The relation between sustainability performance and stock market returns : an empirical analysis of the Dow Jones Sustainability Europe Index

(1)

1

The relation between sustainability performance and

stock market returns:

An Empirical analysis of the Dow Jones

Sustainability Europe Index

Student: Akim van Stekelenburg Student number: 10438831

Supervisor: Dr. Georgios Georgakopoulos Second reader: Professor Brendan O’Dwyer Paper version: Final version

(2)

2 Abstract

This paper investigates the relation between corporate financial performance (CFP) and corporate sustainability performance (CSP). This is done in a two-pronged approach, by first analyzing a sample of European stocks that were added to or deleted from the Dow Jones Sustainability Europe Index (DJSI Europe) over the period 2009–2013, and second by

analyzing a sample of European stocks that were recognized as industry group leaders in CSP by the DJSI Europe over the same period. The impacts are measured in terms of (abnormal) stock returns. For the first analysis no strong evidence could be found that the announcement of the inclusion and exclusion events has any significant impact on stock return. However, on the day of change (CD) and in the period following CD, index inclusion (exclusion) stocks experience a significant but temporary increase (decrease) in stock return. These results seem to support Harris and Eitan’s (1986) price pressure hypothesis, which postulates that event announcement does not carry information and any shift in demand and hence the

corresponding price change is temporary. From the second analysis, on industry group leaders, it can be concluded that the market rewards firms with high CSP. In the period after the day of change, industry group leader stocks experience a permanent and significant positive growth in stock returns. This conclusion can be supported by the resource based perspective, which posits that firms capable of investing heavily in CSP have greater

underlying resources which in turn should produce higher financial performance (Alexander and Bucholz, 1978; Waddock and Graves, 1997; Clarkson et al., 2006).

KEY WORDS: corporate sustainability performance, corporate social responsibility, Dow Jones Sustainability Europe Index, event study, price pressure hypothesis, resource based perspective

(3)

3

Sustainability reporting is a ‘hot topic’ in current academic literature and many related studies have been published in recent years. While most research shows that, while sustainability reporting is not mandatory, more entities choose to report on their social and environmental performance (Solomon and Solomon, 2006). Since 1993, KPMG publishes its International Survey on Corporate Responsibility Reporting, they believe the report to be the largest and most comprehensive survey conducted on this subject and shows the rise in reporting on corporate sustainability performance (CSP). According to the 2011 report approximately 95% of the 250 largest companies worldwide issue sustainability reports, this is an increase from 80% in 2008 and 50% compared to 2005 (KPMG 2011; KPMG 2008; KPMG 2005).

It is clear that it is no longer sufficient for entities to only report on their financial performance, society and stakeholders also expect organizations to report on the impact of their operations on society and the environment. There is a demand for accountability in the form of reports on CSP (Solomon and Solomon, 2006).

Most research conducted regarding the relation between CSP and financial performance outcomes has been focused on the relation between corporate social responsibility (CSR) and corporate financial performance (CFP). Results for the relations between CSP and CFP provide mixed results, research focuses on understanding the type of relation that might exist between the factors (positive, negative, or neutral). This is generally done by investigating whether organizations that are perceived as sustainable outperform or underperform firms which are not perceived in the same regard. A number of papers have been published on this topic, some have suggested a positive relationship (Waddock and Graves, 1997; Orlitzky et al., 2003; Margolis and Walsh, 2001; Cormier and Magnan, 2007), some a neutral impact (Bauer et al., 2005; Becchetti et al., 2008; McWilliams and Siegel, 2000) and one a negative relation (Garcia-Castro et al., 2010). It is clear that recent research shows mixed results on the impact of CSP on CFP, a consensus has not yet been reached.

As shown above there is quite a large number of literature regarding the relation between CSP and CFP. There is however, a rather limited amount of academic research that investigates the relation between CSP or CSR and market returns. Some research has tried to identify whether sustainability-related reporting is associated with market-based measures. For instance Freedman and Patten (2004) examined the market reaction to Toxic Releases Inventory and 10-K report environmental disclosures in the US and found that companies with poorer pollution performance suffered more negative market reactions than companies with better performance. A study done in the UK by Lorraine et al. (2004) investigated the relationship between market reactions and publicity regarding fines for environmental

pollution and recommendations for good environmental achievements. The researchers found a lagged market reaction to the news, especially to the negative news.

(5)

5 Another study conducted in the UK by Brammer, Brooks and Pavelin (2006) examined the relation between CSP and stock returns in the UK. The authors evaluated the interactions between social and financial performance with a set of disaggregated social performance indicators for environment, employment, and community activities instead of using an aggregate measure and surprisingly found that that firms with higher social performance scores tend to achieve lower returns, while firms with the lowest possible CSP scores of zero outperformed the market. The authors recognized as a limitation that they only had one set of social performance indicators at their disposal and concluded that the various aspects of corporate social behaviour must be examined separately in order to achieve an accurate picture of their impact on returns.

Finally, Jones et al. (2007) tried to identify, in the Australian context, whether the level of sustainable reporting is associated with a range of financial and market performance attributes of the firm. They tested the relation between sustainability reporting and market returns by first scoring each individual firms sustainability disclosure with the Global Reporting Initiative (GRI) framework, and then regressed the index they created against the abnormal stock returns of the sample entities. The authors were unable to draw a definitive conclusion as most of the results were statistically insignificant, but most of the data indicated a negative relation between the two factors.

It is clear that the current literature still provides mixed results for the relations between CSP and market returns. No conclusive consent has yet been reached as a result of empirical examinations. Furthermore, almost all research has been conducted in either the US or the UK, while it might prove to be worthwhile to investigate the European market.

Considering the above mentioned facts, the purpose of this paper is to provide evidence on whether a relation exists between CSP and a firm’s market returns, with a focus on the European market.

1.2 Development of research question

As previously mentioned, in the current academic literature there is no conclusive consent regarding the relation between sustainability performance and financial performance in terms of market returns. This due to multiple factors that have been identified by researchers themselves, such as: limited amount of indicators, lack of well-developed theoretical foundation, and a lack of empirical evidence in the field (Jones, et al., 2007). Furthermore, almost all the studies have used different measures for CSP, also most have ignored the leading index for sustainability performance: The Dow Jones Sustainability Index (DJSI), with the most notable exceptions of Cheung (2011), Chih, Chih and Chen (2010) and Lourenço et al. (2012). Additionally most of the research conducted took place in the US or the UK, almost none have looked at this relationship in a European context.

Taking this into consideration, this paper aims to extend research in the area of CSP and market returns, by analysing whether firms that are considered to have a high-CSP

(6)

6 captures the excess returns an investor would have received over a particular event day (event period) if one had invested in security i.

As a result, the following research question has been formulated for this paper:

Does the market reward European firms that have been recognized in having high sustainable performance in terms of abnormal returns?

To put it another way, the purpose of this paper is to investigate whether social and environmental responsible entities can be expected to be rewarded for their sustainable practices.

1.3 Motivation

There are a number of underlying motivations for conducting this research. Firstly, when looking at the contradicting results in the current literature, this paper will try to address some of the problems identified in these recent publications and try to explain to the seemingly inconsistent findings. Namely, by using a consistent and widely recognized measure of CSP; the DJSI Europe.

Secondly, this paper is motivated by the goal of shedding new light on the relationship between CSP and market returns. There is a clear lack of empirical evidence on this subject and this study aims to contribute to this lack by adding new insights to the existing literature, this can be seen as the academic contribution of this paper.

Finally, this study aims to broaden the theoretical background on the discussion of the relation between CSP and market returns, by including well documented theories such as the stakeholder theory. This research will show through an event study on index and exclusion events of the DJSI Europe as well an event study on the DJSI Industry group leaders whether external stakeholders, in this case especially investors’ value CSP and if they take indexes such as the DSJI Europe into their pricing decisions. A number of papers have shown that public sustainability, accountability and ethical (SAE) disclosures are lacking in content to investors (Solomon and Solomon, 2006; Solomon et al., 2011), and that investors rely more on private disclosures with their clients. This study will also contribute to this discussion by showing whether investors take external valuations on CSP into consideration, which can be seen as the societal contribution of this paper.

This paper is organized as follows. The second section provides the results of a theoretical and empirical literature review, and the development of the hypotheses. The third section describes the research design, with reference to sample construction, data collection process and the research methodology, while the fourth presents the descriptive statistics, and the empirical results, together with a discussion of the main findings. The final section summarizes and concludes the paper.

(7)

7 2. Literature review and hypotheses

Where the first section served the purpose of presenting the research question and the

underlying motivations for this paper. This section will provide more in depth explanations on the current state of research in the field of sustainability performance, this will be done from both a theoretical as well as an empirical point of view.

As a result the motivations will also be better highlighted and find more support by providing a theoretical and empirical background, based on which the analysis will be undertaken.

The section is organized as follows. Firstly, the main concepts used in the paper will be defined and explored. This will aid in understanding the terms and concepts that this research is based upon. Secondly, a theoretical framework is provided. Thirdly, an overview of the main studies focused on the CSP – financial performance relation is performed.

Fourthly, the effect on inclusion and exclusion in a reputation index will be explored. Finally, this will lead to the development of two research hypotheses and an explanation hereof.

2.1 Corporate sustainable performance

Wood (2010) considers the CSP domain to be controversial, fluid, ambiguous and difficult to research. This despite the fact that CSP and its sister concepts – corporate social responsibility (CSR), corporate social responsiveness, and corporate citizenship, have been present in

management scholarship for about 45 years, (Wood, 2010). This adds to the issue at hand, that there is no clear agreed upon definition for CSP.

To make matters even more complex, according to McWilliam et al. (2006) CSP is often used as a synonym for corporate sustainable responsibility (CSR), and consider both CSR and CSP as an embryonic concept in the academic world, for which no clear definition has been found. This in turn can also be debated, as according to Wood (1991), CSP can be seen as the application of CSR principles. Carroll and Shabana (2010) have a similar view, they state that CSP has become an established umbrella term which embraces both the

descriptive and normative aspects of CSR, as well as placing an emphasis on all that firms are achieving or accomplishing in the realm of social responsibility policies, practices and results. As the main focus of this study is in regard to CSP, the views of Wood (1991) shall be

adopted. Even though there is a large degree of overlap between the concepts of CSR and CSP, they shall be considered to be two separate notions, and the definition of CSP will form the basis for this paper.

According to Orlitzky et al. ( 2003 p. 411) the definition provided by Wood (1991) ‘is one of the most influential, helpful, parsimonious, and yet comprehensive conceptualizations of CSP’, therefore the choice was made to adopt this definition for the purposes of this study. Wood (1991, p. 693) defines CSP as: ‘a business organization’s configuration of principles of social responsibility, processes of social responsiveness, and policies, programs, and

(8)

8

2.2 Theoretical framework

In the current literature there is no single theory linking the relationship between CSP reporting and financial performance. Through the years there have been attempts to create such a theory. However, in 2010, Wood observed that 25 years after researchers started to try to explain the effects of this relationship, the search is still on.

The search for a theory began in 1985 when Ullman published a paper called ‘Data in search of theory’ that contained an in depth examination of the CSP – CFP relationship, yet the author concluded that ‘the situation pertaining to the relationships among social performance, social disclosure, and economic performance can best be characterized at this time as

empirical data in search of an adequate theory.’

In other words in 1985, there was no clear governing theory on this association and this is still the case 28 years later.

The fact that there isn’t one overarching theory on the relation between CSP and financial performance, be it accounting based performance or market based performance, doesn’t imply that there aren’t any theories that might help to explain this phenomenon. There are in fact several theories that can be linked to sustainable performance and firm performance (McWilliams and Siegel, 2001).

The most commonly used theory related to all factors regarding sustainability

reporting and sustainable performance is the stakeholder theory. A stakeholder is defined by Freeman (1984) as any group or individual who can affect or is affected by the achievement of the organization’s objectives. This theory tries to identify which groups within and outside of an organization are stakeholders, and should be taken into consideration by management. Stakeholder theory plays an important role for an organization, Clarkson (1995) identifies two types of stakeholders. Primary stakeholders are ‘One without whose continuing participation the company cannot survive’ and secondary stakeholders are ‘those who influence or affect, or are influenced or affected by, the corporation, but they are not engaged in transactions with the corporation and are not essential for its survival’ (Clarkson, 1995, pp. 106-107).

The framework created by Ullmann (1985) is especially of importance for this study, based upon the stakeholder approach to strategic management, proposing that stakeholders ultimately control a firm’s access to scarce resources and firms must manage their relation with key stakeholders to ensure that such access to resources is maintained (Roberts, 1992). Furthermore, Ruf et al. (2001) use stakeholder theory as a basis for investigating the

relationship between change in corporate social performance and financial performance. This provides evidence that stakeholder theory is an important basis for empirical analyses relating to CSP. This is also reinforced by a number of more recent studies (e.g. Artiach et al., 2010; Chih, Chih, Chen, 2010; Jones et al., 2009; Ziegler, 2012).

Another interesting theory that complements stakeholder theory and that also might help to explain the relation between CSP and market performance is the resource-based perspective (RBP), since firms may view meeting stakeholder demands as a strategic investment, requiring commitments beyond the minimum necessary to satisfy stakeholders (Ruf et al.,

(9)

9 2001; Lourenço et al., 2012). Under the RPB it is suggested that organizations can generate sustainable competitive advantages by effectively controlling and manipulating their

resources that are, rare, cannot be perfectly imitated, and for which no perfect substitute is available (Lourenço et al., 2012). Furthermore, external benefits of CSP are related to corporate reputation, which has been identified as one of the most important intangible resources that provide a firm sustainable competitive advantage (Roberts and Dowling, 2002; Orlitzky et al., 2003; Lourenço et al., 2012). With a result that companies with high CSP reputation are able to improve relations with external stakeholders such as customers, investors, bankers, suppliers, and competitors. As discussed above stakeholders ultimately control a firm’s access to scarce resources and firms must manage their relationship with key stakeholders to insure that such access to resources is maintained (Roberts, 1992).

In conclusion, sustainable performance, in the long run, can raise benefits through improved relations with stakeholders and reduced cost of conflicts with them and increase reputation creation, which makes an organization more attractive to investors (Laurenço et al., 2012).

2.3 Review of empirical evidence and hypothesis development

As discussed in this paper the two concepts of CSP and CSR are considered as two separate notions. However, as seen in the theoretical framework the two concepts are often used interchangeably within various papers (e.g. Wood, 1991; McWilliam et al., 2006, Carroll and Shabana, 2010; Wood, 2010). Therefore the literature review of empirical results in the following section will reference investigations that study the relation between corporate sustainability performance, corporate social performance or corporate social responsibility and market based financial performance. This due to the fact that when the concepts of CSP and CSR are used in empirical analysis, they are often based on similar measures.

2.3.1 Measures of corporate sustainable performance

According to Waddock and Graves (1997) CSP is a multidimensional construct, which is not only hard to define but also hard to measure. Despite this difficulty, with the increasing interest shown in CSP and CSR activities, the need for measurement has also increased.

In 1985, Ulmann was one of the first who tried to identify and summarize different types of CSP measurements. Based on an assessment of 31 empirical studies conducted in the 70’s and early 80’s, Ulmann (1985) categorized three broad measures for CSP. The first being social disclosures (such as voluntary corporate social reporting and mandatory disclosures on pollution), the second being social performance (such reputational indexes or rankings) and the third and final category being economic performance (such as accounting or market based measures of performance, e.g. shareholder returns, price/earnings ratio, return on equity, net income, or net profit margin).

Orlitzky et al. (2003) also identified a number of broad strategies by which CSP can be measured. The researchers identified four categories, namely: 1) CSP disclosures, 2) CSP reputation ratings, 3) social audits, processes and observable outcomes, and 4) managerial CSP principles and values.

(10)

10 In a similar study to that of Ulmann (1985) published in 2010, Wood conducted an extensive review of the at that time existing literature on CSP. The paper aimed at reviewing the literature on corporate social performance measurement and setting that literature into a theoretical context.

While doing so Wood (2010) conducted an analysis in order to identify the most frequently used measures of CSP in literature. The research showed that one the most frequently used measures of CSP is the KLD rating. In 1988, the investment firm of Kinder, Lydenberg, Domini and Company, Inc. (KLD) began offering a mutual fund of US companies which mirrored the Standard & Poor 500 except for being screened for social responsibility factors. Companies were included for their lack of participation in such ‘irresponsible’ fields as military contracting, South Africa, tobacco products, gambling and nuclear power. Over time, KLD added positive screens to their ratings database and their SOCRATES database is now widely used as a way of calculating CSP in empirical studies Wood (2010). However, KLD and similar ratings are sometimes used as aggregate measures of overall CSP.

As can be seen above there are a large variety of measures that can be used to measure CSP, but there is no consensus as to which method is the ‘best’. An answer to the increasing need for measurement can be seen in the growing popularity of indices such as the Dow Jones Sustainability Index, FTSE4Good and the Domini 400, which have been initiated in the past few years. Inclusion in reputation indices has been recognized as a sound measure for sustainable performance (Ulmann 1985; Orlitzky et al. 2003).

Similar to prior literature, current analysis infers a company’s level of CSP as high (low) by inclusion (exclusion) in a reputation index, in this case Dow Jones Sustainability Europe Index (DJSI Europe).

The DJSI Europe was established in August 2010 to track the performance of the region’s largest companies from developed countries that lead the field in terms of corporate sustainability. These companies are assessed by RobecoSAM using the annual Corporate Sustainability Assessment (CSA). Established in 1999, as the first ever family of global sustainability benchmarks, the Dow Jones Sustainability Indices (DJSI) have become a reference point in Sustainability Investing (DJSI Europe Index Guide, 2013).

The DJSI family uses what they refer to as a best-in-class approach to select sustainability leaders from across all industries based on pre-defined sustainability criteria embedded in the CSA. The DJSI best-in-class approach is based on three factors, namely: 1) no industry is excluded from the indices, with the most sustainable companies in each industry selected for index membership, 2) companies receive a Total Sustainability Score between 0 – 100 and are ranked against other companies in their industry, and 3) only the top 20 % of companies from each industry, based on their sustainability score, are included in the Dow Jones Sustainability Europe Index (DJSI Europe Index Guide, 2013).

As a result companies must continually intensify their sustainability initiatives to be included or remain in the index. Furthermore, a growing number of companies define inclusion in the DJSI as a corporate goal as it publicly endorses their approach to addressing key sustainability issues (DJSI Europe Index Guide, 2013).

(11)

11 The Corporate Sustainability Assessment used by the DJSI consists out of several components, such as identifying which companies are better equipped to recognize and respond to emerging sustainability opportunities and challenges presented by global and industry trends, as well as a Media and Stakeholder Analysis (MSA). RobecoSAM conducts its MSA by the ongoing monitoring of media and stakeholder commentaries and other publicly available information from consumer organizations, NGOs, governments or international organizations to identify companies’ involvement and response to

environmental, economic and social crisis situations that may have a damaging effect on their reputation and core business (DJSI Europe Index Guide, 2013).

Furthermore, the index also factors in an ethical analyses which identifies companies that generate revenue from producing or participation in the production of alcohol, tobacco, gambling, armaments, cluster bombs, landmines, firearms, nuclear, and adult entertainment. Companies that generate revenue from the above listed activities, will be excluded from the subset indices. If more than 5% of a company’s sales are derived from armaments then this company is excluded from the subset index, for all other activities a 0% threshold is applied.

Finally, The DJSI Europe is reviewed annually and rebalanced quarterly to ensure that the index composition accurately represents the top 20% of the leading sustainable European companies. The resulting changes to the index composition are announced on the annual review date in September (DJSI Europe Index Guide, 2013).

2.3.2 The relationship between corporate sustainable performance and firm value

When looking at empirical evidence on the relationship between CSP and firm value, this area of literature is relatively new. It primarily takes an investment perspective and focusses mainly on whether sustainable performance is ‘priced’ in capital markets (e.g. Lo and Shue, 2007) or whether highly sustainable companies outperform other companies in terms of financial performance (e.g. Konar and Cohen, 2001; Lopez et al., 2007; Becchetti et al., 2008; Chih et al., 2010; Artiach et al., 2010; Lourenço et al, 2012).

These studies have produced interesting albeit contradicting results, for instance, Konar and Cohen (2001) found in a study that relates the market value of firms in the S&P 500 to objective measures of their environmental performance, that bad environmental performance is negatively correlated with the intangible asset value of firms. However, in terms of indicators of good sustainable performance, Lopez et al. (2007) found that when analysing the link between performance indicators and sustainable reporting, this link

between these variables is negative. Which suggests that the effect of sustainability practices on performance indicators is negative during the first years in which they are applied.

Somewhat similarly, in an attempt to specify the conditions under which corporations may or may not act in socially responsible ways Chih et al. (2010) found amongst others that firms with larger size are more CSR minded, and that the financial performance and CSR are not related.

By using standard event study methodology, this paper takes a different approach through examining events of index inclusion and index exclusion from the Dow Jones Sustainability

(12)

12 Europe Index, as well as firms recognized as industry group leaders by the index. The main advantage of using this approach over the above mentioned approaches is that by looking into how stock markets respond to these events, it can provide evidence to the question of whether CSP matters to investors or not. Furthermore, by varying the length of the event window, gives an understanding to both short- as well as long-term stock behaviours.

There are a number of studies related to this paper. For instance using an international database containing 103 German, UK and US ethical mutual funds, Bauer et al. (2004) found that after controlling for investment style, there was no evidence of significant differences in risk-adjusted returns between ethical and conventional funds for the period 1990-2001. Contradictory to the findings of Bauer et al. (2004), in a study similar to this paper, Consolandi et al. (2009) performed an event study on the Dow Jones Sustainability Stoxx Index (DJSI Stoxx) and concluded that the evaluation of the CSP performance of a firm is a significant criterion for asset allocation activities over the period 2001-2006. However, in a US context where Cheung (2011) analyses the impacts of index inclusions and exclusions on corporate sustainable firm by studying a sample of US stocks that were added or deleted from the Dow Jones World Index (DJSI World) over the period 2002-2008. They found no

evidence that announcement has any significant impact on stock return. Finally, in a more recent study analysing the effect of the inclusion of German corporations in the DJSI World and the DJSI Stoxx on stock performance, the empirical results of Orbendorfer et al. (2013) suggest that stock markets may penalize the inclusion of a firm in a sustainability index. This finding is mainly driven by a strong negative effect of the inclusion in the DJSI World. 2.3.3 The effect of index inclusions and exclusions

Amongst current literature there are a number of theories and papers that document strong empirical evidence of positive (negative) price impacts upon index inclusion (exclusion) events. Some of the theories that have been formulated to explain this phenomenon are the downward sloping demand curve hypothesis (Shleifer, 1986), the price pressure hypothesis (Harris and Eitan, 1986), the information cost hypothesis (Merton, 1987) and the signalling hypothesis (Jain, 1987; Dhillon and Johnson, 1991; Denis et al., 2003).

Under the first two hypotheses it is assumed that index inclusion and exclusion events do not contain information and thus do not affect security prices. Changes in demand are rather the result from non-information-based portfolio allocation. According to the downward sloping demand curve hypothesis the increase in demand is believed to be permanent.

Whereas, the price pressure hypothesis believes that the increase in demand is of a temporary nature and that the price and volume impacts are also temporary.

In contrast, the other two hypotheses do assume that an event carries information and subsequently has an effect on the fundamental value of a security. Particularly, the

information cost hypothesis argues that index events leads to an increase in investor

awareness and decreases information searching costs, as the index makes more information available to investors. As suggested by the name, the signalling hypothesis believes that index events are interpreted by investors as signals that convey information regarding the future of a security, due to private information possessed by the index institution that results in these

(13)

13 events. Thus, other things being equal, an expected increase (decrease) in the future value of the security leads to an increase (decrease) in share price.

Hence the first hypothesis:

H1a: Firms added to the Dow Jones Sustainability Europe Index, experience a positive market reaction to the news of their inclusion.

And subsequently:

H1b: Firms removed from the Dow Jones Sustainability Europe Index, experience a negative market reaction to the news of their exclusion.

Next to determining the top 20% sustainable performers out of the largest 600 European companies, the DJSI also tracks CSP per industry group. The index has a total of 24 industry groups ranging from the automotive industry to food and beverage producers. For each industry, per year, an industry leader is appointed. These 24 firms have been recognized in having the highest sustainable performance, relative to the rest of their respective peer groups. The industry group leaders are based on the Dow Jones Sustainability Index worldwide, therefore the number of European firms that are industry group leader varies.

As suggested by the resource-based perspective of CSP theory, organizations that can generate sustainable competitive advantages by effectively controlling and manipulating their resources that are, rare, cannot be perfectly imitated, and for which no perfect substitute is available (Laurenço et al., 2012). Furthermore, external benefits of CSP are related to corporate reputation, which has been identified as one of the most important intangible resources that provide a firm sustainable competitive advantage (Roberts and Dowling, 2002; Orlitzky et al., 2003; Laurenço et al., 2012). With a result that companies with high CSP reputation are able to improve relations with external stakeholders such as customers, investors, bankers, suppliers, and competitors. One could make the argument that firms recognized as industry group leaders would experience these benefits.

Hence, the second hypothesis is:

H2: European firms recognized as industry group leader by the Dow Jones Sustainability Europe Index, experience a positive market reaction to the news of their achievement.

3. Data and methodology

This paper focuses on the relation between sustainability performance and market returns, from a European perspective. To examine this relation, the empirical analysis in this paper is based on the organizations included/excluded from the Dow Jones Sustainability Europe Index. The DJSI Europe and respective subsets track the performance of the top 20% of the STOXX Europe 600 Index. The STOXX Europe 600 Index is derived from the STOXX

(14)

14 Europe Total Market Index (TMI) and is a subset of the STOXX Global 1800 Index. With a fixed number of 600 components, the STOXX Europe 600 Index represents large, mid and small capitalization companies across 18 countries of the European region: Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom. (STOXX® Europe 600 Index).

This analysis considers an entities level of CSP as high (low) by inclusion (exclusion) in a reputation index, in this case DJSI Europe. This division is consistent with the analysis techniques used in prior studies (Artiach et al., 2010; Chih, Chih and Chen, 2010; Chueng, 2011; Laurenço et al., 2012; Ziegler and Schröder, 2010; Ziegler, 2012).

This section is organized as follows. First, a description of the research method is provided. Next, detailed reference is made to the sample selection and construction process, as well as data collection. Thirdly, the econometric models to be used are explained, and how they were constructed based on prior literature.

3.1 Research method

This research adopted a quantitative research method using standard event study

methodology. Unlike most research that is conducted with an event study methodology, this paper will use a two pronged approach. In the first stage standard event study methodology is used to determine whether the cumulative average abnormal returns (CAAR) for firms included (excluded) in the DJSI Europe witness a statistically significant increase (decrease) during the event period, which is consistent with prior literature (e.g. Bauer et al., 2005; Ziegler and Schröder, 2007; Curran and Moran, 2007; Cheung, 2011). This study goes one step further with a second approach, the second approach entails conducting a separate event study for firms that have been recognized as industry group leader.

3.2 Sample construction and data collection

As of 5th of July 2011, DJSI constituent data is no longer publicly available on the DJSI’s website (www.sustainability-indexes.com). Students can request access to historical components and weightings data by filling out and submitting an academic request and a Non-Disclosure Agreement. After having filled in and submitted such a form to SAM representatives, approval was obtained and access to relevant data was granted. The

information provided includes constituencies’ lists (composition) of the DJSI Europe over the period of interest.

Daily stock prices data were obtained from Compustat Global Security Daily included in Wharton Research Data Services (WRDS). Historical data of the STOXX 600 was obtained from the stoxx indices website (www.stoxx.com).

3.2.1 Sample and summary statistics for hypotheses H1a and H1b

Panel A of Table Idisplays the frequency of index inclusions and exclusions per year. There are 119 index inclusions and 98 index exclusions events. The total number of events per year

(15)

15 varies from its lowest (37) in 2011 to its highest (47) in 2013.

Panel B and Panel C provide some basic information about the sample firms. The sample consists of 166 firms that were added to or deleted from the DJSI Europe during the period of 2009–2013. The total number of firms (i.e., 166) is less than the total number of events (i.e., 215) because many firms such as AEGON N.V., ArcelorMittal, ING N.V., and Volkswagen were added and deleted in different years of the sampling period. These 166 firms come from various countries and different industry sectors. The largest country present in the sample is the United Kingdom with slightly less than a quarter of the total sample. France is the second largest country present in the sample, with Germany and the Netherlands taking third and fourth place, respectively.

There is also a large spread when looking at the different industry sectors, the majority of firms represented in the sample are from the industrial goods and services sector. The second largest group are banks, insurance providers and other financial services firms, while all other sectors do not come above 10% of the sample, respectively.

Table I.

Number of inclusion and exclusion firms in the sample by year, country and type Panel A

Year No. of index inclusions No. of index exclusions Total

2009 19 25 44 2010 29 16 45 2011 23 14 37 2012 18 24 42 2013 29 18 47 Total 119 98 215 Panel B

Country No. of firms %

Belgium 2 1.20 Denmark 2 1.20 Finland 8 4.82 France 36 21.69 Germany 16 9.64 Italy 9 5.42 Netherlands 14 8.43 Norway 7 4.22 Portugal 2 1.20 Spain 14 8.43 Sweden 6 3.61 Switzerland 10 6.02 United Kingdom 40 24.10 Total 166 100.00

(16)

16 Panel C

Industry No. of firms %

Automobiles & Parts 4 2.41

Banks 14 8.43

Basic Resources 8 4.82

Chemicals 8 4.82

Construction & Materials 8 4.82

Financial Services 5 3.01

Food & Beverage 5 3.01

Health Care 5 3.01

Industrial Goods & Services 33 19.88

Insurance 8 4.82

Media 8 4.82

Oil & Gas 9 5.42

Personal & Household

Goods 9 5.42

Real estate 5 3.01

Retail 10 6.02

Technology 7 4.22

Telecommunications 5 3.01

Travel & Leisure 6 3.61

Utilities 9 5.42

Total 166 100.00

3.2.2 Sample and summary statistics for hypotheses H2a and H2b

The sample for H2a is somewhat smaller than that of H1a and H1b, Panel A of Table II displays the total number of European industry group leaders per year. Over the period 2009 to 2013 there were a total of 65 European firms that had been appointed as group leader by the DJSI. The total number of events per year varies from its lowest 11 in 2011 to generally 14.

Panel B and Panel C provide some basic information about the sample firms. The sample consists of 34 firms that were recognized as the top sustainable performers for their respective industries during the period of 2009–2013. The total number of firms (i.e., 34) is less than the total number of events (i.e., 65) because a number of firms such as Air France-KLM, BMW AG, Roche Holding AG and Swiss Reinsurance Co. were group leaders for a subsequent number of years during the sampling period. These 34 firms come from various countries and different industry sectors. The largest country present in the sample is Germany, with the Netherlands being a close second, representing roughly 21% and 18% of the sample respectively. As with the sample of the first hypothesis the industrial goods & services sector represents the largest industry group in the sample. However, the spread between the industry groups is much smaller for this sample as there can be only one leader per industry.

(17)

17 Table II.

Number of European industry group leaders in the sample by year, country and type

Panel B

Country No. of firms %

Belgium 1 2.94 Finland 2 5.88 France 4 11.76 Germany 7 20.59 Netherlands 6 17.65 Portugal 1 2.94 Spain 4 11.76 Switzerland 4 11.76 United Kingdom 5 14.71 Total 34 100.00 Panel A

Year No. of European industry group leaders

2009 14 2010 12 2011 11 2012 14 2013 14 Total 65

(18)

18 Panel C

Industry No. of firms %

Automobiles & Parts 2 5.88

Basic Resources 2 5.88

Chemicals 2 5.88

Food & Beverage 3 8.82

Health Care 1 2.94

Industrial Goods & Services 4 11.76

Insurance 2 5.88

Media 2 5.88

Oil & Gas 3 8.82

Personal & Household

Goods 3 8.82

Retail 1 2.94

Technology 3 8.82

Telecommunications 1 2.94

Travel & Leisure 2 5.88

Utilities 3 8.82

Total 34 100.00

3.3 Model specification

The sample period is from 2009 to 2013.The following data has been provided by the DJSI Europe and RobecoSAM:

(i) The announcement day of index inclusion and index exclusion events; (ii) The effective day of index exclusion and index inclusion events; (iii) The names of the above companies; and

(iv) Industry group leader firms

Standard event study methodology is used to compare these variables before and after index inclusion (or index exclusion) events. Two sets of event days are used: the announcement days1_{(AD) and the days of change (CD). The length between AD and CD is generally 5} trading days for the years 2010, 2011, 2012 and 2013, with the exception of 10 trading days in 2009.

Following Cheung (2011) the sample is first divided into two periods. The first period is called the estimation period that contains observations from t = ‒250 to t = ‒16 in order to calculate βi (see model 4), while the second period is labelled the event period that starts from t = 0 to t = 60 where a relevant window is examined. For each security, the complete event window runs from 15 days before AD through to 60 days after CD. Based upon research conducted by Lynch and Mendenhall (1997) and Chueng (2011), the complete window is

1_{Every announcement is first published on DJSI website on announcement day at 06:00 GMT and the press}

release is sent to the media on the same day at 07:15 GMT in line with the Swiss market regulations. This means that the announcement is made known to the European financial markets before opening the same day.

(19)

19 further divided into seven sub-windows, which are designed to assess different aspects of stock behaviour around the events. In addition to the standard AD and CD windows, these sub-windows include:

(1) Pre-announcement window that lies between AD‒15 and AD‒1.

(2) Run-up window that spans from the day after AD through to the day before CD.

(3) Three release-related windows that run from CD to CD+10; they include release windows during CD and CD+4 and a post release windows (CD+5, CD+10), respectively.

(4) Temporary price impact windows that cover periods within AD–15 and CD+10. (5) Total permanent price impact windows that fall within AD–15 and CD+60.

The pre-announcement window aims to detect the existence of an anticipation effect before the announcement, while the run-up window is used to test for possible price changes between AD and CD. The three release-related windows allows for an examination ofthe short-term impact of CD on stock prices. The final two sets of windows distinguishes between temporary price changes and permanent ones.

The proxy used for stock price movements is abnormal stock returns. The necessary data are collected from Compustat Global - Security Daily and the CRSP Stock Daily.

Abnormal return of stock i at time t is defined as the difference between realized return and an estimate of its expected (or normal) return in the absence of the event.

𝐴𝑅

_𝑖𝑡

= 𝑅

_𝑖𝑡

– 𝐸(𝑅

_𝑖𝑡

)

Cumulative abnormal returns (CAR) during the event period (t1, t2) are given by:

𝐶𝐴𝑅

_(𝑡₁_,𝑡₂₎

= ∑ 𝐴𝑅

_𝑖,𝑡

𝑡₂

𝑡=𝑡₁

The Cumulative average abnormal returns (CAAR) is calculated as follows:

𝐶𝐴𝐴𝑅

_(𝑡₁_,𝑡₂₎

=

1 𝑁

∑ 𝐶𝐴𝑅(𝑡

1

, 𝑡

2

)

𝑁

𝑖=1

Abnormal returns (cumulative abnormal returns) captures the excess returns an investor would have received over a particular event day (event period) if one had invested in security i. It is assumed that abnormal returns (ARit) are (normally) distributed with mean zero and variance

𝜎

_𝑖2 (MacKinlay, 1997)

.

Following MacKinlay (1997), to compute E(Rit), the market model is used. The market model is a statistical model which relates the return of any given security to the return of the market portfolio. The model's linear specification follows from the assumed joint normality of asset returns. For any security i the market model is:

(1)

(2)

(20)

20

𝐸(𝑅

_𝑖𝑡

) = 𝛼

_𝑖

+ 𝛽

_𝑖

𝑅

_𝑚𝑡

+ 𝜀

_𝑖𝑡

Where:



𝑅

_𝑚𝑡 is the market portfolio



𝜀

_𝑖𝑡 is the zero mean disturbance term;



𝛼

_𝑖

, 𝛽

_𝑖 are the parameters of the market model determined through an OLS regression. Generally, in applications a broad based stock index is used for the market portfolio, with the S&P 500 Index, the CRSP Value Weighted Index, and the CRSP Equal Weighted Index being popular choices (MacKinlay, 1997). This study uses the STOXX Europe 600 as the

benchmark mark, as this paper is focussed on a European context and all firms included in the analyses are represented in the STOXX Europe 600. The market model is one of the most popular methods for computing expected returns in an event study setting. Furthermore, it represents a potential improvement over the other popular methods, such as the constant mean return model. By removing the portion of the return that is related to variation in the market's return, the variance of the abnormal return is reduced. This in turn can lead to increased ability to detect event effects (MacKinlay, 1997). Finally, in a study examining the various models used in an event study context, Ahern (2009) found that the use of multifactor models such as the Fama and French 3 factor model (Fama and French, 1992), and the Carhart 4 factor model (Carhart, 1997) does not decrease the forecast error bias that might be present in simpler methods. Therefore consistent with prior literature (e.g. Freedman and Patten, 2004; Curran and Moran, 2007; Consolandi et al., 2009) the market model was chosen.

In order to test for the significance of cumulative average abnormal returns over the event period, a parametric test, the cross-sectional t statistic and a non-parametric test, Corrado and Zivney’s (1992) sign statistic are used, as recommended by Ahern (2009).

4. Empirical results

This section is organized as follows. First, the descriptive statistics for the (sub)-samples are provided. Second, the results of the three event studies conducted are presented and

interpreted.

4.1 Descriptive statistics

Table III contains several descriptive statistics for the overall sample and each of the

subsamples. Panel A shows the average daily returns and standard deviation, per subsample, per year, whereas Panel B shows additional summary statistics, such as skewness and

kurtosis. When looking at the average daily returns on a yearly basis in Panel A, at first glance there doesn’t seem to be that much of a difference between the inclusion and exclusion stocks, with the exception of 2013. However, when looking at Panel B, it becomes clear that while the samples are approximately symmetric (i.e. the skewness is within normal boundaries), the

(21)

21 kurtosis is rather high, which might suggest non-normal data. This justifies the use of the sign-test statistic for hypothesis testing, as it has proven to be highly suitable for testing with non-normal data (Corrado and Zivney, 1992).

Table III.

Descriptive statistics H1a & H1b

Panel A

Avg. daily returns Std. deviation

Incl. Excl. Incl. Excl.

Year (%) (%) (%) (%) 2009 0.120 0.129 1.943 2.165 2010 0.151 0.112 1.627 1.554 2011 ‒0.083 ‒0.085 3.657 3.540 2012 0.060 0.099 1.557 1.826 2013 ‒0.023 0.057 1.460 1.397

This reports summary statistics of daily return for DJSI Europe stocks used in the sample. The summary statistics are the averages of estimates (mean, standard deviation, skewness and kurtosis).

Table IV for hypothesis H2 generally shows the same picture as for the samples used for hypotheses H1a & H1b, albeit with overall higher average daily returns, which preliminary might suggest that investors react stronger to firms that have been recognized as industry group leaders than those included or excluded from the index. Interesting is that it seems that 2011 was a bad year for all firms in the sample as the average daily returns were negative for all sub-samples during this year. Again the sample has a high kurtosis, only this time

combined with a moderately skewed distribution, which might suggest non-normal data. Once more justifying the use of the sign-statistic for hypothesis testing, especially combined with the smaller sample for this hypothesis.

Panel B

Types Index Inclusions Index Exclusions Whole sample

Daily return

Mean 4.48e–04 7.28e–04 5.64e–04

SD 2.16e–02 2.13e–02 2.15e–02

Skewness ‒0.122 0.073 ‒0.042 Kurtosis 8.433 7.614 8.100

(22)

22 Table IV.

Descriptive statistics H2 Panel A

Avg. daily returns Std. deviation

Year (%) (%) 2009 0.140 1.596 2010 0.167 1.576 2011 ‒0.004 2.708 2012 0.151 1.789 2013 0.142 1.533 Panel B

Types Industry group leaders (whole sample) Daily Return Mean 1.29E–03 SD 1.82E–02 Skewness 0.563 Kurtosis 7.469

This reports summary statistics of daily return for DJSI Europe stocks used in the sample. The summary statistics are the averages of estimates (mean, standard deviation, skewness and kurtosis).

4.2 Index inclusion and exclusion results (H1a & H1b)

In order to ensure that market responses to the events are isolated to the two separate

hypotheses, all index constituent stocks are classified into two groups. The firs group contains those stock newly included to the DJSI Europe, while the second contains those stocks newly deleted from the index.

The trading behaviour is examined separately to see whether there is abnormal trading behaviour after the announcement day (or the day of change). Panel A of Table V shows the cumulative average abnormal returns (CAARs) for the first five trading days after the index inclusion events. Two observations can be made in relation to index inclusion stocks. The first is that there is no evidence to support there is an announcement effect present, due to that (1) the overall result of (AD, AD+4) window is not statistically significant and (2) in other daily windows, no significant results are found. The second observation relates to the CD windows. Here there is some weak statistical evidence of an effective change effect. For the (CD, CD) window both the sign-test as well as the t-test for CAAR are significant, suggesting that investors respond negatively on the day new stocks are included to the DSJI Europe. Furthermore the (CD+3, CD+3) also had a negatively significant sign-test.

When looking at newly excluded stocks, the following observations can be made. Firstly, Panel B of Table V reports that the CAAR in the first 5 trading days after the announcement of their exclusion is made, is statistically not distinguishable from zero.

(23)

23 However, when looking at the day of change it can be seen that the CAAR is significant for (CD+2, CD+2), (CD+3, CD+3) and (CD+4, CD+4) for exclusion stocks. With the first being positively significant and the latter two negatively, which might hint at lagged negative response to the news from the market.

Table V.

Cumulative average abnormal returns (CAARs) in the first 5 days after AD (or CD) Period

2009-2013 Panel A: index inclusions Panel B: index exclusions Event window CAAR

(%) Sign-test statistic t-test CAAR (%) Sign-test statistic t-test (AD, AD) ‒0.171 ‒1.614 ‒1.624 ‒0.203 ‒1.188 ‒0.918 (AD+1, AD+1) ‒0.005 ‒0.861 ‒0.046 ‒0.086 ‒0.517 ‒0.959 (AD+2, AD+2) 0.066 0.268 0.603 0.059 0.154 0.635 (AD+3, AD+3) 0.174 ‒0.297 0.945 ‒0.009 ‒0.294 ‒0.164 (AD, AD+4) 0.127 1.397 0.604 0.141 1.495 0.374 (CD, CD) ‒0.257 ‒2.179** ‒2.234** ‒0.128 ‒0.741 ‒1.396 (CD+1, CD+1) 0.011 0.832 0.100 0.244 ‒0.741 1.300 (CD+2, CD+2) 0.104 0.079 0.787 0.140 2.837*** 0.766 (CD+3, CD+3) ‒0.029 ‒1.802* ‒0.206 ‒0.279 ‒1.188 ‒2.826*** (CD+4, CD+4) 0.006 0.079 0.056 ‒0.205 0.377 ‒1.696* (CD, CD+4) 0.025 ‒0.297 0.118 0.260 1.272 0.637

This presents CAARs in the event window for two different types of stocks in our sample. The whole sample consists of 119 index inclusion stocks and 98 index exclusion stocks. CAAR is the cumulative cross-sectional average of the market model adjusted stock returns. The abnormal return for stock i on day t (ARit) is calculated as follows: ARit = Rit - E(Rit) where Rit is the return on stock i at day t. E(Rit) is the daily expected return from the market model. The table is divided into two main panels. The announcement day (AD) is taken to be day 0 since the announcements are made on AD before European stock markets are open. The first column specifies the days in the event window, where AD (CD) is the actual day on which the actual announcement (change) takes place. The second and third columns show the CAARs for the subsamples, respectively. The second and fourth column shows the sign statistic, while the third and sixth columns provide the cross-sectional t test result. *, **, and *** denote significance at the 10, 5 and 1% level, respectively.

A graphical representation of the time-series behaviour of CAARs in the complete window spanning from AD‒15 to CD+60 is presented in Figure 1. This gives a clearer picture as the data presented in Table V, as this focuses on the impact of an event in the first five trading days only.

A couple of observations can be made in regard to index inclusion stocks. Firstly, there seems to be an anticipation effect due to the fact that the CAAR starts increasing from a negative territory a few days before the announcement. However, there is a loss in momentum as the announcement is made as well as the following day. Secondly, the day of change

(24)

24 Relative Day to the Announcement Day (Day 0)

Figure 1. Cumulative average abnormal returns (CAARs) impact can be clearly seen when looking at the CAARs roughly 6 days after AD2_. Furthermore, it is quite clear that index addition stocks experience an upward price movement. However, this is a highly temporary effect because prices decrease later in the period, which might suggest a price reversal effect.

As for index exclusion stocks, Figure 1 shows that CAARs of index exclusions take a dive after the announcement date. There seems to be a slight movement to recovery roughly 18 days after but AD, but relapse further downward. As was the same with the inclusion stocks the day of change impact can clearly be seen. However, where there was some

evidence of price reversal later on in the period for inclusion stocks, this those not seem to be the case for the index inclusions. This could indicate that the market reacts more heavily on stocks that were excluded from the index compared to those that were included.

One should note that the analysis on Figure 1 does not distinguish the impact of AD from that of CD, and does not indicate whether the persistence or price reversal is statistically significant or not. In order to counter these limitations a detailed analysis on shorter windows is shown in Table VI.

Panel A of Table VI reports that the CAAR on the pre-AD (AD–10, AD–1) window is

positive but statistically insignificant suggesting that the market does not anticipate the events, rather than what was hinted at in Figure 1. The same can be seen when taking a closer look at various AD windows with lengths ranging from 1, 2 to 3 days before or after the

announcement day. While some such as (AD–2, AD+2) and (AD–3, AD+3) show positive CAARs, they are statistically insignificant.

In the run-up window that covers the day after the announcement up to the day before the effective change also no statistically significant CAAR can be found, this while the CAAR is relatively high percentage wise (0.454%) when compared to the other windows.

The same observation applies when it comes to the other run-up windows as well as

2_{Please note that the length between AD and CD varies, from 10 trading days in 2009 to 5 for the other years.}

This means that CD must occur within 10 trading days after AD. -3,00% -2,50% -2,00% -1,50% -1,00% -0,50% 0,00% 0,50% 1,00% 1,50% 0 5 10 15 20 25 30 35 40 45 50 55 C AA R s Index inclusions Index exclusions

(25)

25 the release (CD, CD+4) and post-release window (CD+5, CD+10).

Two sets of windows are used to examine if there is a temporary price effect and two for a permanent price effect. The first two set lies between the announcement day and 10 days after the change date and 15 days before AD and 10 days after CD, respectively. The

permanent price windows are much longer in length, the first spanning from AD to CD+30 and the second covering from 15 days before AD up to 60 days after CD. For the temporary prices, again no significant CAARs can be found for these two sets of windows, while yet again the CAARs themselves are highly positive. However, the first price window does show strong statistical evidence for a positive CAAR movement, it is only at the second permanent price window that the CAAR switches to negative with weak statistical significance. Which could support what was seen in Figure 1. That the overall evidence on stock prices suggests that price changes that are found on AD and CD windows are temporary and do not last long for index inclusion stocks.

Table VI.

Cumulative average abnormal returns in smaller event windows Period

2009-2013 Panel A: index inclusions Panel B: index exclusions Event window CAAR

(%) Sign-test statistic t-test CAAR (%) Sign-test statistic t-test (AD–10, AD–1) 0.046 1.397 0.126 ‒0.583 ‒0.517 ‒1.831* (AD–1, AD+1) ‒0.332 ‒1.426 ‒1.804* ‒0.531 ‒0.965 ‒2.934*** (AD–2, AD+2) 0.178 ‒0.109 0.802 ‒0.184 ‒0.741 ‒1.037 (AD–3, AD+3) 0.047 0.268 0.189 ‒0.231 ‒1.188 ‒0.786 (AD+1, CD–1) 0.421 0.644 1.146 ‒0.178 0.824 ‒0.319 (CD–1, CD+1) ‒0.188 ‒0.485 ‒1.167 ‒0.028 ‒0.157 ‒0.441 (CD–2, CD+2) 0.165 0.456 0.666 ‒0.064 0.824 ‒0.152 (CD–3, CD+3) 0.073 ‒0.297 0.200 0.083 0.154 0.145 (CD, CD+4) 0.025 ‒0.297 0.118 0.260 1.272 0.637 (CD+1, CD+10) 0.454 0.268 0.825 ‒0.137 ‒0.517 ‒0.478 (CD+5, CD+10) 0.211 ‒0.109 0.500 ‒0.390 ‒0.741 ‒1.905* (AD, CD+10) 0.550 1.585 0.842 ‒0.168 ‒0.517 ‒0.329 (AD–15, CD+10) 0.683 1.020 0.829 ‒0.447 ‒0.517 ‒0.597 (AD, CD+30) 1.080 2.149** 1.258 ‒0.204 ‒0.070 ‒0.214 (AD–15, CD+60) ‒2.863 ‒0.109 ‒1.820* ‒2.238 ‒0.070 ‒1.260

The first column specifies the days in the event window, where AD (CD) is the actual day on which the actual announcement (change) takes place. The second and third columns show the CAARs for the subsamples, respectively. The second and fourth column shows the sign statistic, while the third and sixth columns provide the cross-sectional t test result. *, **, and *** denote significance at the 10, 5 and 1% level, respectively.

Panel B of Table VI provides the results of other sub-window analysis on index exclusion stocks. Firstly, the results on the pre-announcement (pre-AD) window show that CAAR is negative and statistically significant, this is also the case for the first AD window (AD–1,

(26)

26 AD+1), suggesting that there is an anticipation effect for newly excluded stocks. However, there is no evidence of an announcement effect because CAARs are negative and statistically insignificant across the other the AD windows. Thirdly, the analysis on the run-up window and the CD windows, show the same results. They are all mixed in terms of negative and positive CAARs, but are all statistically insignificant. The same goes for the CAAR for the release window and the first post-release window (CD+1, CD+10), while negative,

statistically insignificant. The final post-release window (CD+5, CD+10), however, does show weak statistical significance for negative CAAR. Which might suggest the same as found for Table V, a lagged market response on the change date for index exclusion stocks. Contrary to the findings for the longer windows (i.e., temporary price and permanent price windows) for the inclusion stocks, while the CAARs for exclusion stocks are not statistically significant, there is no evidence to believe that there is a price reversal effect for exclusion stocks. Meaning that the limited evidence found that markets penalize firms for being excluded from the index are later reversed.

4.3 Industry group leader results (H2)

The second hypothesis looks at how the market reacts to the announcement of the yearly industry group leaders recognized by the DJSI Europe. These are firms that have been

recognized by the index as the top CSP performers for their respective industry groups. Please note that while these firms are announced on the same day as the inclusion and exclusion firms, they do not experience a change day per se. However, for the sake of comparability between the different hypotheses, the event windows are classified the same as those for hypotheses H1a and H1b. Change day windows should thus be seen as extensions of

announcement day windows i.e. CD falls between AD+5 and AD+10 (see section 3.3 Model specification).

Table VII reports on the CAARs in the first 5 trading days after AD and CD for industry group leader stocks, while a general positive trend can be seen for the group leaders it is clear that there isn’t a case for a short term announcement effect or change effect. As the overall results of the (AD, AD+4) and (CD, CD+4) windows are not statistically significant and in the other daily windows, also no significant results are found.

(27)

27 Table VII.

Cumulative average abnormal returns (CAARs) in the first 5 days after AD (or CD) Period

2009-2013 Industry group leaders

Event window CAAR (%) Sign-test statistic t-test

(AD, AD) ‒0.094 ‒1.081 ‒0.590 (AD+1, AD+1) 0.008 0.179 0.048 (AD+2, AD+2) 0.311 0.935 1.914 (AD+3, AD+3) ‒0.037 0.431 ‒0.234 (AD, AD+4) 0.335 1.187 0.743 (CD, CD) 0.149 0.179 1.031 (CD+1, CD+1) ‒0.088 0.431 1.018 (CD+2, CD+2) 0.170 0.683 ‒0.445 (CD+3, CD+3) 0.156 0.683 1.060 (CD+4, CD+4) ‒0.127 ‒1.585 0.847 (CD, CD+4) 0.259 ‒0.073 0.644

This presents CAARs in the event window for the industry group leader sample. The whole sample consists of 65 stocks. CAAR is the cumulative cross-sectional average of the market model adjusted stock returns. The abnormal return for stock i on day t (ARit) is calculated as follows: ARit = Rit - E(Rit) where Rit is the return on stock i at day t. E(Rit) is the daily expected return from the market model. The announcement day (AD) is taken to be day 0 since the announcements are made on AD before European stock markets are open. The first column specifies the days in the event window, where AD (CD) is the actual day on which the actual announcement (change) takes place. The second and shows the CAARs for the subsamples. The third column shows the sign statistic, while the fourth column provides the cross-sectional t test result. *, **, and *** denote significance at the 10, 5 and 1% level, respectively.

A graphical representation of the time-series behaviour of CAARs in the complete window spanning from AD‒15 to CD+60 is presented in Figure 2. This gives a clearer picture of the data presented in Table VII.

A number of observations can be made in regard to in regard to industry group leader stocks. Firstly, there seems a hint of an anticipation effect due to the fact that the CAAR starts increasing from a slightly negative territory a few days before the announcement. Secondly, unlike with the index inclusion stocks there doesn’t seem to be any loss momentum for the group leader stocks. Finally, while there are some dips along the event period, the abnormal returns show continuous growth, suggesting that the market rewards firms that have been recognized as industry group leader. An analysis on shorter windows is provided in Table VIII, in order to determine whether this continuous growth is statistically significant.

(28)

28 Figure 2. Cumulative average abnormal returns (CAARs)

Relative Day to the Announcement Day (Day 0)

Unlike what was expected from Figure 2, Table VIII reports that the CAAR on the pre-AD (AD–10, AD–1) window is positive but statistically insignificant suggesting that the market does not anticipate the announcement of industry group leaders. The same can be seen when taking a closer look at various AD windows with lengths ranging from 1, 2 to 3 days before or after the announcement day. While some such as (AD–2, AD+2) and (AD–3, AD+3) show positive CAARs, they are statistically insignificant.

Starting with the run-up window that covers the day after the announcement up to the day before the effective change significant CAARs can be found, which was hinted by Figure 2, suggesting that the market reacts to the recognition of industry group leaders after the announcement of their achievement is made.

The same observation applies when it comes to the two of the run-up windows (CD–1, CD+1) and (CD–3, CD+3) as well as the post-release windows (CD+1, CD+10) and (CD+5, CD+10). They all have positive CAARs combined with significant sign- and/or test-statistics.

As with H1a and H1b two sets of windows are used to examine if there is a temporary price effect and two for a permanent price effect. The first temporary price window that lies between the announcement day and 10 days after the change date also shows a positive CAAR, combined with both a positive significant sign- as well as test-statistic, confirming what was seen in Figure 2, that there is a positive price effect relating to the event. The two permanent price windows report the same results. They both have very high CAARs, and are highly significant statistically, proving that unlike with index inclusion and exclusion events, these events have a permanent price impact and witness no price reversal. Providing empirical evidence that the market does respond positively to firms that have been recognized in having ‘high’ CSP. -1,00% -0,50% 0,00% 0,50% 1,00% 1,50% 2,00% 2,50% 3,00% 3,50% 4,00% 4,50% 5,00% 5,50% 0 5 10 15 20 25 30 35 40 45 50 55 60 65 C AA R s

(29)

29 Table VIII.

Cumulative average abnormal returns in smaller event windows Period

2009-2013 Industry group leaders

Event window CAAR (%) Sign-test statistic t-test

(AD–10, AD–1) 0.196 0.431 0.377 (AD–1, AD+1) ‒0.246 ‒0.829 ‒1.157 (AD–2, AD+2) 0.293 0.683 0.949 (AD–3, AD+3) 0.484 0.935 1.270 (AD+1, CD–1) 0.769 1.439 2.414** (CD–1, CD+1) 0.350 1.439 1.934* (CD–2, CD+2) 0.196 0.935 0.817 (CD–3, CD+3) 0.621 2.699** 2.128** (CD, CD+4) 0.259 ‒0.073 0.644 (CD+1, CD+10) 0.611 1.691* 1.394 (CD+5, CD+10) 0.703 1.187 1.870 (AD, CD+10) 1.333 2.195** 2.079** (AD–15, CD+10) 0.862 0.683 0.961 (AD, CD+30) 4.790 2.447** 3.107** (AD–15, CD+60) 4.355 1.691* 2.086**

The first column specifies the days in the event window, where AD (CD) is the actual day on which the actual announcement (change) takes place. The second and shows the CAARs for the subsamples. The third column shows the sign statistic, while the fourth column provides the cross-sectional t test result. *, **, and *** denote significance at the 10, 5 and 1% level, respectively.

5. Additional Analyses

In order to ensure that the results given in the previous section are not biased or driven by specific factors, this section of the paper provides two sets of additional analysis. Firstly, a ‘crisis effect’ test was conducted, in order to investigate whether there is a difference in reaction by investors, to index inclusion and index exclusion before the financial crisis started in 2008. Secondly, robustness checks were conducted on the main hypotheses.

This section is organised as follows. First an identical event study to that of H1a and H1b was conducted, but this time for the period 2004–2007, with an analysis of its results. Finally, the results of two robustness tests are discussed.

5.1 Crisis effect test

Starting with index inclusions, Panel A of Table IX shows the cumulative average abnormal returns (CAARs) for the first five trading days after the index inclusion events. Two

The relation between sustainability performance and stock market returns : an empirical analysis of the Dow Jones Sustainability Europe Index