Supervisor prof. Dr, L.J.R. Scholtens

(1)

A test of the moderating effect of firm risk and environmental risk on the

relationship between corporate social responsibility and corporate financial

performance.

Jasper Roodenburg s2335832

University of Groningen Faculty of Economics and Business MSc International Financial Management

Supervisor prof. Dr, L.J.R. Scholtens

(2)

1

Consumers, organizations and firms are becoming more and more aware of corporate social responsibility (CSR), and the subject is growing both in the world of business and in academia. With the increasing attention that is being diverted to CSR, the body of research that works on finding whether non-financial performance can be converted into financial performance is growing. While the results of two recent meta-analyses (Margolis et al., 2007; Orlitzky et al., 2003) suggest a modest positive relation between CSR and corporate financial performance (CFP), the findings are still heterogeneous. Opponents of CSR argue that the primary responsibility of a firm is to the shareholders. A firm should not use scarce company resources to engage in CSR activities, but should focus these resources on projects that generate shareholder profit instead. Proponents of CSR argue that firms should not satisfy the needs of only shareholders, but also that of stakeholders. Stakeholder management is critical in the survival of a firm, as these stakeholders can affect the firm’s existence (Freeman, 1984). Owing to these inconsistent results, literature has stressed on the importance of contribution to the field in the form of testing mediating and moderating mechanisms in the relationship between CSR and CFP (Margolis and Walsh, 2003).

(4)

3 El Ghoul et al. (2016) argued that market-supporting institutions support the transaction environment where they ensure the effective functioning of the market. The institutions ensure that firms do not incur undue costs or risks by supporting the transaction process. Following this line of thinking, where CSR can be an effective strategy to manage high risk (in this context, by substituting institutions), it can be argued that CSR could also be an effective strategy in managing firm risk. Miller et al. (2009) state that CSR is an effective way to create stable relations with stakeholders. It can create a social capital that can reduce risk, even helping the firm to survive if times get tough (Godfrey, 2005).

This study tries to offer insights into mechanisms through which CSR has an impact on corporate financial performance. I examine the impact of CSR on CFP through the moderating impact of environmental and firm risk. If the mechanism would follow the same line as that of El Ghoul et al. (2016), high risk would strengthen the positive relation between CSR and CFP. This study differs from that of El Ghoul et al. (2016) by taking excess stock return as the CFP measure and extending the research from market-supporting institutions to firm risk. To the best of the author’s knowledge, this study is the first study to examine the moderating effect of risk on the relationship between corporate social responsibility and excess return.

I have measured CSR based on the ESG scores from the ASSET4 Bloomberg database, which is a measure of performance based on environmental, social and governance pillars. Moreover, I utilized different measures of risk: for environmental risk, the level of institutional development was considered based on the governance indicators of the World Bank; and firm risk was measured with the stock’s beta, which measures a stocks sensitivity to market movements, and the Altman’s z-score, which measures the risk of a firm going bankrupt. I measured CFP as a firm’s excess stock return.

(5)

4 environmental risk and firm risk moderates the relationship between CSR and excess stock return.

This study provides some insights into the CSR debate. The argument that CSR activities create value through stakeholder engagement takes center stage in the debate concerning CSR. The results from this paper suggest that CSR engagement results in wealth being transferred from shareholders to other stakeholders, where, from the investor perspective, CSR engagement decreases the returns. The results provide some implications for managers and policy makers. In the international context, CSR does not seem to create higher investor return. No statistically significant results are found that point to the fact that the relationship between CSR and CFP is moderated by risk. Managerial decisions on CSR activities can be directed to other stakeholders under certain circumstances, but are not directly beneficial for investors under any risk profile. Herewith, the study contributes to the line of literature that examines the mechanisms through which CSR affects financial performance in the international context.

This Master’s thesis is written for the Master's degree program in International Financial Management (IFM). Therefore the focus needs to lie on different aspects of financial management in an international business environment. In this paper one of the latest subjects in finance is researched, namely corporate social responsibility. The results of this paper equip managers with understanding of corporate financial responsible management and the financial implications thereof, especially in relation to risk management. The research incorporates the international context by utilizing an international sample with several differences at the country level, such as risk at the country level.

(6)

5 Central research question

‘Is the relationship between CSR and CFP stronger or weaker when firms face high risk?’

Sub-questions

‘Is the relationship between CSR and CFP stronger or weaker when firms face high firm risk?’

‘Is the relationship between CSR and CFP stronger or weaker when firms face high environmental risk?’

2. Literature review and development of hypotheses.

CSR has been the subject of a relatively young academic and corporate debate, first mentioned by Bowen in 1953. Since then, scholars have argued for both the costs and the benefits of CSR initiatives and investments. This empirical quest to find a relationship between CSR and CFP is still ongoing, with the field of business and academia trying to reconcile corporate social responsibility with economic theory, when the classical economic theory dictates that management should set out to maximize shareholder value. Now, even though recent meta-studies have indicated that a positive relationship exists between CSR and CFP (Van Beurden and Gössling, 2008; Orlitzky et al., 2003), the findings are still far from conclusive. Margolis and Walsh (2003) assert that although the meta-analyses found positive relationships, the debate still opens as many questions as it answers. They, therefore, encourage researchers to focus on testing mediating mechanisms and moderating conditions to link CSR and CFP.

(7)

6 CFP. Furthermore, research by Sassen et al. (2016) argues that firm risk is a similar mediating factor in the impact of CSR on CFP. Depending on one’s theoretical perspective, CSR can be seen as an efficient tool in overcoming inefficiencies that translate into risk, where inefficiencies can be defined as factors introduced by a firm’s environment, as in the case of El Ghoul et al. (2016), or uncertainties at the firm level, such as stock volatility, as in the case of Sassen et al. (2016). It can also be argued to be an inefficient use of scarce resources, which is punished with lower CFP. Even when CSR initiatives do not relate to negative financial performance, it can be argued that the firm’s resources can be better diverted to an activity, other than CSR, creating even higher financial performance (Margolis and Walsh, 2003)

2.1. Environmental risk.

As for the impact of risk on the relationship between CSR and CFP, in regard to environmental risk, several studies have found that CSR initiatives can mitigate risk. El Ghoul et al. (2016) use transaction cost theory to argue that CSR initiatives can mitigate risk when market inefficiencies translate into environmental risk. Transaction cost theory states that all firms incur costs of participating in the market and engaging in exchange processes (Williamson, 1985). Minimizing these costs will ensure further profitability. Market inefficiencies introduce high transaction costs. Specialized intermediaries have emerged in order to regulate exchange processes, which decrease transaction costs. When these market inefficiencies translate into increased environmental risk, and specialized intermediaries are lacking, CSR initiatives can serve as substitutes for market-supporting institutions, mitigating the thereby formed risks and decreasing transaction costs.

(8)

7 for stock market efficiency, credit market efficiency and business freedom, this effect is significant at the 10% level. For legal systems & property rights, they found this effect to be significant at the 5% level. In addition, they observed that the economic significance is not trivial, where, if the stock market efficiency is set at the first quartile, a one-standard-deviation increase in CSR is associated with an increase in Tobin’s Q of 0.07; if this is set at the third quartile, this increase is 0.01—a difference of 0.06—an increase of 4.0% relative to the mean. This difference is 0.06 for credit market efficiency, 0.04 for business freedom, and 0.09 for legal systems & property rights. El Ghoul et al. (2016) found that through certain channels, CSR initiatives can reduce transaction costs and substitute market supporting institutions. CSR initiatives can decrease environmental risks when these risks are explained as being institutional voids and market inefficiencies, such as weaker equity and credit markets and so on. El Ghoul et al. (2016) employ a linear regression model with firm-fixed effects and year-fixed effects, where Tobin’s Q is the dependent variable, and the average of the environmental and social performance score from the ASSET4 database is the independent variable. They normalized the score such that it ranges between 0 and 1. The Tobin’s Q definition used by El Ghoul et al. (2016) deviates from the original definition of Tobin’s Q that was designed to indicate whether a firm was undervalued or overvalued by providing the ratio of the market value to the replacement cost. In this paper, deferred taxes are not subtracted from the total market value of the firm. El Ghoul et al. (2016) used Tobin’s Q as a proxy for firm value. An appropriate market measure of corporate financial performance would be excess stock return, which is widely used in literature (Orlitzky et al., 2003).

(9)

8 (2016) found that CSR initiatives create more value for firms in countries with weaker capital markets. But the same mechanism applies when firm-specific risk introduces capital constraints. Through reduced agency costs due to enhanced stakeholder engagement, and through reduced informational asymmetry due to increased transparency, borrowing constraints, which stem from high risk, can be mitigated (Cheng et al., 2014). This suggests that risk, when it increases the inability to obtain finance, can moderate the relationship between CSR and CFP. CSR initiatives signal better engagement with the stakeholders, and reduce the likelihood of opportunistic behavior. In other words, CSR initiatives are used to signal cooperation and reduce potential agency conflict. Cheng et al. (2014) found that engagement in CSR improves access to finance, which is significant at the 1% level, with a corresponding coefficient of -1.034. These results can be interpreted as that firms with better CSR performance face lower capital constraints. Furthermore, they found that the results are unlikely to be driven by reverse causality, where less financially-constrained firms can afford to invest in CSR. They found that the relation is in fact the strongest for constrained firms (Cheng et al., 2014). Cheng et al. (2014) focused on the essential impact that CSR has on idiosyncratic capital constraints of firms. They conducted a research using a sample of 49 countries, in which they used a panel dataset, with the average ESG performance score from the ASSET4 database as the independent variable. They employed a fixed effects regression model.

(10)

9 not are the cheaper option. In conclusion, this indicates that CSR is not an effective strategy to manage risk brought on by market inefficiencies, but CSR initiatives are expected to decrease in such low-sustainable nations, which are often developing nations.

2.2. Firm Risk.

In regard to the impact of firm risk on the relationship between CSR and CFP, by combining transaction cost theory and the resource-based view of a firm, it can be argued that constrained firms can use CSR initiatives to reduce risk and increase financial performance. A meta-analysis on the relationship between CSR and financial risk by Orlitzky et al. (2001), covering 18 US-based studies for the period of 1987-1995, found evidence suggesting that CSR initiatives decrease financial risk. The study also found that the relation between firm risk and CFP is stronger for market risk measures than for accounting risk measures. These measures differ not due to the fact that they measure different conceptual components of risk, but because they measure the same constructs in different ways (Sassen et al., 2016). Examples of accounting risk measures are the standard deviation of a firm’s long-term return on assets (ROA) or return on equity (ROE), or Altman’s Z-score, which measures the risk of a firm going bankrupt. The meta-analysis by Orlitzky et al. (2001) further indicates that most studies employ aggregated CSP measures; these studies rarely use separate measures to investigate the individual effect of either social, environmental or governance scores. Concerning these differences, results are significantly more inconsistent at the disaggregate level. At the aggregate level, the relation between CSR and firm risk is often found to be negative, where engagement in CSR lowers firm risk.

(11)

10

CSR engagement can be theorized to be an efficient strategy to reduce risk through several different channels. First, Godfrey (2005) suggests that, in times of financial crisis or struggle, a firm can use CSR initiatives to gain moral capital amongst stakeholders. By employing an event study, they empirically demonstrate that CSR initiatives can ensure protection for a firm. As the firm is positively assessed, thanks to their CSR initiatives, it can gain loyalty from stakeholders in times of crisis. This is known as the risk management theory (Godfrey 2005). This theory would suggest a moderating effect of firm risk in the positive relationship between CSR and CFP, where, similar to the findings of El Ghoul et al. (2016), high firm risk could strengthen the positive relation between CSR and CFP, suggesting that CSR engagement can be a strong strategy in times of firm crisis. The loyalty generated by CSR initiatives can act as a kind of insurance, increasing CFP and lowering financial risk for future periods. This is supported by the fact that social responsible investors react less sensitively to negative returns than conventional investors (Sassen et al., 2016; Renneboog et al., 2011), implying that investors that highly value CSR are more restrained to withdraw their investments in the case of disappointing financial performance.

(12)

11 Third, another view argues that CSR engagements can mitigate financial constraints, as firm risk can negatively affect the access to finance and increase the cost of debt and equity (Cheng et al., 2014; Goss & Roberts, 2010; El Ghoul et al., 2011; Stellner et al., 2015; Attig et al., 2013; Cheng et al., 2014). El Ghoul et al. (2011) found that CSR is negatively related to the equity cost of a company. Also, Goss and Roberts (2010) found that firms which perform below average on CSR pay up to 20 basis points more on debt. In addition, firms that are more responsible showed to have a significantly lower cost of financing. Stellner et al. (2015) set corporate bond ratings as the dependent variable and found that companies benefit from better ratings if they show superior (above country average) CSR efforts. In turn, this supports the risk mitigation view, suggesting that firms can reduce their risk profile by investing in CSR. Higher risk firms could significantly benefit from CSR initiatives when low credit ratings complicate their access to financing. Market participants are more forthcoming with capital to firms involved in more CSR initiatives, as a result, CSR initiatives can decrease capital constraints (Cheng et al., 2014). This suggests that firms with high capital constraints, possibly partly due to high financial risk, can lower their capital constraints through CSR initiatives.

Sassen et al. (2016) researched how CSR affects several risk measures. They focused on the effect of CSR on firm risk as a determinant of the cost of capital. They found results that suggest that decreasing firm risk mediates the effect that CSR has on decreasing the cost of capital and CFP. Sassen et al. (2016) found that the total risk is significantly and negatively correlated with sustainability scores, suggesting that firms that perform better on CSR have lower firm risk. With regard to economic relevance, they found that increase in social performance by one standard deviation reduces total risk by 4.2%, relative to the mean; in case of systematic risk, this reduction is 3.6%. As CSR is an effective risk strategy to reduce risk, this could indicate that the same case could be made, as found by El Ghoul et al. (2016), concerning firm value. Thus, high firm risk could strengthen the positive relation between CSR and CFP.

(13)

12 effect model, with both firm and year fixed effects, for the independent variable, they took both the aggregate measure of the environmental, social and governance performance from the ASSET4 database, as well as the separate measures with which they conducted a sensitivity test.

The instrumental stakeholder theory and the risk management theory suggests that risk reduction is possible through CSR engagement, where CSR activities could be especially critical for those firms that face crippling risks. These firms have most to gain from risk management, as they are mostly affected by the high firm risk they face, compared to firms with low financial risk, that are relatively less concerned with risk-reduction. On the other hand, the opposite could be the case, and firms with high financial risk could be punished when engaging in CSR. Goss and Roberts (2011) found moderate support suggesting this. When loans are extended to high risk borrowers, agency risks are high and borrowers are punished if they invest in CSR. Stellner et al. (2015) also found statistically significant support for the fact that higher risk firms are likely to be punished when they invest in CSR initiatives as they use up resources that are already relatively scarce. In this case, investors interpret CSR engagement in such high risk firms as risk increasing. This would imply that firms with higher risk, because of liquidity problems, or higher bankruptcy risk, would be punished when they invest in CSR. As these firms are likely to have even scarcer resources, these resources are best invested in business projects instead of CSR investments. Moreover, although firms can profit from CSR engagement, they should be primarily concerned with their main business affairs; wasting scarce resources will be even more damaging for these firms, thereby increasing the risk even further and negatively influencing performance (Stellner et al., 2015). This view, suggesting that firms with higher risk will be punished if they invest in CSR, can be argued to be more relevant for the case of financial firm risk than for environmental risk, when environmental risk implies underdeveloped market-supporting institutions, as researched by El Ghoul et al. (2016). It is likely that scarce firm resources can be effectively used to decrease a firm’s financial risk through more efficient ways than CSR, whereas firm resources might not be able to overcome market inefficiencies.

(14)

13 increasing risk in the long run. They also tend to overinvest in CSR when CFP is low, so as to hide behind CSR initiatives (Bouslah et al., 2013). This strategic use of CSR initiatives to prevent their replacement during periods of low financial performance, or in takeover reorganizations, is known as the entrenchment strategy.

In short, a wide mix of different risk measures, and often outdated data employed in studies, still makes for inconclusive findings concerning the role firms and environmental risk plays in the CSR and CFP relation. To the best of the author’s knowledge, a research that investigates the interaction effect of risk on both the firm and the country level in the relation between CSR, measured with an aggregated ESG score, and CFP, measured as excess stock return, has not been carried out before for an international oriented sample. I set the following hypotheses concerning the impact of risk on the relationship between CSR and CFP.

H1: Overall CSR measured with an aggregated ESG score (ASSET4) affects a firm’s corporate financial performance.

H2: The relationship between CSR and CFP is stronger/ weaker for firms with high firm risk, than it is for firms with low firm risk.

H3: The relationship between CSR and CFP is stronger/ weaker for firms with high environmental risk, than it is for firms with low environmental risk.

(15)

14 sustainability of the country in which the firm operates, I can collect possibly informative data on the moderating role of a country’s sustainability in the investigated relations.

3. Research Design.

This part will elaborate on the variable measurements, data collection, sample, and the models used to answer this paper’s research questions. I test the following general model:

CFP = β0+β1CSR+β2ControlVariables+β3RiskInteractionEffect+year and firm fixed effects

3.1. Variable measurements.

3.1.1. Dependent variable: CFP.

(16)

15 The yearly excess return for company ‘i’ is calculated as follows:

𝐸𝑋𝑅𝐸𝑇𝑈𝑅𝑁 = (𝑅𝐼𝑖𝑡/𝑅𝑖𝑡−1) − (𝑅𝐼𝑚𝑡/𝑅𝐼𝑚𝑡−1)

Where year is 't' and ‘m’ indicates the market index.

Stock returns are a frequently used measure of corporate financial performance in studies concerning CSR (Filbeck and Gorman, 2004; Arx and Ziegler, 2008). Limitations to this measure indicate the fact that it assumes that the investor’s valuation of the financial performance of the firm is an appropriate measure (Tsoutsoura, 2004). McWilliams et al. (2006) criticize the use of excess returns. They argue that stock prices only relate to financial stakeholders and does not incorporate any information from non-financial stakeholders, who are also affected by CSR initiatives. Furthermore, a small positive correlation is found between the risk measures and corporate financial performance measures, confirming the notion that more risk should be compensated with higher returns.

3.1.2. Independent variables: ESG (CSR) measure.

(17)

16 As a sensitivity check, the disaggregated ESG measures will be used, separating the measure of environmental, social, and governance performance, and using these consecutively. The aggregated score is a collective measure of three different ESG dimensions. Therefore, it does not grant the possibility for specific inferences concerning potential distinct impacts of the separate measures of environmental, social, and governance performance. For instance, companies which would have the same aggregate ESG score might have different relations because it is comprised of a different bundling of the three scores (Bouslah et al. 2013). Using the disaggregated ESG measures will shed further light on these different pillars and serve as a sensitivity check.

3.1.3. Interaction variables: risk measures.

To measure environmental risk, this paper uses the indicators of institutional differences and development from the World Bank. These are assessed by taking the mean of the six Worldwide Governance Indicators (to create the variable CORISK). The measurement comprises of the indicator’s voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law, and corruption control. The measure ranges from 0 to 100, and has been used in several other papers to indicate institutional differences (El Ghoul et al., 2016; Rodrigo et al., 2016). Several different sources are used to construct this measure in the Worldwide Governance Indicators, amongst which are several economic risk indicators. This measure is chosen as a proxy for risk because of data availability issues. Other measures, such as the risk ratings by credit rating agencies such as S&P and Moody’s, or the Euromoney environmental risk ratings, were not available for the sample that this study employs.

(18)

17 financial health of the company. A rule of thumb is that a firm is in significant danger of going bankrupt with a Z-score of below 1.81, it is in a dangerous gray area with a score between 1.81 and 2.99 and in good financial health if the Z-score is 3.0 or higher (See; Altman, 1968, p.606). The formula used is as follows;

𝑍 = 1.2𝑋1 + 1.4𝑋2 + 3.3𝑋3 + 0.6𝑋4 + 1.0𝑋5

Where;

X1 = working capital / total assets. This is a measurement of liquid assets in relation to the size of the firm.

X2 = retained earnings / total assets. Measures profitability that reflects the age and earning power of the firm.

X3 = earnings before interest and tax (EBIT) / total assets. Measures operating efficiency, excluding taxes and leveraging factors.

X4 = market value of equity / book value of total liabilities. Adds market dimension, which can indicate security price fluctuation as possible red flags.

X5 = sales / total assets. Standard measure for total asset turnover.

Although the Altman z-score was initially designed for manufacturing companies, it offers a generalizable proxy for firm risk for other industries as well. Other papers use corporate bond ratings and log-spreads as measurements for firm risk (Goss & Roberts, 2010; Stellner et al., 2015, Attig et al., 2013). These were unfortunately not available for the sample used in this study.

(19)

18 shareholders holding diversified portfolios, and is frequently employed as a risk measure (Orlitzky & Benjamin, 2001). The firm’s beta is downloaded from Datastream, and is based on between 23 and 35 consecutive end of the month percentage changes.

3.1.4. Control Variables.

In order to isolate the effects that this paper sets out to measure, several other explanatory variables must be used. The control variables on firm level are size (SIZE), leverage (LEV), research and development expenditures (R&D), sales growth (SG) and capital expenditure (CAPEX). On the country level, I control for GDP per capita (GDP) and country sustainability (SUS). These control variables are commonly used in literature (El Ghoul et al., 2016; Goss & Roberts, 2010). All financial data has been collected from the Datastream database.

(20)

19 I control for economic development (GDP), as economic development has a significant impact on firm performance, and is measured as the natural logarithm of GDP per capita in US dollars. Country level factors often have a stronger influence than firm level factors. Stellner et al. (2015) found that the value of CSR depends strongly on the sustainability level of the country in which the firm operates. To control for country sustainability (SUS), I divide my sample into two subsamples: one sample for the countries that score below average on the country sustainability ranking (henceforth referred to as ‘sample below’), and one sample for countries that score above average on the country sustainability ranking (henceforth referred to as ‘sample above’). In contrast to Stellner et al. (2015), I use the RobecoSAM scores and not the Bloomberg country ESG scores, which were not available to me. I split the sample based on the measurement of May 2016, as country sustainability is not expected to change significantly in this period sample. Phillis et al. (2011) found that economic development is an important factor influencing country sustainability. They further found that economic development is not expected to change dramatically over time; therefore, I feel safe to make this assumption.

Table 1 Definition of Variables.

All financial data is collected from DataStream and all financial values are in US dollars.

Variables Explanation

CSR CSR score obtained from Asset4, measured as the equally weighted

average of pillar scores CGS, SOS, and ENV.

EXRETURN Individual stock return minus return off accompanying market index

ZSCORE Altman’s z-score, measure of predicting bankruptcy risk.

BETA Stock beta, is calculated based on the standard CAPM model.

CORISK Average of the 5 governance pillars from the World Bank

SIZE Natural logarithm of total assets

LEV Total debt/total assets

R&D Intangible assets/total assets

(21)

20

SG Annual percentage change in total sales volume

GDP Natural logarithm of GDP per capita

SUS RobecoSAM country sustainability scores of May 2016

3.2. Data and sample.

The data sample consists of firms that are registered on the Thompson Reuters Asset4 database, as the ESG scores from the Asset4 database are used to measure CSR. The firm list is the same list as used by Jousma (2017). The sample period used for this research is 2006-2014. The Asset4 database is relatively young, and including years before 2006 would decrease the sample significantly. For the most recent years, not all variables have yet been included in their respective databases, after which I conclude that 2006-2014 ensures the most complete sample period, while still covering a long enough period for the purpose of research. The ESG scores are offered on yearly bases, and therefore, an annual interval will be used.

The research will be conducted using a sample that includes all countries for which sufficient information is available. I chose the world as the locus of this study, and did not limit the sample for two reasons. First, to the best of my knowledge, such a comprehensive research on both firm and environmental risk has never been conducted for the worldwide context. Second, this ensures a broad data range concerning environmental risk, due to which, countries that are not leaders in CSR are also covered. Firms in the financial industries have been excluded from the sample. The Altman z-score is not recommended to be used for financial companies because of the opaqueness of their balance sheets. Furthermore, due to their unique regulatory qualities, they are frequently excluded in literature (Goss & Roberts, 2010; Stellner et al., 2015). The sample in this study consists of 17258 firm-year observations.

(22)

21 run for the total sample, and sample 2, representing the above average sustainable countries, thus stems from these three countries. With only 5 observations, the lowest number of observations is recorded from firms from China, and firms from Greece with 14 observation, and firms from Portugal with 69 observations.

Table 2 Observations per country.

This table presents the observations per country, with the count of observations and the percentage of the whole sample.

Country Count Country Count

Australia 1189 Mexico 110

Austria 93 Netherlands 188

Belgium 129 New Zealand 77

Canada 1274 Norway 129

China 5 Philippines 79

Denmark 145 Poland 81

Finland 197 Portugal 69

France 592 South Africa 306

Germany 531 Spain 251

Greece 14 Sweden 286

Hungary 18 Switzerland 351

Indonesia 110 Thailand 83

Ireland 70 Turkey 97

Italy 215 United Kingdom 1766

Japan 2812 United States 5175

Korea, Rep. 438

Malaysia 190 Total 17258

Below sample: China, Greece, Indonesia, Korea, Rep., Malaysia, Mexico, Philippines, South Africa, Thailand, Turkey.

Above sample: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Hungary, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Poland, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States.

(23)

22 values with the percentile value. At 0.01 level, minimal identical values are replaced, while still ensuring robust results, because output corrupting outliers are deleted from this sample.

The CFP variable (excess return) shows a mean of 0.032, indicating an overall positive excess return for the sample of firms in the ASSET4 database. The standard deviation is 0.320, indicating a large variation across firms regarding excess return. El Ghoul et al. (2016) used a comparable sample set with 2,445 unique firms from 53 countries, and presented comparable descriptive statistics for the control variables, LEV, SG and GDP. The sample used in this paper shows a higher mean for R&D than that of El Ghoul et al. (2016), 0.183 versus 0.030, possibly due to the different measurement method we have used to control for R&D (intangible assets/total assets). The mean for SIZE is also found to be higher for the sample used in this paper, 15.126 versus 8.630, but the mean for SIZE is similar to the one found in the European panel dataset used by Sassen et al. (2016). The mean for the BETA is also found to be comparable. The overall CSR score is found to be slightly higher in the dataset used by Sassen et al. (2016) (61,131 over 54,788).

Table 3: Descriptive Statistics.

This table presents the sample descriptive statistics for all variables.

Mean Median Maximum Minimum Std. Dev. Skewness Kurtosis Observations

BETA 0.876 0.800 4.000 -0.520 0.863 1.148 4.647 17258 CAPEX 5.290 2.960 49.621 0.000 7.340 2.933 14.325 17258 CGV 55.344 64.580 95.270 1.780 30.067 -0.469 1.802 17258 CORISK 85.101 85.737 98.030 35.036 9.957 -2.416 10.339 17258 CSR 54.788 55.122 93.337 6.530 22.348 -0.168 2.166 17258 ENV 54.892 59.795 94.860 9.450 31.926 -0.152 1.390 17258 GDP 10.582 10.611 11.073 8.542 0.322 -2.533 12.040 17258 LEV 0.196 0.178 0.760 0.000 0.156 0.847 3.647 17258 R&D 0.183 0.108 0.776 0.000 0.198 1.136 3.394 17258 ROA 5.597 4.620 38.881 -45.149 9.352 -0.266 9.317 17258 EXRETURN 0.032 0.044 1.068 -1.121 0.320 -0.382 5.128 17258 SG 9.940 6.086 327.848 -56.021 33.143 4.831 42.556 17258 SIZE 15.456 15.417 18.817 9.823 1.433 -0.084 3.218 17258 SOS 54.126 56.615 97.840 2.463 29.424 -0.166 1.718 17258 ZSCORE 2.415 2.226 7.942 -2.743 1.542 0.700 4.994 17258

(24)

23 Table 4 shows the correlation matrix. The critical value for correlation is 0.7 according to Brooks (2014). No variables with a correlation equal to, or higher than, this critical value are simultaneously used in one regression. Higher correlation between CSR and CFP (excess stock return) is to be expected if a significant relation between the two is hypothesized. Orlitzky et al. (2003) also found low correlation for market-based measures, and found that CSP is more highly correlated with accounting-based measures of CFP than with market-based measures of CFP.

(25)

(26)

25

3.3. Research Methods.

The regression model is most frequently used to investigate the relation between CSR and CFP (Margolish & Walsh, 2001; Orlitzky et al., 2003). In line with most research, ordinary least square (OLS) is used to estimate the effect of CSR on CFP. I employed a regression model and performed several tests, which indicate, in line with the findings of El Ghoul et al. (2016) and Sassen et al. (2016), that fixed effects are appropriate. In the extended models, the interaction variables have been included.

𝐶𝐹𝑃 = 𝛽0+ 𝛽1𝐶𝑆𝑅𝑖𝑡−1+ 𝛽2𝐿𝐸𝑉𝑖𝑡−1+ 𝛽3𝑅&𝐷𝑖𝑡−1+ 𝛽4𝐶𝐴𝑃𝐸𝑋𝑖𝑡−1 + 𝛽5𝑆𝐼𝑍𝐸𝑖𝑡−1+ 𝛽₆𝐺𝐷𝑃_𝑖𝑡−1+ 𝛽₇𝑆𝐺_𝑖𝑡−1+ ∑𝑦𝑒𝑎𝑟𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + ∑𝑓𝑖𝑟𝑚𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀_𝑡𝑗 (1) 𝐶𝐹𝑃 = 𝛽0+ 𝛽1𝐶𝑆𝑅𝑖𝑡−1+ (𝛽2𝑟𝑖𝑠𝑘𝑓𝑖𝑟𝑚𝑖𝑡−1+ 𝛽3𝐶𝑆𝑅𝑖𝑡−1∗ 𝑟𝑖𝑠𝑘𝑓𝑖𝑟𝑚𝑖𝑡−1) + 𝛽4𝐿𝐸𝑉𝑖𝑡−1+ 𝛽₅𝑅&𝐷_𝑖𝑡−1+ 𝛽₆𝐶𝐴𝑃𝐸𝑋_𝑖𝑡−1+ 𝛽₇𝑆𝐼𝑍𝐸_𝑖𝑡−1+ 𝛽₈𝐺𝐷𝑃_𝑖𝑡−1+ 𝛽₉𝑆𝐺_𝑖𝑡−1+ ∑𝑦𝑒𝑎𝑟𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + ∑𝑓𝑖𝑟𝑚𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀_𝑡𝑗 (2) 𝐶𝐹𝑃 = 𝛽0+ 𝛽1𝐶𝑆𝑅𝑖𝑡−1+ (𝛽2𝑟𝑖𝑠𝑘𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖𝑡−1+ 𝛽3𝐶𝑆𝑅𝑖𝑡−1∗ 𝑟𝑖𝑠𝑘𝑐𝑜𝑢𝑛𝑡𝑟𝑦𝑖𝑡−1) + 𝛽4𝐿𝐸𝑉𝑖𝑡−1+ 𝛽5𝑅&𝐷𝑖𝑡−1+ 𝛽6𝐶𝐴𝑃𝐸𝑋𝑖𝑡−1 + 𝛽7𝑆𝐼𝑍𝐸𝑖𝑡−1+ 𝛽8𝐺𝐷𝑃𝑖𝑡−1+ 𝛽9𝑆𝐺𝑖𝑡−1+ ∑𝑦𝑒𝑎𝑟𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + ∑𝑓𝑖𝑟𝑚𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀_𝑡𝑗 (3) 𝐶𝐹𝑃 = 𝛽₀+ 𝛽₁𝐶𝑆𝑅_𝑖𝑡−1+ (𝛽₂𝑟𝑖𝑠𝑘𝑐𝑜𝑢𝑛𝑡𝑟𝑦_𝑖𝑡−1+ 𝛽₃𝐶𝑆𝑅_𝑖𝑡−1∗ 𝑟𝑖𝑠𝑘𝑐𝑜𝑢𝑛𝑡𝑟𝑦_𝑖𝑡−1) + (𝛽3𝑟𝑖𝑠𝑘𝑓𝑖𝑟𝑚𝑖𝑡−1+ 𝛽4𝐶𝑆𝑅𝑖𝑡−1∗ 𝑟𝑖𝑠𝑘𝑓𝑖𝑟𝑚𝑖𝑡−1) + +𝛽5𝐿𝐸𝑉𝑖𝑡−1+ 𝛽6𝑅&𝐷𝑖𝑡−1+ 𝛽7𝐶𝐴𝑃𝐸𝑋𝑖𝑡−1 + 𝛽8𝑆𝐼𝑍𝐸𝑖𝑡−1+ 𝛽9𝐺𝐷𝑃𝑖𝑡−1+ 𝛽10𝑆𝐺𝑖𝑡−1+ ∑𝑦𝑒𝑎𝑟𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + ∑𝑓𝑖𝑟𝑚𝑓𝑖𝑥𝑒𝑑𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀_𝑡𝑗 (4)

3.3.1. Methodology and Models.

(27)

26 examine whether pooled ordinary least squares (OLS) is the appropriate model (Appendix A1). The test statistics suggest unobserved heterogeneity and indicate that pooled OLS would not be the appropriate model. For unbiased estimates, entity fixed effects and time fixed effects are used in accordance with the results from redundancy test and Hausman test. This involves estimating a parameter for each firm. The use of a fixed effects model controls for all time-invariant characteristics and tackles any endogeneity problems that source from an omitted variable bias. To control for changing economic conditions, time fixed effects are used at the year level. White standard errors are used in all models to overcome potential heteroscedasticity problems.

Endogeneity could also result from simultaneous causality. In this research paper, it has been assumed that CSR affects CFP, although reverse causality might be another source of endogeneity. It is proposed that prior corporate financial performance could provide resources for subsequent investments in corporate social responsibility (Orlitzky et al., 2003). Scholtens (2008) found that financial returns mostly drive CSR initiatives. There are only a few instances where CSR dimensions Granger-causes financial returns. To check for reverse causality, a Granger causality test is performed, between the CSR measure and CFP measure, and the main model, with CSR as the dependent variable, is estimated (Appendix A3). The Granger causality test indicates Granger causality and there is reason to believe that there exists a bidirectional relationship in which CSR impacts CFP, and CFP impacts CSR. In line with the research by El Ghoul et al. (2016), the right-hand-side variables are lagged by one period to mitigate bias stemming from reverse causality and simultaneity bias. The use of lagged explanatory variables does not completely eliminate simultaneity bias (Sassen et al., 2016), but it does diminish any problems stemming from simultaneity as past CSR and current CFP are not determined in the same period.

(28)

27

4. Results.

This section consists of two parts: In the first part, the results from the OLS regression, which tests the existence of a relationship between CSR (aggregate ESG score) and CFP (excess stock return), are presented. Then the results for the main research objective are presented, which comprises of the results of the interaction effect of CSR*firm risk (ZSCORE and BETA) and CSR* environmental risk (CORISK).

4.1. The relation between corporate social responsibility (CSR) and corporate financial performance (CFP).

As described in the methodology section, I performed ordinary least square regressions with firm and year fixed effects to examine the impact of CSR on CFP. Hypothesis 1 stated that ‘overall CSR measured with an aggregated ESG score (ASSET4) affects a firm’s corporate financial performance’. The therewith associated null hypothesis states that there is no relation between the aggregated ESG score and a firm’s corporate financial performance. Table 4 reports the results of testing Hypothesis 1. I found that the coefficient (β=-0.001) of CSR is negative and based on the p-value, which is found to be statistically significant at the 1% level for the full sample and the above sample. On the basis of this result, I rejected the null hypothesis of no relation between CSR and CFP. When the coefficient is negative, it seems that firms with increasing CSR initiatives find themselves punished with decreasing excess share returns. A 1 standard deviation increase of the CSR variable is ceteris paribus followed by a -0.045 standard deviation increase of the excess return variable.1 This would support the view on CSR stemming from the shareholder theory, suggesting that investing in CSR initiatives destroys valuable recourses, which should be used for value-creating business projects.

These findings are in contrast to the findings of Orlitzky et al. (2003) and Beurden & Gössling (2008), who found that the relation between CSR and CFP is positive. These contrasting findings will only be discussed briefly, as the emphasis of this research is on the interaction effect of risk. Brammer et al. (2006) found the impact of corporate social responsibility on stock returns to be negative. They found that firms with higher social performance scores achieve lower returns.

1_{Calculated as follows: the regression coefficient (𝛽) divided by (}𝑠𝑦

𝑠_𝑥), where 𝑠𝑦 is the standard deviation of the

(29)

(30)

29

Table 4 Model 1. Model(1) tests the overall effect of CSR on excess stock return in isolation. The full sample represents the

regression for the full data sample, the below sample represents the sample of below average scoring countries on sustainability, the above sample represents the above average scoring countries on sustainability. CSR is the average ESG score, control variables are respectively leverage as total debt/total assets, research and development as intangible assets/total assets, capital expenditure as capital expenditure/total assets, size as the natural logarithm of total assets, GDP as the natural logarithm of GDP per capita, sales growth as annual percentage change in total sales volume. The sample period is 2004-2014 and all right hand side variables are lagged by one period. End of the year values are taken for all variables. Note: Statistical significance at a 10%, 5% and 1% level is indicated by the display of *,** or ***. Standard errors are reported in brackets[] beneath the coefficients.

ESG country rating

Model (1) Full Sample Below Above

CSR -0.001*** [0.000] 0.000 [0.001] -0.001*** [0.000] LEV -0.118 [0.049] -0.355** [0.162] -0.099* [0.051] R&D 0.155*** [0.052] 0.255* [0.163] 0.156*** [0.056] CAPEX -0.005*** [0.001] -0.010* [0.006] -0.004*** [0.001] SIZE -0.235*** [0.014] -0.086** [0.039] -0.254*** [0.015] GDP -0.058 [0.088] -0.024 [0.151] -0.221* [0.128] SG 0.001*** [0.000] 0.001* [0.000] 0.000*** [0.000] Constant 4.319*** [0.924] 1.614 [1.426] 6.351*** [1.383] Adj. R-squared 0.102 0.116 0.107 Observations 17441 2341 15100

Year Fixed Effects YES YES YES

Firm Fixed Effects YES YES YES

4.2. Risk on both the firm and country level: interaction variables.

(31)

30 sample. The coefficients indicate the economic significance. For the full sample, the economic significance indicates the following: ceteris paribus, a 1 standard deviation increase of the z-score, is followed by a 0.18 standard deviation increase of the firm’s excess return, indicating that when firms improve their financial health, they are rewarded with an increase in excess returns. The BETA shows a negative coefficient. The p-values indicated that the BETA is statistically significant at the 10% level for the full sample (β=-0.082), at the 1% level for the below sample (β=-0.802), and not statistically significant for the above sample (β=-0.056), indicating that when a stock’s volatility increases, the excess returns fall. This is in line with the findings of Sassen et al. (2016), who found a negative relation between BETA and CFP (measured as ROA). However, the CAPM dictates that excess returns increases with the beta.

(32)

31

Table 5 Model 2. Model(2) tests the interaction effect of ZSCORE*CSR and BETA*CSR. The full sample represents the

regression for the full data sample, the below sample represents the sample of below average scoring countries on sustainability, the above sample represents the above average scoring countries on sustainability. CSR is the average ESG score, ZSCORE is Altman’s z-score, and BETA is stock beta calculated based on the CAPM model, using yearly excess stock return. Control variables are respectively leverage as total debt/total assets, research and development as intangible assets/total assets, capital expenditure as capital expenditure/total assets, size as the natural logarithm of total assets, GDP as the natural logarithm of GDP per capita, sales growth as annual percentage change in total sales volume. The sample period is 2004-2014 and all right hand side variables are lagged by one period. End of the year values are taken for all variables. Note: Statistical significance at a 10%, 5% and 1% level is indicated by the display of *,** or ***. Standard errors are reported in brackets[] beneath the coefficients.

CSR -0.001** [0.000] -0.000 [0.000] -0.000 [0.001] -0.001 [0.001] -0.001** [0.000] -0.000 [0.000] LEV -0.041 [0.051] -0.112** [0.049] -0.348* [0.183] -0.270 [0.186] -0.016 [0.052] -0.104** [0.050] R&D 0.233*** [0.055] 0.144*** [0.054] 0.408** [0.185] 0.207 [0.205] 0.229*** [0.057] 0.152*** [0.055] CAPEX -0.005*** [0.001] -0.005*** [0.001] -0.009 [0.006] -0.008 [0.007] -0.005*** [0.001] -0.005*** [0.001] SIZE -0.222*** [0.015] -0.238*** [0.015] -0.055 [0.043] -0.095** [0.044] -0.240*** [0.016] -0.253*** [0.015] GDP -0.041 [0.092] -0.097 [0.120] -0.012 [0.159] 0.386 [0.330] -0.180 [0.133] -0.206 [0.134] SG 0.001*** [0.000] 0.001*** [0.000] 0.000 [0.000] 0.001*** [0.000] 0.001*** [0.000] 0.001*** 0.000 ZSCORE 0.026* [0.010] 0.053** [0.027] 0.037 [0.011] ZSCORE*CSR 0.000 [0.000] 0.000 [0.000] 0.000 [0.000] BETA -0.082* [0.050] -0.802*** [0.273] -0.056 [0.050] BETA*CSR -0.000 [0.049] 0.005* [0.002] -0.000 [0.000] Constant 3.833*** [0.978] 4.857*** [1.274] 1.627 [1.210] -2.044 [3.117] 5.601*** [1.432] 6.240*** [1.436] Adj. R-squared 0.107 0.095 0.127 0.108 0.113 0.100 Observations 16716 16092 2529 1416 14626 14676 Year Fixed Effects

YES YES YES YES YES YES

Firm Fixed Effects

(33)

32 Table 6 reports the results of model 3, which tests the interaction effect at the country level. When testing the interaction effect of CORISK*CSR, the adjusted R-squared for the full sample falls slightly compared to the adjusted R-squared with the ZSCORE for model 2, indicating that less variance in the dependent variable is predicted by the independent variable when ZSCORE is excluded and CORISK is included. For model 1 and 2 the adjusted R-squared is higher for the above sample. For model 3 and 4 the adjusted R-squared is higher for the below sample. CORISK seems to hold higher explanatory power for the above sample. For the below sample, the CORISK variable is statistically significant at the 10% level according to the p-value (β=-0.011), indicating a negative relationship between environmental risk and excess return. When institutional development increases (which was the used proxy for environmental risk), excess return falls within this model. Regarding the above sample, the CORISK variable is significant at the 5% level (β=-0.009). This is not in line with the results of El Ghoul et al. (2016), who found a positive relationship between CFP (measured as ROA) and institutional development, significant at the 1% level. This could be due to the different measure used for CFP. Market measures are found across literature to react differently than accounting measures.

(34)

33

Table 6 Model 3. Model(3) tests the interaction effect The full sample represents the regression for the full data sample, the

below sample represents the sample of below average scoring countries on sustainability, the above sample represents the above average scoring countries on sustainability. CSR is the average ESG score, CORISK is the average of the governance indicators from the World Bank. control variables are respectively leverage as total debt/total assets, research and development as intangible assets/total assets, capital expenditure as capital expenditure/total assets, size as the natural logarithm of total assets, GDP as the natural logarithm of GDP per capita, sales growth as annual percentage change in total sales volume. The sample period is 2004-2014 and all right hand side variables are lagged by one period. End of the year values are taken for all variables. Note: Statistical significance at a 10%, 5% and 1% level is indicated by the display of *,** or ***. Standard errors are reported in brackets[] beneath the coefficients.

CSR -0.002*** [0.000] -0.002 [0.000] -0.018*** [0.004] LEV -0.102** [0.050] -0.301* [0.180] -0.086* [0.051] R&D 0.158*** [0.055] 0.099 [0.190] 0.182*** [0.057] CAPEX -0.005*** [0.001] -0.010 [0.006] -0.004*** [0.001] SIZE -0.239*** [0.015] -0.086** [0.041] -0.261*** [0.016] GDP -0.101 [0.092] 0.081 [0.164] -0.474*** [0.140] SG 0.001*** [0.000] 0.001* [0.000] 0.001*** [0.000] CORISK -0.002 [0.003] -0.011* [0.006] -0.009** [0.004] CORISK*CSR 0.000 [0.000] 0.000 [0.001] 0.000*** [0.000] Constant 4.991*** [0.967] 1.167 [1.473] 9.924*** [1.562] Adj. R-squared 0.094 0.084 0.103 Observations 16468 2047 14421

Year Fixed Effects YES YES YES

Firm fixed Effects YES YES YES

(35)

(36)

35

Table 7 Model 4. Model (4) tests the interaction effect The full sample represents the regression for the full data sample, the

below sample represents the sample of below average scoring countries on sustainability, the above sample represents the above average scoring countries on sustainability. CSR is the average ESG score, ZSCORE is Altman’s z-score, and BETA is stock beta calculated based on the CAPM model, using yearly excess stock return, CORISK is the average of the governance indicators from the World Bank. control variables are respectively leverage as total debt/total assets, research and development as intangible assets/total assets, capital expenditure as capital expenditure/total assets, size as the natural logarithm of total assets, GDP as the natural logarithm of GDP per capita, sales growth as annual percentage change in total sales volume. The sample period is 2004-2014 and all right hand side variables are lagged by one period. End of the year values are taken for all variables. Note: Statistical significance at a 10%, 5% and 1% level is indicated by the display of *,** or ***. Standard errors are reported in brackets[] beneath the coefficients.

CSR -0.001** [0.001] -0.001*** [0.002] -0.001 [0.004] -0.001 [0.005] -0.002*** [0.004] -0.002*** [0.004] LEV -0.029 [0.052] -0.097* [0.050] -0.330* [0.199] -0.130 [0.214] -0.006 [0.053] -0.096* [0.051] R&D 0.237*** [0.058] 0.144*** [0.055] 0.276 [0.219] -0.194 [0.207] 0.254*** [0.059] 0.169*** [0.057] CAPEX -0.006*** [0.001] -0.005*** [0.001] -0.008 [0.007] -0.008 [0.009] -0.005*** [0.001] -0.005*** [0.000] SIZE -0.227*** [0.015] -0.240*** [0.015] -0.063 [0.046] -0.080* [0.047] -0.247*** [0.016] -0.257*** [0.016] GDP -0.104 [0.095] -0.233* [0.130] 0.095 [0.171] 0.108 [0.449] -0.437*** [0.142] -0.432*** [0.146] SG 0.001*** [0.000] 0.001*** [0.000] 0.000 [0.000] 0.001** [0.000] 0.001*** [0.000] 0.001*** [0.000] CORISK -0.001 [0.003] 0.001 [0.003] -0.011 [0.006] 0.009 [0.009] -0.008* [0.004] -0.008** [0.004] CORISK*CSR 0.000 [0.000] 0.000** 0.000 0.000 [0.000] 0.000 [0.000] 0.000*** [0.001] 0.000*** [0.000] ZSCORE 0.036*** [0.002] 0.046* [0.027] 0.038*** [0.000] ZSCORE*CSR -0.000 [0.000] 0.000 [0.000] 0.000 [0.000] BETA -0.092* [0.000] -1.363*** [0.405] -0.069 [0.051] BETA*CSR -0.000 [0.000] 0.007* [0.004] -0.000 [0.000] Constant 4.341*** [1.001] 6.246*** [1.374] 0.592 [1.559] -0.099 [3.991] 9.130*** [1.585] 9.420*** [1.641] Adj. R-squared 0.100 0.090 0.079 0.068 0.110 0.098 Observations 15864 15319 1843 1151 14021 14168 Year Fixed Effects

(37)

36

Firm Fixed Effects

YES YES YES YES YES YES

4.3. Sensitivity tests.

In order to check the sensitivity of my results for several small variations, different robustness tests were conducted. First, the sensitivity of the results to a different corporate financial performance measure was tested. A regression was run for model 4 in which an accounting measure was taken for CFP (ROA). The table that reports the results can be found in the appendix (B1). The regression results show a higher adjusted R-squared for the regressions in which ROA is the dependent variable than for the regressions in which excess stock return is the dependent variable. It indicates that the model explains more of the variance of ROA than it does of excess return. The relationship between CSR and CFP is also found to be negative, with little economic significance. Only the full sample, which takes ZSCORE as firm risk measure, indicates CSR to be significant at the 5% level (β=-0.055). The results concerning Hypothesis 1 do not change dramatically when an accounting measure for CFP is applied.

(38)

37 the returns which employ excess stock return, where the results show that decreasing risk (either risk of bankruptcy or stock volatility) positively relates to excess stock returns.

The second sensitivity test runs the regression for model 4, where the overall three ESG pillar score is replaced by the separate pillar scores, such that each separate score is tested consecutively. The results are robust to this sensitivity test. For each separate ESG score, the relationship with excess return was found to be negative, similar to the aggregate score. Therefore, I did not test the interaction effect for separate ESG scores.

Furthermore, Appendix A3 presents the results of the regression that tries to answer the question of reverse causality. It shows that CSR has a significant negative effect on excess stock returns, whereas excess stock returns have a significant positive effect on CSR, thus suggesting that while financially superior performing firms invest more in CSR, the CSR initiatives in itself do not reward them with higher excess stock returns compared to firms that invest less in CSR.

5. Evaluation and conclusion.

(39)

38 return, while taking an investor perspective. And no research has investigated the moderating role of firm risk on this relationship.

Using a sample of 17,258 firm-year observation from 32 countries, and a fixed effects model, the obtained results indicate a relation between CSR and CFP. CFP, measured with excess returns, is negatively related to CSR, which indicates support for the shareholder theory, which argues that investing in CSR wastes valuable resources and often results in decreasing financial performance. The excess return holds information on the investor appreciation of CSR initiatives, thus suggesting that this appreciation is negative. However, a sensitivity test with ROA does show that the results are fairly robust to this change

(40)

39 Although this research provides some significant insights regarding the investor appreciation of CSR initiatives and the moderating effect of risk, some limitations can be noted. First of all, I performed a Granger causality test, which indicates a bidirectional relation between CSR and CFP. This makes it difficult to assess the effect CSR has on CFP. This simultaneity problem between excess returns and CSR, as indicated by the Granger causality test, can be worth considering as a subject of future research, for example, by employing an event study. Additionally, the sample focused on 32 countries, where the sample was split into countries scoring below- and above-average in terms of sustainability, to acquire insights on the differences country sustainability holds for the relations, as most research focuses on high sustainable samples, such as firms based in the U.S. and Europe. However, the below sample suffers from data limitations. Only little data can be collected for the countries in this sample, severely restricting the size of the sample. Little significance is found and little inferences can be made. Future research could focus on such a sample and extend it. Analyzing the trends in these countries can add further understanding to the analysis. Furthermore, future research can focus on how differences in the country context influence the CSR-CFP relationship. Differences such as culture and economic development can be expected to moderate the relationship. Future research could clarify how these mechanisms work in different countries.

6. References.

Alexander, G. J., & Buchholz, R. A., 1978. Corporate social responsibility and stock market performance. Academy of Management Journal, 21(3), 479-486.

Altman, E. I., 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The Journal of Finance, 23(4), 589-609.

Attig, N., El Ghoul, S., Guedhami, O., & Suh, J., 2013. Corporate social responsibility and credit ratings. Journal of Business Ethics, 117(4), 679-694.

(41)

40 Bénabou, R., & Tirole, J., 2010. Individual and corporate social responsibility. Economica, 77(305), 1-19.

Bouslah, K., Kryzanowski, L., & M’Zali, B., 2013. The impact of the dimensions of social performance on firm risk. Journal of Banking & Finance, 37(4), 1258-1273.

Bowen, H. R., & Johnson, F. E., 1953. Social responsibility of the businessman Harper.

Brammer, S., Brooks, C., & Pavelin, S., 2006. Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial Management, 35(3), 97-116.

Callan, S. J., & Thomas, J. M., 2009. Corporate financial performance and corporate social performance: An update and reinvestigation. Corporate Social Responsibility and Environmental Management, 16(2), 61-78.

Cheng, B., Ioannou, I., & Serafeim, G.,2014. Corporate social responsibility and access to finance. Strategic Management Journal, 35(1), 1-23.

Dalal, D. K., & Zickar, M. J., 2012. Some common myths about centering predictor variables in moderated multiple regression and polynomial regression. Organizational Research Methods, 15(3), 339-362.

Dhaliwal, D. S., Li, O. Z., Tsang, A., & Yang, Y. G., 2011. Voluntary nonfinancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review, 86(1), 59-100.

(42)

41 El Ghoul, S., Guedhami, O., Kwok, C. C., & Mishra, D. R., 2011. Does corporate social responsibility affect the cost of capital? Journal of Banking & Finance, 35(9), 2388-2406.

Eskeland, G. S., & Harrison, A. E., 2003. Moving to greener pastures? multinationals and the pollution haven hypothesis. Journal of Development Economics, 70(1), 1-23.

European Commission. Directorate-General for Employment., 2001. Promoting a European framework for corporate social responsibility: Green paper Office for Official Publications of the European Communities.

Filbeck, G., & Gorman, R. F., 2004. The relationship between the environmental and financial performance of public utilities. Environmental and Resource Economics, 29(2), 137-157.

Freeman, R. E., 1984. Stakeholder management: Framework and philosophy. Pitman, Mansfield, MA,

Godfrey, P. C., 2005. The relationship between corporate philanthropy and shareholder wealth: A risk management perspective. Academy of Management Review, 30(4), 777-798.

Gonenc, H., & Scholtens, B., 2017. Environmental and financial performance of fossil fuel firms: A closer inspection of their interaction. Ecological Economics, 132, 307-328.

Goss, A., & Roberts, G. S., 2011. The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance, 35(7), 1794-1810.

Gramlich, D., & Finster, N., 2013. Corporate sustainability and risk. Journal of Business Economics, 83(6), 631-664.

(43)

42 Hillman, A. J., & Keim, G. D., 2001. Shareholder value, stakeholder management, and social issues: What's the bottom line? Strategic Management Journal, 22(2) , 125-139.

Hubbard, R.G., 1998. Capital-market imperfections and investment. Journal of Economic Literature, 36(1), 193-225.

Jo, H., & Na, H., 2012. Does CSR reduce firm risk? evidence from controversial industry sectors. Journal of Business Ethics, 110(4), 441-456.

Jousma, M., 2017. What is the financial impact of corporate social responsibility? The moderating effect of national context. Unpublished working paper, University of Groningen, Groningen.

Margolis, J. D., Elfenbein, H. A., & Walsh, J. P., 2007. Does it pay to be good? A meta-analysis and redirection of research on the relationship between corporate social and financial performance. Unpublished working paper, University of Michigan, Ann Arbor.

Margolis, J. D., & Walsh, J. P., 2003. Misery loves companies: Rethinking social initiatives by business. Administrative Science Quarterly, 48(2), 268-305.

McWilliams, A., Siegel, D. S., & Wright, P. M., 2006. Corporate social responsibility: Strategic implications. Journal of Management Studies, 43(1), 1-18.

Miller, D., Lee, J., Chang, S., & Le Breton-Miller, I., 2009. Filling the institutional void: The social behavior and performance of family vs non-family technology firms in emerging markets. Journal of International Business Studies, 40(5), 802-817.

(44)

43 Orlitzky, M., & Benjamin, J. D., 2001. Corporate social performance and firm risk: A meta-analytic review. Business & Society, 40(4), 369-396.

Orlitzky, M., Schmidt, F. L., & Rynes, S. L., 2003. Corporate social and financial performance: A meta-analysis. Organization Studies, 24(3), 403-441.

Phillis, Y. A., Grigoroudis, E., & Kouikoglou, V. S., 2011. Sustainability ranking and improvement of countries. Ecological Economics, 70(3), 542-553.

Prieto‐ Carrón, M., Lund‐ Thomsen, P., Chan, A., Muro, A., & Bhushan, C., 2006. Critical perspectives on CSR and development: What we know, what we don't know, and what we need to know. International Affairs, 82(5), 977-987.

Renneboog, L., Ter Horst, J., & Zhang, C., 2011. Is ethical money financially smart? nonfinancial attributes and money flows of socially responsible investment funds. Journal of Financial Intermediation, 20(4), 562-588.

Sassen, R., Hinze, A., & Hardeck, I., 2016. Impact of ESG factors on firm risk in Europe. Journal of Business Economics: Zeitschrift für Betriebswirtschaft, 86(8), 867-904.

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.

Teoh, S. H., Welch, I., & Wazzan, C. P., 1999. The effect of socially activist investment policies on the financial markets: Evidence from the South African boycott. The Journal of Business, 72(1), 35-89.

(45)

44 Van Beurden, P., & Gössling, T., 2008. The worth of values–a literature review on the relation between corporate social and financial performance. Journal of Business Ethics, 82(2), 407-424.

Waddock, S. A., & Graves, S. B., 1997. The corporate social performance-financial performance link. Strategic Management Journal, 18(4), 303-319.

Williamson O. E., 1985. The Economic Institutions of Capitalism Firms Markets Relational Contracting.

7. Appendix.

Appendix A1: Probability results redundancy test and Hausman test for regression models and robustness test 1.

Redundant fixed effects test Hausman test

Cross-section F Cross-section Chis-square Cross-section random

Excess Returns ROA Excess Return ROA Excess return ROA

Model (1) 0.000 0.000 0.000 0.000 0.000 0.000

Model (2) 0.000 0.000 0.000 0.000 0.000 0.000

Model (3) 0.000 0.000 0.000 0.000 0.000 0.000

Model (4) 0.000 0.000 0.000 0.000 0.000 0.000

Appendix A1 shows the results (p-values) of the redundancy tests and the Hausman tests for all four models. A significant value (p-value <0.01) for the redundant fixed effects indicates that

fixed effects are necessary and a significant value (p-value <0.01) for the Hausman test indicates that random effects are not appropriate. The results show that for all performed regression fixed effects is the appropriate model.

Appendix A2: Unit Root test: (1) Fischer ADF individual unit root test (2) Levin, Lin & Chu common unit root test.

The null hypothesis is defined as the presence of a unit root.

(46)

45 ROA 0.00 0.00 EXRETURN 0.00 0.00 SG 0.00 0.00 SIZE 0.00 0.00 SOS 0.00 0.00 ZSCORE 0.00 0.00 BETA*CSR 0.00 0.00 CORISK*CSR 0.00 0.00 ZSCORE*CSR 0.00 0.00

Appendix A3: Multicollinearity test: Centered VIF scores for full sample.

VIF Model(1) Model(2) Model(3) Model(4)

CSR 1.0 2.0 30.0 (1.0) 42.47 (2.0) SIZE 1.2 1.2 1.2 1.3 LEV 1.0 1.1 1.0 1.1 R&D 1.1 1.1 1.1 1.1 SG 1.0 1.0 1.0 1.0 CAPEX 1.0 1.0 1.0 1.0 GDP 1.0 1.0 1.1 1.1 CORISK 1.3 1.3 CORISK*CSR 36.2 (1.3) 41.4 (1.7) ZSCORE 3.9 4.1 ZSCORE*CSR 6.3 6.7 BETA 1.5 1.6 BETA*CSR 2.4 2.7

Where CORISK is mean centered the VIF score is between brackets.

Supervisor prof. Dr, L.J.R. Scholtens

A test of the moderating effect of firm risk and environmental risk on the

relationship between corporate social responsibility and corporate financial

performance.

Contents