ESG揭露程度對公司風險之影響 - 政大學術集成

全文

(1)國立政治大學財務管理研究所碩士學位論文. 治政大 ESG 揭露程度對公司風險之影響立 ‧. ‧ 國. 學. An Analysis of ESG Disclosure Score and Firm Risks. n. er. io. sit. y. Nat. al. Ch. i n U. v. 指導教授e :n 陳鴻毅 c h i 博士研究生 : 楊舒婷撰. 中華民國一百零八年六月 DOI:10.6814/NCCU202000150.

(2) ESG 揭露程度對公司風險之影響. 摘要本研究在探討公司對環境、社會、公司治理(ESG)的揭露程度是否影響其財務風險及下檔風險。財務風險以總風險、系統風險、非系統風險為代表，下檔風險則以股價暴跌風險(NCSKEW)及財務限制指數(KZ、WW、SA 指數)表示。我們發現 ESG 揭露程度愈高，財務風險皆愈低，此結果與過去研究相符，並支持公司從事 ESG 活動有助於降低資訊不對稱問題，進而增進利害關係人的信. 政治大表 ESG 揭露程度可降低下檔風險。以各別揭露程度來看，E 和 G 之揭露程度對立任、降低公司風險。此外，ESG 揭露程度與 WW 指數、SA 指數為負相關，代. ‧ 國. 學. 降低公司風險的效果更為顯著。然而，ESG 揭露程度雖有助於降低公司風險，卻無法藉此提升股價報酬。. ‧ sit. y. Nat. n. al. er. io. 關鍵字 : 企業社會責任;公司風險;財務風險;下檔風險;股票報酬. Ch. en chi. i n U. v. II. DOI:10.6814/NCCU202000150.

(3) An Analysis of ESG Disclosure Score and Firm Risks. Abstract In this study, I try to investigate whether a firm’s environmental, social, and governance (ESG) disclosure score affects its financial risk and downside risk. Specifically, I use total risk, systematic risk, and idiosyncratic risk to measure the financial risk, and the stock price crash risk (NCSKEW) and financial constraint indexes (KZ, WW, and SA index) to capture the downside risk. The empirical findings. 政治大 that ESG engagement can. show a significantly negative association between ESG disclosure score and financial risk measures, suggesting. 立. mitigate information. asymmetry problems and subsequently reduce a firm’s financial risk. In addition, I. ‧ 國. 學. find that ESG disclosure score is negatively associated with WW and SA indexes.. ‧. Further, disclosure scores of E and G are relatively important in reducing firm. y. Nat. risks. However, the risk reduction of ESG engagement cannot improve the stock. n. al. er. io. sit. return in the future.. i n U. Ch. v. en chi Keywords : ESG; firm risk; financial risk; downside risk; stock return g. III. DOI:10.6814/NCCU202000150.

(4) Contents. List of Tables ............................................................................................................V 1.. Introduction ..................................................................................................... 1. 2.. Literature Review ............................................................................................ 4 2.1 ESG and Risk .............................................................................................. 5 2.2 ESG and Downside risk ............................................................................... 5. 政治大. 2.3 ESG and Stock Price Crash Risk .................................................................. 6. 立. 2.4 ESG and Financial Constraint ...................................................................... 6. ‧ 國. 3.. 學. 2.5 ESG and Stock Return ................................................................................. 7 Sample, Variables, and Models ....................................................................... 9. ‧. 3.1 Sample ......................................................................................................... 9. Nat. sit. y. 3.2 Variables ..................................................................................................... 9. al. n. 4.. er. io. 3.3 Models ....................................................................................................... 13. i n U. v. Empirical Results ........................................................................................... 15. Ch. en chi. 4.1 Descriptive Statistics and Bivariate Correlation .......................................... 15 4.2 Empirical Results ....................................................................................... 16 5.. Conclusion and Discussions ........................................................................... 22. References .............................................................................................................. 24. IV. DOI:10.6814/NCCU202000150.

(5) List of Tables Descriptive statistics .................................................................................................27 Pearson correlation coefficients ................................................................................ 28 The relation between ESG and financial risk ............................................................ 29 The relation between ESG and downside risk ........................................................... 30 The relation between ESG individual components and financial risk ........................ 31 The relation between ESG individual components and downside risk ....................... 33. 政治大. The relation between ESG individual components and downside risk (small sample). 立. .................................................................................................................................35. ‧ 國. 學. The relation between ESG and stock return .............................................................. 37. ‧. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. V. DOI:10.6814/NCCU202000150.

(6) 1. Introduction In recent years, the growing popularity in environmental, social and governance issue has been astounding both in academic and business field. The influence of ESG issue from different dimensions can be far-reaching. Firms that are more active in ESG engagement might attract a group of loyal stockholders because of their unique core value. Besides, as ESG issue becomes the mainstream in the area of management, ESG idea dramatically changes the way managers think. When managers are trying to maximize firm value, they must consider wider stakeholders. They not only have to. 政治大 government, society and environment. ESG engagement leads managers to choose a 立 take the responsibility for shareholders, but also for employees, suppliers, customers,. different investing decision. Therefore, stakeholders have shown great interest on ESG. ‧ 國. 學. issue.. ‧. Prior studies shed light on the link between ESG engagement and risk. ESG. y. Nat. engagement may reduce firm risk through different channels. Dufresne and Savaria. er. io. sit. (2004) believe that better ESG commitment can reduce idiosyncratic risk and enhance quality of life in the workplace. They also indicate that firms may suffer from unethical. al. n. v i n C hoccurred in 1989 inUAlaska causes a great loss of issues. For example, the oil spill en chi punitive damages to Exxon Valdez. Cases like Enron and Toshiba can be convincing. evidences that governance risk is costly (CFA Institute, 2015). To assess the impact of ESG engagement on firm risk, it is important to recognize that from what dimension of risk channel is ESG disclosure score delivering. In this article, I first examine the relationship between ESG disclosure score and total risk, systematic risk, and idiosyncratic risk. Total risk consists of systematic risk and unsystematic risk. Systematic risk refers to risk factors that affect the entire financial market such as natural disasters, new investment policy or economic recession. It is unable to be controlled and cannot be mitigated. On the other hand, unsystematic risk is a firm1. DOI:10.6814/NCCU202000150.

(7) specific risk. Idiosyncratic risk is controllable. Change of the management, new strategic policy, the entry of new competitors, etc. all influence a firm’s idiosyncratic risk. Idiosyncratic risk can be divided into two parts, upside risk and downside risk. If the impact of ESG on total, systematic, and idiosyncratic risk is confirmed, I further investigate whether ESG causes a negative influence on downside risk. Downside risk is the financial risk associated with future losses. Firms with serious downside risk are prone to suffer from bad financial performance, market share decreasing, and even bankruptcy. I proxy downside risks by stock price crash risk (NCSKEW) and financial. 政治大 and downside risk. Finally, I would like to test whether the reducing risk effect of ESG 立 constraint indexes (KZ, WW, SA index). I expect a negative relationship between ESG. can bring higher stock return in the future. In other words, under the situation that ESG. ‧ 國. 學. engagement reduces financial risks, can higher ESG score firms achieve higher stock. ‧. return because of lower downside risk?. y. Nat. I collect ESG disclosure score from Bloomberg database during the period from. er. io. sit. 2003 to 2017. To measure the predictability of ESG, dependent variables are computed one year ahead. The stock return data from 2004 to 2018 is obtained from CRSP and. al. n. v i n Cishfurther computed U accounting data from Compustat as financial constraint index. To en chi. meet all data requirements, it yields a final sample of 21,426 firm-year observations across 3,305 US firms from 2004 to 2018. My study gives two important conclusions. First, I conclude that ESG score are negatively correlated with firm risks. When we look into ESG individual components, E score and G score play a crucial role in reducing firm risks. Second, the relationship between ESG and stock return is still unclear. Lower downside risk is not well transformed into higher stock return. My findings should provide information to both managers and market participants. For company managers, the results show that ESG engagement could significantly reduce firm risks, especially, downside risks. For market participants, the findings suggest the 2. DOI:10.6814/NCCU202000150.

(8) role of ESG stocks in a portfolio. Section 2 discusses the existing literature. Section 3 discusses the data, variables and model. Section 4 presents the result. I conclude in section 5.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. 3. DOI:10.6814/NCCU202000150.

(9) 2. Literature Review ESG issue has become popular in academic field in recent years. Literatures about ESG in corporate finance, investing, behavior finance and management allow us to build our knowledge about ESG issue. The puzzles of ESG have been resolved greatly. Attig et al. (2016) shed light on ESG firm characteristics. They find that the degree of internalization of a firm is positively correlated with its ESG rating. As for the benefits of ESG, many studies mention that ESG investments not only bring benefits to the society, but also the enterprises. Albuquerque et al. (2018) assume that CSR. 政治大 believe that firms with higher CSR scores will have lower systematic risk, higher firm 立. investments can bring product differentiation effect. Based on the assumption, they. value and smoother profits. The empirical evidence supports their prediction. Friede et. ‧ 國. 學. al. (2015) gather more than 2000 empirical studies and analyze with vote-count studies. y. Nat. performance.. ‧. and meta-analyses. They find a positive and stable ESG impact on corporate financial. er. io. sit. There are studies that examine whether ESG score can impact firm value. El Ghoul et al. (2017) find that when the “institutional voids” is more serious in the market, CSR. al. n. v i n Cvalue activities benefits more on firm growth. In a risk management U h e nand external i h c. perspective, the use of currency derivatives for well-governed firms is associated with a significant value premium because well-governed firms tend to use derivatives to hedge currency exposure, rather than managerial reasons (Allayannis et al. 2012). Fatemi et al. (2018) conclude that ESG strength is increasing firm value while ESG concern is damaging firm value. Besides, there is a moderating effect of ESG disclosure on firm valuation. They argue that ESG disclosure can mitigate negative influence that ESG concerns bring because ESG disclosure send positive ESG commitments to investors. On the other hand, ESG disclosure can mitigate positive influence that ESG strengths bring because investors will interpret disclosures as firms trying to convince 4. DOI:10.6814/NCCU202000150.

(10) the markets an overinvestment on ESG. However, the conclusion is different from prior papers.. 2.1 ESG and Risk As mentioned above, ESG engagement is beneficial to enterprises. I assume that one of the most important benefits ESG brings is reducing risk. Benlemlih et al. (2018) find a significant negative relationship between a firm’s E and S disclosure score and its total and residual risk. The findings support the claim that extensive E and G. 政治大 problem, firms can build positive reputation among stakeholders. There are evidence 立. disclosure can enhance corporate transparency. By eliminating information asymmetry. that ESG engagement also helps to reduce risk in controversy industries (Jo and Na. ‧ 國. 學. 2012). The results reflect that firms in controversial industries engage in CSR for risk. ‧. reducing, instead of window dressing. Other studies focus on the impact of different. y. Nat. CSR aspects on risk. Oikonomou et al. (2012) present two main findings. First, CSR. er. io. sit. aggregated components can better capture corporate social performance (CSP) effect on market risk than individual components (diversity, environment, community etc.). al. n. v i n C hsocially irresponsible Second, during financial distress, firms which has more CSP en chi U concerns will face more serious stock price volatility. On the other hand, when the. market fluctuation is comparably small, CSR strengths can lead the firms to lower stock price volatility.. 2.2 ESG and Downside risk The ESG score has much to do with downside risk. In a strategic view, Husted (2005) believes CSR investments can be seen as a real option to control ex ante risks because it limits a firm’s downside outcomes. Godfrey (2005) argues that ESG engagement of firms can generate moral capital, which has an insurance-like protection 5. DOI:10.6814/NCCU202000150.

(11) function. It protects shareholder relational wealth because of good trust and reputation among stakeholders. Thus, moral capital mitigates the loss of shareholder value when bad news happens. In this paper, downside risk is further presented in the form of stock price crash risk and financial constraints respectively.. 2.3 ESG and Stock Price Crash Risk Hoepner et al. (2018) find that large institutional investors’ engagement in ESG issue can lower firms’ downside risk. The downside risk is measured by LPM (0,2),. 政治大 (third) order are 1.2% (1.4%) lower at firms engage in ESG, which is economically 立 LPM (0,3) and VaR. Compared to matched firms, Lower partial moments of the second. significant. However, the downside risk effect is only in governance and environment. ‧ 國. 學. aspect. They do not find downside risk effect in social aspect. Hutton et al. (2009). ‧. present that financial transparency such as decreasing earning management can lower. sit. y. Nat. crash risk. Kim et al. (2014) show that a firm’s CSR performance is negatively. io. er. associated with future crash risk because of higher standard of transparency and less bad news hoarding. With further examination, they argue that the mitigating effect of. al. n. v i n CSR on crash risk is especiallyCimportant when governance mechanisms, including hen chi U. monitoring by boards and institutional investors, could not function desirably. There are different opinions about the impact of CSR on crash risk. Becchetti et al. (2015) find that CSR is positively correlated with idiosyncratic volatility. They claim that CSR engagement firms attract specific (smaller) investor population, and thus become less flexible to shocks.. 2.4 ESG and Financial Constraint Many studies explore the impact of ESG engagement on firms’ financial access. Cheng et al. (2014) find a negative relationship between CSR performance and capital 6. DOI:10.6814/NCCU202000150.

(12) constraints. They give two reasons. First, they believe that superior CSR performance can enhance stakeholders’ engagement and thus lower agency costs. Second, firms with better CSR performance tend to be more transparent and thus reduce information asymmetry problems. These all contribute to better access to finance. El Ghoul et al. (2011) conclude that higher ESG score and cost of equity are negatively correlated. It implies firms with higher ESG score face lower capital constraints. Besides, there are other studies researching whether a firm’s financial access influence its CSR activities, such as Hong et al. (2012) and Chan et al. (2016). By applying current-year, three-year,. 政治大 in financial distress, firms tend to not engage in CSR activities. There is a negative 立 and five-year averages financial constraint proxies, Chan et al. find that when firms are. relationship between CSR activities and financial distress. This phenomenon is. ‧ 國. 學. especially significant when the financial constraint is assessed by the KZ index.. ‧ sit. y. Nat. 2.5 ESG and Stock Return. io. er. Renneboog et al. (2008) review prior research of SRI studies. They conclude that SRI funds do not outperform non-SRI funds. However, SRI investors are still willing. n. al. Ch. to invest in CSR company because they can. en chi. iv n gain U non-financial. utility which. compensate for lower stock return. Rather than stock return of SRI fund, Brammer et al. (2006) focus on firm level because they argue that fund performance is affected by fund manager’s stock picking ability. With disaggregate measures, they conclude that CSR score is negatively correlated to stock returns. Makni et al. (2009) also find a significant negative relationship between environment score and stock returns in Canadian market. They argue two reasons for the result. First, compared to US market, Canadian market is too small to be a desirable market for environmental investment. Second, their analysis period is only 2004-2005 and the long-term relationship should be further examined. 7. DOI:10.6814/NCCU202000150.

(13) Building on the literature on ESG, I believe that ESG can influence firm risks. However, there are still many questions to be solved. In this study, I try to clarify the relationship between ESG and different risk measures, and the relationship between ESG and return through the role of risk.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. 8. DOI:10.6814/NCCU202000150.

(14) 3. Sample, Variables, and Models 3.1 Sample I obtain ESG data from Bloomberg database. The ESG data is from 2003 to 2017. Market data is obtained from Center of Research in Security Prices (CRSP). I only include firms whose share code is 10 or 11 and exchange code is 1, 2, 3, 31, 32, or 33 in my sample. In other words, only common stocks traded on NYSE, AMEX, NASDAQ are included. Daily stock return is further computed as total risk, idiosyncratic risk and stock price crash risk in year. Besides, I obtain accounting data. 政治大 index, WW index and SA index, respectively) and control variables. After matching 立. from Compustat. Accounting data is further computed as financial constraint index (KZ. firm-year observations for the period 2004 to 2018.. ‧ sit. y. Nat. 3.2 Variables. 學. ‧ 國. Bloomberg ESG data with data from CRSP and Compustat, my final sample is 21,426. io. er. Bloomberg ESG Disclosure Scores. Instead of MSCI ESG database, which is extensively used in academic research, I. al. n. v i n choose to use Bloomberg ESGCdatabase. Bloomberg h e n c h i U uses a different score rating. method from MSCI. Bloomberg gives ESG score according to the enterprise’s ESG disclosure level based on annual reports, standalone sustainability reports, company websites, etc. Each score category ranges from 0 to 100, and the score has standardized cross-sector and industry-specific metrics characteristics. In this study, I use ESG combine disclosure score (ESG score), E disclosure score (E score), S disclosure score (S score), and G disclosure score (G score) to examine which category can significantly influence enterprises’ financial performance.. 9. DOI:10.6814/NCCU202000150.

(15) Measures of Financial Risk First, I measure financial risk by the total risk. Following Benlemlih et al. (2018), total risk is calculated as the standard deviation of daily stock returns from CRSP. Based on CAPM, total risk consists of systematic risk and unsystematic risk. Systematic risk such as interest rate changes or wars are hard to avoid. It affects the overall market. Idiosyncratic risk is a firm’s specific risk, such as pipeline damage, management corruption, bankruptcy, etc. I take systematic risk as my second financial risk measure and idiosyncratic risk as my third financial risk measure. I develop the risk measure. 政治大. from Eq. (1).. 立. (1). ‧ 國. 學. 𝑅𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑖 ∗ 𝑅𝑚,𝑡 + 𝑒𝑖,𝑡. ‧. where 𝑅𝑖𝑡 is a security’s daily stock return at time t. 𝑅𝑚,𝑡 is value-weighted daily. y. Nat. market return at time t. 𝑒𝑖,𝑡 is the error term. Based on CAPM, I regress the daily stock. er. io. sit. return on the daily market return over the year and the risk measures can be presented in the following format of Eq. (2).. n. al. 𝜎𝑖2. =. 𝛽𝑖2. ∗. 2 𝜎𝑚. +. 𝜎𝜀2. Ch. en chi. i n U. v. (2). Total risk is the standard deviation of a firm’s daily stock return (𝜎𝑖 ). Systematic risk is defined as a firm’s beta times the standard deviation of the daily market return (𝛽𝑖 ∗ 𝜎𝑚 ) and the idiosyncratic risk is defined as the standard deviation of a firm’s residual term (𝜎𝜀 ).. Measure of Stock Price Crash Risk Following Kim et al. (2014), I employ negative condition skewness (NCSKEW) 10. DOI:10.6814/NCCU202000150.

(16) to proxy stock price crash risk. NCSKEW is developed from expanded market model regression and allows me to capture the asymmetry in risk.. 𝑅𝑖,𝑡 = 𝛼𝑖 + 𝛽1,𝑖 ∗ 𝑅𝑚,𝑡−2 + 𝛽2,𝑖 ∗ 𝑅𝑚,𝑡−1 + 𝛽3,𝑖 ∗ 𝑅𝑚,𝑡 + 𝛽4,𝑖 ∗ 𝑅𝑚,𝑡+1 + (3). 𝛽5,𝑖 ∗ 𝑅𝑚,𝑡+2 + 𝜀𝑖,𝑡. where 𝑅𝑖,𝑡 is a security’s daily stock return at time t. 𝑅𝑚,𝑡 is value-weighted daily market return at time t. I put one and two-day lead (lag) to allow nonsynchronous. 政治大 return). As shown in Eq. (4), I further calculate 𝑤 as the natural logarithm of one 立. trading (Dimson, 1979). 𝜀𝑖,𝑡 is the security’s residual return at time t (firm-specific 𝑖,𝑡. ‧ 國. 學. plus the residual return in the expanded market model regression. Next, I make 𝑤𝑖,𝑡 to the second and third power, and put it into Eq. (5) to get NCSKEW. N in Eq. (5) is. y. Nat. (4). sit. io. er. 𝑤𝑖,𝑡 = 𝑙𝑛(1 + 𝜀𝑖,𝑡 ). ‧. the number of daily returns during year t.. 3. al. v ni. 3. n. 3 2 2 )] 𝑁𝐶𝑆𝐾𝐸𝑊𝑖,𝑡 = −[𝑛 ∗ (𝑛 − 1)2 ∗ ∑𝑤𝑖,𝑡 ] / [(n − 1) ∗ (n − 2) ∗ (∑𝑤𝑖,𝑡. Ch. en chi U. (5). Measures of Financial Constraint I use KZ index (Kaplan and Zingales 1997, Lamont et al. 2001), WW index (Whited and Wu 2006), and SA index (Hadlock and Pierce 2010), to proxy financial constraints respectively. The equation is computed as follows:. KZ index = −1.002 ∗ 𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 ⁄𝐾 + 0.283 ∗ Q + 3.139 ∗ 𝐷𝑒𝑏𝑡⁄𝑇𝑜𝑡𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 − 39.368 ∗ 𝐷𝑖𝑣⁄𝐾 − 1.315 ∗ 𝑐𝑎𝑠ℎ⁄𝐾. (6). 11. DOI:10.6814/NCCU202000150.

(17) where cash flow is defined as income before extraordinary items plus depreciation; Q is defined as market value of assets divided by book value of assets; Debt is defined as short-term debt plus long-term debt; Total Capital is defined as short-term debt plus long-term debt plus stockholder’s equity; Div is defined as preferred dividend plus common dividend; cash is defined as cash and short-term investments; K is defined as net property, plant and equipment. All K is lagged. Higher KZ index value implies a more serious financial constraint problem.. 政治大. WW index = −0.091 ∗ CF − 0.062 ∗ DIVPOS + 0.021 ∗ TLTD − 0.044 ∗ LNTA + 0.102 ∗ ISG − 0.035 ∗ SG. 立. (7). ‧ 國. 學. where CF is defined as the ratio of income before extraordinary items plus depreciation. ‧. divided to total assets; DIVPOS is a indicator. If a firm pays cash dividend then. y. Nat. DIVPOS equals to 1, and zero otherwise. TLTD is the ratio of long-term debt to total. er. io. sit. assets. LNTA is defined as the natural log of total assets. ISG is the average three-digit industry sales growth. SG is firm sales growth. Higher WW index value implies a more. n. al. Ch. serious financial constraint problem.. en chi. i n U. v. SA index = −0.737 ∗ 𝑠𝑖𝑧𝑒 + 0.043 ∗ 𝑠𝑖𝑧𝑒 2 − 0.040 ∗ 𝑎𝑔𝑒. (8). where size is defined as the logged value of total assets inflation-adjusted to year 2004’s total assets and is winsorized above at 4.5 billion. Age is the number of years that the firm has a non-missing stock price on Compustat and is winsorized above at 37 years. Again, higher SA index value implies a more serious financial constraint problem.. 12. DOI:10.6814/NCCU202000150.

(18) 3.3 Models I use the following model to examine whether ESG disclosure score can affect a firm’s financial risk and downside risk. Financial Risk 𝑖,𝑡 or 𝑁𝐶𝑆𝐾𝐸𝑊𝑖,𝑡 = α + 𝛽1 ∗ 𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒 𝑆𝑐𝑜𝑟𝑒𝑖,𝑡−1 + 𝛽2 ∗ ln(𝑠𝑖𝑧𝑒)𝑖,𝑡−1 + 𝛽3 ∗ 𝑀𝐵𝑖,𝑡−1 + 𝛽4 ∗ 𝐿𝐸𝑉𝑖,𝑡−1 + 𝛽5 ∗ 𝑅𝑂𝐴𝑖,𝑡−1 + ∑𝑚 𝛽𝑚 ∗ (9). 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑦𝑚 + ∑𝑛 𝛽𝑛 ∗ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑦𝑛 + 𝜀𝑖,𝑡. In Eq. (9), financial risk is one of the risk measures, namely total risk, idiosyncratic risk.. 政治大 2014), I use several control variables. Ln(size) is defined as the natural log of the book 立 Following prior literature (e.g. Benlemlih et al. 2018; Hutton et al. 2009; Kim et al.. value of assets. Market-to-Book ratio (MB) is defined as the ratio of market value of. ‧ 國. 學. equity to the book value of equity. The next control variable is leverage (LEV),. ‧. calculated as total liabilities divided by total assets. Profitability measured by return on. sit. y. Nat. assets (ROA) is calculated as income before extraordinary items divided by total assets.. io. er. To examine the predictability of independent variable, independent variables are oneyear lagged to dependent variables. In addition, I also control industry and year effects. n. al. in all regressions.. Ch. en chi. i n U. v. Financial Constraint 𝑖,𝑡 = α + 𝛽1 ∗ 𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒 𝑆𝑐𝑜𝑟𝑒𝑖,𝑡−1 + 𝛽2 ∗ ln(𝑠𝑖𝑧𝑒)𝑖,𝑡−1 + ∑𝑚 𝛽𝑚 ∗ 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑦𝑚 + ∑𝑛 𝛽𝑛 ∗ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑦𝑛 + 𝜀𝑖,𝑡. (10). In Eq. (10), financial constraint is one of the constraint measures, namely KZ, WW, or SA index. Following Cheng et al. (2014), the control variable ln(size) is defined as the natural logarithm of total assets. Again, the control variables are lagged and industry and year fixed effect are controlled in all regressions.. 13. DOI:10.6814/NCCU202000150.

(19) 𝑅𝐸𝑇𝑖,𝑡 = 𝛼 + 𝛽1 ∗ 𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒 𝑆𝑐𝑜𝑟𝑒𝑖,𝑡−1 + 𝛽2 ∗ 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑅𝑖𝑠𝑘𝑖,𝑡−1 + 𝛽3 ∗ 𝑆𝑐𝑜𝑟𝑒𝑖,𝑡−1 ∗ 𝑅𝑖𝑠𝑘𝑖,𝑡−1 + 𝛽4 ∗ 𝑅𝐸𝑇𝑖,𝑡−1 + 𝛽5 ∗ ln(𝑠𝑖𝑧𝑒)𝑖,𝑡−1 + 𝛽6 ∗ 𝑀𝐵𝑖,𝑡−1 + 𝛽7 ∗ 𝐿𝐸𝑉𝑖,𝑡−1 + 𝛽8 ∗ 𝑅𝑂𝐴𝑖,𝑡−1 + ∑𝑚 𝛽𝑚 ∗ 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑦𝑚 + ∑𝑛 𝛽𝑛 ∗ (11). 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑦𝑛 + 𝜀𝑖,𝑡. In Eq. (11), 𝑅𝐸𝑇𝑖,𝑡 is a security’s yearly stock return at time t, computed by monthly return from CRSP. 𝑆𝑐𝑜𝑟𝑒𝑖,𝑡−1 ∗ 𝑅𝑖𝑠𝑘𝑖,𝑡−1 is a factor to test that given specific risk, can. 政治大 between return and score*risk. To examine the predictability of control variables, all 立 higher ESG disclosure score bring higher stock return. I expect a positive relationship. control variables are one-year lagged by dependent variables.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. 14. DOI:10.6814/NCCU202000150.

(20) 4. Empirical Results 4.1 Descriptive Statistics and Bivariate Correlation Table 1 provides descriptive statistics for the variables used in the analysis. To prevent the results from distortion by outliers, I winsorize the sample except Bloomberg ESG disclosure score at the bottom and top 0.5% levels. There are 21,426 firm-year ESG disclosure scores in the full sample, which represent 3,305 firms. Apart from the combined score, I also collect separate environment, social, and governance score in Bloomberg ESG database. It yields 4,586 firm-year E score representing 694 firms,. 政治大 representing 3,297 firms. The mean values of disclosure scores, ESG score, E score, S 立 12,747 firm-year S score representing 2,012 firms, and 21,258 firm-year G score. score, and G score, are 17.13, 21.04, 15.84, and 51.26, respectively. The average total. ‧ 國. 學. risk is 0.026, the average systematic risk is 0.011, and the average idiosyncratic risk is. ‧. 0.022. The average firm in the sample has a NCSKEW of 0.086, a KZ index of -14.864,. y. Nat. a WW index of -0.248, a SA index of -3.283, and a year return of 0.143. In addition,. er. io. sit. the mean value of control variables such as ln(size), measured as natural log of total asset, is 7.268 (about 1.433 billion), market-to-book ratio is 3.519, leverage ratio is. n. al. Ch. 0.550, and return on assets is 0.003.. en chi. i n U. v. < Insert Table 1 >. Table 2 presents the Pearson correlations. I find a high correlation between total risk and systematic risk (0.48) and idiosyncratic risk (0.97), which are similar to the correlation in Jo and Na (2012). Besides, among a total of 7 risk measures and financial constraint measures, 5 out of 7 measures show a negative and significant correlation 15. DOI:10.6814/NCCU202000150.

(21) with ESG disclosure score. The correlation between ESG score and total risk is -0.29, systematic risk is -0.08, idiosyncratic risk is -0.30, WW index is -0.19 and SA index is -0.35. However, ESG score and NCSKEW and KZ index are weakly correlated (correlation below 0.1). The correlation matrix presents that ln(size) is highly correlated with ESG score (0.59) and other control variables. Therefore, in the following regressions, I report regressions with and without ln(size) factor respectively. Finally, other control variables are weakly correlated. Multicollinearity problems are unlikely to happen in the model.. 立. 政治大 < Insert Table 2 >. ‧. ‧ 國. 學 sit. y. Nat. 4.2 Empirical Results. io. er. The Effect of ESG on firm risks. Table 3 reports results from regression analysis of the relation between ESG score. al. n. v i n and total risk, systematic risk, C and idiosyncratic risk.U h e n c h i Firm characteristics that might. affect risks such as ln(size), MB, LEV, ROA are controlled. The use of lagged ESG score and other controlled variables is to predict future risks. I expect a negative relationship between ESG score and financial risk measures. Columns 1, 3, and 5 include ln(size) factor. High correlation between ln(size) and ESG score might cause multicollinearity problem. In these three regressions, only column 3 has a negative coefficient on ESG score. To prevent from multicollinearity, I redo the regressions without ln(size) factor in columns 2, 4, and 6. In these 3 regressions, all the coefficients on ESG score are negative and significant at 1% level, which matches my prediction and is in line with prior studies. I believe that ESG engagement can eliminate 16. DOI:10.6814/NCCU202000150.

(22) information asymmetry by enhancing corporate transparency. Besides, engaging in ESG activities can build a strong connection between ESG firms and stakeholders. It also helps to attract good quality employees and make product differentiation among competitors. All these factors potentially contribute to reducing financial risk. Table 3 provides strong evidence that higher ESG disclosure score can lower financial risks.. < Insert Table 3 >. 立. 政治大. I further investigate from what portal can ESG engagement affect a firm’s. ‧ 國. 學. idiosyncratic risk. Idiosyncratic risk such as the high possibility of stock price crash,. ‧. and difficult access to finance, which causes serious downside risks to a firm. Therefore,. sit. y. Nat. I use NCSKEW, KZ index, WW index, SA index to proxy downside risks to examine. io. er. the relationship between ESG disclosure score and downside risks. Table 4 reports the results of the regression. Columns 1 and 3 include ln(size) factor in the regressions and. al. n. v i n it shows a negative coefficientCof ESG disclosure score. h e n c h i U However, after removing. ln(size) factor due to the concern of multicollinearity, I find a positive correlation between ESG score and NCSKEW and KZ index. The 𝑅2 is comparatively small in NCSKEW regressions, which is similar to the 𝑅2 in Kim et al. (2014). In the real business world, the greater the market frictions are, the higher the cost of external financing. Improving corporate transparency can solve market frictions problem. With credible internal control system and disclosure reports, ESG firms are able to raise funds at a lower cost either from equity market or from bond market. Therefore, ESG firms can better capture investment opportunities. In contrast, firms with lower ESG score may end up forgo investment opportunities and damage 17. DOI:10.6814/NCCU202000150.

(23) shareholder value due to serious financial constraints. I argue that lower ESG score firms face higher downside risk than higher ESG score firms do. In WW and SA regressions (columns 5 and 6), I find that ESG score is negatively correlated with the downside risk measures and significant at 1% level. I do not control ln(size) variable in WW and SA index regression (Columns 5 and 6) because natural log of asset is considered in the calculation of the indexes.. 政治大. < Insert Table 4 >. The Effect of ESG individual components on firm risks. 學. ‧ 國. 立. ‧. In this section, I examine the effect of ESG individual components on firm risks.. sit. y. Nat. I collect Bloomberg environmental, social, and governance disclosure score and. io. er. combine it with total risk, systematic risk, and idiosyncratic risk to run regressions respectively. Environmental issue includes CO2 emissions, recycled policy, renewable. al. n. v i n C h minorities in workforce energy usage, etc.; social issue includes policy, training policy, en chi U. human rights policy, etc.; governance issue consists of board structure, the independence of the board, etc. I find that each individual ESG components contributes to lower firm risks. E, S, G scores are negatively correlated with all 3 risk measures at 1% level. It suggests that higher disclosure score can build positive reputation among stakeholders and enhance its transparency, which reduce firm risks. Moreover, In columns 4, 8, and 12, it shows that environment and governance engagement have a more pronouncing effect to risks if a firm discloses all 3 dimension scores. Besides, the relationship between ROA and risk are negative and significant in all regressions.. 18. DOI:10.6814/NCCU202000150.

(24) < Insert Table 5 >. The relationship between ESG individual components and downside risks are reported in table 6. From column 9 to 12 (WW index measure) and 13 to 16 (SA index measure), I find significantly negative coefficients on all individual disclosure scores. ESG is a protection to firms’ downside risk. Higher disclosure score firms can enjoy. 政治大 ability to borrow but also to issue new equity. More external financing choices supports 立 better access to finance and are less subject to financial distress. They not only have the. the survival of higher disclosure score firms. Hence, they can pursue a growing prospect. ‧ 國. 學. in the future. However, in NCSKEW regressions, the coefficients on ESG individual. ‧. scores are contrary to my expectation. The relationship between ESG score and KZ. n. al. er. io. sit. y. Nat. index is not clear in table 6.. C<hInsert Table 6 > U n i en chi. v. For NCSKEW and KZ index measures, we cannot conclude that higher ESG (individual) scores bring lower risks to enterprises. Therefore, I further investigate if a firm contribute to environment, social, and governance dimensions at the same time, could individual score play an important role to risks. And if so, which dimension has a more pronouncing effect? I present the results in table 7. In NCSKEW regressions, the coefficient on ESG, E, and S scores are still positive. I argue that only when the annual disclosure score is growing steadily can prevent the stock price from falling 19. DOI:10.6814/NCCU202000150.

(25) sharply, and thus avoid stock price crash risk. However, since the ESG disclosure score fluctuates every year, the risk reducing effect of ESG cannot be fully performed. In contrast, in the case of KZ index, when I limit the sample to firms engaging in all three dimensions of ESG, I find that contribution from each individual component can lower financial constraint significantly in column 7 to 12. This gives two implications. First, participating in ESG activities comprehensively is a more efficient way to lower firm risk than partial ESG engagement. The reducing risk effect can be correctly conveyed when firms contribute to environment, society, and governance simultaneously. Second,. 政治大 to firms engaging in all three dimensions of ESG, it yields a similar firm-year sample 立 E score is the least among three individual scores. Therefore, when I limit the sample. with individual E score sample. This explains the importance of environment. ‧ 國. 學. contribution to risk reduction.. ‧. n. al. er. io. sit. y. Nat. < Insert Table 7 >. Ch. en chi. i n U. v. The Effect of ESG on Stock Return. The results above in this study provide strong evidence that ESG engagement can reduce firm risks. Besides, the reducing risk effect is contributed to lower financial constraints, which is a measure of downside risk. Lower financial constraints allow firms to enjoy better access to finance and borrowing at lower costs. Firms can undertake profitable investments and pursue superior financial performance. Therefore, in this section, I examine whether firms with more extensive ESG activities can achieve higher stock return caused by downside risk reduction. The result is reported in table 8. I expect a positive coefficient of variable score * risk in Eq. (11), which implies under 20. DOI:10.6814/NCCU202000150.

(26) the same level of risk, higher ESG score firms with lower downside risk can enjoy higher stock return. However, in columns 1, 5, and 9, none of the 3 ESG score regressions has significant coefficients on score* financial risk. Therefore, I conclude the relationship between score * risk and return is far from clear. The transfer effect of ESG from risk to return is uncertain in total risk, systematic risk and idiosyncratic risk regressions. I believe this result is due to two opposite effects countering each other. Lower downside risks that ESG engagement brings can raise future stock return. However, lower risks usually accompany lower stock return in the real financial market.. 政治大 other control variables in my regression, we find that financial risk and LEV are 立. As table 8 shows, the overall effect of the two contradicting forces is not clear. For. positively correlated with return. Ln(size) and MB are negatively correlated with return.. ‧ 國. 學. The findings are consistent with prior studies.. ‧. n. al. er. io. sit. y. Nat. < Insert Table 8 >. Ch. en chi. i n U. v. 21. DOI:10.6814/NCCU202000150.

(27) 5. Conclusion and Discussions In this article, I examine the empirical impact of ESG engagement on firm risks. Using a comprehensive sample of U.S. firms from 2004 to 2018, I find a significantly negative association between firm’s ESG disclosure score and one-year-ahead financial risk, measured by total risk, systematic risk, and idiosyncratic risk. Further, I examine the relationship between ESG engagement and downside risk. Downside risk is measured by NCSKEW, KZ index, WW index, and SA index. It shows that the coefficient on ESG disclosure score is significantly negative when WW index and SA. 政治大 which firms engaging environment, society, and governance issues simultaneously, the 立 index are the downside risk measures. After controlling our sample to a smaller data. coefficient on ESG disclosure score becomes significant negative in KZ index. ‧ 國. 學. regression. It implies that ESG engagement can reduce downside risk more efficiently. ‧. when firms participating in ESG activities comprehensively. However, it is reported. sit. y. Nat. that ESG disclosure and NCSKEW are positively correlated. I argue that only when. io. er. ESG disclosure score is steadily increasing every year can it reduce stock price crash risk. Besides, when ESG database is more completed in the future, future studies can. al. n. v i n employ alternative risk measuresCsuch as bankruptcy (there h e n c h i U are only 6 bankruptcy firms in the sample).. My study sheds light on ESG individual components. Comparing 3 individual components, I conclude that E disclosure score and G disclosure score play a more important role of risk reducing. In this study, I obtain ESG data from Bloomberg ESG database. Rather than rating each item according to its performance, Bloomberg collects public resources and gives ESG score by the firms’ ESG disclosure level. Besides, it shows that Bloomberg ESG disclosure score is highly correlated with size factor. To better control size factor when considering the effect of ESG on risks, future research could use alternative ESG data to examine the relationship between ESG score and firm 22. DOI:10.6814/NCCU202000150.

(28) risk. Last, the link between ESG disclosure score and stock return is still unclear. I expect that ESG engagement could bring higher stock return due to lower downside risk. However, I find no evidence to support the claim. I argue that there are two opposite effects contradicting each other. One is “ high risk, high return” effect. Under this structure, higher ESG score will cause lower stock return. The other is the transfer effect of ESG from risk to return. Through this channel, ESG engagement is positively correlated with stock return when firms are less subject to downside risk. The overall. 政治大 better capture the influence of ESG on stock return. 立. effect is uncertain. In this context, future research could separate the two effects and. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. 23. DOI:10.6814/NCCU202000150.

(29) References Albuquerque, R., Koskinen, Y., & Zhang, C. (2018). Corporate social responsibility and firm risk: Theory and empirical evidence. Management Science, forthcoming. Allayannis, G., Lel, U., & Miller, D. P. (2012). The use of foreign currency derivatives, corporate governance, and firm value around the world. Journal of International Economics, 87, 65-79. Attig, N., Boubakri, N., El Ghoul, S., & Guedhami, O. (2016). Firm internationalization and corporate social responsibility. Journal of Business Ethics, 134, 171-197.. 政治大 stakeholder risk, and idiosyncratic 立 volatility. Journal of Corporate Finance, 35, 297-. Becchetti, L., Ciciretti, R., & Hasan, I. (2015). Corporate social responsibility,. ‧ 國. 學. 309.. Benlemlih, M., Amama, S., Qiu, Y., Trojanowski, G. (2018). Environmental and social. ‧. disclosures and firm risk. Journal of Business Ethics, 152, 613-626.. sit. y. Nat. Boutin-Dufresne, F., & Savaria, P. (2004). Corporate social responsibility and financial. al. er. io. risk. The Journal of investing, 13, 57-66.. v. n. Brammer, S., Brooks, C., & Pavelin, S. (2006). Corporate social performance and stock. Ch. en chi. i n U. returns: UK evidence from disaggregate measures. Financial Management, 35, 97116. CFA Institute (2015). Environmental, social, and governance issues in investing: A guide for investment professionals. Chan, C. Y., Chou, D. W., & Lo, H. C. (2016). Do financial constraints matter when firms engage in CSR? North American Journal of Economics and Finance, 39, 241259. Cheng, B., Ioannou, I., & Serafeim, G. (2014). Corporate social responsibility and access to finance. Strategic Management Journal, 35, 1-23. 24. DOI:10.6814/NCCU202000150.

(30) Dimson, E., 1979. Risk measurement when shares are subject to infrequent trading. Journal of Financial Economics, 7, 197-226. El Ghoul, S., Guedhami, O., & Kim, Y. (2017). Country-level institutions, firm value, and the role of corporate social responsibility initiatives. Journal of International Business Studies, 48, 360-385. El Ghoul, S., Guedhami, O., Kwok, C., & Mishra, D. (2011). Does corporate social responsibility affect the cost of capital? Journal of Banking & Finance, 35, 23882406.. 政治大 moderating role of disclosure. Global Finance Journal, 28, 45-64. 立. Fatemi, A., Glaum, M. & Kaiser, S. (2018). ESG performance and firm value: The. Friede, G., Busch, T. & Bassen, A. (2015). ESG and financial performance: Aggregated. ‧ 國. 學. evidence from more than 2000 empirical studies. Journal of Sustainable Finance &. ‧. Investment, 5, 210-233.. sit. y. Nat. Godfrey, P. C. (2005). The relationship between corporate philanthropy and. io. Review, 30, 777-798.. er. shareholder wealth: A risk management perspective. The Academy of Management. al. n. v i n Hoepner, A. G. F., Oikonomou,CI., Sautner, Z., Starks, h e n c h i U L. T., & Zhou, X. Y. (2018). ESG shareholder engagement and downside risk. AFA 2018 paper.. Hong, H., Kubik, J. D., & Scheinkman J. A. (2012). Financial constraints on corporate goodness. Working paper. Husted, B. W. (2005). Risk management, real options, and corporate social responsibility. Journal of Business Ethics, 60, 175-183. Hutton, A. P., Marcus, A. J., & Tehranian, H. (2009). Opaque financial reports, R2, and crash risk. Journal of Financial Economics, 94, 67-86. Jo, H., & Na, H. (2012). Does CSR reduce firm risk? Evidence from controversial industry sectors. Journal of Business Ethics, 110, 441–456. 25. DOI:10.6814/NCCU202000150.

(31) Kim, Y., Li, H., & Li, S. (2014). Corporate social responsibility and stock price crash risk. Journal of Banking & Finance, 43, 1-13. Makni, R., Francoeur, C., & Bellavance, F. (2009). Causality between corporate social performance and financial performance: Evidence from Canadian firms. Journal of Business Ethics, 89, 409-422. Oikonomou, I., Brooks, C., & Pavelin, S. (2012). The impact of corporate social performance on financial risk and utility: A longitudinal analysis. Financial Management, 41, 483-515.. 政治大 Institutional aspects, performance, and investor behavior. Journal of Banking & 立. Renneboog, L., Horst, J. T., & Zhang, C. (2008). Socially responsible investments:. Finance, 32, 1723-1742.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. en chi. i n U. v. 26. DOI:10.6814/NCCU202000150.

(32) Table 1 Descriptive statistics Variables. Observations. Number of. mean. STD. 25th. median. 75th. min. max. 11.16. 13.22. 16.67. 0.88. 75.62. 15.50. 34.11. 0.78. 82.17. 8.77. 19.30. 1.00. 86.67. 51.79. 51.79. 1.00. 85.71. 0.0222. 0.0313. 0.0081. 0.0886. 0.0093. 0.0135. 0.0002. 0.0420. 0.0188. 0.0272. 0.0066. 0.0846. firms 21,426. 3,305. 17.1325. 10.1499. E score. 4,586. 694. 21.0431. 17.5769. S score. 12,747. 2,012. 15.8356. G score. 21,258. 3,297. 51.2633. 6.20 治政 13.1228 8.33 大 5.4911 48.21. Total risk. 21,426. 3,305. 0.0257. 0.0135. 0.0162. Systematic risk. 21,426. 3,305. 0.0110. 0.0072. 0.0065. Idiosyncratic risk. 21,426. 3,305. 0.0223. 0.0129. 0.0133. NCSKEW. 21,426. 3,305. 0.0855. 1.7405. -0.6717. -0.050. 0.6213. -6.4945. 7.6286. RET. 21,426. 3,305. 0.1433. 0.4645. -0.1206. 0.1002. 0.3344. -0.8011. 2.5490. Ln(size). 21,426. 3,305. 7.2681. 2.0285. 5.8440. 7.2019. 8.5974. 2.5962. 13.1162. MB. 21,426. 3,305. 3.5189. 5.7653. 1.2327. 2.0094. 3.6417. 0.2413. 57.0580. LEV. 21,426. 3,305. 0.5497. 0.2457. 0.3616. 0.5538. 0.7496. 0.0420. 0.9751. ROA. 21,426. 3,305. 0.0286. 0.0699. -1.0145. 0.3188. KZ index. 15,828. 2,794. -14.864. -2.6651. 0.1630. -510.3455. 7.4702. WW index. 15,828. 2,794. -0.2475. -0.3170. -0.2179. -0.6057. 2.4135. SA index. 15,828. 2,794. -3.2832. -3.2458. -2.8817. -3.9258. -1.9991. y. sit. io. n. er. Nat. al. 0.0033. ‧. ‧ 國. 立. 學. ESG score. iv n C h50.0759 -10.1909 i U e n c h-0.3995 0.3379 0.1625. 0.0031. 0.4878. -3.8858. This table reports descriptive statistics for all variables of my study. The sample period is from 2004 to 2018. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels.. 27. DOI:10.6814/NCCU202000150.

(33) Table 2 Pearson Correlation coefficients No.. Variables. 1. 2. 3. 4. 5. 1. ESG score. 1.000. 2. Total risk. -0.285*. 1.000. 3. Systematic risk. -0.082*. 0.481*. 1.000. 4. Idiosyncratic. -0.301*. 0.967*. 0.257*. 1.000. 立. risk. 6. 7. 8. 9. 10. 11. 12. 政治大. 0.018*. -0.018*. -0.007. -0.009. 6. KZ index. 0.075*. -0.066*. 0.037*. -0.086* 0.018*. 1.000. 7. WW index. -0.190*. 0.264*. -0.019*. 0.304*. 0.016*. -0.097*. 1.000. 8. SA index. -0.350*. 0.404*. 0.024*. 0.449*. -0.006. -0.116*. 0.266*. 1.000. 9. RET. -0.006. -0.062*. -0.058*. -0.050* -0.412*. -0.047*. -0.050*. -0.024*. 1.000. 10. Ln(size). 0.587*. -0.456*. 0.090*. -0.533* 0.040*. 0.152*. -0.369*. -0.474*. -0.002. 1.000. 11. MB. 0.050*. 0.015. 0.001. 0.023*. -0.094*. 0.095*. 0.081*. -0.016. -0.037*. 1.000. 12. LEV. 0.210*. -0.160*. 0.009. -0.183* -0.012. 0.123*. -0.205*. -0.194*. 0.051*. 0.467*. 0.255*. 1.000. 13. ROA. 0.146*. -0.453*. -0.035*. 0.097*. -0.256*. -0.304*. -0.009. 0.318*. -0.092*. 0.025*. 1.000. n. Ch. y. sit. io. -0.493* 0.041*. er. Nat. 0.033*. ‧. ‧ 國. NCSKEW. 學. 5. al. 13. v ni. 1.000. This table reports Pearson correlation coefficients among variables for the 15,828 firm-year observations from 2004 to 2018. * Indicates statistical significance at 5 % level or less.. en chi U g. 28. DOI:10.6814/NCCU202000150.

(34) Table 3 The relation between ESG and financial risk. ESG score. (2). (3). (4). (5). (6). 43.2500***. 31.5100***. 3.8100***. 8.4900***. 44.5500***. 29.8400***. (28.74). (20.14). (30.34). (18.94). 0.0232***. -0.2500***. (2.85). (-33.64). (11.26) 治政-0.1045*** 大-0.0175*** -0.2297*** (-31.12) (-25.58) (-4.93) 立. -0.0119. -3.0300***. -0.0918***. 0.0041. 0.0676***. 0.0293***. -0.1205***. -0.0002. (-7.35). (0.31). (11.04). (4.68). (-9.88). (-0.02). 4.1400***. -3.4300***. -1.5000***. 1.5200***. 5.0200***. -4.4600***. (10.10). (-8.60). (-7.48). (7.91). y. (-62.00). ‧. ‧ 國. (39.31). (12.56). (-11.12). -33.4400***. -5.2400***. -2.7100***. -25.9100***. -33.8400***. (-72.36). (-23.04). (-12.17). (-58.02). (-72.76). Yes. Yes. v ni. Yes. Yes. e nYes c h i U Yes. Yes. Yes. 21426. 21426. 21426. 21426. 0.6117. 0.5833. 0.5289. 0.4434. -27.1100***. io. (-59.22). Nat. ROA. 0.9665***. n. LEV. (5.16). 學. -2.4200*** (-48.24). MB. Idiosyncratic risk (×1000). (1). (-1.43) Ln(size). Systematic risk (×1000). al. Industry FE. Yes. Yes. Year FE. Yes. Yes. Observations. 21426. 21426. R square. 0.5434. 0.4932. Ch. sit. Intercept. Total risk (×1000). er. Dependent variables. This table displays OLS regressions for the sample over the period of 2004–2018. As measures of financial risk (dependent variables), I employ the total risk (Columns 1–2), systematic risk (Columns 3–4), and idiosyncratic risk (Columns 5–6), respectively. All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels and all coefficients reported have been multiplied by 1000 due to variable scaling issues. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively. 29. DOI:10.6814/NCCU202000150.

(35) Table 4 The relation between ESG and downside risk (1). (2). (3). (4). (5). (6). 0.2682. 0.5799**. -39.7736***. -17.6835**. -0.1957***. -3.4114***. (0.95). (2.07). (-4.88). (-4.13). (-49.26). -0.0035**. 0.0023*. -0.0052***. -0.0139***. (-2.23). (1.74). (-22.32). (-40.98). Yes. Yes. Yes. Yes. 15828. 15828. 0.3493. 0.3341. (12.96) 0.0080***. (4.49). (3.45). -0.2335***. -0.0325. (-3.03). (-0.46). 0.3993***. 0.5673*** (6.86). io. (4.63). ‧. 0.0106***. Nat. ROA. 3.5269***. 學. LEV. (-2.21) 治政-0.1735*** 大0.1884*** (-3.62) (4.81). 立. 0.0643*** (6.81). MB. SA index. y. Ln(size). WW index. sit. ESG score. KZ index. Yes. Year FE. Yes. Yes. Observations. 21426. 21426. R square. 0.0338. 0.0317. Yes. n. Yes. al. Industry FE. Ch. er. Intercept. NCSKEW. ‧ 國. Dependent variables. Yes. v ni. Yes Yes en chi U 15828 15828 0.1641. 0.1551. This table displays OLS regressions for the sample over the period of 2004–2018. As measures of downside risk (dependent variables), I employ the NCSKEW (Columns 1–2), KZ index (Columns 3–4), WW index (Column 5) and SA index (Column 6), respectively. All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively. 30. DOI:10.6814/NCCU202000150.

(36) Table 5 The relation between ESG individual components and financial risk (1). (2). (3). 21.8900***. 30.9300***. (12.69). (15.28). (7). (8). 45.8800***. 9.0000***. 8.7600***. 12.6000***. (26.43). (9.37). (10.52). (10.45). 立. 治 11.8400*** 政24.6000*** (11.46) (12.26) 大. -0.0156***. (-4.90). (-7.09). (-2.73). -0.0118. (-19.85). (-1.08) -0.3565***. -0.0702***. (-26.51). (-3.10). Nat. 0.0340**. (0.63). (2.00). -0.5063. -6.4000***. (-0.65). (-12.15). -32.2600***. -36.7100***. Ch -34.0800***. (-20.93). (-57.56). (-73.33). Industry FE. Yes. Yes. Year FE. Yes. Observations R square. -0.0106***. -0.0091. (-2.94). (-1.48) -0.0124*. -0.0178. (-1.92). (-1.40). 0.0122. 0.0297***. 0.0016. (1.51). (4.69). (0.13). 0.0033. 0.0011. 0.0052. (0.25). (0.05). (0.44). 0.3732. 0.5208. 1.7700***. 1.4100***. 0.8263*. (1.20). (7.10). (7.33). (1.85). -3.5800***. -2.8200***. -10.6300***. io. 0.0133. -4.1200***. al. n. ROA. -0.0259***. -0.1510***. G score. LEV. -0.0497***. ‧. ‧ 國. S score. (5). 學. (-11.45). MB. (6). -0.0747***. (4). (-10.27). (0.47). e -31.2300*** n chi. y. E score. Systematic risk (×1000). sit. Intercept. Total risk (×1000). er. Dependent variables. iv n U-11.0700***. (-19.68). (-12.84). (-11.84). (-12.67). (-11.93). Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. 4586. 12747. 21258. 4476. 4586. 12747. 21258. 4476. 0.6104. 0.4989. 0.4872. 0.6174. 0.7602. 0.6191. 0.5830. 0.7636. (continued) 31. DOI:10.6814/NCCU202000150.

(37) Table 5 (continued) Idiosyncratic risk (×1000). Dependent variables intercept E score. (9). (10). (11). (12). 17.7600***. 28.3800***. 45.9400***. 20.6800***. (11.58). (14.03). (26.29). (10.85). -0.0729***. -0.0490***. (-12.58). (-5.44). S score. -0.1625***. -0.0081. (-21.39). (-0.84). G score MB. 0.0115. 0.0352**. -0.3965***. -0.0764***. (-29.29). (-3.80). -0.0012. -0.0003. (2.08) (-0.09) 治政 -1.1700* -7.7600*** -5.1900*** 大 (-1.70) (-14.75) (-12.85) 立 (0.61). Nat. R square. -34.5000***. (-22.44). (-58.18). (-73.75). (-21.18). Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. 4586. 12747. 21258. 0.4915. 0.4431. 0.4368. y. Observations. -37.0700***. ‧. Year FE. (-0.46). -30.7400***. 學. Industry FE. ‧ 國. ROA. -0.3233 -29.8200***. 4476 0.4981. sit. LEV. (-0.01). n. al. er. io. This table displays OLS regressions for the sample over the period of 2004–2018. As measures of financial risk (dependent variables), I employ the total risk (Columns 1–4), systematic risk (Columns 5–8), and idiosyncratic risk (Columns 9–12), respectively. All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels and all coefficients reported have been multiplied by 1000 due to variable scaling issues. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively.. Ch. en chi. i n U. v. 32. DOI:10.6814/NCCU202000150.

(38) Table 6 The relation between ESG individual components and downside risk (1). (2). (3). (4). (5). (6). (7). (8). -0.2276. 0.2332. 0.3599. -0.1209. -6.8939**. -8.5174. -33.4126***. 0.8362. (-0.58). (0.66). (1.17). (-0.25). (-2.32). (-1.05). (-3.70). (0.21). (1.33). -0.0529*** 大 (-4.03). 政0.0031治. 0.0031**. 立. (2.11) 0.0019. 0.0018. (1.41) G score. (0.74) 0.0048**. -0.0044. (2.00). (-0.86). 0.0068**. 0.0081***. 0.0046. (0.98). (2.31). (3.46). (0.94). 0.0662. 0.0488. -0.0337. 0.1145. (0.38). (0.53). (-0.47). (0.64). 1.1627***. 0.7703***. (3.34). (6.94). (6.89). Industry FE. Yes. Yes. Yes. Year FE. Yes. Yes. Yes. Observations. 4586. 12747. R square. 0.0607. 0.0364. LEV. io. al. 0.5672***. n. ROA. Nat. 0.0047. Ch. 1.2796*** (3.57). e nYes c h i. (-0.10) 0.1148***. -0.0261. (3.53). (-1.20). ‧. MB. 學. ‧ 國. S score. -0.0021. 0.3732***. -0.1437***. (5.04). (-3.13). y. E score. KZ index. sit. Intercept. NCSKEW. er. Dependent variables. i n U. v. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. 21258. 4476. 3605. 9557. 15709. 3520. 0.0316. 0.0601. 0.5208. 0.1936. 0.1553. 0.5283. (continued) 33. DOI:10.6814/NCCU202000150.

(39) Table 6 (continued) (9). (10). (11). (12). (13). (14). (15). (16). -0.3571***. -0.2738***. 0.1271**. -0.3009***. -3.7669***. -3.6081***. -2.4352***. -3.4910***. (-5.16). (-4.58). (2.37). (-3.29). (-45.84). (-41.89). (-31.04). (-32.43). 治政 -0.0014***. -0.0020***. 立. (-6.56) -0.0037***. -0.0004. (-15.31). Year FE. Yes. Yes. Observations. 3605. 9557. R square. 0.2644. 0.3260. -0.0011. (-18.73). (-1.03). Yes. Yes. Yes. Yes. Yes. Yes. 15709. 3520. 3605. 0.2629. 0.4055. al. 0.3424. n. Yes. -0.0083***. io. Yes. Nat. Industry FE. (-0.73). Ch. n en chi U. iv. -0.0026*** (-4.59) -0.0096***. -0.0031***. (-27.63). (-5.15) -0.0240***. -0.0039***. ‧. G score. 學. ‧ 國. S score. (-2.90). -0.0056*** 大 (-15.41). (-37.29). (-3.14). Yes. Yes. Yes. y. E score. SA index. Yes. Yes. Yes. 9557. 15709. 3520. 0.3496. 0.3245. 0.4205. sit. Intercept. WW index. er. Dependent variables. This table displays OLS regressions for the sample over the period of 2004–2018. As measures of downside risk (dependent variables), I employ the NCSKEW (Columns 1–4), KZ index (Columns 5–8), WW index (Column 9-12) and SA index (Column 13-16), respectively. All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively.. 34. DOI:10.6814/NCCU202000150.

(40) Table 7 The relation between ESG individual components and downside risk (small sample) (2). (3). (4). (5). (6). -0.2511. -0.3903. -0.3332. -0.3427. -0.4476. -0.1209. (-0.55). (-0.95). (-0.82). (-0.84). (-0.97). (-0.25). 0.0047**. 0.0039**. (2.06). (1.97). E score. 0.0030**. 0.0031. (2.02). (1.33). S score G score Ln(size). -0.0170 (-0.72). (0.74) 0.0030. -0.0044. (0.81). (-0.86). 0.0039. 0.0045. 0.0045. 0.0043. 0.0042. 0.0046. (0.79). (0.92). (0.93). (0.88). (0.87). (0.94). 0.1497. 0.1108. 0.1182. 0.1074. 0.1404. 0.1145. (0.80). (0.62). (0.66). (0.60). ‧. ROA. (1.79). 政治大. 立. ‧ 國. LEV. 0.0018. 學. MB. 0.0030*. (0.64). y. ESG score. (1). (3.72). (3.57). Yes. Yes. Yes. Yes. 4476. 4476. 0.0591. 0.0601. (0.78). 1.3094*** 1.2805*** 1.2860*** 1.2846*** 1.3300*** 1.2796***. Industry FE. Nat. Year FE. Yes. Observations. 4476. R square. 0.0600. (3.58). (3.60). (3.59). Yes. Yes. Yes. Yes. Yes. Yes. Yes. io. (3.63). n. al. sit. Intercept. NCSKEW. er. Dependent variables. 4476 4476 4476i v n Ch U 0.0599 0.0599 0.0597 en chi. (continued). 35. DOI:10.6814/NCCU202000150.

(41) Table7(continued) Dependent variables Intercept ESG score. KZ (7). (8). (9). (10). (11). (12). -3.7056. -5.9145*. -7.5312**. -6.8776**. 2.2922. 0.8362. (-0.99). (-1.88). (-2.44). (-2.21). (0.62). (0.21). -0.0716***. -0.0835***. (-3.46). (-4.79). E score. -0.0568***. -0.0021. (-4.24). (-0.10). S score G score Ln(size). -0.2272. -0.0261. (-4.47). (-1.20). 政治大. 立. (-1.08). -0.0661***. Yes. Yes. Yes. Year FE. Yes. Yes. Yes. Yes. Observations. 3520. 3520. 3520. 3520. R square. 0.5272. 0.5271. 0.5264. 0.5266. -0.1437***. (-5.47). (-3.13). Yes. Yes. Yes. Yes. 3520. 3520. ‧. ‧ 國. Yes. 學. Industry FE. -0.1828***. 0.5280. 0.5283. n. al. er. io. sit. y. Nat. This table displays OLS regressions for the sample over the period of 2004–2018. The dependent variable is NCSKEW in columns (1) to (6) while the dependent variable is KZ index in columns (7) to (12). All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively.. Ch. en chi. i n U. v. 36. DOI:10.6814/NCCU202000150.

(42) Table 8 The relation between ESG and stock return RET (×1000). Dependent variables. (1). (2). (3). (4). Panel A. Total risk as a measure of financial risk Intercept ESG score. -153.0000**. -108.5900. -290.3200***. -119.7300. (-2.16). (-1.19). (-3.37). (-1.33). 0.2340 (0.35). E score. 1.2400** (2.04). S score. 1.2900** (2.09). G score. 立 -27.8200. -3159.5000. (-0.05). (5.36). (1,98). (-1.53). 10.6900. -51.5500*. -55.7700**. 64.4500. (0.37). (-1.95). (-2.14). (1.59). -9.3400***. -8.2200*. -3.3000. (-3.84). (-1.71). (-1.19). -1.3300**. 0.9452. -1.3200*. (-2.30). (0.97). (-1.87). 119.00***. -14.4200. 91.9900***. -9.0400***. y. -1.5100**. sit. io. (-3.86) (-2.57). v i (-0.39) (3.84) n C h -209.7300*** U-74.5200** 27.6400 en chi. 122.4800***. (1.21). (-2.85). (-2.48). (1.07). Industry FE. Yes. Yes. Yes. Yes. Year FE. Yes. Yes. Yes. Yes. Observations. 20681. 4548. 12428. 20521. R square. 0.1981. 0.3151. 0.2314. 0.1982. ROA. a (6.25)l n. LEV. 987.0400**. Nat. MB. 4450.2100***. ‧. Ln(size). (-0.51). er. Score* financial risk. -0.5821. 學. ‧ 國. Financial risk. 政治大. (6.40) 24.3400. (continued). 37. DOI:10.6814/NCCU202000150.

(43) Table 8 (continued) RET (×1000). Dependent variables. (5). (6). (7). (8). Panel B. Systematic risk as a measure of financial risk Intercept ESG score. -163.1800**. -55.5000. -268.9900***. -266.9700***. (-2.36). (-0.62). (-3.20). (-3.11). 1.2600** (2.06). E score. 1.4100*** (2.66). S score. 1.0900** (1.97). 政治大 1786.4400** 7709.6400*** 2271.2300** 立 -102.3500***. -74.3400*. -112.5900*. (-1.58). (-2.82). (-1.92). (-1.65). -10.6000***. -11.5500**. -5.2900**. -10.4800***. (-4.42). (-2.45). (-2.00). (-4.62). -1.4200**. 0.9040. -1.3900**. -1.6200***. (-2.46). (0.92). (-1.96). (-2.76). 121.1200***. -16.8900. 93.2400***. 124.2200***. io. LEV. -67.1900. al. y. Nat. MB. 6420.4600*. (2.49). ‧. Ln(size). (2.24). (5.79). sit. Score* financial risk. 2.3700**. (1.96). 學. ‧ 國. Financial risk. er. G score. (1.82). v i n 28.3700 C -255.2900*** -76.2600*** h e n c h i U(-2.67) (1.31) (-3.54). (6.50). Industry FE. Yes. Yes. Yes. Yes. Year FE. Yes. Yes. Yes. Yes. Observations. 20681. 4548. 12428. 20521. R square. 0.1982. 0.3157. 0.2315. 0.1982. n. (6.37) ROA. (-0.45). (3.90). 26.9800 (1.24). (continued). 38. DOI:10.6814/NCCU202000150.

(44) Table 8 (continued) RET (×1000). Dependent variables. (9). (10). (11). (12). Panel C. Idiosyncratic risk as a measure of financial risk Intercept ESG score. -145.9100**. -95.7700. -285.0600***. -87.0400. (-2.06). (-1.04). (-3.30). (-0.99). -0.2431 (-0.37). E score. 0.8642 (1.41). S score. 1.1800** (1.98). G score. 立. -496.2500. -42.4700. -64.1000**. 114.4400**. (1.27). (-1.22). (-2.04). (2.51). -9.1600***. -7.9100. -3.4000. -8.5700***. (-3.63). (-1.63). (-1.19). -1.3000**. 0.9698. -1.3200*. (-2.25). (0.99). (-1.87). (-2.53). 118.2300***. -11.7600. 92.4100***. 121.8100***. al. (-2.44). y. sit. io. LEV. 45.6100. Nat. MB. -5608.9500**. (1.78). ‧. Ln(size). (-1.16). (4.25). er. Score* financial risk. -1.2700. (-0.82). 學. ‧ 國. Financial risk. 政治大 4226.1500*** 953.0000*. (-3.52) -1.4800**. v i n 27.3300 C -212.8100*** -77.8700*** h e n c h i U(-2.59) (1.20) (-2.87). (6.35). Industry FE. Yes. Yes. Yes. Yes. Year FE. Yes. Yes. Yes. Yes. Observations. 20681. 4548. 12428. 20521. R square. 0.1981. 0.31. 0.2314. 0.1983. n. (6.19) ROA. (-0.32). (3.85). 24.2800 (1.06). This table displays OLS regressions for the sample over the period of 2004–2018. The independent variable of financial risk is total risk, systematic risk, and idiosyncratic risk in panel A, B, and C, respectively. All models include industry and time fixed effects. All variables except Bloomberg ESG disclosure score are winsorized at the bottom and top 0.5% levels and all coefficients reported have been multiplied by 1000 due to variable scaling issues. ***, **, and * indicate statistical significance at the 1, 5, and 10 % level, respectively.. 39. DOI:10.6814/NCCU202000150.

(45)