!
"#"
!
!
!
!
!
Does it Pay to be Green?
An empirical analysis of the relation between environmental and
financial performance in the European electric utility industry
Olaf Gubler Master’s Thesis Msc. BA Finance
Faculty of Economics and Business University of Groningen
!
"$"
!
!
!
!
!
Does it Pay to be Green?
An empirical analysis of the relation between environmental and
financial performance in the European electric utility industry
Master’s Thesis Msc. BA Finance
Faculty of Economics and Business University of Groningen
Author: O.S. Gubler Student ID: 1322427
Contact details: olafgubler@gmail.com
Supervisor: Dr. L. Dam
Second Marker: Prof. dr. L.J.R. Scholtens Date: June 21, 2009
Keywords: Financial Performance, Environmental Performance, Electric Utilities, Panel Data JEL Classification: C33 (Panel Data), L94 (Electric Utilities), Q42 (Alternative Energy
Sources), Q50 (General Environmental Economics
!
"%"!
Abstract
There are contrasting theoretical views regarding the impact of environmental performance on financial performance. Some academics for example believe strong environmental performance can only come at a cost, whilst alternative theories predict a positive relation between environmental and financial performance. On top, prior empirical research has not been able to show one clear answer neither. In this thesis I first review the relevant literature, after which I empirically study the relationship between environmental and financial performance for the European electric utility industry. The results imply that there is limited evidence of a significant impact of either the percentage of renewable energy generated or the level of CO2 emissions on financial performance. This confirms the cost-benefit perspective
!
"&"!
Table of Contents
Abstract ... 3
List of Abbreviations ... 6
1. Introduction... 7
1.1 Background ... 7 1.2 Problem Statement ... 7 1.3 Purpose... 8 1.4 Outline ... 82. Theoretical Background ... 9
2.1 Traditional Perspective ... 9 2.2 Resource-Based Perspective ... 9 2.3 Cost-Benefit Perspective... 11 2.4 Market-Based Perspective ... 11 2.5 Discussion ... 123. Empirical Literature ... 13
3.1 Portfolio Studies ... 13 3.2 Event Studies ... 143.3 Cross-Section and Panel Data Studies ... 15
3.4 Meta-Analyses ... 18
3.5 Discussion ... 19
4. Research Question and Hypotheses ... 20
4.1 Research Question ... 20
4.2 Hypotheses... 20
5. Methodology... 22
5.1 Panel Data Analysis ... 22
5.2 Financial Performance ... 24
5.3 Environmental Performance ... 27
5.4 Control Variables ... 28
6. Data and Descriptive Statistics ... 31
6.1 Selection Criteria ... 31
6.2 Sample Selection... 31
6.3 Descriptive Statistic ... 32
!
"'"!
7. Results... 36
7.1 Analyses with Return on Assets ... 36
7.2 Analyses with Tobin’s q ... 38
8. Conclusion ... 40
8.1 Summary ... 40
8.2 Limitations and suggestions for future research ... 41
References... 43
Literature... 43
Internet ... 46
Appendices... 47
Appendix A: Overview of the regression studies ... 47
Appendix B: Correlations of ROA measures ... 49
Appendix C: Sources of renewable energy... 50
Appendix D: Overview of the dataset... 51
!
"("!
List of Abbreviations
BV Book Value
CO2 Carbon Dioxide
CSR Corporate Social Responsibility EDF Electricité de France
EIA Energy Information Administration
EU European Union
FRDC Franklin Research and Development Corporation g/kWh Gram per kilowatt hour
gWh Gigawatt hour = 1,000 megawatt hour = 1,000,000 kilowatt hour IEA International Energy Agency
IPCC Intergovernmental Panel on Climate Change IRRC Investor Responsibility Research Centre kWh Kilowatt hour
MV Market Value
MW Megawatt = 1000 Kilowatt OLS Ordinary Least Squares ROA Return On Assets ROE Return On Equity
!
")"!
1. Introduction
1.1 Problem Statement
As evidences of environmental challenges such as global warming and climate change become more prevailing it leaves little doubt that businesses all around the world shall be affected. Environmental risks such as emission reduction regulation and changes in consumer preferences continue to transform the traditional business landscape. As a result, many firms begin to include environmental issues and criteria into the decision-making process (Shrivastava, 1995). For that reason it is not surprising that the relation between environmental and financial performance has gained much attention in the academic literature. Even so, there are several contrasting theoretical perspectives. Some academics for example believe that strong environmental performance can only come at a cost, whilst alternative perspectives predict that environmental performance enhances financial performance. On top, the empirical literature can neither provide one clear answer. From this point of view it is interesting the re-examine whether it ‘pays to be green’.
1.2 Background
This thesis specifically examines the relationship between environmental and financial performance in the European electric utility industry. This industry is especially relevant to examine, as it is the most polluting sector in terms of carbon dioxide (CO2) emissions1.
According to the International Energy Agency (IEA) the electric utility industry accounted for 41% of the global CO2 emissions in 2004. The industry is such a heavy emitter, given that
mainly coal-fired power plants are used for generating electricity. The European Union (EU) has acknowledged the situation and has established a common energy policy in order to reduce the effects of climate change caused by power generation. This policy entails that by 2020 renewable energy should account for 20% of the energy generation in the EU. As a result the electric utility sector is expected to transform impressively, considering that in 2005 only 8.5% of the electricity generated could be labeled renewable. It can be observed that during the last couple of years European electric utilities have invested heavily in renewable energy generation. In addition, a number of renewable energy companies and renewable business units have been successfully listed on the stock market for the attraction of capital. For instance, Electricité de France (EDF) attracted !340 million with the listing of 22.5% of their renewable energy subsidiary, EDF Energies Nouvelles, in November 2006. According to
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 CO
2 is widely accepted as one of the main causes of global warming. See for example the 2007 Fourth
!
"*"!
the company’s press release, the listing has ‘generated extremely high interest’ among institutional investors. This can be perceived as a sign that financial markets put a value on environmental aspects in the European electric utility industry.
1.3 Purpose
All in all, the changing landscape in the European electric utility industry raises the question whether or not investing in renewable energy and CO2 reduction makes sense from a financial
perspective. It would be interesting to find a positive relationship between environmental behavior and financial performance, as it would legitimize strong environmental performance on solid economic grounds. This would allow firms to practice the good in order to contribute more broadly to the well-being of the environment and at the same time improve their financial performance (Margolis, Elfenbein and Walsh, 2007). Furthermore, a positive relation may have important implications for the function that firms ‘can be expected to play
in promoting pollution reduction efforts and the use of cleaner technology’ (Cohen, Fenn and
Konar, 1997). If on the other hand strong environmental performance leads to lower financial performance levels, then an environmentally friendly strategy is not necessarily in the company’s best interest from a financial perspective. In that case, there is need for stringent public regulation to protect the well-being of the natural environment, as companies have no financial incentive to do so.
1.4 Outline
!
"+"!
2. Theoretical Background
There are several theoretical arguments on whether and how environmental performance impacts financial performance. For example, the conventional view believes strong environmental performance can only come with a cost, whilst alternative theories predict a positive or neutral relation between environmental and financial performance. Below, I shall summarize the major lines of reasoning for the arguments made.
2.1 Traditional Perspective
For a long time the traditional perspective has been that achieving strong environmental performance can only come at the expense of financial performance (e.g. Palmer, Oates and Portney, 1995; Walley and Whitehead, 1994) This view is based on the idea that environmental expenditures, whether on waste treatment or pollution reduction, demand significant portions of a firm’s financial resources. For example, Walley and Whitehead (1994) state that responding to environmental challenges is a costly and complicated proposition for managers. Moreover, the authors stress that the number of cases of financial benefits is probably limited, because ‘environmental costs are skyrocketing with little
economic impact in sight’. From an investor’s perspective, the traditional perspective implies
that one can either invest in a profitable firm or in a ‘green’ firm that protects the environment, since no company is likely to be both. (Margolis et al., 2007)
In researching the generation costs for various energy resources for electricity generation, Sims, Rogner and Gregory (2003) find certain evidence supporting the traditional perspective. Although generation costs are very plant and site specific, the author finds that in general renewable energy resources have higher generation costs than fossil resources. This is in line with Hofman (2002), who claims that electric utility companies have long stayed away from the use of renewable energy because of high investment and generation costs.
2.2 Resource-Based Perspective
In contrast with the traditional perspective, the resource-based perspective implies that environmental performance is actually positively related to financial performance.According to Russo and Fouts (1997), the resource-based perspective of competitive advantage addresses ‘the fit between what a firm has the ability to do and what it has the opportunity to
do’. A firm’s resources can be classified as tangible, intangible and personnel-based, and
!
"#,"!
capabilities in attracting high-skilled human resources. One by one, these factors will improve the financial results of the firm in question. Accordingly, Russo and Fouts (1997) consider firms that go beyond compliance towards environmental policies to have better abilities to generate profits. Below their three arguments are explained in more detail.
(i) Tangible assets
Russo and Fouts (1997) indicate that firms with proactive environmental standards are able to use their tangible assets more efficiently as they would likely be involved in the adaption of new efficient technologies. This can be beneficial from a cost perspective if these new technologies outperform equivalent assets hold by competitors. Porter and van der Linde (1995) apply the same line of reasoning and point out that stringent regulation for pollution reduction provides future cost savings by increasing asset efficiency. According to the authors, environmental pollution is often a form of economic waste and can be related with unnecessary and inefficient use of resources. Stringent environmental regulation can stimulate innovation and efficiency, lower the total costs of a product and making the firm as a whole more competitive.
(ii) Intangible assets
!
"##"!
(iii) Human resources
As a final reason, Russo and Fouts (1997) state that a better corporate environmental image is expected to influence a company’s human resources policy, and accordingly its capability in attracting and retaining high quality human resources. This is verified by Dechant and Altman (1994) who show that organizations with ‘a poor environmental record will find it
increasingly difficult to recruit and retain high caliber staff’. In the end, firm financial
performance is likely to improve, as productivity improvements will be observable when employing high quality human resources.
2.3 Cost-Benefit Perspective
The traditional perspective and the resource-based perspective are harmonized in the cost-benefit framework. For example, McWilliams and Siegel (2001) state that for each firm there is an optimal level of Corporate Social Responsibility (CSR)2. This optimal level can be determined, as with any other investment, by applying a cost-benefit analysis. Accordingly, on the demand side, firms should take into account the benefits of CSR, e.g. premium pricing by product differentiation. Then on the supply side, firms should consider the costs of resources allocated to satisfying the demand for CSR. Palmer et al. (1995) share this view, arguing that the financial appeal of environmental regulation should be determined by a comparison of the costs and the benefits. The implication of this cost-benefit perspective is that environmental responsible firms incur higher costs, but these are offset by the fact that consumers are willing to pay higher prices (Elsayed and Paton, 2005). Thus, at least when adopting the optimal level of environmental performance, there can be expected a neutral relationship between environmental and financial performance.
2.4 Market-Based Perspective
Whilst the former three perspectives merely focus on a firm’s ability to generate accounting profits, one can also consider a market-based perspective. Lash and Wellington (2007) indicate that each business face numerous environmental risks, such as emission reduction legislation and environmentally concerned consumers. From that perspective, firms that perform well environmentally can be considered as a less risky investment compared to those with a poor environmental track record (Guenster, Derwall, Bauer, and Koedijk, 2005). As a result, in the risk-return framework, investors shall demand a lower expected return on the
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
$! CSR is perceived as a multidimensional construct in which environmental factors play a significant role
!
"#$"!
shares of environmental leaders. This in turn, implies that those environmental leaders have a relatively lower cost of capital vis-à-vis the environmental laggards.
This notion is shared with Ambec and Lanoie (2007), although the authors argue that the access to capital markets matters for the cost of capital. In their view, it is becoming evident that greener firm have an easier access to capital markets through the expansion of ‘green mutual funds’. According to the authors, these green mutual funds only invest in firms that meet certain environmental criteria. Furthermore, Ambec and Lanoie (2007) indicate that firms with strong environmental performance can borrow more easily from banks. Currently more than 60 international financial institutions have adopted the equator principles, committing themselves to financing projects that are ‘developed in a manner that is socially
responsible and reflect sound environmental management practices’.3 In the same line are the
carbon principles, committing financial institutions to ‘evaluate and address carbon risks in
the financing of electric power projects’.4 Subsequently, for the reason that environmental
leaders have easier access to capital markets, they face a lower cost of capital. This again is in line with the vision of Guenster et al. (2005).
In essence, the cost of capital is a key value driver in valuation, as it is the discount rate assigned to the expected future cash flows (Koller, Goedhart and Wessels, 2005). Thus the market-based perspective implies that environmental leaders, ceteris paribus, have relatively higher financial valuations as a result of a lower cost of capital.
2.5 Discussion
All in all, the existing theoretical literate is undecided as to the expected relation between environmental and financial performance. Arguments have been made for a positive, a negative and a neutral relation. This supports the need for empirical studies to decide between the different theoretical models.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 From the Preamble of the Equator Principles, www.equator-principles.com
!
"#%"!
3. Empirical Evidence
The theoretical debate regarding the relationship between environmental and financial performance has led to various empirical studies on the subject. The empirical literature can, in terms of methodology, be divided in three categories: portfolio studies, event studies and regression studies (Derwall, Guenster, Bauer and Koedijk, 2005). In addition, two meta-analysis studies integrate the findings of several individual studies on this topic. An overview of the empirical analyses per category is provided below.
3.1 Portfolio Studies
Portfolio analysis entails the comparison of the financial performance of two or more mutually exclusive portfolios. The most commonly used proxy for financial performance is a risk-adjusted share return, such as Jensen’s Alpha or the Sharpe ratio (Ambec and Lanoie, 2007). In order to construct the portfolios, a specific firm characteristic acts as a discriminating variable. Only few studies have applied firm environmental characteristics as this discriminating variable (Guenster et al. 2005). Amongst the few exemptions are Cohen et al. (1995) who constructed two industry-balanced portfolios distinct by eight environmental measures in order to evaluate the financial performance between the environmental laggards and leaders of the S&P 500. On the whole, the results suggest neither penalties nor many strong significant premiums for investing in the environmental leaders portfolio. Nonetheless, the authors expect any positive relation to become stronger in the future, as at the time of study global concern for the environment was relatively new and a fast changing issue.
More recent studies indeed propose that there are considerable benefits of including environmental criteria in the investment process. Derwall et al. (2005) for example constructed two portfolios based on eco-efficiency5 ratings provided by Innovest Strategic
Value Advisors. The results show that the more eco-efficient portfolio provided significantly higher risk-adjusted stock returns relative to its less eco-efficient peer over the period 1995-2003. In the same line, Blank and Daniel (2002) found significant higher Sharpe ratios for the eco-efficient portfolio than for the S&P 500 index during the 1997-2001 period. Similarly, Gluck and Becker (2004) found that eco-efficiency ratings have ‘significance in the stock
selection process’. Lastly, Filbeck and Gorman (2004) specifically examine the electric utility
industry. The authors constructed two portfolios of environmental leaders and laggards, utilizing compliance data from the Investor Responsibility Research Centre (IRRC). The
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 According to Derwall et al. (2005) corporate eco-efficiency can be interpreted as ‘the economic value a company
!
"#&"!
results, however, indicate that there are no significant relations between environmental performance and (risk-adjusted) share returns.
In summary, the portfolio studies indicate that there are no penalties for investing in an environmental leaders portfolio. However, the studies are undecided whether there is a neutral or positive relation between environmental performance and risk-adjusted share returns. Even so, portfolio analysis can only be used as informative indication, as it merely focuses on correlations and concern little about causation (King and Lenox, 2001).
3.2 Event Studies
Event studies can be used to examine whether a particular environmentally related event has an effect on a firm’s stock price (King and Lenox, 2001). The stock price’s reaction to an environmental event is obtained by comparing the expected and actual returns within a narrow time frame around the event date. If there is a significant difference between those two returns, it can be concluded that the event had a significant impact on the stock price (Ambec and Lanoie, 2007). Lorraine, Collison and Power (2004) believe that it is difficult to draw general conclusions from the event study literature, since there is much variation in the research methodology. They indicate that there have been large differences in for example, data sources used, window length, industry types studied, firm size examined and in calculating expected share returns. Despite these differences in methodology, event studies typically show that environmentally related events, whether positive or negative, imposes a significant reaction on the stock market.6
The majority of the event studies make use of firm specific environmental news, such as information about illegal spills, pollution fines, emission data or environmental awards. For instance, Klassen and McLaughin (1996) found evidence of significant negative abnormal returns in response to negative news on oil and chemical spills. Furthermore, the authors found that receiving environmental awards is associated with positive abnormal returns. Similarly, Hamilton (1995) reported that news of high levels of toxic emissions, proclaimed by the Toxic Release Inventory (TRI), resulted in significant negative abnormal returns upon the first release of the information. In addition, Lorraine et al. (2004) studied 32 events of pollution fines and environmental awards over the period 1995-2000. The results show significant abnormal returns for pollution fines, at one week after the news is published.
In contrast with firm specific environmental events, some authors choose to consider the intra-industry effect of a major environmental disaster. Blacconiere and Patten (1994), for example, study the market reaction of 47 chemical firms following the Bhopal disaster in
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 For a comprehensive overview of environmentally related event-studies, see Lorraine et al. (2004) or Ambec and
!
"#'"!
December 19847. On the whole, the results denote that there was ‘a significant, negative industry market reaction within the ten days following the Bhopal chemical leak’. Likewise,
Hill and Schneeweis (1983) show significant negative intra-industry market reactions for electric utilities after the Three Mile Island nuclear accident in 19798. The impact was less severe for non-nuclear firms than that for nuclear-based utilities.
Despite these consistent results, the use of event study methodology has been criticized in the literature for its limitations. To begin, Konar and Cohen (2001) indicate that event studies cannot examine long-term trends or environmental performance measures that are not fixed to a specific date. Additionally, as pointed out by Dam (2008), the significance of the environmental event impact depends on whether or not the event is actually providing new information to investors. Most importantly, the findingf short-term price reaction does not offer compelling proof of a long-term systematic effect of environmental performance on financial performance. (Koehler, 2006)
3.4 Cross-Section and Panel Data Studies
Cross-section and panel data regression analysis can be employed in order to evaluate the long-term systematic effect of environmental performance on financial performance. The method attempts to explain movements in financial performance relative to movements in environmental performance and other control factors. (Brooks, 2002) First of al, it is difficult to generalize results from individual regression studies, as there is much difference in the methodology employed. This is emphasized in Table A-1 in Appendix A, which summarizes the methodology and findings of this section of this part of the empirical literature. As can be seen from the table, there is vast variation in the measurement of environmental performance. That is, a number of studies consider subjective rankings, whilst others examine objective emission data. (Cohen et al., 1997).
Subjective measures of environmental performance, such as environmental scores, ratings or rankings, are often used in the empirical literature. Early research in this area is mostly based on pollution control records provided by the Council of Economic Priorities (CEP) in the 1970’s. For example, Chen and Metcalf (1980) examined the pollution control ratings for a sample of 18 firms in the paper and pulp industry over the period 1968-1973. The regression results show that there are no significant relations between the environmental
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7 At 3 December 1984, a Union Carbide pesticide plant in Bhopal, India, leaked 42 tonnes of toxic chemicals,
resulting in approximately 4,000 deaths and 200,000 injuries (Bacconiere and Patten, 2004)
8 The accident at the Three Mile Island nuclear power plant near Middletown, Pennsylvania on March 28, 1979,
!
"#("!
ratings and various measures of financial performance. Even so, the authors stress that the results do not imply that investors are indifferent regarding environmental factors. Rather they are only concerned when pollution control is below satisfactory standards, as that might endanger firm earnings. Nonetheless, the relevance of the early studies has often been criticized for the reason that at that time there was little concern for corporate environmental responsibility. (Elsayed and Paton, 2005; Konar and Cohen, 2001)
A more recent research by Russo and Fouts (1997) makes use of environmental ratings by the Franklin Research and Development Corporation (FRDC). With an analysis of 234 firms over two years, the authors examined the relation between these ratings and return on assets (ROA). The results show that two variables are positively related and that industry growth is found to strengthen this relationship. Guenster et al. (2005) found similar results and conclude that there is a positive, yet non-linear relationship between eco-efficiency and financial performance. More specifically, stronger environmental performance does not necessarily translate in higher valuations and ROA, but rather weaker environmental performance is related to lower market valuations and profitability.
In contrast, Elsayed and Paton (2005), using static and dynamic panel data techniques on a dataset of 227 British listed companies for the years 1994 to 2000, find no significant relation between environmental ratings and financial performance. Their environmental ratings are based on the community and environmental responsibility scores in Management’s Today survey of Britain most admired companies. The authors employ ROA, return on sales, and Tobin’s q to proxy for financial performance. In order to compare the results with Russo and Fouts (1997), Elsayed and Paton (2005) also applied the between effects and the pooled model on their dataset. The estimates of these models imply that environmental performance has a significant positive impact on financial performance. Elsayed and Paton (2005) explain this difference to unobservable firm effects that are important in explaining financial performance. For that reason, ignoring firm heterogeneity may lead to incorrect inferences and more sophisticated panel data techniques are more appropriate. In a similar research, Salama (2005) also use Management’s Today community and environmental scores to proxy for environmental performance. The author criticizes the statistical properties of the conventional Ordinary Least Squares (OLS) regression technique in the presence of outliers and firm heterogeneity. For that reason, he prefers using of median regression techniques. The OLS regression results show that environmental ratings have a neutral impact on financial performance, whilst the median regression estimates indicate a positive relation between the variables.
!
"#)"!
examining a sample of 321 manufacturing firms from the S&P 500, the authors find that the amount of toxic emissions per dollar revenue and the number of environmental lawsuits have a significant negative effect on the value of a firm’s intangible assets and on Tobin’s q. Specifically, the results show that the sample has an average environmental liability of approximately $380 million in market value. In a comparable research, Hart and Ahuja (1996) study the relationship between environmental and firm performance for 127 manufacturing firms drawn from S&P 500. The authors measure environmental performance as the percentage change of the emissions by sales ratio from 1988 to 1989. The results imply that reductions in emissions have a significant positive effect on accounting-based measures of financial performance. This effect is most pronounced in two years following the emission reduction. King and Lenox (2001) criticize the work of Hart and Ahuja (1996) for the reason that the results are difficult to interpret. According to the authors, it is not clear whether it pays to be green or to operate in clean industries. In their analysis, King and Lenox (2001) employ a similar environmental performance measurement based on emissions of toxic chemicals. Additionally, in order to capture the effect of differences in environmental management within industries, the authors include emissions relative to a firm’s industry. By examining 625 manufacturing firms over the time period 1987 to 1996, the results indicate a significant negative association between pollution levels and values of Tobin’s q in the following year. Interesting is that the results show that firms with lower pollution levels relative to their industry tend to have significantly higher financial performance. On the other hand, operating in cleaner industries by itself does not result in higher valuations. For that reason, the authors conclude that there is a significant industry effect that must be controlled for. Expanding on their previous research, King and Lenox (2002) distinguish the effects of several methods for pollution reduction. The conclusions from their work suggest that firms may mainly improve their financial performance by engaging more in waste prevention. On the other hand, simple end-of-pipe pollution reduction does not influence Tobin’s q.
!
"#*"!
recycling. The results of the research indicate that there is a significant, negative relationship between environmental pro-activism and financial performance, which implies that security analysts seem to foresee lower earnings in the short term for pro-active firms. This may be an indication, that in the short term the costs of waste treatment and recycling outweigh the longer-term financial benefits.
On the whole, these regression studies show contradictory results regarding the relationship between environmental and financial performance. There may be several reasons why the results may be conflicting. Most importantly, there is much variety in the methodology used in the studies. This is especially the case for the environmental performance measure, as there is no single best reliable definition of environmental performance (Elsayed and Paton, 2005; Salama 2005) The objective measures of environmental performance show in general that there is a positive relation, whilst the subjective measures may also have a neutral impact on financial performance. The results of the studies may also be dependent on the specific dataset used. Schaltegger and Synnestvedt (2002) indicate that any relation may amongst others fluctuate depending on the type of industry, time span, consumer behaviour, and regulatory regime.
Nevertheless, there are some main conclusions that can be drawn from this part of the literature. Firstly, as the results of King and Lenox (2001) indicate there is a significant industry effect that must be controlled for in future research. Secondly, as Elsayed and Paton (2005) found, firm heterogeneity is especially relevant in these part of the empirical literature. For that reason, it may be better to employ specific panel data regression techniques to provide a more powerful evidence base. Thirdly, the results of King and Lenox (2002) and Dowell et al. (2000) imply that proactive environmental performance has more significance than end-of-pipe pollution reduction or solely adopting to less stringent regulation. Nevertheless, this effect may only hold in the long-term (Cordeiro and Sarkis, 1997), where the financial benefits of environmental performance can be experienced.
3.5 Meta-Analysis
!
"#+"!
interest. The second meta-analysis by Margolis et al. (2007), consisting of 44 studies, also shows a positive yet small effect size (r = 0.11). The authors indicate that the effect is somewhat larger for subjective self-reported (r=0.19) than for objective environmental measures (r=0.09).
3.5 Discussion
Overall, the empirical literature shows inconsistent results regarding the relation between environmental and financial performance. Nonetheless, the majority of the studies seem to support the hypothesis that strong environmental performance is related with better financial performance, or at least not worse performance (Salama, 2005; Ambec and Lanoie, 2007). This is in contrast with the conventional view that environmental performance can only come with a cost.
!
"$,"!
4. Research Question and Hypotheses
4.1 Research Question
This thesis deals with the question whether firm financial performance is influenced by environmental performance in the European electric utility industry. The subsequent research question is the following:
‘What is the impact of environmental performance on financial performance in the European electric utility industry?’
The choice to study one specific industry is based on the conclusions of Lars and Wellington (2007). They argue that the most significant distinctions when regarding environmental performance are not between sectors but within sectors, where a firm’s environmental strategy can create competitive advantages. Furthermore, Hart and Ahuja (1996) indicate that results may be more significant for specific industries where emissions and effluents are more relevant. For that reason, the electric utility industry is especially relevant to examine, as by their very nature as producers of energy, electric utilities produce substantial amounts of pollution (Filbeck and Gorman, 2004).
4.2 Hypotheses
Based on the theoretical and empirical literature, the following null and alternative hypotheses can be formulated. Firstly, the null hypothesis is based on the cost-benefit perspective that the benefits of environmental investments equal the costs. For that reason, a neutral relationship is expected between environmental and financial performance. In the empirical literature the results Cohen et al. (1995), Filbeck and Gorman (2004), and Elsayed and Paton (2005) confirm the null hypothesis.
H0: Environmental performance has no impact on financial performance
Secondly, the first alternative hypothesis finds its ground in the conventional perspective. For the reason that environmental performance can only come at a cost, the variable has a negative impact on financial performance. Only Cordeiro and Sarkis (1997) find evidence supporting the conventional view.
!
"$#"!
Finally, both the resource-based and the market-based perspective provide theoretical foundations for a positive relation between environmental and financial performance. The most pronounced evidence supporting these views can be found in the event-studies. But also the majority of the regression studies show support for this second alternative hypothesis.
!
"$$"!
5. Methodology
In order to answer the research question, panel data regression techniques will be applied. This section starts with an overview of the basics behind panel data analysis and the most commonly applied models for panel data analysis. The latter part of this section shall explicate how the variables in the models are calculated.
5.1 Panel Data Analysis
A great deal of preceding studies (e.g. Harth and Ahuja, 1996; Russo and Fouts, 1997) has relied on cross-section or pooled datasets. As indicated by Elsayed and Paton (2005), the disadvantage of using such datasets is that conclusions based on cross-sectional analysis are likely to be invalid in the presence of significant firm heterogeneity. The use of panel data allows controlling for firm specific effects, and thus offers a more prevailing evidence base. Furthermore, as pointed out by Baltagi (1995), panel data provides more informative data so that one can produce more reliable parameter estimates. For these reasons, panel data regression techniques will be applied to answer the research question.
In panel data there are basically two types of information. First, there is cross-sectional information reflected in the differences between firms, and secondly, there is time-series information reflected in the changes within firms over time. (Baltagi 1995) Particular panel data regression techniques, such as the ‘fixed effects’ and ‘random effects’ models, allow one to take advantage of both types of information. In contrast, it is also possible to use ordinary multiple regression techniques on panel data, i.e., the ‘pooled regression’ and the ‘between
effects’ model. These models do not take into account the two information types and may
therefore not be optimal. Also the estimates of the coefficients derived from these regressions may be subject to the omitted variable bias. This problem arises when there is some unknown variable that cannot be controlled for that affects the dependent variable. By applying the fixed effects and random effects models, it is possible to control for some types of omitted variables even without observing them, by examining changes in the dependent variable over time. This controls for omitted variables that differ between cases but are constant over time. In addition, it is also possible to use these panel data techniques to control for omitted variables that vary over time but are constant between cases. The four different models, i.e.,
!
"$%"!
(i) Pooled model
The pooled model for panel data is nothing more than a standard regression model in which all observations are treated as though they came from a single regression model. (Koop 2008) The pooled model has constant coefficients, as it does not control for firm or time specific effects. The general equation for the pooled model looks the following:
(1)
!
Yit="+ #Xit+$it
Where Yit and Xit denote the observations of respectively the dependent and independent
variable for firm i at time t. Additionally, ! and " are the intercept and slope coefficient in that order. Both remain equal for all firms across time. In many applications, though, the pooled model may not be appropriate as it is often the case that different firms will have different regression lines. Ignoring this fact can lead to biased estimates of both the intercept and coefficients (Koop, 2008).
(ii) Between effects model
The between effects model is the second method that employs standard multiple regression techniques. Performing a between effects regression is equivalent to a cross-section regression using the means of each variable over the sample period. (Elsayed and Paton, 2005) Unlike the pooled model, the between-effects model controls for omitted variables that change over time but are constant between cases (Website: Princeton University). The disadvantage of using the between effects is that all the time-series information is dropped. For that reason, its use is not widespread in the empirical literature and the between effects model will not be applied in this research.
(iii) The fixed effects model
In contrast with the previous models, the fixed effect model controls for omitted variables that differ across individuals but are constant over time. The model allows the firms to have regression lines with the same slope coefficient but each firm having an individual intercepts. (Koops, 2008) In order to model this individual effect, a dummy variable for each firm is included in to the regression equation. The fixed-effects model thus has the following regression equation:
(2)
!
Yit="1Dit(1)+"2Dit(2 )+ ... +"NDit(N)+#Xit+$it
!
"$&"!
variable for firm i at time t. Additionally, " is the slope coefficient of the dependent variable and denotes the dummy variable for firm 1. In contrast with the pooled model, the intercept ai is firm specific. The drawback of using the fixed effects model is that due to the
inclusion of dummy variables for N firms, the number of dependent variables increases with
N as well. This will lower the degrees of freedom, causing the standard errors to increase and
reduces the efficiency of the model. Also, as pointed out by Dowel et al. (2000), the fixed effects model can only be used if there is considerable variation in the independent variables.
(iv) The random effects model
The random effects model is closely related to the fixed effects model. This model also controls for omitted variables that differ across individual but are constant over time. However, unlike the fixed effects model, the random effects model does not use dummy variables, but assumes that the individual effect is a random variable. (Koop, 2008). The random effects model can be written as:
(3)
!
Yit="+ #Xit+$it with
!
"it= vi+ uit
Where Yit and Xit stand for the observations of respectively the dependent and independent
variable for firm i at time t. Furthermore, the intercept a and the slope coefficient " are constant over time and across firms. To this point the model looks similar to the pooled model. However, in the random effect model the error term #it consist of an additional random
variable vi, which refers to the time-invariant unobservable individual effect. (Baltagi, 2005)
The remainder disturbance term uit varies over time and across firms. Koop (2008) indicates
that the random effects model leads to more efficient and accurate estimations, as there are fewer coefficients to estimate than in the fixed effect model. Nonetheless, the random effects model may yield biased estimates when the regression errors are correlated with the independent variables. Thus, there is a trade-off between consistency and efficiency between the fixed effects and random effects models. A Hausman test can be employed in order to determine whether the fixed or random effects model is appropriate.
5.2 Financial Performance
!
"$'"!
decision-making capabilities and managerial performance’ because they depend on the
distribution of resources to projects and policy choices. The second category, market-based measures, comprises of, for example, (risk-adjusted) stock returns and market-to-book value ratios such as Tobin’s q. The use of market-based measures in studies ‘reflect the notion that
shareholders are the primary stakeholder group’ (Orlitzky et al. 2003) and follows the
rationale of Jensen (2000) that firms should have to objective to maximize value instead of profits.
As indicated by Elsayed and Paton (2005), there is extensive literature on the appropriate measurement of financial performance. Nonetheless, there has not been much agreement on the best method to exercise. For instance, Hart and Ahuja (1997) and Russo and Fouts (1997) only consider accounting based measures of financial performance in their studies, whilst on the other hand Konar and Cohen (2001) and Dowell et al. (2000) only employ market-based measures. Cordeiro and Sarkis (1997) indicate that accounting-based measures may be ‘deficient as they reflect only past performance, and are thus unsuited to
fully assessing firms’ strategic outcomes and performance’. Then again, market-based
measures may be noisy, as they are ‘influenced by market or economy wide forces that are
well beyond management forces’. Also, information asymmetry between managers and
shareholders may imply that the share price can only reflect information available to the shareholders, and do not necessarily represent true value (Cordeiro and Sarkis, 1997). Martin (1993) harmonizes both views and proposes that accounting- and market-based measures should be considered as ‘complements rather than substitutes’, since both financial performance measures contain information about market power. Therefore, the author believes there is no rationale that ‘either type of measurement dominates the other’.
In view of the fact that no consensus has achieved on the proper financial performance measurement and that they should rather be regarded as complements, this study shall consider both alternative measures of financial performance. In particular, (i) return on assets and (ii) Tobin’s q will be employed.9 Both are commonly used and generally accepted
measurements of firm performance. This approach is not uncommon: several empirical studies have also utilized these two financial performance measures in examining the environmental-financial performance link. (See for example, Guenster et al., 2005, Elsayed and Paton, 2005, and King and Lenox, 2002) Below, the two financial performance measures are discussed in more detail.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
9 The choice not to include (risk-adjusted) stock returns in this study is made in order to compare the results with
!
"$("!
(i) Return on assets
Return on assets is an accounting-based measure of financial performance that measures the efficiency of which a firm exploits it assets in order to generate earnings. (Cohen et al., 1997) The higher the value the better a firm operates its assets and hence the better it performs financially. According to King and Lenox (2002) and Russel and Fouts (1997) return on assets is employed commonly in finance and strategy research. Even though its popularity, return on assets has been calculated by various methods in the empirical literature. For example, King and Lenox (2002) use profit before interest as a proxy for return, whilst Elsayed and Paton (2005) employ profit before taxes. Nonetheless, all methods use the same basic information from the income statement to calculate the returns on assets. For that reason, the results of the study will not severely be impacted by the choice of method how to calculate ROA.10 This study follows and uses reported profit before interest and taxes divided
by total assets as a proxy for ROA. This choice is made following the rationale of Cohen et al. (1997) that it is the operating income regardless of capital structure or taxes that matters.
(ii) Tobin’s q
Tobin’s q is a market-based measure of financial performance and can be defined as the ratio of the market value of a firm to the replacement costs of its assets. (Lindenberg and Ross, 1981; Chung and Pruitt, 1994; and Dowell et al., 2000) The ratio has a value of unity in an equilibrium situation. If the Tobin’s q ratio is greater than this, then investment in stimulated. (Elsayed and Paton, 2005). On the other hand, a value of Tobin’s q below one indicates that the costs to replace a firm’s assets is greater than the value of its equity, and then the firm is undervalued in the market. In this thesis, the simplified method of calculating Tobin’s q as proposed by Chung and Pruitt (1994) shall be applied. The authors created their method since the ‘highly complex and cumbersome’ procedure by Lindenberg and Ross (1981) had not been used in practice often. In contrast with Lindenberg and Ross (1981), the simple formula for approximating Tobin’s q requires only basic financial and accounting information. The formula for the simple approximation of Tobin’s q is as follows:
(4)
!
Tobin 's q =MW (E ) + BV (PS ) + BV (LTD) + BV (CL) " BV (CA)
BV (TA )
Where MV(E) denotes the market value of equity, BV book value, PS the preferred stock, LTD the long-term debt, CL the current liabilities, CA the current assets and TA the total
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
10 As can be seen from Table B-1 in Appendix B, the several methods of approximating ROA are highly
!
"$)"!
assets.Using this simplified method of estimating Tobin’s q, Chung and Pruitt (1994) found little qualitative difference with the complex method by Lindenberg and Ross (1981). According to their results at least 96.6% of the variability of Tobin’s q is explained by their simple approximation of q. For that reason, the simplified Tobin’s q is a generally accepted method in the empirical literature. For instance, Elsayed and Paton (2005), and King and Lenox (2000) have employed this financial measure to test the relation between environmental and financial.
5.3 Environmental Performance
Environmental performance is the focal independent variable in this study. As indicated by Cohen et al. (1997), there is lack of objective measures to assess environmental performance. A great deal of studies relied on subjective ratings and rankings by public interest groups. (Guenster et al., 2005; Russo and Fouts, 1997; Elsayed and Paton, 2005). Methodological differences in measuring the various ratings make it difficult to understand the data and make objective comparisons (Ilinitch, Soderstrom and Thomas, 1998). Besides subjectivity, another problem with these rankings and ratings is that they can only serve as a proxy of the underlying environmental performance (Elsayed and Paton, 2005). This thesis overcomes these problems by focusing on two objective measures of environmental performance. The first shall be the use of renewable energy sources in electricity generation and the second deals with the emissions of carbon dioxide per kilowatt hour (kWh) of electricity generated. Both indicators can be perceived as reliable, quantifiable and objective measures of environmental performance.
(i) Renewable Energy
The Energy Information Administration (EIA)11 defines renewable energy as electricity that is
generated by sources that are ‘naturally replenishing but flow limited’. In other words, it means that these sources are virtually unlimited in duration, but are limited in the amount of energy that is available per unit of time. The most widely employed renewable energy sources in Europe in 2006 are hydraulic power (63.4%), wind power (18.1%), biomass (16.8%), and solar power (0.5%).12 Appendix C describes these renewable energy sources in more detail.
Renewable energy sources have in common that their use to generate electricity causes no
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
11 The EIA is the institution that provides the official energy statistics from the U.S. Government.
!
"$*"!
emission of environmental hazardous chemicals in the air.13 For that reason, it can thus easily
be hypothesized that electric utilities with a more renewable generation mix have stronger environmental performance. The renewable energy variable that will be used in this thesis is defined as the total electricity generated by renewable sources over total electrical output generated.
(5)
!
Renewable Energy (%) =Electricity Generated by Renewable Sources (GWh )
Total Electricity Generated (GWh ) * 100%
Where renewable energy sources are defined as wind, solar, biomass and hydraulic power. The higher the percentage renewable energy is, the stronger the environmental performance of the company. The data on the electricity generated by source has been collected by consulting companies’ websites, annual reports, CSR reports and environmental reports.
(ii) Carbon dioxide emissions
Another quantifiable proxy for environmental performance is the amount of waste a firm generates. Since CO2 emissions are widely recognized as the cause of global warming and
climate change, firm specific CO2 emissions can be a good proxy for environmental
performance for its direct impact on the environment. Specifically, in order to control for size, this thesis shall focus on the level of CO2 in kilograms (kg) that is emitted per kWh of
electricity generated. For a number of firms the variable is stated directly in annual, CRS or environmental reports. If this was not the case, the variable had to be computed manually according to the following formula:
(6)
!
CO2 Emissions (kg /kWh ) = Total CO2 Emissions (t)
Total Electricity Generated (mWh )
The higher the variable is, the more the firm pollutes relative to its size. As renewable energy sources do not emit carbon dioxide, this variable will have the value of zero for pure renewable energy companies.
5.4 Control variables
A number of variables will be included in the analysis in order to control for potential influences on environmental and financial performance. Based on a review of the relevant
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
13 Although burning biomass causes CO
2 emissions, it is generally considered CO2 neutral since biomass can be
!
"$+"!
environmental literature, it has been decided that the following control variables could be important determinants of financial performance: (i) firm size, (ii) leverage and (iii) growth. Although McWilliams and Siegel (2000) showed the importance of research and development (R&D) expenditures in CSR research, this variable has not been included as the data was not available. Nonetheless, the omitted R&D variable bias may be less likely in this research, since companies that invest in renewable energy can be perceived as more innovative. The last and more sector specific variable is (iv) a dummy for firms that perform multiple utility activities. These firms are called multi-utilities. Below I shall summarize briefly the set of control variables used.
(i) Firm size
As indicated by Margolis et al. (2007) firm size is a worthwhile control variable for the reason that larger firms may have more resources for environmental investments. A greater base of resources may create economies of scale inherent to those environmental investments (Elsayed and Paton, 2005). In addition, larger firms may attract greater pressure to engage in corporate social behavior. There are several industry specific methods to measure firm size. One can take for instance input factors such as the installed capacity to estimate firm size. But one can also look at output factors such as total sales or total electricity generated. Most common is on the other hand to measure firm size as the total book value of assets in a given year. According to Hall and Weis (1967) such a measurement is superior to any input or output measurement, because ‘it is the difficulty of financing large amounts of assets that
limits entry to certain fields’. I follow the majority in the literature (e.g. King and Lenox,
2001; Dowell et al., 2000; and Elsayed and Paton, 2005) and use the natural logarithm of the year-end book value of total assets as a proxy for firm size. The natural logarithm is employed, as the distribution of firm assets is unlikely to be normally distributed (Elsayed and Paton, 2005).
(ii) Leverage
!
"%,"!
Ahuja (1996).14(iii) Growth
Growth is the third factor to control for in this study. As pointed out by Konar and Cohen (2001) recent revenue growth is found to be positively related with accounting-based profitability ratios. Furthermore, Koller et al. (2005) and Guenster et al. (2005) share the opinion that growth is one of the key value drivers in the stock market. For these reasons, recent growth in revenues is included in the empirical analysis. Following King and Lenox (2001), growth is measured as the annual percentage change in revenues.
(iv) Multi-utility
A dummy is included in order to control for other firm utility activities that may affect firm financial performance. A multi-utility firm is one that combines electricity generation with other utility services such as waste management, or the supply of gas and water. This dummy may be relevant when other utility services have better or worse operating margins or are valued differently in the stock market.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
#&!-./!greater part of the empirical literature, however, measures leverage as the ratio of debt over assets (See for
!
"%#"!
6. Data and Descriptive Statistics
This section begins by explaining in what manner the data of the research has been collected. Then the sample and the descriptive statistics of the variables will be discussed. The final part of this section will show how the variables in the dataset are correlated with each other.
6.1 Selection Criteria
For the reason that there was no existing dataset available on the European electric utility sector, it had to be constructed by hand. In order to compose a decent dataset, four selection criteria were formulated. To begin, the generation of electricity had to be the focal activity for each firm included in the dataset. Secondly, the firm’s core generating facilities had to be based within Europe. Thirdly, given that market-based measures of financial performance will be used, the firm had to be publicly listed. Finally, in order to employ panel data regression techniques, at least two consecutive years of company financial and environmental data had to be available over the period 2001 to 2007.
6.2 Sample Selection
The starting point for the selection of the firms has been the Amadeus database. A search query within Amadeus on publicly listed and electricity-generating companies in Europe provided 65 companies. From these companies a total of 31 have been eliminated, as they did not fit the above mentioned selection criteria. Firstly, twelve firms were dropped, as they are primarily focused on the management of the electricity grid instead of generating electricity. Another six companies were removed because they are suppliers of the renewable energy sector and do not generate electricity themselves.15 After that, five firms have been eliminated
from the dataset as their operations are primarily outside Europe. Finally, eight companies have been emitted, as the required environmental data was not available. The Amadeus search thus resulted in only 34 relevant companies that fitted the four criteria mentioned above.
Above all, the Amadeus dataset appeared to be highly incomplete, as some of Europe’s largest electricity firms were excluded. For instance, it did neither contain Electricité de France (EDF) nor the German utility E.ON. Accordingly, another search query had to be conducted; this time using Internet resources. The websites of all the national Stock Exchanges, the Google Finance website, and the energy portal website of the European Union had been consulted. The first two sources have been used in order to detect missing listed energy companies, whilst the third source provides data on market sizes and market shares of
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
"%$"!
the energy sector for all European countries. Each country has been examined to verify whether no important companies had been forgotten. All in all, this time-consuming procedure increased the dataset by another 25 companies that fitted the four criteria.
In the end, the total dataset comprises of an unbalanced panel of 318 annual observations for 59 publicly listed European electricity generating companies. A detailed overview of the dataset can be found in Table D-1
in Appendix D. Form that table, it can be derived that the maximum of seven yearly observations (2001 to 2007) is found for a total 28 companies. On the other hand for nine companies only two yearly observations could be determined. For seven of these companies the reason behind the limited number of observations is that they have only been recently introduced to the public stock market. The other two companies lacked historical environmental data.
A geographical overview of the companies is shown in Table II. The first thing to notice from the table is that the United Kingdom, Germany and Switzerland are most often represented in the dataset. Another point to note is that not all European countries are represented in the dataset. The underlying reason is that the energy sector has not been fully liberalized in many European countries. Government owned electric utility companies can still be found in, for example,
Sweden (Vattenfall), Denmark (Dong), Netherlands (Nuon and Essent), Ireland (ESB) and in a number of Eastern-European countries.
6.3 Descriptive statistics
All the financial data of the companies has been collected from Thomson’s DataStream. The environmental data on the other hand has been collected manually by consulting annual, CSR and environmental reports. The descriptive statistics for each of the variables are shown in Table III. Please note that the number of total observations differs for some of the variables. This is primarily the result of missing data on environmental performance. In addition, sales
Table II
Number of companies by country
!
"%%"!
growth could only be estimated when the preceding year’s revenue was available.
The descriptive statistics show that the mean (median) ROA is 6.3% (6.2%) annually. There is also quite some variation for this variable, as can be seen from the extremes and the standard deviation. Novera Energy experienced the lowest ROA of -14.9% in 2006, whilst Arendals Fossekompani showed the highest ROA of 39.1% in 2003. The values for Tobin’s q show similar results: the mean (median) value of Tobin’s q is 0.812 (0.755), which indicates that on average the European electric utility companies are values below the replacement value of their assets. Greentech Energy Systems was relatively valued highest in the market with a Tobin’s q of 5.770 in 2005, whilst on the other hand Plambeck sees the lowest Tobin’s
q of -0.320 in 2002. This negative value origins because for this particular observation the net
current assets outweighed the sum of the market value of equity and the book value of long-term debt.
Table III – Descriptive statistics of the variables for the whole sample
Variable Obs. Mean Median Std. Dev. Minimum Maximum
Financial Performance Return on Assets 318 0.063 0.062 0.062 -0.149 0.391 Tobin’s q 313 0.812 0.755 0.591 -0.320 5.770 Environmental Performance Renewable Energy (%) 298 0.550 0.385 0.426 0.000 1.000 CO2 Emissions (kg/kWh) 274 0.234 0.051 0.287 0.000 1.005 Control Variables Size (million !) 318 15,120 2,407 30,340 7 184,540
Size (LN Total Assets) 318 14.39 14.69 2.55 8.86 19.03
Sales Growth (%) 302 0.496 0.088 2.646 -0.880 40.824
Leverage 318 1.094 0.853 1.683 -19.094 8.863
Note: (i) All variables are based on annual data; (ii) for variable definition see Table E-1 in Appendix E; (iii) Obs. Denotes the number of observations and Std. Dev. stands for the Standard Deviation;(iv) N = 59 companies for all variables, with the exception of Renewable Energy (N=58), and CO2 Emissions (N=52)
Particularly interesting for this study, however, are the environmental performance measures. By comparing the mean and median of the two variables, it can be seen that both variables have a heavily skewed distribution. Also, as the high values of the standard deviations indicate, a large proportion of the observations lie in the extremes. For example, the maximum score of 100% renewable energy is shown for a total of 128 annual observations belonging to 26 different companies. This is approximately 43% of all the observations in the dataset. On the other hand, a total of 4 companies (Bedzin, British Energy Group, Centrica, and Drax Group) generate no renewable energy at al. The CO2 emissions
