Socially Responsible Investment and Market Performance: The Case of Energy and Resource Companies

Do financial markets reward the energy and resource companies for adopting socially responsible practices? In this study, we investigate the stock market performance of major international energy and resource firms, classified within the socially responsible investment (SRI) category, from 2005 to 2016. We simulate investments in the portfolios of the SRI energy and resource companies stocks during this 11-year period and we further assess their risk-adjusted performance. The returns of the energy and resource SRI portfolio as a whole were neither consistently superior nor inferior to those of the benchmark indices. However, there exist substantial differences across the individual sub-sectors. The overall results show that markets do not reward or penalize the energy and resource firms for their SRI attitudes. We also find that the crude oil price consistently had a significant influence on the stock returns of the SRI energy and resource companies.


INTRODUCTION
The global and national energy and environmental policy debates are increasingly shaped by the need to balance the competing objectives of economic efficiency, sustainability and affordability. Many energy and resource firms have noted the social and political changes in their environment and view the pursuit of profit for shareholders combined with social and environmental responsibility as part of their long-term corporate strategies. The recent developments in global climate agreements (e.g. COP21) and the emergence of the notion of 'stranded carbon' are examples of such contextual changes. The new operating environment represents a major departure from the "business-as-usual" conduct of business as these firms move from a production function of only private goods towards joint production of private and public goods.
From the theoretical point of view, firms undertake sustainable investments to improve their image, secure comparative advantage and maximise profits for their shareholders. This is particularly the case for energy and resource firms since they increasingly find themselves at the centre Open Access Article of the sustainability debate. However, empirical evidence about their performance in financial markets remains scarce. It is, therefore, worthwhile to examine whether the market rewards or penalizes this departure from the conventional profit maximisation model.
In order to achieve sustainable energy economy objectives, it is important to decouple energy use and its related emissions and environmental impacts from economic activity. Therefore, not only the governments but also energy and resource firms can have a crucial role to play through their actions and investments (see, e.g., IEA, 2014 and. In recent years, many major companies have adopted Socially Responsible Investment (SRI) principles as a strategic tool and self-regulation mechanism for improving corporate image and gaining competitive advantage. SRI has grown drastically over the past three decades. Forum for Social Investment reports that the assets invested in SRI companies in the US have increased by over 1260% to $8.72 trillion between 1995 and 2016 (a compound annual growth of 13.25%) representing nearly 22% of the $40.3 trillion total assets under management (USSIF, 2017). The number and value of SRI funds have also increased significantly, which has led to the creation of SRI indices, such as: Calvert Social Index, Domini400 Social Index, FTSE4GOOD Social Index and MSCI ESG Social Indices etc.
However, it is not clear from the literature whether investments according to the SRI principles provide higher, lower or similar returns in comparison with conventional stocks (see, e.g., the review studies by Margolis and Walsh (2003), Orlitzky et al. (2003) and more recently by Revelli and Viviani (2013)). In particular, the literature about the effect of SRI on performance of energy and resource firms is remarkably scarce (see, e.g., Jenkins and Yakovleva (2006), Frynas (2009) and Zhao (2015) for rare exceptions) and the available findings are inconclusive.
Our paper contributes to the literature on SRI investments and firms financial performance in general and in the case of energy and resource firms in particular. To the best of our knowledge, this paper is the first such study to analyse SRI investments in energy and resource companies on a global scale using international data from several markets in different geographical regions covering all six continents. We present novel empirical findings on the performance of international energy and resource SRI stocks. The findings are important for energy market and financial market researchers. In particular, they are of relevance for energy policymakers and for the investors in energy and resource firms.
We analyse the performance of energy and resource SRI companies on the international stock market and we simulate an investment in portfolios of such firms. We calculate raw returns of the energy and resource SRI stocks portfolios and analyse their performance using Fama-French (1992 and Carhart (1997) multi-factor models. Furthermore, we control for changes in the oil price by including the crude oil price returns in our Fama-French and Carhart estimations. We also measure the performance of the portfolio using risk-adjusted techniques, such as the Modified Sharpe Ratio (MSR) and the Certainty Equivalent (CEQ) returns. Moreover, by evaluating the profitability of stocks portfolios in the variants with and without dividends, we can extract the effect of dividends on their total returns. Finally, we analyse the performance in individual sub-sectors and examine the relation between the investigated stocks returns and the changes in the levels of the crude oil price.
The performance of energy and resource SRI stocks portfolios is assessed by comparisons with major global benchmarks, including broad market indices as well as the energy market, the SRI market and the alternative energy market sector indices (S&P Global 1200, MSCI World Energy, FTSE4GOOD and FTSE ET50 Index). The study encapsulates bull and bear market phases and allows the assessment of the impact of those market conditions on the profitability of energy and resource SRI stocks portfolios. We identify bull and bear market periods using the concept of non-overlapping "bull" and "bear" phases based on major peaks and troughs in the stock market indices, presented in Gooding and O'Malley (1977) and in Woodward and Anderson (2009), i.e. based on the price variability of indices and their long-term trends. Our sample is composed of global energy and resource stocks, hence we rely on the examination of bull and bear market phases of the S&P Global Index and the MSCI World Energy Index.
The next section presents the conceptual framework of our study. Section 3 reviews the relevant literature, which relates mainly to market performance of stocks and portfolios within the context of social responsibility. Section 4 presents a theoretical discussion of the effect of limiting the size of stocks portfolio due to imposition of the stocks exclusion criteria. Section 5 provides an overview of the data and the methodology. Section 6 presents and discusses the empirical results. Section 7 concludes the paper.

CONCEPTUAL FRAMEWORK
The conceptual framework of our analysis relies on two competing theoretical views about the profitability of investments in the SRI stocks.
The literature pointing towards a negative relationship between SRI and stock returns proposes two possible explanations. First, the cost of social responsibility is an extra expense for firms that reduces profitability. However, the SRI supporters argue that, over time, this extra cost is traded off by gains in reputation. Second, focusing on SRI companies as a subset of stocks reduces benefits of diversification (e.g., when stocks of tobacco companies are excluded from portfolios), which may result in lower risk-adjusted returns. On the other hand, the proponents of SRI argue that the excluded companies are engaged in unsustainable products or services that will make them less profitable anyway over time. These arguments are supported by many empirical studies that do not find meaningful differences between the performance of SRI and non-SRI stocks (see, for example, the results in Revelli and Viviani, 2013).
There is also a stream of literature that advocates a positive relationship between SRI and stock returns. The conceptual argumentation in this case is related predominantly to the instrumental stakeholder theory and the slack resources theory. Instrumental stakeholder theory postulates that companies aim to satisfy various stakeholder groups and that the resulting stakeholder-management relationships serve as monitoring and enforcement mechanisms leading to various positive side-effects, such as the increased efficiency of the firm's adaptation to external demands as well as better overall financial performance (Freeman and Evan, 1990;Hill and Jones, 1992;Jones, 1995;Clarkson, 1995).
Slack resources theory argues, in turn, that positive financial performance allows companies to become more socially responsible because it provides them with additional resources necessary to engage in corporate social responsibility, which usually requires availability of substantial excess funds (see Ullmann, 1985;McGuire et al., 1988;Waddock and Graves, 1997).
Other theoretical and conceptual views postulate that socially responsible companies are likely to benefit from different "mediating effects", such as improvement of reputation, better relations with financial institutions and investors as well as easier access to capital or even lower cost of capital (Spicer, 1978;Fombrun and Shanley, 1990). Further positive consequences of reputational effects, such as increase in employees' goodwill, may lead to improvement of the firm's financial performance (Davis, 1973;McGuire et al., 1988;Waddock and Graves, 1997).
There exist different channels through which financial performance of the SRI companies can be improved, for example through higher sales, better profitability or achieving lower cost of Open Access Article capital etc. Moreover, in the theoretical literature there are two other themes related to the conceptual discussions and their respective theoretical arguments that are tested empirically, i.e. the existing studies have also attempted to verify: (1) whether the socially responsible and ethical attitudes increase costs of firms' operations, which leads to negative impact on their financial performance and (2) whether social responsibility can be afforded by firms that already have good financial performance, which leads to feedback effects and further improvement of their financial situation.
In this study, we examine the main theoretical conjectures discussed in the above section. We empirically analyse the performance of portfolios composed of the SRI energy and resource companies stocks relative to benchmark portfolios using the data from major international financial markets.

RELEVANT EMPIRICAL LITERATURE
Theories and concepts of SRI have been evolving over time. In a review about social responsibility research, Lee (2008) found that the SRI debates have moved from macro level to micro (organisational) level over the last six decades. The literature in the 1950s and 1960s viewed social problems as a matter for politicians and civil society only. In the 1970s and 1980s, however, the literature began to investigate the relationship between social responsibility of firms and their financial performance. The practice of financial investments regarding the SRI attitudes has evolved and triggered more research. In a 2010 survey of 107 money managers on socially responsible investment, at least half of respondents saw social responsibility as a way to manage portfolio risk or to improve long-term performance (Voorhes and Humphreys, 2011).
The early research on the relationship between SRI and financial performance includes the seminal studies by Moskowitz (1972) and Vance (1975). While Moskowitz (1972) found a positive relationship between social responsibility and financial performance, Vance (1975) reported a negative relationship between them. However, both studies did not include the analysis of risk-adjusted returns which was later carried out by Alexander and Buchholz (1978), who used social responsibility ranking data from Vance (1975) and applied CAPM models to capture the market risk factor, yet they did not find a statistically significant relationship between social responsibility and stock market performance.
Our paper compares the performance of portfolios which are possible to construct by a private investor (i.e. stocks meeting certain screening criteria related to socially responsible investment). Thus, we next focus on the literature on market return and performance of stocks and portfolios within the context of socially responsible business. Margolis and Walsh (2003) and Orlitzky et al. (2003) reviewed the studies about the performance of SRI stocks and portfolios. They found that 54 papers showed a positive relationship with financial performance, while 28 others did not evidence any statistically significant relationship. Further 20 papers showed mixed findings, whereas seven papers found a negative relationship. Or-litzky et al. (2003) used a meta-analysis of 52 studies yielding a sample size of 33,878 observations and found a higher correlation between social responsibility and financial performance although the evidence appeared stronger for accounting based financial performance indicators compared to market based indicators. Derwall et al. (2005) used eco-efficient screening criteria of creating more goods and services using fewer resources and yielding less waste and pollution. Their study covering US data from 1995 to 2003, found that the high eco-efficiency portfolio provided substantially higher average returns than the low eco-efficiency portfolio. Differences in market sensitivity, investment style or industry-specific factors could not explain the performance differential and the results remained significant for transaction costs up to 200 bps. Derwall et al. (2005) suggested that the superior performance of a portfolio, constructed using environmental considerations as a key factor, could be a case of the market mispricing information on the ecological performance of companies. Kempf and Osthoff (2007) presented a trading strategy in which they simulated trades relying on buying stocks with higher ratings for social responsibility and selling those with lower ratings. They found an alpha of 8.7% per annum for investors employing the "best-in-class" screening approach. The increased performance continued even after taking into account reasonable transaction costs. Likewise, Statman and Glushkov (2009) found that stock portfolis with high ratings of a broad range of social responsibility characteristics outperformed those with low ratings. Their study showed community, employee and environment as some of the key screening factors that influenced performance. Ambec and Lanoie (2007) examined several studies in which portfolio analysis was applied to examine whether SRI funds (or indices) exhibit different performance from funds in a more general investment context. A majority of them (11 out of 16 papers) did not find statistically significant differences between the performance of the SRI funds and conventional ones, while in five studies the SRI funds outperformed. Ambec and Lanoie (2008) found companies benefitting from environmental performance. They showed positive links between environmental and economic performance citing examples of better opportunities for cutting costs and increasing revenues by environmentally friendly companies. Humphrey et al. (2012) investigated whether corporate social performance ratings have a systematic effect on the market based financial performance and risk of the firms. They applied the test for the UK companies over the period 2002-2011. They found no difference in the risk-adjusted performance of portfolios among firms which had high and low corporate social performance ratings. Galema et al. (2012) concluded that when considering the entire efficient frontier and not imposing any short sales restrictions, socially responsible US investors are generally worse off in mean-variance terms. However, they suffer only due to foregone risk reduction opportunities and not because of foregone returns. In addition, when short sale constraints are introduced, investors are no longer worse off by engaging in socially responsible investing activities. Brzeszczyński and McIntosh (2014) analysed the performance of the British SRI stocks in the period 2000-2010. Using the "Global 100 Most Sustainable Corporations in the World" list to select sustainable companies, they found average returns of SRI firms to be higher than those of market indices. The positive performance was also evidenced by risk-adjusted measures (certainty equivalent returns and modified Sharpe ratio) and a simple trading strategy did beat the market indices even after the inclusion of different levels of transaction costs.
In a recent meta-analysis of 85 studies and 190 experiments, Revelli and Viviani (2013) investigated whether inclusion of CSR and ethical criteria in the portfolio construction processes Open Access Article is more profitable than conventional investment policies. They found that, compared with conventional investments, the consideration of CSR in stock market portfolios is neither a weakness nor a strength.
The analysis of the SRI samples used in the existing literature further highlights that previous studies have applied data for stocks from different industries, which is likely to have an impact on the results. Kempf and Osthoff (2007) and Statman and Glushkov (2009) exploited data for stocks from KLD ratings, which consist of firms from different industry sectors. Kempf and Osthoff (2007) divided the companies into 10 industries for their best-in class approach of positive screening policy. Similarly, in Humphrey et al. (2012) the sample includes firms from 19 industries and Brzeszczyński and McIntosh (2014) also investigated stocks from more than 15 industry sectors. 1 Recent empirical studies from international markets, including those which analysed the performance of the SRI funds, the SRI stocks or the portfolios composed of SRI stocks (see, e.g., Lean et al., 2015;Auer, 2016;Auer and Schuhmacher, 2016;Syed, 2017;or Wu et al., 2017), also show mixed results.
Moreover, a recent paper by Riedl and Smeets (2017), based on surveys and incentivized experiments, found that both social preferences and social signalling effects can explain the SRI decisions, whereas financial motives play less important role. Socially responsible investors expected to earn lower returns on SRI funds than on conventional funds and also to pay higher management fees. Hence, the results from Riedl and Smeets (2017) suggest that investors are willing to sacrifice some financial performance if they invest according to their social preferences. These findings support some of the theoretical considerations, which we discussed above, and they are also consistent with much of the empirical evidence available in the literature from different international markets.
In summary, the review of the empirical SRI studies shows that the findings about the performance of SRI investments are inconclusive. Some of the existing evidence points towards superior performance of SRI investments (e.g. Derwall et al., 2005;Kempf and Osthoff, 2007;Statman and Glushkov, 2009), while many other available results differ (e.g. in Humphrey et al. (2012) a superior risk-adjusted performance could not be supported based on a range of market performance models) and those other papers do not confirm consistent outperformance.

THE EFFECT OF LIMITING THE PORTFOLIO SIZE DUE TO EXCLUSION CRITERIA
The effect of the exclusion of stocks from the pool of all the stocks available in any given market is a priori uncertain and, in fact, such decision can lead to either lower returns or higher returns of the constructed portfolio (or it can result in no change at all). Below we provide a theoretical discussion regarding this issue and we also demonstrate what happens if the exclusion criteria are imposed on a group of stocks (in the case, which is the subject of investigation in our study, on the non-SRI energy and resource companies stocks) under different possible scenarios.
Let i R denote the return for stock i in a market composed of a total number of I stocks, where i = 1, 2, 3, ... I. The stocks are classified into two groups: SRI energy and resource companies 1. Methodologically, it is not clear how the effect of performance of stocks from different industries (which may, again, have different degrees of social responsibility etc.) is captured by the commonly applied tools, such as through the estimations of multi-factor models. We simplify this problem by using only companies that are focused on the production and supply of energy and related resources (e.g., oil, gas, water and minerals) etc. whereas all of them are characterised by substantial social and environmental responsibility and have been screened as SRI. This sample selection allows us to observe the performance of large and well established SRI firms making our study novel and different from others in the existing literature. stocks, denoted by j = 1, 2, 3, ... J, and all other stocks which do not meet the SRI energy and resource companies stocks selection criteria, denoted by k = 1, 2, 3, ... K. These two sets are mutually exclusive and fully complementary, i.e.: the relation J + K = I must always hold.
The return achieved from the market portfolio p R composed of all stocks I available in any given market is: where: i R are the returns of stocks i, j R are the returns of the SRI energy and resource companies stocks j and k R are the returns of the non-SRI energy and resource companies stocks k. The respective weights are: i w (for i = 1, 2, 3, ... I), j w (for j = 1, 2, 3, ... J) and k w (for k = 1, 2, 3, ... K) and they must always sum up to 1 within each group, i.e.: ∑ I i =1 w i = 1, ∑ J j =1 w j = 1 and ∑ K k =1 w k = 1. The effect of the exclusion of one of the two groups of the distinguished stocks, in this case the removal of the stocks which do not meet the SRI energy and resource companies stocks selection criteria (k = 1, 2, 3, ... K), on the portfolio return p R is as follows: , then the exclusion of the non-SRI energy and resource companies stocks is beneficial for the portfolio performance, because p R increases after the removal of stocks k = 1, 2, 3, ... K.
, then the exclusion of the non-SRI energy and resource companies stocks is detrimental to the portfolio performance, because p R decreases after the removal of stocks k = 1, 2, 3, ... K.
, the exclusion of the non-SRI energy and resource companies stocks does not have any effect on the return p R of the portfolio.
Note that the above relationship holds under all possible combinations of returns, including the cases when ∑ . Moreover, this relation is also true regardless of the size of the groups, i.e. for any possible values of I, J and K. For , the exclusion of the non-SRI energy and resource companies stocks (all stocks k = 1, 2, 3, ... K) is detrimental to the portfolio performance, because p R decreases in all these cases.
, the exclusion of the non-SRI energy and resource companies stocks (all stocks k = 1, 2, 3, ... K) is beneficial for the portfolio performance, because then p R increases in all these cases. However, an important practical issue to consider here that also determines the relation between ∑ J j =1 w j • R j and ∑ I i =1 w i • R i is the method of weighting stocks, i.e. how the weights i w (for i = 1, 2, 3, ... I), j w (for j = 1, 2, 3, ... J ) and k w (for k = 1, 2, 3, ... K) are assigned. For example, the assumption of equal weights will inevitably lead to different values of ∑ J j =1 w j • R j and ∑ K k =1 w k • R k , and consequently to different p R values, than the assumption of unequal weights. The portfolios constructed in practice by stock market investors may have either equal weights or weights different from an equal structure, i.e. they can be allocated through the optimization procedures (e.g. based on the mean-variance relationship) or they can be determined by the size of the stocks (usually measured by their market capitalization) or some other criteria (e.g., the value of such indicators as P/E, P/BV or D/Y ratios etc.).

Open Access Article
Hence, we can conclude that the effect of the exclusion of any one of the two groups of stocks from the broad market portfolio (in the case discussed in this paper the removal of the stocks which do not meet the SRI energy and resource companies stocks criteria) depends predominantly on the sign of the relation defined as: but also on the method of weighting stocks and the assumptions regarding the weights, which ultimately determine the return R p as well as the values of ∑

Data
The sample selection process required us to analyse first the scope of business activity of all 344 companies from the Global-100 list, which appeared on it during the first 11 years since its launch in 2005. As the focus of this study is on energy and resource SRI stocks, from the Global-100 list we chose companies that: (a) produce energy, minerals and water, (b) produce energy related materials for consumption in energy or transport industry and (c) supply energy, minerals and water. 2 This selection led to the identification of the following industry groups: (1) Alternative Energy, (2) Electric Utilities, (3) Electricity, (4) Energy Equipment and Services, (5) Gas, Water and Multiutilities, (6) Industrial Engineering, (7) Mining, (8) Oil Equipment, Services and Distribution and (9) Oil and Gas Producers.
We used the energy and resource SRI stocks data from the compilation published by Corporate Knights based in Toronto, Canada, which prepares annually the "Global 100 Most Sustainable Corporations in the World" list. The basis for it is mainly quantitative as it provides scores against 12 different KPIs 3 for all global publicly traded companies exceeding market capitalisation of at least US$ 2 billion. For example, the screening requires all the companies to pass nine different Piotroski F-Score (Piotroski, 2000) tests, which confirm the sound financial position of the firms. Similarly, in the sustainability disclosure principle, companies that fail to disclose at least 75% of the priority indicators for their respective GICS Industry Group are eliminated. As such, the results follow a rulesbased construction methodology and ensure sustainable financial index of firms. The listing criteria also include screening against product category (e.g. tobacco is not included), sustainability-related fines, penalties or settlements.
We filtered all SRI companies based on the above categories and this procedure identified 56 SRI energy and resource companies in the 11 years period between February 2005 and January 2016. Table 1 presents the constituents of the SRI energy and resource companies stocks portfolio (henceforth, referred to as: 'SRI E&RC stocks portfolio') used in this study. It also provides information about the country of origin, area of operation, number of employees and year of establishment.
2. The purpose of our study was to analyse the performance of stocks which simultaneously meet the SRI and energy / resource industry membership criteria. Therefore, we deliberately chose stocks which are at the intersection of energy /resource and SRI selection rules. We do not test the performance of separate portfolios of energy / resource stocks and SRI stocks, however we investigate it indirectly by comparing the results of our portfolios with pure SRI and pure energy sector indices.
3. These key performance indicators (KPIs) are: Energy Productivity, Carbon Productivity, Water Productivity, Waste Productivity, Innovation Capacity, Percentage Tax Paid, CEO to Average Worker Pay, Pension Fund Status, Safety Performance, Employee Turnover, Leadership Diversity and Clean Capital Pay Link. More details are available at: www.corporateknights.com. Source: Data collated by authors from companies' websites, annual reports and from Bloomberg.

Open Access Article
As shown in Table 1, the list of the constituent companies in our SRI E&RC stocks portfolio consists of long established large firms. For example, BP Plc, Lonmin Plc, PG & E Corp, Teck Resources, Tokyo Gas, Umicore and Wartsila OYJ are more than a century old. There are few companies that were founded more recently but have a long history. For example, the newest company on the list, Cenovus Energy Inc. formed in 2008, is a split from Encana which descends from the 19th century Canadian Pacific Railway. Similarly, BHP Billiton was incorporated in 2001 but it was a merger of Billington and BHP that were established in 1860 and 1885, respectively. Likewise, Aluminia Limited, established in 2002 is a demerger from WMC Resources which had a history dating back to 1950s.
Many of these companies have grown large over time and they have a presence in many countries (e.g., BP has operations in 80 markets). These firms contribute to the national economies and provide employment in communities. They produce gas, oil, minerals and electricity with a range of local and global environmental impacts. Therefore, these firms are widely believed to bear important social, economic and environmental responsibilities. The companies in our sample have more than 25,000 employees on average. Those firms with relatively fewer employees, such as Cairn Energy from the United Kingdom which officially had 156 employees as of year-end 2016, as mentioned in the annual report, also had several hundreds contractors in 2016.
In terms of geographical distribution, the 56 stocks in our database come from 19 countries of which the highest number of firms is from the UK (11 companies) followed by Canada (9 companies). There are 7 companies form the US and 4 from Finland and Spain. Further, Australia and Brazil have 3 companies each. France, Japan and Norway are represented by 2 companies each and the remaining 9 countries have 1 company each. Given that most countries in the world have at least one energy company, the Global-100 ranking concentration in less than 10% of all countries worldwide is an indication that in many of them the SRI related criteria do not seem to be considered to a sufficient extent by energy companies there. Figure A1 in the Appendix shows the countries and number of SRI energy and resource companies in the SRI E&RC stocks portfolio. Table 2 presents the constituent companies in the Global-100 list broken into numbers for each year.
The source of all the stock price and dividend data for the constituents of the analysed SRI E&RC stocks portfolio is Bloomberg.
We used the ticker symbol of the respective stock exchange, so the price at first was obtained in the currency of the country of the exchange and then we used the Bloomberg currency converting function to change both the stock price and dividends data into US dollars in order to maintain uniformity and consistency for the calculation purposes. Where stock price and dividends were not quoted in full currency value (e.g., pound sterling quoted in pence), we converted them into the respective unit of currency (e.g., to pound sterling) before applying the USD conversion.
Similar to the approach used by Brzeszczyński and McIntosh (2014), the returns of the SRI portfolios were compared with the returns of various stock market indices. However, we extend this type of analysis by utilizing a larger number of comparable benchmarks. We employ four benchmark indices as opposed to only two (FTSE100 as the broad market and FTSE4GOOD as the SRI index) in Brzeszczyński and McIntosh (2014). Our selection of benchmarks captures stocks globally and covers the broad market as well as energy market, SRI and alternative energy market sectors, which creates a broader perspective for the comparison purposes. Barrick Gold Corp Cenovus Energy Inc Note: 'x' means that the respective company appeared on the Global-100 list in the indicated year(s) and, therefore, it is accordingly included in the sample for the analysis.
Open Access Article

(1) Broad Market
As the broad market index, we employ the S&P Global 1200, which is a composite index comprising seven regional and country indices: S&P 500, S&P Europe 350, S&P/TOPIX 150 (Japan), S&P TSX 60 (Canada), S&P/ASX 50 (Australia), S&P Asia 50 and S&P Latin America 40. The S&P Global 1200 is calculated in US dollars. The index captures 70% of the global market capitalisation covering 30 countries inclusive of the countries of origin of the stocks in our SRI E&RC stocks portfolio only except for the stocks from India and South Africa. The main selection criterion for S&P Global 1200 is company size measured by its stock market capitalisation. Hence, it contains predominantly large blue-chip firms. Additional selection criteria are stocks liquidity, which is revised at a monthly frequency based on such indicators as stock's annual value traded, its float turnover and the number of days traded. The S&P Global 1200 index takes into account also sectoral classifications and ensures balance between 10 main broad economy sectors with respect to Global Industry Classification Standard (GICS).

(2) Energy Market Sector
We include the MSCI World/Energy Index as a benchmark for the energy sector. It is designed to capture the large and mid-cap segments across 23 Developed Markets (DM) countries (16 of which are common with the countries of origin of our SRI energy and resource companies stocks). Moreover, it maintains sectoral classifications among seven energy categories that are again common to our SRI E&RC stocks portfolio. The selection criteria are based on index construction approach with a strong emphasis on index liquidity, investability and replicability, which allows for cross regional comparisons across all market capitalisation sizes, sectors and style segments. Similar to S&P 1200 Global index, securities in MSCI World Energy Index are classified in the energy sector following the Global Industry Classification Standard (GICS).

(3) SRI Market Sector
In the SRI category, we use the FTSE4GOOD Global 100 (referred to, henceforth, as: FTSE4GOOD) index as a benchmark. It includes companies with high environmental, social and governance (ESG) ratings. The FTSE4GOOD index is designed to measure the performance of companies that meet globally recognised corporate responsibility standards. The selection criteria are revised on regular basis to meet market expectations and reflect the new developments in the CSR practice. They rely on extensive market consultation process and are approved by an independent committee of experts. The FTSE4GOOD inclusion criteria are split into five areas: (i) environment, (ii) human and labour rights, (iii) supply chain labour standards, (iv) countering bribery and (v) climate change. Each of them is further divided into three categories: (i) policy, (ii) management and (iii) reporting. Subsequently, there are indicators assigned to each of the policy, management and reporting subdivisions. The number of the indicators that a company must meet depends on whether it is classed as having high, medium or low impact in a particular area. Moreover, FTSE4GOOD index excludes the companies with business interests in some industries, such as: tobacco producers or weapons manufacturers.

(4) Alternative Energy Market Sector
In the case of alternative energy market sector, we employ the FTSE ET50 index, which is composed of global companies that are involved in clean energy related businesses. It is designed for the creation of index tracking funds, derivatives and used as a performance benchmark. The selection criteria of the index lead to a diversified mix of clean energy production and clean energy technology and equipment provider companies. Therefore, during the selection process the stocks are screened and weighted to ensure that the index is investable and also sufficiently liquid for trading purposes. The FTSE ET50 index consists of companies from the list of 17 countries (9 of which are common with the countries of domicile of our SRI energy and resource companies stocks). Furthermore, it maintains sectoral classifications among 8 industries including oil and gas, materials and utilities that are, again, common with the industry types of the companies in our SRI E&RC stocks portfolio.
We evaluate the performance of our portfolios against the four indices described above both at price and total return definition levels.
First, we compare the results of the investment in the SRI E&RC stocks portfolio with the 'price index' (PI) versions of the four indices mentioned above. However, the SRI E&RC stocks portfolio includes dividend payments, which is the income to investors holding these stocks. Therefore, we also analyse the returns of SRI E&RC stocks portfolio against the 'total return index' (TRI) versions of the four indices (i.e. the versions of the indices which include dividend payments) in order to ensure that the comparison is conducted on equal ground. On the other hand, the 'total return' versions of the indices are not commonly used by investors as conventional benchmarks. Hence, we also perform direct comparison between the 'price index' versions of the indices and the SRI portfolios without dividends.

Methodology
The Global-100 list was used to construct portfolios of global socially responsible energy companies over the period from February 2005 to January 2016 (11 annual sub-periods) and their returns were compared to the returns of the respective benchmark indices. Since the Global-100 list is announced at the end of January each year, right before the meeting of the World Economic Forum (WEF) in Davos, we assumed that the first portfolio was constructed on the 1 st February 2005. The portfolios were then rebalanced each year on the last working day of January.
The selection procedure of stocks entering the portfolios was as follows. The companies identified on the Global-100 list entered the portfolio in the first year and the portfolio was held until the next Global-100 list was announced a year later. Stocks that no longer appeared on the Global-100 were removed from the portfolio and the companies new to the Global-100 list were included. Effectively, this means that we simulate the trades relying on buying stocks that appeared on the list and selling those that were removed from it. This procedure was repeated every year until the last year in our sample period.
As the Global-100 was an unranked list for a number of years (ranking has been only provided since the year 2010) rather than an index, it had to be assumed that each stock has an equal weighting in the SRI portfolios. This means that a stock which remains in the portfolio from one year to the next when the total number of stocks in the portfolio changes requires an adjustment (either additional purchases or sells) in order to maintain the same equal weighting.
When a company was taken over and disappeared from the stock market in the period of the duration of our portfolios, we assumed that the proceeds were kept in a non-interest bearing Open Access Article account until the portfolio was rebalanced. The reason for this assumption is that private investors are less likely to insist on reinvesting the proceeds and may keep them in their current account until the portfolios are re-shuffled.
The stock price data and dividend payments data were collected and included in the analysis of the SRI E&RC stocks portfolio performance. Data on prices and dividends was imported from Bloomberg.
As mentioned above, similarly to Kempf and Osthoff (2007) and Brzeszczyński and McIntosh (2014), the returns of the SRI portfolios are compared to the returns of market indices. The annual simple holding period returns for the SRI portfolios in two versions (with dividends and without dividends) as well as for the following indices: S&P Global 1200 (price index), S&P Global 1200 (total return index), MSCI World/Energy (price index), MSCI World/Energy (total return index), FTSE4GOOD (price index), FTSE4GOOD (total return index), FTSE ET50 (price index) and FTSE ET50 (total return index) were calculated for all 10 individual years and average annual returns were computed for five-year sub-periods and for the overall ten-year period. In addition, we analyse returns in both bull and bear market periods.
The results in these sub-periods allow to conduct a deeper analysis of the profitability of SRI portfolios and to perform further robustness checks. The annual return was determined as a simple holding period return with any dividends added. For the one-, five-and 11-year periods, the average annual returns using the annual data were calculated. For other sub-periods, returns were calculated using monthly data and then annualised to make them comparable with other periods. Whether the differences between returns of the SRI E&RC stocks portfolio and the benchmark indices were statistically significant was assessed by a t-statistic.
We also analyse the performance of the SRI E&RC stocks portfolio by using the most important risk-adjusted measures, such as the modified Sharpe ratio of Israelsen (2005) and the Certainty Equivalent returns (see, e.g., DeMiguel et al. (2009)), which were calculated for both versions of the SRI E&RC stocks portfolio (with and without dividends) and both versions of all four indices (total return indices with dividends and price indices without dividends).
The Sharpe ratio (Sharpe, 1966 and1994) measures excess return per unit of total risk. However, the classical definition of the Sharpe ratio suffers from inaccuracy errors and incorrect assessment of risk when returns are negative in some sub-periods, so we calculated the modified Sharpe ratio (MSR) of Israelsen (2005): where ER is the excess return defined as mean monthly difference between the portfolio (or index) return and the risk-free return computed for n equal to 12, 60 or 132 months, respectively, and SD is the sample standard deviation of the monthly differences of returns. MSR is a commonly used measure to address the problem of negative returns and alleviates the problems with the traditional Sharpe ratio.
Certainty Equivalent (CEQ) returns are defined as: where μ k and 2 σ k are the mean and variance of excess returns of a given portfolio or an index k and γ is the risk aversion parameter. The formulation of CEQ in (3) assumes a multi-period investor with quadratic utility. The 'normal' level of risk aversion is at the level γ =1, while higher (lower) values of γ indicate higher (lower) levels of risk aversion.

Socially Responsible Investment and Market Performance / 31
Open Access Article Finally, we estimate parameters of the Fama-French three-factor model (Fama and French, 1992;: R pt -R ft = α p + β 1p RMRF t + β 2p SMB t + β 3p HML t + ε pt (4a) and the Carhart (1997) four-factor model: where R pt is the return of the SRI portfolio in period t; R ft is the risk-free return in period t; R mt is the return of the world stock market index in period t and RMRF t = R mt -R ft ; SMB t is the difference in return between small-cap and large-cap portfolios in period t; HML t is the difference in return between high book-to-market stocks (i.e. value stocks) and low book-to-market stocks (i.e. growth stocks) in period t; MOMENTUM t is the difference in return between portfolio of stocks classified as those that have strong momentum and stocks classified as those that have weak momentum and ε pt is the error term. The data for the explanatory variables used in models (4a) and (4b), i.e. for R ft , R mt , RMRF t , SMB t , HML t and MOMENTUM t , were obtained directly from the Ken French database. 4 Defined as Fama/French Global Factors and Portfolios, the factors data is constructed from the portfolios of stocks of 23 different countries. We adopted the factor data from Fama/French Global Factors because 16 out of 19 stocks in our portfolio are from the countries in the list of Fama/French Global Factors.
Market factor is defined as the return of a region's value-weighted market portfolio minus the US one month T-bill rate. SMB is the equally weighted average of the returns of the three small stock portfolios for the region minus the average of the returns of the three big stock portfolios:

SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1/3 (Big Value + Big Neutral + Big Growth) (5)
HML is the equally weighted average of the returns for the two high book-to-market (B/M) portfolios for a given region minus the average of the returns for the two low B/M portfolios:

HML = 1/2 (Small Value + Big Value) -1/2 (Small Growth + Big Growth)
MOMENTUM is the equally weighted average of the returns for the two winner portfolios for a given region minus the average of the returns for the two loser portfolios:

MOMENTUM = 1/2 (Small High + Big High) -1/2 (Small Low + Big Low).
( We also perform estimations of the Carhart (1997) model with crude oil returns as additional control variable based on the following model: where: OIL t is the return of the Brent oil price. Finally, we explore the impact of the crude oil price on the portfolio returns and we estimate the parameters of the following model: Open Access Article as well as the model where the dependent variable is defined as the excess return of the SRI portfolio relative to the international stock market benchmark: (10) where ER pt is the excess return defined as the difference:

ER pt = α p + β p OIL t + ε pt
and where R mt is the return of the world stock market index in period t.
In the next section, we present the results of the analysis of the raw returns of our SRI E&RC stocks portfolio and assess its performance relative to the selected benchmark indices as well as using the risk-adjusted measures described above.

Raw Returns
The results of the preliminary analysis based on raw returns for the entire portfolio of the SRI energy and resource companies stocks show that it did not outperform the broad market index as well as other energy sector, SRI and alternative energy market indices in the 11-year sample period from February 2005 to January 2016. Table 3 presents average annual returns for our whole sample period of 2005-2016 based on the simulation of investment in the energy and resource companies from the Global-100 list compared to all four benchmark indices and reports also the values of the respective t-statistics.
Panel A in Table 3 reports the returns of the SRI E&RC stocks portfolio with dividends and the returns of the benchmark indices in their price index version. Such comparison with stock market indices is often used by financial market investors as well as business media, however it is not entirely accurate because price indices by definition do not include dividends, while any investment in stocks (e.g., in a portfolio of SRI energy and resource companies stocks such as the one investigated in our study) in practice will benefit from the dividends paid out by companies. Nevertheless, we start with such comparison because stock market performance relative to price version of stock indices is often discussed in business media etc., so regardless of their validity for this purpose, they are important benchmarks to which our results should first be referred.
As Panel A in Table 3 shows, there is a slight outperformance of the SRI stocks portfolio relative to all four benchmarks. The average annual return is 1.89% and it is higher than the respective average annual returns equal to 1.20%, -0.24%, -0.63% and -0.42%, although the differences are not statistically significant. However, the outperformance in the individual years in terms of the numbers of the annual periods characterized by the superior results is broadly equal in all 5 cases: the SRI E&RC stocks portfolio has outperformed the index benchmarks 2 times, while the indices S&P GLOBAL 1200 TR Index, MSCI WORLD ENERGY TR Index, FTSE4GOOD TR Index and FTSE ET50 TR Index have outperformed others 2, 3, 2 and 2 times, respectively. Panel B in Table 3 presents the returns of the SRI E&RC stocks portfolio with dividends and the returns of the benchmark indices in their total return versions. This comparison in Panel B is the most relevant one from the practical point of view, because it allows (unlike the results in Panel A) for direct assessment of the same type of returns (in this case: the returns including the dividends) and it reflects the actual investments outcomes (unlike the results in panel C which do not take into account the dividends).
Panel B in Table 3 shows that there is no clear pattern of outperformance by either the SRI stocks portfolio or any of the benchmark indices. In the full period from 2005 to 2016 the SRI stocks portfolio achieved the average annual return equal to 1.89%, while the return of the S&P GLOBAL 1200 TR Index was 3.64%, the return of the MSCI WORLD ENERGY TR Index was 2.44%, the return of the FTSE4GOOD TR Index was 1.13% and the return of the FTSE ET50 TR Index was 0.43%. The average value of the returns of these four benchmarks is 1.91%, which is almost exactly the same as the return of 1.89% for the SRI stocks portfolio. Moreover, as in case of the results in Panel A, the outperformance in the individual years in terms of the numbers of the annual periods characterized by the superior results is also very similar in all 5 cases: the SRI E&RC stocks portfolio has beaten the index benchmarks 2 times, while the indices S&P GLOBAL 1200 TR Index, MSCI WORLD ENERGY TR Index, FTSE4GOOD TR Index and FTSE ET50 TR Index have beaten others 2, 3, 2 and 2 times, respectively.
Finally, Panel C in Table 3 presents the returns of the SRI E&RC stocks portfolio without dividends and the returns of the benchmark indices in their price index version. The purpose of this comparison is to examine and compare the relative performance of our stocks which, at the same time, illustrates the impact of dividends on the SRI stocks portfolio and on the indices.
In the full period from 2005 to 2016 the SRI stocks portfolio without dividends achieved negative average annual return equal to -1.36%, while the corresponding value for the best performing benchmark S&P GLOBAL 1200 Price Index was positive and equal to 1.20%. The average value of the returns of the four benchmarks is -0.02%, which shows that dividends played a relatively more important role in the performance of the investigated portfolio than in case of the benchmark stock market indices.
Apart from the calculation of the average annual returns and the returns in the single annual periods, we also investigated the performance in other sub-samples, i.e. in the rolling 5-year long periods and in the bull and bear market phases. 5 Overall, the performance during the multiple-year periods and during the bull and bear market conditions was mixed and the differences in returns were not statistically significant, although the S&P GLOBAL 1200 Price Index has beaten others most often in the 5-year long rolling samples and also in the bear market phases, while FTSE ET50 TR Index was the best performer in the bear market phases.
Despite the fact that there is no clear evidence of the overall outperformance that could be detected in our results presented in all three panels in Table 3, the variation of returns over time across the individual years shows an interesting pattern that we found in our study. Such effect can be explained by two major events on the global market: the global financial crisis and the changes in the level of the crude oil price.
First, the performance of the analysed SRI E&RC stocks portfolio was the worst in the annual period 2008/2009, which directly follows the global financial crisis of 2007/2008. The total return of the portfolio was -42.39%, while the return excluding dividends was -44.36%. Nevertheless, it needs to be emphasized that this result was still comparable with the changes of the benchmark 5. Bull and bear market periods were have been identified using the idea of non-overlapping 'bull' and 'bear' phases based on major peaks and troughs found in the stock market indices, presented in Gooding and O'Malley (1977) and Woodward and Anderson (2009)

Table 3: Average annual returns for the whole 11-year period (February to January) from 2005 to 2016 for the SRI E&RC stocks portfolio and for the benchmark indexes: 1) Global broad market (S&P Global 1200), 2) Global energy market (MSCI World Energy), 3) Global SRI market (FTSE4GOOD) and 4) Global alternative energy market (FTSE ET50).
Open Access Article Notes for Table 3 ( (Table 3, Table 4, Table 5, Tables 7a, 7b, 7c and 7d, Appendix Tables A1 and B1-B4) bull and bear market periods have been identified using the idea of non-overlapping 'bull' and 'bear' phases based on major peaks and troughs found in the stock market indices, presented in Gooding and O'Malley (1977) and more recently in Woodward and Anderson (2009) indices over the same period, which also suffered severe losses. Hence, this negative performance is clearly related to a broader stock markets trend after the global financial crisis. Second, the worsening performance of portfolios starting from the annual period 2011/2012 onwards coincides in time with a decline of the crude oil price, which started to slide down from its peak in 2011 (which was the second peak in our whole sample period after its previous peak in 2008). The decrease of the crude oil price was first gradual and then substantially accelerated after 2013. More importantly, the performance of portfolios has been worsening also in relative terms after 2011. We interpret this effect as the impact of declining crude oil price on the profitability of many companies, which business directly (or indirectly) depends on crude oil price levels and which stocks were part of our portfolios. 6

The Effect of Dividends
Our calculations allow us also to extract the impact of dividends on the SRI E&RC stocks portfolio performance, which can be directly conducted by comparing the results for the variants of portfolios with and without dividends that are contained in the respective panels in Table 3.
The average annual return of the SRI E&RC stocks portfolio in the variant where the dividends were included in the calculations is 1.89%, while in the variant where the dividends were excluded it is -1.36%. This result has a very straightforward interpretation and also practical implications for stock market investors, which are as follows.
First, the difference in the annual average return that is substantially over 3% (i.e. 1.89% minus -1.36% equal to 3.25%) is large, which indicates that dividend payments matter to investors who allocated their funds in the stocks from our SRI E&RC stocks portfolio. Second, as mentioned in the previous section, the dividends play a relatively more important role in the performance of the investigated portfolio than in case of the benchmark stock market indices. 7 Third, in terms of the qualitative conclusions it makes also considerable difference whether dividends are added or excluded from the calculations, because the annual average return is either positive or negative in these two cases, hence leading to either the overall investment profit or the overall investment loss.
Therefore, dividends appear to matter in the performance of the analysed SRI E&RC stocks portfolio and its individual stocks.
Our results also mean that the SRI energy and resource companies tend to pay relatively large dividends, which is another important finding of this study.
6. This finding is clearly supported subsequently by the estimation results of the parameters of the Carhart model with crude oil price returns as a control variable (discussed later in this paper). The respective estimates of the parameter for crude oil returns are statistically significant and positive, which means that the SRI E&RC stocks portfolio returns are related indeed to oil price returns in the same direction. Therefore, the negative oil price returns starting from the year 2011 are associated with negative returns of the SRI energy and resource stocks portfolios.
7. The comparison of data for average dividend yield indicators for different industries also supports this effect. For example, in our sample period the average dividend yield for the energy industry stocks from the MSCI World Energy Sector Index and S&P Global 1200 Energy Sector Index was 2.84% and 2.94%, respectively, while it was as a rule lower for other industries, e.g. 2.38% for the financial industry stocks from the S&P 500 Financials Sector Index, 1.33% for the information technology stocks from the S&P Global 1200 Information Technology Sector Index or 2.16% from the MSCI World Health Care Index etc.
Open Access Article

Risk-Adjusted Performance
In the next step we turn towards the analysis of the risk-adjusted measures, such as the modified Sharpe ratio (MSR) and Certainty Equivalent (CEQ) returns, as well as the evaluation of the portfolio performance based on the Fama-French and Carhart models.
The values of the modified Sharpe ratio (MSR) are presented in Table 4. They show similar pattern of worsening performance of the SRI E&RC stocks portfolio over time, which is also consistent with the evolution path of the crude oil price.
The values of Certainty Equivalent (CEQ) returns are presented in Table 5 for three variants representing normal risk aversion of investors (γ =1), lower risk aversion (γ =0.5, i.e. half of normal risk aversion level) and higher risk aversion (γ =2, i.e. double the normal risk aversion level). Similarly to Tables 3 and 4, they illustrate the same pattern of results for the profitability of the SRI E&RC stocks portfolio with superior performance in the first two 5-year sub periods (2005-2010 and 2006-2011) and then a substantial deterioration with subsequent underperformance in the next periods.
In the next step we move to the analysis of the Fama-French three-factor model and Carhart four-factor model, which are the most widely used multi-factor models for explaining the performance of investment funds or stock portfolios. Due to space considerations, we focus here on the presentation and discussion of the more extended specification of the and Carhart four-factor model, which encompasses the Fama-French three-factor model, however all the results are available upon request.
In all regressions we first tested for the presence of possible seasonality. Next we performed tests for autocorrelation and heteroscedasticity. For autocorrelation we used Ljung-Box Q test and for heteroscedasticity we applied the ARCH test of Engle (1982). When heteroscedasticity was present in a model, it was addressed by estimating an appropriate GARCH class model. Autocorrelation was removed by adding autoregressive (AR) and/or moving average (MA) terms. Table 6a presents the estimation results of the parameters of Carhart four-factor model represented by equation (4b). In the whole sample only the market factor RMRF t is statistically significant (estimate of 1.07 significant at the 1% level). In the sub-samples, the RMRF t variable is significant in all the 5-year long sub-periods and in most of the single-year sub-periods. The SMB t, HML t and MOMENTUM t factors are mostly insignificant in the sub-samples.
The estimation results of the alpha (constant) parameter presented in Table 6a show that it is negative but statistically not significant in the full period from 2005 to 2016. In the shorter 5-year sub-periods, its estimates are positive and statistically significant 8 in the first two sub-sam-8. The positive and significant estimates of the alpha may imply market inefficiency, however we found this effect only at the beginning of the whole analysed period, so the conclusions about market efficiency have to be carefully formulated. Moreover, a comprehensive investigation of market efficiency would require access to very detailed microstructural data for individual trades and this was beyond the scope of our study. Therefore, we can only conclude that our results might suggest some market inefficiency, although a clear lack of consistency in overperformance points towards the validity of the adaptive market efficiency hypothesis (AMH) proposed by Lo (2004 and (some more recent evidence on AMH is provided e.g. by McGroarty (2014 and2016) or Manahov and Hudson (2014), among others) rather than efficient market hypothesis (EMH). Adaptive market hypothesis incorporates the principles of evolution, such as: adaptation or natural selection, to explain financial markets mechanisms. It is consistent with the evolutionary model of individuals adapting to a changing environment using heuristics. In the context of our study, according to the AMH, the stock prices reflect the information that combines environmental conditions and their movements are the result of interaction of different distinct groups of investors. Under the AMH, the degree of market efficiency is a function of such factors as the number and type of competitors in the market or their adaptability to the evolving market conditions. There are also important theoretical implications of AMH in light of our research and the results reported in this paper provide empirical support for them: (1) relation ples (2005-2010 and 2006-2011), however they are becoming negative and statistically significant in the last three sub-samples (2009-2014, 2010-2015 and 2011-2016). This pattern is entirely consistent with the findings presented earlier in Tables 3, 4 and 5 for the raw returns and for other  risk-adjusted measures. 9  Table 6b presents the results from the estimation of the Carhart model with the fifth variable, i.e. the crude oil returns, which serve as the control variable. Its estimate for the entire period is positive and equals 0.1016 (statistically significant at the 1% level). With other estimation results for other variables broadly unchanged in comparison with Table 6a, this finding means that the crude oil price was an important factor in explaining stock returns of the companies from the SRI E&RC stocks portfolio, which is not very surprising given that many of them are directly involved in crude oil business or their financial situation heavily relies (directly or indirectly) on the crude oil price.
However, an interesting effect is the pattern of results for the crude oil returns estimates across all the 5-year sub-periods, which shows a clear decline in the value of the estimated parameter over time (and loss of significance) from 0.0884 (significant at the 5% level) to 0.0014 (not significant), although they are positive in all these sub-samples. This finding means that the crude oil price movements have been an important factor in explaining the SRI E&RC stocks portfolio returns, but their influence weakened over time.
We further explore the role of the crude oil price movements using additional models in sections 6.3 and 6.4.
In the next section 6.2, we also investigate in more details the performance of the SRI E&RC stocks within different sub-groups.

Results for the Sub-groups within the SRI E&RC Stocks Portfolio
In the next step, we inspect more closely what actually happens inside the entire SRI E&RC stocks portfolio by investigating the performance of stocks from the individual sectors. Subsequently, we focus on the analysis of two broader groups of stocks: oil related companies and nonoil related companies.
The results across the distinguished 9 sectoral groups differed quite substantially. The best performing sectors were Alternative Energy and Gas, Water and Multiutilities, which stocks achieved the highest returns equal to 9.44% and 7.17%, respectively, while the worst performing sector was Mining characterised by negative return equal to -16.55%. All returns for the whole 11year period (February to January) from 2005 to 2016 for the individual sub-sectors within the SRI E&RC stocks portfolio (in the variant with dividends) are presented in the Appendix in Table A1. between risk and reward is unlikely to be stable over time, (2) investment strategies perform better in certain environments and worse in others, (3) profit and utility maximization are secondary objectives for investors, whereas their primary goal is survival and (4) survival is achieved through innovation (given that the risk-reward relationship is time-varying in nature, adaptation to changing market conditions is a natural way to behave and to achieve a desired level of expected returns in financial markets).
9. Although the Fama-French and Carhart models are time series models and they are based on time series data (in case of this paper on data at the monthly frequency of observations), such databases can be treated also as panel data, if the portfolio returns are disaggregated into individual stocks returns. Therefore, as a robustness check, we created such database in panel data format and estimated the Fama-French and Carhart models using panel data estimations. The results were qualitatively very similar to the traditional time series approach where portfolio returns were not disaggregated into individual stocks returns (i.e. the estimates of the parameters of the Fama-French and Carhart variables were very similar in terms of value and statistical significance). We do not report those results due to space limitation, but they are available upon request. We would like to thank two anonymous referees for suggesting this interesting idea.

Table 4: Modified Sharpe ratios (MSR) and Standard Deviations (SD) from 2005 to 2016 for the SRI E&RC stocks portfolio (with dividends) and for the total return index versions of benchmark indices.
Notes: 1) The modified Sharpe ratio was calculated based on the formula from Israelsen (2005): MSR = ER/SD (ER/absER) , where ER is excess return defined as mean monthly difference between the portfolio (or index) return and risk-free return computed for n equal to 12, 60 or 132 months, respectively, and SD is the sample standard deviation of the monthly differences of returns. 2). Bold numbers indicate positive MSR figures. 3) Cells highlighted in grey identify the portfolio or index with the highest MSR ratio for that period. 4) Single-year period covers 12 months from 1 st February to 31 st January. 5) Multiple-year period covers five consecutive single-year periods.

Open Access Article
These sectoral differences in performance, as well as the findings from the previous section about the statistical significance of the crude oil returns variable, prompted us further to examine the performance of the SRI E&RC stocks divided into two broader groups: oil related companies and non-oil related companies.
The selection of stocks to these two groups was based on companies that are: (1) energy and resource stocks which are largely oil related and (2) energy and resource stocks which are not oil related. In the case of the former group, companies from the mining industry, oil and gas industry and oil equipment, services and distribution industries were chosen. The companies in the latter group were formed from industries such as alternative energy, electricity and gas, water and multi-utilities.
The results depicting performance of the oil related companies and non-oil related companies are reported in Tables 7a -7d and they reveal very interesting additional patterns.
Tables 7a and 7b present the returns for the whole 11-year period (February to January) from 2005 to 2016 for the oil related companies stocks portfolio (with dividends), for the non-oil related companies stocks portfolio (with dividends) and for the total return versions of the benchmark indexes.
A direct comparison of the average annual return for the whole sample period from 2005 to 2016 reveals a striking result: an investment in the oil related stocks portfolio would have led to the average annual loss equal to -4.27%, while the non-oil related stocks portfolio would have delivered average annual profit equal to 4.61%. The result of the oil related stocks portfolio was consistently worse than all the benchmark indices returns, whereas the result of the non-oil related stocks portfolio was consistently better than all the benchmark indices returns (which were: 3.64%, 2.44%, 1.13% and 0.43%).
Moreover, the oil related stocks recorded only 5 positive returns out of all 11 annual sub-periods and 2 positive returns out of 7 in the 5-year long sub-periods. On the other hand, the non-oil related stocks recorded 8 positive returns out of 11 annual sub-periods and 5 positive returns out of 7 in the 5-year long sub-periods.
Similar picture emerges from Tables 7c and 7d presenting the values of the modified Sharpe ratios for the oil related companies stocks portfolio (with dividends), for the non-oil related companies stocks portfolio (with dividends) and for the total return versions of the benchmark indexes. The average annual modified Sharpe ratios for the group of oil related stocks for the entire period from 2005 to 2016 is negative and equal to -0.01, while for the group of non-oil related stocks it is positive and equal to 0.18.
The results for the variant of the SRI E&RC stocks portfolio without dividends for the oil related companies stocks portfolio, for the non-oil related companies stocks portfolio and for the price index versions of the benchmark indexes are presented in the Appendix in Tables B1 -B4 and they show similar patterns as those discussed above in this section. 10 10. Furthermore, these results provide additional interesting evidence regarding the impact of dividends on portfolio performance. The average annual return of the oil related stocks portfolio is -7.00%, while the version with dividends achieved -4.27%, which means a difference of -2.37%. However, in the case of the non-oil related stocks portfolio, the average annual return is 0.57% while its version with dividends achieved 4.67%, which means a much larger difference of -4.04%. This comparison shows that dividends mattered considerably more in case of the non-oil related companies than for oil related companies.

Table 7a: Average annual returns for the 11-year period (February to January) from 2005 to 2016 for oil related companies stocks portfolio (with dividends) and for total return versions of the benchmark indexes: 1) Global broad market (S&P Global 1200), 2) Global energy market (MSCI World Energy), 3) Global SRI market (FTSE4GOOD) and 4) Global alternative energy market (FTSE ET50).
Notes: 1) *** -means statistical significance at the 1% level, ** -means statistical significance at the 5% level, * -means statistical significance at the 10% level.
2) The t-statistic was calculated based on the paired difference test. 3) Bold numbers indicate positive figures. 4) Cells highlighted in grey identify the portfolio or index with the highest average annual return for the analysed period.
Open Access Article Table 7b: Average annual returns for the 11-year period (February to January) from 2005 to 2016 for non-oil related companies stocks portfolio (with dividends) and for total return versions of the benchmark indexes: 1) Global broad market (S&P Global 1200),
2) The t-statistic was calculated based on the paired difference test. 3) Bold numbers indicate positive figures. 4) Cells highlighted in grey identify the portfolio or index with the highest average annual return for the analysed period.

Table 7c: Modified Sharpe ratios (MSR) and Standard Deviations (SD) from 2005 to 2016 for oil related companies stocks portfolio (with dividends) and for the total return versions of the benchmark indexes.
Notes: 1) The modified Sharpe ratio was calculated based on the formula from Israelsen (2005): MSR = ER/SD (ER/absER) , where ER is the excess return defined as mean monthly difference between the portfolio (or index) return and the risk-free return computed for n equal to 12, 60 or 132 months, respectfully, and SD is the sample standard deviation of the monthly differences of returns.

MSR) and Standard Deviations (SD) from 2005 to 2016 for the non-oil related companies stocks portfolio (with dividends) and for the total return versions of the benchmark indexes.
Notes: 1) The modified Sharpe ratio was calculated based on the formula from Israelsen (2005): MSR = ER/SD (ER/absER) , where ER is the excess return defined as mean monthly difference between the portfolio (or index) return and the risk-free return computed for n equal to 12, 60 or 132 months, respectfully, and SD is the sample standard deviation of the monthly differences of returns.

Impact of Crude Oil Price on Performance of SRI E&RC Stocks Portfolio
The results presented in previous sections clearly point towards the existence of a relationship between the SRI E&RC stocks portfolio financial performance and the dynamics of oil price returns. Therefore, in this section we specifically focus on this issue and we investigate it deeper by trying to answer the question how the crude oil price movements affect the returns of our SRI E&RC stocks portfolio. We also provide evidence regarding how this relation evolved over time. Table 8a presents the estimation results of parameters from model (9) for the SRI E&RC stocks portfolio returns with the crude oil return as the explanatory variable. It shows that in the full period the estimate for the crude oil returns is positive and it equals 0.3004 (significant at the 1% level). Moreover, the estimates of this parameter are also always significant in all sub-samples and they are consistently significant in all 5-year sub-periods and in all single-year periods.
A closer inspection of the evolution of the estimates for crude oil returns across all 5-year sub-periods proves that they increased from the level of 0.2053 in 2005-2010 to 0.4862 in 2009-2014 and then declined to 0.2634 in the last period 2011-2016. This pattern confirms the results reported and discussed earlier indicating the importance of the crude oil overall and it illustrates its declining role over time towards the end of our sample period. Table 8b presents the estimation results of parameters from model (10) for the SRI E&RC stocks portfolio excess returns (defined as the difference between the SRI E&RC stocks portfolio returns and the returns of the world market index) with the crude oil return again as the explanatory variable. It shows that, similarly to Table 8a, the crude oil returns are also statistically significant and positive in the entire sample period as well as in most sub-periods. Furthermore, Table 8b reveals the same effect as previously detected and discussed, i.e. that the crude oil price importance weakened at the end of the analysed sample period. The estimates of crude oil return variable were quite stable in the first six 5-year long sub-periods from 2005-2010 to 2010-2015 at the level between 0.09 and 0.13 (estimates significant in all these cases), but subsequently they dropped to 0.05 (estimate not significant) in the last sub-period 2011-2016.
This finding for the SRI E&RC stocks portfolio excess returns means that the crude oil price directly affects not only the returns of the stocks of SRI energy and resource companies, but it has also an impact on their returns measured relative to the general market conditions (as captured by the world stock market index). This is an important conclusion from this study. The relationships discussed in this section are shown in Figure 1.

Relevance of the Crude Oil Price for Performance of Oil Related Stocks and Non-oil Related Stocks from the SRI E&RC Portfolio
Finally, in this last section we investigate the performance of stocks divided into two groups, i.e. oil related and non-oil related companies, and the relationship between their returns (referred to henceforth as: oil R pt and non-oil R pt , respectively) and the crude oil returns as the explanatory variable. Table 9a reports the estimated parameters of models for the oil related stocks. The estimate for the whole period of the crude oil return variable is positive and equal to 0.4318 (significant at 1% level). In the 5-year periods the crude oil return increases from 0.4211 to 0.5604 and then drops at the end of the sample to 0.3070. This pattern of estimates is similar also in case of the models for the non-oil related stocks presented in Table 9b, but their values are roughly twice as low. The estimate for the whole period of the crude oil return variable is positive, however it is equal to only 0.1897 (significant at 1% level).       Notes: 1) Standard errors are included in brackets. 2) Statistical significance is indicated as: *** significant at 0.01 level, ** significant at 0.05 level and * significant at 0.1 level. 3) Sample size is reported as the number of months in the respective samples. 4) All regressions are based on time series models. Guide to estimation methods: OLS = Ordinary Least Squares, ARCH = AutoRegressive Conditional Heteroscedasticity and GARCH = Generalised AutoRegressive Conditional Heteroscedasticity. 5) The reported diagnostic tests include the value of the F-test statistic, the value of the Ljung-Box Q statistic with 10 lags as the test for autocorrelation and the value of the LM statistic with 10 lags as the test for any remaining ARCH effects (their respective p-values are reported in brackets).
In the 5-year periods it displays the same pattern, but it increases from 0.2257 to 0.3266 and then drops at the end of the sample period to just 0.1425. These findings show that the movements in the crude oil price had more influence on the performance of oil related stocks rather than on the non-oil related stocks, which is not surprising. However, the results in Tables 9a and 9b allow us also to measure the magnitude of this difference: it appears that crude oil price returns are related twice as strongly to the performance of the SRI E&RC oil related stocks than to the performance of the SRI E&RC non-oil related stocks. 11

CONCLUSIONS
The main objective of this study was to examine whether the performance of the stocks of the SRI energy and resource firms is superior relative to the major benchmarks and whether portfolios composed of such companies can outperform the market.
We first calculated the raw returns and assessed the performance of the portfolios relative to the broad, energy sector, SRI and alternative energy market indices. We report that in the entire 11-year period (February 2005 -January 2016) the annual average performance of the SRI E&RC stocks portfolio was neither consistently superior nor consistently inferior compared to the corresponding returns of all the benchmark indices. Overall, we found that the market does not penalize or reward the energy and resource companies for adopting the SRI practices 12 , however their performance relies heavily on the changes in the crude oil price.
When the entire sample is divided into oil related stocks and non-oil related stocks, we discovered that an investment in the oil related stocks portfolio would have led to the average annual loss of -4.27%, while the non-oil related stocks portfolio would have delivered average annual profit of 4.61%. The performance of the oil related stocks portfolio was consistently worse than all the benchmark indices returns, whereas the result of the non-oil related stocks portfolio was consistently better than all the benchmark indices.
Another important finding from our study is that the dividends mattered quite a lot for the analysed SRI energy and resource stocks portfolios and the individual stocks, because their inclusion in the calculation of the total returns substantially increased their performance.
The analysis of models of the SRI E&RC stocks portfolio returns with crude oil return as the explanatory variable shows that in the full sample period the estimate for crude oil returns is positive and significant at 1% level. However, across all 5-year sub-periods the estimate of crude oil return variable first increased between 2005-2010 and 2009-2014 periods and then declined in the last period 2011-2016. This pattern confirms our other results indicating the importance of the oil price, but also illustrates its declining role over time towards the end of our data sample. 11. It is noteworthy that the literature has not yet extensively focused on the effects of oil price changes on firm-level stock returns, which would allow for a more in-depth analysis (given that firms within the same sector naturally exhibit heterogeneous responses to oil price changes). The lack of empirical evidence in this area is related to the fact that there exist very few studies that rely on the firm-level portfolios construction (see a review by Degiannakis et al. (2014)). Only a very limited number of papers report results based on firm-level data. They show that individual firms' stock returns respond to changes in oil prices (see Boyer and Filion (2007), Scholtens and Wang (2008), Narayan andSharma (2011) or Tsai (2015)), yet the evidence about the nature, magnitude and variation of these reactions is scarce. Hence, our paper also contributes to this particular new line of literature by addressing the gap in research on the effects of oil price changes on company-level stock returns, which was identified in a review paper by Degiannakis et al. (2014).
12. Given that firms chase multiple objectives and the maximisation of their share price (or maximisation of shareholder value) is just one of them, the situation within which the companies analysed in this study are not being penalised for adopting the SRI principles may already be sufficiently satisfactory for them if they gain, indeed, social acceptance etc.

Open Access Article
In the models with the excess returns, we also found similar effects, which implies that crude oil price directly affects not only the returns of the SRI energy and resource companies stocks, but it has also an impact on their returns measured relative to the general market conditions (as captured by the international stock market index). This is another important conclusion from this study.
Our findings also provide evidence that the movements in the crude oil price had more influence on the performance of oil related stocks rather than on the non-oil related stocks. Furthermore, we measured the magnitude of this difference: it appears that crude oil price returns are related twice as strongly to the performance of the SRI E&RC oil related stocks than to the performance of the SRI E&RC non-oil related stocks.
Finally, our analysis shows that the group of SRI energy and resource companies from the Global-100 list in the 11-year period 2005-2016 has been limited to 19 countries of origin including 16 developed nations. This indicates that in many emerging economies, where production and consumption of energy and natural resources are substantial and steadily growing, the SRI related criteria are yet to be fulfilled by the firms from these countries.
The findings from this study have broad important policy implications for financial market regulators and environmental protection agencies in addition to the investors who allocate their funds in energy and resource companies stocks (including alternative energy firms). They also should raise awareness among stock market investors to mobilise capital in more sustainable ways and, possibly, to channel it towards more sustainable methods of energy production.

APPENDIX
Results in the Appendix depict the returns for the individual sub-sectors within the entire SRI E&RC stocks portfolio as well as the returns for the oil related companies (without dividends) and the non-oil related companies (without dividends). Table A1 presents the returns for the whole 11-year period (February to January) from 2005 to 2016 for the individual sub-sectors within the SRI E&RC stocks portfolio (with dividends).
Tables B1 -B4 present the returns for the whole 11-year period (February to January) from 2005 to 2016 for the oil related companies stocks portfolio (without dividends) and the nonoil related companies stocks portfolio (without dividends) and for the price index versions of the benchmark indexes: 1) Global broad market (S&P Global 1200), 2) Global energy market (MSCI World Energy), 3) Global SRI market (FTSE4GOOD) and 4) Global alternative energy market (FTSE ET50).
Open Access Article Table B1: Average annual returns for the 11-year period (February to January) from 2005 to 2016 for oil related companies stocks portfolio (without dividends) and for price index versions of the benchmark indexes: 1) Global broad market (S&P Global 1200), 2) Global energy market (MSCI World Energy), 3) Global SRI market (FTSE4GOOD) and 4) Global alternative energy market (FTSE ET50).
2) The t-statistic was calculated based on the paired difference test. 3) Bold numbers indicate positive figures. 4) Cells highlighted in grey identify the portfolio or index with the highest average annual return for the analysed period.  Notes: 1) *** -means statistical significance at the 1% level, ** -means statistical significance at the 5% level, * -means statistical significance at the 10% level.
2) The t-statistic was calculated based on the paired difference test. 3) Bold numbers indicate positive figures. 4) Cells highlighted in grey identify the portfolio or index with the highest average annual return for the analysed period.
Open Access Article Table B3: Modified Sharpe ratios (MSR) and Standard Deviations (SD) from 2005 to 2016 for oil related companies stocks portfolio (without dividends) and for the price index versions of the benchmark indexes.
Notes: 1) The modified Sharpe ratio was calculated based on the formula from Israelsen (2005): MSR = ER/SD (ER/absER) , where ER is the excess return defined as mean monthly difference between the portfolio (or index) return and the risk-free return computed for n equal to 12, 60 or 132 months, respectfully, and SD is the sample standard deviation of the monthly differences of returns. 2) Bold numbers indicate positive MSR figures. 3) Cells highlighted in grey identify the portfolio or index with the highest MSR ratio for that period. 4) Single-year period covers 12 months from 1 st February to 31 st January. 5) Multiple-year period covers five consecutive single-year periods.

Table B4: Modified Sharpe ratios (MSR) and Standard Deviations (SD) from 2005 to 2016 for non-oil related companies stocks portfolio (without dividends) and for the price index versions of the benchmark indexes.
Notes: 1) The modified Sharpe ratio was calculated based on the formula from Israelsen (2005): MSR = ER/SD (ER/absER) , where ER is the excess return defined as mean monthly difference between the portfolio (or index) return and the risk-free return computed for n equal to 12, 60 or 120 months, respectfully, and SD is the sample standard deviation of the monthly differences of returns. 2) Bold numbers indicate positive MSR figures. 3) Cells highlighted in grey identify the portfolio or index with the highest MSR ratio for that period. 4) Single-year period covers 12 months from 1 st February to 31 st