ESG INVESTING AND PORTFOLIO PERFORMANCE: EVIDENCE FROM EMERGING MARKETS
Author: Supervisors:
SJOERD VAN HAAREN1 Dr. A. PLANTINGA
Dr. S.J. DRIJVER
A thesis submitted to the Faculty of Economics and Business in partial fulfillment of the requirements for the degree of Master of Science in Finance
December 13, 2017
2
Abstract
Although there is abundant literature about the financial performance of strategies based on environmental, social and governance (ESG) criteria in developed markets, ESG investing in emerging markets has been studied to a much lesser extent. This paper aims to fill this gap in the literature. Using a best-in-class approach, portfolios are created with stocks of emerging market firms with the highest environmental, social, governance and weighted average ESG scores. Moreover, two equity portfolios are formed with stocks of firms whose ESG rating recently improved or deteriorated. Findings indicate that the risk-adjusted returns of all these portfolios are not significantly different from zero. This could imply that emerging markets are more efficient than generally believed.
Keywords: ESG investing; socially responsible investing; financial performance; emerging
markets
3
1.
Introduction
In recent decades, socially responsible investing (SRI) has emerged as a major field of interest among both investors and financial researchers. The United Nations defines SRI as
“an approach to investing that aims to incorporate environmental, social and governance (ESG) factors into investment decisions, to better manage risk and generate sustainable, long-term returns” (UN Principles for Responsible Investment, 2017a). “Impact investing”, “ESG investing” and “values-based investing” are other terms to describe strategies aimed at creating positive societal changes, although there are some small distinctions (Snider, 2015).
Already in 1971, the first mutual SRI fund was created in the US. This Paw World Fund was designed for opponents of the Vietnam War and did not invest in the arms industry (Renneboog, Ter Horst and Zhang, 2008). In the subsequent 20 years, SRI remained a relatively unknown phenomenon, but since the early 1990s SRI has grown tremendously (Renneboog et al., 2008). As shown in Figure 1, at April 1, 2017 institutional assets worth $68.4 trillion were managed by Principles for Responsible Investment signatories. This represents a more than tenfold increase since April 1, 2006. The principles are shown in Table 1.2
Fig. 1: Assets under Management (in US$ trillion) of UN Principles for Responsible Investment. Numbers at April 1 of every year. Source: Website UN PRI (UN Principles for Responsible Investment, 2017b)
2 Kotsantonis, Pinney and Serafeim (2016) point out that signing up to the UN Principles for Responsible Investment is voluntarily and that the UN has no enforcement power. Although the numbers in Figure 1 are an indication of the levels of SRI, it is not synonymous to actual incorporation of ESG issues into business practices.
4
Table 1
Principles for responsible investment
Principle 1 We will incorporate ESG issues into investment analysis and decision-making processes. Principle 2 We will be active owners and incorporate ESG issues into our ownership policies and principles. Principle 3 We will seek appropriate disclosure on ESG issues by the entities in which we invest.
Principle 4 We will promote acceptance and implementation of the principles within the investment industry. Principle 5 We will work together to enhance our effectiveness in implementing the principles.
Principle 6 We will each report on our activities and progress towards implementing the principles.
Note: Source: Website UN PRI (UN Principles for Responsible Investment, 2017b)
Issues such as climate change, corporate scandals and the rise of microfinance have contributed to the growing interest in ESG investing (Renneboog et al., 2008). Moreover, the nature of firms has changed over time. As Figure 2 illustrates, only 17% of the assets of S&P 500 companies consisted of intangible assets in 1975, but in 2015 87% of the assets was intangible. Snider (2015) argues that this trend causes factors such as reputation, human resources and R&D to become more important for the evaluation of firms. ESG data can help to evaluate risks and opportunities. By using ESG analyses in addition to financial statement analyses, investors can improve decision making.
Fig. 2: Assets composition (according to market value) for S&P 500 firms Source: Website Ocean Tomo (Ocean Tomo, 2015).
5 Apart from the higher demand from the investor side, firms also face regulatory pressure. For instance, EU Directive 2014/95/EU requires large companies to disclose information about the way they manage environmental and social issues (European Commission, 2014). Moreover, stock exchanges are introducing ESG reporting requirements (Kotsantonis et al., 2016). As a consequence, ESG disclosure has risen strongly over the past years. As Figure 3 shows, 20% of the S&P 500 companies issued sustainability reports in 2011. Five years later, this number has increased to 82%.
Fig. 3: Share of S&P 500 firms issuing sustainability reports, 2011-2016
Source: Website Governance & Accountability Institute (Governance & Accountability Institute, 2016).
The link between SRI and financial performance has been widely researched; Friede, Busch and Bassen (2015) identify more than 2,200 studies conducted between 1970 and 2015 about this relationship. At the firm level, higher environmental, social and governance standards could lead to higher operating costs and thus lower profitability (Cheung et al., 2010). However, these higher standards could also translate into a better reputation, which can boost profitability. Moreover, firms with higher ESG standards are more likely to avoid fines for environmental violations and financial losses due to labor disputes (Nagy, Kassam and Lee, 2015). At the portfolio level, it is often claimed that ESG performance is negatively associated with financial performance because investment strategies based on ESG criteria limit the investment universe. This is especially a concern when negative screens are applied and certain industries are eliminated (Sauer, 1997). On the other hand, ESG screening can also result in value-relevant information, which investors can use to construct portfolios with better risk-return characteristics. Renneboog et al. (2008) identify two channels through which this effect can work. Firstly, high environmental, social and governance scores could be an
6 indication of high levels of managerial skill, which ultimately leads to high levels of financial performance. Secondly, ESG selection reduces the risk of incurring high expenses when corporate scandals or environmental disasters occur. Note that ESG-based portfolios can only generate positive risk-adjusted returns if equity markets undervalue environmental, social and governance information. Therefore, believers in market efficiency would argue that ESG investing does not deliver superior returns.
Existing ESG research has mainly focused on developed markets because of the availability of ESG data. Many ESG rating agencies rely largely on data provided by the companies they rate when constructing ESG scores (Novethic, 2013). Although disclosure of SRI-relevant information is still higher among developed market firms, emerging market firms have made a greater enhancement between 2011 and 2015 (Lampl, Bardoscia & Munge, 2016). This increase in data availability opens up possibilities for ESG research that focuses on emerging markets.
This study aims to answer the question whether investing based on ESG criteria can add value in emerging markets. The environment in which emerging market firms operate is fundamentally different from the environment of developed market firms. Moreover, the nature of emerging market firms is also different from the nature of their counterparts in developed markets. Therefore, firms in emerging markets also have to deal with different ESG issues (NN Investment Partners, 2017). For instance, governance practices of these firms are different, as emerging markets are often characterized by a high number of state-owned enterprises and family-owned businesses (Cambridge Associates, 2016). As a consequence, ESG research that focuses on developed markets is not necessarily a good guide for emerging markets.
Since the end of the 1990s, a dynamic ESG rating market has emerged (Novethic, 2013). ESG rating agencies generally use both company provided public information and information provided by NGOs, governments and other third parties to analyze ESG issues. This paper will use data from MSCI, one of the main rating agencies, to explore the relationship between ESG performance and financial performance in emerging markets. MSCI ESG Research was created in 2010, after MSCI acquired RiskMetrics. RiskMetrics itself has taken over the rating agencies Innovest and KLD in 2009 (Novethic, 2013).
7
2.
Theoretical Framework
2.1 Research methods
There are four categories of empirical research on the relationship between socially responsible investing and financial performance. Firstly, researchers compare the historical returns of socially responsible mutual funds to returns of conventional funds or market indices. In general, returns of SRI funds are not different from conventional funds (e.g. Hamilton, Jo & Statman, 1993; Statman, 2000; Kreander et al., 2005). Auer and Schuhmacher (2016) point out that this type of research suffers from several disadvantages. For example, the fact that a fund is labelled as “socially responsible” does not guarantee that a fund actually follows SRI principles, because managers tend to adjust fund holdings during experiences of market turmoil. Moreover, mutual fund performance is not only determined by the performance of the fund’s securities, but also by management fees. Finally, researchers generally use Jensen’s alpha to measure the financial performance of mutual funds. However, this is only a suitable performance measure for well-diversified portfolios. Some SRI funds exclude certain sectors and thus limit diversification opportunities.
The second category uses event studies as the main tool of analysis. Hamilton (1995) finds that listed companies have negative abnormal returns following their first publication of toxics release inventory (TRI) pollution figures. Furthermore, Dasgupta, Laplante and Mamingi (2001) find evidence for positive market movements in response to the announcement that a firm has won an environmental award. They also report negative share price reactions in response to citizens’ complaints about specific firms. According to Renneboog et al. (2008), these findings do not necessarily imply that investors are willing to pay for environmental performance, because environmental events can also be correlated to future cash flows.
Thirdly, researchers examine the relationship between SRI standards and financial performance directly. For instance, Dowell, Hart and Yeung (2000) find that firms committed to a global environmental standard have higher market values. However, the problem of this research design is that correlation does not imply causation and that the direction of the relationship is difficult to determine (Renneboog et al., 2008).
8 finds that the application of a social screen has little influence on portfolio performance. Moreover, Guerard Jr. (1997) finds that the return of a socially screened portfolio is not significantly different from the return of an unscreened portfolio between 1987 and 1996. The mixed conclusions are an indication that the relationship depends on the particular screen, time and market.
Instead of focusing on socially responsible portfolios, some researchers do exactly the opposite: they build portfolios consisting of sin stocks. Sin stocks are stocks of firms that are involved in activities which are considered to be socially irresponsible. Several authors report significantly positive alphas of sin stock portfolios (e.g. Salaber, 2007; Fabozzi, Ma & Oliphant, 2008; Hong and Kacperczyk, 2009). A common explanation for this effect is that these stocks are underpriced because many investors avoid these stocks (Fabozzi et al., 2008). However, recent research from Blitz and Fabozzi (2017) questions these results. They show that portfolios consisting of sin stocks of firms from the US, Europe, Japan and the global developed markets have significantly positive CAPM alphas. However, after including control variables for size, value, momentum, profitability and investment the alphas are no longer significantly different from zero.
Apart from selecting stocks based on some absolute measure of ESG performance, recent evidence has also focused on selecting stocks based on changes in ESG scores. This strategy is closely linked to the momentum effect identified by Jegadeesh and Titman (1993). They show that a strategy that involves buying past winners and selling past losers generates significantly positive abnormal returns over holding periods between three and 12 months. Nagy, Kassam and Lee (2015) demonstrate that a strategy that overweighs stocks of firms whose ESG score increased during the past 12 months also delivers a positive abnormal return. They argue that an ESG rating upgrade signals that a firm is better able to avoid ESG risks. As a consequence, potential future liabilities are lower. Since it might take some time before market participants incorporate this signal into the share price, a strategy based on overweighing companies whose rating recently improved can be profitable.
2.2 Differences emerging markets and developed markets
9 markets such as India (Poshakwale, 1996) and Bangladesh (Mobarek, Mollah and Bhuyan, 2008) rejects the null hypothesis of weak-form efficiency. If environmental, social or governance information is undervalued in emerging markets, then an investment strategy based on ESG criteria can be profitable.
Ortas, Moneva and Salvador (2012) identify several characteristics of emerging markets which lead to ESG challenges. Firstly, rapid population growth and advancements in living standards cause an increase in the consumption of resources such as water and oil. This highlights the importance of sustainable business solutions. Consequently, firms who do not implement sustainable business practices limit their business opportunities. Secondly, government-owned firms are more common in emerging markets than in developed markets.3 This often has negative consequences, such as limited director independence, which could negatively affect investor trust. Thirdly, the limited access to means of long-term financing in emerging markets can restrict growth opportunities for firms that are in need of capital. Since providers of finance may prefer to deal with sustainable firms, these firms face this problem to a much lesser extent.
2.3 SRI in emerging markets
A relatively small number of papers has focused on responsible investing in emerging markets specifically. Cheung et al. (2010) find a positive relationship between SRI scores and market valuation of Asian firms. However, their SRI score is simple and depends on firm disclosure. Mishra and Suar (2010) focus on Indian companies and find that favorable perceptions of managers towards SRI are positively associated with both financial and non-financial performance. However, at the time of the study there was no objective measure of SRI available for Indian firms. The SRI scores are based on questionnaires and are therefore subjective. Finally, Ortas, Moneva and Salvador (2012) investigate the performance of the Brazilian Corporate Sustainability Index. They find that the BCSI is less risky and delivers a return similar to its benchmark, the Bovespa Index, during stable market periods.
In a recent report, NN Investment Partners (2017) investigates the momentum effect of ESG scores in emerging markets specifically. They find that a strategy that buys stocks of firms who have improved ESG practices outperforms a strategy based on selecting stocks of firms whose ESG rating has declined. However, this analysis simply compares the Sharpe ratios of the two portfolios and makes no attempt to control for other risk factors. Moreover, the report is not transparent about when the portfolio composition is adjusted and how long stocks are included in the portfolio.
10 2.4 MSCI Emerging Markets ESG Leaders Index
In June 2013, MSCI launched the MSCI Emerging Markets ESG Leaders Index. This index aims to give exposure to firms with high ESG ratings relative to other companies in their sector (MSCI, 2017b). Appendix A provides a detailed description of the index methodology.
Fig. 4. Cumulative price returns MSCI Emerging Markets Index and MSCI Emerging Markets ESG Leaders Index Index values measured in local currency.
Figure 4 shows the cumulative returns of this index and its benchmark, the MSCI Emerging Markets Index. Over the period from June 2013 until August 2017, the cumulative return of the ESG Leaders index was 16.3 percentage points higher than the return of its benchmark.
The investment consulting firm Cambridge Associates investigates the performance of this index until June 2016 (Cambridge Associates, 2016). Using an attribution analysis, they show that investment style and sector exposure accounted for part of the active return, but stock-specific sources explain most of the outperformance. Cambridge Associates argues that the high prevalence of government-owned and family-owned firms in emerging markets may explain this finding, because these businesses often face stakeholders who are not solely focused on generating profits. However, they emphasize that examining the influence of each of the ESG pillars is beyond the scope of their study. This paper is therefore an attempt to fill a gap in the literature. By constructing portfolios based on the various ESG dimensions, the effect of applying environmental, social and governance criteria will be analyzed separately.
-20% -10% 0% 10% 20% 30% 40% 50% 06-2013 06-2014 06-2015 06-2016 06-2017 MSCI EM Index
11
3.
Data & Methodology
3.1 Portfolio construction
The ESG data analyzed in this paper is provided by MSCI. MSCI has rated 6,400 firms based on their environmental, social and governance risks and opportunities (MSCI, 2017a). It collects and processes three categories of public information: specialized datasets, company disclosure and media disclosure. The environmental, social and governance pillars receive an individual score which is comparable across industries. The individual scores are combined into a weighted average score, on scale ranging from zero to ten. MSCI also converts this score into an ESG rating (on a scale ranging from CCC to AAA), which is relative to industry peers (MSCI, 2017a). This means that theoretically an AAA-rated firm can have a lower absolute score than a B-rated firm from another industry. A more elaborative description of the rating process is presented in Appendix B.
MSCI has provided its historical ESG databases from the first days of every month from January 2007 until June 2017. Both equity and fixed income issuers are included in these databases. This study focuses on firms from countries classified as “emerging markets” and it will be taken into account that MSCI has changed this classification over time. For instance, Israeli firms are only eligible for portfolio inclusion until 2010, because MSCI changed the classification of Israel from “emerging market” to “developed market” in May 2010. For a historical overview of which countries are categorized “emerging markets”, see Appendix C.
12 The remaining number of firms is shown in Figure 5. Coverage of emerging market firms increased enormously during 2012, prior to the launch of the MSCI Emerging Markets ESG Leaders Index in June 2013. Constructing portfolios based on ESG criteria requires a wide universe of rated firms. Therefore, this paper mainly focuses on the results from the analysis of the period starting from January 2013.
Fig. 5. Number of observations after cleaning process
Numbers per January 1 of every year. Includes publicly listed firms from countries that are classified as “emerging markets” by MSCI at the time of the rating. Only firms that received a rating in the past 12 months are included.
Following a similar approach as Derwall et al. (2005), firms from emerging markets will be ranked from high to low based on their environmental, social, governance and weighted average score, respectively. A best-in-class strategy will be followed, by creating four equity portfolios consisting of the highest ranked firms which cover 20% of the market capitalization of all rated firms (hereafter referred to as “ESG portfolio”,4 “Environmental portfolio”, “Social portfolio" and “Governance portfolio”, respectively). These “level portfolios” are formed in January 2007. Rebalancing takes place annually, based on updated ESG scores and market capitalization.
Two other equity portfolios are formed, consisting of firms whose ESG score increased (hereafter: “Improvers portfolio”) and decreased (hereafter: “Decliners portfolio”), respectively, during the past 12 months. ESG score refers to the final letter rating. If a firm was both upgraded and downgraded during the past 12 months, the most recent change determines in which portfolio it will be placed. These “change portfolios” are formed in January 2008 and also rebalanced annually.
13 3.2 Performance measurement
For the purpose of measuring portfolio performance, first the Sharpe ratios of the ESG-based portfolios will be compared to the Sharpe ratio of the MSCI Emerging Markets Index. The standard Sharpe ratio is defined as follows:
𝑆ℎ𝑎𝑟𝑝𝑒 𝑟𝑎𝑡𝑖𝑜 = 𝑅̅𝑝−𝑅̅𝑓
𝜎𝑅𝑝−𝑅𝑓 , (1)
where 𝑅̅𝑝 denotes the average monthly portfolio return; 𝑅̅𝑓 is the average risk-free rate and 𝜎𝑅𝑝−𝑅𝑓 is the standard deviation of the monthly portfolio excess returns.
Returns are monthly price returns, measured in euros.5 The risk-free rate is the one month Euribor rate. By measuring all variables in euros, the results can be analyzed on a common basis.
During the financial crisis excess returns often turned negative, which results in negative Sharpe ratios. Interpreting negative Sharpe ratios can be counterintuitive.6 Therefore, this paper applies the modification suggested by Israelsen (2009). He adds an exponent to the denominator of the Sharpe ratio. This exponent is equal to the average excess return divided by the absolute value of the average excess return.
𝑅𝑒𝑓𝑖𝑛𝑒𝑑 𝑆ℎ𝑎𝑟𝑝𝑒 𝑅𝑎𝑡𝑖𝑜 = 𝐸𝑅̅̅̅̅ 𝜎𝐸𝑅 ( 𝐸𝑅 |𝐸𝑅|) , (2) where 𝐸𝑅̅̅̅̅ = 𝑅̅𝑝 - 𝑅̅𝑓
The consequence of this modification is that the magnitudes of the outcomes are less meaningful, but it produces an appropriate ranking of the portfolios on a risk-adjusted basis.
5
Datastream converts portfolio values into euros using the WM/Reuters local currency/euro closing exchange rate of that day.
14 This paper follows the approach introduced by Jobson and Korkie (1981) to test for the significance of the differences between the Sharpe ratios. This means that the test statistic is given by the transformed difference:
𝑆𝐻̂𝑖𝑛 = 𝑠𝑛𝑟̅𝑖− 𝑠𝑖𝑟̅𝑛 , (3) where si, sn, r̄i and r̄n denote the standard deviation and the mean of the excess returns of portfolio i and n, respectively.
The asymptotic distribution is normal with mean 𝑆𝐻̂𝑖𝑛 and the sample variance is given by:
ψ =1 𝑇[2𝑠𝑖 2𝑠 𝑛2− 2𝑠𝑖𝑠𝑛𝑠𝑖𝑛+ 0.5𝑟̅𝑖2𝑠𝑛2+ 0.5𝑟̅𝑛2𝑠𝑖2+ 𝑟̅ 𝑟𝑖̅̅̅𝑛 2𝑠𝑖𝑠𝑛(𝑠𝑖𝑛 2 + 𝑠 𝑖2𝑠𝑛2)] (4)
Jobson and Korkie (1981) show that the approximate Z-value is a suitable instrument to test the null hypothesis that the transformed difference is equal to zero:
𝑍𝑠𝑖𝑛 = 𝑆𝐻̂𝑖𝑛
√ψ (5)
Secondly, the Fama-French three-factor model (Fama and French, 1993) will provide evidence on portfolio performance after controlling for factor risks. This is relevant because the literature shows that socially responsible investors have different investment styles compared to conventional fund investors (Bauer, Koedijk & Otten, 2015). The MSCI Emerging Markets Index, converted into euros, is taken as the market proxy and the one month Euribor rate is the risk-free rate. A complicating issue is that the HML and SMB returns are only available for developed markets in the Kenneth R. French Library. Since Cakici, Fabozzi and Tan (2013) show that local factors perform much better than global factors in emerging markets, this paper will not use the developed market factor returns.
15 Instead, an alternative approach will be followed to construct factor returns for emerging markets. The SMB returns, which ought to represent the size effect, are approximated by subtracting the historical returns of the MSCI Emerging Markets Large Cap Index from the returns of the MSCI Emerging Markets Small Cap Index. Similarly, the HML factor returns are approximated by subtracting historical returns of the MSCI Emerging Markets Growth Index from the returns of the MSCI Emerging Markets Value Index.7
Moreover, the analysis of Cambridge Associates (2016) shows the relevance of industry exposure for ESG investing. The best-in-class approach of this study might lead to considerable sector biases. Therefore, this paper will also control for industry factors, by following Pástor and Stambaugh (2002) and applying a principal component analysis. This is a method to simplify j original variables to k orthogonal derived variables. Firstly, the excess returns of the 11 MSCI Emerging Markets industry indices will be regressed on a constant, the excess market return and the SMB and HML factors. Secondly, a principal component analysis will be performed on the part that cannot be explained by the model (the intercept and error term). In accordance with the Kaiser criterion, only the k factors with an eigenvalue greater than one will be retained (Kaiser, 1960). Appendix D provides a more detailed description of this procedure and the results of the analysis. Equation 6 shows the multifactor regression model.
𝑟𝑖𝑡− 𝑟𝑓𝑡 = 𝛼𝑖+ 𝛽0𝑖𝑅𝑀𝑅𝐹𝑡+ 𝛽1𝑖𝑆𝑀𝐵𝑡+ 𝛽2𝑖𝐻𝑀𝐿𝑡+ 𝛽3−(𝑘+3),𝑖𝐼𝑃1−𝑘,𝑡+ 𝑒𝑖𝑡 , (6) where rit – rft is the monthly portfolio return in excess of the one month Euribor rate in period t; RMRFt is the excess return on the MSCI Emerging Markets Index in period t; SMBt and HMLt are the returns on diversified value-weighted factor-mimicking portfolios for size and value in period t, and IP1-k,t are principal components which control for industry exposure in period t.
7
16 3.3 Robustness tests
The multifactor regression model will be rerun using some alternative specifications to test the robustness of the results.
Transaction costs of 25, 50 and 100 basis points will be introduced.
Instead of examining the top 20% with the highest scores, portfolios of the top 10% and top 30% firms will be created.
17 3.4 Descriptive statistics
3.4.1 ESG scores
Figure 6 shows the weighted average scores (on a scale ranging from zero to ten) of all firms in the database of this study. Overall, there is no clear trend in the scores. This does not necessarily mean that ESG practices have not improved over time, because the number of rated firms has also increased. In other words, it is possible that the early rated firms incorporated ESG practices relatively well, and that the absence of an upward trend is caused by the fact that the companies with lower levels of ESG performance entered the database over time.8
Fig. 6. Average ESG Scores of publicly listed emerging market firms
These are the scores of firms that are included in the MSCI database per January 1 of each year. Only publicly listed firms that received their most recent rating in the past 12 months are taken into account.
Figure 7 shows the average environmental, social and governance scores of the portfolios. As expected, the level portfolios score higher on the dimension which has been used as selection criterion. Furthermore, the change portfolios score lower than the level portfolios on all pillars.
8
One might argue that an analysis of only the scores of the firms that were already in 2007 included in the database would clarify this issue. However, there might be a survivorship bias in the results because ESG data is not available for all these firms until 2017.
0 1 2 3 4 5 6 7 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
18 Fig. 7. Average environmental, social and governance scores in portfolios
3.4.2 Portfolio returns
Table 2 presents the summary statistics of the monthly portfolio returns. The mean monthly returns of the level portfolios vary between -0.1% and 0.5%, whereas the MSCI Emerging Markets Index has a mean return of 0.4%. The mean returns of the Improvers portfolio and the Decliners portfolio are 0.0% and 0.4%, but these numbers are not directly comparable to the level portfolio means because they are measured over a different time span. Standard deviations vary between 5.3% and 5.9% (level portfolios) and 5.9% and 6.5% (change portfolios). For all portfolios and the index, there is evidence of excess kurtosis. Consequently, the Jarque–Bera test rejects the null hypothesis of normality in all cases at the 5% significance level. This is consistent with evidence from previous studies (e.g. Hwang et al., 2013; Balli et al., 2015) and the popular perception that emerging market equity portfolios are characterized by volatile returns. Moreover, for five of the six portfolios and the Emerging Markets Index, the null hypothesis of no first-order autocorrelation is rejected at the 10% significance level. The existence of autocorrelation in emerging market returns, which contradicts the efficient market hypothesis, is also consistent with the existing literature (e.g. Mobarek et al., 2008; Poshakwale, 1996).9
9
The presence of autocorrelation and the non-normality of the dependent variable are in itself not problematic issues, since OLS requires only normality of and absence of autocorrelation in the error term.
0 1 2 3 4 5 6 7 8 WEIGHTED AVERAGE PORTFOLIO ENVIRONMENTAL PORTFOLIO
SOCIAL PORTFOLIO GOVERNANCE PORTFOLIO
IMPROVERS PORTFOLIO
DECLINERS PORTFOLIO
Table 2
Summary statistics of monthly portfolio returns (%)
ESG portfolio1 Environmental portfolio1 Social portfolio1 Governance portfolio1 Improvers portfolio2 Decliners portfolio2 Emerging Markets Index1 Mean 0.258 0.547 0.303 -0.106 -0.049 0.387 0.374 Std. dev. 5.582 5.887 5.427 5.296 5.886 6.520 5.502 Maximum 20.192 21.666 20.135 12.404 14.994 36.550 16.337 Minimum -19.205 -19.285 -17.936 -16.650 -22.379 -21.638 -18.428 Skewness 0.067 0.037 0.070 -0.307 -0.806 0.954 -0.320 Kurtosis 4.512 4.466 4.665 3.787 4.762 10.900 4.208 Jarque-Bera 12.280*** 11.493*** 14.891*** 5.314** 27.556*** 319.210*** 9.962*** Q(1) 4.331** 2.794* 3.933** 0.835 8.329*** 8.004*** 8.275**
Notes: Returns are monthly percentage changes. 1) Sample period is January 2007-August 2017. 2) Sample period is January 2008-August 2017. All series are in euros.
3.4.3 Country weights
Figure 8 shows the average country weights of all portfolios. For each portfolio the five countries with the highest weights are shown. It appears that the country weights of the level portfolios are fairly similar, but there are large differences with the change portfolios and the Emerging Markets Index. Remarkable is the difference in weights on Chinese firms between the Emerging Markets Index and the level portfolios (this weight is 3%, 4%, 5% and 10% in the ESG, Environmental, Social and Governance portfolio, respectively, versus 19% in the EM index). This underweighting of Chinese stocks is caused by their relatively low ESG scores.10
10
21 Fig. 8. Average country weights per portfolio
22% 15% 14% 13% 13% 23% ESG portfolio South-Korea South-Africa Brazil Taiwan India Other 25% 16% 14% 11% 11% 23% Environmental portfolio South-Korea Brazil Taiwan India South-Africa Other 16% 16% 15% 13% 12% 29% Social portfolio Brazil South-Korea Taiwan South-Africa India Other 17% 16% 13% 12% 12% 30% Governance portfolio South-Korea Brazil Taiwan South-Africa India Other 19% 16% 11% 11% 8% 35% Improvers portfolio China Brazil South-Korea India Taiwan Other 19% 14% 13% 10% 10% 34% Decliners portfolio China South-Africa South-Korea Russia India Other 19% 15% 12% 11% 8% 35%
MSCI Emerging Markets Index
22 3.4.4 Currency effects
Figure 9 shows the fluctuations in the exchange rates of the currencies of the major emerging markets to the euro. In the period 2007-2017, the exchange rate variations have been considerable. Therefore, there are two effects, working in opposite directions, which could potentially explain a large share of the portfolio returns. Firstly, there is a direct effect of exchange rate changes when converting the invested amount (in local currency) back to euros. Secondly, there is an indirect effect because exchange rate changes also affect a firm’s competitiveness and thus its share price. For instance, an appreciation of the Brazilian Real means that an amount invested in Brazil will be converted to a higher amount in euros, all else equal. However, the appreciation also has a negative effect on the competitiveness of Brazilian firms, which negatively influences the returns of these firms.
Fig. 9. Cumulative changes in the spot exchange rate of the currencies of six major emerging markets to the euro Base date: January 1, 2007. Source: Thomson Reuters Datastream.
-40% -20% 0% 20% 40% 60% 80% 100% 120% 140% 160% 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
23 Throughout this paper, spot exchange rates are used to convert investments from local currencies into euros. However, in practice investors may choose to hedge currency risks, for instance by entering into futures (or forwards) contracts. In the absence of arbitrage opportunities, a futures contract which fixes the exchange rate at which a particular amount in some local currency is converted into euros (three months from now) is priced as follows:
𝐹𝑢𝑡𝑢𝑟𝑒𝑠𝑃𝑟𝑖𝑐𝑒𝑒𝑢𝑟/𝑙𝑜𝑐 ,3𝑚𝑜𝑛𝑡ℎ𝑠 = 𝑆𝑝𝑜𝑡𝑅𝑎𝑡𝑒𝑒𝑢𝑟/𝑙𝑜𝑐∗1+𝑖𝑒𝑢𝑟, 3 𝑚𝑜𝑛𝑡ℎ𝑠
1+𝑖𝑙𝑜𝑐, 3 𝑚𝑜𝑛𝑡ℎ𝑠 , (7)
where FuturesPriceeur/loc, 3 months is the price to convert one unit of local currency into euros three months from now, against the current spot exchange rate; SpotRateeur/loc is the current euro/local currency spot exchange rate; ieur, 3 months is the three months interest rate in the Eurozone and iloc, 3 months is the three months interest rate in the local country.
In recent years, interest rates in emerging markets have been relatively high compared to rates in developed markets (Center for Global Development, 2017). This means that the futures price will exceed the spot rate. In other words, even without taking the transaction costs of entering into currency future contracts into account, hedging currency exposure in emerging markets can be quite costly from the perspective of a euro investor.
24 3.4.5 Number of stocks in portfolios
Table 3 shows the number of stocks in the portfolios. The large increase in the number of stocks in 2013 (level portfolios) and 2014 (change portfolios) compared to the previous year is caused by the expansion in rated firms starting in 2012 (see Figure 4).
Although Evans and Archer (1968) conclude that a portfolio consisting of 15 randomly picked stocks already eliminates virtually all unsystematic risk, more recent evidence suggest that a larger number of stocks is required. For example, Campbell et al. (2001) show that before 1985 a portfolio of 20 randomly chosen stocks had an excess standard deviation of about five percent. However, for the period between 1986 and 1997, almost 50 stocks were needed to reduce excess standard deviation to this level.11 Statman (2002) suggests that even more than 120 stocks are required for optimal diversification. Therefore, the subperiods 2007-2012 versus 2013-2017 (level portfolios) and 2008-2013 versus 2014-2017 (change portfolios) will be analyzed separately.
Table 3
Number of stocks in portfolios
ESG portfolio Environmental portfolio Social portfolio Governance portfolio Improvers portfolio Decliners portfolio 2007 13 15 13 12 n.a. n.a. 2008 39 42 37 44 14 8 2009 15 18 18 17 19 3 2010 17 21 16 22 26 16 2011 11 12 11 17 26 6 2012 25 15 28 33 7 8 2013 187 149 223 202 26 30 2014 186 167 240 256 186 108 2015 232 166 254 281 167 87 2016 216 232 272 284 174 131 2017 258 216 255 297 168 86 11
25
4.
Results
4.1 Sharpe ratios
The refined Sharpe ratios reported in Tables 4 and 5 give an initial impression of the financial performance of the portfolios.
Table 4 shows that during the period from January 2007 until August 2017 the Environmental portfolio had a higher Sharpe ratio than the benchmark, whereas the ESG portfolio, Social portfolio and the Governance portfolio performed worse. All Sharpe ratios are negative, which is caused by on average negative excess returns. As the analyses of the subperiods show, these negative Sharpe ratios are caused by the returns in the period from 2007 until 2012, when the global financial crisis occurred and excess returns turned negative. The analyses from the period after 2013 show that all portfolios have a higher Sharpe ratio than the benchmark. These results are the most meaningful since ESG data is available for a considerable number of firms during this time period.
Table 4
Refined Sharpe ratios: level portfolios and Emerging Markets Index
Full period 2007-2012 2013-2017
𝑅̅𝑝
̅̅̅̅ − 𝑅̅𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio 𝑅̅̅̅̅̅ − 𝑅̅𝑝 𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio 𝑅̅̅̅̅̅ − 𝑅̅𝑝 𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio
ESG portfolio -0.779% 6.216% -0.00048 -1.853% 7.390% -0.00137 0.603% 3.908% 0.154
Environmental portfolio -0.490% 6.347% -0.00031 1.199% 7.664% -0.00092 0.422% 3.960% 0.106
Social portfolio -0.734% 6.024% -0.00044 -1.643% 7.236% -0.00119 0.435% 3.704% 0.117
Governance portfolio -1.142% 5.915% -0.00068 -2.390% 6.882% -0.00165 0.462% 3.872% 0.119
EM Index -0.663% 6.038% -0.00040 -1.497% 7.177% -0.00107 0.409% 3.950% 0.104
Notes: Full period denotes period between January 1, 2007 and September 1, 2017. 2007-2012 denotes the period between January 1, 2007 and December 31, 2012.
2013-2017 denotes the period between January 1, 2013 and September 1, 2013-2017. EM Index is MSCI Emerging Markets Index. Returns are monthly price returns; risk-free rate is one month Euribor rate.
Table 5
Refined Sharpe ratios: change portfolios and Emerging Markets Index
Full period 2008-2013 2014-2017
𝑅̅𝑝
̅̅̅̅ − 𝑅̅𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio 𝑅̅̅̅̅̅ − 𝑅̅𝑝 𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio 𝑅̅̅̅̅̅ − 𝑅̅𝑝 𝑓 𝜎𝑅𝑝−𝑅𝑓 Refined Sharpe ratio
Improvers portfolio -0.764% 6.493% -0.00050 -1.842% 7.407% -0.00136 0.998% 4.126% 0.242
Decliners portfolio -0.329% 6.971% -0.00023 -1.048% 8.037% -0.00084 0.848% 4.588% 0.185
EM Index -0.494% 6.160% -0.00030 -1.262% 7.032% -0.00089 0.762% 4.150% 0.184
Notes: Full period denotes period between January 1, 2008 and September 1, 2017. 2008--2013 denotes the period between January 1, 2008 and December 31, 2013.
Table 6 shows the test statistics of the Jobson and Korkie (1981) test and the Z-values with associated probabilities. This test has only been performed on the Sharpe ratios of the second subperiod because the test is designed to evaluate differences between positive Sharpe ratios. Moreover, ESG coverage in emerging markets was very low during the first period, so the results obtained for the second period are much more reliable. The results show that none of the transformed differences is statistically different from zero. In other words, there is no statistical evidence that the risk-return characteristics of the constructed portfolios are different from those of the Emerging Markets Index.
Table 6
Jobson and Korkie (1981) test on difference in Sharpe ratios
Transformed difference Z-values Probability > |Z| Panel A: Level portfolios (2013-2017)
ESG portfolio 0.00008 0.991 0.321
E portfolio 0.00000 0.055 0.956
S portfolio 0.00002 0.239 0.811
G portfolio 0.00002 0.334 0.739
Panel B: Change portfolios (2014-2017)
Improvers portfolio 0.00010 0.718 0.473
Decliners portfolio 0.00000 0.018 0.985
Notes: Probabilities and Z-values are based on the null hypothesis that the transformed difference in Sharpe
28 4.2 Regression results
The regression model shown in Equation 6 is run with the excess returns of the constructed portfolios as dependent variable. The residuals are tested for heteroskedasticity and autocorrelation (extensive results are presented in Appendix E). Since heteroskedasticity is an issue, the regressions are rerun using Newey-West robust standard errors. Results are reported in Table 7 (level portfolios) and Table 8 (change portfolios).
Table 7
Performance level portfolios in a multifactor regression model
α Rm - Rf SMB HML R 2 Panel A: 2007-2017 ESG portfolio -0.111 (-0.57) 0.968*** (28.76) -0.215 (-1.57) -0.009 (-0.07) 0.901 E portfolio 0.185 (1.01) 0.990*** (31.12) -0.248* (-1.93) -0.091 (-0.51) 0.889 S portfolio -0.090 (-0.51) 0.945*** (28.96) -0.153 (-1.31) -0.013 (-0.15) 0.902 G portfolio -0.543** (-2.49) 0.902*** (19.49) -0.050 (-0.36) -0.035 (-0.24) 0.865 Panel B: 2007-2012 ESG portfolio -0.251 (-0.72) 1.001*** (21.82) -0.285 (-1.34) 0.267 (1.41) 0.903 E portfolio 0.464 (1.68) 1.036*** (24.73) -0.393** (-2.26) -0.038 (-0.10) 0.896 S portfolio -0.025 (-0.09) 0.993*** (26.14) -0.286 (-1.66) 0.035 (0.27) 0.923 G portfolio -0.942*** (-2.71) 0.911*** (15.04) -0.133 (-0.71) 0.205 (1.05) 0.857 Panel C: 2013-2017 ESG portfolio 0.145 (0.87) 0.858*** (15.12) -0.105 (-1.08) -0.179* (-1.77) 0.940 E portfolio -0.043 (-0.30) 0.826*** (11.93) -0.092 (-0.69) -0.078 (-0.68) 0.898 S portfolio 0.056 (0.29) 0.791*** (13.92) 0.021 (0.20) -0.087 (-0.94) 0.913 G portfolio 0.050 (0.34) 0.935*** (18.62) 0.289*** (2.91) -0.068 (-0.54) 0.926
Notes: Dependent variable is the monthly return of the portfolios in the most left column in excess of the one
29 The coefficients of the industry components are not reported, as these are difficult to interpret. Nonetheless, it is interesting to note that for 11 of the 12 regressions, the null hypothesis that the coefficients of these four components are all equal to zero is rejected at the 10% significance level. This proves that the components indeed have explanatory power.
The most important result from Table 7 is that all alphas from the 2013-2017 regressions are not significantly different from zero. There is evidence of a growth effect for the ESG portfolio (at the 10% significance level) and a small firm effect for the Governance portfolio.
Furthermore, the alpha of the Governance portfolio is significantly negative for the full period and for the first subperiod. At first glance, this seems to imply that the costs of improving corporate governance (for instance by improving transparency and installing more control mechanisms) are larger than the benefits. However, for the period 2013-2017 this alpha is not significantly different from zero. As explained in Section 4.1, due to higher ESG coverage the results from this period are much more reliable. Therefore, there is not enough evidence to conclude that firms with a better governance structure have a lower level of financial performance.
30
Table 8
Performance change portfolios in a multifactor regression model
α Rm - Rf SMB HML R 2 Panel A: 2008-2017 Improvers portfolio -0.243 (-0.93) 0.981*** (16.27) -0.150 (-1.34) 0.076 (0.29) 0.854 Decliners portfolio 0.183 (0.57) 0.950*** (10.17) -0.163 (-1.05) 0.816** (2.61) 0.742 Panel B: 2008-2013 Improvers portfolio -0.478 (-1.44) 1.019*** (15.47) -0.312** (-2.16) 0.478 (1.30) 0.884 Decliners portfolio -0.014 (-0.03) 0.965*** (7.86) -0.235 (-0.93) 1.426*** (3.06) 0.740 Panel C: 2014-2017 Improvers portfolio 0.323 (0.73) 0.903*** (16.76) 0.394* (1.81) -0.050 (-0.33) 0.843 Decliners portfolio 0.083 (0.26) 1.073*** (15.17) 0.162 (1.20) 0.032 (0.16) 0.890
Notes: Dependent variable is the monthly return of the portfolios in the most left column in excess of the one
month Euribor rate; Rm - Rf represents the return of the MSCI Emerging Markets Index in excess of the one month Euribor rate; SMB and HML are the returns on diversified value-weighted factor-mimicking portfolios for size and value. Coefficients of industry components are not reported. 2008-2017 denotes the period between January 1, 2008 and September 1, 2017. 2008-2013 denotes the period between January 1, 2008 and January 1, 2014. 2014-2017 denotes the period between January 1, 2014 and September 1, 2017. T-statistics between parentheses, those are based on Newey-West standard errors. ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.
Table 8 shows the regression results for the change portfolios. The insignificant alphas indicate that there is no evidence of either superior or inferior financial performance of these portfolios. The full period analysis reveals that there is a relatively large number of value firms in the Decliners portfolio. However, for the period 2014-2017, this effect is not statistically significant. For the Improvers portfolio, there is evidence for some small-cap effect for the period from 2014 onwards.
32 4.3. Robustness tests
4.3.1 Transaction costs
In this section transaction costs of 25, 50 and 100 basis points are introduced to strengthen the practical relevance of the analysis. Figure 10 presents the annual portfolio turnover ratios, both for the full period and the subperiod 2014-2017. The turnover ratio is defined as follows:
𝐴𝑛𝑛𝑢𝑎𝑙 𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑟𝑎𝑡𝑖𝑜 = 𝑀𝐼𝑁(𝑎𝑛𝑛𝑢𝑎𝑙 𝑏𝑢𝑦𝑠;𝑎𝑛𝑛𝑢𝑎𝑙 𝑠𝑒𝑙𝑙𝑠)𝑁𝐴𝑉𝑏𝑒𝑔𝑖𝑛+𝑁𝐴𝑉𝑒𝑛𝑑 2
(8)
where NAVbegin and NAVend denote the Net Asset Value at January 1 and December 31, respectively.
The figure shows that the turnover ratios are particularly high for the change portfolios, whereas there is more consistency in the composition of the level portfolios. Furthermore, turnover ratios of the level portfolios are much lower for the period after 2014, when a wide range of firms with ESG ratings was available.
Fig 10. Average annual portfolio turnover ratios
Full period refers to January 2007-August 2017 for the level portfolios (ESG portfolio, Environmental portfolio, Social portfolio and Governance portfolio) and to January 2008-August 2017 for the change portfolios (Improvers portfolio and Decliners portfolio). 2014-2017 refers to January 2014-August 2017.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
33 Tables 9 and 10 report the alphas of the regression model shown in Equation 6 under different transaction costs scenarios.
Table 9
Alphas of multifactor regression model under different transaction costs scenarios (level portfolios)
No transaction costs 25 bp 50 bp 100 bp Panel A: 2007-2017 ESG portfolio -0.111 (-0.57) -0.136 (-0.71) -0.162 (-0.85) -0.214 (-1.13) E portfolio 0.185 (1.01) 0.162 (0.88) 0.138 (0.76) 0.090 (0.50) S portfolio -0.090 (-0.51) -0.117 (-0.67) -0.144 (-0.84) -0.199 (-1.17) G portfolio -0.543** (-2.49) -0.571*** (-2.64) -0.600*** (-2.80) -0.659** (-3.10) Panel B: 2007-2012 ESG portfolio -0.251 (-0.72) -0.276 (-0.79) -0.301 (-0.87) -0.351 (-1.02) E portfolio 0.464 (1.68) 0.443 (1.61) 0.422 (1.54) 0.379 (1.39) S portfolio -0.025 (-0.09) -0.049 (-0.17) -0.074 (-0.25) -0.122 (-0.43) G portfolio -0.942*** (-2.71) -0.968*** (-2.80) -0.993*** (-2.89) -1.045*** (-3.07) Panel C: 2013-2017 ESG portfolio 0.145 (0.87) 0.126 (0.76) 0.107 (0.64) 0.068 (0.40) E portfolio -0.043 (-0.30) -0.062 (-0.43) -0.082 (-0.57) -0.121 (-0.83) S portfolio 0.056 (0.29) 0.032 (0.17) 0.008 (0.05) -0.040 (-0.22) G portfolio 0.050 (0.34) 0.024 (0.16) -0.001 (-0.01) -0.054 (-0.34)
Notes: Monthly alphas from Equation 6. Dependent variable of the regression is the monthly return of the
34
Table 10
Alphas of multifactor regression model under different transaction costs scenarios (change portfolios)
No transaction costs 25 bp 50 bp 100 bp Panel A: 2008-2017 Improvers portfolio -0.243 (-0.93) -0.284 (-1.10) -0.325 (-1.27) -0.409 (-1.62) Decliners portfolio 0.183 (0.57) 0.142 (0.44) 0.101 (0.32) 0.019 (0.06) Panel B: 2008-2013 Improvers portfolio -0.478 (-1.44) -0.518 (-1.57) -0.557* (-1.70) -0.664* (-1.99) Decliners portfolio -0.014 (-0.03) -0.051 (-0.09) -0.088 (-0.16) -0.162 (-0.30) Panel C: 2014-2017 Improvers portfolio 0.323 (0.73) 0.278 (0.64) 0.233 (0.54) 0.142 (0.33) Decliners portfolio 0.083 (0.26) 0.035 (0.11) -0.014 (-0.04) -0.112 (-0.34)
Notes: Monthly alphas from Equation 6. Dependent variable of the regression is the monthly return of the
portfolios in the most left column in excess of the one month Euribor rate; 2008-2017 denotes the period between January 1, 2008 and September 1, 2017. 2008-2017 denotes the period between January 1, 2008 and September 1, 2017. 2008-2013 denotes the period between January 1, 2008 and January 1, 2014. 2014-2017 denotes the period between January 1, 2014 and September 1, 2017. T-statistics between parentheses, those are based on Newey-West standard errors. ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.
35 4.3.2 Different cut-off rates
This section presents the consequences of adjusting the selection criteria of the portfolios. For the level portfolios, portfolios consisting of the 10% and 30% firms with the highest ESG scores (instead of 20%) will be formed. For the change portfolios, portfolios of the improvers and decliners from the past six months and the past three months (instead of the past 12 months) will be formed.
Increasing the cut-off rate from 20% to 30% leads to a higher number of stocks in the portfolios, thereby increasing diversification benefits. On the other hand, the extra stocks are from firms in the top 20%-30% range of ESG scores. Since these firms have lower ESG scores than the top 20% companies, the true effect of ESG performance on firm performance might become more difficult to detect.
Similarly, considering only changes in rankings in the past six and three months means that portfolios will be smaller and diversification benefits will decrease. However, when an ESG momentum effect exists, it should become more pronounced when more recent changes are examined.
36
Table 11
Alphas of multifactor regression model under different transaction cut-off rate scenarios (level portfolios) 10% 20% 30% Panel A: 2007-2017 ESG portfolio 0.133 (0.57) -0.111 (-0.57) 0.059 (0.39) E portfolio 0.013 (0.05) 0.185 (1.01) 0.220 (1.43) S portfolio -0.032 (-0.13) -0.090 (-0.51) -0.048 (-0.31) G portfolio -0.505** (-2.10) -0.543** (-2.49) -0.349** (-2.18) Panel B: 2007-2012 ESG portfolio 0.203 (0.48) -0.251 (-0.72) 0.093 (0.36) E portfolio 0.203 (0.47) 0.464 (1.68) 0.372 (1.59) S portfolio 0.053 (0.12) -0.025 (-0.09) 0.052 (0.21) G portfolio -0.705* (-1.72) -0.942*** (-2.71) -0.545** (-2.17) Panel C: 2013-2017 ESG portfolio 0.377* (1.76) 0.145 (0.87) 0.145 (1.06) E portfolio 0.161 (0.76) -0.043 (-0.30) 0.073 (0.56) S portfolio 0.221 (1.04) 0.056 (0.29) 0.060 (0.33) G portfolio 0.026 (0.13) 0.050 (0.33) -0.014 (-0.10)
Notes: Monthly alphas from Equation 6. Dependent variable of the regression is the monthly return of the
portfolios in the most left column in excess of the one month Euribor rate; 2007-2017 denotes the period between January 1, 2007 and September 1, 2017. 2007-2012 denotes the period between January 1, 2007 and January 1, 2013. 2013-2017 denotes the period between January 1, 2013 and September 1, 2017. T-statistics between parentheses, those are based on Newey-West standard errors. ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.
37 The positive alpha from the ESG portfolio is in line with the findings of Cambridge Associates (2016) and calls for some further analysis. The results shown in Table 11 are calculated without taking transaction costs into account. However, as shown in Table 12, the alpha is no longer significantly positive after correcting for transaction costs of 25 basis points or more. Therefore, it is highly doubtful whether a trading strategy based on weighted average scores is profitable in practice.
Table 12
Alphas of multifactor regression model under different transaction costs scenarios (level portfolios)
No transaction costs 25 bp 50 bp 100 bp ESG portfolio (10% cut-off rate) 0.377* (1.76) 0.358 (1.67) 0.339 (1.58) 0.299 (1.40)
Notes: Monthly alphas from Equation 6. Dependent variable is the monthly return of the ESG portfolio with a 10%
cut-off rate in excess of the one month Euribor rate; Analyzed period is period between January 1, 2013 and September 1, 2017. T-statistics between parentheses, those are based on Newey-West standard errors. ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.
38 4.3.3 Different momentum time spans
Table 13 presents the alphas from the regression model shown in Equation 6 when only rating changes in the past six months and the past three months, respectively, are considered. Results are only presented for the period 2013-2017, since there are not sufficient rating changes in the period before 2013.12
Table 13
Alphas of multifactor regression model for rating changes in different time periods (change portfolios)
12 months 6 months 3 months
Improvers portfolio 0.490 (1.37) 0.178 (0.44) -0.146 (-0.26) Decliners portfolio -0.009 (-0.04) 0.020 (0.06) 0.249 (0.71)
Notes: Monthly alphas from Equation 6. Dependent variable of the regression is the monthly return of the
portfolios in the most left column in excess of the one month Euribor rate; Analyzed period is January 2013-August 2017; t-statistics between parentheses, those are based on Newey-West standard errors. ***, ** and * denote significance at the 1%, 5% and 10% level, respectively.
Although analyzing momentum using different time frames leads to some sign switches of the alpha coefficients, all alphas are still not significantly different from zero.
12
39
5.
Conclusions
This paper has analyzed the performance of several emerging market equity portfolios based on both absolute ESG scores and changes in ESG scores. Although data is available from 2007 onwards, this paper focuses on the results from the period after January 2013, because during the first few years only a small number of emerging market firms was rated by MSCI. The results indicate there is no significant difference between the Sharpe ratios of portfolios consisting of firms with high environmental, social, governance and weighted average scores and the Sharpe ratio of the MSCI Emerging Markets Index. Moreover, portfolios consisting of firms whose ESG score recently improved and declined, respectively, do also not show significant out- or underperformance compared to the Emerging Markets Index. These main conclusions do not change after controlling for size, value and industry exposure. There is some evidence that the risk-adjusted returns of a portfolio consisting of the 10% stocks with the highest weighted average ESG scores are significantly positive. However, when transaction costs are taken into account, this effect disappears. Furthermore, as outlined in Section 3.4.4, emerging market currencies are highly volatile and therefore investors might opt for hedging currency risks, which would lead to additional costs. Therefore, a trading strategy based on ESG scores in emerging markets appears to be not profitable in practice.
The results suggest that emerging markets may be more efficient than previously thought. Both the weighted average score and the scores of individual ESG pillars are not a predictor for future share price performance. This is in accordance with standard finance theory, which suggests that only factors which are proxies for risk should earn abnormal returns. Similarly, the finding that the momentum portfolios have insignificant alphas suggests that market participants quickly incorporate new information into share prices.
40 However, the results should be interpreted with some caution because this study has several limitations. Firstly, it focuses on a relatively short time period. This could explain the absence of statistical significant coefficients. Therefore, it is advised to repeat this analysis in the future when ESG scores of emerging market firms are available for a longer period of time.
Moreover, this paper follows the approach of selecting a subset of firms rated by MSCI to create the equity portfolios. This means that the portfolio composition depends on the rating policy of MSCI. This is especially concerning for the first period: when the rated firms share some specific characteristics, this might cause a bias in the results.13
Indeed, the results for the period from 2007 to 2012 are somewhat different from the results for the later period. Another obvious explanation for this effect is the limited number of rated firms in this early period. As a consequence, portfolios contain a relatively small number of stocks and might be exposed to unsystematic risk. Finally, it is also possible that stocks of firms with high ESG scores behave differently during periods of financial stress. Since it is impossible to distinguish between these effects, it is difficult to interpret the results for the first period.
Apart from data availability, data quality is also a major challenge for ESG research in emerging markets. Andreas Feiner, head of ESG research at Arabesque Asset Management, said he believed that “the data quality of ESG is only at 10% of where it will be in the future” (Bloomberg, 2016). Moreover, Feiner argues that the quality of ESG data is particularly low in emerging markets.
The correlations between ESG scores of different agencies are low, which is another indication of poor data quality. Chatterji et al. (2016) examine the rating methodologies of six ESG rating agencies. They find that these agencies disagree on ESG definitions (i.e. beliefs about what should be measured) and that commensurability (i.e. the extent to which agencies obtain similar results when they measure the same) is low. Therefore, improving the quality of ESG scores remains a challenge for the upcoming years. More research about correlations between ESG scores and environmental, social and governance scandals might help to assess the quality of these scores.
13
41
Acknowledgements
I would like to express my sincere gratitude to everybody who has helped me during the process of writing my Master’s thesis.
Firstly, I would like to thank TKP Investments for providing me the opportunity to write my thesis at the department Investment Strategy. This has enabled me to become part of a great team of people and obtain a valuable impression of the daily routine within a financial company. In particular, I would like to thank my supervisor at TKPI, dr. Sibrand Drijver, for his guidance and feedback during the process of my thesis writing.
Furthermore, I want to thank my supervisor from the University of Groningen, dr. Auke Plantinga. Our discussions about currency hedging were very inspiring and his comments and feedback on preliminary versions have greatly improved the quality of my work. All remaining errors in this thesis are on my own.
42
References
Auer, B., Schuhmacher, F., 2016. Do socially (ir) responsible investments pay? New evidence from international ESG data. The Quarterly Review of Economics and Finance 59, 51-62.
Balli, F., Hajhoj, H., Basher, S., Ghassan, H., 2015. An analysis of returns and volatility spillovers and their determinants in emerging Asian and Middle Eastern countries. International Review of Economics & Finance 39, 311-325.
Bauer, R., Koedijk, K., Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. Journal of Banking & Finance 29, 1751-1767.
Blitz, D., Fabozzi, F., 2017. Sin Stocks Revisited: Resolving the Sin Stock Anomaly. The Journal of Portfolio Management 44, 105-111.
Brown, S., Warner, J., 1985. Using daily stock returns: The case of event studies. Journal of financial economics 14, 3-31.
Cakici, N., Fabozzi, F., Tan, S., 2013. Size, value, and momentum in emerging market stock returns. Emerging Markets Review 16, 46-65.
Campbell, J., Lettau, M., Malkiel, B., Xu, Y., 2001. Have individual stocks become more volatile? An empirical exploration of idiosyncratic risk. The Journal of Finance 56, 1-43.
Chatterji, A., Durand, R., Levine, D., Touboul, S., 2016. Do ratings of firms converge? Implications for managers, investors and strategy researchers. Strategic Management Journal 37, 1597-1614.
Cheung, Y., Tan, W., Ahn, H., Zhang, Z., 2010. Does corporate social responsibility matter in Asian emerging markets?. Journal of Business Ethics 92, 401-413.
Dasgupta, S., Laplante, B., Mamingi, N., 2001. Pollution and capital markets in developing countries. Journal of Environmental Economics and management 42, 310-335.
Derwall, J., Guenster, N., Bauer, R., Koedijk, K., 2005. The eco-efficiency premium puzzle. Financial Analysts Journal 61, 51-63.
Diltz, J., 1995. Does social screening affect portfolio performance?. The Journal of Investing 4, 64-69.
