results………..17
4.1 Event study results………...…18
4.2 Cross-sectional analysis results………20
4.3 Robustness………...22
4.3.1 Robustness of the event study……….………22
4.3.2Robustness of the cross-sectional analysis………..……….23
5. Conclusion and Discussion……….24
6. References………...……….25
7. Appendices………...29
A. Background literature………29
A.1 Bonuses………..29
5
B. Tables………31
C. Event study robustness using a different methodology………...…………..33
C.1 Methodology………...……….…33
C.2 Results………...…...………..34
D. Robustness by testing the assumptions of multiple linear regression……...………...35
D.1 Testing for linearity……..……….…35
D.2 Testing for multicollinearity………..………36
D.3 Testing for normality………….……….………...37
6
1. Introduction
The relation between share price performance and executive pay has been a research topic in many executive pay studies. It reflects shareholders’ objectives and tries to reduce agency problems by aligning the objectives of executives to those of the shareholders (Buck, Liu and Skovorode, 2008). According to Kerr & Slocum (2007), awarding bonuses to top executives is a way to stimulate them to reach their goals. They found that awarding bonuses have an positive effect on the performance of the company.
However, since the recent financial crisis in 2007/2008, more and more top executives started to refuse their bonus. The main reason mentioned by top executives for refusing their bonus is that they felt that they did not deserve the bonus. But there are for no doubt more factors that have affected their decision in refusing the bonus such as public pressure, layoffs, company scandals and government support.
Since the refusal of bonuses by top executives is only a recent phenomenon, no prior research on this specific topic is done. Nevertheless, using agency theory, stewardship theory and literature in respect to market reactions hypotheses are formed.
The purpose of this study is therefore to close this gap by examining the effect of refusing a bonus by a top executive on the companies’ share price. This will be examined by conducting an event study on 34 companies listed on the New York Stock Exchange (hereafter: NYSE) and London Stock Exchange (hereafter: LSE), using an event window of 10 days prior to the refusal of the bonus announcement till 10 days after the announcement. Over this event window the abnormal returns are calculated and statistical significance is tested with a t-test. Thereafter, the abnormal returns are regressed on other explanatory variables to examine the relationship more rigorously.
The results of this study show more negative than positive observations. However, the positive observations are found to be more often significant than the negative observations. This might be explained by other negative company news that in some cases comes together with the refusal of bonuses and causes these negative observations. Additionally, results with regard to the cross-sectional analyses indicate that banks and firms with better performance experience higher abnormal returns.
The paper is organized as follows. Section 2 will provide a literature review on agency theory, stewardship theory and theories on market reaction. Thereafter, hypotheses from these theories are derived. In section 3 the employed methodology and data used in analysis will be described. Next, section 4 will present the results. Finally, section 5 concludes.
7
2. Literature review
Investors and other financial parties are interested in the reaction of the market on an economic event. The first part of the literature review is a paragraph about aligning agency and stewardship theory to the refusal of bonuses by top executives. Thereafter, the second paragraph contains theories of market reactions on the refusal of a bonus by a top executive. Finally, an overview of the theory will be given and hypotheses based on the theories are developed. In addition, extra background literature on bonuses and reputational effects is provided in Appendix A.
2.1 Agency and stewardship theory
In this section of the literature review, two competing theories are discussed. The competing theories are agency and stewardship theory and these theories are used to form the basis of the formulated hypothesis..
2.1.1 Agency theory
In the literature, agency theory is adopted as an appropriate standard for analysis of executive pay (Gomez-Mejia, Wiseman, & Dykes, 2005). In addition, according to Garen (1994), it is clear that principal agent-considerations play a role in the executive compensation. In agency theory there is a so-called principal-agent problem in which one party which is the principal delegates work to the other party which is the agent. As a result of the delegated work, conflicts of interest occur between the principal and the agent. Agency theory describes these conflicts of interest and is concerned with solving the problems which result from this relationship. Firstly, it tries to align the desired goals of the agent and the principal when they are in conflict and secondly it tries to solve the problem of monitoring the agent, which is often difficult or expensive (Jensen & Meckling, 1976). These conflicts also exist between shareholders and executives, in this case the conflict of interest that occurs is that executives will not act to maximize the returns to shareholders but instead act to maximize their own utility. Agency theory is therefore used to propose solution to alleviate principal-agent problems between those parties by for example using performance pay and bonuses (Jensen & Meckling, 1976). As mentioned before, agency theory suggests that a top executive will only refuse his bonus to maximize his own utility and not to maximize overall wealth. As a result, the refusal of the bonus will not improve the top executives’ reputation and investors’
8
expectations about the future earnings of the company. So in lines with agency theory, the refusal of a bonus by a top executive yields a first hypothesis:
H1A: Refusing a bonus by a top executive has no effect on the companies’ share price.
2.1.2 Stewardship theory
In contrast to agency theory, stewardship theory argues that shareholder interests are maximized by shared incumbency of these parties (Davis & Donaldson, 1991). Stewardship theory arises from moral considerations and shows with empirical evidence that most people try do a good job. According to this theory, executives and managers act as stewards wealth through a natural sense of fiduciary duty which is reinforced by threat of legal sanctions (Davis & Donaldson, 1991). Besides, stewardship theory insist that people are collective self-actualizers and seek organizational achievement instead of that they are individualistic and maximize their own utility (Davids, Donaldson, & Schoorman, 1997). So since stewardship theory states that refusing a bonus by a top executive is done to maximize overall wealth and to maximize shareholder returns (Davis & Donaldson, 1991). Attributing stewardship theory to the refusal of bonuses suggests a second hypothesis:
H1B: Refusing a bonus by a top executive has a positive effect on the companies’ share price.
2.2 Market reaction
For a long time a debate has been going on about whether changes in stock prices reflect rational responses to changes in the fundamentals of the company or that also some other factors determine the share price.
2.2.1 Irrational behaviour of investors On the one hand, Keynes (1964) argues that investment is like a game of chances where investors try to find the average expectations of all the investors and try to anticipate to have a favourable return. Besides, there is more evidence that investors behave irrational. Edmans, Garcia and Norli (2007) investigated the effect of international soccer matches on the stock market. They find that there is a significant negative loss effect and an insignificant positive win effect. Besides, they find that the loss effect is bigger for more important matches like world cup matches. Another event study on irrational behaviour of investors is done by
9
Hirshleifer and Shumway (2003) have studied the impact of morning sunshine on daily market index returns. Their research shows that sunshine is strongly and significantly correlated with stock returns. In my opinion, the refusal of bonus by a top executive shows characteristics of the top executive and this could have an effect on rational and irrational expectations of investors. Since the top executive acts in an unselfish way by refusing his, he acts as a steward of wealth just like stewardship theory suggests and in contradiction with agency theory which states that people are individualistic (Davids, Donaldson, & Schoorman, 1997). This will improve the reputation of the CEO and the company which could have an effect on the company’s share price (Hannon & Milkovich, 1996).
2.2.2 Efficient market hypothesis
On the other hand, some academics and researchers are following the efficient market hypothesis (Basu, 1990; Fama, 1970). The efficient market hypothesis suggest that stock prices are volatile due to changes in the discount rate or due to changes in investors’ expectations about the future earnings of the company. The changes in discount rate have an effect on public expectations about the future course of the economy. These changes affect economic activity by influencing the expectations of economic actors such as businesses and financial institutions (Waud, 1990). Furthermore, this theory implies that in any point time share prices reflect all the available information (Fama, 1970). Following the efficient market hypothesis, there should only be a change in the share price of the company as a result of refusing a bonus due to the cash that stays within the company. Because by refusing a bonus this money can be reinvested or paid out to investors in the form of dividend. However, this effect of refusing a bonus will probably not be significant, because the salary expense of a top executive is a very small portion of the total expenses made by the company. So this effect will probably be negligible in the most companies.
3. Data and Methodology
3.1 DataThe datasets, consist of announcements of executives who refused their bonus during the period 2008-2015. Of these 34 observations are collected, whereof 16 companies are listed on London Stock Exchange (LSE) and 18 are listed on the New York Stock Exchange (NSYE). Companies from just these two exchanges have been chosen, because these are two of the largest stock exchanges (Chiang & Jeon, 1991) and both have the Anglo-Saxon model
10
of capitalism (Harvie, Karunanayake & Valadkhani, 2013). To identify the executives which refused their bonuses and the associated companies, financial publications of newspapers are used. The companies and the accompanying announcement dates can be found in Appendix B. The associated data on share prices for all companies is collected from DataStream. To collect the companies’ share prices, all company information is retrieved using equity price (P, adjusted-default). To collect the price indexes of the S&P 500 and FTSE 100, the datatype price index (PI) in DataStream is used.
The data that is used for the cross-sectional analysis is collected using Wharton Research Data Services (WRDS), financial reports, LexisNexis database and related news articles. The data for the measures of the variables LN(FirmSize) and Firm Performance are retrieved from the Compustat database in WRDS. For the variable LN(FirmSize) the total assets of the firms are collected and for the variable Firm Performance which is measured by the return on assets also the net income of the firms is collected. The missing data of some observations is assembled through financial reports. Furthermore, the data for the variable Public Pressure which is measured by the amount of related news articles is retrieved from the LexisNexis database and websites of financial newspapers.
3.2 Event study
To test the effect of refusing a bonus by a top executive on the companies’ share price, an event study is done. An event study is used to measure the effects of an economic event on the firms’ value. It thus measures the effect of a specific event on the value of the firm. The following steps describe how to peform an event study (Mackinlay, 1997):
The first step of an event study is to define the event of interest and to determine the event window and estimation period. The event window is the period over which the share price involved in the event is examined. Secondly, the selection criteria have to be determined for inclusion of firms. Subsequently, the abnormal return has to be calculated, which is the actual ex-post return of the security over the event window minus the expected return of the firm during the event window. To get the cumulative abnormal returns (hereafter: CARs), the abnormal returns have to be accumulated. Finally, statistical significance of the CARs is tested with a t-test.
11 3.2.1 Event of interest
The event date is this event study is the date when it is announced that the top executive refuses his bonus. A common way to collect the event date is to find financial publications from financial papers such as the Wall Street Journal (Mackinlay, 1997). It is this approach that I follow as well.
3.2.2 Event window and estimation period
The event window is usually larger than the specific event of interest. In practice, the event window is expanded to multiple days to capture the price effects of announcements which occur after the stock markets closes on the announcement day and due to that the market may already acquire information about the event prior to the announcement day (Mackinlay, 1997). In Figure 1, the announcement day is defined as 0, while t0 to t1 represents the
estimation window, the event window is between t1 to t2 and the post-event window is from t2
to t3.
Estimation Event Post-event
window window window
t0 t1 0 t2 t3
Figure 1. Timeline for an event study
In this study, an event window of 10 days prior to the refusal of the bonus announcement till 10 days after the announcement is used, where 0 represents the event date. Hsu, Reed and Rochell (2010) also use this event window (-10,10) and find statistically significant results with their event study. Additionally, a second analysis on an shorter event window is done to verify the robustness of my results. This second event window lasts from a day prior to the refusal of the bonus announcement to a day after the announcement. This event window (-1,1) of 3 days is also used by Singal (1996) and Moeller, Schlingemann & Stulz (2005).
According to Mackinlay (1997), it is typical for an estimation period and event window not to overlap, because the event returns could in these case affect the normal returns. An estimation window of 250 days before the event date is recommended in his study and therefore the same estimation window is used in this study to estimate the market model.
12
3.2.3 Selection criteria
To measure the effect of refusing a bonus by a top executive on the companies’ share price this study uses these announcements in the period of 2006-2015. This period is chosen since the refusal of bonus by top executives is a recent phenomenon. Besides, only companies from the NSYE and LSE are chosen, because these are two of the largest stock exchanges (Chiang & Jeon, 1991) and both have the Anglo-Saxon model of capitalism (Harvie, Karunanayake & Valadkhani, 2013).
3.2.4 Calculating the Abnormal Return
The abnormal return is the actual ex-post return of the security over the event window minus the expected return of the firm during the event window. The abnormal return is given by:
𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− 𝐸(𝑅𝑖𝑡|𝑋𝑡) (1)
Where:
- 𝐴𝑅𝑖𝑡 isthe abnormal return for firm i at time t - 𝑅𝑖𝑡 is the actual return for firm i at time t
- 𝐸(𝑅𝑖𝑡|𝑋𝑡) is the expected return for firm iat time t based on characteristics x
The actual return for all firms is retrieved from DataStream. Furthermore, the expected return can be calculated using multiple models such as the market model, Capital Asset Pricing model and Arbitrage Pricing Theory. Results from Brown and Warner (1980) show that a simple methodology based on the market model is well-specified and relatively powerful under a wide variety of conditions. Moreover, the gains of using multiple factor models for event studies are limited due to that the marginal explanatory power of the added factor is small and there is little reduction in variance (Mackinlay, 1997). According to Knapp (1990), the market model is generally used in event studies. The market model is also used in research of Hsu, Reed and Rochell (2010) and Moeller, Schlingemann & Stulz (2005). These are the reasons why also in this study is chosen to calculate the expected returns by using the market model. The market model is given by:
13
Where:
- 𝑅𝑖𝑡 is the return for firm iat time t
- 𝑅𝑚𝑡 is the return of the market portfolio at time t
- 𝛼𝑖 is the intercept of the market model for firm i - 𝛽𝑖 is the slope of the market model for firm i - 𝜀𝑖 is the error term of the market model for firm i
To compute the expected return for firm i on day t, the parameters of the market mode have to be estimated and by means of an Ordinary Least Squares (OLS) regression. The return of the firm 𝑅𝑖𝑡 and the return of the market portfolio 𝑅𝑚𝑡 are retrieved from DataStream. The error
term 𝜀𝑖 and the variance of the error term are considered to be equal to zero (Mackinlay, 1997). When all parameters are known, formula (2) can be filled in to compute the expected return.
Finally, in order to get the abnormal return the normal return has to be subtracted from the actual return, as in equation (1).
3.2.5 Calculating the Cumulative Abnormal Return (CAR)
The Cumulative Abnormal Return (hereafter: CAR) is computed by accumulating all the abnormal returns of the event window.
𝐶𝐴𝑅𝑖(𝑡1, 𝑡2) = ∑𝑡𝑡=𝑡2 1𝐴𝑅𝑖𝑡 (3)
Where:
- 𝐶𝐴𝑅𝑖 is the cumulative abnormal return of firm i for the event window (𝑡1, 𝑡2) - 𝐴𝑅𝑖𝑡 is the abnormal return of stock i at time t
- (𝑡1, 𝑡2) is the event window where 𝑡1 is the first day of the event window and 𝑡2 the last day.
14 3.2.6 T-test
To test if the cumulative abnormal returns of a single firm are statistically different from zero a t-test is done using the following test-statistics and hypothesis:
𝐻0 : 𝐶𝐴𝑅(𝑡1, 𝑡2) = 0 (agency theory) 𝐻1 : 𝐶𝐴𝑅(𝑡1, 𝑡2) > 0 (stewardship theory) 𝑡 = 𝐶𝐴𝑅(𝑡1, 𝑡2) 𝑆𝐷(𝐶𝐴𝑅(𝑡1, 𝑡2)) (4) 𝑡 = 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) 𝑆𝐷(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2)) (5)
Formula 4 is used as the main t-test, using this test the statistical significance of the CAR of every firm is tested. In addition, to verify the robustness of the results also a t-test using the average of the CARs and average of the standard deviations of all observations is used, as in equation (5).
3.3 Cross-sectional analysis
To examine the relation more rigorously, a cross-sectional analysis is performed to explore if the cumulative abnormal returns are significant.
𝐶𝐴𝑅 = 𝛼 + 𝛽1𝐹𝑖𝑟𝑚 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖 + 𝛽2𝐷_𝐵𝑎𝑛𝑘𝑖 + 𝛽3𝐿𝑁(𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒)𝑖+ 𝛽4𝐷_𝑆𝑡𝑜𝑐𝑘𝑒𝑥𝑖 + 𝛽5𝑃𝑢𝑏𝑙𝑖𝑐 𝑃𝑟𝑒𝑠𝑠𝑢𝑟𝑒 + 𝜀𝑖
Where:
- 𝐹𝑖𝑟𝑚 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖 is the Return on Assets (hereafter: ROA) of the firm i. - 𝐷_𝐵𝑎𝑛𝑘𝑖 is a dummy variable and 1 if the company is in the finance sector. - 𝐿𝑁(𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒)𝑖 is the size of the size of the firm i.
- 𝐷_𝑆𝑡𝑜𝑐𝑘𝑒𝑥𝑖 is a dummy variable and 1 if the company is listed on NYSE. - 𝑃𝑢𝑏𝑙𝑖𝑐 𝑃𝑟𝑒𝑠𝑠𝑢𝑟𝑒 is measured by the amount of related news articles.
15
Following the literature, Firm Performance measured by ROA is often used as a financial control and is for example also used by Chen (2012) and Qiu and Yu (2009).
The dummy variable D_Bank controls for a diverged effect on banks. Since almost 40 percent of the observations are banks and banks were among the most criticized companies in the recent financial crisis. Banks were especially criticized by politicians and the public opinion for awarding enormous bonuses ( Fassin & Gosselin, 2011).
LN(FirmSize) is added since abnormal returns are more often detected in large firms
and are positively cross-correlated (Kothari & Wasley, 1989).
D_Stockex is added to check whether different effects occur between the firms listed
on the NYSE and LSE.
Firm Performance is used as a control variable since it is likely that top executives
refuse their bonus more often when they are under public pressure. As a result of that the amount of public pressure might affect the abnormal returns.
3.4 Descriptive statistics
[Table 1: Descriptive statistics event study]
CAR(-10,10) N 34 Positive observations 14 Negative observations 20 Maximum 25.91% Minimum -48.22% Mean -3.22% Standard deviation 14.96% P-value 0.43371
1 The p-value corresponds to the t-value, calculated as in equation (5). This test was done to verify the
16
Table 1 illustrates the descriptive statistic of the sample. One can observe that there are more negative than positive observations and that there is a negative mean in the event window (-10,10), which contrasts hypothesis 1B. One reason for this could be that announcement of executives refusing there bonus is often related to other negative news about the company. This news might overrule the market reaction on the effect of the refusal of a bonus by top executive and cause negative abnormal returns in the most cases. Furthermore, the event window (-1,1), which is used as an robustness check shows slightly more positive observations and a less negative mean. This could be explained by the fact that that in some cases other negative news about the company might be filtered out in this shorter event window.
The small mean of the abnormal returns is a result of having slightly more negative observations than positive ones and is also caused by some observations that consists very negative abnormal returns. Table 1 also illustrates a high standard deviation of the CARs, which can be explained by that almost the same amount of positive and negative observations that are found. Finally, The p-value1 indicates no statistical significance.
[Table 2: Descriptive statistics cross-sectional analysis]
Firm
performance
D_Bank LN(FirmSize) D_Stockex Public
Pressure N 34 34 34 34 34 Maximum 11.93% 1 14.5735 1 2 Minimum 18.99% 0 8.2271 0 24 Mean 1.39% 0.3824 0.4933 11.7341 0.5294 9.3235 Standard deviation 5.22% 2.0003 0.5066 5.9377
In Table 2, the descriptive statistics of the variables that are used in the cross-sectional analysis are shown. The variable Firm Performance is measured by the return on assets (ROA) over the year of the event. For calculating the ROA, the net income was divided by the
17
total assets of the firm. The mean of the variable is 1.39% with a standard deviation of 5.22%. Since the maximum and especially the minimum differ a lot from the mean, there will be tested for outliers in Appendix D.4.
The variable D_Bank is a dummy variable that is 1 if the company is a bank. This variable is added in the cross-sectional analysis since 38.24% of the companies used in this event study are banks. A different relation is expected since banks were very criticized by politicians and the public opinion for awarding enormous bonuses during the recent financial crisis ( Fassin & Gosselin, 2011).
Firm size is used as a control variable and measured as the natural logarithm of the total assets of the company. The companies used in this event study are all listed on a major stock exchange and feature a considerable amount of assets. However, since banks feature a much higher asset base than other companies, the natural logarithm is used.
D_Stockex is a dummy variable that is 1 when a company is listed on the NYSE and
the descriptive statistics show that the variable has a mean of 0.5294 which indicates that 52.94% of the companies are listed on the NYSE, while the other 47.06% of the companies are listed on LSE.
The final variable is Public Pressure and is measured by the total amount of related news articles that are found corresponding to the refusal of a bonus by a top executive. The related data is found in Appendix B. It seems plausible that this variable has an effect on the abnormal returns, but the measure might not be the best measure since the amount of news articles that is found is also related to brand awareness and to the year of the announcement since more and more news is distributed by the internet. This is also indicated by a standard deviation of 5.9377 which shows a wide diversity in the amount of related news articles that are found.
4. Results
As described earlier, there are two opposing views found in the literature with regard to the refusal of bonuses by top executives. On the one hand, a positive effect on the companies’ share price is expected based on stewardship theory and irrational behaviour of investors. On the other hand, no effect on the companies’ share price is predicted by agency theory and the efficient market hypothesis. The results are shown below.
18 4.1 Event study results
[Table 3: Event study results]
CAR(-10,10)
Percentage of positive significant observations at 1% level 78.57%
Percentage of positive significant observations at 10% level 92.86%
Percentage of positive, but insignificant observations 7.34%
Total number of positive observations 14
Number of negative significant observations at 1% level 40.0%
Number of negative significant observations at 5% level 60.0%
Number of negative significant observations at 10% level 65.0%
Number of negative insignificant observations 35.0%
Total number of negative observations 20
Total Number of observations 34
Table 3 is a summarized Table of Table 6, which can be found in Appendix B. Table 6 illustrates the exact company names and the corresponding t-values that are found. The results from Table 3 show that more negative observations are found than positive ones, which contrasts the literature and hypotheses. However, it is quite interesting that 92.86% of the positive observations are found to be significant, while only 65% of the negative observations are significant. Furthermore, 78.57% of the positive observations are significant at a 1% level, whereas only 40% of the negative observations are found to be significant at a 1% level. Besides, Table 3 also illustrates that much more negative observations are found to be insignificant than positive ones.
The results indicate that the positive observations are more significant than the negative observations. This is in line with Keynes (1964) theory on investment and it also indicates that people act indeed as stewards of wealth (Davids, Donaldson, & Schoorman, 1997).
19
some of the refusals of bonuses by top executives are related to other negative news about the company, such as bad performance, company scandals, layoffs and government support. For example, the CEO of Barclays announced the refusal of his bonus after a £290 million fine for rigging the Libor rate( Edmonds, 2014). This news about the company probably affects the share price much more than the refusal of the bonus and explain the negative significant t-statistic that is found. Positive observations that are significant at a 1% level are mostly companies where the negative news that was related to the refusal of the bonus was already known before the announcement. These companies only announced the refusal of the bonus on the event date. So the news related to the refusal of the bonus did probably already affect the share price and could explain the positive significant abnormal returns of these companies.
Figure 2 represents the average of the CARs, maximum and minimum during the event window (-10,10). The average of the CARs does not indicate any large abnormal returns. However, some individual observations do indicate abnormal returns and that is the reason why also the maximum (Lloyds Banking Group) and the minimum (Citigroup) of the CARs are represented in Figure 2. The maximum represents Lloyds Banking Groups’ cumulative abnormal returns during the event window. A clear peak is observed just before the announcement, which might suggest that insider trading occurred. The minimum is represented by the Citigroups’ cumulative abnormal returns during the event window. One can observe that not much fluctuation occurred around the announcement date. However, the large cumulative abnormal returns are observed 6 days after the announcement, but these could be explained by signals of a break up of Citigroup (Dash & Story, 2009).
Interestingly, the robustness check with a shorter event window of one day before the announcement to one day after the announcement shows slightly more positive than negative returns. The shorter event window could therefore filter out some of the other negative company news. Though, these observations are in relation to the main event window(-10,10) less significant, which could be due to that the shorter event window captures less of the announcement effect.
20
Figure 2. The averages of the CARs, maximum and minimum during the event window (-10,10)
4.2 Cross-sectional analysis results
To examine the effect of refusing a bonus by a top executive on the companies’ share price more rigourously, the results of the cross-sectional analysis on the event window (-10,10) are shown in the following Table:
[Table 4: Cross-sectional analysis results]
CARi, Event window (-10,10) (1) (2) (3) (4) (5)
Constant -0.445 (-1.56) -0.054 (1.56) 0.161 (0.95) 0.159 (0.94) 0.172 (0.94) Firm Performance 0.905* (1.71) 0.946* (1.79) 1.040* (1.85) 0.997* (1.78) 0.906 (1.31) D_Bank 0.022 (0.39) 0.084 (1.34) 0.089 (1.50) 0.097* (1.69) LN(FirmSize) -.020 (-1.26) -0.022 (-1.28) -0.024 (-1.23) -0,6 -0,5 -0,4 -0,3 -0,2 -0,1 0 0,1 0,2 0,3 -10 -8 -6 -4 -2 0 2 4 6 8 10 Cu m u lativ e Ab n o rm al R e tu rn s(% ) Event window (-10,10)
Average of the CARs Max
21 D_Stockex 0.026 (0.53) 0.029 (0.55) Public Pressure 0.002 (0.40) R2 N Prob > F 0.100 34 0.097 0.105 34 0.214 0.140 34 0.245 0.147 34 0.316 0.151 34 0.399
*,**,*** denote significance at a 10%, 5% and 1% level, respectively.
In regression 1, the control variable Firm Performance has been regressed on the CARs of the firms. The variable is positive and found to be significant at a 10% level which indicates that Firm Performance affects the CARs in positive way. The coefficient of Firm Performance is around one and indicates that the variable Firm Performance moves in the same way as the CARs which is argumentative since firm performance affects investors’ expectations about future earnings. As a results of these changes in investors’ expectations also stock prices are influenced (Fama, 1970).
In the second regression, the dummy variable D_Bank has been added to the model. The variable which is 1 if the company is a bank. The positive coefficient indicates that banks have a more positive cumulative abnormal return than other companies. Nevertheless, the dummy variable is not significant.
The third regression model adds the variable LN(FirmSize), which has an negative effect on the CARs. This is in contrast with research of Kothari and Wasley (1989), where abnormal returns in relation to larger firms were found to be positively cross-correlated. However, the coefficient of LN(FirmSize) is not significant.
In the fourth regression, the dummy variable D_Stockex is added to the model. This variable is 1 in the case of an company listed on the NYSE and 0 for a company listed on the LSE. The variable is added to check whether there are different results on the CARs between the two stock exchanges. The variable is insignificant, which might be exemplified by the fact that the countries of these stock exchanges feature the same Anglo-Saxon model of capitalism, and thus are not so different from each other.
Finally, the full regression model also includes Public Pressure. This variable has a small but insignificant coefficient which is in contradiction to the expectation. The measure of the variable might me biased since brand awareness and the year of the announcement also
22
influence the amount of related news articles with regard to the bonus refusal. Besides, in the completed model the D_Bank variable is significant which means that the refusal of a bonus by a top executive has more effect on the share prices of banks. However, this coefficient is only marginally significant at the 10% level in this model.
Taken together, the results of the regressions in Table 4 show that few coefficients are significant and that those who are significant are only significant at a 10% level. This can be exemplified by the low R2 which means that the regressors do explain little of the variance of the dependent variable. In addition, a high prob > F is found which indicates that the regression model does not have much explanatory power, potentially due to the small sample. However, the same regression on the shorter event window (-1,1) for robustness shows a much higher R2 and lower prob > F. This means that the model that can be found in Table 8 can explain more of the CARs in the shorter event window. Furthermore, the variables
D_Bank and Firm Performance are found to be significant at 1% level in those regressions.
So in the shorter event window these independent variables do influence the CARs. One reason for this might be that other company news that also influences the share price is filtered out in the shorter event window. In this case the results thus indicate that firm performance and if the company is a bank do influence the abnormal returns occurring from the refusal of a bonus by a top executive.
4.3 Robustness
To verify the robustness of the results, a shorter event window (-1,1) is used. In this section the results of the event study and cross-sectional analysis using the shorter event window are discussed. In addition, robustness of the event study results is verified by performing an event study with a different test-statistic in section C of the Appendices. Furthermore, in part D of the Appendices robustness tests for the assumptions of multiple linear regression are performed as provided by previous literature.
4.3.1 Robustness of the event study
The shorter event window (-1,1) shows more positive and less negative observations than the main event window(-10,10). This can be explained by that some other negative company news is filtered out in the shorter event window. The results of this shorter event window are found in Table 6 in section B of the Appendices. Besides, Table 6 also illustrates more positive significant observations than negative significant observations in the shorter event
23
window. However, the main findings of this robustness test are pretty similar to the results of the main event window (-10,10).
4.3.2 Robustness of the cross-sectional analysis
The results of the cross sectional analysis using the shorter event window (-1,1) are given by following table:
[Table 8: Cross-sectional analysis of shorter event window]
CARi, Event window (-1,1) (1) (2) (3) (4) (5)
Constant -0.016* (-1.89) -0.031*** (-2.79) -0.012 (-0.24) -0.011 (-0.22) -0.023 (-0.39) Firm Performance 0.645*** (2.49) 0.712*** (3.35) 0.721*** (3.38) 0.742*** (3.33) 0.829*** (3.74) D_Bank 0.037*** (2.71) 0.042*** (2.55) 0.040** (2.29) 0.033* (1.80) LN(Firm Size) -0.002 (-0.45) -0.001 (-0.30) 0.001 (0.21) D_Stockex -0.012 (-0.92) -0.015 (-1.05) Public Pressure -0.002 (-1.03) R2 N Prob > F 0.398 34 0.018 0.509 34 0.004 0.511 34 0.007 0.525 34 0.015 0.549 34 0.020
*,**,*** denote significance at a 10%, 5% and 1% level, respectively.
The results of this cross-sectional analysis are comparable to the results of the main event window (-10,10). Both the variables Firm Performance and D_Bank are also found to be positive and significant. The variables are even more significant in the shorter event window.
24
Firm Performance and D_Bank are found to be significant at 1% level in these regressions.
The variables are LN(FirmSize), D_Stockex and Public Pressure are still not significant and these results indicate no effect of these variables on the abnormal returns when a top executives refuses his bonus. However, the cross-sectional analyses in this study indicate that firm performance and if the company is a bank do influence the abnormal returns that occur when a bonus is refused by a top executive.
5. Conclusion and Discussion
This paper examines the announcement effect of refusing a bonus by a top executive on the companies’ share price in the period of 2008 until 2015. During this period the announcement effect is examined by multiple event studies and a cross-sectional analysis. Agency theory, stewardship theory and other related literature are used to formulate predictions about the effect of refusing a bonus by a top executive. In lines with agency theory and the efficient market hypothesis, no effect is expected, while stewardship theory and theory on irrational behaviour of investors predict positive effects.
The results of the event study first of all show more negative observations than positives ones. However, it is quite interesting that 92.86% of the positive observations are found to be significant, while only 60% of the negative observations are found to be significant. Furthermore, the positive observations are mostly significant at a 1% level. This indicates that the refusal of a bonus by a top executive has a positive effect on the companies’ share price and that this announcement effect is more in line with stewardship theory and theories on irrational behaviour of investors. Though, lots of negative observations are found which might be explained by the fact that the announcement of the refusal of a bonus by a top executive in some case comes together with negative company news.
The results with regard to the cross-sectional analyses show little significant variables. Only the coefficients of D_Bank and Firm Performance are significant on a 10% level. This is a result of the low R2 and high prob > F of the model which potentially is a result of the small sample. However, these coefficients are significant at a 1% level in the shorter event window and this indicates that the variables Firm Performance and D_Bank have influence on the CARs. So banks and firms with better performance experience higher abnormal returns as a result of a top executive refusing his bonus.
A possible explanation for the limited significance of some coefficients might be the small sample size. Since no specific literature on this topic exists, this study cannot be
25
compared to other studies. However, further research might explain more about the effect of a bonus refusal by a top executive and the results found in this study.
6. References
Ang, J.S., Cheng, A. & Wuh Lin, J. (2005). Ascertaining the Effects of Employee Bonus Plans. Journal of Applied Economics, 37(12), 1439-1448.
Basu, S. (1997). Investment Performance of Common Stocks in Relation to their Price-Earnings Ratios: A Test of the Efficient Market Hypothesis. The Journal of Finance, 32(3), 663-682.
Battalio, R., Ellul, A., & Jennings, R. (2007). Reputation Effects in Trading on the New York Stock Exchange. The Journal of Finance, 62(3), 1243-1271
Bertrand, M. & Mullainathan, S. (2001). Are CEOs Rewarded for Luck? The Ones without Principals are. The Quarterly Journal of Economics, 116(3), 901-
Boyd, B.L., Bergh, D.D. & David, J. (2010). Reconsidering the Reputation-performance Relationship: A Resourced-based View. Journal of Management, 36(3), 588-609.
Brown, S. & Warner, J. (1980). Measuring security price performance. Journal of Financial
Economics, 8(3), 205-258.
Bruce, A., Buck T., Fattoruso, J. & Skovoroda, R. (2007). Executive Bonus and Firm Performance in the UK. Long range planning, 40(3), 280-294.
Buck, T., Liu X. & Skovoroda R. (2008). Top Executive Pay and Firm Performance in China.
Journal of International Business Studies, 39(5), 833-850.
Chen, D. (2012). Classified boards, the Cost of Debt, and Firm Performance. The Journal of
26
Chiang, T.C. & Jeong, B.N. (1991). A System of Stock Prices in World Stock Exchanges: Common Stochastics Trends for 1975-1990. Journal of Economics and Business, 43(4), 329-338.
Dash, E. & Story L. (2009, January 8). Rubin Leaving Citigroup; Smith Barney for Sale. The
New York Times. Retrieved from http://www.nytimes.com/2009/01/10/business/10rubin.html.
Davis, J.H., & Donaldson, L. (1991). Stewardship Theory or Agency Theory: CEO Governance and Shareholder Return. Australian Journal of Management, 16(1), 49-6.
Davis, J.H., Donaldson, L. & Schoorman, F.D. (1997). ‘Toward a Stewardship Theory of Management’. Academy of Management Review, 22(1), 20-47.
Dykes, B.J., Gomez-Mejia, L.R. & Wiseman, R.M. (2005). Agency Problems in Diverse Contexts: a Global Perspective. Journal of management studies, 42(7), 1507-1517.
Edmans, A., Garciá, D. & Norli, O. (2007). Sports Sentiment and Stock Returns. The Journal
of Finance, 62(4), 1967-1998.
Edmonds, L. (2014, February 3). Barclays boss gives up £2.7m bonus: Chief executive refuses to take money because of ‘very significant costs’ of scandals that have hit the bank.
The Daily Mail. Retrieved from
http://www.dailymail.co.uk/news/article-2551173/Barclays-boss-2013-bonus-Chief-executive-refuses-money-significant-costs-scandals-hit-bank.html.
Fahlenbrach, R. & Stulz R.M. (2011). Bank CEO Incentives and the Credit Crisis. Journal of
Financial Economics, 99(1), 11-26.
Fama, E.F., & Malkie, B.G. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance, 25(2), 383-417.
Fassin, Y. & Gosselin, D. (2011). The Collapse of a European Bank in the Financial Crisis: An Analysis from Stakeholder and Ethical Perspectives. Journal of Business Ethics, 102(2), 169-191.
27
Garen, J.E. (1994). Executive Compensation and Principal-agent Theory. Journal of Political
Economy,102(6), 1175-1199.
Hannon, J.M. & Milkovich, G. T. (1996). The Effect of Human Resource Reputation on Share Prices: an Event Study. Human Resource Management, 53(3), 405-424.
Harvie, C., Karunanayake I. & Valadkhani, A. (2013). Global Output Growth and Volatility Spillovers. Applied Economics, 45(5), 637-649.
Hirshleifer, D. & Shumway, T. (2003). Good Day Sunshine: Stock Returns and the Weather.
The Journal of Finance, 58(3), 1009-1032.
Hsu, H., Reed, A.V. & Rocholl, J. (2010). The New Game in Town: Competitive Effects of IPOs. The Journal of Finance, 65(2), 495-528.
Jensen, M., & Meckling, W. (1976) Theory of The Firm: Managerial Behavior, Agency Costs, and Ownership Structure. Journal of Financial Economics, 3, 305-360.
Jensen, M.C. & Murphy, K.J. (1990). Performance Pay and Top-management Incentives.
Journal of Political Economy, 98(2), 225-264.
Kerr, J. & Slocum, J. W. (1987). Managing Corporate Culture through Reward Systems. The
Academy of Management Executive, 1(2), 99-107.
Keynes, J.M. (1964). The General Theory of Employment, Interest, and Money. New York, United states of America: Harcourt Brace Jovanovich.
Knapp, W. (1990). Event Analysis of Air Carrier mergers and Acquisitions. The Review of
Economics and Statistics, 72(4), 703-707.
Kothari, S. P. & Wasley, C.E. (1989). Measuring Security Price Performance in Size-clustered Samples. The Accounting Review, 64(2), 228-249.
28
Nguyen, N. & Leblanc, G. (2001). Corporate Image and Corporate Reputation in Customers’ Retention Decisions in Services. Journal of retailing and Consumer Services, 8(4), 227-236.
Mackinlay, A.C. (1997). Event Studies in Economic and Finance. Journal of economic
literature, 35(1), 13-39.
Meijer, M.M. & Kleinnijenhuis, J. (2006). News and Corporate Reputation: Empirical Findings from the Netherlands. Journal of Management Studies, 32(4), 341-348.
Moeller, S.B., Schlingemann, F.P., & Stulz, R.M. (2005). Wealth Destruction on a Massive scale? A Study of Acquiring Firm-returns in the Recent Merger Wave. The Journal of
Finance, 60(2), 757-782.
O’brien, R. M. A Caution of Regarding Rules of Thumb for Variance Inflation Factors.
Quality & Quantity, 41(5), 673-690
Qiu, J. & Yu, F. (2009). The Market for Corporate Control and the Cost of Debt. The Journal
of Financial Economics, 93(3), 505-524
Roberts, P. W., & Dowling, G. R. (2002). Corporate Reputation and Sustained Superior Financial Performance. Strategic Management Journal, 23(12), 1077-1093.
Singal, V. (1996). Airline Mergers and Competition: An Integration of Stock and Product Price Effects. The Journal of Business, 69(2), 233-268.
Waud, R.N. (1990). Public Interpretation of Federal Reserve Discount Rate Changes: Evidence on the ’’ Announcement Effect’’. Econometrica, 38(2), 231-250.
Weiss, A. M., Anderson, E., & MacInnis, D. J. (1999). Reputation management as a Motivation for Sales Structure Decisions. The Journal of Marketing, 74-89.
29
7. Appendices
A: Background literature
In this section of the Appendices some extra background literature on bonuses and reputational effects is provided. This is done to clarify how share prices are affected by bonuses and to expound how share prices can be affected as a result of refusing a bonus.
A.1 Bonuses
Jensen and Murphy (1990) have studied the effect of performance pay on incentives of top managers. They found that there is an empirical relation between performance pay and the incentives of top managers. Their study shows a positive and significant relationship, however the effect of performance pay on management incentives was small. A $3.25 change in CEO wealth was found for every $1000 increase in shareholder wealth. To increase shareholder wealth, shareholders are often seeking to align their own motivation with that of the top executives by awarding bonuses that are linked to share price performance. A problem by linking bonuses to share price performance is that the share price may be a noisy measure. That is why it is important for shareholders to optimally design a pay package and to align performance measures to this package which maximize the firms value (Bertrand & Mullainathan, 2001). Buck, Liu and Skovoroda (2008) had similar findings when studying Chinese top executive pay and firm performance. Their results show that executive pay and performance affect each other through both reward and motivation. Research on executive bonus and firm performance is also done on UK firms and also a positive link was found between bonuses and shareholder returns (Bruce, Buck, Fattoruso, & Skovoroda, 2007)
In contrast, Fahlenbrach and Stulz (2011) find some evidence that the CEOs of banks whose incentives were better aligned to that of the shareholders performed worse during the recent financial crisis. However, bank CEOs who received higher option compensation and a larger fraction of cash compensation did not performance worse during the crisis. Bank CEOs did also not reduce their holding in shares anticipating on the crisis and during the crisis.
30 A.2 Reputational effects
In the case that investors are irrational and there is an effect of refusing a bonus by top executives on the companies’ share price then this effect is caused by an improvement in reputation of the top executive. Reputation also has a significant impact on performance. Reputation can be seen as an intangible asset of complementary and reinforcing relationships. Reputation creates synergies which have a positive implication on performance (Boyd, Bergh and Ketchen, 2010). Moreover, the effect of customer esteem and regard and its impact on marketing outcomes such as sales and margins has been studied. The results of the study suggest that reputation has a significant relationship with marketing outcomes and marketing strategy. So reputation is also used as a marketing strategy device (Weiss, Anderson, & MacInnis, 1999). Besides, consumers are more loyal to a company with a better reputation (Nguyen & Leblanc, 2001). Furthermore, research shows that firms with relatively good reputations are more profitable over time. These firms have a competitive advantage over other firms, which is hard to replicate by the competitors. (Roberts & Dowling, 2002). In addition, Battalio, Ellul, and Jennings (2007) results suggest that reputation plays an important role in the liquidity provision process on the New York Stock Exchange. Hannon and Milkovich (1996) find that human resource reputation in the business press has an effect on the companies’ share price. A favourable human resource reputation affects the future earnings of the company. This is supported by Meijer and Kleinnijenhuis (2006) who indicate that positive news in the press contributes to a stronger and more positive reputation. Nguyen and Leblanc (2001) also find that negative news in the press can damage the reputation of the company. Accepting a bonus when the company of the executive is not performing well might anger stakeholders which thus could damage the company’s reputation, share price and future earnings. In contrast, refusing a bonus by a top executive will probably be seen as acting in the good way and as a steward of wealth by investors and stakeholders. As a result of refusing the bonus, the executives will build up a favourable reputation. As mentioned above, reputation has a positive implication on firm performance, so this could be a reason for executives to forgo their bonus. It might even benefit in them in the future if they hold a reasonable amount of shares or in the way of receiving higher future bonus or salary.
31 B: Tables
In this section tables are shown that contain data and results of the study.
[Table 5: Announcement dates and number of related news articles per company]
Company: Announcement date: Number of related news
articles:
Listed on NYSE:
Aflac 24-02-2009 9
Bank of Montreal 02-02-2009 5
Coca-Cola Company 05-04-2012 16
Fiat Chrysler Automobiles 07-03-2013 8
General Electrics 18-02-2009 11
Gold Fields 22-08-2013 4
Goldman Sachs Group 17-11-2008 9
Hartford Financial 05-04-2012 5
Home Depot 30-03-2009 7
ING 22-03-2011 5
IBM 21-01-2014 24
Jones Lang Lasalle 15-04-2011 2
Morgan Stanley 18-12-2009 8
Plum Creek Timber 18-12-2014 15
Royal Bank of Canada 02-02-2009 6
Sony 01-05-2013 17
United Continental Holdings 04-01-2010 9
UBS 04-03-2011 8 Listed on LSE: Barclays 03-02-2014 23 British Airways 10-06-2010 8 BG Group 05-04-2013 4 BHP Billiton 03-08-2012 22 Centrica 04-11-2013 8 Citigroup 31-12-2008 7 Commerzbank 15-02-2013 8
Eversource Energy (formerly known as Northeast Utillities)
13-03-2012 2
First Group 27-06-2013 3
G4S 28-08-2012 2
Hibu (formerly known as Yell Group)
21-06-2009 3
Lloyds Banking Group 13-01-2012 9
Morrisons 27-03-2014 16
Royal Bank of Scotland 28-01-2012 14
Standard Chartered 04-03-2015 12
Rio Tinto 09-02-2012 8
[Table 6: The t-values of the CAR t-test]
Company: t-value CAR (-10,10): t-value CAR (-1,1):
32
Aflac -2,836*** -4.005***
Bank of Montreal -0.528 0.477
Coca-Cola Company 3.648*** 9.243***
Fiat Chrysler Automobiles -0.274 1.871*
General Electrics -2.879*** -1.759*
Gold Fields 1.878* -2.768***
Goldman Sachs Group -0,448 -1.559
Hartford Financial -2.065** -3.139***
Home Depot 2.887*** 1.180
ING 0.067 2.947***
IBM -2.085** -1.955**
Jones Lang Lasalle -0.915 -0.853
Morgan Stanley -1.980* -3.383***
Plum Creek Timber 2.647*** 2.852***
Royal Bank of Canada -2.091** 1.969**
Sony 2.980*** 0.983
United Continental Holding 3.544*** -0.035
UBS -1.227 0.245 Listed on LSE: Barclays -2.732*** -2.084** British Airways 2.522*** 0.882 BG Group -3.103*** 3.624*** BHP Billiton 3.024*** -0.128 Centrica -3.577*** 2.228** Citigroup -5.159*** -1.173 Commerzbank -3.554*** 1.621
Eversource Energy (formerly known as Northeast Utillities)
3.239*** -0.776
First Group 1.855* 1.115
G4S -1.528 -1.499
Hibu (formerly known as Yell Group)
-2.414** -2.708***
Lloyds Banking Group 2.884*** 4.804***
Morrisons -7.272*** 2.610***
33
Standard Chartered 2.849*** 1.469
Rio Tinto -1.410 -1.194
*,**,*** denote significance at a 10%, 5% and 1% level, respectively.
C: Event study robustness using a different methodology
C.1 Methodology
The abnormal returns for this robustness test are calculated using the same formula and the market model as described in section 3.2.4. Thereafter, the methodology deviates from the methodology described in sections 3.2.5 and 3.2.6. For robustness the average of all the abnormal returns is computed and used to perform a t-test on the average of the CARs of all observations. This test uses the short event window (-1,1) and also the longer and main event window (-10,10). The following formulas are used for this t-test:
Firstly, the average of the cumulative abnormal returns are calculated as follows:
𝐶𝐴𝑅 ̅̅̅̅̅̅(𝑡1, 𝑡2) = 1 𝑁∑ 𝐶𝐴𝑅 𝑡2 𝑡=𝑡1 (6) Where:
-𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) is the average of the cumulative abnormal return of the firms for the event window (𝑡1, 𝑡2).
- CAR is the cumulative abnormal return of firm i for the event window (𝑡1, 𝑡2)
- N is the number of events
- (𝑡1, 𝑡2) is the event window where 𝑡1 is the first day of the event window and 𝑡2 the last
day. The event windows are (-10,10) and (-1, 1).
Secondly, the standard deviation of the average of the CAR’s can be found by:
𝑆𝐷(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2)) =𝑁1∑𝑁𝑖=1𝛿𝑖(𝑡1, 𝑡2) (7)
34
- 𝜎𝑖2 is the standard deviation of the return of firm I
At last, to test if the average CARs are statistically differ significant, a t-test is used with the following test-statistic and hypothesis:
𝐻0 : 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) = 0 𝐻1 : 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) > 0 𝑡 = 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) 𝑆𝐷(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2)) (5) C.2 Results
The results of the t-test on the average of the CARs are found in the following table:
[Table 7: Robustness using the average of CARs t-test]
CAR(-10,10) CAR(-1,1)
N 34 34
Positive observations 14 18
Negative observations 20 16
Average of the CARs -0.0322 -0.0075
0.01635 Average of CARs standard deviations 0.0411
t-value -0.7829 -0.4608
The results of the t-test on both the event windows show negative and insignificant t-values. This is contradiction to the results of the t-tests which is done on the single firms, which are often significant within particular the positive observations. This can be explained by the fact that around the same number of positive and negative observations is are found and that they counteract each other. This causes a mean close to zero and a higher standard deviation whereby no significance is found using this methodology.
35
D. Robustness by testing the assumptions of multiple linear regression
In this part of the Appendices the robustness of the model is tested as provided by previous literature. Linearity, multicollinearity and normality are tested in this section. For homoscedasticity no test is performed since the robust standard error regression which is performed in STATA corrects for homoscedasticity. In addition, there will also be tested for outliers.
D.1 Testing for linearity
The ordinary least squares estimator will try to fit a straight line on the data and that is the reason why linearity is assumed for the ordinary least squares regressions in STATA. So to test for linearity the residuals are plotted to the predictor variables in STATA. The plots in the Figures below do not indicate a clear departure from linearity. So it is assumed that the relationships between the response variable and the predictor variables are linear.
Figure 3. Residuals Firm Performance
-. 4 -. 2 0 .2 .4 R e si d u a ls -.2 -.1 0 .1 ROA
36
Figure 4. Residuals LN(FirmSize)
Figure 5. Residuals Public Pressure
D.2 Testing for multicollinearity
Multicollinearity occurs when independent variables are highly correlated and this can affect the outcome of the regression. Multicolinearity can be tested by using the variance inflation factor. According to O’brien (2007), a variance inflation factor of 10 or higher indicates
-. 4 -. 2 0 .2 .4 R e si d u a ls 8 10 12 14 16 lnFIRMSIZE -. 4 -. 2 0 .2 .4 R e si d u a ls 0 5 10 15 20 25 Public Pressure
37
multicollinearity. The results of the variance inflation factor of the different variables are given in Table 9. All the variances inflation factor are found to be below 10. So in this regression analysis no multicollinearity problems occur.
[Table 9: Multicollinearity test results]
Variables: Variance inflation factor 1/variance inflation factor Firm Performance 1.40 0.7144 D_Bank 2.45 0.4083 LN(Firm Size) 2.59 0.3859 D_Stockex 1.09 0.9147 Public Pressure 1.55 0.6454
Mean of variance inflation factor 1.82
D.3 Testing for normality
For valid hypothesis testing the normality of residuals is required. The normality has to be valid for assuring that the p-values for t-tests and f-tests are valid. To check for normality in this section a plot of the kernel density estimate will be shown in Figure 5 and in addition a Shapiro-Wilk W test for normality is done with the results shown in Table 10. Both the plot and high prob > Z value indicate that the residuals are normally distributed.
38
Figure 5. Kernel density estimate
[Table 10: Normality test results]
Shapiro-Wilk W test for normal data
Variable Observation W V Z Prob > Z
Residuals 34 0.9679 1.112 0.2410 0.4049
D.4 Testing for outliers
Outliers are observations with a large residual which have an unusual influence on the predictor estimators. So outliers might influence the regression and therefore it is preferable to remove outliers from the observations. To identify outliers studentized residuals are used. Studentized residuals with a value that exceeds +2 or -2 indicate outliers. The observations with the 10 highest studentized residuals are given in Table 12 and the observations with the 10 lowest studentized residuals are given in Table 11. Looking at the Tables 11 and 12, one can observe no value of the studentized residuals that exceeds +2 or -2, so the data does not consist any outliers.
0 1 2 3 4 D e n sit y -.4 -.2 0 .2 .4 Residuals
Kernel density estimate Normal density
kernel = epanechnikov, bandwidth = 0.0476
39 [Table 11: The lowest 5 studentized residuals]
5 lowest observations
Firm Performance
D_Bank LN(FirmSize) D_Stockex Public
Pressure Studentized residuals 1 -0.0143 1 14.4774 0 7 -0.3976 2 0.1898 0 9.1498 0 3 -0.1945 3 0.0178 0 11.3398 1 9 -0.1942 4 -0.0141 0 13.5694 0 11 -0.1809 5 0.0736 0 9.5815 0 16 -0.1795
[Table 12: The highest 5 studentized residuals]
5 highest observations
Firm Performance
D_Bank LN(FirmSize) D_Stockex Public
Pressure Studentized residuals 30 0.0037 1 13.4952 0 12 0.1604 31 -0.0405 0 8.8951 1 4 0.1677 32 -0.0043 1 14.5735 0 14 0.1707 33 -0.0084 0 11.9110 1 17 0.2228 34 -0.0015 1 14.2233 0 9 0.3223
