The effect of ESG screening on the performance of
American stock portfolios
Emma Witteveen (S3125092) University of Groningen Master’s Thesis – MSc Finance Focus Area: Sustainable Society
Supervisor: dr. A. Dalò Final Thesis – June 4, 2020
Abstract
1
1. Introduction
Socially responsible investing (SRI) is growing in popularity among investors. They are increasingly aware of the consequences of decisions made by governments and companies on society and the environment and their investment choices reflect this. Investors’ interest in SRI has increased in recent years. According to the US Sustainable Investment Forum, the total US-domiciled assets managed with SRI strategies increased by 38%, compared to 2016, to $12 trillion at the start of 2018. There are different ways to manage an SRI portfolio and this study investigates whether it is possible to achieve higher risk-adjusted returns by focusing on one of the environmental, social, and governance (ESG) dimensions or total ESG scores and using a positive screening method to build a portfolio.
The goal of this study is to answer whether ESG portfolios can result in higher risk-adjusted returns and whether the results are significantly different from each other. Significant different performances in the dimension specific portfolios would show which dimension is most important for risk-adjusted returns and it could give investors a direction for constructing their SRI portfolios. Moreover, it could provide companies that want to improve their ESG performance with a starting point.
From the ASSET4 database, I collect ESG scores for a sample of 3082 American companies and construct portfolios based on the total ESG score and the separate ESG dimensions: environmental, social, and governance. The portfolios are weighted by market value and consist of the best performing companies for each score, with cut-off points ranging from 10% to 90%. The Carhart (1997) four-factor model evaluates the performance of the portfolios. Then, their performance is compared with the other ESG screened portfolios and a benchmark portfolio that receives no screening treatment. The main results are: The ESG screened portfolios do not perform significantly better than the benchmark. The dimension specific portfolios show significant differences in performance and the governance dimension seems the most dominant. Lastly, the governance portfolio can outperform the total ESG score portfolio, showing that focussing on this dimension may be more beneficial to performance than total ESG scores. The results on the differences between the dimension specific portfolios and the total ESG and benchmark portfolios form this paper’s main contribution to existing literature on the influence of SRI on portfolio performance.
The rest of the paper is structured as follows. The literature review summarizes existing literature on the topic of SRI and using ESG screens in portfolio construction. The data section describes the data sources and provides a summary of the sample’s statistics. In the methodology section, the portfolio construction process is explained, as well as the performance evaluation method. The next section covers the empirical results. Lastly, the conclusion summarizes the most important results and key takeaways.
2. Literature review
2.1 Socially responsible investing (SRI)
2 of social and environmental goals with financial goals, such as obtaining a financial return close to the market.
Investors’ interest in SRI has increased in recent years. A study by Apostolakis et al. (2016) shows that Dutch pension beneficiaries would be willing to accept a lower pension for investing in a socially responsible portfolio. If investors are willing to pay for SRI, there must be non-financial motives driving them to do so. Bénabou and Tirole (2001) identify three possible motives for individual social responsibility: intrinsic altruism, material incentives, and social and self-esteem concerns.
The study by Apostolakis et al. (2016) shows that investors are willing to accept a lower pension when they invest in an SRI portfolio. However, does SRI always imply that the investor must compromise on financial performance, e.g. risk and return? Renneboog et al. (2008) provide a literature review for various topics surrounding SRI. When it comes to SRI and financial performance, there are varying results. Renneboog et al. (2008) argue that SRI funds could outperform non-SRI funds, because good social and environmental behaviour implies good managerial quality and SRI screening could reduce the possibility of incurring high costs during bad social or environmental events. When comparing the performance of SRI funds to non-SRI funds, most studies find no significant differences in alphas between the two (see, e.g. Kreander et al. 2005). Bauer et al. (2005, 2006, and 2007) find significant underperformance of US ethical funds, significant outperformance for UK domestic ethical funds, and no significant differences between SRI and non-SRI funds for German, Australian and Canadian domestic funds.
2.2 SRI strategies
Kempf and Osthoff (2007) have found a strategy to obtain significant risk-adjusted returns. They have a different approach than the literature discussed above, because they do not compare the performance of SRI funds to non-SRI funds. Instead, they create their own portfolios based on different screening methods and compare their financial performance. They find that a high-low strategy, buying stocks with high ratings and selling stocks with low ratings, leads to high risk-adjusted returns. They reach the highest risk-adjusted returns, when combining this strategy with the best-in-class screening strategy.
The abovementioned best-in-class strategy is one of the strategies to implement SRI. Sparkes and Cowton (2004) identify the following SRI implementation strategies: best in class, positive/negative selection, and shareholder activism. Best-in-class screening involves screening companies for positive ESG performance relative to industry peers. Shareholder activism is the active involvement of a shareholder to change a company’s practices. A positive screening strategy involves including the best performing companies and a negative screening strategy involves excluding the worst. Another strategy is ESG integration (Kotsantonis et al., 2016), which focuses on using information regarding a company’s environmental, social and governance (ESG) performance indicators in making investment decisions.
3 than 5% of the firms are excluded. Only the governance portfolio achieves significantly better performances up until a cut-off point of 20%.
2.3 ESG dimensions
When it comes to the differences between the three dimensions in portfolio construction, there are not many studies that compare them, besides the previously mentioned study by Auer (2016). However, there are plenty of studies that consider the dimensions separately and look at specific subdimensions. The following sections concern research on the dimensions and their relation with financial performance.
2.3.1 Environmental dimension
Regarding the environmental dimension, there is evidence that portfolios constructed with a focus on environmental responsibility perform similarly to normal portfolios. Climent and Soriano (2011) evaluated the performance of US environmental mutual funds between 2001 and 2009 and found that they did not perform significantly better or worse than conventional funds. They did report a significantly worse performance for the green funds when considering a larger time frame between 1987 and 2009. Chang et al. (2012) find that green mutual funds perform worse than comparable conventional funds. Other research on the environmental dimension explores the relation between environmental performance and firm performance. There are several reasons that explain a positive relation between the two. Firstly, making production processes more efficient and reducing the amount of necessary resources can make the process more environmentally friendly as well as cost effective (Epstein and Roy, 2001). Bostian et al. (2016) found a positive relation between environmental investments in energy efficiency, environmental performance, and overall productivity. Furthermore, improving waste management practices decreases the strain on the environment and can provide financial benefits through cost reduction and the elimination of regulatory fines (Reinhardt, 1999). El Ghoul et al. (2018) also found lower costs of capital financing among firms with responsible environmental practices. Lastly, good environmental performance has a positive effect on a company’s reputation and Doh et al. (2010) found that companies with a reputation for good environmental performance experience less negative impact on their share price after a negative environmental event.
2.3.2 Social dimension
4 performance is linked to financial performance. For example, focussing on employee relations, layoffs are negatively related to returns (Chen et al., 2001), good social performance improves a company’s attractiveness to future employees, potentially improving a company’s workforce quality (Turban and Greening, 1997), and poorly managed pension liabilities are reflected in lower debt ratings, increasing the cost of debt for companies with this problem (Carroll and Niehaus, 1998).
2.3.3 Governance dimension
For the governance dimension, there is evidence that screening companies for their governance policies in portfolio construction can lead to outperforming comparable conventional portfolios. Auer (2016) finds that the governance selection portfolio is the only dimension specific portfolio that consistently outperforms the benchmark portfolio for all cut-off percentages between 5% and 20%. Gompers et al. (2003) use a high-low strategy, buying stocks from the best governance category and selling stocks from the worst governance category, and realize risk-adjusted returns of 8.5% per year. This indicates that there should be a clear relation between corporate governance and firm performance. Bhagat and Bolton (2008, 2019) emphasize the importance of director stock ownership as it is positively linked to future performance and it allows for replacing management in poorly performing companies. They suggest that board stock ownership provides more incentives to effectively monitor management and thus results in improved performance.
2.4 Hypotheses
The following hypotheses are derived from the previously discussed literature. The goal of this study is to answer the question whether ESG screening and/or the dimension specific focus can lead to significantly higher risk-adjusted returns. Thus, the first hypothesis is:
H1: Applying ESG screening in portfolio construction achieves higher risk-adjusted returns than no screening.
Secondly, this study investigates whether the ESG dimension specific portfolios are significantly different from each other in terms of performance to see if one of the dimensions could be more important than the others. Thus, the second hypothesis is:
H2: The ESG portfolios are significantly different from each other in performance. Lastly, I also examine whether using a total ESG score to base the screening is better than using the best performing dimension’s score. Based on the idea that a total ESG score could dilute the effect of the most important dimension, the third hypothesis is as follows:
5
3. Data
The portfolio composition is based on companies’ scores for the environmental dimension, the social dimension, and the governance dimension. The environmental score captures a company’s impact on the environment and its ability to manage environmental risks and opportunities. The social score captures a company’s relationship with its employees and other stakeholders, and it reflects its reputation in society. The governance score measures a company’s ability to align the interests of board members, executives, and shareholders. Data on companies’ ESG performance are retrieved from the Thomson Reuters ASSET4 ESG database. Thomson Reuters retrieves information from annual reports, sustainability reports, NGO’s and other news sources when assigning ESG scores. The database contains ESG information on close to 9000 companies and has worldwide coverage. The first ratings date back to fiscal year 2002. This study covers the ratings between 2002 and 2017. The companies used for this study are American companies which are included in the full ASSET4 universe list. The list consists of 3228 companies. Both active and dead companies are included to avoid survivorship bias. From the 3228 companies, 146 are excluded from this study’s investment universe, because they have zero observations for the ESG scores or the excess returns during the studied period. This leaves an investment universe of 3082 companies.
After the portfolios are constructed, their financial performance is evaluated and compared using the Carhart (1997) four-factor model, which is described in detail in the methodology section. For all companies included in the sample, I retrieve the monthly return index from the Thomson Reuters Datastream database. The return index includes the effect of reinvestment of dividends and is therefore used to measure financial performance. The Carhart (1997) four-factor model requires information on the risk factors, risk-free rate and performance of the market portfolio, and the performance of the size, book-to-market value, and momentum factor portfolios. The data on these variables are retrieved from Kenneth R. French online data library1.
Excess returns are computed using the return index and risk-free rate. Each firm has 192 returns observations, covering the period between July 2003 and June 2019. Table 1 shows summary statistics on the mean and standard deviation of the sample’s ESG scores and excess returns. The sample is quite unbalanced, mostly regarding the environmental scores, as can be seen from the number of observations (N). I did not exclude companies from the sample for not having (enough) observations for one or two of the ESG scores. This would exclude companies that have been added to the database recently and it could also bias the sample towards certain industries, as some firms are only rated for the ESG scores that are relevant for their business. For example, the sample contains firms from the services industry that do not receive an environmental score.
Firstly, the environmental scores range between 0.27 and 89.73. On average, the sample has a mean environmental score of 25.83. The standard deviation of the mean shows that firms differ quite a lot in their means. Looking at the standard deviation’s characteristics in panel B of table 1, for some firms, the environmental score does not change at all, whereas others experience large jumps. The mean social scores range from 1.18 to 92.04 and average at 33.43. The mean standard deviation of the social scores shows that firms tend to vary less in
6 their social scores than their environmental scores. The mean governance scores have a higher average than the environmental and social scores, indicating that companies in the sample perform better on corporate governance issues than environmental and social issues. The governance scores range between 0.36 and 97.76 and have a mean of 39.97. The mean standard deviation shows that companies’ governance scores change more than the environmental and social scores. Next, the combined ESG scores range between 2.02 and 87.38 and have a mean of 33.99. The mean standard deviations show that combining the individual ESG scores leads to a lower overall standard deviation of the total ESG score. The positive skewness indicates that the scores’ distributions have fatter right tails. The mean environmental and governance scores report a kurtosis statistic lower than three, which indicates that there are fewer extreme values in both ends of the distribution than for a normal distribution. The mean social and ESG scores have kurtosis statistics above three, which means that the ends of the distribution contain more extreme values than a normal distribution.
The mean excess returns range from -3.40% to 6.09%, and on average, they are 1.18%. The standard deviations range between 1.91% and 58.80% and they average at 11.60%. The skewness of the excess returns shows that the distribution is skewed to the right. The kurtosis exceeds 3, which indicates that there are some extreme observations in both tails of the distribution, even after the outliers in the bottom 1% and top 1% for mean returns have been removed from the sample. These statistics suggest that the excess returns are not normally distributed.
Table 1 Descriptive statistics - Excess returns and ESG scores
This table summarizes the mean and standard deviations of the environmental scores, social scores, governance scores, ESG scores and excess returns of the sample with their mean, median, minimum/maximum value, and standard deviation. The outliers in the bottom 1% and top 1% in mean returns were removed from the sample. The skewness and kurtosis statistics describe the distribution of the scores/returns. The last column shows the number of observations per variable. Panel A describes the means and panel B describes the standard deviations.
Variable Mean Median Min Max St. Dev. Skewness Kurtosis N (Stocks) Panel A: Mean Environmental score 25.83 20.45 0.27 89.73 20.56 0.84 2.78 1750 Social score 33.43 31.24 1.18 92.04 15.45 0.70 3.37 3018 Governance score 39.97 39.58 0.36 97.76 19.33 0.11 2.24 3020 ESG score 33.99 32.65 2.02 87.38 13.47 0.66 3.52 3020 Excess Returns 1.18 1.12 -3.40 6.09 1.09 0.11 6.39 3020
Panel B: Standard deviations
7
4. Methodology
To construct the different portfolios, positive screening is applied based on the ASSET4 ESG scores. Positive screening involves selecting the best performing companies in the portfolio construction process. In total, there are five different portfolio categories: environmental, social, governance, total ESG, and benchmark. The environmental portfolios are screened based on companies’ environmental scores. The social portfolios are screened based on companies’ social scores. The governance portfolios are screened based on companies’ governance scores. The total ESG portfolios are screened based on companies’ general ESG scores. This portfolio is included to see if including all dimensions and screening based on a general score, dilutes the effect from one of the dimensions. I decide not to use ASSET4’s total ESG score, in which the dimensions are weighted according to their number of included indicators. Instead, the total ESG score is an equally weighted average of the environmental, social and governance score. This way, no dimension is assigned a higher importance in the overall ESG score. The benchmark portfolio is not screened and is constructed to compare with the other portfolios.
The construction of the portfolios occurs as follows. For example, the environmental portfolio is constructed by first ranking the yearly assigned environmental scores. Then, the stocks in the top 10% are included in a value weighted portfolio. Every year, the portfolios are rebalanced based on the new environmental scores for that year. The social, governance and total ESG portfolios are constructed in the same way. Furthermore, the portfolio construction is repeated for different cut-off points, so there are also portfolios where the top 20%, 30%, 40%, 50%, 60%, 70%, 80% and 90% are included. The benchmark portfolio is constructed as a value weighted portfolio of all the firms in the sample.
The portfolios’ performance is evaluated using the Carhart (1997) four-factor model, which extends the Fama-French (1993) three-factor model with a momentum factor. This model controls for the effect of market risk, size, book-to-market, and momentum. The following regression is used to evaluate the performance of the different portfolios:
𝑅𝑝𝑡 𝑑 − 𝑅𝑓𝑡 = 𝛼𝑝 𝑑 + 𝛽
1𝑝 𝑑 (𝑅𝑚𝑘𝑡− 𝑅𝑓𝑡 ) + 𝛽2𝑝 𝑑 ∗ 𝑆𝑀𝐵𝑡+ 𝛽3𝑝 𝑑 ∗ 𝐻𝑀𝐿𝑡+ 𝛽4𝑝 𝑑 ∗ 𝑀𝑂𝑀𝑡+ 𝑢𝑝𝑡(1)
The left-hand side of the equation represents the portfolio return as the excess of the return over the risk-free rate. The 𝛼𝑝 𝑑 is the risk-adjusted return of the portfolio p for ESG dimension d.
8
5. Empirical results
5.1 Descriptive statistics portfolios
Table 2 contains descriptive statistics for the monthly excess returns of each value-weighted portfolio for the ESG dimensions and the benchmark portfolio. The portfolios are named such that the first letter(s) describe(s) the dimension and the following numbers point to the cut-off percentage that was used during the positive selection process. For example, E10 is the value-weighted portfolio screened for the environmental score using a cut-off of 10%, and ESG60 is the value-weighted portfolio screened for the total ESG score using a 60% cut-off. The benchmark portfolio is value-weighted but not screened for any of the dimensions.
Panel A of table 2 describes the excess returns of the environmental portfolios with the 10% to 90% cut-off rates. The mean monthly excess returns increase slightly when the selection becomes less strict as the cut-off increases from 10% to 90%. Overall, the monthly excess returns range from -17.43 to 25.72. The skewness statistics show that in all the environmental portfolios, the monthly excess returns are skewed to the left, which suggests that risk is underestimated. The kurtosis statistics show that the distributions have fatter tails than a normal distribution would. Combined, the skewness and kurtosis statistics show that the portfolios’ returns are not normally distributed. The last column shows the number of stocks that are included in the portfolio during the last construction year. This explains why the benchmark portfolio does not include the total 3082 stocks, as some of the stocks die before the studied period ends. As the environmental dimension has the least amount of stocks with scores assigned, the number of stocks in these portfolios are lower than for their equivalents in the other dimensions.
Panel B of table 2 presents the monthly excess returns for the social dimension portfolios with 10% to 90% cut-off rates. The mean excess return increases from the S10 portfolio to the S40 portfolio, which has the highest average return of all the portfolios. The portfolios with cut-offs from 50% to 90% have slightly lower returns than S40, but higher returns than the portfolios with cut-offs between 10% and 30%. Overall, the excess returns for the social score portfolios range between -17.18 and 24.39. The skewness and kurtosis statistics show that the distribution of the monthly excess returns is skewed to the left, has fat tails and is, thus, not normal. In general, the mean monthly excess returns for the social score portfolios are higher than those for the environmental score portfolio with the same cut-off point.
Panel C summarizes the monthly excess returns for the governance dimension portfolios with 10% to 90% cut-off rates. The mean excess returns do not show a consistently increasing or decreasing pattern as the cut-off point increases. The portfolios with 10% to 50% cut-off rates have higher returns than the portfolios with cut-off rates from 60% to 90%. The G40 portfolio has the highest mean excess return, followed by the G20 portfolio. Overall, the excess returns range between -17.38 and 23.62. Again, the distribution of the returns is not normal. The mean monthly returns for the governance score portfolios are higher than those for the equivalent environmental and social portfolios, except for the G80 portfolio, which has a lower mean excess return than S80, but a higher mean excess return than E80.
9 has a slightly lower mean excess return than ESG80. The skewness and kurtosis statistics indicate that the means are not normally distributed. The mean excess returns are lower than those from the equivalent governance portfolios, except for ESG80, and some of the equivalent social portfolios, but they are higher than those from the equivalent environmental portfolios. This indicates that there might be a dimension that is more dominant in generating returns and its effect is diluted when combined with the other scores in the total ESG score.
Lastly, the benchmark portfolio’s statistics are described in panel E. They are similar to the statistics for the other portfolios. Again, the excess returns are not normally distributed. When it comes to the portfolio’s mean return, it is higher than the mean returns of all the environmental and some of the social and ESG portfolios’ returns. The governance portfolios do have consistently higher mean excess returns than the benchmark.
Table 2 Descriptive Statistics – Portfolio excess returns
This table summarizes the excess returns of the constructed portfolios with their mean, median, minimum/maximum value, and standard deviation. The skewness and kurtosis statistics describe the distribution of the returns. The last column shows the number of stocks that are included in the portfolio during the last construction year. The portfolios are named such that the first letters describe the dimension and the following numbers point to the cut-off percentage that was used during the positive selection process. Panel A describes the environmental portfolios, panel B the social portfolios, panel C the governance portfolios and panel D the ESG portfolios. Panel E contains the results for the benchmark portfolio, which was constructed without any screening strategy.
Portfolio Mean Median Min Max St. Dev. Skewness Kurtosis N (Stocks)
Panel A: Environmental portfolios
E10 0.60 1.09 -17.43 25.72 4.76 -0.11 8.38 144 E20 0.66 1.16 -16.14 20.23 4.41 -0.33 6.15 287 E30 0.70 1.25 -16.15 20.11 4.44 -0.37 6.28 433 E40 0.72 1.19 -15.69 19.77 4.43 -0.38 6.09 574 E50 0.75 1.11 -15.37 20.41 4.48 -0.34 6.24 719 E60 0.76 1.23 -15.68 20.40 4.52 -0.35 6.29 863 E70 0.77 1.26 -15.94 20.30 4.55 -0.38 6.31 1007 E80 0.78 1.30 -16.24 20.18 4.60 -0.40 6.32 1162 E90 0.78 1.26 -16.74 20.03 4.61 -0.43 6.36 1295
Panel B: Social portfolios
10 Panel C: Governance portfolios
G10 0.81 1.33 -17.01 23.62 4.58 -0.38 8.07 255 G20 0.83 1.23 -16.49 20.64 4.46 -0.50 6.85 510 G30 0.81 1.27 -16.33 20.24 4.52 -0.49 6.58 767 G40 0.83 1.27 -16.68 20.72 4.57 -0.50 6.82 1021 G50 0.81 1.25 -16.40 20.59 4.59 -0.45 6.63 1277 G60 0.80 1.18 -16.63 20.23 4.60 -0.45 6.53 1533 G70 0.78 1.19 -16.79 20.22 4.63 -0.44 6.48 1789 G80 0.78 1.21 -17.13 19.86 4.64 -0.47 6.46 2044 G90 0.79 1.22 -17.38 20.13 4.68 -0.47 6.53 2296
Panel D: ESG portfolios
ESG10 0.71 1.21 -16.53 22.30 4.46 -0.33 7.34 255 ESG20 0.75 1.19 -16.03 21.52 4.47 -0.37 6.95 510 ESG30 0.76 1.19 -15.85 20.90 4.51 -0.37 6.58 767 ESG40 0.76 1.13 -16.58 20.76 4.57 -0.40 6.63 1021 ESG50 0.77 1.17 -16.56 20.48 4.59 -0.42 6.53 1277 ESG60 0.78 1.16 -16.61 20.50 4.61 -0.42 6.55 1533 ESG70 0.78 1.18 -16.70 20.36 4.62 -0.42 6.51 1789 ESG80 0.79 1.21 -16.93 20.15 4.63 -0.45 6.48 2044 ESG90 0.79 1.18 -17.27 20.05 4.67 -0.46 6.48 2296
Panel E: Benchmark portfolio
Benchmark 0.78 1.16 -17.17 19.45 4.53 -0.45 6.78 3008
5.2 ESG portfolio performance: the Carhart (1997) four-factor model
Table 3 contains the results for the Carhart (1997) four-factor model for portfolio performance for the portfolios based on positive screening for total ESG scores with cut-off points between 10% and 90% and the benchmark portfolio.
All ESG portfolios show positive risk-adjusted returns that, consistently with the descriptive statistics from table 2, tend to increase as the cut-off point increases from 10% to 90%. However, they are not significantly different from zero, so the screening method for these portfolios does not lead to significant risk-adjusted returns. The benchmark portfolio has a higher risk-adjusted return than any of the ESG portfolios, but this result is also insignificant.
The second column reports the market risk coefficients, which indicate the level of volatility of the portfolio compared to the market. The ESG10 and ESG20 portfolios report significant market risk coefficients that are below one, which suggests that the portfolios are less volatile than the market. The benchmark portfolio also has a market beta lower than one. For the ESG portfolios with cut-off points from 30% to 90%, the market beta is larger than one, and the portfolios are thus more volatile than the market.
11 portfolios with more lenient cut-off rates contain relatively less large capitalization stocks. As the portfolios are value-weighted, the portfolios are generally biased towards large capitalization stocks. The fourth column shows the coefficients for the book-to-market, HML, factor. The HML coefficients are positive for all portfolios, though only significant for the ESG10 and ESG20 portfolios. This means that these portfolios contain relatively more value stocks than growth stocks and the excess returns are partially explained by the value premium associated with the value stocks. As the cut-off points increase, the HML coefficients decrease, indicating that the value stocks are represented less in these portfolios. The coefficients for the momentum factor, MOM, are displayed in the fifth column. All portfolios have a negative coefficient for the momentum factor, but they are only significantly for the portfolios with cut-off points from 50% to 90% and the benchmark portfolio. The negative coefficient for the momentum factor indicates a negative correlation between current returns and past returns. In other words, when these portfolios perform bad, they tend to start performing better in the future, which is also described as a reversal effect.
Table 3 already shows some differences between the portfolios that are positively screened for their ESG score and the benchmark portfolio in terms of performance and composition. The next section explores these differences with a regression of difference portfolios in which each ESG portfolio is compared to the benchmark portfolio.
Table 3 Carhart (1997) four-factor model – ESG portfolios
This table summarizes the results from the Carhart (1997) four-factor model for portfolios constructed with positive screening on total ESG score. The portfolios are named such that the first letters describe the dimension and the following numbers point to the cut-off percentage that was used during the positive selection process. The last row contains the results for the benchmark portfolio, which was constructed without any screening strategy. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
ESG10 0.032 0.989*** -0.288*** 0.161** -0.061 0.795 ESG20 0.052 0.999*** -0.216*** 0.114* -0.064 0.810 ESG30 0.048 1.017*** -0.183*** 0.071 -0.072 0.823 ESG40 0.045 1.022*** -0.160** 0.066 -0.086 0.824 ESG50 0.053 1.023*** -0.139** 0.055 -0.092* 0.827 ESG60 0.055 1.025*** -0.124* 0.052 -0.096* 0.829 ESG70 0.058 1.026*** -0.111 0.043 -0.102** 0.831 ESG80 0.064 1.028*** -0.098 0.039 -0.102** 0.833 ESG90 0.057 1.033*** -0.085 0.033 -0.106** 0.834 Benchmark 0.075 0.980*** -0.007 0.035 -0.113** 0.831 Significant at: *=0.1, **=0.05, ***=0.01
5.3 ESG screening vs. Benchmark
12 portfolios are compared through difference portfolios to see whether there are any significant differences. Table 4 shows the results from the Carhart (1997) four-factor model for the difference portfolios. Before the Carhart (1997) model estimation, the difference portfolios are constructed as the excess returns of the benchmark minus the excess returns of the ESG portfolio.
The first column displays the difference portfolios’ alphas. None are significant, which means that there are no significant differences between the ESG portfolios and the benchmark portfolio in terms of risk-adjusted returns. This result rejects the first hypothesis that portfolios with ESG screening achieve higher risk-adjusted returns than portfolios without screening.
Looking at the market beta, there are significant differences between the benchmark portfolio and some of the ESG portfolios. The significant negative coefficients for the difference portfolios comparing the benchmark to the ESG portfolios with cut-off points from 30% to 90% show that the benchmark portfolio is less exposed to market risk than these portfolios. The ESG portfolios do not display the reduction in market risk that has been observed in previous research by Albuquerque et al. (2019), which presents ESG as a product differentiation strategy that can decrease systematic risk as ESG levels increase.
The next column displays the coefficients for the size factor for the difference portfolios. All difference portfolios have a significant positive coefficient. In terms of portfolio composition, the ESG screened portfolios are weighted significantly more towards large capitalization stocks than the benchmark portfolio. This contradicts previous literature in which SRI portfolios are often observed to be biased towards small capitalization companies and tend to display higher exposure to the size factor compared to non-SRI portfolios (Geczy et al., 2006). On the other hand, Drempetic et al. (2019) find a positive relation between firm size and the ESG scores assigned by Thomson Reuters ASSET4, which would explain the large portion of large capitalization stocks in the ESG portfolios.
Table 4 Difference portfolios - Benchmark vs. ESG
This table summarizes the results from the Carhart (1997) four-factor model for the difference portfolios that compare the portfolios constructed with positive screening on total ESG score to the benchmark portfolio. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
13 Next, the book-to-market coefficients are significantly negative for the difference portfolios including ESG portfolios with cut-off rates from 10% to 40%, which means that these ESG portfolios contain significantly more value stocks than the benchmark. ESG companies tend to have lower book-to-market ratios due to a higher demand for their stocks, so ESG portfolios tend to show a lower exposure to the book-to-market factor than non-ESG portfolios (Galema et al., 2008). In this case, it is the benchmark portfolio that has a lower exposure to the HML factor and is relatively less oriented towards value stocks.
Lastly, the coefficients for the momentum factor show a significant difference between the benchmark portfolio and the ESG portfolios with cut-off points ranging from 10% to 30%. The benchmark portfolio experiences a significantly larger reversal effect regarding its returns than the ESG screened portfolio with 10% to 30% cut-off points.
To summarize, the difference portfolios show that there are indeed significant differences between the ESG screened portfolios and benchmark. However, these differences are not observed in terms of risk-adjusted returns, and thus the first hypothesis is rejected. The practical implications for investors are that they can introduce SRI into their portfolios with positive screening for ESG scores without having to compromise on risk-adjusted returns. However, the ESG portfolios may have a higher exposure to the risk factors.
5.4 Dimension specific portfolio performance: the Carhart (1997) four-factor model
Table 5 contains the results of the Carhart (1997) four-factor model for portfolio performance for the portfolios based on positive screens for environmental, social and governance scores, and the benchmark portfolio.
For the environmental portfolios (Table 5, Panel A), the portfolios with cut-off rates between 30% and 90% have positive alphas, and the E10 and E20 portfolios have negative alphas. However, these risk-adjusted returns are not significantly different from zero. Compared to the benchmark, the environmental portfolios obtain lower alphas. Regarding the market risk, all environmental portfolios have a market beta that exceeds one, which indicates that these are more volatile than the market. All environmental portfolios have significant negative exposure to the size factor, which means that they contain relatively many stocks with high capitalization. Only the E10 portfolio has significant exposure to the book-to-market portfolio, which suggests that this portfolio is weighted more towards value stocks compared to the other portfolios. The E80’s and E90’s coefficient for the momentum factor shows that they tend to experience a reversal effect in their returns.
14
Table 5 Carhart (1997) four-factor model – Dimension specific portfolios
This table summarizes the results from the Carhart (1997) four-factor model for portfolios constructed with positive screening for environmental, social and governance scores. The portfolios are named such that the first letters describe the dimension and the following numbers point to the cut-off percentage that was used during the positive selection process. Panel A describes the environmental portfolios, panel B the social portfolios and panel C the governance portfolios. Panel D contains the results for the benchmark portfolio, which was constructed without any screening strategy. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
Panel A: Environmental portfolios
E10 -0.092 1.009*** -0.263*** 0.152** -0.097 0.751 E20 -0.035 1.001*** -0.246*** 0.081 -0.052 0.809 E30 0.009 1.000*** -0.241*** 0.052 -0.073 0.806 E40 0.024 1.001*** -0.219*** 0.037 -0.076 0.812 E50 0.039 1.012*** -0.199*** 0.041 -0.074 0.817 E60 0.048 1.020*** -0.180*** 0.041 -0.075 0.819 E70 0.053 1.025*** -0.163** 0.040 -0.078 0.822 E80 0.053 1.031*** -0.151** 0.039 -0.088* 0.825 E90 0.051 1.034*** -0.144** 0.034 -0.092* 0.827 Panel B: Social portfolios
S10 0.013 1.027*** -0.249*** 0.127* -0.079 0.789 S20 0.038 1.027*** -0.199*** 0.122* -0.059 0.807 S30 0.067 1.017*** -0.180** 0.105 -0.070 0.813 S40 0.084 1.030*** -0.167** 0.077 -0.082 0.821 S50 0.066 1.028*** -0.144** 0.064 -0.091* 0.828 S60 0.060 1.032*** -0.127* 0.058 -0.098* 0.829 S70 0.053 1.033*** -0.109 0.051 -0.101** 0.831 S80 0.057 1.030*** -0.099 0.041 -0.103** 0.834 S90 0.059 1.032*** -0.086 0.031 -0.109** 0.835 Panel C: Governance portfolios
G10 0.124 1.004*** -0.312*** 0.210*** -0.051 0.782 G20 0.138 0.998*** -0.246*** 0.145** -0.062 0.814 G30 0.102 1.016*** -0.195*** 0.102 -0.066 0.822 G40 0.112 1.024*** -0.167** 0.091 -0.075 0.827 G50 0.097 1.021*** -0.139** 0.086 -0.084* 0.827 G60 0.076 1.022*** -0.117* 0.070 -0.091* 0.831 G70 0.057 1.026*** -0.102 0.052 -0.101** 0.832 G80 0.054 1.028*** -0.093 0.044 -0.102** 0.834 G90 0.058 1.034*** -0.085 0.034 -0.106** 0.833 Panel D: Benchmark portfolio
Bench 0.075 0.980*** -0.007 0.035 -0.113** 0.831
15 The momentum factor negatively affects the social portfolios with cut-off points between 50% and 90%, which indicates a reversal effect for their returns.
The results for the governance portfolios are described in panel C of table 5. Consistently with the descriptive statistics from table 2, the governance portfolios show positive risk-adjusted returns. They tend to be higher than those for equivalent portfolios from the environmental and social dimension and even exceed the benchmark portfolio’s risk-adjusted return. However, they are not significantly different from zero. All portfolios, except for G20, have market betas higher than one, which indicates that the portfolios are more volatile than the market. The governance portfolios with cut-off points from 10% to 60% have significant negative coefficients for the size factor, which suggests that they contain relatively more large capitalization stocks. The G10 and G20 portfolios also contain relatively more value stocks, as is indicated by their book-to-market factor exposure. Part of their excess returns are explained by the value premium. The momentum factor negatively affects the governance portfolios with cut-off points between 50% and 90%, which indicates a reversal effect for their returns.
In this section, I have described the performance and composition of the dimension specific portfolios. The next section focuses on comparing these portfolios to the benchmark to test for significant differences.
5.5 Dimension specific screening vs. Benchmark
Table 5 showed that the separate dimension portfolios and the benchmark portfolio differ from each other in terms of performance and exposure to the risk factors. In this section the benchmark and dimension specific portfolios are compared through difference portfolios to see whether there are any significant differences. Table 6 shows the results from the Carhart (1997) four-factor model for the difference portfolios.
Panel A of table 6 compares the benchmark portfolio to the environmental portfolios. All the alphas are positive, which indicates that the benchmark tends to outperform the environmental portfolios in terms of risk-adjusted returns, but this difference is not significant. Regarding market risk, the environmental portfolios with cut-off points between 50% and 90% are significantly more volatile than the market compared to the benchmark portfolio. All difference portfolios have a significant positive coefficient for the size factor. In terms of portfolio composition, the environmental portfolios are weighted significantly more towards large capitalization stocks than the benchmark portfolio. E10 is the only environmental portfolio with a significantly larger amount of value stocks than the benchmark. Portfolios with cut-off points of 20% and between 50% and 90% are significantly less affected by a reversal effect on their returns than the benchmark.
16
Table 6 Difference portfolios - Benchmark vs. dimension specific portfolios
This table summarizes the results from the Carhart (1997) four-factor model for the difference portfolios that compare the portfolios constructed with positive screening for separate dimension scores to the benchmark portfolio. In panel A the benchmark portfolio is compared to the environmental portfolios, in panel B to the social portfolios and in panel C to the governance portfolios. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
Panel A: Benchmark vs. environmental portfolios
Bench-E10 0.167 -0.029 0.256*** -0.117** -0.016 0.139 Bench-E20 0.110 -0.021 0.239*** -0.047 -0.060** 0.257 Bench-E30 0.067 -0.019 0.234*** -0.017 -0.039 0.292 Bench-E40 0.051 -0.021 0.212*** -0.002 -0.037 0.292 Bench-E50 0.036 -0.032* 0.192*** -0.007 -0.039* 0.300 Bench-E60 0.027 -0.040*** 0.173*** -0.007 -0.038* 0.306 Bench-E70 0.022 -0.044*** 0.156*** -0.005 -0.035** 0.318 Bench-E80 0.022 -0.050*** 0.144*** -0.005 -0.025** 0.314 Bench-E90 0.024 -0.054*** 0.137*** 0.001 -0.021** 0.315 Panel B: Benchmark vs. social portfolios
Bench-S10 0.062 -0.047** 0.242*** -0.093*** -0.033 0.260 Bench-S20 0.037 -0.047** 0.192*** -0.088*** -0.053* 0.290 Bench-S30 0.008 -0.036** 0.173*** -0.071** -0.043* 0.307 Bench-S40 -0.009 -0.049*** 0.160*** -0.042* -0.031 0.346 Bench-S50 0.009 -0.047*** 0.137*** -0.029 -0.021 0.340 Bench-S60 0.016 -0.051*** 0.120*** -0.024 -0.015 0.372 Bench-S70 0.022 -0.053*** 0.102*** -0.016 -0.011 0.378 Bench-S80 0.018 -0.050*** 0.091*** -0.006 -0.010 0.372 Bench-S90 0.017 -0.051*** 0.079*** 0.003 -0.003 0.399 Panel C: Benchmark vs. governance portfolios
17 Panel C of table 6 compares the benchmark portfolio to the governance portfolios. Here, the governance portfolios with cut-off points between 10% and 60% seem to outperform the benchmark in terms of risk-adjusted returns, as is indicated by the negative alpha for the difference portfolio. All portfolios with cut-off points between 30% and 90% are more volatile than the market compared to the benchmark. All governance portfolios also contain relatively more large capitalization stocks than the benchmark. Lastly, the governance portfolios with cut-off points between 10% and 60% also contain more value stocks and experience smaller reversal effects on their returns than the benchmark.
The observed significant differences between the benchmark and the dimension specific portfolios are consistent with the differences between the benchmark and the ESG portfolios that were reported in table 4. The ESG-screened portfolios have a higher exposure to market risk than the benchmark, which shows that the focus on ESG does not reduce systematic risk, as is suggested by Alburquerque et al. (2019). Furthermore, the dimension specific portfolios also tend to be focussed more on large capitalization stocks than the benchmark, which is the opposite of what is observed by Geczy et al. (2006). However, the positive relation between ESG scores assigned by Thomson Reuters ASSET4 and firm size (Drempetic et al., 2019) provides a plausible explanation for the focus on large capitalization stocks in the screened portfolios. Lastly, the difference in exposure to the book-to-market factor contradicts previous studies (e.g. Galema et al., 2008, Bauer et al., 2005) that find that SRI portfolios tend to be more oriented towards growth stocks. In this study, the ESG screened portfolios have a higher exposure to the book-to-market factor, indicating that they are relatively more oriented towards value stocks.
Regarding the hypotheses, the lack of significant differences in alphas between the benchmark portfolio and separate dimension portfolios further supports rejecting the first hypothesis that portfolios with ESG screening achieve higher risk-adjusted returns. On the other hand, the analysis above does confirm that the separate dimension portfolios are not the same across equivalents regarding their performance and composition. These results support the second hypothesis that the separate ESG portfolios are different from each other in performance. The next section is aimed at answering the third hypothesis whether total ESG portfolios can be outperformed by some dimension specific portfolios.
5.6 Difference portfolios – Dimension specific portfolios vs. ESG portfolios
The goal of this section is to test the third hypothesis that the total ESG portfolios can be outperformed by dimension specific portfolios. The dimension specific portfolios are compared to each other with difference portfolios to see whether there are any significant differences and to detect if there is a dominant dimension. Table 7 shows the results from the Carhart (1997) four-factor model for the difference portfolios. The difference portfolios are constructed as the excess returns of one dimension specific minus the excess returns of another dimension specific portfolio. Comparing the social and environment portfolios to each other does not give significant differences in risk-adjusted returns, so this difference portfolio is not included in table 7.
18 portfolios obtain significantly higher risk-adjusted returns than the environmental portfolios. Governance portfolios with cut-off rates between 30% and 90% tend to contain less large capitalization stocks. Governance portfolios with cut-off rates between 20% and 60% also have a higher exposure to the book-to-market factor, which shows that these governance portfolios include more value stocks than the equivalent environmental portfolios.
In panel B of table 7, the governance portfolios are compared to the social portfolios. For the portfolios with cut-off rates of 10% and 20%, the governance portfolios obtain significantly higher risk-adjusted returns than their counterparts from the social portfolios. The G10 and G20 portfolio are also weighted more towards large capitalization stocks. Furthermore, the G20 portfolio is significantly less exposed to market risk than S20. The G10 portfolio is also significantly exposed to the book-to-market factor than the S10 portfolio. Beyond the 20% cut-off rate, there are no significant differences between the governance and social portfolios.
Table 7 Difference portfolios - Dimension specific portfolios
This table summarizes the results from the Carhart (1997) four-factor model for the difference portfolios that compare the different dimension specific portfolios. In panel A, the governance portfolios are compared to the environmental portfolios and in panel B to the social portfolios. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
Panel A: Governance vs. environmental portfolios
G10-E10 0.216** -0.006 -0.050 0.058 0.046** 0.013 G20-E20 0.173*** -0.004 0.001 0.064** -0.010 0.024 G30-E30 0.094* 0.017 0.047** 0.050** 0.008 0.068 G40-E40 0.088** 0.023* 0.052*** 0.055*** 0.000 0.133 G50-E50 0.058* 0.009 0.060*** 0.045*** -0.009 0.176 G60-E60 0.028 0.002 0.063*** 0.028*** -0.016 0.206 G70-E70 0.004 0.001 0.061*** 0.013 -0.022*** 0.277 G80-E80 0.001 -0.003 0.058*** 0.005 -0.014*** 0.180 G90-E90 0.007 0.000 0.059*** -0.001 -0.014*** 0.190 Panel B: Governance vs. social portfolios
19 Table 7 shows that the governance portfolios obtain significantly higher risk-adjusted returns than some of their counterparts from the environmental and social portfolios. As the environmental and social portfolios do not significantly outperform each other, the governance dimension seems the most dominant in generating risk-adjusted returns. To test the third hypothesis whether the most dominant dimension can outperform the total ESG portfolios, a last set of difference portfolios is constructed where the governance portfolios are compared to their ESG portfolio counterparts. Table 8 contains the results from the Carhart (1997) four-factor model for these difference portfolios.
In table 8, the difference portfolios with cut-off rates between 10% and 50% show significant positive alphas. This indicates that these governance portfolios outperform their equivalent ESG portfolio on risk-adjusted returns. These governance portfolios also contain a significantly larger amount of value stocks compared to their equivalent ESG portfolios. The results from table 8 support the third hypothesis that portfolios screened for the most dominant dimension, in this case the governance dimension, can outperform corresponding ESG portfolios. This confirms that, in some cases, the total ESG portfolio may dilute the positive effects of screening for the governance dimension.
Table 8 Difference portfolios - ESG vs. Governance portfolios
This table summarizes the results from the Carhart (1997) four-factor model for the difference portfolios that compare the governance portfolios to the ESG portfolios The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
G10-ESG10 0.092* 0.014 -0.024 0.050* 0.010 0.016 G20-ESG20 0.086** -0.002 -0.029* 0.031* 0.002 0.018 G30-ESG30 0.054* -0.000 -0.012 0.031** 0.006 0.009 G40-ESG40 0.067** 0.003 -0.007 0.025** 0.010 0.011 G50-ESG50 0.044** -0.002 0.000 0.031*** 0.009 0.056 G60-ESG60 0.021 -0.003 0.006 0.018*** 0.005 0.026 G70-ESG70 -0.000 0.000 0.009* 0.009* 0.002 0.028 G80-ESG80 -0.010 -0.001 0.005 0.005 0.000 -0.004 G90-ESG90 0.001 0.001 0.000 0.000 0.000 -0.016 Significant at: *=0.1, **=0.05, ***=0.01 5.7 Robustness
20 In terms of alphas, table 9 shows similar results for the screened equally weighted portfolios as table 3 and 5 did for the screened value-weighted portfolios. The alphas are somewhat higher for the equal weighted portfolios, but they are not significant. Only the benchmark portfolio obtains a significant alpha when switching from value-weighting to equal weighting. Next, on average the betas are slightly higher for the equally weighted portfolios than for the value-weighted portfolios. As the larger and more stable companies have less weight in an equally weighted portfolio, this is not a surprising result. The equal weights also change the effect of the size factor on the portfolios. In value-weighted portfolios, large capitalization stocks make up a larger portion of a portfolio, which is shown by the negative SMB coefficients in table 3 and 5. In table 9, the significant coefficients for the size factor are positive, which indicates that the equally weighted portfolios experience the “small firm effect” and the larger amount of small capitalization stocks in the portfolio has a positive effect on excess returns. Regarding the book-to-market factor, the results are not drastically different. In general, the portfolios have a positive coefficient for this factor, indicating that the portfolios contain a relatively large portion of value stocks. The momentum factor holds a lot more significance in the equally weighted portfolios and indicates that the portfolios tend to experience larger reversal effects in returns.
The value weighted ESG screened portfolios do not always align with previous research regarding their exposure to the risk factors. The robustness check shows that this is partially due to the use of market value to weight the portfolios. However, the significance of the risk-adjusted returns is not affected by the chosen weighting method. Therefore, the main conclusion remains that ESG screenings do not result in significant risk-adjusted returns.
Table 9 Carhart (1997) four-factor model - Equally weighted portfolios
This table summarizes the results from the Carhart (1997) four-factor model for portfolios constructed with positive screening for environmental, social, governance and ESG scores. The portfolios in this table are equally weighted. The portfolios are named such that the first letters describe the dimension and the following numbers point to the cut-off percentage that was used during the positive selection process. Panel A describes the environmental portfolios, panel B the social portfolios, panel C the governance portfolios, panel D the ESG portfolios. Panel E contains the results for the benchmark portfolio, which was constructed without any screening strategy. The results include the risk-adjusted return (Alpha), market beta (Mkt-RF), coefficients for the size factor (SMB), book-to-market factor (HML), momentum factor (MOM) and the model's adjusted R2.
Portfolio Alpha Mkt-RF SMB HML MOM R2
Panel A: Environmental portfolios
21 Panel B: Social portfolios
S10 0.126 1.106*** 0.037 0.106 -0.196*** 0.825 S20 0.156 1.089*** 0.113 0.112 -0.191*** 0.835 S30 0.161 1.095*** 0.170** 0.123* -0.207*** 0.845 S40 0.182 1.101*** 0.209** 0.115 -0.215*** 0.843 S50 0.176 1.089*** 0.268*** 0.128* -0.231*** 0.847 S60 0.157 1.096*** 0.292*** 0.128* -0.228*** 0.848 S70 0.157 1.098*** 0.315*** 0.136* -0.220*** 0.851 S80 0.162 1.093*** 0.328*** 0.141* -0.217*** 0.852 S90 0.163 1.092*** 0.342*** 0.134* -0.217*** 0.852 Panel C: Governance portfolios
G10 0.133 1.074*** 0.077 0.160* -0.252*** 0.825 G20 0.150 1.078*** 0.123 0.145 -0.223*** 0.837 G30 0.161 1.095*** 0.170** 0.123* -0.207*** 0.845 G40 0.155 1.079*** 0.234*** 0.170** -0.195*** 0.845 G50 0.169 1.068*** 0.264*** 0.190** -0.186*** 0.846 G60 0.168 1.069*** 0.292*** 0.170** -0.199*** 0.848 G70 0.163 1.082*** 0.311*** 0.164** -0.210*** 0.853 G80 0.167 1.088*** 0.327*** 0.157** -0.209*** 0.854 G90 0.170 1.095*** 0.343*** 0.147* -0.211*** 0.852 Panel D: ESG portfolios
ESG10 0.109 1.081*** -0.037 0.072 -0.211*** 0.835 ESG20 0.144 1.063*** 0.071 0.092 -0.199*** 0.838 ESG30 0.157 1.071*** 0.140* 0.129* -0.204*** 0.841 ESG40 0.164 1.076*** 0.199** 0.145* -0.214*** 0.840 ESG50 0.165 1.081*** 0.247*** 0.150* -0.214*** 0.844 ESG60 0.168 1.079*** 0.282*** 0.159** -0.215*** 0.846 ESG70 0.173 1.078*** 0.308*** 0.152** -0.211*** 0.850 ESG80 0.171 1.085*** 0.326*** 0.145* -0.206*** 0.850 ESG90 0.165 1.090*** 0.343*** 0.142* -0.212*** 0.852 Panel E: Benchmark portfolio
Benchmark 0.440*** 1.016*** 0.623*** 0.216*** -0.188*** 0.851
Significant at: *=0.1, **=0.05, ***=0.01
6. Conclusion
Investors increasingly incorporate SRI into their investment decisions. This paper investigates if investors can increase their portfolios’ performance by implementing positive screening based on ESG ratings. Furthermore, it considers at how the separate ESG dimensions influence performance and whether their effects are different from each other. The goal of this approach is to see if focussing on one of the separate dimensions may be more beneficial to performance than screening for total ESG scores.
22 governance and total ESG scores for cut-off points ranging from 10% to 90%, and they are weighted by market value. The performance of these portfolios is evaluated with the Carhart (1997) four-factor model. They are compared to each other and to an unscreened benchmark portfolio with difference portfolios.
In general, the portfolios that receive positive screening for ESG, environmental, social and governance scores do not generate significant risk-adjusted returns. When I compare them to the benchmark portfolio, there is no significant difference in risk-adjusted returns. Therefore, the first hypothesis that ESG screened portfolios outperform non-screened portfolios is rejected. Regarding their exposure to the risk factors, the screened and non-screened portfolios do display significant differences. The screened portfolios tend to have higher exposure to market risk, lower exposure to the size factor, and higher exposure to the book-to-market and momentum factor. However, the factor loadings for the portfolios, especially for the size and book-to-market factor, change when the portfolios are equally weighted rather than value-weighted and depend thus on the chosen weighting method.
When the dimension specific portfolios are compared to each other, the governance portfolios can outperform some of their equivalents from the social and environmental portfolios in terms of risk-adjusted returns. This result supports the second hypothesis that the dimension specific portfolios are significantly different from each other. The governance dimension seems to be the most dominant dimension for generating risk-adjusted returns. Comparing the governance portfolios to the total ESG portfolios, the governance portfolios also outperform some of the equivalent ESG portfolios, which indicates that focussing on the governance dimension separately instead of screening for total ESG scores could potentially improve the performance of a portfolio.
Overall, the results suggest that investors can introduce SRI into their investment decisions without having to compromise on financial performance. Furthermore, focussing on the governance dimension seems to have positive benefits for returns. The differences between the dimensions and their individual impact on risk and return provide an interesting direction for future research.
23
References
Albuquerque, R., Koskinen, Y., Zhang, C., 2019. Corporate social responsibility and firm risk: Theory and empirical evidence. Management Science 65(10), 4451–4469.
Apostolakis, G., Kraanen, F., Van Dijk, G., 2016. Examining pension beneficiaries’ willingness to pay for a socially responsible and impact investment portfolio: A case study in the Dutch healthcare sector. Journal of Behavioral and Experimental Finance 11, 27-43. Auer, B., 2016. Do socially responsible investment policies add or destroy European stock
portfolio value? Journal of Business Ethics 135(2), 381–397.
Bauer, R., Koedijk, K., Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. Journal of Banking and Finance 29, 1751–1767. Bauer, R., Derwall, J., Otten, R., 2007. The ethical mutual funds performance debate: New
evidence for Canada. Journal of Business Ethics 70 (2), 111–124.
Bauer, R., Otten, R., Tourani Rad, A., 2006. Ethical investing in Australia: Is there a financial penalty? Pacific-Basin Finance Journal 14 (1), 33–48.
Bhagat, S., Bolton, B., 2008. Corporate governance and firm performance. Journal of Corporate Finance 14(3), 257-273.
Bhagat, S., Bolton, B., 2019. Corporate governance and firm performance: The sequel. Journal of Corporate Finance 58, 142-168.
Bostian, M., Färe, R., Grosskopf, S., Lundgren, T., 2016. Environmental investment and firm performance: A network approach. Energy Economics 57, 243-255.
Carhart, M., 1997. On persistence in mutual fund performance. The Journal of Finance 52(1), 57-82.
Carroll, T., Niehaus, G., 1998. Pension plan funding and corporate debt ratings. The Journal of Risk and Insurance 65(3), 427-443.
Chang, C., Nelson, W., Witte, H., 2012. Do green mutual funds perform well? Management Research Review 35(8), 693-708.
Chen, P., Mehrota V., Sivakumar, R., Yu, W., 2001. Layoffs, shareholders’ wealth, and corporate performance. Journal of Empirical Finance 8(2), 171-199.
Climent, F., Soriano, P., 2011. Green and good? The investment performance of US environmental mutual funds. Journal of Business Ethics 103(2), 275–287.
Cowton C., 1994. The development of ethical investment products. In: Prindl, A.R., Prodhan, B. (Ed.), The ACT Guide to ethical conflicts in finance, Blackwell Publishing, Oxford, pp. 213-232.
Doh, J., Howton, S., Howton, S., Siegel, D., 2010. Does the market respond to an endorsement of social responsibility? The role of institutions, information, and legitimacy. Journal of Management 36(6), 1461-1485.
Drempetic, S., Klein, C., Zwergel, B., 2019. The influence of firm size on the ESG score: Corporate sustainability ratings under review. Journal of Business Ethics
El Ghoul, S., Guedhami, O., Kim, H., Park, K., 2018. Corporate environmental responsibility and the cost of capital: International evidence. Journal of Business Ethics 149(2), 335 -361
24 Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of
Financial Economics 33(1), 3–56.
Galema, R., Plantinga, A., Scholtens, B., 2008. The stocks at stake: return and risk in socially responsible investment. Journal of Banking and Finance 32(12), 2646–2654.
Geczy, C., Stambaugh, R., Levin, D., 2006. Investing in socially responsible mutual funds. Working Paper, Wharton School, University of Pennsylvania.
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. The Quarterly Journal of Economics 118(1), 107-155.
Kempf, A., Osthoff, P., 2007. The effect of socially responsible investing on portfolio performance. European Financial Management 13(5), 908-922.
Kotsantonis, S., Pinney, C., Serafeim, G., 2016. ESG integration in investment management: Myths and realities. Journal of Applied Corporate Finance 28(2), 10-16.
Kreander, N., Gray, G., Power, D., Sinclair, C., 2005. Evaluating the performance of ethical and non-SRI funds: A matched pair analysis. Journal of Business, Finance and Accounting 32 (7), 1465–1493.
Reinhardt, F., 1999. Market failure and the environmental policies of firms. Journal of Industrial Ecology 3(1), 9–21.
Renneboog, L., Ter Horst, J., Zhang, C., 2008. Socially responsible investments: Institutional aspects, performance, and investor behaviour. Journal of Banking and Finance 32(9), 1723-1742.
Sauer, D., 1997. The impact of social-responsibility screens on investment performance: Evidence from the Domini 400 social index and Domini equity mutual fund. Review of Financial Economics 6(2), 137–149.
Sparkes, R., 2001. Ethical investment: Whose ethics, which investment? Business Ethics: A European Review 10(3), 194-205.
Sparkes, R., Cowton, C., 2004. The maturing of socially responsible investment a review of the developing link with corporate social responsibility. Journal of Business Ethics 52(1), 45-57.
