Listing and (non)-delisting on the Dow Jones Sustainability World Index:
implications on market return
An event and regression analysis of the listing, delisting and non-delisting of US
firms on the DJSWI
MSc Finance: Quantitative Finance Wessel Joren – 11865520 Academic Year 2017 – 2018
i Abstract
This paper investigates the effect of a listing, delisting and continuation on the Dow Jones Sustainability World Index (DJSWI) on the stock market return. Subsequently, the effect of a change in competition on the firm’s profitability and CARs is studied. The event study uses the Fama-French Three Factor Model and studies the abnormal returns of US firms listed, delisted and continued to the DJSWI during 2005-2016. The result indicate a significant positive effect on market return when a firm is listed on the DJSWI. However, both a delisting and continuation do not show significant results. Therefore, the stock market values a listing more than it devalues a delisting and thus the effect is asymmetric. Furthermore, regressions applying competition variables as a proxy for the level of sustainability regulations show that a change in competition is not of any significant influence on the CARs obtained in the event analyses. The domestic competition variable is expressed as the HHI and a proxy import tariff. The overall results of this study imply that improving corporate sustainability has the potential to increase shareholders’ wealth and the level of competition does not change the shareholder’s return during the DJSWI events.
ii Abbreviations
ACAR Average Cumulative Abnormal Return
AD Announcement Date
AR Abnormal Return
CAAR Cumulative Average Abnormal Return
CAPM Capital Asset Pricing Model
CAR Cumulative Abnormal Return
CRSP The Centre for Research in Security Prices
CS Corporate Sustainability
CSP Corporate Sustainability Performance
DJSI STOXX Dow Jones Sustainability Index Europe
DJSI Dow Jones Sustainability Index
DJSWI Dow Jones Sustainability World Index
ED Effective Date
FF3 Fama-French Three Factor model
FTSE4Good Financial Times Stock Exchange for Good
NGO Non-Governmental Organisation
OLS Ordinary least squares
SAM Sustainable Asset Management Group
iii Table of Contents
Abstract ... i
Abbreviations ...ii
1 Introduction ... 1
2 Background and Hypotheses ... 4
2.1 Corporate social responsibility ...4
2.2 Corporate sustainability & Financial performance ...4
2.3 Sustainability indices ...5
2.4 Dow Jones Sustainability Indices & Event studies ...7
2.5 Hypotheses development ... 9
3 Methodology and Data ... 12
3.1 Corporate social responsibility ... 12
3.2 Corporate sustainability & Financial performance ... 13
4 Results and Discussion ... 17
4.1 Event analyses ... 17
4.1.1 Listings on the DJSWI ... 17
4.1.2 Delistings from the DJSWI ... 19
4.1.3 Continuations on the DJSWI ... 21
4.2 Regression analyses ... 23
4.2.1 Listings on the DJSWI ... 23
4.2.2 Delistings from the DJSWI ... 26
4.2.3 Continuations on the DJSWI ... 27
4.2 Limitations ... 29 5 Conclusion ... 31 5.1 Event analysis ... 32 5.2 Regression analysis ... 33 Appendices ... References ...
1 1 – Introduction
For the past decades there has been an increase in corporate demand to apply policies with a focus on sustainability. This sustainability view discards the classical economic theory which focusses on maximizing shareholders’ value as defined by Friedman (Friedman, 1970). Instead, attention is aimed to stakeholders since its success depends on stakeholder’s satisfaction, which requires sustainable development (Lopez et al., 2007). Sustainable development is defined by Baumgartner and Ebner (2011) as the protection of the environment and a fight against poverty. When a firm applies these aspects its defined as corporate sustainability (CS). The level of CS is known as corporate sustainability performance (CSP). CSP measures the firm’s adoption of social, economic, environmental factors into its operations and subsequently quantifies the impact these adoptions exerts on both the firm itself and society (Artiach et al., 2010). In order to quantify the effect of CSP on the short term shareholder’s wealth, I study the change in CSP and subsequent listing, delisting or continuation on a sustainability index (i.e. The Down Jones Sustainability World Index). I find that a listing leads to a higher shareholder’s return, while a delisting and continuation show no significant influence.
An increasing demand of CSP has led to social responsible investments, which refer to a strategy where investors make investment decisions based on their social and ethical principles (Ortas et al, 2011). The idea behind these investments stems from the theory that environmental and social performance can determine a firm’s public image and therefore can influence the relations towards the public and investors. Furthermore, this could lead towards greater efficiency within a company due to innovations that lower costs such as the implementation of renewable energy, which could significantly lower a company energy costs (Lopez et al., 2007).
The fast increase of social responsible investments has led to the development of stock indices with a focus on sustainability. These indices track the financial performance of firms that belong to the top in their industry in regards to CSP. The rise of these social indices has made it easier for investors to track and diversify their portfolio based on sustainability an socially responsible benchmarks. The idea behind adding sustainable companies to a portfolio is based on the added value social investments brings on top of regular investments. Unsurprisingly, some studies have found that portfolio diversification towards sustainable firms instead of non-sustainable firms in a similar industry has led to a higher return (Consolandi et al., 2009; Lopez et al., 2007).
The availability of social indices has not only led to an increase in investor’s value but from a business perspective a firm listed said index can potentially obtain alternative funds from non-conventional investors concerned about sustainable development and social well-being (Curran and Moran, 2007). There are several well-known social indices available that all have their own set of criteria for
2 evaluating a firm’s CSP. However, the most popular of these indices is the Dow Jones Sustainability World Index (DJSWI).
The DJSWI was introduced on 8 September 1999 and it was the first global index to track the CSP of companies that scored high on sustainability. The DJSWI is internationally recognized for its informational transparency and objectivity, and is well received by international investment communities (Oberndorfer et al., 2013). Each year more than 2,000 companies are able to file for listing on the index. The index list only the top 10% sustainability leaders within their own respective industry. These leaders are chosen using a large set of criteria defined in the filing process. In summary, the DJSWI assesses the CS risks and opportunities in economic, environmental and social dimensions (Hawn et al., 2016).
Due to an increasing CS and the availability of tracking the sustainability performance through the use of the DJSWI, studies trying to quantify financial performance and listing on sustainability indices have increased. However, for the most part these studies focus on either listing or delisting, use small sample sets (e.g. 2008-2010) and use data before 2010. Furthermore, only one study also investigates a continuation on a similar index. Therefore I have developed the following research question: What are the implications on a firm’s stock return of a listing, delisting and continuation on the DJSWI? To evaluate the research question I implement an event study. This event study applies a regression using the Fama-French Three Factor Model with and without an import tariff variable used as a competition proxy. The event analysis quantifies the effect of a company being added, deleted or continued on the DJSWI of leading sustainable companies in the US. Data on a total of 236 company listing, delisting and continuations on the DJSWI over the period 2005-2016 is collected. This data is used to estimate the average cumulative abnormal return (ACAR) to quantify the DJSWI effect. The normal returns are estimated with a regression by using the realized stock return data of the collected companies. The realized returns are acquired using S&P500 stock return data.
The event analysis implements several event windows. A total of five event windows are implemented following a similar study by Hawn et al. (2016). The event windows are as followed: (1) directly before and after the announcement of change in DJSWI; (2) between the announcement and the effective change in DJSWI; (3) directly before and after the effective change in DJSWI; (4) directly before to several weeks after the effective change in DJSWI; (5) directly before to several weeks after the effective change in DJSWI.
3 I find that the market rewards a listing on the DJSWI, however it does not penalize a delisting, nor does it rewards a continuation. The results show that both the announcement and effective listing result into a significant positive ACAR which is also sustained after the listing becomes effective.
A change in sustainability regulations can lead to a change in shareholder’s value. For example, some companies are focussed heavily on the use of sustainable fuels because this is less regulated than fossil fuels. However, when these regulations for fossil fuels become more lenient this can potentially lead to a loss in market return. A change in said regulations could be seen as a change in competition since the market becomes more or less accessible for companies with a different focus on sustainability. I order to quantify the effect of a change in competition I apply the Herfindahl–Hirschman Index (HHI) and import tariff variables in a regression on a firm’s profitability and CAR. This tariff variable serves as a proxy for the change in competition and its influence on the DJSWI effect. To my knowledge this has not yet been studied.
I find that the both the HHI and import tariff have a significant influence on profitability of a firm but not the CARs obtained from the previously studied event analyses. However, because apply data of the second quarter (Q2) of each year, several months before the DJWSI events happen, this can potentially lead to biased results.
Although similar event analyses have been studied and described in several papers, this paper provides several contributions to the already available research. First, as described by Dilling (2008) the positive effect of CSP on the market value of a firm has increased over the past decades due to an increasing sustainability demand. However, most studies regarding this topic are either conducted before 2010 or use data from before that year. I will not only use the most recent data but also an extensive timeframe to study the potential changes in attitude towards CSP. Second, all studies on the DSJWI focus on either a listing, delisting or a combination of both. As described above this paper also studies the effect of a continuation on the DJSWI which has not yet been studied for this specific index (Ortas & Moneva, 2011). Thirdly, I will implement the HHI and change of import tariffs in an OLS regression as a proxy for the effect of change in competition on the DJSWI effect. I have not found a study which has studied the DJSWI effect by also taking into consideration the change in competition. Lastly, I will use a set of event windows which also study the prolonged effect of a listing on the index. On the other hand, the already available research predominantly focusses on the direct short-term effects of either the announcement, effective change or a combination of both.
4 2 – Background & Hypotheses
2.1 – Corporate social responsibility
Over the last decades companies are becoming more aware of the benefits of contributions to sustainable development by reorienting their operations and processes. This reorientation assumes that the firm obtains results that are beneficial to its survivability since that is the firm’s primary focus. As explained by Michael and Gross (2004), this increase in survivability is partially due to larger long-term profits for shareholders are ensured by means of corporate management applying both economic and sustainability criteria. This sustainable focus of firms is termed corporate social responsibility (CSR). At first CSR seemed a passing trend but it has grown its own cultural dimension of business and is now a crucial factor for companies to take into consideration. Numerous studies have described this change in corporate behaviour as the legitimation (legitimation theory). This theory stated that in order for a company to survive it’s necessary to gain stakeholder’s (in most cases this refers to society in general) approval (Campbell et al., 2002; Deegan, 2000; Deegan and Gordon, 1996; Deephouse, 1996). This stakeholder focus is in sharp contrast with the classical shareholder theory, which argues that the main (social) responsibility of a business is solely to maximize its shareholder value (Friedman, 1970). However, the legitimacy theory developed by Freeman in 1984 stated that by ignoring the interests of the stakeholders a firm cannot maximize its shareholder value in the first place. Therefore, legitimacy is viewed as a balance between society and a business’ value system. If this balance is not achieved by disregarding the stakeholders the survivability of a firm might become jeopardized. Organisations have therefore evolved in placing themselves to meet the demanded requirements set by the communities in which they are developing their business activities. These requirements are interrelated and should not be approached independently. Together they define the level of CRS a company applies
2.2 – Corporate sustainability & Financial performance
The level of corporate sustainability (CS) or CSR is expressed by corporate sustainability performance (CSP). Companies engage in CSP because it can be beneficial for financial performance based on neoclassical micro-economics. Using this theory it is suggested that external effects brought up by business are not fully resolved by governments and also not by a competitive market environment (Heal, 2005). CS can therefore act as a replacement for missing market and regulations when these lead to external costs and are therefore able to decrease potential conflicts between a firms and its stakeholders e.g. governments, the public, NGOs, competitors, employees, or clients. This theory explains in depth why stakeholder theory is especially viable when it comes to the environment. This theory also can also be used to explain a potential increase in corporate financial performance when
5 increasing CSP since this causes a reduction in the stated conflicts and can thus e.g. increase corporate profits and subsequently stock returns. Strategic management (theory) further corroborates the value of CS by suggesting that management by their very nature has to satisfy groups with interests (stake) in a firm and can influence the firm’s outcome. In short, this means that is can be financially beneficial to engage in CS with a significant degree of CSP because otherwise these interest groups could cease their support for the business (Barnett et al., 2006). An often used example is the avoidance of child labour since it could cause aggressive campaigns of NGOs and could thus results in profit loss. According to Barney et al. (1991), the resource-based view of the firm suggests that stakeholder management liabilities, in this case the level of CSP, can be regarded as a competitive advantage and an important organizational resource. He further explains that new technologies that are installed due to proactive corporate environmental activities are a further example for a strategic resource if these technologies cannot be easily imitated by competitors.
The discussed arguments for applying a degree of CSP refer to corporate environmental and social activities, which can lead to financial benefits. Not applying CSP (negative news) such as child labour can be relatively easily observed and could lead to negative consequences. However, the majority of CS is not easily observed by the public for example. Companies therefore try and signal to stakeholders by for example applying for certification of environmental management systems according to ISO 14001 (XXX et al). Another possibility to signal a relative high level of CSP is by inclusion on a sustainability stock index. Inclusion on such an index is especially viable for investors or institutional clients who are sensitive to CS, which could lead to increased profits. A good CS reputation increase its employee retention rate and additionally attract highly skilled and thus more productive employees. Regarding the embedding in the resource-based view of the firm, a good reputation is a further example for an intangible resource that is valuable, rare, and difficult to imitate or substitute.
2.3 – Sustainability indices
As mentioned above, sustainability indices are an efficient method to indicate a firm’s level of CSP. According to Karlsson and Cheung (2008), Dilling (2008) and Cheung (2009) these indices are considered an appropriate indicator for environmental and social activities, CSP and CSR. Because of this studies have tried to quantify the benefit of a firm to be listed on a sustainability index. These studies predominantly applied an event analyses focusing on the price and volume effects associated with changes in such an index. The three most well-known sustainability indices are the Dow Jones Sustainability World Index (DJSWI), the FTSE and the Domini 400. Although the majority of the papers focus on listings and delistings on the DJSWI, I also discuss two papers that focus on the FTSE and Domini 400.
6 Of the available research on the question whether CSR pays off Karlsson and Chakarova (2008) have written the most comprehensive paper. In order to quantify the positive effect CSP they apply an event analysis of DJSWI changes using a sample of listings from 2002 to 2007. They thus use DJSWI listings as a proxy for CSP. Their research concludes that a positive or negative change in a company’s dedication to CSP does not generate significant abnormal returns.
This seems to contradict the expectation that a degree of CSP lead to increased financial performance of a business. The findings of Karlsson and Chakarova (2008) are corroborated by a similar study conducted by Cheung (2009) who investigates changes in the DJSWI with a listings sample from 2002 to 2008. Cheung concludes that the announcements per se do not have any significant impact on stock return and risk.
Although both Karlsson and Cheung (2008) find almost identical results when they decompose their samples across geographic or time then they find different results. They first thing they find is that the impact of CSP on stock return does actually vary over times but in a trendless motion. Similar results have been found by Dilling (2008). He also studies the changes of DJSWI from listings however using a smaller sample period from 2002 to 2005. He observers that the market reaction on a listing is becoming less positive. This effect was theorized to be due to a continues growing complexity of CSP. Another finding of Karlsson and Cheung (2008) are the geographical differences of the impact of a listing on the DJSWI. It was found that in some regions the abnormal returns are significantly different from others. Especially the abnormal returns from US and UK listings and for Japanese delisting are significantly different. Dilling (2008) on the other hand found no difference in market reactions decomposed by different regions. However, a significant difference in market reaction can be found for corporations in different industries. More specifically, he reports that the share prices of corporations in the consumer product, healthcare, technology, and utilities industry seem to react more positively to the DJSI World inclusion announcement than the share prices of corporations in the basic material, financial, and industrial product industry.
Where the DJSWI covers sustainable companies around the entire world, the Domini 400 Social Index focusses solely on US based companies. Becchetti et al. (2007) have studied the reaction of the stock market to listings and delistings between 1990 and 2004. They test two hypothesis. The first hypothesis tests whether the changes of the index cause significant abnormal returns. The second hypothesis tests whether these changes show a trend over time. A trend is expected because of a growing interest in CSP over the past decades. They mention steadily increasing SRIs in ethical funds as a proxy for the increasing interest. Becchetti (2007) concludes that CSP has risen over time and that the abnormal
7 returns that occur in case of an exit from the index is significantly negative. Becchetti (2007) further concludes that there is an asymmetric response to listings and delistings, whereas delistings cause only a minor negative effect while listings cause a significant positive effect. This is identical to results described in this paper.
A final study of Curran et al. (2007), conducts similar research as described above however with a focus on the FTSE4Good UK Index. Listings and delistings are again used as a proxy for CSP. Although Curran finds that a listing lead to a positive market reaction and a delisting to a negative market reaction the findings are not statistically significant. Curran states that the insignificance is caused by the idea that investors rather want direct financial information rather than needing to make a link themselves between good CSP and a change of the social index. However, it is clear from both theory and empirical studies that investors could benefit from a good reputation due to a significant level of CSP which could come about through a listing on the UK social index. Nonetheless, the authors conclude that companies do not benefit, nor become penalized, form a listing or delisting, respectively.
2.4 – Dow Jones Sustainability Indices & Event analyses
Of the available social indices the Dow Jones Sustainability Indices (DJSI) are the most well-known. The DJSI was launched in 1999 as collaboration between the Sustainable Asset Management group (SAM) and the Dow Jones Index. The DJSI was the first social index that covered the entire globe and thus did not focus on a single nations like its peer indices did. The DJSI comprises a set of indices with the DJSWI being most used. However, the index family also contains regional focussed indices such as the DJSI STOXX which focusses on Europe and the Asia Pacific index. The sustainability assessment criteria by SAM include the three major scopes: the economic, the environmental and the social dimension. This integrated assessment has a strong focus on long-term shareholder value.
Each year the DJSWI assessment questionnaire is handed out to 2000 of the largest companies in their respective industry. The companies are then compared to their peers and subsequently ranked. Only the 10% of list are included on their index. Because CS itself is a dynamic construct so are the questionnaires. Each year the assessments itself is assessed and new sustainability trends are incorporated, likewise redundant assessments are removed from the list. This method ensures that companies need to stay up to date with the current CS trends and cannot rely solely on passed results otherwise they will simply be removed from the list when a peer shows better results. Examples of such trends are changes in: food and water scarcity, local environment dynamics, global climate change and accountability and health. Next to the questionnaire which is completed by the company itself, the assessment is extended by media and stakeholder analyses. Industry analysts review media,
8 articles, press releases and commentary of particular stakeholders regarding the company that has been collected over the years.
In order to estimate the effect of a change in one of the DJSI indices numerous researched have published articles with applying empirical event analyses of reactions to listings and delistings. Most of these studies are focussed on the DJSWI or the DJSI STOXX. An overview of the prior research regarding this topic is presented in Table 1.
Table 1 – An overview of past event analyses studying firm listings, delistings and continuations on the DJSWI and
DSJI STOXX.
Paper Index Window Model Event Abn. Return
Consolandi et al. (2009) DJSI STOXX 2002-2006 Market Model Listing Positive Delisting Not sign. Cheung et al. (2011) DJSI STOXX 2002-2008 Fama & French 3F Listing Positive
Delisting Negative Lackman et al. (2011) DJSI STOXX 2000-2008 Market Return Listing Positive
Delisting Oberndorfer et al. (2011) DJSWI 1999-2005 Fama & French 3F Listing Not sign.
Delisting
Ortas et al. (2011) DJSI STOXX 2002-2007 Multivariate Regression Listing Not sign. Delisting Not sign. Continuation Positive Robinson et al. (2011) DJSWI 2003-2008 Market Return Listing Positive Delisting Negative
Baas et al. (2016) DJSWI 2006-2010 CAPM Listing Positive
Delisting Not sign.
Hawn et al. (2016) DJSWI 1999-2007 CAPM Listing Positive
Delisting Negative
As discussed in the last section, studies of the different social indices show variable results. From Table 1 it also becomes clear that different studies of solely the DJSWI or the STOXX also differ. However, most studies on show that a listing on the index results in a positive shock reaction (Lackmann et al., 2011; Robinson et al., 2011; Hawn et al., 2016; Consolandi et al., 2009). A second study of Cheung (2011) is also presented in more detail. As can be observed Cheung also finds a positive market response from a listing, however this is short-lasted and quickly corrected to the mean return. One other study conducted by Baas (2016) uses two different time samples, one from 2006-2010 and the other form 2009-2010. The longer-term sample does not find any significant effects. However, the short timeframe does find a significant positive effect. Baas theorizes that this is potentially due to the financial crisis and a change of investor’s perspective regarding CSP. The remaining studies on the list either find a non-significant result or a negative result. Negative reactions are potentially a
9 consequence that some investors see an increase in CSP as a decrease in short-term profits due to higher investment costs.
Of the available research the predominant amount is focussed on a listing analysis instead of a delisting analysis. That being said, the studies that do also study the effects of a delisting all conclude that the effects are either negative or not significant. Again Cheung (2011) concludes that the effects are only temporary and are quickly corrected back to the mean return.
Of the entire set of available literature there is only a single study conducted to find the effect on the stock market of a non-delisting or continuation on the DJSI STOXX. This study conducted by Ortas & Moneva (2011) concludes that the market response significantly positive to a non-delisting.
2.6 – Hypotheses development
Throughout section 2 I describe several theories that explain the possible positive and negative effects when I company decides to develop CS and starts implementing social responsible investments. Although some authors like Cheung (2013) suggest that investors could perceive a significant degree of CSP as value destroying (especially short-term value), the majority concludes that overall investors should react positive to CSP (Consalendi, 2009; Lackman, 2011; Robinson, 2011; Baas, 2016; Hawn, 2016).
In order for companies to easily effectively signal their commitment in CSP they can apply to be listed on social indices such as the DJSWI. Several researchers tried and estimate the quantitative effect on stock price of such a listing. An overview of several of said studies are presented in Table 1. A downside of the available literature is the fact that most sample data used is before 2010 while several studies conclude that the effects of CSP are actually increasing. Furthermore, only a single study also takes a non-delisting or continuation on the DJSI into consideration but only for the DJSI STOXX. Additionally, the majority of the past literature applies short timeframe event analyses and does not look at the relatively long run abnormal returns. Lastly, none of the studies discuss the effect of changes in sustainability regulations and subsequent changes in competition on the abnormal returns from a DJSWI event. Together, these comments have led me to develop the following hypotheses:
H1: A listing on the DJSWI will have a significant positive effect on the stock performance which is sustained at least two weeks after the listing becomes effective. The effect will show a yearly increase. H2: A delisting on the DJSWI will have a significant negative effect on the stock performance which is sustained at least two weeks after the listing becomes effective. The effect will show a yearly decrease.
10 H3: A continuation on the DJSWI will have a significant positive effect on the stock performance which is sustained at least two weeks after the listing becomes effective. The effect will show a yearly increase.
H4: An increase in competition (decrease in HHI or import tariff) will have a significant negative effect on the profitability and CAR of a firm listed or continued on the DJSWI but a positive effect on the delisted firms.
11 3 – Methodology and data
3.1 – Event analyses
In order to quantify the effect of a change in the DJSWI on stock return I implement an event study. Using this analysis I estimate the cumulative abnormal returns (CARs) of US based companies. A sample of the announcement dates (AD) and effective dates (ED) of listings, delistings and non-delistings on the DJSWI is collected over the period of 2005 to 2016. The RobecoSAM database is used to collected to data of ADs and EDs of listings and (non)-delistings events and the names of the companies corresponding to these events.
As described in Ortas & Moneva (2010) one should take into consideration during the application of event analyses that other news events can influence the CAR. Therefore the Lexis-Nexis news group is used to conduct searches for potential other news effects, such as mergers or expansions, of specific companies in specific years. In case such an event is found the company is removed from the test sample to prevent any estimation bias.
Table 2 – The yearly announcement and effective dates of firms listing, delisting and continuing on the DJSWI
from 2005-2016
Announcement date Effective date Listing Delisting Non-delisting Total
September 8, 2005 September 19, 2005 8 8 60 76 September 14, 2006 September 25, 2006 13 10 39 62 September 13, 2007 September 24, 2007 9 8 42 59 September 11, 2008 September 22, 2008 16 5 46 67 September 10, 2009 September 21, 2009 12 8 35 55 September 9, 2010 September 20, 2010 11 3 37 51 September 8, 2011 September 19, 2011 8 5 43 56 September 13, 2012 September 24, 2012 13 7 44 64 September 12, 2013 September 23, 2013 9 8 46 63 September 11, 2014 September 22, 2014 7 12 40 59 September 10, 2015 September 21, 2015 4 6 36 46 September 8, 2016 September 19, 2016 7 6 34 47 Total 117 86 502 705
An overview of the ADs and EDs of the listings, delistings and non-delistings are presented in Table 2. All firms that are included in the sample are traded on the S&P500, NASDAQ and NYSE. As explained above, the yearly firm samples have been screened for contemporary news using the Lexis-Nexis News Group. The firms that could have been affected by particular news in a given year were removed from the sample. This resulted in the deletion of a total of 68 observations. An overview of the deletions and corresponding news headline is presented in Appendix 2.
12
The final sample consists of 117 listings, 86 delistings and 502 non-delistings across the sample period of 2005-2016. A full list of every corporation included in the final sample is presented in Appendix 1.
Having defined the event dates and samples, an event window needs to be established. In order to make the ACAR estimations with a significant degree of robustness a large event window was chosen for each sample. According to Lorraine et al. (2004), the choice of an event window is extremely important. The show that longer event windows can increase the uncertainty of obtained abnormal returns because a longer timeframe increases the chances on influencing events happening. They term this as event clustering. Therefore they conclude that event windows ought to be surrounding the event date tightly and propose an event window of two days surrounding the specific event date.
An overview of the event windows is presented in Figure 1 and are defined as follows:
1 – [AD – 1 : AD + 2] to study the immediate stock reaction to the announcement of inclusion, exclusion or non-exclusion. Following Lorraine et al. (2004) and MacKinlay (1997), this event period starts one trading day before the actual announcement, because of potential insider trading. This event period ends two trading days after the announcement because this ensures that all information has been made available to the stock markets.
2 – [AD – 1 : ED – 1] to study if the potential immediate stock reaction after the announcement is maintained until the effective date, and not corrected by the market. Again, one trading day before the actual announcement is incorporated within the event period because of potential insider trading.
3 – [ED – 1 : ED + 2] to study the immediate stock reaction to the new effective index. Because information regarding the new index is already publicly available around this date, there should be no need to add extra days to the event period to account for potential insider trading.
4 – [ED – 1 : ED + 10] to study if the potential immediate stock reaction to the new effective index is sustained for two weeks (20 trading days), and is not corrected by the market within this period. 5 – [AD – 1 : ED + 10] to study if there is a significant and sustained stock reaction over the entire observation period, from the day prior to the announcement until two weeks after the effective date.
13 Figure 1 – A visual overview of the estimation window, trading gap, and the five different event windows
The different event windows are used to estimate different set of CARs. The CAR can then be compared to the estimated daily abnormal return (ARit). In doing so one can assess whether the index changes
had an significant effect on the stock returns. However, in order to estimate the daily abnormal return the realized return (Rit) and the expected value for the normal return E(Rit) are required. The complete
formula for ARit is defined by the following formula:
𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− 𝐸(𝑅𝑖𝑡) (1)
The Center for Research and Security Prices (CRSP) is used to obtain the realized returns for the US firms that are listed and (non)-delisted on the DJSWI. The estimated or normal returns are computed on the basis of asset pricing models. The main approached are the market model and the Capital Asset Pricing Model (CAPM). The CAPM model which is based on a corporation or stock i in day t is defined by the following formula:
𝐸(𝑅𝑖𝑡) = 𝛼𝑖+ 𝛽𝑖(𝑅𝑚𝑡− 𝑅𝑓𝑡) + 𝑅𝑓𝑡+ ɛ𝑖𝑡 (2)
Rit and Rmt present the returns for firm i and the market portfolio at the end of period t. Rft is the
risk-free interest rate at the beginning of period t, and εit is the error term with an expectation E(εit)=0 and
variance Var(εit)=σ2εi. Finally, αi and βi are the unknown parameters.
However, many studies show that the three-factor model of Fama and French, which includes two additional factors to explain the excess returns Rit−Rft, has more explanatory power than the one-factor
model. The structure of this three-factor model for a corporation or stock i in day t is as follows:
𝐸(𝑅𝑖𝑡) = 𝛼𝑖+ 𝛽𝑖(𝑅𝑚𝑡− 𝑅𝑓𝑡) + 𝑅𝑓𝑡+ 𝛽𝑖2𝑆𝑀𝐵𝑡+ 𝛽𝑖3𝐻𝑀𝐿𝑡+ ɛ𝑖𝑡 (3)
In this equation, SMBt (small minus big) is the difference between the returns on diversified portfolios
of small and big stocks, and HMLt (high minus low) is the difference between the returns on diversified
14 Substituting equation 3 into equation 1 gives a clear picture of the computation of abnormal return:
𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− (𝛼𝑖+ 𝛽𝑖(𝑅𝑚𝑡− 𝑅𝑓𝑡) + 𝑅𝑓𝑡+ 𝛽𝑖2𝑆𝑀𝐵𝑡+ 𝛽𝑖3𝐻𝑀𝐿𝑡+ ɛ𝑖𝑡) (4)
The unknown parameters in E(Rit) can be estimated by OLS on the basis of the three-factor model for
all days t in the time interval [T0,…,T1], i.e. in the estimation window. The Rmt variable is estimated by
using the returns of a representative market portfolio. In this case the S&P500 will be used. The length of estimation window will follow the model implemented by Farag (2009) of 200 trading days, which is a widely used general duration window. In order to increase robustness of the event analysis and decreased bias a gap between the estimation window and start of the event analysis is normally implemented. Just as with the estimation window there are no set criteria so I will use the gap set by Farag (2009), which is two weeks or 10 trading days. This summarizes to an the following estimation window:
0 – [AD – 210 : AD – 11]
Subsequently the CAR is calculated for each event window. CAR during the event period (t1, t2) is defined by:
𝐶𝐴𝑅𝑖(𝑡1,𝑡2)= ∑ 𝐴𝑅𝑖𝑡 𝑇1
𝑇2 (5)
After the computation of the CAR within a given event window for all firms within a given year, the CAR is divided by N i.e. the number of firms, which results in the Average Cumulative Abnormal Return (ACAR): 𝐴𝐶𝐴𝑅(𝑡1,𝑡2)= 1 𝑁∑ 𝐶𝐴𝑅𝑖(𝑡1,𝑡2) 𝑁 𝑖=1 (6)
The final ACAR value will be tested using a two-sided standardized cross-section student’s t-test. In this case null and alternative hypotheses are:
𝐻0: 𝐴𝐶𝐴𝑅(𝑡1,𝑡2)= 0
𝐻1: 𝐴𝐶𝐴𝑅(𝑡1,𝑡2)≠ 0
3.2 – Regression analyses
Because I want to study the effect of the level of competition on the previously calculated firm listing, delisting and continuation CARs of the entire event analysis period (event window 5), a regression analysis is implemented. First, I analyse the effect of competition on the profitability of the companies
15 used in the event analyses. Firm profitabilities are computed as the ratio of profit level to the value of total assets. The level of profit is defined as the difference between sales revenue and total operating expenses. Total costs include labour, material and opportunity costs of capital (Burja, 2011). To study the mentioned effect a multivariable OLS regression is used. In this case the dependent variable is profitability whereas the independent variables are cash, leverage and firm size. These variables will function as control variables as described in Burja (2011), Fu (2011) and Frésard (2010). The level of competition variable will be expressed as the Herfindahl–Hirschman Index (HHI), which is a measure of the size of firms in relation to the industry and an indicator of the amount of competition amongst them. It is defined as the sum of the squares of the market shares of the firms within the industry, where the market shares are expressed as fractions. Therefore, an increase in the HHI value will result into a decrease in the level of competition because fewer companies take a market share. Since the HHI index is thought to be potentially endogenous, a proxy for the level of competition is also implemented namely the level of US import tariff. Reductions (increase) in import tariffs decrease (increase) the cost of entering U.S. product markets substantially and thus increase (decrease) the competitive pressure on domestic producers. I obtain data on imports between 2005 and 2016 from the Centre for International Data. The actual percentage tariff is calculated by dividing the total yearly value of duties collected by the total yearly value of imports.
The subsequent regression is identical to the profitability regression only in this case the previously collected CAR values are used as the dependent variable. Resulting in the following general regression equation:
𝑌𝑖𝑡 = 𝛼𝑖+ 𝛽1𝐶𝑎𝑠ℎ𝑖𝑡+ 𝛽2𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝛽3𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽4𝑋𝑖𝑡+ µ𝑡+ ɛ𝑖𝑡 ` (7)
Where Yit is either the profitability or CAR value. Cash is calculated by dividing the firm’s cash holdings
by total assets. Leverage is the firm’s debt divided by total assets and lastly firm size indicates the natural log of total assets (Frésard, 2010). Xit is either the HHI or level of import tariff. β1-4 are unknown
parameters, µt are time fixed effects and ɛit is the error term. The regression used for the continuation
data will also implement firm fixed effects because firms can continue on the index for multiple years. A summary of the variables is presented in Table 3.
Variable data is obtained from the COMPUSTAT database using the firm PERMNOS collected and used in the event analyses over a range from 2005 to 2016. Because I want to use data as close as possible to the actual event analyses dates of event window 5, I collect quarterly fundamental data. On average the event period is from 10 September until 21 September, therefore I use data of the second quarter of each year, which is roughly two months before the event happens. Because listings and delistings
16 only happens once for every firm the amount of data points will generally equal the number of firms. However, continuation of firms on the index can happen several years in a row and therefore N will differ from listings and delistings.
A total of six regression tables are presented in results and discussion section, comprising a profitability and CAR regression per (de)-listing and continuation. Every table has six columns or six regressions, similar to the methodology of Frésard (2010). The first regression applies solely the HHI variable as independent variable. Second, solely the tariff variable. The third and fourth regression apply the control variables on the HHI and tariff variable using time fixed effects. The regression for the continuation data will also implement firm fixed effects.
Table 3 – A summarizing overview of the different dependent and independent variables used in the regression analyses Profitability 𝑹𝒆𝒗𝒆𝒏𝒖𝒆𝒊𝒕− 𝑻𝒐𝒕𝒂𝒍 𝒐𝒑𝒆𝒓𝒂𝒕𝒊𝒏𝒈 𝒆𝒙𝒑𝒆𝒏𝒔𝒆𝒔𝒊𝒕 𝑻𝒐𝒕𝒂𝒍 𝒂𝒔𝒔𝒆𝒕𝒔𝒊𝒕 CAR ∑ 𝑨𝑹𝒊𝒕 𝑻𝟏 𝑻𝟐 Cash 𝑪𝒂𝒔𝒉𝒊𝒕 𝑻𝒐𝒕𝒂𝒍 𝒂𝒔𝒔𝒆𝒕𝒔𝒊𝒕 Leverage 𝑫𝒆𝒃𝒕𝒊𝒕 𝑻𝒐𝒕𝒂𝒍 𝒂𝒔𝒔𝒆𝒕𝒔𝒊𝒕 Firm Size 𝑳𝒏(𝑻𝒐𝒕𝒂𝒍 𝒂𝒔𝒔𝒆𝒕𝒔) HHI ∑ 𝑺𝒊𝒕𝟐 𝑵 𝒊=𝟏 Tariff 𝑽𝒂𝒍𝒖𝒆 𝒐𝒇 𝒅𝒖𝒕𝒊𝒆𝒔𝒕 𝑽𝒂𝒍𝒖𝒆 𝒐𝒇 𝒊𝒎𝒑𝒐𝒓𝒕𝒔𝒕
17 4 – Results and Discussion
4.1 – Event analyses
The section described the results of the event analyses. I start by describing the listing event analysis after which I describe the event analyses of the delistings and continuations.
4.1.1 – Listings on DJSWI
Table 4 presents the regression results for the listings on the DJSWI of the five different time windows. The first event window (1), directly around the announcement, shows that the ACAR value of the total sample is positive at a 90% significant level. The next event window (2), announcement until just before the effective date, is not only positively higher than the first window but also has a higher significance level at 95%. The fact that the values increase could be explained by a slow adoption of information by the market.
Table 4 – An overview of the Average Cumulative Abnormal Returns over the different event windows using the
2005-2016 DJSWI listings sample data
AD – 1 : AD + 2 AD – 1 : ED – 1 ED – 1 : ED + 2 ED – 1 : ED + 20 AD – 1 : ED + 20 2005 0.70% (1.20) 1.53 %** (2.16) 0.20% (0.60) 1.14%** (2.11) 1.15% (0.89) 2006 0.81% (1.42) 1.20% (0.96) 0.18% (0.31) -0.24% (-1.29) 0.89%** (2.12) 2007 -0.01% (-0,24) -0.37% (-0.85) 0.01% (0.19) 0.33%* (1.78) 1.01% (0.34) 2008 -0.12% (-0.55) -0.24% (-0.12) 0.98%** (2.19) -2.02%** (-2.14) -0.57% (1.01) 2009 -0.26% (-1.00) 0.00% (0.04) 1.13% (0.87) 1.89% (1.55) -0.09% (1.11) 2010 0.90% (1.44) -0.38% (-0.60) 1.94%** (2.17) 3.01%* (1.75) 1.11% (1.20) 2011 0.21% (0.24) 2.00% (1.41) 0.54% (0.81) 0.20% (0.33) 2.01% (1.56) 2012 -0.56% (-0.72) 0.40% (0.44) 0.88% (1.01) 0.02% (0.45) 0.12% (0.23) 2013 0.65% (1.43) -0.49% (-0.67) 1.16%** (2.18) -2.70%* (-1.78) -1.91% (1.21) 2014 0.98% (0.96) 1.40% (1.59) 0.42%** (2.01) -0.43% (-0.99) 1.18% (1.58) 2015 -0.20% (-0.48) 1.79%** (2.23) 1.14% (0.55) 1.48% (0.91) 2.22% (1.43) 2016 1.20%* (1.77) 1.34% (0.70) 1.69%* (1.77) 1.36% (0.22) 2.53% (1.27) Total 0.45%* (1.67) 0.91%** (1.99) 1.10% (0.97) 0.99% (0.54) 1.26%** (2.12)
18
Note: The percentages indicate the ACAR values for a given time window and year. The t-values are shown
between parentheses. * Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
Another explanation for this rise could be found in the index effect theory. According to Kappou et al. (2009), the index effect indicates the pressure on the price which is observed when a stock is listed or delisted on (any) an index. If this index is well-known and widely used then a trader could make a profit by taking action before the index change becomes effective. For example, by front running and buying the stock of the future listed firm before the index funds one can sell the position shortly after the listing after the index fund demand has increased. In summary, the index effect assumes that a listing on an index will results in a short term price increase due to trading opportunities by rapid volume changes and an increase in demand which is later reversed.
The total sample result of the third event window (3), directly around the effective date, is positive with a significance level of 90%. The next even window (4), one day before listing until two weeks after the listing becomes effective, although positive is not significant. The final event window, just before the announcement until two weeks after the effective date, shows a positive results with a significance level of 90%.
The results of event windows 3 and 4 can be explained by again the index effect, which is short lived and quickly reversed. This fits the results since the direct effect (3) is significantly positive but the sustained effect (4) is not, which could indicate that the index effect has reversed. Another explanation could be found in the notion that relative longer event windows result in lower reliability or significance of an event study as suggested by MacKinlay (1997). Although both the index effect and the theory suggested by MacKinlay (1997) could explain the results found with event window 3 and 4 there is another phenomenon which could be of influence. Figure 2 presents the yearly regression results of event window 5. There are two years when the results are significantly negative i.e. in 2008 and 2013. The negative result in 2008 were to be expected due to the financial recession, however the relative large significant negative value in 2013 seems to be an anomaly. Furthermore, the first three event windows are significantly positive in 2013 whereas the final two are not. This suggests that during the two week period after the effective date some event happened with a significant negative effect on the stock performance. Furthermore, the figure does indicate a visual rising trend in the listing effect over time which is corroborated by Bechetti (2007).
After an extensive search of events shortly after the effective date in 2013 (23 September) I found out that a US government shutdown was present in 2013 from October 1 until October 16. In that year
19 congress was unable to allocate the appropriate funds for 2014. This shutdown happened exactly during the two trading weeks after the effective date of the DJSWI of 2013. Both Burwell (2014) and Hankel (2015) discuss the shutdown in 2013 and fund that it had a significant negative impact on stock performance. Therefore, in order to avoid biased results I have also performed the event analysis of the total sample without including the data of 2013.
Figure 2 - A visual overview of the yearly listing ACAR values for event window (5) with an added trend line
Table 5 presents the total sample regression results of without the use of the 2013 data. As hypothesized, when the 2013 observations are excluded the ACAR result for window 4 not only increases in value but also become significant at a 95% level. Furthermore, the values of windows 3 and 4 also increase in both value and significance. The values of windows 1 and 2 only differ marginally and the significance does not change. These findings indicate that the shutdown (or the anticipation towards the shutdown) had a significant effect on the stock performance and could therefore lead to biased results. Therefore, the remaining tables in this report will also present data without the use of data observations from 2013.
Table 5 – An overview of the total Average Cumulative Abnormal Returns of over the different event windows
using the 2005-2016 DJSWI listings sample data with and without the 2013 sample
AD – 1 : AD + 2 AD – 1 : ED – 1 ED – 1 : ED + 2 ED – 1 : ED + 20 AD – 1 : ED + 20 Total incl. 2013 0.45%* (1.32) 0.91%** (1.99) 1.10% (0.97) 0.99% (0.54) 1.26%** (2.12) Total excl. 2013 0.49%* (1.35) 0.98%** (1.80) 1.20%* (1.40) 1.15 %* (1.30) 1.34%*** (2.70) -3.00% -2.00% -1.00% 0.00% 1.00% 2.00% 3.00% 2004 2006 2008 2010 2012 2014 2016 AC AR Sample Year
20
Note: The percentages indicate the ACAR values for a given time window and year. The t-values are shown
between parentheses. * Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
In summary, the results indicate that a listing on the DJSWI leads to a significant positive ACAR following both the announcement of the listing as well as the effective date. Furthermore, the positive effect of the listing is sustained until at least two weeks after the effective date and also throughout the entire tested period. Additionally, Figure 2 shows an increasing ACAR over time thus indicating that that CSP is potentially becoming more important. These findings support Hypothesis 1.
4.1.2 – Delistings from the DJSWI
Table 6 presents the event analyses results for the delistings from the DJSWI. Again, the five different time windows are analysed and the resulting ACAR values are presented. The regression results of the total sample are both presented with and without the inclusion of the 2013 observations. A handful results are significantly positive or negative but there does not seem to be a clear trend present nor clear anomalies, which was the case for listing results of the year 2013. Furthermore, the total sample regression results are all not significant both with and without the 2013 sample. Figure 3 presents the yearly regression results of event window 5. It shows three things: the ACARs are very close to zero, not necessarily negative and there does not seem to be a decreasing time trend present.
Table 6 – An overview of the Average Cumulative Abnormal Returns over the different event windows using the
2005-2016 DJSWI listings sample data. The total ACAR values are shown with and without the use of the 2013 sample data AD – 1 : AD + 2 AD – 1 : ED – 1 ED – 1 : ED + 2 ED – 1 : ED + 20 AD – 1 : ED + 20 2005 -0.05% (-1.10) -0.09% (-0.99) 0.11% (0.57) 0.06% (1.11) 0.00% (1.05) 2006 0.11% (0.44) 0.01% (0.23) 0.02% (0.31) 0.09% (0.21) 0.10% (0.29) 2007 -0.01% (-0.13) -0.10% (-0.19) 0.02% (0.20) -0.05% (-0.04) -0.04% (-0.2) 2008 -0.20% (-0.30) -0.38% (-0.19) -0.21% (-0.18) -0.39% (-0.07) -0.28% (-0.11) 2009 -0.09% (-0.76) -0.10% (-1.14) -0.09% (-0.99) -0.03% (-1.01) -0.04% (-1.03) 2010 -0.39% (-1.13) -0.25% (-0.61) 0.35% (1.09) -0.15% (-1.01) -0.40% (-0.91) 2011 -0.12% (-0.25) 1.21% (1.95)* 0.24% (0.78) 0.79% (0.99) 0.22% (2.01)** 2012 0.01% (0.56) 0.03% (0.44) -0.01% (-0.23) -0.02% (-0.29) 0.01% (0.13) 2013 -0.02% (-0.66) -0.45% (-1.10) 0.15% (1.02) -0.50% (-0.97) -0.40% (-0.89)
21 2014 0.13% (0.32) 0.11% (0.27) -0.08% (-0.38) -0.50% (-1.65)* -0.25% (-0.90) 2015 -0.12% (-1.12) -0.05% (-0.45) 0.06% (0.39) 0.02% (-0.31) -0.10% (-0.48) 2016 0.00% (-0.23) 0.05% (-0.11) 0.19% (0.46) 0.15% (0.17) 0.20% (0.04) Total incl. 2013 -0.08% (-0.50 0.01% (0.08) 0.10% (1.10) -0.06% (-0.33) -0.08% (-0.12) Total excl. 2013 -0.10% (-0.55) 0.03% (0.40) 0.12% (0.86) -0.05% (-0.31) 0.04% (0.12) Note: The percentages indicate the ACAR values for a given time window and year. The t-values are shown
between parentheses. * Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
The ACAR result of the first event window (1) for the total sample are negative. Although this is in agreement with Hypothesis 1 it is not significant, which is corroborated by Ortas (2011), Baas (2016) and Consolandi (2009). This indicates that the market values a listing and delisting asymmetrically and a delisting is punished as much as a listing is rewarded. This could be explained by the fact that the DJSWI does not necessarily removes firms based on poor sustainability results but simply selects the 10% market leaders per industry. This means that if a firm is delisted this can also be caused because a different firm, which is also to be listed on the DJSWI, is simply larger than an already listed firm causing it to be delisted. In short, a delisted firm does not necessarily has lower CSP than when it was listed on the DJSWI. Furthermore, a delisting from the World index does not mean by default that this firm is also delisted from any of the several regionals indices e.g. DJSI North America.
Figure 3 - A visual overview of the yearly delisting ACAR values for event window (5) with an added trend line.
An interesting finding is that the ACAR values, although not significant, are positive around the effective dates i.e. event window 3. These findings are corroborated by Ortas (2011), which state that
-1.00% 0.00% 1.00% 2004 2006 2008 2010 2012 2014 2016 AC AR Sample Year
22 investors expect a higher spending on more profitable projects which do not necessarily require a high level of CSP. Although, this might be a valid theory, this does not explain why the values surrounding the announcement date are negative, be it not significant. The price-pressure hypothesis might offer an explanation stating that the delisting announcement has increased short term trading volume causing a short term price increase, which is not sustained after the effective date as indicated by windows 4 and 5. This theory is supported by Kappou et al. (2009), who state that trading activities rebalance directly surrounding and after the event date and thus causing a short term increase to the mean.
In summary, the total sample regression results with and without the inclusion of the 2013 observations are al not significant. Furthermore, Figure 3 does not show a decreasing time trend. Therefore, the findings do not support Hypothesis 2.
4.1.3 – Continuation on the DJSWI
Table 7 presents the event analyses results for the continuations on the DJSWI. Again, the five different time windows are analysed and the resulting ACAR values are presented. The regression results of the total sample are both presented with and without the inclusion of the 2013 observations.
Table 7 – An overview of the Average Cumulative Abnormal Returns over the different event windows using the
2005-2016 DJSWI continuations sample data
AD – 1 : AD + 2 AD – 1 : ED – 1 ED – 1 : ED + 2 ED – 1 : ED + 20 AD – 1 : ED + 20 2005 0.47% (1.55) 0.09% (0.01) 0.12% (0.59) 0.22% (1.43) 0.10% (-0.41) 2006 -0.22% (-0.99) 0.10% (1.59) 0.01% (0.13) 0.87% (0.36) 0.17% (0,24) 2007 0.12% (0.36) 0.25% (1.41) 0.06%* (1.67) 0.13% (1.27) 0.09%* (1.81) 2008 -0.23% (-0.96) -0.15% (-1.59) -0.31%** (-2.17) -0.24% (-0.99) -0.18 % (-1.58) 2009 0.15% (0.22) 0.33% (0.70) 0.11% (0.87) 0.21% (1.17) 0.17% (1.00) 2010 -0.20% (-1.20) -0.26% (-0.44) 0.21% (0.60) -0.15% (-0.89) -0.19% (1.20) 2011 -0.12% (-0.57) 0.10% (-0.12) 0.30% (2.19) 0.46% (-2.14) 0.27% (1.01) 2012 0.04% (-1.00) 0.00% (0.09) 0.03% (0.87) 0.15%* (1.65) 0.09% (1.11) 2013 -0.10% (1.44) -0.22% (-0.60) -0.05%** (2.16) -0.03%* (-1.76) -0.12% (1.20) 2014 0.12% (0.24) 0.15% (1.41) 0.01% (0.81) -0.10% (-0.33) -0.05% (1.56)
23 2015 -0.27% (-0.72) -0.01% (-0.44) 0.10% (1.01) 0.03% (0.45) -0.09% (-0.23) 2016 0.13% (-1.43) 0.04% (-0.67) 0.10%** (2.36) 0.15%* (-1.66) 0.19% (1.20) Total incl. 2013 -0.10% (-0.48) -0.02% (-0.90) 0.10% (1.12) 0.18% (0.60) -0.08% (-0.25) Total excl. 2013 -0.08% (-1.04) -0.01% (-0.30) 0.12% (0.88) 0.18% (0.21) -0.06% (-0.61) Note: The percentages indicate the ACAR values for a given time window and year. The t-values are shown
between parentheses. * Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
Similar to the delisting results the continuation results also show both significant positive and negative ACAR values at several time windows. However, again none of the analyses using the total sample are significant also without the 2013 sample.
The first event window results (1), directly around the announcement, show that the ACAR values are negative although insignificant. The second event window (2), the announcement until one day before effective listing, shows insignificant positive results. These findings are in contrast with Hypothesis H3A and results of Ortas et al. (2011) who performed a similar study using DJSI STOXX data instead of DJSWI data and obtained significant positive results using similar time windows. Although this might be an anomaly a possible explanation might simply be that the market already rewarded the listing on the DJSWI and does not further reward a continuation on the index or a continuation of a high level of sustainability. Figure 4 presents the yearly regression results of event window 5. It shows two things: the ACARs are not necessarily positive and there does not seem to be an increasing time trend present.
Figure 4 - A visual overview of the yearly continuation ACAR values for event window (5) with an added trend line. -1.00% 0.00% 1.00% 2004 2006 2008 2010 2012 2014 2016 AC AR Sample Year
24 The results of third and fourth event windows (3 & 4) are do not support hypothesis H3B, however they are in accordance with the findings of Ortas et al. (2011). A positive ACAR is found for event window 3, directly surrounding the effective listing, although not significant. Event window 4, just before effective listing until two weeks after listing, results in a negative non-significant ACAR. Especially, the level of significance in corroborated by Ortas et al. (2011), who state that the index effect does not play a role in continuation on the index and thus the should have no effect.
The final event window implements a regression of the entire period, just before announcement until two weeks after the effective listing. The ACAR values are negative and positive without the inclusion of the 2013 sample. Again, the findings are insignificant.
In summary, the event analyses results for the continuation on the DJSWI do not support Hypothesis 3. All the results are insignificant, negative in some cases and there are no indications for an increasing event effect on the ACAR. These findings are similar to those of Ortas et al. (2011).
4.2 – Profitability and CAR regression
This section will discuss the effect of the level of competition on firm profitability and the previous calculated CAR values across event window 5. A total of six tables are presented, two for each event i.e. listing, delisting and continuation. The competition variable will function as a proxy for the change in sustainability regulations in the US, since e.g. a lowering of sustainability regulations will open the market to less sustainable firms. The competition variable will be expressed by the HHI and an exogenous proxy, the US import tariff rate. Using this proxy is based on the notion that when the tariffs are relatively lower (higher) it is less expensive (more expensive) for firms to import thereby increasing (decreasing) domestic competition. In my hypotheses I explain that I expect the level of competition to significantly influence both the profitability and the CAR values for the different analyses. This is based on the possibility that e.g. increasing competition will lower the positive value of a e.g. listing because this signals investors that the priority of a company should be retaining a strong foothold not increasing spending on sustainability. However, this section concludes that the level of competition does not have a significant influence on the overall CAR regression results. Section 4.2.1 describes the regression results for the listings in great detail after which I describe the delistings and continuation, section 4.2.2 and 4.2.3 respectively, regression results in a more condensed manner since the overall results of all sections are similar.
25 4.2.1 – Listing regression results
Table 8 and 9 presents the profitability and CAR regression results for companies that were listed on the DJSWI, respectively. The regression does not use the 2013 data because of the significant CAR outlier values obtained in the event analyses.
Table 8 – The regression results for the profitability of the listings data set Profitability (ex. 2013) (1) (2) (3) (4) HHIit 0.0066 0.0043* (0.96) (1.74) Tariffit 0.0058 0.0061* (0.87) (1.65) Firm Sizeit 0.0269** 0.0272** (1.97) (1.99) Leverageit -0.0223*** -0.0210*** (2.72) (2.79) Cashit 0.0166** 0.0155** (2.12) (2.14) No. of Obs 74 74 74 74 R2 0.013 0.31 0.012 0.33
Time FE Yes Yes Yes Yes
Note: The student t-value is presented between parentheses.
* Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
The results of Table 8 indicate that the both the tariff rate and the HHI value with use of control variables are significant at a 90% level with the value being positive as expected. The control variables are all of significant influence with leverage being negative in both regression 2 and 4 with a 99% significance level, which is corroborated by Burja (2011). Firm size and cash are also significant at a 95% level and are as expected positive and negative, respectively.
Although the studied firms all belong to the largest top 10% of their respective industry, firm size is apparently still of significant influence with a value of roughly 2.7% in regression 2 and 4. This was expected because the size of the firm can correlate to variables such as rating, age and financial constraints, which on their turn can correlate to investment experience of a firm and methods of funding. Therefore, the larger the firm size, the higher potential profitability (Frésard, 2010).
Leverage is of influence with a value of roughly -2.2% in regression 2 and 4. Again, this was expected because the level of leverage correlates directly to the level of debt which could be of influences on financial constraints. So a higher leverage can result into a lower profitability due to increasing debt and risk.
26 Cash attains a value of roughly 1.6% in the control regressions. This can be explained that a higher amount of cash results in lower financial constraints, lower risk and higher investment opportunities as explained by Lamont et al. (2001).
When both the HHI and tariff rate are used as sole independent variable they are not of significant influence. However, with the addition of the control variables they both attain a 90% significance level with HHI being 0.43% and the tariff variable having a value of 0.61%. Although both values do differ significantly, they both show a positive influence. The difference between the HHI and tariff result could be explained by potential endogeneity between HHI and the other variables.
Table 9 – The regression results for the CARs of the listings data set
CAR (ex. 2013) (1) (2) (3) (4) HHIit 0.0012 0.0014 (0.22) (0.30) Tariffit 0.0019 0.0021 (0.17) (0.31) Firm Sizeit 0.0069* 0.0072* (1.68) (1.69) Leverageit -0.0079* -0.0099* (1.79) (1.75) Cashit 0.0030* 0.0035* (1.71) (1.70) No. of Obs 74 74 74 74 R2 0.008 0.12 0.007 0.13
Time FE Yes Yes Yes Yes
Note: The student t-value is presented between parentheses.
* Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
Table 9 differs significantly from Table 8. Both the HHI and import tariff are not significant in all four regressions even though they are positive as expected. The control variables are significant at a 90% level. These results are corroborated by Burja (2011) and Fu (2011) who run similar regressions using merger and acquisitions CAR values.
Although the addition of the tariff variable functions as exogenous competition proxy, it could be the case that the results are biased because the listing events take place several months of Q2.
In summary, the results partially support Hypothesis 4. The competition values are positive and do influence the level of profitability and are positive (decreasing competition) but not significant for the CAR level.
27 4.2.2 – Delisting regression results
Table 10 and 11 presents the profitability and CAR regression results for companies that were delisted from the DJSWI, respectively. The regression does not use the 2013 data because of the significant CAR outlier values obtained in the event analyses.
Table 10 – The regression results for the profitability of the delistings data set Profitability (ex. 2013) (1) (2) (3) (4) HHIit 0.0071 0.0079* (0.78) (1.65) Tariffit 0.0052 0.0050* (0.85) (1.65) Firm Sizeit 0.0280** 0.0269** (2.01) (2.03) Leverageit -0.0201*** -0.0220*** (2.72) (2.90) Cashit 0.0155** 0.0152** (1.98) (2.11) No. of Obs 62 62 62 62 R2 0.011 0.36 0.012 0.39
Time FE Yes Yes Yes Yes
Note: The student t-value is presented between parentheses.
* Significant at the 90% level, ** Significant at the 95% level, ***Significant at the 99% level.
The results of Table 10 are similar to Table 8. The significance level of are identical to those found in Table 8 and the variables have similar values. Again, both the HHI and the tariff variable are significant at a 90% level with the use of said variables. This was not entirely unexpected because the data used for the regressions is that of Q2 and thus several months before any DWJSI event. Therefore using these regressions I am simply comparing the profitability of companies not yet listed and soon to be delisted. As explained in Section 4.2.2, companies can simply be deleted from the index because they are overtaken in size by a competitor, so this does not mean that the company is less sustainable.
Table 11 – The regression results for the CARS of the delistings data set CAR (ex. 2013) (1) (2) (3) (4) HHIit 0.0005 0.0006 (0.16) (0.28) Tariffit 0.0007 0.0008 (0.14) (0.28) Firm Sizeit 0.0032* 0.0030* (1.66) (1.78) Leverageit -0.0033* -0.0041* (1.70) (1.71) Cashit 0.0024* 0.0029*
