Does the market value corporate sustainability?
An event-study analysis with emerging market companies.
Name: Rosanne Korpershoek
Student number: 10646051
Programme: Economics and Business Track: Finance and Organisation Supervisor: R. C. Sperna Weiland
2 Statement of Originality
This document is written by Student Anne Hendrika Korpershoek who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
3 Abstract
This thesis empirically analyses the impact of index inclusion on sustainable companies by studying a sample of emerging market companies that were added to the Dow Jones Sustainability Emerging Markets Index over the period 2013-2015. The impact is measured in terms of stock return and systematic and idiosyncratic risk by means of an event study. The event study presents significant negative cumulative average abnormal returns around the announcement. The cumulative average abnormal returns subsequently revert to pre-announcement levels, which provides support of the price pressure hypothesis. Additionally, a larger percentage of the sample experiences a decrease in risk rather than an increase. However, not enough evidence is found to provide compelling support for the hypothesis that inclusion reduces risk. Taken together, the results indicate negative valuation of corporate sustainability in emerging markets.
Keywords: sustainability, emerging markets, Dow Jones Sustainability Emerging Markets Index, DJSI, index inclusion, event study
4 Table of contents
Abstract ... 3
1 Introduction ... 5
2 Literature review & hypotheses ... 6
2.1 Corporate sustainability... 6
2.2 Index inclusion effects ... 8
2.3 Emerging markets ... 10
2.4 CS-risk relationship ... 12
3 Methodology and data... 15
3.1 Dow Jones Sustainability Emerging Markets Index ... 15
3.2 Data ... 17 3.3 Event-study methodology ... 18 3.4 Risk measures ... 20 4 Results ... 22 4.1 Summary statistics... 22 4.2 Abnormal returns... 22 4.3 Risk... 24 4.4 Robustness checks ... 26
5 Discussion and conclusion ... 29
Appendix ... 32
5 1 Introduction
Renneboog, Ter Horst, and Zhang (2008) have noted the extraordinary growth of socially responsible investments (SRI). According to Renneboog et al. (2008) total assets under management of SRI portfolios in the US have grown 1200 percent in the decade preceding the year 2005. This extraordinary growth reflects the increasing awareness of investors to social, environmental, ethical, and corporate governance issues. With the increasing public interest in sustainability, the question arises whether this public interest is shared by investors. Furthermore, do these investors appreciate it when companies invest time and resources into sustainable activities?
Various studies assess the valuation of corporate sustainability performing an event study on inclusion in a sustainability index. Most studies have examined the inclusion effect for US or European firms. Despite the enormous flow of portfolio investments to emerging markets that was already mentioned by De Santis & İmrohoroǧlu (1997), no event study has yet been performed on sustainability index inclusion of emerging market firms. This may be due to the relatively recent launch of the first index for emerging markets, the Dow Jones Sustainability Emerging Markets Index (DJSI EM). Previously no data was available to perform such a study. The DJSIEM was launched in 2012, hence inclusion data is now available for three subsequent years. This thesis therefore aims to answer the question: Does the stock market value the inclusion of emerging market companies in a sustainability index?
This thesis takes a slightly different approach to answering this question. Similar to other studies a standard event-study analysis will first be performed to assess the presence of abnormal returns following inclusion in the DJSI EM. In line with Cheung’s (2011) opinion, the analysis of the profitability of stocks is not complete without taking into account the associated risks. The second part of the analysis therefore consists of a comparison between the levels of risk, both systematic and idiosyncratic, before and after the announcement of inclusion.
The remainder of this thesis is organised in the following way: Section 2 provides an in-depth analysis of the literature related to this subject and culminates in the formulation of the hypotheses. Following that, the research methodology is described and a brief overview of the sample is given. Then the results are presented in section 4. This thesis then concludes with a discussion of the results and some limitations and ideas for future research.
6 2 Literature review & hypotheses
This thesis analyses the effect of inclusion in a sustainability index on stock performance to determine whether investors value corporate sustainability. To come up with well-founded hypotheses it is important to establish what exactly is meant with corporate sustainability. Moreover, what does the literature put forward on index inclusion effects in general and sustainability indices in specific. As this thesis focuses on companies from emerging markets, this section will also discuss the theory in specific application to emerging markets. Lastly, an overview of the literature on risk related to sustainable investments will be given.
2.1 Corporate sustainability
In the literature on sustainability practices performed by companies multiple terms seem to be used. Some scholars use the term Corporate Social Responsibility (CSR), whereas others prefer to refer to these practices with the term Corporate Sustainability (CS). Indeed, more often than not scholars use both terms interchangeably. Van Marrewijk (2003) has made an effort to provide an overview of the meaning of the two terms, their relationship and their separate definitions. The first proposal mentioned by Van Marrewijk (2003) argues that the ambiguity associated with this concept is of linguistic origin. This view originates from Göbbels (as cited in Van Marrewijk, 2003) and his proposed solution is to rename the concept to corporate societal accountability. Van Marrewijk (2003) then justifiably remarks that it will be difficult for everyone involved to get used to this new notion. The following proposal argues that CS can be viewed as an all-encompassing concept including corporate responsibility (CR). Corporate responsibility (CR) is said to consist of three dimensions: economic, environmental, and social responsibility. CSR is thus one of the dimensions within the greater sphere of CS. The last proposal discussed by Van Marrewijk (2003) states that CS and CSR are merely two sides of one coin. The general conclusion that can be drawn from the three proposals is that CS is either equal to CSR or transcends CSR. Therefore the term used in this thesis will be CS as this will at least encompass as much as or more than the notion of CSR.
The next important issue to address is the relationship between CS and financial performance (FP). The two key theoretic perspectives are shareholder theory and stakeholder theory. Shareholder theory is often viewed as the traditional perspective and attributed to Milton Friedman. Friedman (1962) states that a company’s only social responsibility is to its shareholders by maximising the net present value of future profits. Any activities that are not purely aimed at profit maximisation will therefore negatively affect the company’s financial
7 performance and harm shareholders. Friedman argues his case in the context of the fundamental theorems of welfare economics. Not only is there no justification for companies to engage in CS, but this means CS should also not be necessary as the resource allocation is already Pareto efficient (Lerner, 1944). However, an important underlying assumption is the absence of externalities. When externalities are indeed present, Greenwald and Stiglitz (1986) have shown that the resource allocation is not Pareto efficient. Therefore, shareholder theory need not be seen as being at odds with CS (Heal, 2005). CS can then be viewed as activities aiming to reduce these externalities which will in turn improve companies’ profitability. Correspondingly, Besley and Ghatak (2007) show with their model for for-profit companies operating in competitive markets, that CS is in line with profit-maximisation. Moreover, they find that more responsible companies earn higher profits. Besley and Ghatak (2007) attribute this to a reputational premium these companies receive as support for their good behaviour.
Stakeholder theory uses the reduction of externalities in a similar way to argue for a positive effect from CS on FP. Stakeholder theory, pioneered by Freeman (1984), goes beyond shareholder theory by stating that companies not only have a responsibility towards their shareholders, but also to other interest groups. Companies have an ethical obligation to balance the interests of their stakeholders (Garriga & Melé, 2004).
This rather broad concept has been refined by Donaldson and Preston (1995). They have made a division into three distinct aspects: normative, descriptive/empirical, and instrumental. The normative aspect encompasses the fundamental basis of stakeholder theory: Stakeholders are all those people with an interest in the affairs of the company and their interests have intrinsic value. The descriptive/empirical aspect finds justification for stakeholder theory by describing the perspectives on CS held by actual managers (de la Cruz Déniz-Déniz & Garcia-Falcon, 2002; Quazi & O’Brien, 2000; Rojšek, 2001). Lastly, the instrumental aspect of stakeholder theory entails the examination of the relationship between the practice of stakeholder management and corporate performance. For instance, Allen, Carletti, and Marquez (2007) show that societies with stakeholder-oriented companies have higher firm values compared with shareholder-oriented societies. The prospect of higher firm values may induce companies to voluntarily adopt a stakeholder orientation. Similarly, Orlitzky, Schmidt, and Rynes (2003) provide confirmation of the positive association between CS and financial performance by performing a meta-analysis.
The idea behind the positive relationship between CS and financial performance is the following: CS allows a company to establish a reputation with its stakeholders. This reputation is built based on the honouring of the implicit contracts (also called self-enforcing relational
8 contracts) the company has with its stakeholders (Jones, 1995). A good reputation helps a company to achieve beneficial outcomes from negotiations with the various stakeholders. A company will be better able to attract and retain skilled employees but also to avoid certain negative externalities, such as boycotts or strikes, thereby increasing its profitability (Oberndorfer, Schmidt, Wagner, & Ziegler, 2013).
Donaldson and Preston (1995) also included a section regarding the managerial implications of stakeholder theory. The normative aspect requires managers to balance the interests of a multitude of stakeholders. This obligates managers to decide on necessary trade-offs among the various stakeholder interests. This is the foundation for Jensen’s (2001) critique of stakeholder theory. When managers have multiple objectives it becomes impossible to make reasoned decisions on trade-offs and it may induce opportunistic behaviour. Jensen (2001) therefore proposes a different version of stakeholder theory, namely enlightened stakeholder theory. It recognises the value-adding abilities of CS but emphasises that managers need a measure to make trade-offs. This measure should be a balanced scorecard based on long-term company market value maximisation. Enlightened stakeholder theory recognises the importance of stock value as a determinant of long-term company value, but points out that shareholders ‘are not some special constituency that ranks above all others’ (Jensen, 2001, p. 310).
2.2 Index inclusion effects
Based on the aforementioned stakeholder theory a positive relationship between CS and FP is expected but the question remains whether this will be borne out by the event that will be analysed. An extensive amount of literature has been produced on the effects of index inclusion in general. In this section five general theories regarding the effect of inclusion in an index will be discussed. Subsequently, this section will zoom in on the effects expected from inclusion in a sustainability index.
Both the downward sloping demand curve hypothesis (Shleifer, 1986) and the price pressure hypothesis (Harris & Gurel, 1986) assert that index inclusion events do not contain any new information. Therefore they should not have any effect on the share price. Shleifer (1986) finds evidence for the downward sloping demand curve by examining inclusions in the S&P 500 over nearly 20 years. His results show that the significant and permanent change in prices are due to permanent change in demand. In a similar manner, Harris and Gurel (1986) find compelling evidence for the price pressure hypothesis. However, the price change is not
9 permanent as the increase in demand is not of permanent nature either. Their results show a consistent reversal to pre-announcement share prices within on average three weeks.
In contrast to these first two theories, the next three theories claim that index inclusion events do indeed carry new information that will affect the fundamental value of the stocks. Dhillon and Johnson (1991) provide evidence that contradicts the results that Harris and Gurel (1986) and Shleifer (1986) found. They find that their evidence is consistent with the information hypothesis. Similarly, Denis, McConnell, and Ovtchinnikov (2003) find evidence that index inclusion events carry information. They subsequently argue in favour of the signalling hypothesis. Despite denials from Standard & Poor’s that it uses non-public information to decide on inclusions and deletions, they may still unintendedly send a signal to the market. Perhaps investors perceive an addition as an increase in the quality of management because of the close monitoring activities by S&P (Jain, 1987). It could also be the case that investors perceive S&P to possess superior analytical abilities (Denis et al., 2003). These and other reasons can explain a rational upward revision of investors’ expectations of companies following an index inclusion.
Merton’s (1987) information cost hypothesis takes into account the costs that investors need to incur to become aware of a certain company. Due to these information costs investors only trade in a limited subset of all traded securities. Merton (1987) shows that an increase in a company’s investor base leads to a reduction of the company’s cost of capital, which in turn results in a higher market value. Hence, inclusion in an index reduces information asymmetry and brings a company into prominence. The resulting effect is an increase in its investor base and subsequently higher stock returns.
The last important theory for explaining the increase in stock returns following index inclusion is the liquidity hypothesis. According to the liquidity hypothesis a permanent increase in trading volume should lead to a permanent decrease in the bid-ask spread (Stoll, 1978). The decrease in the bid-ask spread represents a decrease in the direct cost of trading. Hedge and McDermott (2003) find a consistent relationship between abnormal returns and the decrease in the bid-ask spread. Thus, liquidity hypothesis entails that the abnormal returns following an index inclusion result from a decrease in the direct cost of trading.
The aforementioned theories all support the notion that index inclusion will lead to positive abnormal returns, either temporarily or permanently. However, these theories were predominantly tested with the S&P 500. Therefore, the effect for a sustainability index might be different. Theoretically, an argument can be made in line with the signalling and information cost hypotheses. Reliable information on corporate sustainability practices is not abundantly
10 available or easily observable. Negative news (e.g. child labour or environmental pollution) is relatively easy to observe and evaluate, but positive CS activities are much harder to identify (Oberndorfer et al., 2013). A sustainability index can therefore be a useful tool to reduce the cost of finding this reliable information. Index inclusion instantly sends a positive signal to the market about the quality of a company’s corporate sustainability.
Where the theoretical perspective provides a clear picture, the empirical evidence is rather ambiguous. Some studies assess the relationship between CS and FP using accounting measures of performance. Becchetti, Di Giacomo, and Pinnacchio (2008) examine the CS-FP relationship using the Domini Social Index. They find that ‘firms produce ‘larger cakes’ but a smaller portion of these cakes goes to shareholders’ (p. 555). However, shareholders are compensated by lower volatility. In a similar manner, Ziegler (2012) does not find a very pronounced positive relationship either. The effect of inclusion in the DJSI World on return of assets is insignificant for the UK and Ireland, and weakly positive for other European countries. Similarly, Oberndorfer et al. (2013) and Cheung (2011) perform event studies but find insignificant or negative effects of inclusion in sustainability indices. In contrast, Lackman, Ernstberger, and Stich (2012) present significant positive reactions to inclusion of companies in the DJSI STOXX. They argue that inclusion in the DJSI STOXX represents an increase in reliability of sustainability information, which they find to be positively valued by investors. Correspondingly, Consolandi, Jaiswal-Dale, Poggiani, and Vercelli (2009) compare the DJSI STOXX to a benchmark index which the DJSI STOXX slightly outperforms. Additionally, they perform an event-study analysis that provides positive results, albeit limited. Finally, Orlitzky et al. (2003) acknowledge the ambiguous evidence on the CS-FP relationship and perform a meta-analysis to find more robust and unambiguous results. Their conclusion is that CS is positively correlated with FP across studies, thus CS activities are likely to pay off.
2.3 Emerging markets
The discussed literature on sustainability index inclusions focuses on U.S. or European companies. Nearly twenty years ago De Santis and İmrohoroǧlu (1997) already mentioned the enormous flow of portfolio investments to emerging markets. Nevertheless, it is safe to say that the research gap is enormous. The (surprising) lack of research into CS within emerging markets has been indicated by several studies (Belal, 2001; Jamali & Mirshak, 2007; Ortas, Moneva, & Salvador, 2012).
Admittedly, this research area is not completely bereft of emerging market studies. However, these studies all zoom in on a specific emerging country (Belal, 2001; Jamali &
11 Mirshak, 2007; Park & Choi, 2015) and fail to provide evidence for emerging markets as a whole. Moreover, they focus on the descriptive aspect of stakeholder theory, whereas this thesis focuses on the instrumental relationship between CS and stakeholder theory.
However, some general inferences can be drawn from the study by Belal (2001). Belal (2001) stresses the valuable insights that can be obtained from research into emerging market CS practices because of the difference in socio-economic context. In his paper he links the socio-political and economical background of Bangladesh to the level of reporting of CS by companies. Linking back to the different dimensions of CS, social responsibility tends to improve as the industrialisation of emerging countries continues. Industrialisation is linked to unionisation of the labour force and subsequent improvement of working conditions and reduction of child labour. Thus, the expectation is that the social dimension is positively valued by both investors in emerging countries as well as by foreign investors.
Regarding the economic dimension, Belal (2001) observes increasing pressure within emerging countries on companies to disclose more information regarding their CS practices. This is caused by the widespread corruption that is often present in emerging countries and of which the corporate sector is increasingly accused. The inclusion in a globally recognised sustainability index may provide in the need for more disclosure by ratifying the high level of CS of the included companies.
Where there are clear benefits to local investors from the social and economic dimensions of CS, the environmental Kuznets curve provides a different perception of CS. The environmental Kuznets curve hypothesises the relationship between environmental degradation and per capita income as an inverted U-shape (Grossman & Krueger, 1995). Before the turning point, an increase in per capita income will increase environmental damage. Not until after the turning point, is per capita income high enough for citizens to want to reduce their environmental footprint (Krugman, Obstfeld, & Melitz, 2012). This could indicate a lack of interest within emerging countries to focus on environmental responsibility. As emerging countries industrialise further, their per capita income will increase and the interest in environmental responsibility will increase correspondingly. Hence, a high level of CS may not yet be valued by local emerging country investors. However, this does not necessarily imply that the overall valuation of the environmental dimension is trivial. Belal (2001) points at the considerable amount of foreign investors who may demand environmental responsibility. Due to foreign pressure it may still be valuable for emerging market companies to invest in the environmental dimension of CS.
12 Some of the empirical evidence on the relationship between CS and FP within emerging markets can be found in three studies examining Brazil and Korea. Ortas et al. (2012) compare the performance of the Brazilian Corporate Sustainability Index (BCSI) with several general indices but find inconclusive results: The BSCI slightly outperforms some of the indices, but underperforms in comparison to the rest of the indices. The studies by Choi, Kwak, and Choe (2010) and Oh and Park (2015) both research the link between CS and FP in Korea using accounting measures of performance: Both studies find a robust positive effect from CS on FP. This is a preliminary indication that the marginally positive results found in studies for developed countries can be extended to companies in emerging markets. Research into index inclusion effects for emerging market companies has been performed by Hacibedel (2008). Hacibedel (2008) intuitively expects the price impact to be magnified and indeed provides evidence of a positive price effect of greater magnitude than in developed countries.
In summary, stakeholder theory suggests a positive relationship between CS and FP. Enlightened stakeholder theory slightly adjusts the focus towards long-term value maximisation, but still gives an important role to CS in this process. The relevance of CS for FP can even be shown to be reconcilable with shareholder theory. At first sight, the empirical evidence seems to be ambiguous, but further studies and meta-analyses tend to support the theoretical positive relationship between CS and FP. Even though not all dimensions of CS may be valued by local investors, the presence of wealthier foreign investors may make up for this lack of valuation. The review of the literature surrounding CS thus provides a solid basis for formulating a hypothesis regarding the effect of sustainability index inclusion on stock returns.
Hypothesis 1: Inclusion in the DJSI EM has a positive effect on stock return.
2.4 CS-risk relationship
The majority of research regarding CS has focused on the relationship with financial performance. This has either been measured using accounting measures of performance or by performing an event study and analysing the abnormal returns. However, these analyses overlook a crucial part of the overall profitability of companies to investors, namely the risk associated with these returns. Modern portfolio theory teaches that higher returns are compensation for higher risks (Markowitz, 1952). A decrease in a company’s returns may reflect investors’ response to lower risk by requiring lower return. On the other hand, an
13 increase in a company’s return may not be caused by outstanding CS, but merely by an increase in risk due to the CS activities. Thus, when analysing abnormal returns following an index inclusion, it is imperative that the associated risks are taken into account.
The theoretical argument is that improved relationships with the company’s stakeholder decrease risks as the possibilities of negative externalities are mitigated and positive externalities are strengthened (Besley & Ghatak, 2007; Heal, 2005; Oberndorfer et al., 2013). High levels of CS (as reflected by inclusion in a sustainability index) enable companies to reduce the costs and risks associated with sustainability issues. This is also termed the risk insurance hypothesis which states that firms use CS to control risks (Bouslah, Kryzanowski, & M’Zali, 2013). Some studies argue that CS only affects systematic risk based on the crucial insight that the CAPM has provided: When holding a well-diversified portfolio, the unsystematic or idiosyncratic risk becomes irrelevant (Sharpe, 1964).
Several studies therefore examine the relationship between CS and systematic risk. Herremans, Akathaporn, and McInnes (1993) find that companies with superior CS reputations outperform their competitors by providing higher returns and lower risk. Lee, Faff, and Langfield-Smith (2009) improve on the methodology of Herremans et al. (1993) by examining a broader sample and using a more comprehensive measure of CS: Contrastingly, they do not find a significant relationship between CS and systematic risk. Similarly, Cheung (2011) performs an event study and compares systematic risk before and after the event, but he does not find a significant change. Although Luo and Bhattacharya (2009) mainly focus on the relationship between CS and idiosyncratic risk, they also test the relationship between CS and systematic risk. Remarkably, they find a significant reduction of systematic risk as a consequence of CS. Salama, Anderson, and Toms (2011) find similar results by examining UK firms: Their research provides evidence for a significant inverse relationship between CS and systematic risk. Equivalently, Oikonomou, Brooks, and Pavelin (2012) show a (weak) negative relationship between CS and systematic risk.
Despite all the research relating to systematic risk, a sizeable body of research exists claiming idiosyncratic is not as irrelevant as the CAPM would suggest (Levy, 1978; Merton, 1987; Bessembinder, 1992; Malkiel & Xu, 1997; Huang, Liu, Rhee, & Zhang, 2009). A similar focus in the literature regarding CS can be found. With Boutin-Dufresne and Savaria (2004) being among the first to analyse the relationship between CS and idiosyncratic risk, this research area has gained traction. The argument is that only idiosyncratic risk is affected by CS because CS is firm specific (Bouslah et al., 2013). Several studies examine the link between CS and idiosyncratic risk (Boutin-Dufresne & Savaria, 2004; Lee & Faff, 2009; Lee, Faff, &
14 Langfield-Smith, 2009; Mishra & Modi, 2013): They all provide matching results indicating a negative relationship. Bouslah, Kryzanowski, and M’Zali (2013) determine the effects of several dimensions of CS on total risk and idiosyncratic risk, and go a little further by separately examining the effect of CS strengths and concerns on a company’s risk. They find that the majority of CS strengths (concerns) have a negative (positive) impact on risk. Similar to the evidence on the CS-systematic risk relationship, there is a strong expectation that a negative relationship exists between high levels of CS and idiosyncratic risk.
In addition to the discussed empirical evidence, it is important to establish whether similar results can be expected for emerging markets. However, as limited as the research on CS in emerging markets is, the literature on the CS-risk relationship is even sparser. One study is found that investigates the risk-adjusted performance of a sustainability index in an emerging country: Ortas et al. (2012) investigate a Brazilian sustainability index and find some results indicating that the sustainability index is associated with lower risk levels. Furthermore, it is widely known that volatility is generally higher in emerging markets (Harvey, 1995; Bekaert & Harvey, 1997). In a similar way to Hacibedel’s (2008) expectation of a magnified price impact, the impact of CS on risk measures may also be magnified. Due to the lack of further related research, this thesis adopts the same hypotheses as would be adopted for developed markets. The results found in this thesis may therefore represent a much needed contribution to the currently available literature.
Hypothesis 2a: Inclusion in the DJSI EM reduces systematic risk. Hypothesis 2b: Inclusion in the DJSI EM reduces idiosyncratic risk.
Hypothesis one states that there will be a positive effect on stock returns and the second hypotheses state that there will be a reduction in risk. Taken together, the expectation is that index inclusion has a positive effect on stock return even after adjusting for risk. In other words, the Sharpe ratio before and after inclusion will be analysed and a positive change in the Sharpe ratio is expected. This prediction is reflected in the following and therewith last hypothesis.
15 3 Methodology and data
This thesis performs an event study to analyse the effects of inclusion in a sustainability index on the stock prices of emerging market companies. This section elaborates on the research method used in this thesis. Firstly, some information is provided on the specific index that is used, namely the Dow Jones Sustainability Emerging Markets Index. After that, the summary statistics of the data are provided. Lastly, the steps taken to perform the event study are described.
3.1 Dow Jones Sustainability Emerging Markets Index
Due to multiple factors, a substantial amount of sustainability indices have become available in the past 15 years. Fowler and Hope (2007) attribute this increasing supply of sustainability indices to the increase in funds under management by Socially Responsible Investment (SRI) funds, the publicity resulting from corporate accounting scandals, and investors’ demand for recognised sustainability benchmarks. Some of the sustainability indices currently available are from MSCI (formerly KLD Analytics), Calvert Investments, Dow Jones, ECPI Group, Vigeo, FTSE Group, and Jantzi Research. Only Dow Jones and MSCI offer an emerging markets based index. Moreover, the DJSI construction methodology appears to meet the desirable traits of consistency, verifiability, logicality, and replicability (Fowler & Hope, 2007). Therefore, the sustainability index that is used in this thesis is the DJSI EM.
The DJSI EM was launched in 2013 by S&P Dow Jones Indices and is calculated and composed by RobecoSAM. The DJSI family aims to track the world’s leading companies in terms of economic, environmental, and social criteria (S&P Dow Jones Indices, 2015). To this end, the DJSI employs a best-in-class approach to select its constituents. In other words, only the most sustainable companies are selected for membership of the index. To assess the level of sustainability of a company, a Total Sustainability Score (TSS) is calculated based on RobecoSAM’s annual Corporate Sustainability Assessment (CSA). The companies that are invited to complete the CSA questionnaire are the 800 largest emerging market companies in the S&P Global BMI. These companies constitute the ‘Invited Universe’. Subsequently, the ‘Assessed Universe’ consists of those companies from the Invited Universe that have a TSS. Finally, the ‘Eligible Universe’ is constructed from the Assessed Universe: Companies with a TSS that is less than 40% of the TSS of the highest scoring company are excluded and it is ensured that there are sufficient companies within each industry. The best companies (top 10%) per industry are then selected from the Eligible Universe to form the DSJI EM. The weightings
16 of the constituents within the index are based on their market capitalisation and are capped at 10%.
The annual CSA questionnaire forms the basis of the constituent selection process. A CSA features approximately 80 to 120 questions (RobecoSAM, 2015) on financially relevant economic, environmental, and social dimensions. The focus of these questions is on sustainability factors that can have an impact on companies’ long-term value creating potential. The CSA questions capture both general and industry-specific criteria covering the three dimensions. Each dimension consists of on average 8 criteria, which in turn contain approximately 6 questions. Each criterion is assigned a weight and may be worth up to 100 points. The combined weight of the criteria within each dimension make up the dimension weight. The general dimension weights are 27%, 38%, and 35% for respectively the economic, environmental, and social dimensions. This shows a slight overemphasis on the environmental dimension of CS, but overall conveys the balanced assessment of a company’s level of CS by the DJSI EM.
Table 1
Number of Inclusions per Year
Year Inclusions 2013 20 2014 19 2015 13a Total 52 Source: RobecoSAM
a Includes a company that was also included in
2013, namely ASUR.
Table 2
Number of Inclusions per Country
Country Inclusions Percentage
Brazil 7 13.46 Chile 2 3.85 China 2 3.85 Colombia 4 7.69 Greece 1 1.92 India 6 11.54 Malaysia 2 3.85 Mexico 5 9.62 South Africa 6 11.54 Taiwan 5 9.62 Thailand 10 19.23 Turkey 1 1.92
United Arab Emirates 1 1.92
Total 52 100.00
17 3.2 Data
Since a few years ago RobecoSAM no longer publicly publishes the constituents of its indices or the changes thereof. To obtain the necessary data it is required to fill in an academic request and sign a non-disclosure agreement. This procedure was followed for this thesis and subsequently the data were received via email from RobecoSAM.
Table 3
Number of Inclusions per Sector
Sector Inclusions Percentage
Automobiles & Components 1 1.92
Banks 6 11.54
Capital Goods 5 9.62
Consumer Services 1 1.92
Diversified Financials 1 1.92
Energy 5 9.62
Food & Staples Retailing 1 1.92
Food, Beverage & Tobacco 4 7.69
Health Care Equipment & Services 1 1.92
Household & Personal Products 2 3.85
Insurance 1 1.92
Materials 2 3.85
Media 1 1.92
Real Estate 4 7.69
Retailing 2 3.85
Semiconductors & Semiconductor Equipment 1 1.92
Software & Services 2 3.85
Technology Hardware & Equipment 1 1.92
Telecommunication Services 2 3.85
Transportation 5 9.62
Utilities 4 7.69
Total 52 100.00
Source: RobecoSAM
Table 1 displays the number of inclusions per year. One of the companies, ASUR, was newly included in the index in both 2013 and 2015. Since this has no confounding effects on
18 either inclusion, this company remains in the sample and is analysed twice. From table 2 can be seen that more than half of the newly included companies are based in only four countries, namely Thailand, Brazil, India, and South Africa. The majority of countries that are part of the assessed universe (20 countries) are represented in this sample (13 countries). Furthermore, table 3 shows that inclusions are fairly evenly spread over the different sectors, with a slight overrepresentation of the Banking, Capital Goods, Energy, and Transportation sectors.
The individual stock returns for each company are obtained from Datastream. The index used as benchmark for the market model is the S&P Emerging BMI of which the total return data are also collected from Datastream. With these data the daily logarithmic returns are calculated as follows:
𝑅𝑡 = 𝑙𝑛 ( 𝑃𝑡
𝑃𝑡−1) (1)
3.3 Event-study methodology
Event studies are used to analyse the effect of a certain event by examining abnormal returns. The event-study methodology is based on three assumptions. An important assumption is that capital markets are sufficiently efficient to respond to events (Brown & Warner, 1980). This assumption may be violated in the case of emerging capital markets, which is an important notion to keep in mind when assessing the results. For now, it is assumed that the capital markets in the respective emerging markets are sufficiently efficient. Secondly, the event is assumed to be unforeseen. This way the abnormal returns accurately capture the reaction of investors to the event. Lastly, there may be no other important events during the event window that can confound the results for the event that is studied. Confounding effects are largely avoided when performing a short-term event study because only a short window is studied (McWilliams, Siegel, & Teoh, 1999). The companies in this thesis have been reviewed and no potentially confounding events were found.
In this thesis the standard event-study methodology as described by MacKinlay (1997) will be employed. A multitude of models are available to estimate the normal returns with, however MacKinlay (1997) notes that more extensive multifactor models provide limited gains. Moreover, availability of risk factor data is often problematic with multifactor models, which is especially true for emerging markets (e.g. Kenneth French’s online data library does not include daily factor data for emerging markets). Therefore, this thesis uses the market model as benchmark to estimate the normal returns according to the following equation:
19 This equation is estimated using the data of the estimation window with a length of 100 trading days ending 25 days prior to the event. Previous research into the effects of inclusion in a sustainability index specify two event days. The first event day is the day on which the changes in the index are announced (AD), whereas the second event day is the day of the actual change in the composition of the index (CD). Therefore, the estimation window is based on the first event day resulting in an estimation period from t = AD –125 to t = AD –26.
Subsequently, the abnormal returns and average abnormal returns are estimated as follows: 𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡− 𝐸(𝑅𝑖,𝑡) (3) 𝐴𝐴𝑅𝑡 = 1 𝑁∑ 𝐴𝑅𝑖,𝑡 𝑁 𝑖=1 (4) It is assumed that the abnormal returns are normally distributed with mean zero and variance σi2. The cumulative abnormal returns (CAR) and cumulative average abnormal returns (CAAR) are given by:
𝐶𝐴𝑅(𝑡1,𝑡2)= ∑ 𝐴𝑅𝑖,𝑡 𝑡2 𝑡=𝑡1 (5) 𝐶𝐴𝐴𝑅(𝑡1,𝑡2) = 1 𝑁∑ 𝐶𝐴𝑅(𝑡1,𝑡2) 𝑁 𝑖=1 (6) The overall event window runs from AD –10 to CD +5. The total event window is split up in several smaller event windows to analyse different effects related to the inclusion. This thesis mainly follows the division used by Consolandi et al. (2009). The first pre-announcement window ranges from AD –10 to AD –1 to test for any leakage or anticipation of information. Subsequently, the announcement day itself is tested, which is the first trading day following the announcement. In addition to this, a three-day window surrounding AD is examined: The existence of different time zones may lead to some differences in the actual announcement day between different countries. Following the announcement, the time in between the announcement day and the effective change day is analysed for any abnormal returns. Then the change day is tested to determine whether there are any abnormal returns after the effective revision of the index. Lastly, a window after the change day is specified to see whether there is any lagged impact or slow assimilation of the information carried by the event.
20 Each year RobecoSAM announces the annual rebalancing after the close of trading on the second Friday of September. Therefore the event day associated with the announcement (AD) is set on the Monday following the announcement. The change in the index takes place after the close of trading on the third Friday of September. Thus, the change day (CD) is also set on the Monday following the actual change. Table 4 shows the corresponding dates for each year.
Boehmer, Masumeci, & Poulsen (1991) show that event-induced variance leads the most commonly-used test statistics to reject the null hypothesis too frequently. They consequently propose a combination of Patell’s (1976) t-statistic and the ordinary cross-sectional t-test resulting in the most accurate rejection of the null hypothesis. Their test-statistic is called the standardised cross-sectional t-test and will be used in this thesis, alongside the normal cross-sectional t-test, to test the significance of the CAARs. In addition to this, a nonparametric test will be used because they are free of specific assumptions made regarding the distribution of the abnormal returns. Since the normality of daily returns is often a very questionable assumption, the use of a nonparametric test helps to overcome the issue of non-normality. This thesis uses the rank test as proposed by Corrado (1989). All the test statistics are explained in more detail in the appendix.
3.4 Risk measures
When analysing returns it is important to take into account the risk associated with these returns. Therefore, in this thesis systematic risk, idiosyncratic risk, and the Sharpe ratios before the event are compared with their levels after the event. Following Cheung (2011), systematic risk is measured by the beta estimated with equation 1 and idiosyncratic risk is measured by the variance of the residuals. The market model is estimated for two new windows to allow for
Table 4
Event Windows of the Analysis
Pre-AD Broad AD AD Post-AD CD Post-CD
2013 02/09–13/09 13/09–17/09 16/09 17/09–20/09 23/09 24/09–30/09 2014 01/09–12/09 12/09–16/09 15/09 16/09–19/09 22/09 23/09–29/09 2015 31/08–11/09 11/09–15/09 14/09 15/09–18/09 21/09 22/09–28/09
Note. The dates reported in this table are based on the publication dates regarding the index rebalancing that
21 a comparison. These windows are a pre-AD and a post-AD comparison window. Cheung’s (2011) post-CD period consists of 60 trading days. In a similar fashion, the post-AD comparison window ranges from AD to AD +60. The pre-AD comparison window is chosen to be of similar length from AD –60 to AD –1. To detect any significant changes in systematic risk, the betas of the two separate regressions in the comparison windows are tested with a Chow test (Appendix A). Rejection of the null hypothesis means that the coefficients of the two regressions, their betas, are not the same which implies a change in systematic risk following the index inclusion. Similarly, to test for a significant change in idiosyncratic risk, an F-test (Appendix A) is used to test the null hypothesis that the idiosyncratic risk (variance of the residuals) before and after AD are equal.
To generate an accurate picture of the potential increase in profitability following an index inclusion the Sharpe ratio is calculated in the different comparison windows. The well-known formula for the Sharpe ratio is:
𝑆𝑅𝑖 =
𝜇𝑖− 𝑟𝑓
𝜎𝑖 (7)
where μi is the mean of the logarithmic returns of a given company, σi is the standard deviation of the logarithmic returns of a given company and rf is the risk-free rate. In the case of emerging markets, one overall risk-free rate does not exist. Moreover, government bonds that are generally used as the risk-free rate are often per definition not risk-free in emerging markets because of the higher probability of sovereign defaults. Since the purpose of calculating the Sharpe ratios in this thesis is to compare them across time and not to provide a precise estimation at one point in time, the most commonly used risk-free rate will be used, namely the 4-week U.S. Treasury Bill rates. For the years 2013-2015 these risk-free rates are obtained from Datastream.
22 4 Results
In this section the results that are found with the event study are provided and interpreted. Firstly, some summary statistics regarding the market and the sample are provided and shortly discussed. Subsequently, the main results, namely the presence of abnormal returns, are analysed. Lastly, the results regarding the risk associated with the index inclusions are presented.
4.1 Summary statistics
The summary statistics of the market returns and the sample companies returns are useful indicators of the distribution of the daily returns. The parametric tests assume a normal distribution of the returns, however this is often not warranted for daily returns. Table 5 indeed shows that the normality assumption may be violated. The skewness of the market and the companies included in the sample (just) fall within the rule of thumb specified by Bulmer (1979) in both windows. Bulmer’s (1979) rule of thumb is that data with skewness within the –0.5 to 0.5 range is approximately symmetric. However, in most cases the kurtosis notably exceeds the normal distribution kurtosis value of three, which implies the data are leptokurtic. Although, the deviation from the normal distribution may not severely affect the parametric test statistics, the use of a nonparametric test will provide the necessary robustness of the results.
Table 5
Summary Statistics
Estimation window Event window
Daily return Market Sample Market Sample
Mean (%) –0.011 0.021 –0.087 –0.001
SD (%) 0.776 1.736 0.914 1.682
Skewness –0.462 –0.089 0.441 0.219
Kurtosis 4.264 4.947 3.610 4.383
Note. This table reports the summary statistics obtained from an estimation window of 100 trading days and
an event window 21 trading days. The reported values are the averages of the estimates (mean, standard deviation, skewness, and kurtosis). The summary statistics of the market are based on the S&P Emerging BMI. The sample summary statistics include the 52 companies that were included in the years 2013-2015.
4.2 Abnormal returns
This thesis analysis the effect of inclusion in a sustainability index by performing an event study. The abnormal returns are calculated based on estimates from the market model. Multiple
23 event windows are analysed to determine the potentially different effects of the announcement day and the effective change day. Table 6 displays the results of this analysis. The first observation is that the announcement has a significant and relatively large negative effect on the returns of the included companies. This is most clearly seen from the CAAR in the broader AD window [AD –1:AD +1] which has a –0.767% cumulative average abnormal return. This window is more representative than the one-day window at the announcement day due to time-zone differences between the sample companies and the RobecoSAM headquarters. The ordinary and the standardised cross-sectional t-tests for the broad AD window are significant at respectively the 10% and 5% level. The Corrado rank test is also significant at the 5% level, providing robustness to this negative result.
Secondly, no significant effect is found in relation to CD. This implies that the market incorporates the new information when it is first announced, and is not subsequently affected by the actual revision of the index. This is in line with efficient markets as the revision of the index does not carry any new information about the newly included companies. Moreover, the significant negative effect is limited to the small window around AD and there is no sign of leakage. The pre-AD window [AD –10:AD –1] carries only a small positive CAAR and is not significant.
Table 6
Cumulative Average Abnormal Returns Calculated Based on an Estimation Window of 100 Trading Days
Event window CAAR (%)
Percentage positive CAR tc test Standardised tc test Corrado rank test AD –10:AD –1 0.160 53.85 0.206 0.154 –0.812 AD –1:AD +1 –0.767 36.54 –1.850* –2.462** –2.446** AD –0.427 38.46 –1.746* –1.821* –1.570 AD +1:CD –1 0.525 51.92 1.057 0.906 0.055 CD 0.289 50.00 1.084 0.968 0.800 CD +1:CD +5 –0.142 50.00 –0.340 –0.358 –0.002 AD –10:CD +5 0.406 50.00 0.354 0.178 –0.705
Note. This table displays the cumulative average abnormal returns estimated using the market model with an
estimation window of 100 trading days. The percentage positive CAR represents the percentage of the 52 sample companies that display a positive CAR. Two parametric test statistics are provided, namely the ordinary cross-sectional t-test and the standardised cross-sectional t-test. For robustness, a non-parametric test statistic is also included, namely the Corrado rank test.
24 Remarkably, the significant negative effect seems to be reversed. Although the CAAR of the complete event window [AD –10:CD +5] is not significant, it does display a positive overall CAAR of 0.406%. The reversal can also be seen from the two windows after AD, namely the post-AD and the one-day CD window: The respective CAARs are 0.525% and 0.289%. In addition to this, the effects found do not seem to uniformly hold for the majority of the companies. Except for the AD and the broad AD windows, half of the companies indeed experience the same sign for their CAR as the CAAR sign, but the other half of the companies display CARs with an opposite sign. This could indicate a difference in the valuation of CS among emerging markets. Although such a country difference is not observed in this sample, it may still be an interesting topic for further research.
The negative effect around the announcement day is even better visible in the graphical representation of the CAARs as displayed in figure 1. The graph shows a very pronounced depression of the CAARs starting just before AD until just after AD. Subsequently the reversal effect is visible: After AD the CAAR returns to approximately the same level as before AD.
Figure 1. Cumulative average abnormal returns across the 52 sample
companies for the entire event window [AD –10:CD +5] as estimated using the market model with an estimation window of 100 trading days.
4.3 Risk
The valuation of CS is not complete without taking into account the associated risks. Following modern portfolio theory the risk associated with a company consists of systematic and idiosyncratic risk. Since many CS activities are firm specific, the modern portfolio theory irrelevance argument due to diversification is disregarded and idiosyncratic risk is not excluded from the analysis.
-0.40% -0.20% 0.00% 0.20% 0.40% 0.60% 0.80% AD -10 A D -5 AD -1 AD AD + 1 C D -1 CD C D + 1 C D + 5 C A A R Relative Day to AD (CD)
25
Table 7
Changes in Systematic and Idiosyncratic Risk
Summary statistics No. (%) of firms with significant Mean
difference
Median
difference SD Increase Decrease No change
Systematic
risk –0.223 –0.220 0.510 2 (3.85%) 11 (21.15%) 39 (75.00%)
Idiosyncratic
risk –0.003% –0.001% 0.016% 10 (19.23%) 15 (28.85%) 27 (51.92%)
Note. Systematic risk is estimated by the beta from the market model. Idiosyncratic risk is estimated by the
residual variance of the market model. The risk measures are compared between two comparison windows: [AD –60:AD –1] and [AD:AD +60]. The number of significant increases (decreases) are significant at the 5% level for a one-sided F-test.
The levels of systematic risk and idiosyncratic risk before AD are compared to their levels after AD (table 7). The mean and median systematic risk difference between the pre-AD and post-AD period are very similar and negative, which indicates a decrease in systematic risk, as measured by the beta from the market model, of approximately 0.223. However this decrease in sensitivity to the market is only significant for 11 out of 52 companies. The majority (75%) experiences no significant change in systematic risk. Therefore, the conclusion must be that the index inclusion does not appear to uniformly decrease sensitivity to the market.
The mean reduction in idiosyncratic risk, as measured by the residual variance of the market model, is 0.003%. Approximately half the companies experience a change in their idiosyncratic risk, but not all of these experience a reduction. Around 20% of the newly included companies see their idiosyncratic risk rise after the announcement. A slightly larger fraction (29%) indeed experiences a reduction, this might suggest that at least some companies are seen to be less risky due to their high levels of CS. However, this analysis does not provide convincing evidence that index inclusion leads to a lower level of idiosyncratic risk.
In addition to the separate analyses of the index inclusion effect on stock return and risk, a comparison of the Sharpe ratio before and after AD is included to facilitate a complete analysis of the risk-return trade-off associated with inclusion in the DJSI EM. Table 8 provides the values of the Sharpe ratio in the pre-AD period [AD –60:AD –1] and the post-AD period [AD:AD +60]. A mixed result emerges when looking at the three individual years. In 2013 and 2014 the average Sharpe ratio is initially positive, but experiences a large fall to a negative value after AD. In both years this is the case for the majority of the 52 companies included in the sample. In contrast to 2013 and 2014, the Sharpe ratio is initially negative in 2015 and experiences a rise after AD. This rise is however not enough to change the Sharpe ratio to a
26 positive value, thus the observation remains that investors would have been better off with a risk-free asset than sustainable emerging market companies. The average Sharpe ratio over the entire three-year period starts off very low but positive, but experiences a substantial decrease resulting in on average a negative Sharpe ratio after inclusion. Again it is important to note that only a 60% majority experiences a decrease in its Sharpe ratio. The other 40% may represent certain emerging markets with different characteristics. As such it would be interesting to undertake a more detailed study differentiating between multiple emerging markets.
Table 8
Comparison of the Sharpe Ratio before and after AD
Mean Median SD
Before After Before After Before After Increase Decrease 2013 –0.050 –0.060 –0.039 –0.078 0.081 0.101 30.00% 70.00% 2014 –0.022 –0.038 –0.009 –0.046 0.109 0.139 36.84% 63.16% 2015 –0.061 –0.033 –0.059 –0.032 0.094 0.082 61.54% 38.46% 2013–2015 –0.012 –0.045 –0.001 –0.046 0.103 0.111 40.38% 59.62%
Note. The Sharpe ratio before AD is compared with the Sharpe ratio after AD. The comparison window
before AD is [AD –60:AD –1] and after AD is [AD:AD +60]. An increase in the Sharpe ratio constitutes an improvement of the risk-return trade-off.
4.4 Robustness checks
To verify the robustness of the results presented in this thesis, some additional analyses are performed. Firstly, the length of the estimation window is changed to encompass 200 instead of 100 trading days. This change results in an even more profound negative effect of inclusion in the DJSI EM (table 9). In comparison to the results that use a 100-day estimation window, the results mostly carry the same sign, but both more negative and more significant. Again, no significant effect is found in the run-up window [AD –10:AD –1]. However, significant negative abnormal returns are found in the post-CD event window which would suggest a lagged impact of the information carried by the event. In addition to this, in every event window only a minority of the companies display positive CARs indicating that the negative effect is more widespread among the companies.
A last important note is that the CAAR over the complete event window is now significantly negative and there is no sign of price reversal anymore. Interestingly, this negative effect is driven by a difference in the year 2014. The overall CAAR in the year 2013 (2015) is consistently negative (positive): The length of the estimation window does not influence the
27 sign of the CAAR. The same cannot be said for the year 2014. Using the short estimation window the CAAR is 1.151%, but the CAAR severely drops to –5.697% when the longer estimation window is used. Although this changes the sign and significance of the overall CAAR, the results found in the sub-windows around AD remain roughly the same. Therefore the main result that inclusion has a negative effect does not change.
Table 9
Cumulative Average Abnormal Returns Calculated Based on an Estimation Window of 200 Trading Days
Event window CAAR (%)
Percentage positive CAR tc test Standardised tc test Corrado rank test AD –10:AD –1 –0.635 42.31 –0.781 –0.810 –1.406 AD –1:AD +1 –1.327 26.92 –3.107*** –3.674*** –2.903*** AD –0.610 32.69 –2.545** –2.647*** –2.071** AD +1:CD –1 0.372 44.23 0.759 0.766 –0.100 CD –0.110 44.23 –0.397 –0.496 –0.682 CD +1:CD +5 –1.142 38.46 –2.653*** –2.728*** –1.499 AD –10:CD +5 –2.126 38.46 –1.693* –1.810* –2.347**
Note. This table displays the cumulative average abnormal returns estimated using the market model with an
estimation window of 200 trading days. The percentage positive CAR represents the percentage of the 52 sample companies that display a positive CAR. Two parametric test statistics are provided, namely the ordinary cross-sectional t-test and the standardised cross-sectional t-test. For robustness, a non-parametric test statistic is also included, namely the Corrado rank test.
* p < .10. ** p < .05. *** p < .01.
Once more, the visualisation of the progression of the CAARs over time provides an even clearer view of the index inclusion effects. Figure 2 displays a roughly consistent downward trend of the CAARs over the entire event window. Figure 2 displays a remarkable difference with figure 1: The negative effect is consistent over time and is not reversed within the 21-day event window.
Based on the related literature, a positive effect was hypothesised. However, large significant negative results are found. Due to these large significant negative effects, it is in order to briefly analyse the performance of the DJSI EM compared to its benchmark, the S&P Emerging BMI. This comparison shows that, although both indices performed poorly, the average daily return of the DJSI EM is most often below the market return (table 10). Only in 2014 does the DJSI EM on average provide a better return. The underperformance in daily return terms could explain the negative reaction to inclusion in the DJSI EM: It might be an
28 indication that the company is now part of a group of underperforming sustainable companies. The comparison of the Sharpe ratios of the DJSI EM with the market conveys a similar message. Although the risk-adjusted returns of the DJSI EM have on average not been below that of the market in the three-year period, the only year that actually shows a higher Sharpe ratio for the DJSI EM is 2014. Investors may be rather pessimistic and more easily remember the bad, therewith overemphasising past negative performance and extrapolate this to their expectations of the future performance. This could explain the unwillingness of investors to keep holding the stocks of newly included companies and cause the downward pressure on these stocks resulting in negative abnormal returns.
Figure 2. Cumulative average abnormal returns across the 52 sample
companies for the entire event window [AD –10:CD +5] as estimated using the market model with an estimation window of 200 trading days.
Table 10
Performance Comparison Between the Market Index and the DJSI EM
Daily return SD Excess return Sharpe ratio
Market (%) DJSI EM (%) Market (%) DJSI EM (%) Market (%) DJSI EM (%) Market DJSI EM 2013 –0.016 –0.040 0.794 0.879 –0.063 –0.087 –0.079 –0.099 2014 –0.008 0.009 0.686 0.907 –0.036 –0.019 –0.052 –0.021 2015 –0.066 –0.088 1.004 1.143 –0.101 –0.122 –0.100 –0.107 2013–2015 –0.030 –0.040 0.838 0.983 –0.066 –0.076 –0.079 –0.078
Note. This table displays several performance indicators of the market benchmark (S&P Emerging BMI) and
the DJSI EM.
-2.50% -2.00% -1.50% -1.00% -0.50% 0.00% 0.50% 1.00% 1.50% AD -10 AD -5 AD -1 AD AD + 1 C D -1 CD C D + 1 C D + 5 C A A R Relative Day to AD (CD)
29 5 Discussion and conclusion
This thesis studies the effect of index inclusions of sustainable companies originating from emerging markets in the Dow Jones Sustainability Emerging Markets Index. Based on the shareholder view a company’s CS activities are at the expense of other value maximising activities. Thus inclusion in a sustainability index that indicates CS excellence has a negative effect on the profitability of a company and is negatively valued. Stakeholder theory opposes this view by arguing that focusing on CS will increase a company’s long-term value creating ability. Through CS a company takes into account all its stakeholders which enables it to reduce potential conflicts and establish a superior reputation. This results in reduced risks and increased profitability. In turn this implies that an inclusion in a sustainability index is positively valued.
The effect of inclusion in the DJSI EM on stock returns is analysed by means of an event study and some additional analyses are performed to assess the effect on systematic and idiosyncratic risk. The initial abnormal return results based on an estimation window of 100 trading days indicate a significant negative effect of the inclusion following the announcement. Contrary to other studies examining the relationship between CS and FP, excellence in CS does not yet seem to be positively valued for companies from emerging markets. Moreover, an apparent price reversal before the end of the 21-day event window seems to support the price pressure hypothesis (Harris & Gurel, 1986). The event does not carry any actual information and stock returns revert relatively quickly to their pre-announcement levels.
However, this statement does not hold for the results obtained from using a longer estimation window. The development of the CAARs over time is visibly negative and no reversal is found within the total event window. It is possible that this reversal takes place after the end of the specified event window (Harris and Gurel (1986) find that price reversal takes place within on average three weeks), but the marked difference between the results remains. Regardless of the cause of this difference, the results based on both estimation windows remain negative around AD. Although this is not hypothesised, this negative result is not unique as is evidenced by other studies. For instance, Oberndorfer et al. (2013) similarly find negative inclusion effects and Cheung (2011) finds mostly insignificant effects that are subsequently reversed.
These negative results may be explained in different ways. The first obvious explanation is that stakeholder theory does not hold true, instead shareholder theory represents the correct view on CS. If this is the case, these negative results are robust and are not likely to change in the future. A different explanation might be one that is related to the environmental
30 Kuznets curve. Per capita income may in (many) emerging markets not yet have surpassed the highest point of the curve. Because of this it is hypothesised that the citizens of these countries are not yet at a point where they care for the active reduction of environmental degradation. Consequently, they do not yet positively value CS practices performed by companies in their countries. In other words, if they are the main investors of the companies studied in this thesis then one can reasonably expect a negative effect of inclusion in the DJSI EM.
The decrease in stock returns following the announcement of inclusion in the DJSI EM may be compensated by reduced risks. The change in both systematic and idiosyncratic risk is examined. In contrast to the results found by Cheung (2011), but in line with other literature relating systematic risk to CS, this study finds a significant reduction of systematic risk for 21% of the companies. Unfortunately, 75% of the companies do not experience a significant reduction of their systematic risk. Therefore hypothesis 2a is not strongly supported. A higher percentage of firms (29%) experiences a significant decrease in their idiosyncratic risk. Yet, the percentage of firms that experiences a significant increase in their idiosyncratic risk is also higher (19%). This leaves about half the companies without a significant change in their idiosyncratic risk. Hence, no compelling evidence is found to support hypothesis 2b.
Despite the lack of unambiguous results regarding the change in risk, it is still possible that investors are better off after the announcement based on risk-adjusted return. To this end, the Sharpe ratios before and after AD are compared. An increase in the Sharpe ratio is hypothesised, but this is not confirmed by the data. On average over the years 2013-2015 a slight majority of the companies (60%) actually experiences a decrease in their Sharpe ratio.
All these results combined paint a rather sad picture for highly sustainable companies. On a more positive note, the results may be driven by rather few companies with very large negative abnormal returns. For most of the event windows, a sizable portion of the companies experience positive CARs despite the overall negative results. This could indicate a difference between the emerging markets. For instance, some countries may have already surpassed the turning point on the environmental Kuznets curve. Or there could be other factors at play which would make a study that differentiates between the emerging markets very interesting.
It is important to point out some of the limitations associated with this thesis. First of all, MacKinlay (1997) points out the issue of event-day clustering which may invalidate results. This is clearly the case in this thesis as the announcement and revision of the index take place on pre-specified dates each year. Every year the newly included companies all have the same event day. However, two of the three test statistics used in this thesis are proven to be robust
31 against event-day clustering. As such, this limitation is already mostly resolved within this thesis.
Another limitation is one that cannot easily be resolved due to the nature of the index used. The companies that are invited to participate in the annual CSA are only the largest emerging market companies that are included in the S&P Global BMI. Hence, many potentially highly sustainable companies are never assessed and cannot be part of the index unless they are of significant size. This naturally incorporates a bias in the companies included in the sample and must be kept in mind. Building on this, another limitation is the small sample size which also cannot yet be resolved simply because the DJSI EM was launched only several years ago. Moreover, the lack of data on risk factors in emerging markets prohibits the use of more extensive benchmark models. The use of a different benchmark model could provide additional robustness to these results.
Lastly, although the results found in this thesis seem to make a bad case for corporate sustainability, this is not the entire story. McWilliams et al. (1999) point out that an event-study approach may be rather limited, since it only investigates the valuation of CS by one group of stakeholders, namely the shareholders. It may indeed be the case that the other stakeholders positively value the high level of CS and supersede the negative valuation by the shareholders.
These limitations provide input for several future research possibilities. Evidently, as time progresses the sample size for this specific index will increase. It will be interesting to perform a similar study in a few years’ time to find out if the negative results persist or whether they will change as the DJSI EM matures. Similarly, the effect of index exclusions may provide additional valuable insights into the valuation of CS. Due to lack of data this is not yet possible, but this issue will naturally be resolved over time. In conclusion, the results currently indicate that CS is not valued for companies in emerging markets. Yet these results are likely to change and more closely resemble results found for developed markets as emerging markets further develop and the DJSI EM matures.
