The relation between CSR performance and the credit rating of a firm
Marjan Schuitema*Abstract
This study examines the relation between Corporate Social Responsibility (CSR) and the credit rating of a firm. It also investigates the relation in different economic cycles. The findings indicate that CSR performance is positively related to credit ratings, especially for firms in the United States. The relation is weaker for European firms. The results of the regression based on individual dimensions of CSR show that the economic and environmental scores influence the credit rating positively, whereas corporate governance and social scores do not show to be related to credit ratings. Finally, there is no evidence found that the relation between CSR performance and credit ratings is unstable between different economic cycles.
Key words: CSR performance, credit ratings, economic cycles, economic, environmental, social and corporate governance performance
Word count: 8,985 January 11, 2018
2 1. Introduction
The importance of Corporate Social Responsibility (CSR) has increased intensively during the last years. The forum for Sustainable and Responsible Investment (2016) reports that out of 40.3 trillion dollars of total Assets Under professional Management (AUM)1 in the United States (U.S.), 8.72 trillion dollars are classified as Social Responsible (SR) assets. Besides the magnitude of the share in AUM, the growth rate since 2014 is 33%. These numbers suggest that not only financial performance (FP) and creation of shareholder wealth are important for firms but also non-financial performance. The forum for Sustainable and Responsible Investment specifies the institutional mission, personal values and goals, and the demand of clients as motivations for the increase in magnitude of CSR. An example of a Dutch company with a social responsible mission besides their commercial goal is Tony Chocolonely. This company sells fair trade chocolate and their mission reads as follows: “together we make 100% slave free the norm in chocolate’’ (Tony Chocolonely, 2017, pp. 17). To achieve this mission the company only works together with fair trade suppliers and exclude non-fair trade suppliers. The FP of Tony Chocolonely is overwhelming. The success of this mission is displayed in the revenue that rises from 2,4 million in 2011 to 44,9 million in 2017 (Tony Chocolonely, 2017). Another example is about the customer demand. The Dutch insurance company a.s.r., which received a credit rating of BBB+ in 20172, applies an investment policy that does not invest in countries or industries that do not meet their SRI requirements based on customer preferences. These exclusions do impact the costs of capital for the excluded firms due to the lower access to capital. For aforementioned practices the need for CSR performance is increasingly desired by various stakeholders.
A growing amount of studies focus on this development and tries to examine the relation between CSR performance and FP. However, this study focuses not on the FP, but examines the relation between CSR performance and credit ratings of firms. Credit ratings are indications of the creditworthiness of a firm. Higher credit ratings can result in higher stock prices but also better access to capital markets and lower costs of debt (Weber, 2012). Therefore, it is relevant to know whether CSR performance is related to credit ratings. In addition to this, different economic cycles are taken into account because Ryana et al. (2017) find that announcements of credit rating changes have more impact during a recession. For this reason it is relevant for firms to know whether the influence of CSR performance on credit ratings differ between economic cycles.
1
Assets under management are the total market value of assets that is managed by financial professionals as financial institutions and pension funds.
2
3
Attig et al. (2013), Jiraporn et al. (2014), and Oikonomou et al. (2014) have already examined the relation between CSR and credit ratings. They find that CSR performance positively affects credit ratings, but these studies focus all on the U.S. By contrast, Stellner et al. (2015) focus on Europe and this study does not find a significant relation between CSR performance and credit ratings. As far as known, this thesis is the first that examines the relation between overall CSR performance and individual dimensions of CSR on credit ratings. The individual dimensions are the economic, environmental, corporate governance and social performances. The data set consists of U.S. and European firms3 over different economic cycles. To be specific, this study makes use of an unbalanced panel data set consisting of 663 firms and 5,273 firm-year observations during the time period ranging from 2004 to 2013. The following research question is formulated: what is the effect of CSR
performance on the credit ratings of firms during different economic cycles?
The results show that CSR performance is positively related to credit ratings. For firms in the U.S. the relation is highly significant, but for European firms it is only significant at a 10% significance level. The regressions including the individual dimensions of CSR reveal that the economic score and the environmental score significantly influence credit ratings, whereas the corporate governance and social scores do not. In the U.S. the relation for the environmental score is positive and for the European firms the economic score turns out to be significantly related to credit ratings. Additionally, for the overall CSR score based on the full sample there is no evidence that the relation is unstable between different economic periods.
The remainder of this paper is structured as follows. Section 2 presents an overview of the relevant literature related to the central question of this paper. The end of the section states the hypotheses. In Section 3 the ordered logistic regression model is explained just like the data collection procedure. The section finishes with the descriptive statistics. Section 4 provides the results. The article finishes with the conclusion and the limitations.
2. Theoretical background and prior research
This section gives an overview of previous literature related to the research question. Both papers and theories will be discussed to build a framework for defining the hypotheses of this research that are given in Section 2.3.
2.1 Corporate Social Responsibility
Attention for integrating CSR as part of the business strategy is becoming increasingly important. The effect of implementing non-financial goals in addition to the creation of
3
4
shareholder wealth becomes more often an important issue for firms (Oikonomou et al., 2014). The European Commission defines CSR as: ‘’the responsibility of enterprises for their impacts on society to integrate social, environmental, ethical, human rights and consumer concerns into their business operations and core strategy in close collaboration with their stakeholders” (European Commission, 2017).
Although it sounds favorable to implement activities that improve social concerns, different opinions about this role are present. A well-known view on the responsibilities of firms comes from Friedman (1970) and states that the only responsibility of firms is profit maximization. In his opinion, activities related to CSR are a waste of resources and therefore not favorable for shareholder wealth. In contradiction with this view is the risk mitigation theory or the stakeholder theory. The main point of these theories is that companies can reduce their risk profile by engaging in CSR activities. This can be achieved by developing and maintaining close relationships with important stakeholders and the creation of valuable internal resources and intangibles. According to Freeman (2010, pp 25) a stakeholder is defined as: “anyone who is associated with the firm and is (or can be) affected by the behavior of the firm’’. Examples of stakeholders of firms are employees, investors, suppliers, creditors, and customers.
Due to the enormous growth in Socially Responsible Investing (SRI) an increasingly number of studies tries to answer questions related to this. Most of these papers study the relation between CSR and FP. To investigate the conclusions from these studies Margolis et al. (2009) conduct a meta-analysis of 251 studies in the time period 1972 to 2007. The results show a mildly positive relation between CSR and FP, but there are low significant correlations. Another meta-analysis by Orlitzky et al. (2003) uses 52 existing studies and the results show that, in line with Margolis et al. (2009), social and environmental performance are positively related to FP. More recently, Eccles et al. (2014) show that companies that adopted pioneering social and environmental standards, before they become legally required, perform better than companies that underperform on sustainability.
5
Attig et al. (2013) study the relation between CSR and a firm’s credit rating. They use CSR scores of MSCI ESG Research (formerly known as KLD STAT) as their measurement of CSR performance. Their credit ratings come from Standard & Poor’s (S&P). The final sample consists of 11,662 U.S. firm-year observations over the period 1991 to 2010. Their results show a significant positive relation between CSR and credit ratings. Besides this aggregate score they find that social issues and environmental issues positively affect credit ratings. This finding is in line with Sharfman and Fernando (2008) who find that improved environmental performance results in lower costs of capital. Chava (2014) concludes that firms with environmental concerns are penalized with higher interest rates. This is corresponding to the results of Oikonomou et al. (2014) who conclude that good CSR performance is rewarded and bad behavior is penalized through higher bond spreads and lower credit ratings. Jiraporn et al. (2014) use the same proxies for credit rating and CSR performance as Attig et al. (2013). Their findings are in line with Attig et al. (2013). To be specific, an increase in CSR by one standard deviation improves the firms’ credit rating by as much as 4.5%. Additionally, they show that the CSR policy of a firm is significantly influenced by the CSR policies of U.S. firms in het same three-digit zip code. Another aspect of CSR is corporate governance. Ashbaugh-Skaife et al. (2006) study this dimension and their findings indicate a positive relation between corporate governance activities and U.S. firms’ credit ratings. The recent study of Schultz, Tan and Walsh (2017) suggests that corporate governance activities are correlated, but not causally related to the financial health of the firm.
When the literature that focuses on Europe is analyzed the results are different. Stellner et al. (2015) focus on the European corporate bond market. They extend the approach of Jiraporn et al. (2014) by considering the equivalent of CSR on the country level. They use Thomson Reuters ASSET4 ESG ratings in twelve countries from the Economic and Monetary Union and they obtain ratings from S&P and Moody’s. Their results are based on a sample of 872 bonds between 2006 and 2012. They find only weak evidence that superior CSR performance systematically reduces credit risk but they do find evidence that superior CSP is rewarded in countries with above average ESG performance. An important limitation of this study is that the individual dimensions of CSR are not taken into account, but solely the aggregated CSR score.
2.2 The financial crisis
6
accuracy of credit ratings. The study of Ryana et al. (2017) examines in which way the value of credit ratings of U.S. based firms change across economic cycles and across different bond types. The main finding is that announcements of credit rating changes are more valuable during a recession. This means that in such a period a credit rating downgrade (upgrade) results in greater negative (positive) abnormal returns during a recession compared to periods of economic growth.
2.3 Hypothesis
Arguments supporting the hypothesis that good CSR performance can reduce a firms’ financial risk and enhance credit ratings are based on the following arguments: first, in favor of stakeholder theory, Waddock and Graves (1997) state that good management and CSR are positively correlated and that CSR activities improve relationships with stakeholders. Second, good CSR performance results in increased customer loyalty, increased ability to attract and retain high-quality employees (Greening and Turban, 2000), and enhances the competitive position. The creation of aforementioned intangible assets is a reason to expect that good CSR performance positively influences the long-term sustainability of the firm and reduces the probability of default. Besides this, high levels of CSR performance are expected to decrease the probability of law suits and legal fines, and to lower capital constraints because more investors are willing to invest in firms with good CSR performance (see, e.g., Cheng et al., 2014; Spicer, B., 1978). This improves the perception of product quality among consumers. Based on aforementioned arguments and the literature discussed the first hypothesis is:
Hypothesis 1: CSR positively affects credit ratings.
This study expands the existing literature by looking at the individual dimensions of CSR in addition to the overall CSR performance. Four dimensions are investigated: the economic, environmental, corporate governance and the social dimension. The economic score is expected to influence credit ratings positively because in general good economic performance lowers the probability of default.
7
The third dimension, corporate governance could influence the creditworthiness of a firm via the following two dimensions according to Bhojraj and Sengupta (2003): agency risk and information risk. The first dimension represents the risk of a selfish management that does not act in a way that maximizes firm value. A well-documented example of such behavior is taking actions that focus on short-term profits instead of long-term profits (see, e.g., Murphy and Zimmerman, 1993; Dechow and Sloan, 1991). This can negatively influence the continuity of the firm and in turn lower the creditworthiness of the firm. The second dimension they mention is information risk. This describes the risk that the management of the firm has private information that could affect the default risk of the loan and therefore the creditworthiness of the firm. Accurate disclosure of information is an example of a mechanism that can help reducing information risk. The study of Beasley (1996) reports that the higher the proportion of outsiders taking part in the management the less likely is the probability of financial statement fraud. This has in turn a positive effect on credit ratings.
The last dimension is social performance. The ASSET4 Thomson Reuters database defines the social performance measure as: “the performance of a company on the following domains: workforce, human rights, community, and product responsibility’’. It is a reflection of the company's reputation and the health of its license to operate, which are key factors in determining its ability to generate long-term shareholder value. Examples of channels through which good social performance can influence the creditworthiness of a firm are the attractiveness of the firm towards skilled labor and the loyalty of consumers towards the firm. These expectations result in the following hypothesis:
Hypothesis 2: Economic, environmental, corporate governance and social performances
positively affect the credit rating of a firm.
Based on the results of Ryana et al. (2017) it is interesting to take different economic cycles into account. This is the first study that examines the relation between CSR performance and credit ratings between different time periods, namely: pre-, during- and post-crisis. Risk management theory introduced by Godfrey (2005) states that good CSR performance results in moral capital by shareholders. This moral capital results in enhanced loyalty towards the company. When stakeholders become more loyal to the firm it decreases the possibility of stakeholder sanctions, even in times of crisis. Therefore, the future cash flows are more stable over time. For this reason the expectation is that the effect of CSR on credit ratings is the same or even higher in times of crisis compared to non-crisis periods. This results in the following hypothesis:
Hypothesis 3: The relation between CSR and credit ratings is stable across different economic
8 3. Methodology and data collection
This section starts with describing the methodology used followed by a sample overview and the data description. After that, the control variables are given and the section concludes with the descriptive statistics of the data.
3.1 Regression model
In general, the empirical tests try to examine the relation between CSR performance and credit ratings. More precisely, the purpose of this research is how the measure of CSR performance – the Thomson Reuters ASSET 4 ESG scores - impacts Fitch’s credit ratings. The credit rating of a firm, the dependent variable, is characterized by an ordinal scale ranging from 1 to 24. A credit rating of AAA, assigned a numerical value of 24 is not ‘twice as good’ as a rating of BB-, assigned a numerical value of 12. Similarly, the difference between a score of 19 and 20 cannot be assumed to be the same as the difference between a score of 9 and 10. This implies that the scale has a natural ordering and therefore only the ordering can be interpreted and not the actual numerical values. For this reason an ordered logit model is applied. A description of this model is given in Appendix A. In addition to CSR performance (measured in ESG scores), company-specific control variables that are expected to influence credit ratings are included in the model as well. Section 3.3 discusses these variables in detail. The baseline model incorporates CSR performance and control variables (firm characteristics) as independent variables. In addition to this, year fixed effects and country fixed effects are included in the model to control for differences over time and between countries. The following formula is used as baseline model:
𝑟𝑎𝑡𝑖𝑛𝑔𝑖𝑡 = 𝛽1(𝐶𝑆𝑅 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡) + 𝛽2(𝑓𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠𝑖𝑡) + 𝛽3(𝑦𝑒𝑎𝑟) +
𝛽4(𝑐𝑜𝑢𝑛𝑡𝑟𝑦) + 𝜀𝑖𝑡 (1)
Where 𝑟𝑎𝑡𝑖𝑛𝑔𝑖𝑡 is the Fitch’s credit rating for firm i in year t and depends on CSR performance measured by 𝛽1 and firm characteristics represented by 𝛽2. Year and country fixed effects are given by 𝛽3 and 𝛽4. The coefficients are denoted as odd ratios. 𝜀𝑖 is the disturbance term that is assumed to follow a cumulative standard logistic distribution.
The other models include variables as period which describes credit ratings during different economic cycles and is represented by 𝛽3. Additionally, interaction terms are included to examine the relation between CSR performance and credit ratings over different economic cycles as stated in hypothesis 3. The effect is given by 𝛽4. The following formula is used:
9
Where period represents the pre-crisis period or the during-crisis period and 𝛽4 gives the effect of the interaction between period and ESG performance. Clustered robust standard errors are applied in each model to correct for correlated errors for a firm over the sample period. The impact of CSR performance on credit ratings is examined on a yearly basis because the frequency of ESG score publications are once a year.
3.2 Sample overview and data description
The sample consists of firms that are included in the ASSET4 ESG database acquired by Thomson Reuters. In total there are 4,456 companies from all over the world present in this database. The first selection criterion is that firms have to be located in the U.S. or in the European Union (referred to as Europe). The reason for this criterion comes from the purpose of this research to observe the influence of the Financial Crisis of 2007 on the relation between CSR performance and credit ratings. The expectation is that the influence of the Financial Crisis will be most observable in the U.S. and Europe and therefore firms have to be located in those regions. When this selection criterion is applied the sample size reduces to 2,206 firms. The following EU-countries are taken into account: Austria, Belgium, Czech Republic, Germany, Denmark, Espana, Finland, France, United Kingdom, Greece, Ireland, Italy, Luxembourg, The Netherlands, Poland, Portugal, and Sweden. Due to data availability the other 9 EU-countries are not included in the sample. The second criterion is that the firms that defaulted during the sample period are excluded. This results in a sample of 1,915 firms. Finally, these firms are matched with the other independent variables which yield a final sample of 663 firms.
This study expands the literature by studying the time period before, during and after the Financial Crisis. The time periods used are similar to that of De Haan (2017) who takes the period 2004 to June 2007 as the “pre-crisis” period, the period July 2007 to June 2009 as the “during-crisis” period and July 2009 to 2013 as the ‘’post-crisis’’ period. However, as mentioned before the ASSET4 ESG database consists of yearly data. For this reason the time frame for this study ranges from 2004 to 2013 and the sub-periods for the “pre-crisis” period, “during-crisis” period, and ‘’post-crisis’’ period are 2004 to 2006, 2007 to 2009, and 2010 to 2013, respectively.
10
between 1 and 6). The reason for this specific distribution is that each category includes an equal amount of rating classes.
ASSET4 ESG performance data acquired by Thomson Reuters is used as a measure for CSR performance. ASSET4 collects data and scores firms on ESG dimensions since 2002. The coverage of the database is worldwide and includes 4,456 firms. There are several reasons to choose this database. First of all, it is a good proxy for CSR performance because it covers the main elements of the definition of CSR defined by the European Commission (2017). The second reason is the availability of the ESG scores and the different dimension scores it offers. Previous studies use KLD data (see e.g., Sharfman and Fernando, 2008; Oikonomou et al., 2012; Attig et al., 2013) or ASSET4 (Stellner et al., 2015). In contrast to ASSET4, KLD data is not available for this research. This constrain enforces the use of ASSET4. The last reason is the magnitude of the countries and companies that are included in the database. This makes it possible to compare differences between the U.S. and European countries.
Research analysts collect 500 ESG data points that are used as inputs to calculate more than 180 Key Performance Indicators (KPI). Every KPI is in the database twice, reported in two different measures: value and score. The value is the value that is calculated by making use of the data points. The score is normalized and a Z-scored version of the indicator’s value. This implies that a firm with a score above fifty performs better than other companies in the database for that specific indicator. Besides this, the database includes an overall ESG score, and four different pillars: economic, environmental, corporate governance and social. Because all the pillars are included in the ESG score and this research will examine the relation between the overall ESG score and the different dimensions, the economic score is taken into account as well.
3.3 Control variables
11
The first control variable is size. The expectation is that size influences credit ratings positively, because large firms tend to face lower business and financial risks and therefore a lower probability of default. Size is measured as the natural logarithm of total assets in millions of U.S. dollars4. The second control variable is leverage. It is the ratio of total debt to total assets. The relation between leverage and credit ratings is expected to be negative, because a firm with more debt relative to total assets is expected to be less creditworthy because of the higher default risk. Another control variable is margin. Higher operating margin should result in reduced default risk so the expected sign is positive. Additionally, loss is expected to have a negative influence on the creditworthiness of a firm. Loss is a dummy variable that is assigned a value of one if net income is negative in the current and prior fiscal year, otherwise it is zero. The reason for including this variable as control variable is that a loss increases the probability of default. Interest coverage is also a control variable. It is measured as the ratio of Earnings Before Interest and Taxes (EBIT) plus interest expense divided by interest expense. The expected relation is positive because higher interest coverage reduces default risk and in turn enhances the firms’ credit rating. The sixth variable is Return On Assets (ROA) and is measured as the net income before extraordinary items dividend by total assets. The expectation is that a higher ROA results in a lower default risk and therefore a positive relation is expected. Finally, financial utility is included. This dummy variable is equal to one if a firm is assigned to be financial and zero otherwise. The SIC code is used to decide whether a firm is financial or not. The expected relation is positive because financial firms are regulated more strictly so therefore they are expected to be more creditworthy. Despite the fact that the beta of a firm and the capital intensity are expected to influence the credit rating of a firm, these variables are not used as control variables initially. The reason for this is that it will reduce the sample extensively as can be seen in Appendix C. The observations for which these variables are present are used as robustness tests.
Aforementioned criteria result in a final unbalanced panel data set consisting of 663 firms and 5,276 firm-year observations.
3.4 Descriptive statistics
This subsection provides an overview of the descriptive statistics of the variables and the sample discussed in the previous subsections. Fig. 1 shows the distribution of credit ratings in the full sample. The majority of the firm-year observations firms are assigned a
4
Data points that are not denoted in U.S. dollars are converted to U.S. dollars via historical yearly exchange rates obtained by the European Central Bank.
12
BBB-rating (877), and the second largest category consists of BBB+-ratings (807). There are fewer observations with lower credit ratings. The dark color in the bars in Fig. 1 represents the amount of observations in the U.S. and the lighter color presents the observations in Europe. The amount of European observations is lower (1,823 observations) but the shape of the distribution is comparable to that of the U.S. (3,453 observations).
Figure 1
Frequency per rating
This graph shows the distribution of the credit ratings for the full sample. Fitch’s company credit ratings are converted to an ordinal scale according to the following schedule: 24 (AAA), 23 (AA+), 22 (AA), 21 (AA-), 20 (A+), 19 (A), 18 (A-), 17 (BBB+), 16 (BBB), 15 (BBB-), 14 (BB+), 13 (BB), 12 (BB-), 11 (B+), 10 (B), 9 (B-), 8 (CCC+), 7 (CCC), 6 (CCC-), 5 (CC+), 4 (CC), 3 (CC-), 2 (C), and 1 (D). The sample consists of 5,276 observations. The dark color in the bars is the observations in the U.S. and the lighter color represents the European observations.
In addition to the distribution of credit ratings it is also useful to look at the distribution of credit ratings over the different time periods. This is displayed in Table 1. To get a clear overview of the distribution this table shows only the four categories as discussed in Section 3.2. Appendix D.1 includes a comprehensive table of the distribution of all 24 credit rating categories in each year. What is already observed in Fig. 1 is the relatively low amount of observations with poor credit ratings. In the pre-crisis period the amount of observations with poor ratings is only one (0.08%). The total amount of observations per time period is increasing over time as can be seen in the bottom row. Whereas the proportion of the excellent rating category is decreasing, the magnitudes of the other categories are increasing over time.
13 Table 1
Amount of credit rating observations over different time periods Period
Rating Pre-crisis During-crisis After-crisis
Excellent 475 (39.98%) 539 (33.46%) 605 (24.42%)
Good 657 (55.30%) 956 (59.34%) 1,655 (66.81%)
Moderate 55 (4.63%) 111 (6.89%) 209 (8.44%)
Poor 1 (0.08%) 5 (0.31%) 8 (0.32%)
Total 1,188 1,611 2,477
This table presents the distribution of credit ratings over different time periods. It is based on the full sample of 5,273 firm-year observations. The ‘’pre-crisis’’ period ranges from 2004-2006, the ‘’during-crisis’’ period ranges from 2007-2009, and the ‘’post-‘’during-crisis’’ period includes the years 2010-2013. The four rating categories are as follows: “excellent” (ratings between 19 and 24), “good” (ratings between 13 and 18), “moderate” (ratings between 7 and 12), and “poor” (ratings between 1 and 6). The bottom row sums the total firm-year observations per period.
Appendix D.2 shows the distribution of firm-year observations per country. As already mentioned, the majority of the observations are obtained from firms located in the U.S. The European countries with the highest amount of observations are the United Kingdom (582), France (320), and Germany (233). Countries with the lowest amount of firm-year observations are Czech Republic (7), Ireland (8), and Greece (9).
After describing the distribution of the credit ratings over time and geographical location it is also relevant to look at the descriptive statistics of the variables of interest. Table 2 shows the summary statistics of the following variables: credit rating, ESG score, economic score, environmental score, corporate governance score, and social score. The scores are ranging from 0 to 100, except for the credit rating. Row 1 of Table 2 shows that the mean credit rating for the full sample is 17 (BBB+). The maximum rating in the full sample is AAA+ and the minimum is a rating of D.
14 Table 2
Summary statistics of regression variables
This table reports the mean, median, maximum (max.), minimum (min.), standard deviation (st. dev.), skewness (skew.) and kurtosis of the following variables: credit rating, ESG score, economic-, environmental -, corporate governance-, and social scores. The results are based on the full sample of 5,273 firm-year observations. The data is obtained via DataStream. Fitch’s company credit ratings are converted to an ordinal scale according to the following schedule: 24 (AAA), 23 (AA+), 22 (AA), 21 (AA-), 20 (A+), 19 (A), 18 (A-), 17 (BBB+), 16 (BBB), 15 (BBB-), 14 (BB+), 13 (BB), 12 (BB-), 11 (B+), 10 (B), 9 (B-), 8 (CCC+), 7 (CCC), 6 (CCC-), 5 (CC+), 4 (CC), 3 (CC-), 2 (C), and 1 (D).
The skewness and kurtosis are also reported in Table 2. Skewness measures the asymmetry of the distribution of the series around its mean. The negative skewness of credit ratings implies that the distribution of credit ratings has a long left tail. This is in line with Fig. 1 which shows that the amount of moderate and poor credit ratings is low. By contrast, kurtosis measures the peakedness or flatness of the distribution of the series. A kurtosis of three would indicate that the distribution is normal. As can be seen in the last column of Table 2 the variables do not have a normal distribution. However, because of the large sample size this will not cause any problems in the regressions.
Appendix E shows a deviation between the summary statistics for the variables based on the U.S. and Europe. For the observations in the U.S. the mean credit rating is 16.689 and for Europe it is 17.602. The striking observation from Appendix E is that the mean ESG score in the U.S. (61.317) is much lower than for European firms (79.982). A t-test indicates that this difference is statistically different. The same holds for the four individual dimensions, except the corporate governance score which is higher for U.S. firms.
In addition to Table 2, Appendix F.1 shows the same statistics for the control variables. It is however, more interesting to look at the pair-wise correlation between the regression variables. Appendix F.2 reports that credit ratings are positively correlated with the ESG score and the four individual dimensions. This implies that better ESG performance is related to a higher credit rating. Additionally, the signs of all of the control variables are as expected.
15
Finally, variance inflation factors (VIFs) are calculated to test for multicollinearity. If the value of a VIF is higher than ten this could affect the results. However, the VIFs calculated for this data set do not exceed 1.76. This implies that multicollinearity does not affect the results.
4. Results
As described in the previous section on methodology this study performs ordered logistic regressions to examine the relation between CSR performance of firms and their credit ratings.
4.1 Regression results
Table 3 shows the coefficients for six different models, which have to be interpreted as odd ratios. The odd of an event happening is defined as the probability that the event occurs divided by the probability that the event does not occur. In this case, an odd ratio shows by how much the odds of the dependent variable (credit rating) will change for each unit change in the independent variable. An odd ratio smaller than one indicates that the odds decrease as the independent variable increase and for an odd ratio above one it is the other way around. An odd ratio of one indicates that there is no relation between the variables. The first two models show the results based on 24 cutoff points for the full sample for both the ESG score (Model 1) as the four different dimensions (Model 2). Models 3 and 4 show the results for the same regressions as in Models 1 and 2 but with four different cutoff points and in these models interaction terms are included. Models 5 and 6 show the results based on subsamples for the U.S. and Europe.
16
The control variables in Model 1 show the following: size, leverage, loss, ROA and financial utility have a highly significant effect on credit ratings. The relations are as expected. For the variable financial utility this implies that when a firm is financial it is 3.446 times more likely to be in a higher category compared to non-financial firms. Model 1 does not find a significant effect for margin and interest coverage. It is worth mentioning the absence of a constant (intercept) in the results. The reason for this is that this constant would be exactly collinear with the threshold variables which are not shown in the presented results but which are estimated in the ordered model.
Model 2 examines the relations between the four different dimensions of ESG on credit rating. Economic and environmental scores turn out to be positively related to credit ratings at a 1% and 5% significance level, respectively. The social score is also positively related to credit rating but only at a 10% significance level. Corporate governance performance does not have an effect on credit ratings according to the results. The observed positive relations between environmental and social performance is in line with the findings of Margolis et al. (2009) and Orlitzky et al. (2003). The estimated coefficient for governance performance is in line with previous literature as well. The coefficients of the control variables have the same signs and comparable magnitudes as predicted in Model 1.
The Wald test shows that the coefficients of the thresholds are not significantly different from each other so therefore the 24 categories are divided in four subgroups: “excellent”, “good”, “moderate”, and “poor”. Models 3 and 4 show the results for the regression based on these subgroups for which the confidence intervals are significantly different now. The coefficients for the overall ESG score, individual dimensions and control variables are comparable to the results in Models 1 and 2, except for the social dimension which becomes insignificant in Model 4.
Based on the year fixed effects in Models 1 and 2 some differences between the post-crisis period and the other time periods can be expected. To observe this in greater depth, Models 3 and 4 include dummy variables for different time periods. The results for these dummies indicate that credit ratings in the post-crisis period (2010 to 2013) are significantly lower compared to the other periods. A reason for this could be that firms have become less creditworthy due to the crisis and subsequently receive lower credit ratings.
17 Table 3
Ordered logit models of the effect of CSR performance on credit ratings
Sample Full sample U.S. Europe
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 ESG performance ESG score 1.014*** (0.003) 1.020*** (0.004) 1.021*** (0.004) 1.013* (0.007) Economic score 1.006*** (0.002) 1.008** (0.004) Environmental score 1.006** (0.003) 1.010** (0.005) Governance score 0.999 (0.004) 1.003 (0.006) Social score 1.005* (0.003) 1.005 (0.006) Control factors Size 2.098*** (0.120) 2.071*** (0.120) 2.135*** (0.159) 2.092*** (0.159) 1.947*** (0.171) 3.206*** (0.399) Leverage 0.976*** (0.003) 0.976*** (0.003) 0.973*** (0.004) 0.972*** (0.004) 0.968*** (0.005) 0.985** (0.008) Margin 1.000 (0.001) 1.000 (0.001) 0.999* (0.001) 0.999* (0.001) 0.999*** (0.000) 1.047*** (0.014) Loss 0.316*** (0.074) 0.324*** (0.076) 0.336*** (0.082) 0.328*** (0.079) 0.393*** (0.112) 0.258*** (0.102) Interest coverage 1.001 (0.001) 1.001 (0.001) 1.001 (0.001) 1.001 (0.001) 1.000 (0.001) 1.005 (0.004) ROA 1.089*** (0.014) 1.086*** (0.014) 1.086*** (0.014) 1.087*** (0.014) 1.105*** (0.015) 1.025 (0.016) Financial utility 3.446*** (0.574) 3.550*** (0.610) 4.327*** (0.853) 4.676*** (0.959) 4.550*** (1.067) 3.153*** (1.190)
Dummy variables and interaction terms
Pre-crisis period 2.586*** (0.823) 4.283*** (1.971) 1.761* (0.602) 13.630*** (8.746) During-crisis period 2.544*** (0.570) 3.346*** (1.149) 2.283*** (0.561) 5.494*** (2.903)
ESG score * pre-crisis 0.998
(0.692)
1.002 (0.005)
0.983** (0.008)
ESG score * during-crisis 0.997
(0.003)
0.997 (0.003)
0.989* (0.006)
Year fixed effects Yes Yes No No No No
Country fixed effects Yes Yes Yes Yes No No
N 5,276 5,276 5,276 5,276 3,453 1,823
Pseudo R-squared 0.131 0.131 0.304 0.309 0.279 0.387
Number of indicator values 24 24 4 4 4 4
18
According to hypothesis 3 it is expected that the relation between ESG performance and credit ratings is stable across different economic cycles. The results indicate that there is no evidence that this relationship is unstable over time, at least for the ESG score as can be seen in Model 3.
The same regression as used in Model 3 is performed for the individual dimensions. Due to the large amount of interactions terms these are not included in model 4, but Model 4.1 in Appendix G do show the results. For the full sample the relation between the economic score and credit ratings is lower during the crisis compared to the post-crisis period. Additionally, the corporate governance score is lower during the pre-crisis period compared to the post-crisis period. However, the results in Models 5.1 and 6.1 in Appendix G indicate that the findings are not robust. Because regressions based on subsamples for firms in the U.S. (Model 5.1) and Europe (Model 6.1) find no evidence for unstable relations between individual dimensions and credit ratings over time.
The results in Model 1 are in line with previous literature based on U.S. samples but in contrast to conclusions from studies based on Europe. To examine this difference in greater depth Models 5 and 6 in Table 3 are performed. Model 5 is based on a subsample in which only U.S. firms are included and Model 6 represents the European firms. The results confirm the expectations based on previous literature. Model 5 shows that the effect of the ESG score is highly significant in the U.S. The coefficient of 1.021 implies that the odd of moving to a higher credit rating category is 2.1% when ESG score increases by one unit.
Model 6, conversely shows that the effect of ESG score on credit ratings in Europe is only significant at a 10% significance level. Oikonomou et al. (2011) state as a possible explanation for this observation that Europe might lag in the recognition of the potential benefits of CSR investments in comparison with the U.S. Another remarkable result is the difference between credit ratings before and during the crisis compared to the post-crisis period in Europe. An explanation for this finding is that a few firms are downgraded significantly over the sample period.
19
European firms than for U.S. firms. The respectively percentages change are 220.6%, and 94.7%.
In addition to Models 5 and 6, tests for the individual dimensions are also performed for subsamples based on U.S. and Europe. The results in Models 5.1 and 6.1 in Appendix G show that the environmental score is the only dimension which significantly influences credit ratings in the U.S. For the sample based on Europe this is the economic score.
4.2 Robustness checks
Due to a large amount of missing variables of beta and capital intensity these variables are not used as control variables in the Models 1 to 6. However, based on previous literature these variables are expected to influence credit ratings. Therefore these two variables are used as robustness tests. The results based on the specification in Model 3 show that when these variables are added to the list of control variables the relation between ESG score does not change.
Stellner et al. (2015) find that bonds’ credit ratings gain more from good ESG performance if the country where the firm is located scores above the average ESG score. Due to the fact that such a country ESG score is not provided in the ASSET4 ESG database this study uses the average overall ESG score instead. To test whether above or below average scores make sense, the sample is divided into two subsamples. The results are presented in Table 4. Model 8 shows the coefficients for the subsample with above average scores (>67.766) and Model 9 include the firms with below average ESG scores (<67.766). The effect of ESG score on credit ratings is highly significant for the subsample with above average ESG scores. In contrast to this, the coefficient in Model 9 is not significant at all. This is in line with the findings of Stellner et al. (2015). It implies that above average ESG performance is more valued in credit ratings than below average ESG performance. Another remarkable observation in Model 8 is the large difference between credit ratings in the pre-crisis period compared to the post-pre-crisis period. The same test as in Models 8 and 9 are applied for the deviation between above and below average firm size but the results are robust to the earlier findings.
20 Table 4
Ordered logit models of the effect of ESG performance on credit ratings
Sample Above average
mean ESG score
Below average mean ESG score
Financial utility = 0
Model 8 Model 9 Model 10
ESG performance ESG score 1.041*** (0.015) 1.007 (0.007) 1.021*** Control factors Size 2.482*** (0.224) 1.789*** (0.166) 1.922*** (0.185) Leverage 0.970*** (0.007) 0.977*** (0.004) 0.960*** (0.005) Margin 1.030*** (0.011) 0.999*** (0.001) 1.037*** (0.010) Loss 0.303*** (0.111) 0.293*** (0.090) 0.396*** (0.117) Interest coverage 1.001 (0.001) 1.001 (0.002) 1.000 (0.001) ROA 1.095*** (0.020) 1.037** (0.016) 1.059*** (0.014) Financial utility 2.810*** (0.838) 6.314*** (1.544) -
Interaction terms and dummy variables Pre-crisis period 82.157*** (134.261) 1.666 (0.737) 1.932* (0.762) During-crisis period 10.301 (15.119) 1.658 (0.514) 2.081** (0.627)
ESG * pre-crisis period 0.960** (0.118)
1.013 (0.011)
1.000 (0.005) ESG * during-crisis period 0.981
(0.016)
1.007 (0.007)
0.998 (0.004)
Year fixed effects No No No
Country fixed effects Yes Yes Yes
N 3,150 2,126 3,687
Pseudo R-squared 0.2983 0.2614 0.2561
Number of indicator values 4 4 4
21
4.3 Endogeneity
As a final robustness check the results are verified against potential endogeneity stemming from reverse causality. It can be argued that firms that perform financially well can invest more in CSR and in turn receive higher credit ratings. Chih et al. (2010) show that firms are more willing to invest in CSR if the financial performance is good. To control for this potential bias, the ESG score in Model 3 is replaced by the ESG score with one lag. The result of the coefficient for the lagged ESG score remains significant and therefore there is no evidence that the result in Model 3 is driven by reverse causality. The positive relations between the economic and environmental scores and credit ratings in Model 4 are tested as well. The lagged variable for the economic score is still significant but the environmental score becomes insignificant. These results strengthen the aforementioned conclusion that reserve causality does not influence the findings.
5. Conclusion and discussion
This study uses an unbalanced panel data set consisting of 663 firms and 5,273 firm-year observations during the time period ranging from 2004 to 2013 to answer the following question: what is the effect of CSR performance on the credit rating of a firm during different
economic cycles?
The results indicate that the risk mitigation view or stakeholder theory is valid. This implies that companies can reduce their risk profile by engaging in CSR activities, because the relation between the ESG performance is positively and significantly related to credit ratings. To be specific, an increase in the ESG score by one unit implies that it is 1.020 times more likely to be in a higher credit rating category. The existing literature finds contradicting results for samples with U.S and European firms. To examine this difference the tests are also performed for subsamples based on the U.S. and Europe. The results are in line with previous literature. Whereas CSR performance does not significantly affect (only at 10%) credit ratings in Europe (Stellner et al. 2015), it has a positive and highly significant effect in the U.S. (see, e.g., Attig et al., 2013; Jiraporn et al., 2014; Oikonomou et al., 2014). Oikonomou et al., (2011) state as a possible explanation for this difference that Europe might lag in the recognition of the potential benefits of CSR investments in comparison with the U.S.
22
credit ratings. For Europe the economic score is positively related to credit ratings and for the U.S. this is the environmental score. It implies that hypothesis 2 only holds partially.
Another important feature of this study is that it includes different economic cycles, defined as follows: the pre-crisis period, the during-crisis period, and the post-crisis period. The results find no evidence against hypothesis 3 which states that the relation between ESG performance and credit ratings is stable over time. For the individual dimensions only the economic and corporate governance scores are slightly different, but the differences are small.
The findings of this study have implications for firms. An increase in credit rating results in better access to capital markets and a lower cost of debt capital. For this reasons it is advisable for firms to improve their CSR performance. Especially, investments that enhance the economic and environmental scores are recommended according to the results. For example, investing in adjustments that reduce emissions or lower the need for natural resources. Another result stemming for the robustness tests is that the impact of CSR performance is more valued for firms with above average CSR performance. This implies that firms are recommended to at least achieve ASSET4 scores of 50.
Although this study carefully performs accurate tests, there are limitations as well. The first limitation of this study is that it uses the Thomson Reuters ASSET4 database for measuring CSR performance. Despite the fact that this database is widely used in other studies and the scores are created professionally it is advisable to test the reliability of these scores with other databases such as KLD Research & Analytics. The second limitation relates to the use of Fitch’s credit ratings as risk measure. The amount of missing observations for this variable is large and this could influence the results. Additionally, the reliability and accuracy of credit ratings is questionable. The reason for this is that firms that received a good credit rating collapsed shortly after receiving such a good rating and therefore the credit ratings before the crisis were probably too high.
23 Appendix A
A.1. Ordered logistic model
𝑅𝑖∗ = 𝑋𝑖𝛽 + 𝜀𝑖 (3) 𝑅𝑖 { 1 𝑖𝑓 𝑅𝑖∗ ≤ 𝜇0 2 𝑖𝑓 𝜇0 < 𝑅𝑖∗ ≤ 𝜇1 3 𝑖𝑓 𝜇1 < 𝑅𝑖∗ ≤ 𝜇2 4 𝑖𝑓 𝜇2 < 𝑅𝑖∗ ≤ 𝜇3 5 𝑖𝑓 𝜇3 < 𝑅𝑖∗ ≤ 𝜇4 … 24 𝑖𝑓 𝑅𝑖∗ > 𝜇23
24 Appendix B
B.1. Variables description
This table provides an overview of all variables used in the regression. Variables are defined in three groups: dependent variables, CSR variables and control variables. The last column shows the source of the variable. Between the brackets the DataStream code is displayed.
Variable Description Source
Fitch Credit rating
The company's credit rating as provided by Fitch. DataStream (ECSLO05V)
CSR variables
Equal-Weighted Rating
The Equal-Weighted Rating is an example of how a company's financial and extra-financial health can be equally weighted based on the information in ASSET4's economic, environmental, social and corporate governance pillars.
Thomson Reuters ASSET4 (A4IR)
Economic Score The economic pillar measures a company's capacity to generate sustainable growth and a high return on investment through the efficient use of all its resources.
Thomson Reuters ASSET4
(ECNSCORE) Environmental
Score
The environmental pillar measures a company's impact on living and non-living natural systems, including the air, land and water, as well as complete ecosystems.
Thomson Reuters ASSET4
(ENVSCORE) Social Score The social pillar measures a company's capacity to generate
trust and loyalty with its workforce, customers and society, through its use of best management practices.
Thomson Reuters ASSET4 (SOCSCORE) Corporate Governance Score
The corporate governance pillar measures a company's systems and processes, which ensure that its board members and executives act in the best interests of its long term shareholders.
Thomson Reuters ASSET4
(CGVSCORE)
Control variables
Size Natural logarithm of total assets in millions of US dollars. DataStream (WC02999) Leverage Ratio of debt to total assets (TA). DataStream
(WC08236) Coverage Earnings before interest and taxes divided by interest
expense on debt.
DataStream (WC08291)
Margin Ratio of operating income to sales. DataStream
(WC08316) Loss Indicator variable and marked one if net income is negative
in the prior and current fiscal year and zero otherwise.
DataStream (WC01551) ROA Net income before extraordinary items divided by TA. DataStream
(WC08326) Financial utility Indicator variable which is marked one when the firm is
financial according to SIC CODES and zero otherwise.
25 Appendix C
C.1. Amount of missing data points per variable
This table presents the missing variables per year and per variable in DataStream. These amounts are based on the sample of 1,915 firms for the time period from 2004 to 2013. The description of these variables can be found in Appendix B.
26 26 18 7 233 23 131 48 320 582 9 8 132 17 106 43 50 70 3453 Appendix D
D.1. Sample breakdown by Fitch’s credit rating and year
Rating 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 24 6 6 6 5 4 4 3 2 2 2 23 3 5 5 5 4 5 4 1 1 1 22 15 16 20 20 17 14 12 9 6 6 21 25 32 42 42 38 28 29 20 16 16 20 43 57 55 57 58 59 50 52 49 42 19 42 49 48 51 65 63 69 75 74 64 18 46 55 57 71 77 84 85 83 78 79 17 39 56 71 72 82 92 93 94 104 104 16 45 64 69 70 88 94 110 115 116 106 15 24 27 26 34 38 44 55 60 72 76 14 11 15 16 17 22 25 30 34 35 36 13 10 12 14 13 15 18 22 25 23 20 12 4 6 12 13 16 18 16 16 19 19 11 2 3 6 4 8 9 10 17 16 13 10 3 2 3 5 8 11 11 12 15 11 9 1 3 3 4 5 7 8 8 6 6 8 3 1 0 0 0 0 0 0 0 0 7 0 1 2 1 1 1 1 0 2 3 6 1 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 1 0 0 3 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 2 2 2 1 2 2 Total 323 410 455 484 549 578 610 625 636 606
This table presents the number of ratings by year. Fitch’s company credit ratings are converted to an ordinal scale according to the following schedule: 24 (AAA), 23 (AA+), 22 (AA), 21 (AA-), 20 (A+), 19 (A), 18 (A-), 17 (BBB+), 16 (BBB), 15 (BBB-), 14 (BB+), 13 (BB), 12 (BB-), 11 (B+), 10 (B), 9 (B-), 8 (CCC+), 7 (CCC), 6 (CCC-), 5 (CC+), 4 (CC), 3 (CC-), 2 (C), and 1 (D). The sample consists of 5,276 firm-year observations.
27 Appendix E
E.1. Comparison summary statistics U.S. and Europe
This table reports the mean, median, maximum (max.), minimum (min.), standard deviation (st. dev.), skewness (skew.) and kurtosis of the following variables: Credit rating, ESG score, economic-, environmental -, governance-, and social score. The results in panel A are based on 3,453 U.S. firm-year observations and panel B on 1,823 European firm-year observations. The data is obtained via DataStream.
Panel A: U.S.
Mean Median Max. Min. St. dev. Skew. Kurtosis Credit rating 16.689 17.000 24.000 1.000 3.015 -0.839 5.487 ESG score 61.317 65.590 98.300 2.960 28.459 -0.288 1.640 Economic score 58.565 61.510 98.890 1.410 27.474 -0.290 1.885 Environment score 50.838 50.290 97.470 8.270 32.386 0.039 1.349 Governance score 75.367 78.700 97.460 1.430 15.822 -1.330 5.477 Social score 53.793 56.210 98.930 3.590 28.907 -0.131 1.644 Panel B: Europe
28 Appendix F
F.1. Summary statistics of control variables
This table reports the mean, median, maximum (max.), minimum (min.), standard deviation (st. dev.), skewness (skew.) and kurtosis of the following variables: size, leverage, margin, loss, interest coverage, ROA, and financial utility. The results are based on the full sample of 5,273 firm-year observations. The data is obtained via DataStream.
Mean Median Max. Min. St. dev. Skew. Kurtosis
29
F.2. Correlation diagram variables
30 Appendix G
G.1. Ordered logit models of the effect of ESG performance on credit ratings
Sample Full sample U.S. EU
Model 4.1 Model 5.1 Model 6.1 ESG performance Economic score 1.008** (0.004) 1.004 (0.004) 1.016** (0.008) Environmental score 1.010** (0.005) 1.013** (0.006) 1.002 (0.011) Governance score 1.003 (0.006) 1.001 (0.009) 0.995 (0.007) Social score 1.005 (0.006) 1.006) (0.007) 1.003 (0.011) Control factors Size 2.092*** (0.159) 1.919*** (0.174) 3.173*** (0.416) Leverage 0.972*** (0.004) 0.967*** (0.005) 0.985* (0.008) Margin 0.999* (0.001) 0.999*** (0.000) 1.047*** (0.014) Loss 0.328*** (0.079) 0.355*** (0.102) 0.304*** (0.124) Interest coverage 1.001 (0.001) 1.000 (0.001) 1.004 (0.004) ROA 1.087*** (0.014) 1.108*** (0.015) 1.028* (0.016) Financial utility 4.676*** (0.959) 4.927*** (1.220) 3.221*** (1.213) Interaction terms and dummy variables
Pre-crisis period 4.283*** (1.971) 1.119 (0.927) 19.435*** (14.983) During-crisis period 3.346*** (1.149) 1.238 (0.817) 8.455*** (5.722) Economic * pre-crisis 1.005 (0.005) 1.009 (0.011) 0.984 (0.013) Economic * during-crisis 0.989*** (0.004) 0.997 (0.009) 0.980 (0.013) Environmental * pre-crisis 0.992 (0.006) 0.986 (0.011) 0.983 (0.018) Environmental * during-crisis 1.004 (0.005) 1.011 (0.011) 0.987 (0.018) Corporate governance * pre-crisis 0.989*
(0.006)
1.005 (0.014)
0.986 (0.012) Corporate governance *
31 Social * pre-crisis 1.005 (0.007) 0.999 (0.013) 1.000 (0.017) Social * during-crisis 1.001 (0.006) 1.009 (0.011) 0.994 (0.019)
Year fixed effects No No No
Country fixed effects Yes No No
N 5,276 3,453 1,823
Pseudo R-squared 0.309 0.279 0.387
Number of indicator values 4 4 4
32 Reference list
Ashbaugh-Skaife, H., Collins, D., La Fond R., 2006. The effects of corporate governance on firms’ credit ratings. Journal of Accounting and Economics 42, 203-243.
Attig, N., El Ghoul, S., Guedhami, O., Suh, J. 2013. Corporate social responsibility and credit ratings. Journal of Business Ethics 117, 679–694.
Baron D., Harjoto A., Jo H., 2011. The economics and politics of corporate social performance. Journal of Business and Politics 13, 1-46.
Beasley, M., 1996. An empirical analysis of the relation between the board of director composition and financial statement fraud. Accounting Review 71, 443–65.
Bhojraj, S., Sengupta P., 2003. Effect of corporate governance on bond ratings and yields: the role of institutional investors and outside directors. Journal of Business 3, 455-475. Blume, E., Lim, F., Mackinlay A., 1998. The declining credit quality of U.S. corporate debt:
myth or reality? The journal of Finance 53, 1389-1412.
Cai, L., Cui, J., Jo. H., 2016. Corporate environmental responsibility and firm risk. Journal of Business Ethics. 139 (3), 563–594.
Chatterji, A.K., Levine, D.I. and Toffel, M.W., 2009. How well do social ratings actually measure corporate social responsibility? Journal of Economics and Management Strategy, 18, 125– 169.
Chava, S., (2014) Environmental externalities and cost of capital. Management Science, 60 (9), 2223-2247.
Cheng, B., Ioannou I., Serafeim, G., 2014. Corporate social responsibility and access to finance. Strategic Management Journal 35, 1-23.
Chih, H-L., Chih, H-H and Chen, T., 2010. On the determinants of corporate social responsibility: international evidence on the financial industry. Journal of Business Ethics 93, 115-135.
Dechow, P., Sloan, R., 1991. Executive incentives and the horizon problem: an empirical investigation. Journal of Accounting and Economics 14 1, 51–89.
33
El Ghoul, S., Guedhami, O., Kwok, C., Mishra, D., 2011. Does corporate social responsibility affect the cost of capital? Journal of Banking & Finance 35, 2388–2406.
European Commission, 2017.
https://ec.europa.eu/growth/industry/corporate-social-responsibility_en
Godfrey, P., Merrill C., Hansen, J., 2009. The relationship between corporate social responsibility and shareholder value: an empirical test of the risk management hypothesis. Strategic Management Journal 30, 425-445.
Goss, A., Roberts, G., 2011. The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance 35, 1794–1810.
Greening, D., Turban, D., 2000. Corporate social performance as a competitive advantage in attracting a quality workforce 39, 254-280.
Haan de, E., 2017. The financial crisis and corporate credit ratings. The accounting review 92, 161-189.
Jiraporn, P., Jiraporn, N., Boeprasert, A., Chang, K., 2014. Does corporate social responsibility (CSR) improve credit ratings? Evidence from geographic identification. Financial Management 43, 505-531.
Margolis, J., Elfenbein, H., Walsh, J., 2009. Does it pay to be good… and does it matter – a meta- analysis of the relationship between corporate social and financial performance. Unpublished working paper, United States.
Mazzotta, R., Veltri, S., 2014. The relationship between corporate governance and the cost of equity capital. Evidence from the Italian stock exchange. Journal of Management & Governance 18, 419–448.
McGuire, J., Sundgren, A., Schneeweis, T., 1988. Corporate social responsibility and firm financial performance. Academy of Management Journal 31, 854-872.
Menz, K., 2010. Corporate social responsibility: is it rewarded by the corporate bond market? A critical note. Journal of Business Ethics 96, 117-134.
Murphy, K., Zimmerman, J., 1993. Financial performance surrounding CEO turnover. Journal of Accounting and Economics 16, 273–315.
34
Orlitzky, M., Schmidt, F., Rynes, S., 2016. Corporate social and financial performance: a meta-analysis. Journal of organization Studies 24, 403-441.
Renneboog, L., Ter Horst, J., Zhang, C., 2008. Socially responsible investments: Institutional aspects, performance and investor behavior. Journal of Banking & Finance 32, 1723-1742.
Ryana, P., Villupuramb S., Zygoc J., 2017. The value of credit rating changes across economic cycles. Journal of economics and business 92, 1-9.
Schultz, E., Tan, D., Walsh, K., 2017. Corporate governance and the probability of default. Journal of Accounting & Finance 57, 235-253.
Sharfman, M., Fernando, C., 2008. Environmental risk management and the cost of capital. Strategic Management Journal 29, 569-592.
Spicer, B., 1978. Investors, corporate social performance and information disclosure: an empirical study. The Accounting Review 53, 94-111.
Stellner, C., Klein, C., Zwergel, B., 2015. Corporate social responsibility and Eurozone corporate bonds: the moderating role of country sustainability. Journal of Banking & Finance 59, 538–549.
Tony Chocolonely, 2017. Tony Chocolonely annual report 2016/2017. Accessed on December 5, 2017. Via https://tonyschocolonely.com/storage/configurations/tonyschocolonely com.app/files/jaarfairslag/2017-2017/tc_jaarfairslag_2016_en_totaal_01.pdf.
