The influence of the Global 100 list on stock returns over time

The Influence of the Global 100 list

on Stock Returns over time

by

Pascalle Veltman

June, 2018

Supervisor:

_Güler

Statement of Originality

This document is written by Student Pascalle Veltman who declares to take full re-sponsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

The worldwide growing awareness on environmental issues not only encourages indi-viduals to go green, it also causes a sustainability movement among companies. The Global 100 list is a yearly ranking of the most sustainable and financially successful companies. This thesis aims to investigate if investors prefer to make money out of sustainable stocks as a result of their increased knowledge of the consequences of envi-ronmental pollution. To do this, the influence of the Global 100 list on the values of companies is compared between 2006 and 2018. The test results, obtained from event study, claim that the stock return of companies included in the list is generally higher than the stock return of companies excluded from the list that same year. Also, the contrast in performance between excluded and included companies seems to decrease over the year. However, the specific significant influence of the Global 100 list on the stock return remains unclear.

Contents

1

Introduction

1

2

Theoretical Framework

3

2.1 Global 100 . . . 3

2.2 Socially Responsible Investment . . . 4

2.3 Environmental Awareness over Time . . . 5

2.4 Existing Empirical Research . . . 6

2.5 Summary and Hypotheses . . . 7

3

Method

9

3.1.1 Event Time Line . . . 9

3.1.2 The Abnormal return . . . 11

3.1.3 Models to calculate the Expected Return . . . 12

3.2 Test for Normality . . . 15

3.3 Testing the CAR . . . 16

4

Data

20

4.2 Factor Data . . . 21

5

Results

23

5.2 Test for Normality . . . 24

5.3 Student t-test . . . 25

Chapter 1

1

Introduction

Over the last decades there has been an increasing awareness regarding environmen-tal issues. Technological advancements, both in academia and communication, have spurred numerous studies into the future of the planet and has raised the concern of the public regarding global health (Dunlap Jorgenson, 2012). The worldwide corporate industry plays an important role in environmental pollution. Many companies still do not take their environmental footprint into account and merely focus on minimizing costs. The pressure on companies to report their social and environmental perfor-mance, mainly from stockholders, has risen (Ballou, Heitger, Landes Adams, 2006). The last ten years there has been a positive change in investments in corporate social responsibility (Flammer, 2017).

Corporate Knights compiled the Global 100 list in 2005 for the first time. This list includes the most sustainable large corporations in the world. It gives investors insight into which companies use their invested money in an eco-friendly way. Reaching a position in the Global 100 can be attractive for investors who are concerned about environmental issues, and therefore may increase the stock value of the firm.

Many studies have investigated the connection between corporate social re-sponsibility and the stock value of a firm. Cheung (2011) for instance shows that an inclusion in the Dow-Jones sustainability Index has a short term positive influence on the stock value of the corporation. This reaffirms the growing trend of environmental awareness under investors. However, results of research done by Bamberg (2003) claim that the growing concern on environmental issues has hardly encouraged people to take action, such as investing in more sustainable corporations, at all. Furthermore, the public is getting more skeptical about environmental issues (Whitmarsh, 2011). This combined with the fact that satisfying sustainability restrictions generally increase costs, makes corporate social responsibility (CSR) not seem attractive for companies (Geczy,

Sambaugh, Levin, 2005).

This paper analyses the influence of growing knowledge of environmental issues over time on the sustainability of investments. To investigate this, it is important to determine the influence of a company reaching a position in the Global 100 list to the stock value of that company. The comparison of this influence between the year 2006, the year after the list was first published, and 2018, shows how this effect has changed over time. The level of impact of an event such as inclusion in the Global 100, can be determined by event studies.

The results of this paper contribute to existing findings because it clarifies the relation between the sustainable ranking and the share value of a company. If investors are getting more interested in the environment and sustainability of companies, CSR will increase the value of the company. However, if there is a negative trend in the influence of sustainability on the choice of investors, then the costs CSR brings may not be worth it.

This thesis is organized as follows: the next part, chapter 2, reviews the theory behind the influence of the Global 100 on stock values. Chapter 3 explains the method-ology used in this paper. The data and data sources are described in Chapter 4 and Chapter 5 analyzes of the obtained results. Finally, Chapter 6 provides a conclusion and discussion of this research.

Chapter 2

2

Theoretical Framework

This chapter gives an overview of all existing literature regarding the influence of the Global 100 list on stock returns over time. The first part is about the definition and compilation of the Global 100 list. The second part explains the extensive meaning of socially responsible investment and the third part discusses existing theories about the connection between growing environmental awareness and SRI over the past decades. The fourth part describes all important existing researches on subjects comparable to the research question of this thesis. Finally, the fifth part summaries the theories and composes a hypothesis.

2.1

Global 100

The Global 100 list of the most sustainable companies worldwide is compiled in 2005 by Corporate Knights for the first time. Corporate Knights is a Canadian research and investment advisory organization which constructs ratings using their research results on the sustainable behaviour of companies. Furthermore, Corporate Knights seasonally publishes a society and business magazine, called Corporate Knights. Their main focus is information symmetry between investors and companies about their ecological, social and economic performance (Staff, 2015a).

Determining the level of sustainability of a company is a complicated task. To compose the Global 100 list, Corporate Knights has decided to investigate companies of consideration on four different manners starting each year at the first of October (Staff, 2015b). Every company worldwide with a market share of at least 2 billion dollars, is selected for the screening process. The first screen is about the disclosure of the sustainability of a company. To succeed, at least 75% of the twelve existing key performance indicators is required. Key performance indicators are criteria distributing

social and environmental health standards (Staff, 2014). The second screen consists of several individual tests about the liquidity of the company. The social responsibility of the product category is evaluated in the third part of the screening. In this part sin stocks, for example Tobacco companies, are eliminated. The fourth screen is about the amount of sustainable-related payments done by a company. The corporations which emerge through all four criteria, are included in the Global 100.

2.2

Socially Responsible Investment

The content of the Global 100 list is likely to be of great interest to investors who aim to make money out of sustainable stocks, a process known as socially responsible investment. The principle of socially responsible investment has an extensive meaning, a history and different categories. The portfolio of a socially responsible investor results from ecological, religious and social considerations as well as financial performances (Kurtz, 2008, p. 250). All these factors of a company are part of the personal screening process of an investor. Many social responsible investors aim to make a difference in the social behavior of companies. Therefore, Kurtz (2008) finds that they generally attach value to a good communication and relation between the investor and the management. SRI is a long-used term and has known many different definitions. Kinder (2005) has done research on the development of the Modern Socially Responsible In-vestment. According to his theory, it started with the concept of shareholders activism by Alinsky and Kodak in 1966 in US together with an increasing influence of religious and moral motivations on investment decisions. In the 1970‘s, the concept of portfolio screening was added to that principle (Kinder, 2005, pp. 8, 9). The growing part of institutional investors and their criticism on social and environmental aspects, has led to the meaning of socially responsible investment as it is today.

Kinder classified three different types of Socially responsible investments based on the named motivations. First of all, there is “Value-based SRI”. This implicates that a portfolio is chosen most importantly driven by moral and religious

character-University of Amsterdam

istics. The second type seeks to find environmental and social aspects that enhance the financial performance of their investments, this is called “Value-seeking SRI”. The third and last type of investor according to Kinder, is the “Value-enhancing SRI” with increasing corporate financial performance and achieving corporate governance through shareholders activism as their greatest interest.

Socially responsible investment does not always have a strong connection with corporate social responsibility, Kurtz (2008) claims, where he defines CSR as operating in a financial focused way with respect to social and environmental aspects. Although the two concepts have much in common, the causality between SRI and CSR is not clearly determined.

2.3

Environmental Awareness over Time

As mentioned, corporate social responsibility does not always directly lead to socially responsible investment. Investors do not automatically follow the sustainability trend of companies. Instead, they are more influenced by personal interests (Kurtz, 2008). It is difficult to determine the development of the public opinion on environmental issues. Because of the growing knowledge of this subject, people not only get more concerned, they also get more critical.

From 1960 the consciousness on environmental issues began to grow (Dunlap Jorgenson, 2012). Starting in the US, people became more critical on industrialization and pollution of water and air. After that, a wide range of concerns about climate risks was rapidly spread worldwide along with the scientific agreement of the human behaviour as a main cause (Doran Zimmerman, 2011). This included matters such as global warming, deforestation, use of pesticides, decreasing animal population and increasing human population. After the year 2000, modern technological research and mass media resulted in even more knowledge about social and environmental issues among the public.

and the human part in that (Whitmarsh, 2011). Whitmarsh shows that more and more people are uncertain about the reality and proof of the actual problem. The main part believes that the problem is exaggerated, but a minority part even rejects the whole concept of global warming. According to Whitmarsh, this skepticism is motivated by personal political and environmental beliefs rather than academic arguments.

2.4

Existing Empirical Research

Aside from contradicting theories about human environmental interests development, there are also statements based on empirical results. The connection between socially responsible investment, corporate social responsibility and the stock value of a firm is a well-known subject among existing research. This part firstly discusses past studies about the general correlation between socially responsible investment, corporate social responsibility and the financial performance of a company. After that, the results of specific research on the effect of a sustainable ranking position on the stock value of a firm are given.

Many past studies have shown a general correlation between the sustainable performance and the return of a company. Curto and Vital (2014) for example, in-vestigated stocks of 4 sustainable indexes and stocks of 10 conventional indexes. They concluded that sustainable indexes generally do better than traditional indexes during the complete research period from 2001 until 2011, even though there was no significant proof of sustainable indexes obtaining higher mean returns. Orlitzky (2003) found that corporate social responsibility positively influences the corporate financial performance as well. King and Lenox (2008) confirm, in addition to results mentioned right before, a negative correlation between pollution and financial return of a company regarding to their research over the period time 1987-1996. However, ethical companies in Aus-tralia tend to underperform during the period 1986-2005 (Frost, Jones, Van der Laan, Loftus, 2007). Also, Stentröm and Thorell (2007) have determined that regular funds outperform SRI funds based on their results obtained from 2001 until 2007.

University of Amsterdam

Besides proof of correlation between SRI and stock-values, there are also re-searches specifically regarding the effect of a position on a sustainability index (such as Global 100) on the stock value of a company. For instance, Murgaia and Lence (2014) studied the influence of the announcement of the Global 100 ranking on the value of an equal weighted portfolio and relative stock prices. They find a positive significant effect on relative stock values, but not on the value of the equal weighted portfolio. Remarkably, their results confirm that reaching one position higher in the Global 100 list, increases the firm value by an average of 11 million dollars. In 2011, Cheung studied the influence of inclusion and exclusion on the Dow Jones Sustainability Index on the value of the firm from 2002 until the year 2008. The DJSI, launched in 1999, contains the most suitable companies for sustainable investment. Cheung concludes from his results that the stock value and liquidity increases temporarily but significant when the firm is added to the list, as well as the stock value decreases for a short pe-riod of time when the firm is deleted from the Dow Jones Sustainability Index. Tsai (2007) however, claims in his research that only a sustainability index exclusion has a significant negative influence on the stock value, but there is no significant proof of inclusions resulting in higher stock values. Additionally, Glushkov and Statman (2008) state that return increase of positive screening does not weight up against the return decrease of negative screening a company. Karlsson and Chakarova (2008) conclude a non-existing significant correlation between inclusion or exclusion from sustainable indexes on abnormal returns, based on their sample research.

2.5

Summary and Hypotheses

Previous theoretical and empirical researches draw contradicting conclusions on the connection between environmental awareness, socially responsible investment and the stock value of a firm. While the public awareness on climate problems has grown over the last thirteen years, the skeptical attitude is spreading at the same time. And even though many researches claim a positive influence of socially responsible investment on

stock values of corporate responsible firms, there are still studies with contradicting results or without any significant proof of correlation at all.

This thesis specifically investigates the difference in effect of a G100 position between 2006 and 2018. The results sustain and clarify existing theories on significant, insignificant, positive or negative consequences of the G100 list. Therefore, contribution of this research is important.

To construct a hypothesis for this thesis, all existing studies are carefully eval-uated based on validity, observation period and date of publish. Several studies claim that sustainable indexes effect the share value. However, the claims of these studies are questionable because they provide insignificant results and contradict other researches. Regarding the varying results in existing literature, in this study it is expected that the stock value is not significantly influenced by a position on the Global 100 list in 2006 or 2018. This means the decision of investors is not influenced by the sustainability of companies, and the excess return of a company does not change after being included in or excluded from the Global 100 list. In this case, the abnormal return, the difference between the expected return and actual return (see Chapter 3.1.2), is predicted to be zero in 2006 and in 2018. Based on these predictions, the null-hypothesis is formed as follows:

Chapter 3

3

Method

This chapter describes the method used to determine the impact of a position on the Global 100 list on stock return and if that impact is significantly different between the year 2006 and 2018. Event study is a commonly used, powerful and suitable method for the investigation of this kind of research question. The first part explains how an event study is done using the Fama and French four-factor model. The second part tests the validity normality assumption of the abnormal return. The third part tests the influence of the G100 over time.

3.1

Event studies

Event study is an extensively used method in existing literature to analyze the exact effect of an unexpected event on the market (Sitthipongpanich, 2011). MacKinley (1997) explains in his study that the effect of an unforeseen event will be reflected directly on the value of company shares if the market circumstances remain unchanged (ceteris paribus). Event study is important because it determines the effect of any event on the stock return. The results of an event study, a firm can decide to change their policies in order to avoid unexpected share decrease.

3.1.1 Event Time Line

First of all, the event and exact date of the event must be noted. Every year around January 22nd, the Global 100 list is released. This list demonstrates what companies are the most successful, sustainable and socially responsible at the same time. It is assumable that sustainable investors are interested in the Global 100 list and may even let their investment decision depend on it. The release of the Global 100 list can be seen as an event which influences the market through investors. Despite the fact that the

Global 100 publish date is approximately the same every year, the content of the list is unknown until Corporate Knight publishes it. To companies and investors, the event of ‘knowing what companies are on the G100 list‘ is unforeseen and therefore the influence of the G100 list can be analyzed via the method of event study. The announcement day (AD) of the event study is in this case January 22nd both for the year 2006 and 2018.

Secondly, the time line of the event study has to be established. This is the period before and after the announcement day which is analyzed to determine the effect of the event. The time line is divided into an estimation period and a test period. The decisions on the exact range of the time line and the length of the different periods specified within this time line, are based on the approach of Sitthipongpanich (2011). The test period, also called event window, is investigated to evaluate the impact of the event on the stock returns. The expected daily return of the event window is calculated using data of the estimation period. The estimation period runs from 120 trading days, approximately 6 months, before AD (T0) until 5 trading days before AD (T1). The

last 5 trading days before the announcement day are not included because of possible information leaks which are likely to manipulate the outcome (MacKinley, 1997). The length will be as follows:

L1 = T1− T0 = −5 + 120 = 115 (2)

The event window runs from a minimum 5 trading days before AD (T1) until a maximum

20 trading days, approximately 1 month, after AD (T2). The AD is identified as τ =0

and is also included in this test period. The length of the test period in this case is defined as:

L2 = T2 − T1+ 1 = 20 + 5 + 1 = 26 (3)

Three different windows within the test period will be investigated to examine a possible change in short-term and long-term influences on the stock value. The length of the various limited event windows investigated in this study are based on research of Cheung (2010) on the influence of the Dow Jones Sustainability world index on share values

University of Amsterdam

of companies. The first window 3-day event window (AD-1, AD+1) and the second 11-day event window (AD-5, AD+5) are specified in order to investigate the short-term influence on the stock return of a position in the Global 100. The last window investigates the complete test period (AD-5, AD+20) concerning the long-time influence of the event.

Figure 1: Time Line

3.1.2 The Abnormal return

After identifying the exact event time line, the obtained abnormal return over the test period can be calculated (MacKinlay, 1997). The formulas used in this chapter are based on research done by MacKinley (1997) and Sitthipongpanich (2011). Event study focuses on the change in stock value before and after the day of the event. This change is expressed as the difference between the actual excess stock return and the expected stock return, resulting into the abnormal return. The abnormal return (AR) is defined as follows:

ARi,τ = Ri,τ − E(Ri,τ|Xτ) (4)

In this formula, AR stands for the abnormal return of company i on day τ of the test period. The Ri,τ stands for the actual excess return of company i on day τ and

E(Ri,τ|Xτ) expresses the expected excess return given the information obtained during

the estimation period. When speaking of return, this research always implies excess return. The excess return is the individual return minus the risk-free rate. This means that the excess return of a company is the obtained return directly resulting from stocks of that specific company. The actual excess return of company i on day τ is determined

by calculating the difference in percentage of the stock closing price and the day before.

Ri,τ =

Pi,τ − Pi,τ −1

Pi,τ −1

− rf,τ = ri,τ − rf,τ (5)

3.1.3 Models to calculate the Expected Return

A common way to calculate the conditional expected return of a company is by using the Capital Asset Pricing Model (CAPM) (Sitthipongpanih, 2011). According to the CAPM, the expected excess return of a company at moment τ is a linear expression of the market return at τ . The linear relation is defined by and β resulting from ordinary least squares (OLS) regression over the estimation period. These α and β are used to forecast the excess return on day τ of the test period. The expected return of company i on time τ according to the CAPM is:

E(Ri,|Xτ) = ri,τ − rf,τ = αi+ βi(rm,τ − rf,τ) + εi (6)

There are many different models besides the CAPM to calculate the conditional expected excess return of a company. In this study, the Fama and French-four factor model is used. This model expands the CAPM model by two Fama and French (1993) factors and one factor created by Carhart (1997). The first two factors were added by Fama and French to compensate the model for large differences in stock sizes and categories. Often, large capitalization stocks are outperformed by small capitalization stocks. The first Fama and French factor, the Small minus Big (SM B) factor, corrects the model for this situation. The formula of this factor stands for the difference between the average return of three small portfolios and the average return of three big portfolios.

SM B = 1

3(smallvalue + smallneutral + smallgrowth) − 1

3(bigvalue + bigneutral + biggrowth)

(7)

Also, growth stocks are often outperformed by value-stocks. For this appearance the second factor, the High minus Low (HM L) factor, is created. Fama and French (1993)

University of Amsterdam

define this factor as the average return on two portfolios of value stocks and two port-folios of growth stocks.

HM L = 1

2(smallvalue + bigvalue) − 1

2(smallgrowth + biggrowth)

(8)

The third factor is the momentum, added to the Fama an French model by Carhart (1997). This factor is based on the tendency of continually falling stock returns or continually rising stock returns. Carhart as well as Jegadeesh and Titman (2001) state in their research, that stocks with positive performance over the past 6 months are likely to increase the return of investors. In the same way, stocks with a bad past performance are more likely to have a bad influence on the future return of investors. To correct the model for the sensibility to strong shares that keep outperforming weak shares, Carhart (1997) introduced the M OM factor. M OM is defined as the obtained excess returns of rising stocks minus the lost excess returns of falling stocks. Adding the SM B, HM L and M OM factors to the CAPM model, results in the Fama and French four-factor model. This model defines the expected excess return as follows:

E(Ri,τ|Xτ) = αi+ βiM RM Rτ+ βiSM BSM Bτ + βiHM LHM Lτ + βiM OMM OMτ+ εi (9)

Where

M Rτ = rm,τ − rf,τ (10)

The model is corrected for potential advantaged and disadvantaged shares, and for that reason the Fama and French four-factor model is a preferred predictor. The first term, α, is a constant. The second term is the sensitivity to the market fluctuations (βM R_{) times the market excess return (M R) of that day. The third term describes the}

sensitivity of company i to the changes in small stocks (βSM B) times the small stock return minus the big stock return on time τ (SM B). The fourth term is the sensitivity of company i to the value stock movements (βHM L) times the the difference between

value stocks and growth stock on time τ . The fifth term is the sensitivity to movements in rising and falling stocks (βM OM) times the difference between rising and falling stocks (M OM ). The α and β‘s are obtained by performing OLS on the excess return values of the estimation period. εi is the standard error of the expected return of company i.

The standard errors, εi are saved for the derivation of the possibility distribution later

in this chapter.

The abnormal return of company i on day τ is given by the actual excess return minus the expected return according to the Fama and French four-factor model. The sum of the abnormal returns of company i over all days of the test period, result in the cumulative abnormal return (CAR). Aggregation of the abnormal return is important in order to draw general conclusions on the effect of the announcement day. In this case there are two dimensions in which the abnormal return can be aggregated: over time (τ ) and over the companies (i). The formula below defines the CAR which adds the abnormal returns from time T1 until time T2:

CARi(T1, T2) = T2

X

τ =T1

ARi,τ (11)

The cumulative abnormal returns show the deviating return of a company during the test window, after and right before the day of publishing the Global 100 list. The mean cumulative abnormal return is the mean over all N companies when investigating a certain sample. This study defines four different samples regarding four different situations. The first sample are the companies losing their position in the Global 100 list of 2006. The second sample contains the companies reaching a position in the G100 in 2006. The third sample are the companies excluded from the G100 in 2018 and the last sample includes companies included in the G100 in 2018. The mean CAR is a value highly informative of the sample and the expected effect of the event on the corresponding situations. The mean cumulative abnormal return is defined as follows:

CART1,T2 = 1 N N X i=1 CARi(T1, T2) (12)

University of Amsterdam

3.2

Test for Normality

Many tests of event study are based on the CAR. In these tests, a student t-test for instance, it is assumed that the AR and CAR are normally distributed. However, MacKinley (1997) emphasises that the assumption of a normally distributed abnormal return is not always well-specified. Because the conclusions of the t-test in this study depend on the normality assumption, it is important that the validity this assumption is tested. The normality of the CAR of the samples in this thesis is tested by the Shapiro-Wilk test, presented by Shapiro and Wilk in 1965. First, a vector is created for each sample containing the aggregated AR of all companies at day τ of the event window with length L2. Under the null-hypotheses, the values of this vector correspond

to the expected normal distributed values. In this case, the sample follows a normal distribution. The null-hypothesis and the alternative hypothesis are formed as follows:

H0 : CARτ ∼ N (µ, σ2) (13)

H0 : CARτ  N (µ, σ2) (14)

Where in this case:

CARτ = N

X

i=1

ARiτ (15)

Let OCAR be the ordered vector of the aggregated AR (CAR1 ≤ ... ≤ CAR26) The

test statistic for normality is defined as:

θ1 =

(PN

i=1aiOCARi) 2

PN

i=1(OCARi − OCAR)2

(16)

Where the constant a is in this case:

a0 = (a1, ..., an) =

m0V−1

(m0_V−1_V−1_m)1₂ (17)

The vector m0 = (m1, ..., mn) expresses the expected values according to standard

nor-mal order statistics. V is the corresponding N x N covariance matrix. The significance of the test statistic θ1 is based on the p-value. A 10% confidence interval is used in this

1% confidence interval will not give any significant results. For this reason, the 10% confidence interval is used which has a bigger null-hypothesis rejection area but is not as exact as smaller confidence intervals.

3.3

Testing the CAR

The CAR and mean CAR are powerful and informative representations of the sample abnormal returns. Because of this, the conclusions regarding the research question are based on these values. The previous part of this chapter tests if the AR and CAR are normally distributed. Based on this normality assumption, it is first tested if obtaining or losing a position in the global 100 list has a significant influence on the stock return. After that, it is tested if this influence has changed between the year 2006 and 2018.

To analyze if stock values are influenced by the global 100 list, the null-hypotheses and alternative null-hypotheses are formulated as follows:

H0 : CART1,T2 = 0 (18)

Ha: CART1,T2 6= 0 (19)

The null-hypothesis claims that the mean cumulative abnormal return of each sample is equal to zero. This means it is expected that there is no significant difference suggested in the return after the announcement, so and the event does not have any effect. Con-trariwise, the alternative hypothesis states a serious change in return during the test period, meaning the release of the G100 influences the stock return of the companies involved. In this part of the study it is assumed that the abnormal return is normally distributed with an expected value of zero. Because of this assumption, it is possible to distract the distribution of the mean cumulative abnormal return of all companies. The variance of the abnormal return of company i on day τ is σ2

AR,i,τ = ε2i where εi is the

standard error of that company obtained by performing OLS on the Fama and French four-factor model. The distribution of the abnormal return and cumulative abnormal return of the sample are:

University of Amsterdam

CARi(T1, T2) ∼ N (0, σ2CAR_i,T1,T2) (21)

σ_CAR2 _i, T1,T2 = T2 X τ =T1 σ_AR,i,τ2 (22)

The possibility distribution of the mean CAR is:

CART1,T2 ∼ N (0, σ 2 CAR_T1,T2) (23) σ_CAR2 T1,T2 = 1 N2 N X i=1 σ2_CAR,i,T1,T2 (24)

Where N is the number of companies in the sample. This can be derived from a statistic with a standard normal distribution:

θ2 = CART1,T2 (σ_CAR2 T1,T2) 1 2 ∼ N (0, 1) (25)

For this asymptotic test it is necessary to assume independence across the included securities (MacKinley, 1997). No clustering is allowed, meaning that the stocks do not have any overlap regarding the event window. Using θ2 we can test the significance of

the abnormal returns of all companies in the sample by performing a student t-test. The null-hypothesis will be rejected if the p-value is below 0.1 in combination with θ2 > 1.645 or θ2 < −1.645. This test is based on a 10% confidence interval for the

same reason as mentioned in Chapter 3.2. Rejecting the null-hypothesis implies that the mean CAR is significantly different from zero and the G100 has a serious influence on share values.

To examine the possible difference of this effect over time, the results of the AR and aggregated abnormal return (CAR) are used for comparison between the year 2006 and 2018. This means that besides the earlier null-hypothesis of the abnormal return to be zero, there is also another hypothesis that has to be tested. Under this null-hypothesis, it is expected that the mean CAR of the test period in 2006 is equal to the mean cumulative abnormal return of the test period in 2018. The null-hypothesis

and the alternative hypothesis are formulated as follows:

H0 : CARE6 = CARE18, H0 : CARIn6 = CARIn18 (26)

Ha : CARE6 6= CARE18, Ha : CARIn6 6= CARIn18 (27)

This hypotheses concerns all three test windows of excluded and included companies. To perform this test, two normally distributed test statistics are combined into a t-statistic. The first refers to the excluded companies in 2006 and 2018, and the second refers to included companies in 2006 and 2018:

θE =

CARE6− CARE18

q s2 E( 1 NE6 + 1 NE18) ∼ t(NE6+ NE18+ 1) (28) θIn =

CARIn6− CARIn18

q s2 In(NIn61 + 1 NIn18) ∼ t(NIn6+ NIn18+ 1) (29) Where s2_E = (NE6− 1)s 2 E6+ (NE8− 1)s2E18 NE6+ NE18− 2 (30) s2_In = (NIn6− 1)s 2

In6+ (NIn8− 1)s2In18

NIn6+ NIn18− 2

(31)

In these formulas E6, E18, In6 and In18 stand for exclusions in 2006, exclusions in 2018, inclusions in 2006 and inclusions in 2018 respectively. N is the sample size and s2 _{is the sample variance. The null-hypotheses is rejected if the values of θ}

E and θI n

are outside the 10% confidence interval (CI). This test also uses a 10% rejection area for the same reason as mentioned when testing θ1 and θ2. The confidence interval is

calculated as follows:

CIE =

h

(CARE 6− CARE18) − t0.05

r s2 E( 1 NE6 + 1 NE18 ),

(CARE 6− CARE18) + t0.05

r s2 E( 1 NE6 + 1 NE18 )i (32) CIIn = h

(CARI n6− CARIn18) − t0.05

r s2 In( 1 NIn6 + 1 NIn18 ),

(CARI n6− CARIn18) + t0.05

r s2 In( 1 NIn6 + 1 NIn18 )i (33)

University of Amsterdam

This test is performed on every event window: (AD-1, AD+1), (AD-5, AD+5) and (AD-5, AD+20).

4

Data

This chapter gives an overview of the data and data sources used in this thesis to investigate the influence of the Global 100 list on stock returns over time. The first part provides more background information about the company samples. The second part summarizes the statistical description of the variables.

4.1

Company Data

Based on their own research results, the Canadian research organization Corporate Knights compiles and publishes the Global 100 list every year. They have collected and saved information about the she sustainable performance and ranking position of every company ever included in the Global 100 since the year 2005. The company samples in this research and their data on environmental efficiency are provided from the database of Corporate Knights. Table 1 demonstrates the company samples and categorizes the companies based on their year of reference, main production division and data availability. The different major industry divisions are based on the firm categories used in the database of Corporate Knights. Not all companies expose their excess return data to public online sources, and some company data is incomplete or incorrect. For this reason the sample sizes in this thesis are smaller than the actual sample sizes. The amounts of companies associated with the different major industry divisions, lead to no suspicion of advantages or disadvantages based on this category. The samples are analyzed independently and not merged, meaning there are four small samples of 20, 24, 44 and 40 companies.

University of Amsterdam

Table 1: Company Sample Information

Year Major Industry Division Exclusions Available Inclusions Available

2006 Health Care 1 1 3 3

Energy 3 2 0 0

Financials and Real Estate 5 3 5 3

Industrials 2 1 4 4 Materials 6 5 5 5 Consumer Discretionary 6 5 6 4 Telecommunication 1 0 1 1 Utilities 3 2 1 1 Information Technology 2 1 2 2 Total 29 20 29 24 2018 Health Care 5 5 5 5 Energy 6 4 1 1

Financials and Real Estate 11 10 11 10

Industrials 6 6 5 5 Materials 2 2 3 2 Consumer Discretionary 9 8 8 7 Telecommunication 2 2 1 0 Utilities 2 2 7 4 Information Technology 6 5 7 6 Total 49 44 49 40

4.2

Factor Data

The main focus of this research is to analyze the difference in stock return after the publish date of the Global 100 list. The data on the excess stock return of the identified samples are derived from Wharton Research data devices. This is an extensively used research platform for global business data analysis. The exact event time line regarding included and excluded companies of the G100 list in 2006 runs from 2005/07/20 to 2006/02/21. For the year 2018, this period runs from 2017/07/20 to 20018/02/20 and the announcement day for both 2006 and 2018 is January 22nd. The data frequency is daily for all samples, which implies 5 trading days per week. The daily excess returns are based on the daily closing prices of shares and formula (5).

the event time line. The return means of the four samples are close to zero and the standard deviations are small, which is not surprising looking at the minimum and maximum values of the sample. The values of skewness and kurtosis also seem normal.

Table 2: Descriptive Statistics of Return

Mean SD Min Max Skewness Kurtosis N

2006 Exclusions 0.001 0.005 -0.013 0.021 0.341 4.724 20

Inclusions 0.001 0.004 -0.011 0.010 -0.203 3.116 24

2018 Exclusions 0.000 0.004 -0.010 0.012 -0.311 4.880 44

Inclusions 0.000 0.006 -0.020 0.017 -0.703 4.697 40

Figure A.1 to A.4 of the Appendix demonstrate a scatter plot including a fitted regres-sion line of every sample. The graphs are in line with the findings in Table 2. Besides a few outliers, the return during the event time line tends to fluctuate normally.

Chapter 5

5

Results

This chapter demonstrates and discusses the results of this research. The first part gives an overview of the obtained CAR. The second part states if a normally distributed CAR can be assumed according to the results of the Shapiro-Walk test. The third part shows the results of the tests regarding the G100 influence over time.

5.1

Cumulative Abnormal Returns

As described in Chapter 3, the abnormal return over the test period is calculated using the Fama and French four-factor model (9). Also, the cumulative abnormal return of N companies at day τ of the event window is computed. The CAR is a clear representation of the effect of the G100 on the share value after the event. Table 3 shows a short descriptive statistic of the CAR during the test period.

Table 3: Statistic Description of CAR

Mean SD Min Max Skewness Kurtosis N

2006 Exclusions 0.004 0.075 -0.190 0.185 -0.064 3.239 20

Inclusions 0.022 0.087 -0.131 0.194 0.081 2.127 24

2018 Exclusions -0.085 0.252 -0.910 0.438 -0.791 4.070 44

Inclusions -0.032 0.270 -0.778 0,485 -0.571 3.122 40

Table A.1 in the Appendix summarizes the AR on day τ of the event window and the CAR until that day. Figure A.5 to Figure A.8 of the Appendix draw a graph and a fitted regression line of the CAR for every sample to visualize the course and possible distribution of the AR. The figures show a course quite similar to the normal one. Every sample contains several outliers right before τ = 20, but this does not influence the direction of the regression line. The regression lines of the year 2006 are generally increasing, while the regression lines of the year 2018 are close to horizontal. The values

in Table 3 substantiate these findings. The obtained means and standard deviations are relatively close to zero. However, the mean CAR of exclusions are smaller than the mean CAR of inclusions in 2006 and 2018. This means that the excluded companies generally obtain a lower AR than the included companies after the event. The skewness of the samples are between -1 and 1 and there is no Kurtosis radically exceeding the normal value of 3. Based on these figures and table, there is no reason to invalidate the null-hypothesis of normally distributed abnormal returns with zero expectation so far.

5.2

Test for Normality

Besides analyzing the descriptive statistics in Table 3, it is important that the nor-mality distribution of the AR is officially tested. Otherwise, no legitimate conclusions can be made based on the results of the student t-test. The normality of the AR is examined by performing a Shaprio-Wilk test (1965). The obtained p-values of θ1 (16)

are demonstrated in Table 4.

Table 4: p-values of normality test

2006 2018

Exclusions Inclusions Exclusions Inclusions

Window p-value p-value p-value p-value

(AD-1,AD+1) 0.851 0.485 0.141 0.043**

(AD-5,AD+5) 0.749 0.673 0.418 0.205

(AD-5,AD+20) 0.983 0.819 0.482 0.432

*, ** or *** after a value stand for significance of the value using a 10%, 5% and 1% confidence interval respectively.

Except for one p-value, all values in Table 4 are above 0.1. This means that only for included companies in 2018 on short term, the null-hypothesis of a normally distributed CAR would is rejected. If the normality test was based on a 5% confidence interval, there still would have been just one significant value. Despite this exceptional p-value, the results in Table 6 confirm the expectations based on Table 3. There is not enough significant proof to reject the null-hypothesis (13) that the CAR after the event is normally distributed.

University of Amsterdam

5.3

Student t-test

To test the significance of the change in CAR after the event, the test statistic θ2 (25)

is formed which is assumed to be normally distributed. The value of θ2 is tested by

a student t-test. Table 5 demonstrates the results of this test. The third and seventh column give the mean CAR of the sample respecting the different event windows. The fourth and eighth column show the percentage of the excluded (included) companies with a negative (positive) CAR during the investigated event window. The fifth and ninth column show the value of the test statistic θ1, finally the p-value of the test is

given in the fourth and eighth column.

Table 5: Results on CAR

Exclusions Inclusions

Window (T1, T2)

CART1,T2 %CAR < 0

θ1 p-value CART1,T2 %CAR > 0 θ1 p-value 2006 (AD-1,AD+1) -0.013 65% -2.271 0.023** -0.014 33% -2.995 0.003*** (AD-5,AD+5) -0.015 65% -1.422 0.155 -0.019 25% -2.130 0.033** (AD-5,AD+20) -0.004 55% -0.226 0.821 0.009 50% 0.674 0.500 2018 (AD-1,AD+1) 0.005 48% 1.620 0.105 0.004 53% 0.765 0.444 (AD-5,AD+5) 0.010 44% 1.589 0.112 0.012 60% 1.231 0.218 (AD-5,AD+20) -0.040 75% -4.297 0.000*** -0.036 30% -2.319 0.020**

The results in Table 5 show that in almost every case, the mean CAR of exclusions is lower than the mean CAR of inclusions of the same test period. On the short term, the stock performance of exclusions is almost equal to the stock performance of inclusions. The difference between exclusions and inclusions seems to increase on the long term. Based on the outcome demonstrated in table 5, the 5 significant p-values provide enough proof to reject the null-hypothesis (18) that the G100 has no influence on the stock value of companies.

Second, the student t-test is performed to examine a significant difference in G100 influence between the year 2006 and 2018. For this test it is also assumed that the AR and CAR are normally distributed. The significance of the test statistics θE

intervals (32) and (33) as described in section 3.2. Table 6 summarizes the test results. The second, third, sixth and seventh column give the mean CAR of exclusions in 2006, exclusions in 2018, inclusions of 2006 and inclusions of 2018 respective. The fourth and eighth column show the value of the test statistic and column five and nine show if the test statistic is below, above or in the confidence interval (CI).

Table 6: Differences in mean CAR over Time

Exclusions Inclusions

Window (T1, T2) CAR2006 CAR2018 θE in CI? CAR2006 CAR2018 θIn in CI?

(AD-1,AD+1) -0.012 0.001 -0.297 < CI -0.008 0.004 -0.249 < CI

(AD-5,AD+5) -0.015 -0.003 -0.313 < CI -0.011 0.012 -0.386 < CI

(AD-5,AD+20) -0.004 -0.074 0.291 > CI 0.009 -0.035 0.384 > CI

Table 6 shows that in general, the CAR of 2006 is significantly lower than the CAR of 2018 on a short term perspective. On the long term, the CAR of shares in 2006 sig-nificantly outperform the CAR of shares in 2018. These statements concern excluded companies as well as included companies. In the Appendix, Figure A.9 shows the differ-ence between exclusions of 2006 and 2018 and Figure A.10 shows the differdiffer-ence between inclusions of 2006 and 2018. Around the event date, the difference is mostly negative and despite the fluctuations this inequality gets more positive in time. Furthermore, the difference between the mean CAR of exclusions and the mean CAR of inclusions in 2006 is bigger than the mean CAR of exclusions and the mean CAR of inclusions in 2018. This implies that the Global 100 list has more influence on the stock value of companies in 2006 than in 2018. Based on the results in table 6, there is enough signif-icant proof to reject the null hypothesis (26) that the CAR of excluded companies in 2006 is equal to the CAR of excluded companies in 2018. Also, the null-hypothesis can be rejected that the CAR of included companies in 2006 are not significantly different from the CAR of included companies in 2018.

Chapter 6

6

Conclusion

This chapter draws a conclusion based on the results of this study and discusses poten-tial further research.

This thesis aims to determine the effect of a position on the Global 100 list on the share value, and the possible change in influence of the list over time. It ex-amines if there is a significant difference between the expected stock return and actual stock return of a company after losing or obtaining a position on the G100 list. In addition, the difference between expected and actual stock return in 2006 is compared to the difference between expected and actual stock return in 2018. To do this, the period around the announcement date (January 22nd) is investigated which runs from 2005/08/01 until 2006/02/21 for 2006, and from 2017/08/01 until 2017/02/20 for 2018. Assuming the stock value of a firm is directly influenced by the preference of investors, the outcome of this research can provide information about the development of investors‘ interests in sustainable companies.

The results of the first test show that the assumption of the difference between expected and actual stock return being normally distributed is valid. Based on these results, the legitimacy of the following tests in this study tests is confirmed.

The second test examines if the influence of the G100 list is significant. Ac-cording to the outcomes, the abnormal return of excluded companies is generally lower than the abnormal return of included companies. Also, the difference between the per-formance of excluded and the perper-formance of included companies increases on the long term.

The third test, investigating the difference between 2006 and 2018 specifically, confirms that the contrast between excluded and included companies is bigger in 2006 than in 2018.

of the test outcomes can be analyzed by further research. For example, the tests can be done using larger samples by adding more years (2007, 2017) and taking extra variables into account such as the expenditures on CSR and the rank of the company on the G100 list. Also, a larger event window could be investigated to see if the difference between 2006 and 2018 and the difference between included and excluded companies keep increasing.

There are several more famous yearly sustainable company ranking lists which are published a few months earlier than the G100, the Dow Jones Sustainability Index for example. It is possible that the influence of G100 list does not reflect the interest of investors on environmental issues, because they already took the content of these other lists into account by making an investment decision.

Concluding, based on the results it can be assumed that the decision of investors is insignificantly influenced by the sustainability of the company. However, despite of the worldwide growing environmental awareness over the past ten years, the effect of the Global 100 list seems to be lower in 2018 than in 2006. This means that people do not necessarily act towards their increased knowledge when it comes to investing money.

7

References

Ballou, B., Heitger, D. L., Landes, C., Adams, M. (2006). The Future of Corporate Sustainability. Journal of Accountancy, 202(6):65-74.

Bamberg, S. (2003). How does environmental concern influence specific environmen-tally related behaviors? A new answer to an old question. A Journal of Environmental Psychology, 23:21-32.

Bansal, P., Roth, K. (2017). Why Companies Go Green: A Model of Ecological Re-sponsiveness. Acadamy of Management, 43(4):717-736.

Chen, H., Noronha, G., Singal, V. (2005). The Price Response to SP 500 Index Addi-tions and DeleAddi-tions: Evidence of Asymmetry and a New Explanation. The journal of American Finance Association, 59(4):1901-1930.

Cheung, A. (2011). Do stock investors value corporate sustainability? Evidence froman event study. Journal of Business Ethics, 99(2):145-165.

Corrado, C. (1989). A Nonparametric Test for Abnormal Security-Price Performance in Event Studies. Journal of Financial Economics, 23(2):385-95.

Curto, J. D., Vital, C. (2014). Socially Responsible Investment: A Comparison be-tween the Performance of Sustainable and Traditional Stock Indexes. University of Lisbon, Journal of Reviews on Global Economics, 22(18):349-363.

Doran, P. T., Zimmerman, M. K., (2011). Examining the Scientific Consensus about Climate change. American Geophysical Union, 90(3):22-23.

Fama, E. F., French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1):3-56.

Flammer, C. (2017). Corporate Social Responsibility and Shareholder Reaction: The Environmental Awareness of Investors. Acadamy of Management, 56(3):758-781.

Frost, G., Jones, S., van der Laan, S., Frost, G., Loftus, J. (2008). The Investment Performance of Socially Responsible Investment Funds in Australia. Journal of Busi-ness Ethics, 80(2):181-203.

Geczy, C., Sambaugh, R., Levin, D. (2005). Investing in Socially Responsible Mu-tual Funds.Wharton School Working Paper, University of Pennsylvania.

Glushkov, D., Statman, M., (2008). The Wage of Social Responsibility. Financial Analysts Journal, 65(4):33-44.

Jegadeesh, N. Titman, S. (2001). Profitability of Momentum Strategies: An Eval-uation of Alternative Explanations. Journal of Finance, 56(2):699-720

Karlsson, J., Chakarova, Y. (2008). Does corporate social responsibility pay off? School of Business, Economics and Law, at Göteborg University.

Kinder, P. (2005). “Socially Responsible investement”: An involving concept in a chang-ing world. KLD Research & Analytics.

King, A. A., Lenox, M. J. (2008). Does It Really Pay to Be Green? An Empirical Study of Firm Environmental and Financial Performance: An Empirical Study of Firm Environmental and Financial Performance. Journal of industrial Ecology.

Kurtz, L. (Crane, A., Matten, D., MacWilliams, A., Moon, J., Siegel, D. S.) (2008) Socially Responsible Investment and Shareholder Activism. The Oxford Handbook of Corporate Social Responsibility.

Lombardo, R., D‘Orio, G. (2012). Corporate and State Social Responsibility: A long-term perspective. Department of Economics and Statistics, University of Calabria.

MacKinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Eco-nomic Literature, 35(1):13-39.

Mill, G. A. (2006). The Financial Performance of a Socially responsible Investment over Time and a Possible Link with Corporate Social Responsibility. Journal of busi-ness ethics, 63:31.

Murgaia, J. M., Lence, S. H. (2015). Investors‘ Reaction to Environmental Perfor-mance: Global Perspective of the Newsweek‘s “Green Rankings”. Environmental and Resource Economics, 60(4):583-605.

Orlitzky, M., Schmid, F. L, Rynes, S. L. (2003). Corporate Social and Financial per-formence: A Meta-Analysis. Sage Journals, 24(3):403.

Shaprio, S. S., Wilk, M. B. (1965). An Analysis for Variance Test for Normality (Com-plete Samples). Biometrika, 52(3/4):591-611.

Staff, C. K. (2014). Key Performance Indicators. Corporate Knights. Geraadpleegd op 22 april 2018, van http://www.corporateknights.com/uncategorized/key-performance-indicators/.

Staff, C. K. (2015a). 2015 Global 100 Results. Corporate Knights. Geraadpleegd op 22 april 2018, van http://www.corporateknights.com/magazines/2015-global-100-issue/2015-global-100-results-14218559/.

Staff, C. K. (2015b). Global 100 Methodology. Corporate Knights. Geraadpleegd op 22 april 2018, van http://www.corporateknights.com/reports/2015-global-100/methodology/.

Stenström, H. C., Thorell, J. J. (2007). Evaluating the Performance of Socially Re-sponsible Investment Funds: A Holding Data Analysis. Stockholm School of economics.

Tsai, C.C., (2007). ‘The Reputation Effect and Value in Corporate Social Respon-sibility’, Department of Finance, Yuan Ze University, Taiwan.

Whitmarsh, L. (2011). Skepticism and uncertainty about climate change: Dimensions, determinants and change over time. Global environmental change, 21(2):690-700.

A

Appendix

Figure A.1: Daily Average Excess Return of excluded companies in 2006

Table A.1: AR and CAR over test period

2006 2018

Exclusions Inclusions Exclusions Inclusions

τ AR CAR AR CAR AR CAR AR CAR

AD-5 -0.118 -0.118 -0.098 -0.098 -0.043 -0.043 0.150 0.150 AD-4 0.033 -0.085 0.011 -0.087 0.129 0.085 0.178 0.327 AD-3 -0.039 -0.124 -0.058 -0.145 -0.061 0.024 -0.160 0.167 AD-2 0.024 -0.100 -0.083 -0.227 -0.149 -0.125 -0.009 0.158 AD-1 -0.030 -0.130 -0.131 -0.358 0.106 -0.019 -0.008 0.150 AD -0.078 -0.208 -0.083 -0.441 0.100 0.081 -0.013 0.138 AD+1 -0.141 -0.349 -0.121 -0.562 0.020 0.101 0.180 0.317 AD+2 -0.032 -0.381 -0.055 -0.617 0.026 0.127 -0.016 0.302 AD+3 0.121 -0.261 0.067 -0.551 0.059 0.185 -0.053 0.249 AD+4 -0.015 -0.275 0.021 -0.530 0.065 0.250 0.102 0.351 AD+5 -0.023 -0.299 0.074 -0.456 0.174 0.424 0.139 0.490 AD+6 0.025 -0.274 0.053 -0.403 -0.001 0.424 0.033 0.522 AD+7 0.087 -0.188 0.008 -0.395 -0.138 0.286 -0.286 0.236 AD+8 0.031 -0.157 0.129 -0.266 -0.302 -0.016 -0.240 -0.004 AD+9 -0.009 -0.166 0.102 -0.164 0.106 0.089 0.162 0.158 AD+10 0.063 -0.102 -0.096 -0.260 -0.129 -0.039 0.015 0.173 AD+11 0.077 -0.025 0.181 -0.079 -0.283 -0.322 -0.233 -0.061 AD+12 -0.083 -0.108 0.139 0.060 -0.070 -0.392 -0.153 -0.213 AD+13 -0.083 -0.191 -0.094 -0.033 0.108 -0.284 0.043 -0.171 AD+14 0.020 -0.171 0.059 0.026 -0.315 -0.599 -0.330 -0.501 AD+15 -0.018 -0.189 -0.034 -0.008 -0.406 -1.001 -0.406 -0.907 AD+16 0.014 -0.175 0.101 0.093 -0.781 -1.787 -0.778 -1.685 AD+17 -0.040 -0.216 0.056 0.148 0.525 -1.262 0.485 -1.200 AD+18 0.185 -0.030 0.042 0.190 -0.370 -1.632 -0.355 -1.554 AD+19 0.148 0.117 0.087 0.277 -0.405 -2.037 -0.332 -1.886 AD+20 -0.190 -0.073 -0.055 0.222 0.272 -1.765 0.468 -1.418

Figure A.7: Exclusions 2018