Do Investors Have Preferences for Socially Responsible Assets?
An Empirical Study for the Chinese Stock Market
Author: Bin Li
Student number: s3701867 Supervisor: dr. Lammertjan Dam Date: 12 June 2019
MSc Finance
Faculty of Economics and Business University of Groningen
Abstract: Investors with preferences for socially responsible assets do not only concern
themselves with risk-related or return-related features. They are also willing to accept lower returns from socially responsible investment (SRI). Most existing literature studies only the investors’ preferences for socially responsible assets in developed markets; this study fills the gap within the Chinese stock market. In this paper, I adopt the Fama-Macbeth two-step procedure to investigate the relation between cross-sectional stock returns and aggregated ESG scores. With a sample of 55 Chinese public companies from July 2009 to December 2017, I do not find any preferences for socially responsible assets in the Chinese stock market. Studies on different asset pricing models and disaggregated ESG scores support my finding.
Keywords: Preferences for socially responsible assets, Socially responsible investment
1
1.
Introduction
Socially responsible investment (SRI) is a relatively new invest ment approach that incorporates environmental (E), social (S), and governance (G) considerations when making investment decisions, and is always related to corporate social responsibility (CSR). Global Sustainable Investing Alliance (GSIA)1 reports the assets managed under SRI approach
reached $30.7 trillion at the end of 2017 in the five largest markets: Europe, the United States, Japan, Canada, and Australia/New Zealand (Global Sustainable Investing Alliance, 2018). Although the SRI concept appeared later in China, it has made great progress in the past ten years. As of August 2017, the SRI investing assets reached $7,288 million in the Chinese market (SynTao Green Finance and Aegon-Industrial Fund, 2018). Previous researches find evidence for investors preferring socially responsible assets on the global market. This study aims to investigate whether investors have preferences for such assets in the Chinese capital market.
SRI develops rapidly on a global scale, however, there is never a consensus on what drives investors to invest in socially responsible assets. Some studies argue that investors are driven by higher returns from SRI; others argue that investors are driven by their preferences for socially responsible assets because SRI generates lower returns. This type of investors does not merely focus on risks and returns but pay more attention to environmental consequences, society contribution and corporate governance. They boycott products and stocks of irresponsible companies and are willing to sacrifice financial benefits to facilitate responsible firms. Ciciretti et al. (2017) demonstrate that investors pay a price equivalent to an annual 4.8% underperformance for their preferences for responsible assets.
However, most studies focus on developed countries where SRI has become the mainstream of capital market. Although the number of studies in emerging markets has increased over recent years, any research on Chinese market is scarce. There are several possible explanations for this phenomenon: 1) As a socialist country, the Chinese government has a big influence on their economy and the development of CSR is more policy-driven. Li et al. (2016) find that Chinese companies tend to obtain government subsidies and tax preferences by taking on socially responsible activities. 2) The number of public CSR reports is small. In 2008, only 62 companies published a CSR report, though the number of CSR report increased to 851 in 2018, the proportion of companies releasing CSR reports is still
2
low. 3) It is hard to find comprehensive CSR/ESG scores for Chinese firms. Per my review, most Chinese rating agencies grade a firm’s CSR performance based on the quality of its CSR report, which, however, is not a good representation of its social responsibility performance. Thus, there are substantial possibilities to obtain biased results from analysis based on the quality of CSR reports. On the other hand, few Chinese companies are graded by international grading agencies due to the poor quality of CSR reports.
This paper studies the financial performance and CSR performance of China A-shares2 at
the firm level, covering a period from July 2009 to December 2017. I first look at risk-adjusted returns for each firm and then adopt the Fama-Macbeth two-step procedure with a 24-month rolling window to catch the time-variant betas. I do not find any significant difference in cross-sectional returns between responsible stocks and irresponsible stocks, meaning investors have no preferences for socially responsible assets in the Chinese market. The development of SRI in China still has a long way to go.
This study focuses on areas that are rarely analyzed by previous literature, mainly contributing to the following aspects. Firstly, I adopt the Thomson Reuters3 ESG score as an
independent variable instead of the score of CSR report quality or self-designed grading system to improve the accuracy and comparability of my study. Secondly, I perform analysis on individual stocks instead of portfolios or funds. By doing so, I can avoid the risk-diversification effect of portfolios and the impact of fund managers. Lastly, I analyze the relation between CSR performance and stock market performance on both overall ESG scores and disaggregated dimension scores.
The remaining section of this paper is designed as follows. In section 2, I provide an overview of previous literature related to my study. Section 3 contains an introduction to the methodology adopted in this paper and the hypothesis that needs to be tested. Followed by section 4 in which I describe the dataset of this paper. Section 5 reports the main results. Finally, in section 6, I present the conclusion.
2.
Literature Review
Many researchers investigate the performance of socially responsible assets , but cannot reach a consensus concerning whether SRI generates higher returns or not. There are three
2 A-shares are issued by Chinese registered companies and listed in China. They are denominated in
Renminbi and subscribed by individuals and institutions in China.
3 Thomson Reuters is a world-leading ESG database which started to provide ESG scores from 2002.
3
prevailing views on this topic: “positive effect”, “negative effect” and “no effect”. The lack of consensus might be related to different SRI strategies, CSR measurements, financial measurements, and political systems.
“Positive effect” means that SRI generates higher expected returns. Advocates of this view believe that good ESG performance improves the relationship among different stakeholders, which in turn improves financial performance (Preston and O'Bannon, 1997). On the other hand, SRI faces many strict restrictions when making investment decisions, but the screening reduces the risk of litigation, thus protecting the interest of the shareholder s. The evidence of "positive effect" at the portfolio level is found by Kempf and Osthoff (2007). They construct portfolios with different SRI strategies in the US market and conclude that SRI earns a prominently higher risk-adjusted return by employing positive screening or best-in-class screening. This finding is supported by Glushkov and Statman (2009). By analyzing the stock market performance of socially responsible firms from 1992 to 2007, they conclude that constructing portfolios with best-in-class approach helps investors to do both well and good, whereas negative screening leads to lower returns.
The second view is the “negative effect”, which believes that SRI generates lower risk-adjusted returns than those of conventional investments. People holding the view of "negative effect" argue that SRI forfeits the opportunities of investing in certain industries, which leads to a weakly diversified investment and relatively higher risks (Rudd, 1981). On the other hand, discarding certain high-yield stocks that do not meet SRI criteria will result in sacrifices on returns (Grossman and Sharpe, 1986). Becchetti and Ciciretti (2009) test the performance of individual stocks and portfolios between 1990 and 2003. They find individual socially responsible stocks underperform significantly with an industry dummy variable than control sample stocks in risk-adjusted returns. However, at the portfolio level, socially responsible portfolios and control sample portfolios do not generate significantly different returns under the buy-and-hold strategy. Hong and Kacperczyk (2009) study with an extended time span from 1962 to 2006 and prove that “sin stocks,” such as tobacco, alcohol and gambling stocks, have been outperforming than their peers for decades. However, these high-yield “sin stocks” are filtered out by socially responsible investors due to higher litigation risks. Ciciretti et al. (2017) study on an updated sample period of July 2005 to June 2014, reveals that annual return from SRI is 4.8% lower on average.
4
significantly lower risk-adjusted returns than their counterparties in most countries. This finding is also valid in the UK where Humphrey et al. (2012) demonstrate that returns from responsible firms and irresponsible firms are not significantly different. Bauer et al. (2006) also prove this pattern in the Australian market during the period of 1992 to 2003. From the perspective of the global market, Auer et al. (2015) find that SRI produces a similar return to the market portfolio in the US and Asia-Pacific region.
In addition, some scholars propose that analysis based on aggregated ESG/CSR measurement would produce biased results because of the interaction between heterogeneous decomposed dimensions. Scholtens and Zhou (2008) observe different relationships between financial performance and different dimensions. Specifically, strengths in corporate governance and product characteristics have positive impacts on financial performance, whereas strengths in employee relationships, workforce diversity, and human rights produce a negative influence on financial performance. Their research gives me inspiration for the robustness check of this paper.
Apart from the comparison of stock market performance, some studies try to compare the financial performance of socially responsible firms and irresponsible firms through accounting data. The study of Dam and Scholtens (2015) reveals that socially responsible firms achieve a higher market-to-book ratio and return on assets. Amini and Bianco (2017) study the relationship between CSR and financial performance for six Latin American countries, concluding that engaging in CSR improves firms’ turnover in less industrialized economies such as Bolivia and Colombia. Mahoney and Robert (2007) find no significant evidence for the relationship between the aggregated measurement of companies' CSR performances and return on assets or return on equity. But they observe a significant relationship between environmental performance and return on assets. Nollet et al. (2016) investigate the performance of S&P500 firms in the period from 2007–2011 with both a linear model and a non-linear model. The linear model reveals a negative correlation between CSR performance and return on capital, however, the non-linear model discovers a U-shaped relation between CSR performance and return on assets/return on capital.
5
conducted by Maqbool and Zameer (2017) provides supportive evidence for the positive impact on profitability and stock return from CSR practices. On the contrary, the research on Nigerian listed companies conducted by Amran and Usman (2015) finds a negative relation between environmental disclosure and financial performance.
As the largest emerging market in the world, the performance of SRI in China also draws the attention of academic researchers. Zhu et al. (2012) find socially responsible indices perform better than the market index in both Shanghai Stock Exchange (SSE) and Shenzhen Stock Exchange (SZSE). But the thirty-month time span and flawed construction method of socially responsible index weaken the robustness of their results. In the construction industry, Xiong et al. (2016) find that better CSR performance on supplier and client dimension results in better financial performances; in turn, better financial performance results in a better outcome in environmental protection. On the other side, Wu and You (2018) prove the "negative effect" in the Chinese stock market through the investigation of social responsibility performance and expected stock returns for Chinese A-shares from 2009 to 2016. They conclude that good socially responsible performance results in lower expected returns, while the different CSR dimensions have different impacts on financial performance. However, Zhang (2014) finds two highly concentrated SRI funds generate insignificant risk-adjusted returns compared with the broad market index, but two well diversified SRI funds generate higher returns. She infers that high-level industry diversification helps SRI funds to outperform than market index, but this inference needs to be tested with more SRI funds. Kao et al. (2018) argue that due to different ownership structures and government intervention, they cannot find any significant connection between CSR performance and financial performance in state-owned enterprises, but they observe positive relations in non-state-owned enterprises.
3.
Methodology and Hypothesis
The Decennial Report on the Responsible Investment in China (SynTao Green Finance and Aegon-Industrial Fund, 2018) reports that 57% of institutional investors take CSR performance into consideration even though it generates lower returns. This percentage declines to 34.3% in individual investors. Based on this survey, I hypothesize Chinese investors do not have preferences for socially responsible assets and that there is no significant relationship between social responsibility performance and stock market performance.
6
company. Since Yang and Chen (2003), Wu and Xu (2004), He (2012) and Xiong and Xu (2009) have proven the validity of the following three asset pricing models in the Chinese stock market: 1) Capital Asset Pricing Model (CAPM - Sharp,1964; Lintner,1965 and Mossin,1966), 2) Fama-French three-factor model (FF3 - Fama and French,1992;1993) and 3) Fama-French-Carhart four-factor model (FFC - Carhart, 1997), therefore, I employ the above models in this paper directly. The models for time-series regressions are expressed as follows:
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝛼𝑖+ 𝛽𝑖𝑚𝑘 𝑀𝐾𝑇𝑡+ 𝜇𝑖,𝑡 (1-CAPM)
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝛼𝑖+ 𝛽𝑖𝑚𝑘 𝑀𝐾𝑇𝑡+ 𝛽𝑖𝑠𝑆𝑀𝐵𝑡+ 𝛽𝑖ℎ𝐻𝑀𝐿𝑡+ 𝜇𝑖,𝑡 ( 1 - F F 3 )
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝛼𝑖+ 𝛽𝑖𝑚𝑘 𝑀𝐾𝑇𝑡+ 𝛽𝑖𝑠𝑆𝑀𝐵𝑡+ 𝛽𝑖ℎ𝐻𝑀𝐿𝑡+ 𝛽𝑖𝑚𝑀𝑂𝑀𝑡+ 𝜇𝑖,𝑡 (1-FFC)
where 𝑅𝑖,𝑡 is the return on stock 𝑖 in month 𝑡; 𝑅𝑓𝑡 is the monthly risk-free interest rate converted from PBOC benchmark interest rate of three-month deposit; 𝛼𝑖 is the risk-adjusted return of stock 𝑖; 𝑀𝐾𝑇𝑡 is the market risk factor in month t, which is calculated by monthly return on market portfolio minus monthly risk-free rate; 𝑆𝑀𝐵𝑡 denotes the size risk factor in month 𝑡, which is the difference in monthly return between small and big capitalization portfolios; 𝐻𝑀𝐿𝑡 denotes the value risk factor in month 𝑡, which is the difference in monthly returns between low B/M and high B/M portfolios; 𝑀𝑂𝑀𝑡 denotes the momentum risk factor in month 𝑡, which is the difference in returns between high and low prior return portfolios (Kenneth R. French Data Library); 𝜇𝑖,𝑡 is the error term for stock 𝑖 in month 𝑡.
In this step, I estimate N time-series regressions of excess returns against corresponding risk factors at the firm level to capture risk exposures for each stock, where N is the number of companies included in the sample. Instead of a widely employed portfolio analysis, I conduct this study at the firm level. As my research sample is relatively small, there are potentially biased estimations from constructing portfolios. Besides, portfolios might diversify the risks of an individual stock.
7
towards heteroscedasticity and autocorrelation; though ordinary least squares (OLS) estimation gives unbiased parameters, the standard errors are always biased, leading to inaccurate t-statistics. To correct biased standard errors caused by autocorrelation and heteroscedasticity in error term, I employ the Newey-West estimator with a six-month lag4
(Newey and West, 1987).
In the second step of the Fama-Macbeth procedure, I estimate T-235 cross-sectional
regressions on excess returns against rolling window betas and aggregated ESG scores. The models for this step are expressed as follows:
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,t+ 𝜆S,t 𝑆𝑖,𝑡+ 𝜆𝑚𝑘,𝑡 𝛽𝑖𝑚𝑘,t−1 + 𝜀𝑖,𝑡 (2-CAPM) 𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,𝑡 + 𝜆𝑆,𝑡 𝑆𝑖,𝑡+ 𝜆𝑚𝑘 ,𝑡 𝛽𝑖𝑚𝑘,𝑡−1 + 𝜆𝑠,𝑡 𝛽𝑖𝑠,𝑡−1+ 𝜆ℎ,𝑡 𝛽𝑖ℎ,𝑡−1 + 𝜀𝑖,𝑡 (2-FF3) 𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,𝑡 + 𝜆𝑆,𝑡 𝑆𝑖,𝑡+ 𝜆𝑚𝑘 ,𝑡 𝛽𝑖𝑚𝑘,t−1 + 𝜆𝑠,𝑡 𝛽𝑖𝑠,𝑡−1+ 𝜆ℎ,𝑡 𝛽𝑖ℎ,𝑡−1 + 𝜆𝑚,𝑡 𝛽𝑖𝑚,𝑡−1 + 𝜀𝑖,𝑡 (2-FFC) where 𝜆0,𝑡 is the intercept at time 𝑡; 𝑆𝑖,𝑡 is the aggregated ESG score for stock 𝑖 at time 𝑡; 𝛽𝑖𝑚𝑘,𝑡−1 , 𝛽𝑖𝑠,𝑡−1, 𝛽𝑖ℎ,𝑡−1 and 𝛽𝑖𝑚,t−1 are the one-month lagged betas of market factor, size factor, value factor and momentum factor for stock 𝑖 from time-series regressions; 𝜆S,t , 𝜆𝑚𝑘 ,𝑡 , 𝜆𝑠,𝑡 , 𝜆ℎ,𝑡 and 𝜆𝑚,𝑡 are the risk premiums for Score, market, size, value and momentum factor at time 𝑡; 𝜀𝑖,𝑡 is the error term.
From this step, I can also obtain time-series averages of intercepts and risk premiums for the period of July 2011 to December 2017.
To test the robustness, I repeat the Fama-Macbeth two-step regression on pillar scores and factor loadings. By doing so, I can eliminate the mutual effects between each pillar score. The specifications for the robust test are shown as below:
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,t+ 𝜆ENV,t 𝐸𝑁𝑉𝑖,𝑡+𝜆SOC,t 𝑆𝑂𝐶𝑖,𝑡+𝜆CGV,t 𝐶𝐺𝑉𝑖,𝑡+ 𝜆𝑚𝑘,𝑡 𝛽𝑖𝑚𝑘,t−1 + 𝜀𝑖,𝑡 (R-CAPM)
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,t+ 𝜆ENV,t 𝐸𝑁𝑉𝑖,𝑡+𝜆SOC,t 𝑆𝑂𝐶𝑖,𝑡+𝜆CGV,t 𝐶𝐺𝑉𝑖,𝑡+ 𝜆𝑚𝑘 ,𝑡 𝛽𝑖𝑚𝑘,𝑡−1 +
𝜆𝑠,𝑡 𝛽𝑖𝑠,𝑡−1+ 𝜆ℎ,𝑡 𝛽𝑖ℎ,𝑡−1 + 𝜀𝑖,𝑡 (R-FF3)
4 I also tried one-month lag, three-month lag and four-month lag for Newey-West adjustment,
estimations with six-month lag generate the smallest standard errors and the biggest t-statistics.
5 As I set a 24-month rolling window in time-series regression, the first 23 months are excluded from
8
𝑅𝑖,𝑡− 𝑅𝑓𝑡 = 𝜆0,t+ 𝜆ENV,t 𝐸𝑁𝑉𝑖,𝑡+𝜆SOC,t 𝑆𝑂𝐶𝑖,𝑡+𝜆CGV,t 𝐶𝐺𝑉𝑖,𝑡+ 𝜆𝑚𝑘 ,𝑡 𝛽𝑖𝑚𝑘,t−1 +
𝜆𝑠,𝑡 𝛽𝑖𝑠,𝑡−1+ 𝜆ℎ,𝑡 𝛽𝑖ℎ,𝑡−1 + 𝜆𝑚,𝑡 𝛽𝑖𝑚,𝑡−1 + 𝜀𝑖,𝑡 (R-FFC) where 𝜆0,𝑡 is the intercept at time 𝑡; 𝐸𝑁𝑉𝑖,𝑡, 𝑆𝑂𝐶𝑖,𝑡 and 𝐶𝐺𝑉𝑖,𝑡 are the corresponding pillar scores of environmental, social and governance for stock 𝑖 at time 𝑡 ; 𝛽𝑖𝑚𝑘,𝑡−1 , 𝛽𝑖𝑠,𝑡−1, 𝛽𝑖ℎ,𝑡−1 and 𝛽𝑖𝑚,t−1 are the one-month lagged betas of market factor, size factor, value factor and momentum factor for stock 𝑖 from the time-series regression; 𝜆S,t , 𝜆𝑚𝑘 ,𝑡 , 𝜆𝑠,𝑡 , 𝜆ℎ,𝑡 and 𝜆𝑚,𝑡 are the risk premiums for Score, market, size, value and momentum factor at time 𝑡; 𝜀𝑖,𝑡 is the error term.
4.
Data and Descriptive Statistics
I collect my research data mainly from three databases. From DataStream, I retrieve monthly Return Index (RI) and yearly Thomson Reuters ESG scores (Score). The cash dividends are considered to be reinvested when calculating stock returns based on RI. Thomson Reuters is a world-leading ESG database that grade ESG performance for more than 7,000 public companies around the world based on firm reports and present ESG scores in percentage. It rates firms' ESG performance on over 400 metrics to reflect companies' ESG performance, commitment, and effectiveness. The acquired 400 plus metrics are divided into10 different themes and then grouped into 3 pillars: environmental (E), social (S) and corporate governance (G). The environmental pillar includes dimensions of resource use, emissions, and innovation, accounting for 34% in the aggregated ESG score. The social category covers workforce, human rights, community, and product responsibility, weighing 35.5% in the overall ESG score. The corporate governance pillar assesses management, shareholders and CSR strategy and accounting for 30.5% in the aggregated score. (Thomson Reuters ESG Scores, 2018).
9
months). As the China Security Regulation Committee (CSRC) requires public companies to issue their annual financial report and CSR report before the end of April in the following year, setting the sample period starts from July also ensures all the financial information and CSR information of last year are priced in. Up to December 2017, there are 3575 A-shares publicly traded on the Chinese stock market. I select stocks based on the following criteria: 1) monthly financial data are accessible for the entire period;
2) annually ESG scores are continuously available from 2011.
I then obtain a balanced panel of 55 firms which contains 5610 observations in total.
According to CSRC, the listed companies are divided into 19 industries6. The companies
included in this study generally fall under 8 industries, indicating the sample is not well diversified. Table 1 shows the number of firms, proportion, average aggregated ESG score and average excess return for each industry. It is clear that the composition of each industry varies significantly. The manufacturing industry includes 18 companies, consisted of the largest part of the research sample. The financial industry is the second largest component which consists of 13 companies, accounting for 23.64 % of the research sample. In contrast to the manufacturing industry, the wholesale and retail industry is the smallest component of the sample, including only one company.
Table 1: Descriptive Statistics for Industry Classification
Industry No. of Firms Prop (%) Avg. ESG Score (%) Avg. Excess Return (%) Construction 4 7.27 23.84 0.73 Finance 13 23.64 50.46 0.69 Manufacturing 18 32.73 35.44 0.60 Mining 8 14.55 54.54 -0.09 Real estate 2 3.64 29.41 1.24
The production and supply of electricity, heat,
gas, and water 3 5.45 36.41 0.10
Transportation, storage and post 6 10.91 45.90 0.52
Wholesale and retail 1 1.82 38.43 1.07
Total 55 100 41.95 0.53
6 Industry Classification Results for Listed Company in the 4th Quarter of 2017
10
The table shows the industry classification based on the ‘Industry Classification Results for Listed Company in the 4th Quarter of 2017’ released by CSRC. It also reports the number of firms,
proportion, average yearly aggregated ESG score and average monthly excess return for each industry. The ESG scores and stock returns are retrieved from DataStream. Except for the number of firms, other items are presented in percentage. Values above average are shown in bold.
The average overall ESG score of the eight industries ranges between 23.84% and 54.54%, with a mean value of 41.95%. Among the eight industries, the finance, mining and transportation, storage and post industries report average overall ESG scores exceeding the average overall ESG score of the whole sample. The reason might be that these industries have higher ESG awareness and are prone to take more ESG opportunities. Another reason might be that some industries are more policy-driven than others. The supportive policies for “Green Finance” and “Beautiful China” stimulate some industries to rank higher importance to corporate social responsibility. Unexpectedly, the mining industry achieves the highest aggregated ESG score. A possible explanation would be that the mining industry is always regarded as a highly polluting industry and is placed under a higher standard regulation. The lowest overall ESG score is reported by the construction industry. In the past decades, the fast-expanding construction activities damage the ecological environment and encroach agricultural land, resulting in severe environmental and social problems (Zhao et al., 2012). There are also severe employee relationship problems in the construction industry. Construction workers work 9.5 to 10 hours per day without overtime pay; only 17.4% of them sign labor contracts with construction companies (Zhihu, 2017). Due to the backward understanding of corporate social responsibility, China's construction enterprises simply equate CSR with donations, ignoring environmental damage, employee rights, construction safety and product quality.
11
exacerbating the hardship.
Table 2 demonstrates the descriptive statistics of average excess return and Fama-French-Carhart risk factors over the time span. The average excess return of the sample is 0.53%, while the median excess return is -0.09%, reflecting the significant difference between high-yielding stocks and low-yielding stocks. We can also infer that more than half of the companies in my sample gain excess returns lower than 0.53%. Among the four risk factors, the size factor has the highest mean value, whereas the value factor has the lowest mean value.
Panel B presents the correlation coefficients between risk factors. As shown in the table, every risk factor is significantly correlated with another. For example, the market factor has a positive correlation with the size factor but a negative correlation with the value and momentum factor. In general, the correlation between each factor is low, indicating the rationality of the models at a preliminary stage.
Table 2: Descriptive Statistics for Monthly Excess Return and Risk Factors
Panel A Panel B (1) Obs (2) Mean (3) Med (4) Min (5) Max (6) SD (7) MKT (8) SMB (9) HML (10) MoM Excess Return 5555 0.53 -0.09 -48.93 116.84 11.14 - - - - MKT 102 0.50 1.02 -23.79 18.80 7.13 1.00 SMB 102 1.10 0.97 -22.50 21.38 5.29 0.14 *** 1.00 HML 102 -0.04 0.10 -15.51 16.38 3.68 -0.10 *** -0.82 *** 1.00 MoM 102 0.55 0.80 -19.06 13.95 4.89 -0.12 *** 0.24 *** -0.30 *** 1.00 The table shows descriptive statistics for monthly excess return and monthly Fama-French-Carhart risk factors: MKT (the excess return of the market), SMB (the size risk factor), HML (the value risk factor), MoM (the momentum risk factor) and the correlation coefficients between risk factors. The monthly risk factors are collected from CSMAR and are presented in percentage.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
12
mechanism (present in Appendix A), the scores can be converted into 12 grades, from D- to A+. The overall ESG scores in the sample are distributed from grade D- to grade A-, but the average overall ESG score of the entire sample is located in grade C+. Comparing with the score in developed markets, the ESG score for Chinese companies is rather low, indicating that although these companies reported their CSR report at an early stage, they did not put enough effort to work on the environment, society and corporate governance when pursuing profit. Taking a look at the decomposed pillar scores, the environmental pillar has the highest average score, followed by the social score. However, the governance score has the lowest mean value, which can be explained from two aspects: on one hand, many Chinese companies' understanding of corporate social responsibility is still at the stage of environmental protection and donation; on the other hand, achievement on environment and society contributes more to building a company’s reputation, therefore, companies allocate more resources to these pillars. Panel B describes the Pearson's correlation coefficients between aggregated ESG score and pillar scores. The overall ESG score is positively correlated to the environmental score, social score and governance score and their correlations are significant at 1% level. Among the three pillars, the social pillar has the strongest explanatory power on overall ESG score while the governance pillar has the lowest explanatory power. This is in line with the weights allocated to the three pillars by Thomason Returns. It is also worth noting that the standard deviations of scores are much higher than risk factors.
Table 3: Firm-Level Descriptive Statistics for Monthly ESG Scores
Panel A Panel B (1) Obs (2) Mean (3) Med (4) Min (5) Max (6) SD (7) Score (8) E (9) S (10) G Score 4290 42.24 40.49 4.81 80.99 18.24 1.00 E 4290 47.33 45.87 8.52 94.40 27.67 0.79*** 1.00 S 4290 42.98 38.95 4.14 95.77 27.50 0.85*** 0.84*** 1.00 G 4290 28.07 24.21 1.64 92.13 21.50 0.65*** 0.52*** 0.62*** 1.00 Panel A presents descriptive statistics for the aggregated ESG score (Score) and the three pillar scores: environmental score(E), social score(S) and corporate governance score(G) from July 2011 to December 2017. Panel B shows the correlation between ESG score and its three pillars. All scores are obtained from Thomson Reuters and presented in percentage.
13
Figure 1 gives a more visualized illustration of the annually cross-sectional average ESG scores. The overall average ESG score was consistent in the first three years, then rose sharply from 2014 and reached 51% in 2017. The environmental score increased every year, but the social score and governance score was more variant prior to 2014. From 2014, the environmental score and the social score rose rapidly, and the governance score started growing at a relatively slower pace, enlarging the gap between environmental score and governance score. 2014 appeared to be a turning point for the development of social responsibility. In that year, the PBOC and the United Nations Environment Program (UNEP) Sustainable Finance Project jointly launched the Green Finance Working Group aiming to stimulate and guide the development of green finance in China. The Green Finance Working Group is an important component of the Green Finance Professional Committee (GFC), which was established in April 2015. This non-profit professional organization builds a bridge in the financial sector, green industry and government, and plays a crucial role in the landing of green financial support policies. Their efforts stimulate Chinese companies to engage in practices related to the environment, society, and corporate governance, leading to the improvement in ESG scores.
Figure1: Annually Cross-sectional Average ESG Scores for Chinese firms
The figure displays the annually aggregated ESG scores and environmental, social and governance pillar scores for Chinese firms from 2011 to 2017. The scores, collected from Thomson Reuters, are presented in percentages. The bars represent the pillar score of environmental, social and corporate governance; the broken line represents the aggregated ESG score.
Figure 2 displays the average overall ESG score, average excess return across the time span and the overall trend of average excess returns from low-score companies to high-score.
0 10 20 30 40 50 60 70 2011 2012 2013 2014 2015 2016 2017 Sc or e (% ) Year
Annually Average ESG Scores
14
I rank the ESG score in ascending order from left to right. The average aggregated ESG score climbs from 7.54% to 75.47%, reflecting the huge difference in CSR performance. The average excess return of each firm varies between -1.0% and 2.40%, but both the peak point and bottom point appear in low-score firms. With the rising of ESG score, the fluctuation in excess returns narrows down gradually. Overall, the trend line reflects the decline in average excess returns from the low-score company to the high-score. This trend is contracting to my hypothesis. There seem to be preferences for socially responsible stocks among Chinese investors.
Figure 2: Time-series Average Excess Return and Average ESG Score at Firm Level
The figure shows the average aggregated ESG scores (%) and the average excess returns (%) over the entire period. It also reflects the overall trend in average excess returns from the low-score company to the high-score. The bars denote the overall ESG scores, the solid line denotes the average excess return, and the dotted line is the overall trend of average excess return. The left vertical axis represents the scale of the score, the right vertical axis stands for the scale of excess return. Returns are retrieved from DataStream; scores are collected from Thomson Reuters.
5.
Results
This section presents the risk-adjusted returns for each firm and relationships between cross-sectional returns, betas and ESG scores (both aggregated ESG scores and disaggregated ESG scores).
First, I estimate risk-adjusted returns for a set of asset pricing models, namely Capital Asset Pricing Model (CAPM), the Fama French three-factor model (FF3) and the
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 3.00 0 10 20 30 40 50 60 70 80 1 3 5 7 9 1113151719212325272931333537394143454749515355 A vg . E x c e ss R e tu rn A vg . E S G S c o re Firm
Avg. Excess Return and Avg. ESG Score
15
Carhart-Fama-French four-factor model (FFC). Started with the market factor, the size factor, value factor and momentum factor are added to the CAPM model step by step. Table 3 presents the estimated risk-adjusted returns and corresponding t-statistics without a rolling window. Firms are ranked on time-series average overall ESG score in ascending order. From this table, I observe a slight downtrend in risk-adjusted returns from socially irresponsible firms to responsible firms. Most risk-adjusted returns are insignificantly different from zero, suggesting the models have good explanatory power in excess return. However, there are six firms with statistically significant nonzero risk-adjust returns in the CAPM model (Panel A, column1). The number of statistically significant nonzero risk-adjust return increases to 13 in the FF3 model (Panel B, column3) and 15 in the FFC model (Panel C, column5). This result indicates that though CAPM, FF3 and FFC model has good explanatory power, they cannot completely conclude the differences in excess returns among high-score firms and low-score firms. The significant risk-adjust returns vary from -0.88% to 2.11%, -2.05% to 2.48% and -2.09% to 2.37% in the CAPM model, FF3 model and FFC model respectively, and most are produced by high-score firms, suggesting that investors might have preferences for responsible stocks. The average risk-adjusted return increases with more risk factors included, suggesting the more risks investors bear, the more financial compensation they come to expect.
16
Table4: Risk-adjusted Returns for Each Firm from July 2009 to December 2017
Panel A Panel B Panel C
17
Table4: Risk-adjusted Returns for Each Firm from July 2009 to December 2017 (continued) C+ 47.71 0.88 [1.45] 2.03*** [3.89] 2.09*** [3.98] B- 50.12 0.96 [0.96] 1.02 [1.03] 0.98 [1.03] B- 50.31 -0.66 [-0.88] -0.63 [-0.85] -0.68 [-0.91] B- 52.16 0.07 [0.22] 0.17 [0.58] 0.18 [0.60] B- 53.49 -0.62 [-0.62] -0.98 [-1.16] -1.29* [-1.82] B- 55.87 0.79 [1.03] 0.3 [0.33] 0.22 [0.26] B- 56.00 0.47 [0.71] 0.65 [0.93] 0.87 [1.35] B 58.45 0.24 [0.46] 0.01 [0.03] 0.21 [0.42] B 59.27 -0.88* [-1.82] -1.04* [-1.90] -1.10** [-2.03] B 59.91 -0.44 [-0.51] -0.75 [-0.82] -0.91 [-1.05] B 62.57 -0.41 [-0.89] -0.47 [-1.36] -0.56 [-1.50] B 63.94 0.09 [0.21] 0.1 [0.25] 0.15 [0.38] B 64.09 0.97 [1.04] 1.27 [1.11] 1.02 [0.95] B+ 67.07 -0.66** [-2.01] -0.28 [-0.85] -0.28 [-0.81] B+ 68.01 0.66 [1.29] 1.05** [2.19] 1.07** [2.11] B+ 71.67 -0.25 [-0.42] 0.09 [0.18] 0.06 [0.12] A- 75.10 0.40 [1.20] 0.65*** [2.72] 0.73*** [2.93] A- 75.47 0.33 [1.04] 0.58** [2.24] 0.58** [2.21] C+ 41.95 0.17 0.27 0.28
This table reports the firm-level risk-adjusted returns (α) and corresponding t-statistics (τ[α]) from time-series regressions (without rolling window) for CAPM, FF3 and FFC model. Firms are ranked from low to high on time-series average of overall ESG scores. The scores, obtained from Thomson Reuters, are converted into letter grades following the methodology proposed by Thomson Reuters (presented in Appendix A). The last row shows the average score for the entire sample, the grade of average score and average risk-adjusted returns, which are presented in bond.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
18
returns.
The premium on Score has a positive sign in the CAPM model but a negative sign in the FF3 and FFC model. It varies in a small range between -0.01 and 0.00, eliminating the concern of multicollinearity problem for ESG score. However, the relation between aggregated ESG score and average cross-sectional return is insignificant in all models, indicating that the overall ESG score does not help in explaining differences in cross-sectional returns between socially responsible firms and irresponsible firms. Therefore, I cannot reject the hypothesis that investors do not have preferences for socially responsible stocks in the Chinese stock market.
The explanatory power of the model in the Fama-Macbeth cross-sectional regression improves from 0.10 in the CAPM model to 0.21 in the FFC model. Compared with the results of Ciciretti et al. (2017), the R2 in this paper is low, which might be caused by omitted risk
factors.
Table 5: Results of Fama-Macbeth Cross-sectional Regression on Aggregated ESG Score for China, 2011-2017
Panel A Panel B Panel C
CAPM FF3 FFC 𝜆0 1.02 1.31 1.36* τ [𝜆0] [1.15] [1.49] [1.68] 𝜆𝑆 0.00 -0.01 -0.01 τ [𝜆𝑆] [0.18] [-0.79] [-0.84] 𝜆𝑚𝑘 -0.35 -0.32 -0.37 τ [𝜆𝑚𝑘] [-0.63] [-0.54] [-0.66] 𝜆𝑠 -0.43 -0.47 τ [𝜆𝑠] [-1.08] [-1.16] 𝜆ℎ 0.58*** 0.66*** τ [𝜆ℎ] [2.46] [2.62] 𝜆𝑚 -0.44 τ [𝜆𝑚] [-0.96] R2 0.10 0.18 0.21
19
Fama-Macbeth cross-sectional regressions on aggregated ESG scores. 𝜆0 is the time-series average of intercept, 𝜆𝑆 is the time-series average of premium on the aggregated ESG score, 𝜆𝑚𝑘 , 𝜆𝑠 , 𝜆ℎ and 𝜆𝑚 are the time-series average of risk premiums for market, size, value and momentum factor. R2 is the time-series average of Newey-West corrected R2 of each cross-sectional regression.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
For the robust test, I also follow the Fama-Macbeth two-step procedure; but in the second step, I estimate cross-sectional regressions on ESG pillar scores instead of the overall ESG score. Table 6 reports the time-series average of intercepts, risk premiums for ENV, SOC, CGV, market, size, value and momentum factors. The results of the robust test are almost the same as regression results on aggregated ESG score. The intercepts are positive, but insignificant in all models. Market beta, size beta and momentum beta have negative relations with cross-sectional returns, but the relations are insignificant in all models. Value betas have significantly positive impacts on cross-sectional returns in FF3 and FFC model. Although the premiums on value factor slightly decrease in the robust test, the t-statistics increase, especially for the FF3 model. In line with previous studies, different ESG dimensions have different effects on cross-sectional returns, but all effects are not significant. The environmental dimension is positively correlated with cross-sectional returns, nevertheless, the social dimension is negatively correlated with cross-sectional returns. The premium on corporate governance score has a negative sign in CAPM model and FF3 model but has a positive sign in FFC model. It is also noted that the R2 is improved in the robust test model,
suggesting the models with disaggregated ESG scores have stronger explanatory power of cross-sectional excess returns.
To summarize, the robust test supports my findings from regressions on aggregated ESG scores. I then arrive at the conclusion that the investors have no preferences for socially responsible assets in the Chinese stock market.
20
financial performance. Last but not least, SRI is not the mainstream in the Chinese capital market, both companies and investors do not pay enough attention to ESG engagement. As a result, the ESG information for Chinese companies is very limited. Furthermore, ESG performance is assessed on an annual basis while financial performance is evaluated monthly. The flat ESG scores might be the main cause of insignificant results.
Table 6: Results of Fama-Macbeth Cross-sectional Regression on Disaggregated ESG Score for China, 2011-2017
(1) (2) (3) CAPM FF3 FFC 𝜆0 0.92 0.97 0.98 τ [𝜆0] (1.16) (1.28) (1.41) 𝜆𝐸𝑁𝑉 0.01 0.01 0.01 τ [𝜆𝐸𝑁𝑉] (0.88) (0.65) (0.92) 𝜆𝑆𝑂𝐶 -0.00 -0.01 -0.01 τ [𝜆𝑆𝑂𝐶] (-0.40) (-0.55) (-0.72) 𝜆𝐶𝐺𝑉 -0.00 -0.00 0.00 τ [𝜆𝐶𝐺𝑉] (-0.42) (-0.22) (0.13) 𝜆𝑚𝑘 -0.31 -0.29 -0.35 τ [𝜆𝑚𝑘] (-0.78) (-0.65) (-0.80) 𝜆𝑠 -0.34 -0.35 τ [𝜆𝑠] (-1.06) (-1.02) 𝜆ℎ 0.48*** 0.55*** τ [𝜆ℎ] (3.06) (2.75) 𝜆𝑚 -0.40 τ [𝜆𝑚] (-0.88) R-squared 0.13 0.21 0.23
21
R2 is the time-series average of Newey-West corrected R2 of each cross-sectional regression. *** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
6. Conclusion
SRI has developed rapidly in the Chinese capital market in the past ten years, but whether or not Chinese investors have preferences for socially responsible assets was rarely discussed. My study fills the gap by examining the relation between cross-sectional returns and ESG performance of Chinese A-shares at firm level. I discard companies that went public after July 2009 and companies with accessible ESG scores less than 7 years and finally reach a study sample of 55 firms with a period from July 2009 to December 2017. I assume investors with preferences for socially responsible assets are willing to sacrifice their returns for SRI. Following the Fama-Macbeth two-step procedure, I estimate time-series regressions with a 24-month rolling window and cross-sectional regressions successively on CAPM model, Fama-French three-factor model and Fama-French-Carhart model. Since Fama-Macbeth two-step procedure only corrects cross-sectional autocorrelation, I adopt Newey-West estimator to solve the problem of autocorrelation and heteroscedasticity in time-series regressions. The regression results show that the aggregated ESG score has a negative but insignificant relation with cross-sectional stock returns, suggesting investors do not have preferences for socially responsible assets in Chinese stock market. This result might be caused by limited data, unsound stock market and lagged awareness of social responsibility in China. Robust test on disaggregated ESG scores supports my result.
22
more non-professional investors.
This study contributes to the existing literature in three aspects. Firstly, I conduct research at the firm level. Study on individual stock eliminates pote ntial influence from the fund manager and risk-diversification effect from portfolios. Secondly, I take ESG scores from Thomson Reuters, a global leading ESG database. It has a comprehensive and mature scoring mechanism, providing convincing and comparable ESG scores. Lastly, I also estimate Fama-Macbeth cross-sectional regressions on disaggregated ESG scores. Employing environmental, social and corporate governance pillar scores as independent variables can avoid the interaction between each pillar. Consistent with previous literature, different dimensions have different relations with cross-sectional returns, but the relations are insignificant in the Chinese stock market. Regression results from disaggregated dimension scores support my findings from aggregated ESG scores.
However, there are some limitations of my study. First, the study sample is small and poorly diversified. Whether the result holds for large and well-diversified samples needs to be further tested. Second, in China, state-owned companies are a special group because they receive more government intervention and have different goals rather than profit maximizing. Different ownership structures might result in different relationships between socially responsible performance and stock market performance. Taking ownership into consideration when investigating socially responsible preferences in China will be an interesting topic .
7. Appendix
Appendix A: Grading Criteria
Score Range (%) Grade
0.0 <= score <= 8.3 D - 8.3 < score <= 16.7 D 16.7 < score <= 25.0 D + 25.0 < score <= 33.3 C - 33.3 < score <= 41.7 C 41.7 < score <= 50.0 C + 50.0 < score <= 58.3 B - 58.3 < score <= 66.7 B 66.7 < score <= 75.0 B + 75.0 < score <= 83.3 A - 83.3 < score <= 91.7 A 91.7 < score <= 100 A +
23
graded D-, with a score above 91.7% are graded A+. Source: Thomson Reuters ESG Scores, 2018.
Appendix B: Descriptive Statistics for factor loadings
Panel A: Descriptive Statistics for Betas
Obs Mean Med Min Max SD
β_MKT 4290 1.10 1.04 -0.22 3.24 0.47 β_SMB 4290 -0.09 -0.05 -3.16 2.88 0.73 β_HML 4290 0.29 0.32 -4.15 4.82 1.13 β_MOM 4290 -0.01 -0.06 -2.38 3.57 0.59 Panel B: Correlation CAPM Score β_MKT Score 1.00 β_MKT -0.18*** 1.00 FF3 Score β_MKT β_SMB β_HML Score 1.00 β_MKT -0.11*** 1.00 β_SMB 0.05*** -0.22*** 1.00 β_HML 0.31*** 0.05*** 0.70*** 1.00 FFC Score β_MKT β_SMB β_HML β_MOM Score 1.00 β_MKT -0.11*** 1.00 β_SMB 0.02 -0.30*** 1.00 β_HML 0.28*** -0.02 0.69*** 1.00 β_MOM 0.02 0.12*** -0.06*** 0.03** 1.00
24 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 20 40 60 80 β_ M K T
Aggregated ESG Score
MKT Beta
ESG score is significantly correlated to all factor loadings in the CAPM model and FF3 model, but the significance of size beta declines in the FFC model. Additionally, the coefficient between score and momentum beta is insignificant in the FFC model.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1.
Appendix C: Relationship Between Average Overall Score and Betas / Risk-adjusted Returns
Panel A: CAPM Model
Panel B: FF3 Model -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 0 20 40 60 80 α
Aggregated ESG Score
Risk-adjusted Return -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 0 20 40 60 80 α
Aggregated ESG Score
Risk-adjusted Return 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 20 40 60 80 β_ M K T
Aggregated ESG Score
25 Panel C: FFC Model -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 0 20 40 60 80 α
Aggregated ESG Score
Risk-adjusted Return -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0 20 40 60 80 β_ S M B
Aggregated ESG Score
SMB Beta -1.5 -1 -0.5 0 0.5 1 1.5 2 0 20 40 60 80 β_ H M L
Aggregated ESG Score
HML Beta 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 20 40 60 80 β_ M K T
Aggregated ESG Score
MKT Beta -1.5 -1 -0.5 0 0.5 1 1.5 2 0 20 40 60 80 β_ H M L
Aggregated ESG Score
26
These figures display the relationship of average aggregated ESG score over the entire period with risk-adjusted returns and factor loadings. Aggregated ESG scores are retrieved from Thomson Reuters; betas are estimated from time-series regressions without rolling window. Panel A shows the relationship in the CAPM model, where the aggregated score has a negative relationship with risk-adjusted return and market beta. Panel B shows the relationship in the FF3 model, in which overall ESG score has a negative relationship with the risk-adjusted return, market beta and size beta, but has a positive relationship with value beta. Those relationships also hold in the FFC model, shown in Panel C. However, I cannot observe any clear relationship between aggregated score and momentum beta, because the slop of momentum beta is too flat. Comparing with the slop in the CAPM model, the slop of risk-adjusted return and market return becomes flatter in FF3 and FFC model.
8. References
Amini, C., Dal Bianco, S., 2017. Corporate social responsibility and Latin American firm performance. Corporate Governance: The international journal of business in society 17, 403-445.
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. Bauer, R., Otten, R., Tourani Rad, A., 2004. Ethical Investing in Australia: Is there a Financial Penalty? Pacific-Basin Finance Journal 14, 33–48.
Becchetti, L., Ciciretti, R., 2009. Corporate social responsibility and stock market performance. Applied Financial Economics 19, 1283-1293.
Becchetti, L., Ciciretti, R., Dalò, A., 2018. Fishing the Corporate Social Responsibility risk
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0 20 40 60 80 β_ M O M
Aggregated ESG Score
27
factors. Journal of Financial Stability 37, 25-48.
Carhart, M. M., 1997. On Persistence in Mutual Fund Performance. The Journal of Finance 52, 57-82.
Chordia, T., Goyal, A., Nozawa, Y., Subrahmanyam, A., Tong, Q., 2017. Are Capital Market Anomalies Common to Equity and Corporate Bond Markets? An Empirical Investigation. Journal of Financial and Quantitative Analysis 52, 1301-1342.
Ciciretti, R., Dalò, A., Dam, L., 2017. The Price of Taste for Socially Responsible Investment. Unpublished working paper.University of Groningen, Groningen.
CSRC, 2017. Industry Classification Results for Listed Company in the fourth Quarter of 2017. http://www.csrc.gov.cn/pub/newsite/scb/ssgshyfljg/201801/t20180119_332958.html. Accessed 9 June 2019.
Dam, L., Scholtens, B., 2015. Toward a theory of responsible investing: On the economic foundations of corporate social responsibility. Resource and Energy Economics 41, 103-121. SynTao Green Finance and Aegon-Industrial Fund, 2018. Decennial report on the responsible investment (in Chinese). Technical Report.
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.
Fama, E., MacBeth, J., 1973. Risk, Return, and Equilibrium: Empirical Tests. Journal of Political Economy 81, 607-636.
Grossman, B. R., W. F. Sharpe., 1986. Financial implications of South African divestment. Financial Analysts Journal 42, 15–29.
Hong, H., Kacperczyk, M., 2009. The price of sin: The effects of social norms on markets. Journal of Financial Economics 93, 15-36.
28
Kao, E., Yeh, C., Wang, L., Fung, H., 2018. The relationship between CSR and performance: Evidence in China. Pacific-Basin Finance Journal 51, 155-170.
Kempf, A., Osthoff, P., 2007. The Effect of Socially Responsible Investing on Portfolio Performance. European Financial Management 13, 908-922.
Kenneth R. French, 2018. Description of Fama/French 5 Factors for Developed Markets https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_5developed.html. Accessed 9 June 2019.
Lewellen, J., Nagel, S., Shanken, J., 2010. A skeptical appraisal of asset pricing tests☆. Journal of Financial Economics 96, 175-194.
Li, S., Li, X., Song, Z., 2015. Property rights, market competition and corporate social responsibility: based on the political rent-seeking theory (in Chinese). Chinese Journal of Population Resources and Environment 01, 162－169.
Lintner, J., 1965. The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets. The Review of Economics and Statistics 47, 13-37. Mahoney, L., Roberts, R., 2007. Corporate social performance, financial performance and institutional ownership in Canadian firms. Accounting Forum 31, 233-253.
Mossin, J. (1966). Equilibrium in a Capital Asset Market. Econometrica 34, 768-783.
Newey, W., West, K., 1987. Hypothesis Testing with Efficient Method of Moments Estimation. International Economic Review 28, 777-787.
Nollet, J., Filis, G., Mitrokostas, E., 2016. Corporate social responsibility and financial performance: A non-linear and disaggregated approach. Economic Modelling 52, 400-407. Preston, L., O'Bannon, D., 1997. The Corporate Social-Financial Performance Relationship. Business & Society 36, 419-429.
29
Governance: The Performance of Socially Responsible Mutual Funds. Journal of Corporate Finance 14, 302–322.
Renneboog, L., Ter Horst, J. and Zhang, C., 2008b. Socially responsible investments: Institutional aspects, performance, and investor behavior. Journal of Banking & Finance 32, 1723-1742.
Rudd, A., 1981. Social responsibility and portfolio performance. California Management Review 23, 55–61.
Scholtens, B., Zhou, Y., 2008. Stakeholder relationships and financial performance. Sustainable Development 16, 213-232.
Sharpe, W., 1964. Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk. The Journal of Finance 19, 425-442.
Thomson Reuters, 2018. Thomson Reuters ESG Scores. Technical report.
Usman, A., Amran, N., 2015. Corporate social responsibility practice and corporate financial performance: evidence from Nigeria companies. Social Responsibility Journal 11, 749-763. Waworuntu, S., Wantah, M., Rusmanto, T., 2014. CSR and Financial Performance Analysis: Evidence from Top ASEAN Listed Companies. Procedia - Social and Behavioral Sciences 164, 493-500.
Wu S., Xu J., 2004. A comparative study on the rational asset pricing model and irrational asset pricing model: evidence from stock market in china (in Chinese). Economic Research Journal 06, 105-116.
Wu Z., You J., 2018. Will social responsibility lead to lower stock returns in the future? Discussion on the causes of corporate social responsibility anomalies (in Chinese). Commercial Research 12, 162-168.
30
Xu H., Xiong C., 2009. Research on the applicability of Carhart's four-factor model in china's stock market (in Chinese). Business Culture 12, 128-137.
Yang W., Chen Z., 2003. Empirical study on the three-factor asset pricing model of Chinese stock market (in Chinese).The Journal of Quantitative & Technical Economics 12, 137-141. Zhang T. Do socially responsible investment funds underperform the market? Evidence from China. International Journal of Management Research and Business Strategy 3, 113-132. Zhihu, 2017. Research on the status quo of Chinese construction workers (in Chinese). https://zhuanlan.zhihu.com/p/25042793. Accessed 8 June 2019.
