Cross-Sector Risk Spillover Analysis: A Study Using DY & BK Spillover Index Bachelor’s Thesis in Finance 15 Credits Authors: Vidar Hammarén Viktor Holmlund Supervisor: Zelalem Abay Spring 2024 Abstract The purpose of this thesis is to deepen the understanding of risk spillover between sectors, an area with limited prior research. It aims to provide an overview of how sectors are connected and interacting with each other. This will be achieved by using the DY and BK spillover indices, these models have been used in prior studies but mostly when regional indices have been studies. The dataset includes eleven different sectors and a total of 4,069 companies between 2014 to 2023. The results reveal significant variations in risk spillover, heavily influenced by the specific sectors involved. These findings highlight the complexity of inter-sector interaction and suggest that traditional diversification strategies may not be as effective as previously assumed. This study lays a foundation for new research avenues in risk spillover and Environmental, Social, and Governance (ESG) factors, providing crucial insights for investors and policymakers. Future research should address dataset limitation by extending the time frame and adjusting models to account for sector-specific characteristics. Additionally, incorporating a benchmark group to compare ESG and non-ESG companies could further elucidate the impact of ESG factors. This thesis contributes valuable insights into sector interactions, opening up possibilities for further exploration in this emerging field. Acknowledgement We would like to thank our supervisor, Zelalem Abay, for the valuable discussions and suggestions during the writing process. As well as our opponents during the seminars, their comments and feedback have been appreciable. 2 Table of Contents Abstract .................................................................................................................................................... 2 1. Introduction ......................................................................................................................................... 5 1.1. Background ................................................................................................................................................... 5 1.2. Problem description and problem analysis .................................................................................................. 7 1.3. Purpose ......................................................................................................................................................... 8 2. Literature Review .................................................................................................................................. 9 2.1. Risk & Risk Spillover .................................................................................................................................... 9 2.2. ESG ............................................................................................................................................................. 10 2.3. Risk spillover within ESG ........................................................................................................................... 12 2.4. Sector dependence/independence ............................................................................................................... 12 2.5. Hypothesis development ............................................................................................................................. 13 2.5.1. Hypothesis Statement:.........................................................................................................................14 3. Method, Model description & Theory .................................................................................................. 14 3.1. Models – DY Spillover Index ...................................................................................................................... 16 3.2. Models – BK Spillover Index ...................................................................................................................... 18 3.3. Comparison to other models (GARCH and CoVaR)................................................................................... 18 3.3.1. GARCH ...............................................................................................................................................19 3.3.2. CoVaR .................................................................................................................................................19 3.3.3 Conclusion models ................................................................................................................................... 19 3.4. Model Discussion .............................................................................................................................. 20 3.4.1. DY – Spillover Index ..........................................................................................................................20 3.4.2 Relation to ESG ...................................................................................................................................20 3.5. Are there any disadvantages of using the DY Spillover Index as a measurement for risk spillover?......... 21 3.5.1. Importance of industry context in ESG investing .................................................................................... 23 3.5.2. Limitations of a one-size-fits-all approach .........................................................................................24 4. Data ................................................................................................................................................... 25 4.1. Descriptive Statistics .................................................................................................................................. 26 5. Empirical Results ................................................................................................................................. 29 5.1. VAR ............................................................................................................................................................. 29 5.2. DY Spillover Index ...................................................................................................................................... 29 5.3. BK Spillover Index ...................................................................................................................................... 31 6. Discussion .......................................................................................................................................... 32 6.1. VAR ............................................................................................................................................................. 32 6.2. DY Spillover Index ...................................................................................................................................... 33 6.3. Limitations .................................................................................................................................................. 35 7. Conclusion .......................................................................................................................................... 37 3 8. References .......................................................................................................................................... 39 9. Appendix ............................................................................................................................................ 44 Appendix 1A - Correlation Matrix ..................................................................................................................... 45 Appendix 2A – ANOVA Output .......................................................................................................................... 45 Appendix 2B – Description of ANOVA Variables .............................................................................................. 46 4 1. Introduction 1.1. Background Diversification and risk management have been fundamental components of investing for decades, if not centuries. Markowitz (1952) argues that rational investors must diversify their investments to optimize their portfolios and minimize risk. In recent years, Environmental, Social, and Governance (ESG) criteria have become increasingly integrated into investment strategies, offering an additional layer of diversification and potential risk mitigation. The three components of ESG can be explained as follows: Environmental: This factor evaluates a company’s disclosure, impact, and initiatives related to the environment, including efforts to reduce carbon emissions. These aspects pose tangible risks and opportunities for both stakeholders and shareholders (S&P Global, 2019). Social: This aspect handles the relationship with employees and customers, encompassing issues such as human rights, labor standards, and data handling (Lipton et al., 2022). Governance: This involves the functioning of the board, protection of shareholder interests, adherence to ethical standards, and compliance with regulations (S&P Global, 2020). ESG has emerged as a prominent aspect of companies' operations, investment decisions, and academic discourse. Companies worldwide (e.g. PwC, NVIDIA, and PayPal) have integrated aspects of ESG into their strategies. This integration reflects a broader trend toward sustainable investment, where investors increasingly seek to allocate their assets into more sustainable businesses (Financial Times; GSIA, 2021). Additionally, many firms now view ESG as a value creator, recognizing its importance in enhancing their long-term performance and reputation. This growing emphasis is evident in corporate reporting practices; for instance, 90% of the companies on the S&P 500 now produce ESG reports (PwC, 2023; Key ESG, 2024). Furthermore, extensive research has been conducted concerning ESG. For instance, Garcia et al. (2017) investigated the relationship between ESG scores and financial performance in BRICS1 countries. They found that profitability is dependent only on the environmental aspect of ESG, with a negative relationship: higher environmental scores are associated with lower profitability. Additional studies have explored the relationship between ESG ratings and returns (e.g., Gavrilakis and Floros, 2023; Engelhardt et al., 2021; Karoui and Nguyen, 2022). 1 Brazil, Russia, India ,China and South Africa 5 However, the effectiveness of diversification strategies can be compromised by interconnections between sectors, leading to a phenomenon known as risk spillover (Eckert, 2020). Risk spillover occurs when events in one sector, country, or market influence another. This phenomenon has been plainly demonstrated by events such as the Covid-19 pandemic and the ongoing war between Russia and Ukraine. As the global economy becomes more interconnected, the potential for risk spillover increases, making it a critical area of study for investors. Diebold and Yilmaz (2009) and Baruník and Křehlík (2018) have developed two prominent methods to measure risk spillover, known as the DY spillover index and the BK spillover index, respectively. The DY spillover index measures the "total" amount of risk spillover one sector transfers to another, such as the effect of a shock in the financial sector on the industrial sector (Diebold and Yilmaz, 2009). The BK spillover index emphasizes different time frequencies, examining how spillover varies depending on the timing of the event (Baruník and Křehlík, 2018). In our study both methods are valuable to use. The DY Spillover Index can provide a more overall picture of the spillover effects within ESG investing but the BK Spillover Index may add more depth showing us how these effects may vary across different time scales. These models initially focused on regional indices but have since been applied to various contexts, including ESG-specific indices. Gao et al. (2022) utilized these models to investigate risk spillover between ESG indices in both developed and developing economies, finding that risk spillover is most significant in developed markets such as North America and Europe. They argue that factors such as the advancement of green finance and the openness of financial markets, in addition to whether an economy is advanced, influence risk transmission. Furthermore, significant global non-financial events, such as the war between Russia and Ukraine, contribute to heightened levels of ESG market risk spillover. This indicates a more pronounced impact, given that previous studies and applications of the methods have focused on regional indices. In our case when applying the indices to sectoral data it can offer us new insights into the dynamics of risk spillover both withing and between sectors. Despite these advancements, limited research has been conducted on risk spillover between sectors. If sectoral risk spillover is significant, traditional diversification strategies may be less effective than previously thought. Understanding which sectors influence each other will provide a more detailed overview of how an investor can reduce risk by selecting sectors with little interconnectedness. This thesis aims to fill this gap by exploring the unexamined 6 relationship between sectors and the potential for risk spillover. Understanding these dynamics can provide valuable insights for investors seeking to diversify their portfolios more effectively. Previous studies offer mixed conclusions on sectoral versus geographical diversification. Laopodis (2016) found that certain sectors can act as predictors for others, suggesting interconnectedness. Hong et al. (2007) reached similar conclusions, finding that the returns of industry portfolios can predict fluctuations in the stock market. In contrast, other studies indicate that sectoral diversification may be more effective than country diversification in reducing risk (Cavaglia et al., 2000; Chen et al., 2006). These conflicting views highlight the need for further investigation into sectoral risk spillover. Therefore, this study aims to answer the following questions: Research Questions: - How does risk spillover vary between different sectors focusing on ESG-companies? - If risk spillover varies between different sectors, what could the explanations be? To address these questions, this thesis will utilize the DY and BK spillover indices. While these models have primarily been applied to examine connectedness between different regional indices, their application to the sectoral environment remains unexplored. By doing so, this thesis will open a new avenue for investors regarding risk management and portfolio diversification. The study will be structured as follows. Firstly, the problem will be described and analyzed. Secondly, the purpose will be stated. Next, a literature review of existing research will be presented, followed by the development and presentation of hypotheses. After that, there will be a detailed description and discussion of the models used. Subsequently, the data will be presented and discussed. Then, the empirical results will be presented and analyzed in connection with the literature. Finally, conclusions will be drawn, and suggestions for further research will be provided. 1.2. Problem description and problem analysis As the world becomes increasingly interconnected, understanding risk spillover is crucial for both stakeholders and investors. Traditional investment strategies often suggest that diversification across different asset classes or sectors is the best way to reduce overall portfolio risk (Markowitz, 1952). However, if significant risk spillover exists between these investments, such strategies may be less effective. Understanding the dynamics of risk spillover enables 7 investors to diversify their portfolios more efficiently by focusing on sectors with lower intercorrelation, thus achieving better risk reduction. Moreover, examining risk spillover provides a nuanced approach to portfolio diversification. Investors with knowledge of these dynamics can make more informed decisions, enhancing asset allocation. This understanding allows for either higher returns for a given risk level or reduced risk for a set return, optimizing investment outcomes. The growing importance of Environmental, Social, Governance (ESG) criteria in contemporary investing underscores the need for such analysis. ESG considerations are increasingly integral to investment decisions, reflecting a broader shift towards sustainable and responsible investing. Most studies in the ESG field focus on the return characteristics of ESG stock, including their ESG rating risk profile (Garcia et al., 2017; Noy et al., 2020; Li et al., 2018). However, few studies have concentrated on the contagion of risk and risk spillover within ESG contexts. Gao et al. (2022) examined these aspects broadly, using cross-regional comparisons. Nonetheless, the regional indices used by Gao et al. (2022) included various industries, making it difficult to isolate how specific sectors are affected by risk spillover. To address this gap, a dataset concentrated on sectors should be utilized with the DY and BK spillover indices. These methodologies are particularly suited for analyzing the dynamic relationships and time-varying nature of risk spillovers across different sectors. This approach would illuminate how different industries vary in terms of the amount of risk received and sent, providing detailed insights into sector-specific risk dynamics. Understanding these dynamics and relationships between sectors will not only inform investment strategies but also be of aid in the development of corporate and public policies aimed at mitigating systemic risks. By identifying and managing sector-specific risk spillover, stakeholders can better navigate the complexities of global financial markets, ultimately contributing to more stable and resilient economies. 1.3. Purpose The purpose of this thesis is to deepen the understanding of risk spillover, specifically between sectors within ESG-focused companies. This will be achieved by testing the hypothesis that risk spillover varies between sectors. Given that this field is relatively unexplored, the research will provide valuable insights into sector relationships and establish a framework for further study. 8 2. Literature Review 2.1. Risk & Risk Spillover The research about risk spillover is quite extensive where a lot of subjects have been touched upon. A common theme within these studies is that the conditional value at risk (CoVaR) 9 method has been used to examine risk spillover in different environments. Adrian and Brunnerheimer (2016) describe CoVaR as a measurement of systemic risk and define it as the alteration in the financial system's value at risk under the condition of an institution experiencing distress compared to its median state. Risk spillover or spillover effect occurs when events, both small and large in scale and for example the Russia-Ukraine war, COVID-19 pandemic or the local estate market, effects the economy of other countries or nearby cities/states (Eckert, 2020). As of recent years, and decades the world has become united in terms of trades thus there is an increasing dependence between countries now in relation to before (WTO, 2023). Thus, it can be argued that, due to the increase of connectedness between countries, events in one country can have an impact in another country. As stated, the CoVaR model has been used in numerous previous studies. For instance, Akhtaruzzaman et al. (2022) investigate the risk spillover manifest from the United States to emerging markets in Africa during the COVID-19 pandemic. They concluded that the characteristics and extent of downside risk exposures in African financial markets parallel those observed in the US market, suggesting that African financial markets are also significantly impacted by the pandemic. Furthermore, the findings indicate that the United States serves as a net source of risk spillovers, while developing African economies act as net recipients during this period of global crisis. Moreover, Kumar (2023) and Jiang et al. (2020) analyses the risk spillover in European travel and leisure sector and oil to BRICS stock market respectively. Kumar (2023) found that there is a negative relationship between travel and leisure stocks and uncertainties. In other words, as uncertainties increase, the return of these stocks decline. Jiang et al. (2020) describes risk spillover as diverse and dependent on factors such as degree of oil dependence and energy policy. As risk spillover is diverse the consequences on the stock markets are also diverse. For example, major oil exporters, there's minimal financial risk correlation and systemic risk spillover between oil and the Brazilian stock markets, while the correlation is strongest between oil and Russian stock markets. 2.2. ESG Moreover, there has been thorough research into ESG. However, most of the studies are studying the relationship between ESG scores/rating and stock returns of companies. For instance, Gavrilakis and Floros (2023) examined ESG performance, herding behavior and stock returns in Europe whereas Engelhardt et al. (2021) analyzed ESG ratings and stock returns in Europe during the COVID-19 pandemic. Gavrilakis and Floros (2023) argue that investors in 10 five out of the six countries used companies with higher returns rather than high ESG scores. However, in Europe as a whole they reason that investors forfeit returns by buying stock of companies with a well implemented ESG strategy. As said, Engelhardt et al. (2021) examined ESG ratings and stock returns during the COVID-19 pandemic. Their research indicates that companies with higher ESG ratings exhibit greater resilience during crises and also deliver higher returns compared to those with lower ESG ratings. Furthermore, Karoui and Nguyen (2022) investigated the relationship between ESG exposure and stock returns in the United States and they found that firms with high social exposure often perform worse, in terms of returns, compared to firms with lower social exposure. Research done by Fatemi et al. (2018) and Li et al. (2018) further analyses the connection between ESG performance and firm value with respect to the disclosure of information. Li et al. (2018) argue that firm value can be improved by transparency, accounting proficiency and stakeholder trust. Fatemi et al. (2018) discovered that ESG information and its disclosure are positively correlated with corporate value, and conversely. Fu (2024) looks at the relationship between ESG score and returns for companies in China. The author finds a negative relationship between ESG and stock returns and argues that it can be explained by the relatively little ESG investment in China. However, the time frame of the data is comparably small and hence, may not show the complete picture. Moreover, Shaikh (2022) finds that firms that invest in ESG disclosure perform better financially, arguing that the firms that are using resources on ESG disclosure are getting more attention. Besides studies examining the relationship between ESG scores and returns there are some concerns around the metric. For instance, Berg et al. (2022) and Chatterji et al. (2016) examine how ESG scores are measured differently depending on which institution is performing the process. Berg et al. (2022) compared six different rating institutions and the discrepancy relies on one hand on the measurement and on the other on the disagreement on the data that gets analyzed. Furthermore, Edmans (2022) sheds additional light on the problems of ESG and mentions that depending on the context, non-ESG companies can be looked at as an ESG company. One example, prior to the Russian invasion of Ukraine, was defense companies They were classified as non-ESG companies. After the invasion, questions were raised if they could be classified as ESG companies. Edmans (2022) further argues that while ESG is necessary and holds value, it should not be idealized or prioritized above essential factors. Another issue with ESG is greenwashing2 and comprehensive research has been done on the subject (see e.g., 2 Greenwashing promotes false solutions to the climate crisis that distract from and delay concrete and credible action 11 Delmas and Burbano, 2011; Testa et al., 2018; Torelli et al., 2020). However, Li et al. (2022) argues that due to the high level of information asymmetry in today’s emerging economies it may be hard to spot greenwashing. 2.3. Risk spillover within ESG Besides studies examining risk spillover in more general settings and the relationship between ESG and stock returns, recent studies have tried to combine the two. Previously mentioned Li et al. (2023) and Gao et al. (2022) are two examples. Kilic et al. (2022) conducted a similar study as Gao et al. (2022) and found different correlation depending on whether the country/economy is developed (for example, the United Kingdom and Germany) or developing countries/economies (for example, Argentina and South Korea). They differ in the aspect that developed countries tend to have a negative relationship between ESG returns and stock returns, while in developing countries the co-movement was positive. Gao et al. (2022) made similar findings in their study; the developed economies are often the source of risk spill over. Furthermore, significant international financial events contribute to an elevation in the level of ESG market risk spillover. Umar et al. (2020) adds additional insight to the subject when they compare how major ESG index handles systematic chocks compared to more traditional equivalents. Their findings are similar to the previously mentioned, developed countries play a substantial role as net contributors to other markets, while developing countries act as net recipients. Thus, one can argue that there is some sort of consensus within the field. Larger actors play the role of distributing while smaller actors often receive and do not influence others. Nevertheless, these studies often take the perspective between countries and market without adding the dimension of how different industries differ and how they get impacted by shocks to the system. 2.4. Sector dependence/independence There has been extensive research on the subject concerning sectors' differences with respect to risk, shocks and how they are connected to each other. For instance, Baca et al. (2000), investigates the impact of country versus industry effects on return variation. The authors found that country effects have declined during the two last decades, meanwhile sector components (United Nations) 12 have either been the same or increased. They argue that, due to this, there is a sign that the economy and sectors have been more dependent and connected to each other. Cavaglia et al. (2000) found similar results and suggest that sector factors are becoming more important than country factors. Furthermore, they propose that diversification across sectors has become more important than diversification across country in order to reduce risk. There can be an argument made, based on Cavaglia et al. (2000) findings, that risk does not transfer between sectors. Chen et al. (2006) compared the country versus sector factors in developed and emerging markets. In the developed markets there was a shift from country effects to sectors effects being more important. In the emerging markets, country-specific factors exerted a predominant influence throughout the entire sample period spanning from 1994 to 2005. Despite the prevailing dominance of country effects in emerging markets, there is a conjecture that this dynamic may evolve, aligning more closely with the patterns observed in developed markets (Chen et al., 2006). Laopodis (2016) investigates the predictive role of industry portfolios on variables such as inflation and market dividend yield. The author posits that sectors, such as oil and financials, can serve as information sources, providing data about the stock market one to two months in advance. The study concludes that certain sectors serve as informational leaders for others, suggesting that developments in one sector can impact and spill over to others. Hong et al. (2007) came to similar conclusions. They found that the return of industry portfolios is able to predict fluctuations in the stock market. 2.5. Hypothesis development As mentioned earlier, there is growing evidence suggesting that sector effects are becoming more significant in mitigating risk compared to country effects. This observation falls in line with the increasing globalization and integration of financial markets, where sector-specific factors often outweigh regional influences. Diversification across has been shown to effectively reduce overall portfolio risk, implying that risk does not uniformly transfer between sectors (Cavaglia, 2000; Chen et al., 2006). However, Laopodis (2016) presents a counterargument, indicating that sector information and performance can predict outcomes in other sectors. This suggests that intersectoral linkages and dependencies may exist, allowing risk to spill over from one sector to another. Such spillovers could be driven by factors like supply chain interdependencies, technological advancements, or macroeconomic condition affecting sectors simultaneously. For instance, 13 Autor (2015) discusses how development in technology will change jobs in the future, for example with machine learning etc. Classens and Kose (2013) examines how financial crisis occur and how they are affecting economies outside the core of the crisis. As previous studies have shown that risk spillover varies, and one country/region is not sending or receiving similar amount of risk as others it is plausible to assume that this study will find similar results. However, it is important to look at the different characteristics between this study and previous studies. As they are looking more at regional differences, this study is focusing on sectors and the geographic aspect is not as prominent as others. Countries from every continent are included in each sector. The contradictory arguments present two options for the hypothesis: either that risk spillover does not differ between sectors or that it does differ between sectors. Given studies demonstrating the efficacy of sector diversification in mitigating risk, the null hypothesis posits that risk spillover does not differ between sectors. 2.5.1. Hypothesis Statement: 𝐻0: Risk spillover among ESG-companies does not differ across specific industries. 𝐻1: Risk spillover among ESG-companies does differ across specific industries. 3. Method, Model description & Theory The models used in this study are models that are widely accepted in the ESG sphere, the study does not suggest any models that have not been previously used. Measuring risk spillover can be done in different ways, this study advocates two main ways to quantify spillover shocks from one asset market to another. As the study will compare several markets regarding risk 14 spillover analysis, there are numerous models to use for comparing markets. A common way to go is to begin with the Forecast Error Variance Decomposition (FEVD) model. This model provides a framework to study the dynamics of stochastic volatility between different selected markets (Lanne and Nyberg, 2016). It allows for the cross examination of how shocks in one market may influence the variance, consequently the risk in another market over time. Hence, this model will especially be useful in understanding the interrelation of different markets and the budding for risk contamination (Fu and Qiao, 2022). The Forecast Error Variance Decomposition works in a way that it may provide a way to quantify a contribution of each determined variable to the forecast error variance of all other variables in a multivariate system (Lütkepohl, 2005). Consequently, this model is a useful tool when being applied in ESG study as it allows for the examination of how shocks or changes in one market may influence the variance, hence the risk in another market over time. When measuring FEVD it is common to break down this model into smaller segments. Diebold and Yilmaz (2009) proposed a model or an index which gives a more comprehensive view of understanding the use of the FEVD model. The index provides a comprehensive measure of the magnitude of risk spillover between several markets. Hence, the index can be useful when understanding the interrelation of financial markets and potential of underlying risk (Diebold and Yilmaz, 2009). The model is in turn based on the vector autoregressive model (Engle and Susmel, 1993). The Diebold & Yilmaz DY spillover index model works in such a way that it describes volatility spillover based on the VAR model. It demonstrates by showing the contribution of volatility in price indices related to the forecast error variance. As with the FEVD, the DY index is based on variance composition. The DY spillover index is the determined part of the forecast error variance for a specific asset that is attributable to shocks on other assets (Diebold and Yilmaz, 2009). In ESG investing, the DY Spillover index is important as it provides a “framework” to apprehend the interrelation of different markets, e.g. for ESG assets (Liu et al. 2021). The DY Spillover index can be a useful tool to analyze how returns and volatility "spill over" from other markets to an ESG index. This helps us understand how external market changes can impact the performance of ESG investments (Liu et al. 2021). The DY Spillover Index can aid investors to understand the potential risk which is associated with ESG investments. It can in turn apprise investments and risk management strategies (Moosawi and Segerhammar, 2022). This is implemented by measuring spillover effects between indices and other benchmark indices for eg. return and volatility. Another advantage 15 of the DY spillover index is that it may give beneficial comprehension of a diversification of a portfolio. Research by Moosawi and Segerhammar (2022) suggests that diversification with various country-level ESG indices within a portfolio leads to lower overall risk. This is because ESG indices can experience different performance patterns compared to traditional assets, reducing the contagious effects of negative shocks on the portfolio. However, there are several problems with this model. These issues may be facilitated by using a similar model but with a different approach. We will therefore introduce a new model with similar characteristics to the DY Spillover index but whereas the methodology and focus differ, hence the information these models provide differ. These problems can be mitigated by using a similar index, the so-called Barhuník and Krehlík Spillover index. The BK Spillover index and the DY Spillover index are in several ways similar however, the difference in methodology and focus may have a great impact on how the indices perform regarding ESG investing (Gao et al. 2022). The DY Spillover Index is focused on the notion of forecast error variance decomposition within a Vector Autoregression framework. It applies a vector autoregression framework to make a decomposition of the forecast error variance of an asset and evaluates the contribution of past volatility shocks in another asset. In other words, it provides a measure of a “total” spillover effect occurring at a specific given point in time. It can therefore calculate how much of an asset’s volatility in the future that may be explained by historical volatility variation in another (Zhao et al. 2023). Hence, this model has its advantages. However, according to same study the BK Spillover Index has an advantage due to the fact that it can measure the magnitude and “direction” of (risk) spillovers across different frequencies. This is important because it may contribute more comprehensive and detailed information about the spillover dynamics (Zhao et al 2023). They mean that the BK Spillover Index is therefore a better method to use when examining time frequency analysis, due to the fact that the BK Spillover Index examines how spillovers change over different time periods such as a short-term, medium-term or longterm perspective (Zhao et al. 2023). 3.1. Models – DY Spillover Index The Diebold and Yilmaz (DY Spillover Index) is calculated based on a forecast error variance decomposition from a Vector Autoregression (VAR) model. We chose to use the following formula for calculating the Vector Autoregression: 16 EQ 1:𝑌𝑡 = 𝐴1𝑌𝑡−1 + 𝐴2𝑌𝑡−2 + ⋯ + 𝐴𝑃𝑌𝑡−𝑝 + 𝑢𝑡 The Yt on the left side of the equal sign represents the return of ESG investments in different sectors. The Yt on the right of the equal sign is a vector containing the returns of ESG investments in different sectors at a specific point in time (t). The returns are considered endogenous due to the fact that the VAR model assumes that they are explained by a combination of how they performed in the past and the past performance of other ESG sectors within the model. To make it easier, one can say that they are not random variables, but they are influenced by internal factors which are captured by the VAR model. Ap are matrices of different coefficients to be estimated and p stands for the lag order of the model and ut is a vector of error terms. This is in turn based on the Forecast Error Variance Decomposition (FEVD). Hence you will perform a FEVD based on the calculated VAR model. The FEVD may be calculated using the following formula: EQ. 2: 𝜎𝑖𝑗2(ℎ) = Σ Σ𝐻ℎℎ𝐻==11𝑒𝑒𝑖𝑗𝑖𝑖22((ℎℎ)) Where 𝜎𝑖𝑗2(ℎ) tells us the proportion of the h-step (number) ahead of the forecast error variance of variable i due to the shocks to variable j. 𝑒𝑖𝑗2 (ℎ) is the forecast error variance of variable i due to shocks related to variable j at horizon h. and where H is the maximum forecast horizon. When both the VAR model and FEVD are calculated you can formulate an appropriate DY Spillover Index. Consequently, the DY Index may be written as: EQ. 3: 𝑆𝑖𝑗 In the DY index 𝑆𝑖𝑗 is the spillover from variable j to variable i. n is the number of variables. In this case it means number of selected/chosen industries. The sums are over each one of the variables except for the variables associated with each other or in other words for the diagonal elements. That is i.e., j ≠ i. This formula provides a measure of the total spillover effects from one industry to another at a given point in time. It can therefore be used to analyze the risk spillover effects in ESG investing across different industries over time. 17 3.2. Models – BK Spillover Index When using the BK Spillover index, it is interesting to calculate the index based on a wavelet decomposition. This allows us to analyze how the spillover effects change over different time scales such a short-term, medium-term or long-term. By using the wavelet decomposition, we are allowed to break down the data into components associated with different scales or frequencies. In the context of ESG investing, a wavelet decomposition could then be used to analyze the time-series data of different ESG investments. The process of a wavelet decomposition will result in a set of wavelet coefficients which may represent the information in the initial set of time series at different scales. When obtaining these coefficients, these could be used to analyze the relationships between different investments over different time periods or at various time scales (Chen et al. 2022). We represent the wavelet decomposition using following formula: EQ. 4: 𝑋 𝑑𝑗𝑘 Ψ𝑗𝑘(𝑡) + 𝑎𝐽0Φ𝐽0(𝑡) Where X(t) is the selected time series data. 𝑑𝑗𝑘 are the wavelet coefficients. Ψ𝑗𝑘(𝑡) are the wavelet functions. 𝑎𝐽0 is the approximation coefficient at the final scale of J0 and Φ𝐽0(𝑡) is the scaling function. Afterwards we calculate the spillover index on the results we have obtained from the wavelet decomposition. We write the BK Spillover Index interpreted as: EQ. 5: 𝑆𝑖𝑗 𝑆𝑖𝑗(𝜏) is the spillover from variable j to variable i at frequency 𝜏 . . 𝜎𝑖𝑗2(𝜏) is the wavelet gamut of variable i due to shocks to variable j at frequency 𝜏 . . Further, the sums are over all frequencies. A higher result, that is a higher spillover index at a specific frequency, indicates a greater degree of risk spillover between variables at that particular frequency. To perform the model and wavelet decomposition we used information and statistics calculated with help of statistical software package STATA. 3.3. Comparison to other models (GARCH and CoVaR) While both the DY Spillover Index and the BK Spillover Index are valuables methods to use for examining risk spillover within the context of ESG investing there are other models or rather techniques that can be used or complemented to these indices to give a more comprehensive understanding of risk transmission in ESG markets. 18 3.3.1. GARCH This Generalized Autoregressive Conditional Heteroskedasticity (GARCH) method created by Danish economist Bollerslev (1986) is a further development of the ARCH method which was created by American economist Engle (1982). This method or rather technique can be used to complement the Vector Autoregression (VAR) model. An important key assumption of the VAR model is homoscedasticity. This means that there is an assumption that the variance of the error terms have constant variance across all observations and across all time periods. However, in real world practice this may not be true. Data may instead exhibit heteroscedasticity. In this case the error-terms will exhibit non constant variance which may lead to biased standard errors, consequently unreliable model and hence misleading results. The GARCH model is designed to confront heteroscedasticity in error terms. The GARCH method works in a way that it captures volatility as a cluster, this will eventually lead to a more “realistic” view of risk. When a VAR model experiences heteroscedasticity, the relationships which are calculated between variables may tend to be inaccurate. If one chooses to incorporate a GARCH model to complement a VAR model to account for changing variance, the VAR itself become more accurate, hence reliable. This will in turn lead to more meticulous forecasts of the variables. 3.3.2. CoVaR Another method to use is Conditional Value at Risk (CoVaR). While the DY Spillover Index and BK Spillover Index can provide information of how much risk spills over, it will not provide us information of how severe of eventual losses at time when spillovers occur. Instead, one could complement with CoVaR. This method is a measure of systemic risk that will quantify the potential for losses in a portfolio. Usually, you use CoVaR when you want to stress test a portfolio or have requirements in risk capital. For example, CoVaR can be used to stress test an ESG portfolio against extreme situations such as scandals or climate related disasters. CoVaR measures the systemic risk within a financial system. CoVaR can therefore provide valuable insights and a more comprehensive view of risk in financial markets than solely relying on spillover indices. 3.3.3 Conclusion models The models used in this study are the DY Spillover Index and the BK Spillover Index. While established models like GARCH are valuable for volatility analysis, they do not directly quantify risk spillover between markets. The DY and BK Spillover Indices were specifically designed for this purpose, making them particularly suitable for investigating how risk 19 transmits across ESG sectors. The DY Spillover Index offers a clear picture of the total spillover effect at a specific point in time, valuable for understanding the overall risk transmission between ESG sectors. The BK Spillover Index goes a step further by analyzing spillover across different frequencies (short-term, medium-term, long-term), which is crucial for ESG as risk dynamics might differ across time horizons. 3.4. Model Discussion 3.4.1. DY – Spillover Index The DY Spillover Index is a valuable tool to use to examine our stated hypotheses. Unlike traditional correlation methods or measures, the DY Spillover Index will consider the time varying nature of risk transmission. This is of crucial importance since risk spillover can fluctuate over time. The DY Spillover Index can capture these changes. It provides more information than solely identifying the presence of spillover. Instead, it also provides information on the direction, that is which industry transmits risk to another industry and the strength of the spillover effect. This helps us understand not only if but also how different industries differ in risk transmission. This allows us to examine if an industry is a net transmitter or net receiver of risk. The DY Spillover Index is designed to analyze risk “connectedness” among several industries simultaneously. In our case this is very interesting since we are comparing risk spillover across different sectors. By applying the DY Spillover Index to industry sets we may identify how risk spillover differs between industries for ESG companies, hence it allows us to test the hypothesis. By calculating the DY Spillover Index for ESG industries, we can compare the index values across industries. If there is a significant difference in the values of the index, it could suggest that industries differ in risk spillover. 3.4.2 Relation to ESG Both the DY Spillover Index and the BK Spillover Index are valuable tools when analyzing risk spillover in the context of ESG investing, however they offer different perspectives. The DY Spillover Index may provide an understanding of the direction and sources of volatility spillovers between ESG assets and other assets in a comprehensible way. It may acknowledge if an ESG asset is a net transmitter of/or a net receiver of volatility (Hasan et al. 2023). The DY Spillover Index can assess and study these effects of correlations. A study also found that the DY Spillover Index may be used to study the contagious effect of networks within ESG investments (Li et al. 2023). This concludes the need for a more resilient and more sustainable financial system (Li et al. 2023). Further studies also showed that the index may provide info 20 regarding portfolio optimization. Studies found that investors could receive information that there was lower volatility spillover between “green” bonds and ESG stocks during both more tranquil and turbulent periods such as the COVID 19 or the war between Russia-Ukraine. For investors this information could be valuable since it could assist investors seeking to hedge their portfolios (Hasan et al. 2023). However, even though the DY Spillover Index provides a valuable method of examining risk transmission in ESG investing it may lack a more detailed view of transmission of risk. Instead, the BK Spillover Index can be useful to use since it allows a time-frequency analysis, which provides a more detailed view of spillover effects over different time periods. Studies found that it can be particularly useful when you want to understand how risk transmission may vary over time over a short, medium and long-term analysis (Agyei et al. 2022). This may be crucial for ESG investing where some risks might have long-term implications to consider. The BK Spillover Index may have a better chance to apprehend a quicker and abrupt change in the volatility spillover effects of an ESG Index. This is especially common during periods of market turmoil. This can in turn assist investors in managing their risk exposure during such periods in a more desirable way (Liu et al. 2021). When using the BK Index, it can differentiate between positive and negative volatility spillovers. This is especially important in ESG investing. For example, good news e.g., within the energy sector could eventually lead to a positive impact on other ESG sectors not directly affected by a breakthrough in the energy sector. Since the BK Spillover Index provides an “insight” into the frequency domain. It can be shown that the preponderance of the return spillover effects of ESG index are concentrated in the short-term. At the same time the majority of volatility spillover effects occur in the long-term. This is interesting for investors since it can advise investors for investments strategies that are bespoken to different horizons (Liu et al. 2021). 3.5. Are there any disadvantages of using the DY Spillover Index as a measurement for risk spillover? From the above text you can conclude that there are some disadvantages using the DY Spillover Index as a tool to measure risk spillover. Firstly, the DY index will provide the reader with a “total spillover effect” at a specific point in time. This could be a problem since it does not capture how risk spillover could fluctuate over both shorter and longer timeframes. This could eventually be a limitation when examining risk spillover within ESG investing since they have 21 both short and long-term implications. Also, since the DY spillover Index provides a measure of total spillover effect it may be difficult to accurately recognize the direction of spillovers. It could therefore be arduous to identify transmitter and or receivers of risk. Furthermore, the DY index solely focus on the direction of risk transmission but only when it is in the context of “negative” spillovers. Hence, it does not differentiate between positive and negative volatility spillovers. In the context of ESG investing this could be a problem since positive developments in one specific ESG sector can have a positive spillover effect on another ESG sector and so on. Another aspect to consider when using the DY Spillover Index is that it is based on different assumptions and decompositions. It primarily focuses on variance decomposition. This may not capture all forms of market contagion. A study found that the DY Spillover Index might overlook other forms of spillover which are not captured by measures within variance (Meng and Chen, 2023). Another issue with the DY Spillover Index is that it is ordinarily calculated pairwise. In our case it means between two industries. This may however not be optimal for capturing spillover dynamics within ESG industries. ESG industries usually have interdependencies that may not fully be captured through pairwise comparisons. Pairwise comparisons may oversimplify the dynamics of risk spillover. This could lead to under –or overestimation of the actual risk within the whole network of industries. Also, systemic risk such as collapse of the entire system is not appropriately addressed by pairwise comparisons. Therefore, the DY Spillover Index might not display the aggregate risk that affects the market as a whole. The DY Spillover Index relies on the assumptions of the VAR model. This could pose some problems that could affect the results of the usage of the model. Since VAR models assume that time series data are stationary. There could be a problem if the data exhibit trends which could be the case when looking at financial time series data. This could lead to the results of the VAR, consequently the DY Spillover Index may be misleading. A VAR model also assumes a linear relationship among variables. If financial markets exhibit market stress or financial crises the relationship may be non-linear which may in turn lead to under–or overestimation of spillover effects. Furthermore, will the VAR model assume that the error terms are exogenous and hence not correlated with past variables. This means that if there is autocorrelation in error terms, the estimates may be biased. Also, if there is an assumption that there are constant variables in error terms. In other words, if there is homoscedasticity. This could also lead to issues when applying the DY Spillover Index. Financial time series data will often exhibit 22 heteroscedasticity. This means that the variance will change over time. This is especially common during high volatile periods. In terms of DY Spillover Index this is a problem since the heteroscedasticity may affect the accuracy of the forecast error variance decomposition (FEVD) which could lead to inaccurate results when applying FEVD to the DY Spillover Index. So even though the DY Spillover Index is a common and reliable method to use when examining risk spillover within financial markets, violations of assumptions may lead to inaccurate spillover measurements which in turn can misinform investors about the risk transmission in financial markets. One way to complement spillover indices is to use as mentioned earlier the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model to have a better understanding of handling volatility clustering. 3.5.1. Importance of industry context in ESG investing The impact of ESG factors can vary significantly between different industries. If you look at environmental factors such as energy, industrial or materials alternatives these industries can have a significant impact on the environment through emissions of greenhouse gases or waste production or the usage of resources. While on the other hand sectors such as financial services or information technology might have less direct environmental impact. However, these industries may still play a role through their investment decisions or the usage of energy. Industries that are labor-intensive, in this study, sectors as consumer staples, health care or consumer discretionary could have to take into consideration significant social issues related to working conditions or labor rights or how they affect communities. In opposite, sector as information technology or financial services might have a greater focus on issues regarding data -or cybersecurity or equal accessibility to services. Even though one should consider that governance factors are a crucial part for almost every sector or company, specific issues may vary. Companies within the financial services sector for instance might have to focus on risk management or regulatory compliance while sectors such as information technology or health care could confront issues related to data privacy or regulations. When conducting a risk spillover analysis within ESG investing, it is of great importance to consider the context, hence also the characteristics of each specific industry. Bespoken approaches that take into account industry specific ESG risks and furthermore opportunities may provide a more genuine and distinct understanding of a company’s ESG performance. This can lead to more enlightened and well-informed decision-making for all interested parties and stakeholders including both investors and regulators. This type of approach may not only help stakeholders to identify companies that do not perform well in the long run, but it could likewise contribute to overall goals of sustainable advancements. Consequently, it could be a win for all 23 parties involved. That is both for the financial world and our society as a whole. So, if you consider the specific context of each industry involved, ESG could be a very forceful tool for both creating value and advancement of positive change. 3.5.2. Limitations of a one-size-fits-all approach There are several ways to address the limitations of a one-size-fits-all approach. These can be used for recognizing different characteristics and hence impacts of each industry. Different industries have different ESG risks and also opportunities. In our study you could for example look at the energy sector or information technology sector. In the energy sector one should expect that carbon emissions could be an important metric to use when evaluating environmental footprint. However, when looking at the information technology sector it could be more interesting to have questions regarding data privacy. It is therefore of great importance to understand the context of which companies operates in. This means that it is important to conclude which ESG environment is of value or interest for that specific company. One could look at both how the regulatory environment is shaped or maybe where the company operates, that is literally where the geographical location of where the company’s operating is located. One could even look at societal norms. When considering these factors, it may have a fundamental impact on a company’s performance withing ESG. Another way to address the limitations of a one size model is to create an industry-specific benchmark. That is, you could compare companies with their industry peers. You could for example develop an ESG criterion which is specified to each industry based on what ESG impact it could have. This criterion should mainly focus on ESG factors that may have a significant impact on the industry’s probability or sustainability. If we look at our chosen industries, eg. the energy sector or information technology sector. An ESG criterion for the energy sector could be the percentage of renewable energy in a company’s portfolio. In contrast to this if you look at the information technology sector it could be more interesting to examine for example how high a transparency scores the industry or company could reach. This could be based on the detail of a company’s privacy policy by analyzing policy’s language, organization or accessibility, or how long it could take for management to resolve and report data breaches. Another important consideration is to analyze whether one could add qualitative data with quantitative data to gain a deeper understanding of how an industry or company is approaching ESG issues. This could eventually be a more accurate assessment by compiling those factors 24 that have the most significant impact on an industry’s sustainability practices. In this study this could be implemented by for example in the energy sector, environmental factors could hold a higher weight compared to information technology sector or those sectors that are more labor intensive, where social factors such as labor practices would be more heavily weighted. 4. Data The primary source of the data collection of this thesis is gathered from Refinitiv Eikon. The total sample size of the study is 4069, these samples are divided into 11 sectors defined by Global Industry Classification Standard3 (GICS) (see table 1.). The reason that GICS is being used is because it is well-known and used in previous studies within the field. Furthermore, it gives an overview of how different companies are classified and what the difference between them. Therefore, it is a well fitted classification system for this thesis. Moreover, there are differences in sample size in each sector (see table 1.). However, the size of each sector is large enough to reduce anomalies in the sample, thereby minimizing the risk of skewing the data and ensuring a more comprehensive and accurate representation. 3 Global Industry Classification Standard was developed in 1999 by MSCI and S&P Down. It offers an overview of 11 sectors, with 24 industry groups. This capture both the complexity and development of sectors as well as offers a tool for comparison (MSCI, 2020) 25 To measure the aspect of ESG, this study includes companies with an ESG score of at least 50. The decision to use 50 instead of a higher threshold is to avoid reducing the sample size of certain sectors, which could compromise the validity of the findings. While this approach may not capture the full extent of ESG influence, it ensures the study remains legitimate by maintaining a sufficient sample size. For instance, when processing the data there were attempts made to increase the threshold to 60 and 70. Nonetheless, the already relatively low sample size in sectors such as energy and communications services was even lower. Furthermore, the data set with a higher limit excluded certain regions and thus it would eliminate important perspectives. Refinitiv Eikon defines ESG-score as follows: “Refinitiv ESG-score is an overall company score based on the self-reported information in the environmental, social and corporate governance pillars.” (Refinitiv Eikon, 2024) 4 . Furthermore, other filters used in the data collection are the YTD for the last ten years, that is the lagged YTD for the previous ten years. The idea of using a ten-year span is to get an overview that is not heavily influenced by certain events. If the time span was lower the findings may be affected and not show the complete picture. 4.1. Descriptive Statistics Table nr. 1. Description of each GICS sector Sector Industry group Sector sample size Communication Services Telecommunication Services, 173 Media & Entertainment Consumer Discretionary Automobiles & Components, 472 Consumer Durables & Apparel, Consumer Services, Consumer Discretionary Distribution & Retail 4 Refinitiv Eikon is now known as LSEG, in the reference list the link is to a LSEG site. https://www.lseg.com/en/dataanalytics/sustainable-finance/esg-scores 26 Consumer Staples Consumer Staples Distribution & 302 Retail, Food, Beverage & Tobacco, Household & Personal Products Energy Energy Equipment & Services, Oil 186 Gas & Consumable Fuels Financial (services) Banks, Financial Services, 581 Insurance Health Care Health Care Equipment & 305 Services, Pharmaceuticals, Biotechnology & Life Sciences Industrial Capital Goods, Commercial & 724 Professional Services, Transportation Information Technology (IT) Software & Services, Technology 388 Hardware & Equipment, Semiconductors & Semiconductor Equipment Materials Chemicals, Construction 473 Material, Containers & Packaging, Metals & Mining, Paper & Forest Products Real Estate Equity Real Estate Investment 297 Trusts, Real Estate Management & Development Utilities Utilities 168 27 Table 1 gives an overview of the different sectors and the industries within each sector and also the sample size of each sector. Table nr. 2. The descriptive statistic of the return data of each sector of the GICS Sector 10-year Price YTD YTD ESG ESG PCT Change Std.de Mean Std.dev Mean Std.dev mean v Communicati 45.4% 3.66 3.1% 0.207 66.66 9.76 on Services Consumer 88.4% 2.88 2.4% 0.245 65.52 9.86 Discretionary Consumer 44.5% 1.66 3.8% 0.182 66.13 10.59 Staples Energy 30.5% 1.97 5.6% 0.285 65.43 10.06 Financial 70.9% 1.97 2.3% 0.193 65.92 10.71 (services) Health Care 141.6% 2.23 4.6% 0.227 65.93 10.83 Industrial 110.0% 2.78 4.1% 0.220 64.91 9.81 Information 304.5% 9.17 6.0% 0.244 64.84 10.02 Technology (IT) Materials 132.7% 4.61 7.2% 0.288 65.99 10.41 Real Estate 20.0% 1.82 2.5% 0.189 65.64 9.75 Utilities 37.3% 0.99 3.7% 65.16 9.90 0.156 Table 2 gives an overview of the YTD returns and the YTD standard deviation. The distribution of the return of each sector is fairly even, with some exceptions such as the materials and IT sector. Furthermore, the 10-year Price PCT Change is included to show the different industries, on average, have developed during the last ten years. It shows the percentage difference between the closing price of the latest completed tradable day and the closing price from 10 calendar year ago. There are certain differences between these sectors, for example IT has an 28 extensive increase and on other hand, real estate and energy still have a positive progress but not on the same level as the IT sector. Regarding the ESG part, the sectors exhibit similar mean and standard deviations. This indicates a consistent level of ESG performance across different sectors. The homogeneity suggests that, on average, the sectors are relatively aligned in their ESG practices and commitments. 5. Empirical Results 5.1. VAR In the initial stages of setting up the DY spillover index a VAR model was estimated (see EQ 1). The VAR model was used with a lag of 1 and the main outcome of the VAR model was that three of the sectors (Financial, Real Estate and Industrial) were omitted from the model and by extension omitted from the spillover index (see Appendix 1A). The reason for them being omitted was due to collinearity. The correlation between these three sectors and the others was too high and including them could give inaccurate results from the spillover index and thus giving a deceptive picture of the situation (see Appendix 1A). 5.2. DY Spillover Index Table 3. - Spillover Index (in %) Commun Utilities Material Health IT Energy Consume Consume From ication Care r r staples Service Discretio nary 29 Commun 21.658 12.351 8.563 21.406 18.770 3.345 7.146 6.761 99.91 ication Service Utilities 17.279 9.945 6.367 32.395 16.624 4.202 6.431 6.757 100 Material 12.222 18.780 29.342 3.676 10.616 7.299 16.739 1.326 99.998 Health 3.022 5.033 8.310 4.815 44.118 13.659 11.509 9.533 99.999 Care IT 4.146 1.763 0.550 3.849 82.378 1.158 5.587 0.568 99.999 Energy 1.749 6.347 24.804 2.806 16.692 23.056 22.719 1.828 100.001 Consume 1.315 0.739 0.412 0.859 1.677 0.428 0.726 93.844 100 r discretio nary Consume 22.204 11.025 3.099 14.607 28.287 7.079 13.524 0.175 100 r Staples To 83.593 65.983 81.0762 88.228 219.162 60.226 84.381 120.784 Net 16.317 34.017 18.9218 11.771 -119.163 39.774 15.619 -20.784 The data presented in the table above shows the outcome of the DY spillover index using EQ 3. The rows are showing the impulse the variable, in other words how the sector is sending risk to other sectors. In the columns are the response variables, meaning what the sector is receiving in terms of risk from another sector. The “To” row shows the aggregate risk spillover a sector is getting from other sectors and the “From” column aggregates the amount of risk spillover a sector is sending to all the other sectors. The “Net” row displays the net difference between the “From” and “To”. 30 The accumulation of the amount of risk spillover that each sector is sending is around 100%. However, there are differences depending on the pair of sectors that are sending/receiving. For instance, there is not a clear pattern that when the impulse and response is the same sector there is an abnormal figure. IT to IT is 82.378% whereas Health Care to Health Care is lower at 4.815%. Furthermore, when Consumer discretionary and Consumer Staples are acting as an impulse the outcome differs a lot. To begin with, Consumer Discretionary is sending relatively little to other sectors other than the sector itself. On the other hand, when Consumer Staples is acting as the impulse variable it is sending more to the other sectors than to itself. In the receiving row “To” there are some differences between every sector. Some sectors, for example IT and Consumer Staples are obtaining the most and Utilities and Energy are collecting the least. Similarly, to “From” there is variation in the risk spillover depending on the pair of sectors that are being looked at. Furthermore, for each sector there is not a regular distribution, meaning that one sector is not sending equal amount to the other sectors. In the “Net” row two of the eight sectors are acting as net recipients, meaning that they are receiving more than they are sending. The two sectors, being IT and Consumer Staples, and IT, due to being the largest recipient, is also the largest net recipient with a net of –119.163%. The other six sectors act as net contributors, they are sending more than they are receiving. Utilities and Energy are the largest net contributors with 34.017% and 39.774% respectively. 5.3. BK Spillover Index Due to the inability to facilitate the BK Spillover Index properly, the model is excluded from the thesis discussion and will not be further evaluated or discussed. 31 6. Discussion The discussion will be divided similarly to the findings section but there will be a part where the findings from the two models will be discussed together. 6.1. VAR The estimated VAR model had an outcome where three of the eleven sectors were omitted. This was due to the collinearity between these three sectors and the other sectors, the reasoning being that they were highly correlated (see Appendix. 1A). Overall, the variables were relatively highly correlated, with the majority of them having a strong positive correlation and only one pair of variables having a negative correlation. The reason for the collinearity among variables could be due to various factors. For instance, the homogeneity of the dataset, where variables share similar characteristics, may contribute to collinearity. However, since the dataset consisted of over 4000 observations spanning ten years, the issue is unlikely to stem from its size. Instead, the time frame of the data might have influenced the collinearity. Widening the time frame could potentially reduce the correlation between the variables. Regarding the possibility of homogeneity in the dataset, efforts were made to ensure variation in the observations. For example, data from all continents were included, although the majority of observations were from Europe, North America, and Asia. 32 The presence of collinearity in the data leads to decreased precision and makes estimation more difficult. This means that the resulting outcome may not accurately reflect the true situation. However, by omitting these variables, the findings of both the DY and BK models could be argued to be more accurate. Nevertheless, the correlation between the remaining variables was still relatively high. Even though they were not omitted, the results may be questionable due to the high correlation among the other variables. This leads back to the earlier discussion where some sectors are able to influence other sectors and even act as a predictor (Laopodis, 2016, Hong et al., 2007). This could be one of the sources for the collinearity in the model. If sectors are able to influence other sectors, a movement in one sector may lead to a movement in another sector and thus the relation will be positive. Lastly, it is important to state the problem is acknowledged and taken into account when the analysis is further conducted, and it will be discussed when the findings of the models are being discussed. 6.2. DY Spillover Index The DY spillover provides an overview of how sectors both receive and transmit risk to themselves and other sectors. There is some variation in both the receiving and net rows, while in the sending column, every sector is transmitting approximately the same cumulative amount of risk. Firstly, the observation that every sector is contributing equal amounts of risk differs from previous studies (see Zhao et al., 2023 and Gao et al., 2022). Neither of these studies reported variables contributing around 100%; in most cases, these percentages were even lower. Possible reasons for the deviation in findings from this study could be the different variables studied and the unique characteristics of regional-wide indices and sectors. Furthermore, differences may arise from variations in the utilization of the models, resulting in different outcomes. Moreover, there exists considerable variation in how sectors perceive risk. One notable example is the diversity in the extent to which each sector influences its own risk. For instance, consumer staples and consumer discretionary sectors demonstrate minimal impact on their own risk, whereas the IT sector exhibits the highest impact at 82.378%. The first observation aligns with the findings of Laopodis (2016) and Hong et al. (2007), who argue that sectors can serve as predictors for one another. This phenomenon suggests interdependence among sectors, as certain sectors are impacted by others, confirming the predictive power of sectoral analysis. Nevertheless, two sectors, consumer staples and consumer discretionary, have abnormal outcomes compared to the other sectors. Thus, this could be due to the characteristics of these 33 two sectors, that are different compared to the other six. Looking into the two sectors shows that the two are heavily connected to the other sectors. For instance, in consumer staples food retail and household products are part of the sector, and in consumer discretionary automobiles and household durables are included. Furthermore, this could also be a reason for the high amount consumer discretionary is sending to consumer staples. Due to dependence of sectors the argument made by Cavaglia (2000) becomes disputed as the author stated that sector diversification leads to a reduction in risk. However, this is in relation to country diversification and thus it does not relate to the risk between sectors in isolation. Furthermore, the amount of risk each sector is receiving is varied. These findings align better with previous studies. Some sectors, for instance utilities and energy, are receiving less than what they are sending. While IT, the sector who is receiving the most, is obtaining around 219%. The other sector that receives more than it is sending is consumer staples. The reason for the relatively high amount in IT could be due to that the IT sector is dependent on other sectors and their demand for IT products. Especially, as of late, when IT has become a more prominent role for corporations. For instance, higher demand for cybersecurity and that companies are developed with more focus on technology and automation (Autor, 2015). However, going back to the previous section, IT is the sector who is sending the most to itself, which is a contributing factor to the high amount. The other sectors are sending approximately equal amounts to IT and thus, the previous argument may be the reason behind it. Moreover, concerning the two sectors, utilities and energy, that are receiving the least amount. Looking into these two sectors, they are sharing some common traits. For instance, they both supply daily needs such as water and electricity. Exploring more in depth shows that the sectors are affected by different sectors. Energy is influenced by itself the most, whereas utilities are affected more evenly by some sectors, for example communication services and materials. Because energy is primarily affected by itself, a similar argument as with IT can be applied. The energy sector has experienced significant development, particularly in the "green" energy field (Ahmed et al., 2022). Furthermore, it could be reasoned that energy's self-influence may be attributed to its relative insulation from disturbances in other sectors. Lastly, the net row shows the difference between from and to. Referring back to previous sections, there are two sectors that receive, in net, more than what they send. These sectors are IT and consumer staples. For the other sectors, they are sending more than what they are getting. The sectors that have the lowest net are communication services, health care and consumer discretionary. Health care is mostly influenced by communication services, utilities 34 and consumer staples. In the consumer staples, drug retail is included and thus it is sensible to spot the relationship between the two sectors. That there is spillover between these sectors. However, the relationship between health care and communication services is not as easy to spot. There could be an argument for the inclusion of advertising in communication services could be a potential explanation for it. Nonetheless, it would suppose that the communication services should be fairly even distributed across all sectors if advertising has a significant impact. As the findings show, that is not the case in the scenario. Compared to previous studies, such as Gao et al. (2022), our findings show a less even distribution. In their study, the net results were evenly split, with four out of seven cases showing a negative net and three out of seven showing a positive net. This discrepancy could be attributed to differences in the datasets, particularly geographical variations. If our study were to distinctly separate regions to better capture the effects of geographical differences, it might yield different outcomes. 6.3. Limitations The study has, as previously discussed, some limitations and mostly concerning the utilization of the model and the consequences that follows. To begin with, there is an assumption that the model fits for every sector, that despite the differences between the sectors the model should be able to set these aside and compute a reliable result. However, as the findings were computed, and the outcome was insignificant, it is reasonable to question if the “one-size- fitsall" approach was suitable for this study. As previously mentioned, the models assume that the characteristics of each sector are similar in regard to the ESG factors. However, as stated, this is not the case and therefore this could be an explanation for the insignificant findings that follow and a potential solution to deal with the issue is to take into account these differences. This could be done by isolating ESG and developing a better systematic approach when dealing with sectors that are operating in the same financial/geographical area. Moreover, additional potential issues with DY Spillover index could be the following. Firstly, one should consider the situation where you could see non-stationary events and a continuously evolving ESG landscape. In this study Table. 3 provides an overview of risk spillover effect between industries but it might not capture the “true” extent of risk spillover if the chosen data exhibits “non-stationarity”. That is if there are rapidly growing, evolving or changing ESG sectors, could show higher or lower spillover effect than the model captures. Another issue is that the DY Spillover Index primarily focuses on variance. However, there might not be a correlation between ESG sectors, that is it might be linear. This is a problem since it could lead 35 to the underestimation of spillover effects at times when society or different industries are facing stress or distress or during periods of consequential change. Furthermore, one limitation of the sector-specific analysis is that shocks to the economic system often impact the economy as a whole rather than isolated sectors. For instance, the Covid-19 pandemic led to nationwide lockdowns, making it challenging to measure how a shock affects a specific sector when the shock impacts the entire system. A solution could be to examine sector-specific shocks and evaluate how these shocks spread to other sectors. This approach would enable a clearer connection between sectors, allowing for a more accurate measurement of risk spillover. In contrast to Gao et al. (2022), who used regional indices to cover regional spillover, this study did not account for regional aspects. Gao et al. (2022) approach allowed them to analyze shocks within each region and how a shock in one region affects another. In this study, each sector includes various countries, regions, and continents, which absorbs region-specific shocks because a region is included in every sector. This leads to limitations in the findings. Sensitivity tests indicated that the results were insignificant (see Appendix 2A). Revisiting the discussion about the VAR models and the issue of a "one-size-fits-all" approach may explain these insignificant findings. Consequently, it is difficult to conclude or argue that risk spillover differs between sectors based on these results. 36 7. Conclusion The purpose of this thesis is to address a critical gap in in the understanding of risk spillover across sectors. Existing research in this area is limited, and this thesis aims to shed light on an important phenomenon. To address this knowledge gap, this thesis will investigate whether risk spillover varies across sectors. Our findings demonstrate that risk spillover does indeed differ significantly between sectors. The variation in risk spillover is substantial and depends heavily on the specific sectors interacting with each other. This result is significant as it lays the foundation for new research avenues in both risk spillover and ESG (Environmental, Social, and Governance) factors, while also providing additional insights into inter-sector interactions. Our study highlights the considerable variation in risk spillover across different sectors, underscoring the complexity of these interactions. This insight is crucial for stakeholders, including investors and policymakers, who must consider sector-specific dynamics when assessing risk. Moreover, our findings suggest that diversification strategies may not be as effective as previously thought. In addressing the dataset limitations of this research, several avenues for future exploration present themselves. Firstly, due to the collinearity issues encountered, extending the time frame of the data set could help mitigate these correlations and provide a more robust analysis. Additionally, adjusting the models to better account for sector-specific characteristics could yield more distinct results, offering a clearer explanation of the connections between sectors. Furthermore, the aspect of ESG remains not fully understood within the scope of this study. Including a benchmark group of non-ESG companies for comparison could reveal any significant differences between ESG and non-ESG firms. This approach would enhance our understanding of ESG’s potential role as a risk reducer/provider. 37 Lastly, the geographical connectedness within the dataset suggests that a more specified dataset could be beneficial. Incorporating events from specific geographical regions could provide more detailed insights, as the current study does not fully integrate the geographical context. By addressing these limitations, future research can build upon the foundations laid and further examine the complex relationship at play. In conclusion, this thesis provides valuable insights into how sectors interact with each other. By highlighting these variations in risk spillover patterns, this thesis provides valuable insights, and by being one of the first studies exploring this field, it opens up possibilities for further research. 38 8. References Adrian, T., & Brunnerheimer, K. M. (2016). CoVaR. American Economic Review, 106(7), 1705-1741. http://dx.doi.org/10.1257/aer.20120555 Agyei, S. K., Owusu, J. P., Bossman, A., Asafo-Adjei E., Asiamah O., & Adam. A. M. (2022). Spillovers and contagion between BRIC and G7 markets: New evidence from timefrequency analysis. PLoS ONE, 17(7), e0271088. https://doi.org/10.1371/journal.pone.0271088 Ahmed, Z., Ahmad, M., Murshed, M., Shah, I. M., Mahmood, H., & Abbas, S. (2022). How do green technology investment, technological innovation, and trade globaliztion enhance green energy supply and stimulate environmental sustainability in the G7 countries? Gondawa Research, 112, 105-155. https://doi.org/10.1016/j.gr.2022.09.014 Aktharuzzaman, M., Benkraiem, R., Boubaker, S., & Zopounidis, C. (2022). COVID-19 crisis and risk spillovers to developing economies: Evidence from Africa. Journal of International Development, 34(4), 898-918. https://doiorg.ezproxy.ub.gu.se/10.1002/jid.3634 Autor, H. D. (2015). Why Are There Still So Many Jobs? The History and Future of Workplace Automation. The Journal of Economic Perspectives, 29(3), 3-30. http://www.jstor.org/stable/43550118 Baca, S. P., Garbe, B. L., & Weiss, R. A. (2000). The Rise of Sector Effects in Major Equity Markets. Financial Analysts Journal, 56(5), 34–40. https://doi.org/10.2469/faj.v56.n5.2388 Baruník, J., & Křehlík, T. (2018). Measuring the Frequency Dynamics of Financial Connectedness and Systemic Risk. Journal of Financial Econometrics, 16(2), 271-296. https://doi.org/10.1093/jjfinec/nby001 Berg, F., Julian F Kölbel, F. J., & Rigobon, R. (2022). Aggregate Confusion: The Divergence of ESG Ratings. Review of Finance, 26(6), 1315-1344. https://doi.org/10.1093/rof/rfac033 Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307-327. https://doi.org/10.1016/0304-4076(86)90063-1 39 Cavaglia, S., Brightman, C,. & Aked, M. (2000). The Increasing Importance of Industry Factors. Financial Analysts Journal, 56(5), 41-54. https://doi.org/10.2469/faj.v56.n5.2389 Chatterji, A. K., Durand, R., Levine, D.I., & Touboul, S. (2016). Do ratings of firms converge? Implications for managers, investors and strategy researchers. Strategic Management Journal, 37(8), 1597-1614. https://doi.org/10.1002/smj.2407 Chen, J., Bennett, A., & Zheng, T. (2006). Sector Effects in Developed vs. Emerging Markets. Financial Analysts Journal, 62(6), 40–51. https://doi.org/10.2469/faj.v62.n6.4352 Chen, Y., Wang, C., Miao, J., & Zhou, T. (2022). Identifying Risk Transmission in Carbon Market With Energy, Commodity and Financial Markets: Evidence From Time-Frequency and Extreme Risk Spillovers. Frontiers in Energy Research, 10, 922808. https://doi.org/10.3389/fenrg.2022.922808 Claessens, S., & Kose, A. (2013). Financial Crises Explanations, Types, and Implications. IMF Working Papers, 2013(28), 1-68. https://doi.org/10.5089/9781475561005.001 Delmas, M. A., & Burbano. V. C. (2011). The Drivers of Greenwashing. California Management Review, 54(1), 64-87. https://doi.org/10.1525/cmr.2011.54.1.64 Diebold, X. F., & Yilmaz, K. (2009). Measuring Financial Asset Return and Volatility Spillovers, with Application to Global Equity Markets. The Economic Journal, 119(534), 158-171. https://doi.org/10.1111/j.1468-0297.2008.02208.x Eckert, C. (2020). Risk and risk management of spillover effects: Evidence from the literature. Risk Management and Insurance Review, 23(1), 75-104. https://doi.org/10.1111/rmir.12139 Edmans, A. (2022). The end of ESG. Financial Management, 52(1), 3-17. https://doiorg.ezproxy.ub.gu.se/10.1111/fima.12413 Engle, F. R. (1982). Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation. Econometrica, 50(4), 987-1007. https://doi.org/10.2307/1912773 Engle, F. R., & Susmel, R. (1993). Common Volatility in International Equity Markets. Journal of Business & Economic Statistics, 11(2), 167-176. https://doi.org/10.2307/1391368 Englehardt, N., Ekkenga, J., & Posch, P. (2021). ESG Ratings and Stock Performance during COVID-19 Crisis. Sustainability, 13(13), 7133. https://doi.org/10.3390/su13137133 40 Fatemi, A., Glaum, M., & Kleiner, S. (2018). ESG performance and firm value: The moderating role of disclosure. Global Finance Journal, 38, 45-64. https://doi.org/10.1016/j.gfj.2017.03.001 Financial Times. The rise and rise of sustainable investment. Financial Times. https://www.ft.com/partnercontent/london-stock-exchange-group/the-rise-and-rise- ofsustainable-investment.html Fu, J. (2024). ESG stock returns and stock volatility: Evidence form Chinese listed companies. Les Ulis: EDP Sciences. https://doi.org/10.1051/shsconf/202418102002 Fu, J., & Qiao, H. (2022). The Time-Varying Connectedness Between China’s Crude Oil Futures and International Oil Markets: A Return and Volatility Spillover Analysis. Letters in Spatial and Resources Sciences, 15, 341-376. https://doi.org/10.1007/s12076-021-00288-z Gao, Y., Li, Y., Zhao, C., & Wang, Y. (2022). Risk spillover analysis across worldwide ESG stock markets: New evidence from the frequency-domain. The North American Journal of Economics and Finance, 59, 101619. https://doi.org/10.1016/j.najef.2021.101619 Garcia, S. A., Mendes-Da-Silva, W., & Orsato, R. J. (2017). Sensitive industries produce better ESG performance: evidence from emerging markets. Journal of Cleaner Production, 150, 135-147. https://doi.org/10.1016/j.jclepro.2017.02.180 Gavrilakis, N., & Floros, C. (2023). ESG performance, herding behavior and stock market returns: Evidence from Europe. Operational Research, 23(3), 1-21. https://doiorg.ezproxy.ub.gu.se/10.1007/s12351-023-00745-1 Global Sustainable Investment Alliance. (2021). Global Sustainable Investment Review 2020. https://www.gsi-alliance.org/wp-content/uploads/2021/08/GSIR-20201.pdf Hasan, M. B., Rashid, M. M., Hossain, M. N., Rahman. M. R., & Amin. M. R. (2023). Using green and ESG assets to achieve post-COVID-19 environmental sustainability. Fulbright Review of Economics and Policy, 3(1), 25-48. https://doi.org/10.1108/FREP-04-2022-0026 Hong, H., Torous, W., & Valkanov. (2007). Do insdustries lead stock markets? Journal of Financial Economics, 83(2), 367-396. https://doi.org/10.1016/j.jfineco.2005.09.010 Jiang, Y., Mu, J., Nie, H., & Wu, L. (2020). Time-frequency analysis of risk spillovers from oil to BRICS stock markets: A long memory Copula-CoVaR-MODWT method. International Journal of Finance & Economics, 27(3), 3386-3404. https://doiorg.ezproxy.ub.gu.se/10.1002/ijfe.2326 41 Karoui, A., & Nguyen, K. D. (2022). Systematic ESG exposure and stock returns: Evidence from the United States during the 1991-2019 period. Business Ethics, the Environment and & Responsibility, 31(3), 604-619. https://doi-org.ezproxy.ub.gu.se/10.1111/beer.12429 Key ESG. (2024 2 February). 50 ESG statistics you need to know in 2024. https://www.keyesg.com/article/50-esg-statistics-you-need-to-know-in-2024 Kilic, Y., Destek, A. M., Cevik, I. E., Bugan, F. M., Korkmaz, O., & Diboglu, S. (2022). Return and risk spillovers between the ESG global index and stock markets: Evidence from time and frequency analysis. Borsa Istanbul Review, 22(2), 141-156. https://doi.org/10.1016/j.bir.2022.11.015 Kumar, D. (2023). European travel and leisure sector and uncertainties: A risk spillover analysis. Tourism Economics, 29(1), 48-67. https://doiorg.ezproxy.ub.gu.se/10.1177/13548166211035954 Lanne, M., & Nyberg, H. (2016). Generalized Forecast Error Variance Decomposition for Linear and Nonlinear Multivariate Models. Oxford Bulletin of Economics and Statistics, 78(4), 595-603. https://doi.org/10.1111/obes.12125 Laopodis, T. N. (2016). Industry returns, market returns and economic fundamentals: Evidence for the United States. Economic Modelling, 53, 89-106. https://doi.org/10.1016/j.econmod.2015.11.007 Li, Y., Gong, M., Zhang, X-Y., & Koh, L. (2018). The impact of environmental, social, and governance disclosure on firm value: The role of CEO power. The British Accounting Review, 50(1), 60-75. https://doi.org/10.1016/j.bar.2017.09.007 Li, W., Li, W., Seppänen, V., & Koivumäki, T. (2022). Effects of greenwashing on financial performance: Moderation through local environmental regulation and media coverage. Business Strategy and the Environment, 32(1), 820-841. https://doiorg.ezproxy.ub.gu.se/10.1002/bse.3177 Li, L., Qin, K., & Wu, D. (2023). A Hybrid Approach for the Assessment of Risk Spillover to ESG Investment in Financial Networks. Sustainability, 15(7), 6123. https://doi.org/10.3390/su15076123 Liu, T., Nakajima, T., & Hamori, S. (2021). Which Factors Will Affect the ESG Index in the USA and Europe: Stock, Crude Oil, or Gold? In: ESG Investment in the Global Economy. 42 SpringerBriefs in Economics. Springer, Singapore. https://doi.org/10.1007/978-981-16- 29907_4 Lipton, M. & Lipton, W. Rosen & Katz. (2022). Stakeholder Capitalism and ESG as Tools for Sustainable Long-Term Value Creation. Harvard Law School Forum on Corporate Governance. https://corpgov.law.harvard.edu/2022/06/11/stakeholder-capitalism-and-esg- astools-for-sustainable-long-term-value-creation/ LSEG. (n.d). LSEG ESG SCORE. Collected 2024-05-12. https://www.lseg.com/en/dataanalytics/sustainable-finance/esg-scores Lütkepohl, H. (2005). New Introduction to Multiple Time Series Analysis. Springer. ISBN: 978-3-540-40172-8 Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77-91. https://doiorg.ezproxy.ub.gu.se/10.2307/2975974 Meng, S., & Chen, Y. (2023). Market Volatility Spillover, Network Diffusion, and Financial Systemic Risk Management: Financial Modeling and Empirical Study. Mathematics, 11(6), 1396. https://doi.org/10.3390/math11061396 Moosawi, S., & Segerhammar. (2022). How do ESG assets relate to the financial market? - A Diebold-Yilmaz spillover approach to sustainable finance [Master’s thesis, Linköping University]. https://www.diva-portal.org/smash/get/diva2:1722164/FULLTEXT01.pdf MSCI. 2023. Global Industry Classification Standard (GICS®). https://www.msci.com/documents/1296102/11185224/MSCI_GICS_Overview.pdf Noy, I., Doan, N., Ferrarini, B., & Park, D. (2020). Measuring the Economic Risk of COVID- 19. Global Policy, 11(4), 413-423. https://doi-org.ezproxy.ub.gu.se/10.1111/17585899.12851 Nvidia. (n.d.). Corporate responsibility. Retrieved 2024-04-10. https://www.nvidia.com/enus/csr/ Paypal. (n.d.). Global Impact. Retrieved 2024-04-10. https://about.pypl.com/values- inaction/reporting/global-impact-report/default.aspx PWC. (2023). Global Private Equity Responsible Investment Survey 2023: Generating upside from ESG: Opportunities for private equity. https://www.pwc.com/gx/en/sustainability/assets/pwc-private-equity-responsible- investmentsurvey-2023.pdf 43 PWC. (n.d.). Sustainability and climate change services. Retrieved 2024-04-10. https://www.pwc.com/gx/en/services/sustainability.html S&P Global. (2019, October 19). Understanding the “E” in ESG. https://www.spglobal.com/en/research-insights/articles/understanding-the-e-in-esg S&P Global. (2020, February 24). Understanding the “G” in ESG. https://www.spglobal.com/en/research-insights/articles/what-is-the-g-in-esg Shaikh, I., (2022). Environmental, Social, and Governance (ESG) Practice and Firm Performance: An International Evidence. Journal of Business Economics and Management, 23(2), 218-237. https://doi.org/10.3846/jbem.2022.16202 Testa, F., Boiral, O., & Iraldo, F. (2018). Internalization of Environmental Practices and Institutional Complexity: Can Stakeholders Pressures Encourage Greenwashing? Journal of Business Ethics, 147, 287-307. https://doi.org/10.1007/s10551-015-2960-2 Torelli, R., Balluchi, F., & Lazzini, A. (2020). Greenwashing and environmental communication: Effects on stakeholders' perceptions. Business Strategy and the Environment, 29(2), 407-421. https://doi.org/10.1002/bse.2373 Umar, Z., Kenourgios, D., & Papathanasiou, S. (2020). The static and dynamic connectedness of environmental, social, and governance investments: International evidence. Economic Modelling, 93, 112-124. https://doi.org/10.1016/j.econmod.2020.08.007 United Nations. (n.d.). Greenwashing – the deceptive tactics behind environmental claims. Retrieved 2024-04-10. https://www.un.org/en/climatechange/science/climateissues/greenwashing World Trade Organization. 2023. World Trade Statistical Review 2023. https://www.wto.org/english/res_e/booksp_e/wtsr_2023_e.pdf Zhao, L., He, W., Wang, A., & Zhu, F. (2023). Time and Frequency Spillovers between the Green Economy and Traditional Energy Markets. Systems, 11(3), 153. https://doi.org/10.3390/systems11030153 9. Appendix 44 Appendix 1A - Correlation Matrix Finan Com Utiliti Materi Real Indust Health IT Energ Consu Consu cial munic es al Estate rial Care y mer mer ation Disc Staple Servic s es Finan 1 cial Com 0.939 1 munic 7 ation Servic es Utiliti 0.927 0.871 1 es 5 4 Materi 0.831 0.826 0.836 1 al 5 1 3 Real 0.979 0.946 0.949 0.860 1 Estate 7 6 6 8 Indust 0.964 0.985 0.898 0.851 0.966 1 rial 7 4 1 8 2 Health 0.441 0.614 0.375 0.255 0.508 0.584 1 Care 4 5 7 0 0 5 IT 0.767 0.847 0.713 0.567 0.795 0.871 0.791 1 7 1 6 3 2 2 4 Energ 0.852 0.717 0.761 0.766 0.792 0.735 0.037 0.372 1 y 2 7 7 9 6 3 3 3 Consu 0.897 0.791 0.795 0.735 0.844 0.862 0.414 0.699 0.718 1 mer 7 3 6 4 6 9 6 3 5 Disc Consu 0.313 0.470 0.359 0.349 0.405 0.509 0.781 0.767 - 0.396 1 mer 4 3 5 0 2 5 3 6 0.146 7 Staple 2 s Appendix 2A – ANOVA Output SOURCE DF F PROB>F MODEL 14 0.54 0.8977 TO SECTOR 7 0.26 0.7847 FROM SECTOR 7 1.08 0.3927 RESIDUAL 49 TOTAL 63 45 Appendix 2B – Description of ANOVA Variables Description of Anova Variables To Sector The variable that is testing the significance of the receiving part From Sector The variable that is testing the significance of the sending part 46