CEO activism – delegated philanthropy or agency problem? by Silvia Dimitrova s2702649 Faculty of Economics and Business Master of Science in Finance UNIVERSITY OF GRONINGEN Thesis supervisor: dr. Swarnodeep Homroy Abstract

(1)

CEO activism – delegated philanthropy or agency problem?

by

Silvia Dimitrova

s2702649

Faculty of Economics and Business

Master of Science in Finance

UNIVERSITY OF GRONINGEN

Thesis supervisor: dr. Swarnodeep Homroy

Abstract

The purpose of this paper is to examine the motives of CEOs’ public announcements on social issues like LGBT rights, climate change, racial or religious discrimination and gender equality, which are unrelated to the core business of the company. We conduct an event study for 115 such CEO announcements, between 2010 and 2018 in the USA. We find no statistically significant effects of the CEO announcements on social issues. However, there are significantly positive returns for investors when we consider only the more recent events or announcements to a live audience. An additional result is that there is a significant increase in the CEO pay after such announcements. However, we do not observe a similar significant effect on the cost of capital. Therefore, it appears CEOs use these announcements for their personal benefit and, as such, can be considered an agency cost.

(2)

2 1. Introduction

The world that we live in grows exponentially, with regard to its population, production and development. We create a lot, we waste a lot and in the past, we have not always considered all the consequences of our actions. This is all changing now, with activists and regulators demanding and posing various requirements regarding the environmental and social impact of companies. Yet, Chatterji and Toffel (2017) argue that it is not enough for companies to make their operations “more green” in order to solve today’s sustainability issues. The amount of social and environmental problems that we are facing nowadays goes beyond the scope of any one company’s corporate social responsibility (CSR) initiatives.

Despite their best efforts, corporations cannot completely eradicate the looming threat of increasing surface temperatures or the eminent discrepancy in pay of men and women. That, however, does not mean that CEOs of companies are powerless in the face of such issues. On the contrary, we see an increasing number of corporate leaders using their platforms to raise awareness and share their support for equal opportunities among different genders, races, religions, and sexual orientations. We see more CEOs discussing their companies’ plans on controlling climate change. Yet, this wave of CEO activism carries with it some new unanswered questions and potential challenges. It raises the question of why the CEOs decide to speak out in the first place – is it delegated philanthropy as defined by Benabou and Tirole (2010) or CEO agency.

(3)

3 In their paper, Benabou and Tirole (2010) outline the different reasons for companies to engage in CSR. They suggest that most companies invest in “green” technology and practices to fulfil the preferences of their stakeholders, the so-called delegated philanthropy strategy. Therefore, if the CEOs of companies engage in activism simply to please their stakeholders, a stock price reaction should be observed following such an event. In contrast, CEO agency, or the capacity of the CEOs to act independently and to follow their own beliefs, is another possible reason for corporate leaders to engage in social issues. This causes divergence between the actions of the company leader and the other shareholders, who want to maximize stock returns. If this is the case, positive abnormal returns will not be observed since the CEOs do not adhere to their stakeholders’ preferences and instead choose to act on their own. For instance, such actions can be driven by a desire for a higher pay, meaning that the CEOs mainly act for their personal benefit. In other words, these CEOs might only support certain social causes if they believe they will receive a higher salary by building their own brand and refining their reputation. Another possibility to explore is that the CEOs engage in social issues for the benefit of the company. For instance, such CEOs might engage in activism in order to improve the company’s opportunity to attract external funds, thus lowering its cost of capital. A study by Almeida et al. (2004) discovers that financially unconstrained firms tend to keep lower cash to cash flow sensitivity ratios, meaning that these companies that attract funds more easily do not feel the need to keep too much cash in reserves. Therefore, appearing more sustainable on the outside, through supporting social issues, could be a way for companies to hold less cash.

The paper’s empirical research begins with an event study, which aims to discover the market reaction to CEO activism, or more specifically, to CEO announcements on social topics. These social topics are placed into four categories, which encompass the majority of CEO activism. These are LGBT rights, climate change, gender equality and racial or religious discrimination. The event study covers a sample of 115 of these events between 2010 and 2018 for all S&P 500 companies. Both parametric and non-parametric tests are conducted to test for any significant abnormal returns, using the market model as a reference (MacKinlay, 1997).

(4)

4 Finally, additional tests are conducted to better assess the reasons behind CEO activism. More specifically, we examine the CEO pay in the years before and after the event to explore whether there is a positive association between CEO activism and pay raise. A following study explores whether there are any differences in the cash flow sensitivity of cash in the years before and after the event occurs. That is, it investigates whether CEOs speak out on social topics to decrease the amount of cash their companies have to hold as reserves, because of better financing sources.

This study does not find any significant results from the event study and thus, it fails to reject the hypothesis that there are no abnormal returns due to the CEO activism. However, the sensitivity analyses yield some interesting findings. When the sample is split into older and more recent events, there are statistically significant positive abnormal returns due to CEO activism for the most recent events. Similarly, the second sensitivity analysis discovers positive abnormal returns for the subsample, which only includes events that occurred in front of live audience. In contrast, no abnormal returns are found for the subsample, which includes televised interviews or newspaper articles.

Moreover, the additional tests find there is a significant difference in the CEO pay before and after the event. That is, there is a statistically significant pay increase for the CEO of the company if they speak out on social issues. Milbourn (2003) suggests that such pay rises might be due to an increase in CEO reputation, which leads to improved bargaining power. Finally, the tests failed to discover any consistent differences in the cash reserves of companies whose CEOs are involved in social issues. Therefore, no support is found for the hypothesis that such events lower the cost of capital.

The paper will begin by examining the literature on the topic and analysing the different methods and results. Then, the data for the study will be explained and summarized, followed by a thorough explanation of the applied methodology. Finally, the results will be presented, a few sensitivity analyses will be conducted and the paper will be concluded.

2. Motivation

(5)

5 Zuckerberg who created a post on Facebook in support of immigrants, sharing his family’s and his wife’s personal experiences of moving to the USA. After Donald Trump’s decision to retreat from the Paris Climate Act in 2017, there was a wave of company CEOs who rushed to share their disagreement with the president, as well as to share their personal take on climate change (McGregor, 2017).

As a result, there is an increased scientific interest in the reasons behind this new wave of CEO activism. Chatterji and Toffel (2017) conduct a field study to investigate one such instance, when Tim Cook gave a supporting speech for the LGBT community. Chatterji and Toffel explore how this public announcement affected customers’ willingness to purchase Apple products. Perhaps not surprisingly, customers who shared Cook’s views on the matter were more willing to purchase Apple products and vice versa. What this tells us is that one simple announcement such as that by Tim Cook can be both rewarding and damaging for the company, depending on people’s personal beliefs.

Another similar study by Selznick (2017) investigates the increasing number of such events from a legal perspective. The paper aims to discover what power the government has when CEOs are using their influence to put their own opinions forward, by looking at both ends of the spectrum – at CEOs who support and who are against certain social issues. The paper concludes on a slightly perplexing note, since the government of the USA is forbidden to undermine anyone’s freedom of speech; however, there is no consensus on how the stakeholders of the company might react to such announcements. Selznick hypothesizes that Starbucks’ decision to train its personnel on how to handle racial and religious bias might be praised by its customers, who share similar values on these issues, and not by others. In addition, Chick-Fill-A’s support of the “traditional American family” can be seen as discriminatory to its stakeholders, who support more versatile definitions of family. However, the lack of quantitative analysis in this study makes it difficult to determine how stakeholders and mostly, investors will react to CEO activism.

(6)

6 (Romm, 2014). Therefore, exploring the reasons behind public activism by the CEO is important for determining why the CEOs engage in such social issues in the first place.

The paper by Benabou and Tirole (2010) investigates and categorizes the different reasons of individuals to engage in individual or corporate social behaviour. They outline three main reasons for engaging in such behaviour – a win-win strategy, delegated philanthropy and insider-initiated philanthropy. They argue that the former two fall under the bigger umbrella of strategic CSR, with companies engaging in sustainability to improve financial performance or to fulfil stakeholders’ preferences, whereas the latter reason stems from the shareholders’ own belief system and is not directly linked to higher returns. Although difficult to say why most companies engage in CSR, Benabou and Tirole suggest that delegated philanthropy might be the most common reason, as it fits the stakeholder view well. A study by Dimson et al. (2015) finds evidence of the win-win and delegated philanthropy views when they investigate how engagements in environmental, social and governance policies affect the stock market returns of companies. They find one-year abnormal returns of +1.8 percent, consisting of +4.4 percent for successful and no significant results for unsuccessful engagements. In other words, CEOs of companies might engage in prosocial behaviour because this is what their stakeholders want them to do.

However, given that the social issues that CEOs support have no relation to the companies’ core businesses, it is interesting to analyse whether their public activism is delegated by their stakeholders, or simply a manifestation of their own agency. In other words, does this CEO activism follow under the delegated philanthropy theory or is it agency.

Clearly, this area of research is recent and thus, there are almost no papers on the topic. However, it is abundantly clear that researchers are taking an interest at the increasing amount of this particular kind of CEO activism. Nevertheless, proper quantitative analysis on the topics is inadequate and this study aims to fill the gap, since the majority of the papers only assess the qualitative aspects of the CEO announcements.

Therefore, the first hypothesis of this study follows the findings of Chatterji and Toffel (2017) that CEO announcements on social issues have an underlying effect on the company’s stakeholders, as well as the paper by Selznick (2017), which theorizes that stakeholders and mainly investors might be impacted by the news.

(7)

7 Finally, additional tests will be conducted to investigate other possible reasons for CEOs’ acts of support for various social topics. To assess the topic thoroughly different considerations about the likely motives of corporate leaders will be made. On the one hand, as discussed above and in line with Dimson et al. (2015) and Benabou and Tirole (2010), positive abnormal returns from the event study indicate that CEOs’ support is driven by purely financial considerations. In other words, if CEOs anticipate their announcements of support for a number of areas to lead to higher stock market returns, these acts fall under a win-win scenario. CEOs become branded as “the good guys” and investors value their companies more. However, if there are no positive abnormal returns, it seems logical to categorize these announcements as manifestation of their own agency. That is, the CEOs use their platforms to speak out on social issues; however, they do so only because they benefit from this act.

For instance, the debate between the conflict-resolution hypothesis and the overinvestment hypothesis sheds some light on the matter (Jo and Harjoto, 2011). While the conflict-resolution theory suggests that CEOs invest in CSR to resolve issues with the company’s stakeholders, the overinvestment theory proposes that CEOs use corporate social responsibility as an excuse to make unnecessary expenses. Cai et al. (2011) investigate this issue in the context of CEO compensation and whether investing in CSR is followed by an increase in CEO pay. Their study finds support for the conflict-resolution hypothesis and fails to discover any evidence that CEO pay is positively associated with investments in CSR. In contrast, Milbourn (2003) argues that an investment in CSR can ultimately lead to a better reputation of the CEO. As a result, the bargaining power of the CEO increases leading to a possible higher compensation, in support of the overinvestment theory. Examining the reverse relationship, Mahoney and Thorn (2006) investigate how CEO compensation affects CSR weaknesses and strengths, and discover that CEO bonuses are significantly positively related to CSR strengths. This result indicates that higher CEO compensation can be a trigger for better social performance by the firm.

Therefore, the next hypothesis of this study deals with CEO compensation and whether it is affected by corporate leaders’ activism through their announcements on social issues.

Hypothesis 2: CEOs’ public announcements on social issues lead to changes in their respective salaries.

(8)

8 from the ability of a company to attract internal and external funding. Almeida et al. hypotheses that firms, which experience difficulties at attracting funds, should observe a positive cash flows sensitivity of cash, in contrast to unconstrained firms. Their study finds systematic evidence for this hypothesis and concludes that the propensity to save cash out of cash flows is different depending on how financially constrained the company is. This finding is supported by Khurana at el. (2009), who further extend the analysis and discover that the cash flow sensitivity to cash diminishes with financial development. Attig et al. (2014) investigate, too, the topic of cash flows sensitivity, while relating it to CSR. Their study assesses to what extent CSR determines a company’s propensity to save cash of cash flows and discovers that a higher degree of investment in CSR is positively associated with a lower cash flow sensitivity of cash. This finding is the strongest for the categories Community, Diversity and Human Rights of the KLD classification system. This finding supports the view that investing in CSR leads to an easier access to financial means and thus, firms with higher CSR experience less financial constraints, leading them to save less cash for future needs.

As a result, the final hypothesis of this paper focuses on whether activism on social issues, as defined in this paper, leads to a decrease in cash flow sensitivity of cash, as argued by Attig et al. (2014)

Hypothesis 3: CEOs’ public announcements on social issues lead to lower cash flow sensitivity of cash.

3. Data

3.1. Event study

This study covers 115 events from 2010 until 2018. All the investigated companies belong to the S&P 500 and the full list of events can be found in Table 1 in the Appendix.

Insert Table 1 here

(9)

9 core business. Interestingly, companies rarely discuss some of these issues for long periods of time. However, there is a noticeable increase of support from CEOs against laws, which limit the freedom of certain groups of people. Such laws include the North Carolina transgender law, or USA’s retraction from the Paris Climate Agreement, etc. Additionally, this study only investigates announcements that are made by the CEO of a company personally, and not by a representative party. As investigated by Chatterji and Toffel (2016), Apple’s CEO Tim Cook’s personal take on a religious freedom law affected the customers of the company with regard to their shopping behaviour, depending on whether their own beliefs aligned with Cook’s. Therefore, the CEO, a public figure with individual beliefs and values, is worth differentiating from the whole company, and statements by a representative body cannot be considered as insider initiated philanthropy, which is what this study investigates. Finally, in order to keep the events relevant and the information reliable, only these announcements that can be verified are hand-picked. The sources of data include The Wall Street Journal, The New York Times, The Guardian, Independent, Fortune and the Huffington Post.

For the event study, I require the daily stock prices of the companies in the sample. This data is obtained from Datastream from the Total Return Index of each company. These stock prices are then transformed into stock returns through the following formula:

𝑆𝑡𝑜𝑐𝑘 𝑟𝑒𝑡𝑢𝑟𝑛𝑡= 𝑙𝑛 (𝑆𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒𝑡 _{𝑆𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒} 𝑡−1

⁄ ) (1)

The index, which is used as a proxy for the market, is S&P 500, which is a suitable choice given that the entire sample is from the same index. This information is, also, obtained from Datastream.

Table 2 illustrates the summary statistics of the stock returns in the event window, for each separate day.

Insert Table 2 here

(10)

10 Table 3 summarizes the descriptive statistics for the parameters of the market model – alpha and beta, as well as the abnormal market and risk-adjusted returns in the estimation window.

Insert Table 3 here

Firstly, alpha is a security’s excess return relative to the return of a benchmark index (Investopedia, 2018). It represents the part of the stock’s return that cannot be explained by the market. As a result, an alpha coefficient of 0, as observed in Table 3, means that the stock has earned a return which is tolerable for the given level of risk.

Secondly, the beta of the market model is the systematic risk that the security has relative to the market return (Investopedia, 2018). Beta represents the tendency of a security's returns to respond to the market’s fluctuations. As a result, a beta coefficient of 1 means the security’s return will move in line with the market return, which is very close to the observed figure in Table 3 of 0.978.

Finally, the abnormal market returns in the estimation window represent the excess of the actual stock returns over the predicted returns by the market model. If there are no computational or other errors, this value should be 0, as observed in Table 3.

3.2. Cross-sectional regression model variables

Following the event study, a cross-sectional regression is run on the cumulative abnormal returns (CARs1) to further explore the data. The control variables that are used for the regression are Total Assets (log), Net Sales (log), ROA, Leverage, Market-to-Book, Board Composition, CEO Tenure and 3 dummy variables – for industry, for year and for media. The former five variables are obtained from Datastream, whereas Board Composition and CEO tenure are from ExecuComp.

The definition for all variables can be found in table 4. The summary statistics follow in table 5.

Insert Tables 4 and 5 here

(11)

11 3.3. Additional tests

Finally, as indicated in the literature overview, several additional tests will be conducted to further examine the issue of CEO activism. Therefore, data is required for the annual cash balances and cash inflows to measure the cash to cash flow sensitivity. Such data can be collected from Datastream. Next, the CEO pay and how it differs on a yearly basis before and after each event is investigated. The data on CEO salary is obtained from ExecuComp.

4. Methodology 4.1. Event study

For the event study, an estimation window and an event window need to be designed in order to assess whether there are any abnormal returns associated with the events (MacKinlay, 1997). Most event studies use an estimation window of one year (or 250 working days before the event window), which is also applied to this study. This estimation period is used for calculating the normal market returns. In contrast, the event window is relatively shorter and it covers the days before and after the event date. In this paper, an event window of 4 days is estimated, for days [-1;2]. Studying one day before the event is done to incorporate any possible news leaks about the CEO announcement and the two days after the event are included to reflect that not all information is always incorporated on the same day of the event (Kothari and Warner, 2006). There are no confounding events, which occur in any of the studied event windows.

Next, the normal performance of all examined companies will be assessed using the market model. The abnormal returns are calculated as the difference between the actual stock returns and the estimated normal returns, using the following formula:

𝐴𝑅_𝑖𝑡 = 𝑅_𝑖𝑡− 𝐸(𝑅_𝑖𝑡|𝑋_𝑡) (2)

where ARit is the abnormal return of each company, Rit is actual return of each company and

E(Rit|Xt) is the estimated normal return.

The normal return is estimated by using the market model, meaning that a linear association between the stock return and the market return is assumed. This relationship can be illustrated by the following formula:

(12)

12 where αi is defined as the non-systematic portion of the model risk, βi is the systematic portion

and Ri and Rmt are defined as the stock return for each company and market return, respectively.

Abnormal returns are calculated for all four days of the event window, for all 115 events. An aggregate figure will provide easily interpretable results, which is why an average abnormal return will be calculated across events, through the following formula:

𝐴𝐴𝑅_𝑡 = 1

𝑁∑ 𝐴𝑅𝑖𝑡

𝑁

𝑖=1 (4)

where AARt represents the average abnormal returns for each day of the event window and N

is the number of events.

Moreover, the abnormal returns can be aggregated through time to calculate cumulative abnormal returns (CAR) for each event for a specified time frame. This is yet another method of presenting the results in a concise manner and the following formula can be applied:

𝐶𝐴𝑅𝑖(𝑡1,𝑡2) = ∑𝑡2𝑡=𝑡1𝐴𝑅𝑖𝑡 (5)

where CARi(t1,t2) is defined as the cumulative abnormal return for each event accumulated for

different event windows.

Finally, all CARs can be averaged to calculate cumulative average abnormal returns for the entire dataset, which offers an easy to interpret number.

Thereafter, the non-parametric2 test that is implemented is the one suggested by Campbell et al. (2010) and is frequently applied to studies, in which the data is not normally distributed. This test is referred to as the generalized sign test and it compares the anticipated fraction of abnormal returns with a particular sign with the actual fraction of these returns.

4.2. Sensitivity analyses

To check the validity of the results through time or across securities, a few sensitivity analyses will be conducted. Firstly, given the increasing frequency of CEO announcements from S&P 500 companies over the years, it is intriguing to assess whether there is any significant difference among the abnormal returns over the years. Therefore, the abnormal returns of the

2_{An additional parametric test is conducted as described by MacKinlay (1997) and is not included in the main}

(13)

13 most recent two years and the oldest four years are compared. This separation of the sample and removal of the events, which occurred around the middle of the investigated period, can help with distinguishing whether any significant changes have occurred in how the market perceives CEO activism. Additionally, more years are investigated from the beginning of the sample (four compared to two), since the events occurred less frequently in that period.

Additionally, Sanghara (2016) suggests that different methods of announcing new information have varying degrees of success, because of the different communication channels and the different audiences. In this dataset, there are announcements via Twitter, via an online publication or through a public appearance such as a press conference. A clear distinction among these types of announcements is that a public appearance allows the audience to have a direct contact with the CEO and observe their body language and emotions first hand. Therefore, it is intriguing to explore whether there is any difference in the market reaction. As a result, the dataset is split into two parts – public appearances versus all others.

Both sensitivity analyses will be assessed through parametric and non-parametric tests, in parallel with the event study.

4.3. Multiple regression model

In order to inspect empirically the relationship as initially explored in the event study, a cross-sectional regression is estimated3:

𝐶𝐴𝑅_𝑖 = 𝛽₀+ 𝛽₁ ∗ 𝑆𝐴𝐿𝐸𝑆_𝑖+ 𝛽₂∗ 𝑇𝑂𝑇𝐴𝑆𝑆𝐸𝑇𝑆_𝑖+ 𝛽₃∗ 𝑅𝑂𝐴_𝑖 + 𝛽₄∗ 𝑀𝑇𝐵_𝑖+ 𝛽₅∗ 𝐿𝑉𝑅𝐺_𝑖 + 𝛽₆∗ 𝐶𝐸𝑂𝑇𝑁𝑅_𝑖 + 𝛽₇∗ 𝐵𝑅𝐷𝐶𝑀𝑃_𝑖+ 𝛽₈∗ 𝐼𝑁𝐷𝑆𝑇𝑅_𝑖 + 𝛽₉∗ 𝑌𝐸𝐴𝑅_𝑖 + 𝛽₁₀∗ 𝑀𝐸𝐷𝐼𝐴_𝑖 + 𝜀_𝑖

The last two dummy variables – YEAR and MEDIA – build on the investigation of the sensitivity analyses. That is, adding these two dummies explores how the CARs are affected by only the last two years of observations or the use of live-audience media.

4.4. Additional tests

Finally, some additional tests will be conducted to assess the relationship among CEO activism and different factors.

(14)

14 Cai et al. (2011) evaluate the CEO pay of different CEOs, before and after their respective companies engage in CSR. Similarly, in this study the average CEO pay of the two years before is compared to the average pay in the two years after the event occurs. The summary statistics and a test for mean equality will provide an adequate distinction between the pay around the event, which might lead to some inferences about how CEO activism affects their pay.

The relationship between CEO pay and CEO activism will be further empirically evaluated. The regression that will be applied is the following:

𝐶𝐸𝑂𝑝𝑎𝑦𝑖,𝑡 = 𝛼 + 𝛽 ∗ 𝐶𝐸𝑂𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 + 𝜙(𝑓𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠) +

𝜙(𝐶𝐸𝑂 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠) + 𝜀_𝑖𝑡 (7)

where, CEOpay is the salary of the CEO of each company for the years 2010-2017, CEO activism is a dummy variable that takes a value of 1 for every year after the event and firm and CEO characteristics are the same as in table 4.

Almeida et al. (2004) assess the cash to cash flow sensitivity by examining the ratio of cash holdings related to cash inflows of companies with different financial constraints. As a result, this ratio is calculated for all companies on an annual basis and the average ratio of two years before the event is compared to the average ratio of two years after the event. Again, any significant differences around the events might indicate that CEO activism is in part responsible for changes in how much cash companies keep as a buffer.

The relationship between cash to cash flow sensitivity and CEO activism will be empirically evaluated with the following regression:

𝐶𝑎𝑠ℎ𝐹𝑙𝑜𝑤𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦_𝑖,𝑡 = 𝛼 + 𝛽 ∗ 𝐶𝐸𝑂𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 + 𝜙(𝑓𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠) + 𝜀_𝑖𝑡 (8)

where, CashFlowSensitivity represents the ratio between cash and net cash flow of a company

for the years 2010-2017, CEO activism is a dummy variable that takes a value of 1 for every year after the event and the firm characteristics are the same as in table 4.

5. Results

5.1. Event study results

(15)

15 the third and fourth column illustrate the t-statistic of Student’s test and the test statistic from the generalized sign test, respectively.

Insert Table 6 here

Based on these results, there are no significant results when applying both the parametric and non-parametric tests. That is, the p-values corresponding to the test statistics of all tests are larger than 0.10 or any other conventional significance level. Therefore, the null hypothesis that the CEO announcements do not have an impact on the firm stock returns cannot be rejected.

Next, table 7 shows the results of the tests on the cumulative abnormal returns.

Insert Table 7 here

As illustrated in table 7, there are no statistically significant results when the cumulative average abnormal returns are tested with either the parametric or the non-parametric tests. Therefore, this result is in agreement with what was discovered before and the null hypothesis that CEO announcement do not have an impact on stock prices cannot be rejected at any significance level.

However, it is crucial to emphasize that the insignificant results from the event study do not suggest that CEO activism on social issues is completely irrelevant for the companies’ performance. It simply suggests that the stock returns of the companies are not affected by these announcements on the days around the event. As a result, one should investigate whether the companies are affected in a different way by implementing several sensitivity analyses and additional tests.

5.2. Sensitivity analyses

Table 8 illustrates the results from the first sensitivity analysis, which investigates whether there are any differences in the how the market reacts to CEO activism in different points of time. To do this, the sample is split into two and the oldest 4 years are compared to the most recent 2 years, followed by a test for mean equality.

(16)

16 The results of the sensitivity analysis are almost the same as for the overall event study with some noteworthy exceptions. Firstly, the average abnormal returns for day 2 for the most recent 35 events is significantly different from zero at the 0.05 and 0.1 significance levels. Therefore, the null hypothesis that CEO activism does not affect stock returns can be rejected on day 2 for the most recent events. This result is interesting and meaningful, as it suggests that the market has a more substantial reaction to the news of CEO supporting social issues in the recent years.

Finally, the test for mean equality shows that on the day before the event occurs there is a significant difference in the AARs between the two samples, at the 0.1 significance level. This result indicates that there is a statistically significant difference in the reaction of the stock returns to the CEO announcements on day -1. However, such a significant difference is not observed on any other day of the event window.

Next, table 9 illustrates the results of the second sensitivity analysis, for which we compare the AARs of events with indirect contact with the audience, such as interviews and public releases in journals or magazines (first subsample) and these with direct contact, such as public speeches (second subsample).

Insert Table 9 here

It can be seen that there are no significant results from the parametric tests, much like when the entire sample is evaluated. However, there are some significant results from the non-parametric tests. More specifically, on days -1 and +2 of the event window, there are significant results at the 0.1 level. Therefore, the null hypothesis of no impact on stock returns due to CEO activism can be rejected for this sample with some caution, given that the p-value is higher than 0.05 and the result cannot be rejected with a high certainty. For both of these days the AARs have positive signs, meaning that the CEO announcements that the people can observe directly have a positive impact on the stock price. However, these results are not observed for the other sample, for which the CEO announcements are indirect. This result is interesting, as it confirms that if one is present or knows someone else who is present during an announcement, this will lead to stronger reactions compared to the case when one simply reads about the announcement online.

5.3. Multiple regression model

(17)

17 Insert Table 10 here

As observed from Table 10, there are no significant results of the first regression, which does not incorporate the two dummies for the sensitivity analyses. Additionally, the F-statistic of the regression is equal to 0.939 and not statistically significant. Therefore, the null hypothesis that any of the independent variables do not affect the cumulative abnormal returns cannot be rejected at any conventional significance level. This finding suggests that the CARs of the event study do not vary according to specific firm or CEO characteristics, as one might expect.

However, when either of the two dummies is added to the model, some significant results are observed, as well as an increase in the R-squared. Firstly, when the YEAR dummy is added, a positive effect on the CARs is observed, statistically significant at the 0.05 level. Therefore, the result from the sensitivity analysis is confirmed, the most recent years seem to have a larger impact on the abnormal returns. Secondly, the MEDIA dummy also has a statistically significant positive impact on the CARs according to the regression. As a result, this confirms the finding that live-audience appearances by the CEOs have a larger effect on the market.

5.4. Additional tests

Now that the results of the event study, along with the sensitivity analyses, were discussed, it is intriguing to understand more about how CEO activism affects different aspects of the company’s operations.

Firstly, as was discussed in previous sections of this paper, the CEO pay of before and after the event will be compared. Table 11 represents the summary statistics of the 2-year average CEO pay for before and after the event. The observations from 2016 onwards have been excluded as there is not enough available data to calculate average CEO pay after the event. However, further research can be conducted once more data can be obtained for the most recent events.

Insert Table 11 here

(18)

18 conventional significance levels. This result is confirmed by other tests for mean equality. Therefore, the null hypothesis that the CEO pay remained the same before and after the event can be safely rejected. Consequently, the average salary of the CEO improved significantly after their respective announcements on various social issues. In line with the reasoning of Tirole and Benabou (2010), the event study discovered no positive significant results due to CEO initiated philanthropy. Yet, interestingly, there are significant positive differences between the CEO pay due to these announcements.

Next, the same analysis will be conducted with the inclusion of several control variables, namely size (assets), market-to-book, leverage, ROA and CEO tenure. This additional analysis aims to discover whether there are any differences in the CEO pay change around the event when the sample is controlled for these variables. Table 12 illustrates the results of this analysis. When controlling for every individual variable, the whole sample of events is split evenly and the summary statistics are presented for each subsample for before and after the event, similar to the previous analysis, followed by a test for mean equality.

(19)

19 differences in CEO pay around the CEO announcement. Finally, the control variable CEO tenure splits the sample in companies with short and long tenure of the company leader. Interestingly, only the subsample of short CEO tenure shows highly significant differences at the 0.01 level in CEO pay around the event. In contrast, there are no significant differences for the subsample of companies with high CEO tenure. Therefore, only CEOs who have recently taken the position as a company leader experience significant changes in their salary when they make an announcement on a social issue.

Table 13 illustrates the results from equation (7). As observed, CEO activism has a highly significant positive impact on the CEO pay. The coefficient of 0.210 suggests that when the CEOs engage in social issues, their salary is increased by 210 thousand dollars. When controlling for various firm and CEO characteristics, this result is magnified.

Finally, the second additional analysis that is conducted aims to investigate the differences in the cash to cash flow sensitivity ratios of the companies before and after the event. Therefore, as in the previous analysis, the before the event two-year average ratio is compared to the after the event two-year average ratio. Table 14 represents the summary statistics of this analysis along with a test of mean equality of the two samples. Extreme outliers have been excluded from the sample.

As observed in table 14, when examining the average cash to cash flow ratio, the higher ratio corresponds to the two years after the event has taken place. When observing the median, the opposite is true. In line with this mismatch, the test of equality does not yield any statistically significant results. Therefore, the null hypothesis that the means of the two samples are the same cannot be rejected at any conventional significance level. That is, this study does not find any association between cash to cash sensitivity and the CEO announcement.

(20)

20 Table 15 illustrates the results from equation (8). Evidently, there are no significant effects of CEO activism on the cash to cash flow sensitivity ratio, both with and without controls for firm characteristics. This analysis confirms the previous finding, that CEO activism does not affect the cost of capital of a company and consequently, the difficulty with which companies attract external funding.

6. Conclusion

This paper has investigated the impact of CEO announcements and public releases on social issues such as LGBT rights, gender equality, climate change and racial or religious discrimination. Overall, 115 public speeches, interviews or online posts between 2010 and 2018 have been selected for the sample, with all companies belonging to the S&P 500. The event study, which was conducted to search for any abnormal stock market returns due to these announcements, did not provide any statistically significant results, from both the parametric and non-parametric tests. This result is interesting on its own, as it suggests that for the most part, the investors do not seem to be influenced by these announcements.

Additionally, the sensitivity analyses provided some intriguing insights for the research. The first sensitivity analysis discovered that the market has a more substantial reaction to CEO activism only in the most recent years. Then, the second sensitivity analysis suggested that the media used for the announcements is important, as making the announcement in front of live audience seems to have a larger impact on the company’s stock price than a newspaper article or a post on social media. Further research is needed to discover whether this is a persistent trait for all CEO activism.

(21)

21 investigate whether the timing of these announcements coincides with any other major events for the company, such as a class action lawsuit.

A limitation of this study is the restricted number of events that are used. The CEO announcements that are investigated are a selected sample from the USA and over the last nine years. Therefore, a more complete international sample might provide more insight into the issue of CEO activism. In addition, the current method of selecting the events, by handpicking them one at a time, is not optimal. A more organized approach that uses a specific database would be more appropriate, so that no important events are omitted. However, to the best of our knowledge, such a database does not yet exist. Therefore, this is a potential area for further research.

(22)

22 References

Almeida, H, Campello, M & Weisbach, M.S. 2004. The Cash Flow Sensitivity of Cash. Journal

of Finance. 59(4), pp. 1777-1804.

Attig et al.. 2014. Corporate Legitimacy and Investment - Cash Flow Sensitivity. Journal of

Business Ethics. 121(2), pp. 297-314.

Benabou, R & Tirole, J. 2010. Individual and Corporate Social Responsibility. Economica. 77(305), pp. 1-19.

Cai, Y, Jo, H & Pan, C. 2011. Vice or Virtue? The Impact of Corporate Social Responsibility on Executive Compensation. Journal of Business Ethics. 104(2), pp. 159-173.

Campbell, C.J, Cowan, A.R & Salotti, V. 2010. Multi-Country Event Study Methods. Journal

of Banking & Finance. 34(12), pp. 3078-3090.

Chatterji, A.K, Levine, D.I & Toffel, M.W. 2009. How Well Do Social Ratings Actually Measure Corporate Social Responsibility?. Journal of Economics & Management Strategy. 18(1), pp. 125-169.

Chatterji, A.K & Toffel, M.W. 2015. Starbucks’ “Race Together” Campaign and the Upside of CEO Activism. Harvard Business School Working Paper.

Chatterji, A.K & Toffel, M.W. 2017. Assessing the Impact of CEO Activism. Harvard Business

School Working Paper. 16(100).

Collins, D. W., Rozeff, M. S. & Dhaliwal D. S. 1981. The Economic Determinants of the Market Reaction to Proposed Mandatory Accounting Changes in the Oil and Gas Industry: A Cross-Sectional Analysis. Journal of Accounting and Economics. 3(1), pp. 37-71.

Dimson, E, Karakas, O & Li, X. 2012. Active Ownership. Review of Financial Studies. 28, pp. 3225-3268.

Jo, H & Harjoto, M. 2011. Corporate Governance and CSR Nexus. Journal of Business Ethics. 100(1), pp. 45-67.

Khotari, S.P & Warner, J.B. 2006. Econometrics of Event Studies. Handbook of Corporate

Finance: Empirical Corporate Finance. A(1), pp. 2-52.

Khurana, I.K, Martin, X & Pereira, R. 2006. Financial Development and the Cash Flow Sensitivity of Cash. Journal of Financial and Quantitative Analysis. 41(4), pp. 787-808. Leftwich, R. 1981. Evidence of the Impact of Mandatory Changes in Accounting Principles on Corporate Loan Agreements. Journal of Accounting and Economics. 3(1), pp. 3-36.

MacKinlay, A.C. 1997. Event Studies in Economics and Finance. Journal of Economic

Literature. 35(1), pp. 13-39.

(23)

23 Margolis, J.D, Elfenbein, H.A & Walsh, J.P. 2009. Does it Pay to Be Good And Does it Matter? A Meta-Analysis of the Relationship between Corporate Social and Financial Performance. SSRN Electronic Journal. [Online]. 7(5), pp. 1-89. [3 May 2018]. Available from:

https://ssrn.com/abstract=1866371.

McGregor, J. 2017. “'Climate Change Is Real': CEOs Share Their Disappointment over Trump's Paris Accord Exit.” The Washington Post, WP Company. www.washingtonpost.com/news/on-

leadership/wp/2017/06/01/ceos-make-final-pleas-to-trump-to-stay-in-paris-climate-agreement/?utm_term=.ccd46786021a.

Milbourn, T. 2003. CEO Reputation and Stock Based Compensation. Journal of Financial

Economics. 68(2), pp. 232-262.

Sanghara, A. 2016. Barack Obama’s Speeches and Addresses: A Narrative and Framing Analysis. Ryerson University Research Paper. pp. 1-89.

(24)

24 Appendix

Appendix A

Table 1. A list of all events, including the CEO, the company’s name, the topic of the announcement, the date of the event and the media used.

No. CEO Company Topic: Date Media

1a _{2 3 4}

1 J. Lechleiter Eli Lilly 1

9-07-10 Press release

2 F. Blake Home Depot 1

3-06-11 Press release 3 H. Schultz Starbucks Corporation 1

4 L. Blankfein Goldman Sachs 1

5 K. Frazier Merck & Co. 1

6 D. McMillon Wal-Mart Stores 1

7 P. Dauman Viacom 1

8 J. Bezos Amazon 1

27-07-12 Press release 9 P. Otellini Intel Corporation 1

10 B. Gates Microsoft 1

11 W. Buffett Berkshire Hathaway 1

28-12-12 Press conference 12 M. Erdoes JP Morgan Chase & Co 1

25-01-13 Press conference

13 K. Powell General Mills 1

28-01-13 Press release 14 H. Schultz Starbucks Corporation 1

15 R. Lance ConocoPhillips 1

16 J.Lechleiter Eli Lilly 1

17 L. Merlo CVS Health 1

5-02-14 Press release 18 C. Nassetta Hilton Worldwide Holdings 1

19 M. Hewson Lockheed Martin 1

20 L.McAdam Verizon Communications 1

21 M. Kent Coca-Cola 1

29-05-14 Interview 22 A. Gorsky Johnson & Johnson 1

23 M. Barra General Motors 1

26-06-14 Interview

24 I. Nooyi PepsiCo 1

19-09-14 Press release 25 D. Simon Simon Property Group, Inc. 1

26 M. Corbat Citigroup 1

(25)

25

27 M. Hewson Lockheed Martin 1

28 M. Benioff Salesforce.com 1

29 T. Cook Apple 1

30 J. Lechleiter Eli Lilly 1

31 M. Parker Nike 1

31-03-15 Press release 32 D. McMillon Wal-Mart Stores 1

1-04-15 Press release 33 J. Immelt General Electric 1

1-04-15 Press release 34 W. Buffett Berkshire Hathaway 1

3-04-15 Press release 35 L. McAdam Verizon Communications 1

36 J. Gorman Morgan Stanley 1

37 R. Goings Tupperware Brands 1

38 D. Reed Sempra Energy 1

39 A. Peck Gap Inc. 1

25-06-15 Press release 40 D. Parker American Airlines Group 1

26-06-15 Press conference 41 S. Mollenkopf Qualcomm Inc. 1

20-07-15 Press release 42 A. Gorsky Johnson & Johnson 1

43 S. Catz Oracle 1

44 H. Grant Monsanto 1

1-12-15 Press release 45 G. Hayes United Technologies Corp 1

46 D. McMillon Wal-Mart Stores 1

47 B. Carrigan Dun & Bradstreet 1

21-01-16 Press conference 48 D. Tarman eBay 1 21-01-16 Press conference 49 M.Zuckerberg Facebook 1 21-01-16 Press conference 50 C. Crane Exelon 1 25-02-16 Press release

16-03-16 Press release 52 B. Krzanich Intel Corporation 1

17-03-16 Twitter

1-04-16 Press release 54 D. Schulman PayPal Holdings 1

55 T. Kennedy Raytheon Company 1

56 B. Cornell Target Corp. 1

57 W. Buffett Berkshire Hathaway 1

58 L. Merlo CVS Health 1

59 R. Stephenson AT&T Inc. 1

(26)

26

60 O. Ishrak Medtronic 1

61 M. Kent Coca-Cola 1

2-12-16 Press conference 62 B. Moynihan Bank of America 1

63 M.Zuckerberg Facebook 1

27-01-17 Facebook

28-01-17 Twitter

66 J.Weiner Microsoft 1

28-01-17 Twitter

67 M. Barra General Motors 1

28-01-17 Twitter

70 J. Immelt General Electric 1

71 M. Parker Nike 1

72 S. Kaufer Trip Advisor 1

29-01-17 Twitter

73 T. Cook Apple 1

74 J. Murdoch 21st Century Fox 1

75 L. Blankfein Goldman Sachs 1

76 M. Fields Ford Motor 1

77 H. Grant Monsanto 1

78 J. Haley Willis Towers Watson 1

79 B. Krzanich Intel Corporation 1

8-03-17 Twitter

80 I. Thulin 3M Company 1

81 S. Wojcicki Alphabet 1

16-03-17 Press release 82 K. Johnson Starbucks Corporation 1 1

83 G. Hayes United Technologies Corp 1

84 J. Bewkes Time Warner Inc. 1 1

15-05-17 Press conference 85 M. Thompson New York Times Company 1 1

15-05-17 Press conference 86 M. Rothblatt United Therapeutics 1 1

87 P. Kibsgaard Schlumberger 1 1

88 D. Woods Exxon Mobil Corp 1

26-05-17 Press release 89 R. Iger The Walt Disney Company 1

1-06-17 Twitter 90 V. Hollub Occidental Petroleum Corp. 1

91 D. Taylor Procter & Gamble Co 1

92 P. Nanterme Accenture 1

93 L. Good Duke Energy 1

25-07-17 Press release 94 D. Abney United Parcel Service Inc. 1

(27)

27

95 H. Joly Best Buy Co. Inc. 1

96 B. Cornell Target Corporation 1

98 J. Dimon JP Morgan Chase & Co 1

99 K. Frazier Merck & Co. 1 1

14-08-17 Press release 100 C. Robbins Cisco 1 1 20-08-17 Interview 101 C.r Crane Exelon 1 20-09-17 Press release 102 L. Fink BlackRock 1 20-09-17 Press conference 103 S. Nadella Microsoft 1 27-09-17 Press conference 104 S. Catz Oracle 1 27-10-17 Press conference

105 K. Chenault American Express 1 1

9-11-17 Press conference 106 D. Parker American Airline Group 1

107 J. Hackett Ford Motor 1

1-01-18 Twitter

109 M. Corbat Citigroup 1

110 S. Narayen Adobe Systems Inc. 1

111 B. Moynihan Bank of America 1

23-01-18 Press conference 112 C. Nassetta Hilton Worldwide Holdings 1

26-01-18 Interview

113 T. Sloan Wells Fargo 1

114 G. Rometty IBM 1

12-02-18 Press conference 115 L. McAdam Verizon Communications 1

2-03-18 Interview

a_{Topics 1, 2, 3 and 4 refer to LGBT rights, climate change, racial or religious discrimination and gender}

(28)

28 Appendix B

Table 2. Summary statistics on stock returns in the event window. Days of event window -1 0 1 2 Mean 0.002 -0.001 0.000 -0.002 Median 0.001 -0.001 -0.001 0.000 Maximum 0.040 0.076 0.039 0.059 Minimum -0.023 -0.056 -0.042 -0.098 Standard deviation 0.009 0.014 0.012 0.018

(29)

29 Table 4. Explanation of all regression variables.

CARi Cumulative abnormal (daily) returns for ith company

SALESi Net sales of the previous year, calculated as the gross sales and other

operating revenue less discounts, returns and allowances of the company. The natural logarithm of net sales is taken.

TOTASSETSi Total assets of the ith company, calculated as the sum of total current

assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets. The natural logarithm of total assets is taken.

ROAi The return on assets of the ith company, calculated as the ratio of net

income (before extraordinary items) and total assets.

MTBi Market-to-Book ratio for the ith company, calculated as the ratio of the

company’s market capitalisation and its common equity.

LVRGi Leverage of the ith company, calculated as the ratio of debt and equity or

more specifically, as (Long Term Debt + Short Term Debt & Current Portion of Long Term Debt) / Common Equity * 100

CEOTNRi The CEO tenure for the ith company, calculated as the number of years

since the appointment of the current CEO.

BRDCMPi Board composition for the ith company, calculated as the proportion of

women in each board of directors.

INDSTRi An industry variable which is represented by a 1/0 dummy variable. It

takes a value of 1 if the ith company is a business to business firm and 0, otherwise.

YEARSi A time variable which is represented by a 1/0 dummy variable. It takes a

value of 1 for all events that occurred in the last 2 years and 0, otherwise.

MEDIAi A media used variable which is represented by a 1/0 dummy variable. It

(30)

30 Table 5. Summary statistics – Variables in regression model. 115 observations.

CARi (%) SALESi TOTASSETSi ROAi (%) MTBi LVRGi(%)

Median 0.138 17.504 16.662 7.447 3.845 62.045

Mean 0.140 17.300 16.646 7.646 7.678 150.885

Standard

Deviation 2.063 1.385 1.343 7.306 24.186 415.114

CEOTNRi BRDCMPi

INDUSTRYi YEARSi MEDIAi

Yes No Yes No Yes No

Median 3.0 0.00 69 46 53 62 34 81

Mean 5.4 0.08 - - - -

Standard

(31)

31 Appendix C

Table 6. Results of parametric and non-parametric tests on average abnormal returns. 115 observations. Day of event window AAR in event window t-statistic parametric test

Test statistic non-parametric test

-1 0.001 0.095 0.746

0 -0.001 -0.066 0.560

1 -0.000 -0.041 0.373

2 -0.001 -0.087 0.373

*, ** and *** represent significant results at the 0.1, 0.05 and 0.01 levels, respectively. Table 7. Results of parametric and non-parametric tests on cumulative average abnormal returns. 115 observations.

Event window CAAR in event window

t-statistic parametric test

Test statistic non-parametric test

[-1;2] -0.001 -0.082 0.000

[0;1] -0.001 -0.104 0.000

(32)

32 Table 8. Results of first sensitivity analysis, including AARs for oldest and most recent 2 years, t-statistic of parametric tests, test-t-statistic of non-parametric tests and test-t-statistic of test for mean equality between the two samples. 77 observations.

Days AAR of oldest 4 years t-statistic parametric test Test statistic non-parametric test AAR of most recent 2 years t-statistic parametric test Test statistic non-parametric test Test statistic of test for mean equality -1 -0.001 -0.133 1.014 0.003 0.263 1.690 -1.922* 0 0.002 0.176 0.338 -0.001 -0.086 0.338 1.132 1 -0.002 -0.247 1.014 -0.002 -0.164 0.676 -0.446 2 -0.001 -0.068 0.676 0.001 0.070 2.366** -0.529

*, ** and *** represent significant results at the 0.1, 0.05 and 0.01 levels, respectively.

Table 9. Results of second sensitivity analysis, including AARs for different announcement methods, t-statistic of parametric tests, test-statistic of non-parametric tests and test-statistic of test for mean equality between the two samples. 115 observations.

Days AAR for interviews, public releases t-statistic parametric test Test statistic non-parametric test AAR for public speeches t-statistic parametric test Test statistic non-parametric test Test statistic of test for mean equality -1 0.000 0,038 0.222 0.003 0.266 1.886* -1.401 0 0.000 0.014 0.667 0.005 0.411 0.174 -1.483 1 0.000 0.017 0.222 -0.002 -0.172 1.200 1.129 2 -0.002 -0.171 0.667 0.001 0.115 1.886* -1.074

(33)

33 Table 10. Cross-sectional regression estimates with CARi as the dependent variable. 115

observations. Coefficients and standard errors (in brackets) are provided.

Model 1a Model 2 Model 3

Constant 0.026 (0.026) 0.005 (0.024) 0.021 (0.021) SALES -0.000 (0.003) -0.002 (0.004) -0.004 (0.004) TOTASSETS -0.002 (0.002) 0.001 (0.004) 0.002 (0.004) ROA 0.021 (0.030) 0.030 (0.023) 0.026 (0.020) MTB -0.000 (0.000) 0.000 (0.000) 0.000 (0.000) LVRG -0.000 (0.000) -0.000 (0.000) -0.000 (0.000) CEOTNR 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) BRDCMP -0.006 (0.015) -0.006 (0.012) -0.005 (0.012) INDSTR 0.003 (0.004) 0.007 (0.005) 0.007 (0.005) YEAR - 0.009** (0.005) - MEDIA - - 0.012*** (0.003) R2 0.075 0.114 0.151 F-statistic 0.939 1.305 2.078**

a_{Model 1, 2 and 3 refer to the model without dummies for year and media, with a dummy for}

year and with a dummy for media, respectively.

(34)

34 Table 11. Summary statistics of 2-year average CEO pay before and after the event (in 10000’s USD). 41a_{observations.}

Before the event After the event

Mean 1245.385 1477.693 Median 1285.258 1500.000 Maximum 3608.334 3800.000 Minimum 81.840 81.840 Standard Deviation 611.504 709.557 Jarque-Bera 64.402 34.560 Probability 0.000 0.000

Test for equality

(Wilcoxon) 2.917***

a_{The number of events in the period 2010-2015. The rest of the events are omitted due to inadequate}

(35)

Appendix B

Table 12. Summary statistics of 2-year average CEO pay before and after the event (in 10000’s USD) in subsamples based on five control variables, namely size (assets), market-to-book, leverage, ROA and CEO tenure. *, ** and *** represent significant results at the 0.1, 0.05 and 0.01 levels, respectively.

Size Market-to-Book Leverage

Small Large Small Large Small Large

Before After Before After Before After Before After Before After Before After

(36)

36 Table 12 continued.

ROA CEO Tenure

Small Large Small Large

Before After Before After Before After Before After

(37)

Table 13. Regression of CEO activism on CEO pay, a panel regression with fixed firm effects. Dependent variable is CEO pay and independent variable is/are CEO activism dummy variable (1) or CEO activism dummy variable and various control variables (2).

(1) No other control variables (2) Other control variables CEO activism 0.210*** (0.120) 0.401*** (0.120) F-statistic 14.905*** 16.438*** R2 0.730 0.762 Number of observations 319 319

Table 14. Summary statistics of 2-year average cash to cash flow sensitivity before and after the event. 41a observations.

Before the event After the event

Mean 0.726 0.939 Median 0.695 0.607 Maximum 3.176 8.979 Minimum 0.009 0.009 Standard Deviation 0.606 1.453 Jarque-Bera 69.216 903.360 Probability 0.000 0.000

Test for equality

(Wilcoxon) 0.079

a _{The number of events in the period 2010-2015. The rest of the events are omitted due to inadequate}

(38)

38 Table 15. Regression of CEO activism on cash flow sensitivity of cash, a panel regression. Dependent variable is cash flow sensitivity of cash and independent variable is/are CEO activism dummy variable (1) or CEO activism dummy variable and various control variables (2). (1) No other control variables (2) Other control variables CEO activism -1.296 (1.575) -1.671 (1.809) F-statistic 0.677 1.066 R2 0.001 0.145 Number of observations 508 508