1
The Effect of Corporate Social Responsibility on Stock Return in
the US Market: A Comparative Industry Analysis
Abstract
This paper examines the effect of corporate social responsibility (CSR) on stock return across three industries (healthcare, mining and oil). In line with the literature the results show a significant positive relationship between CSR and stock return for all industries combined. The comparative industry analysis shows that the effect of CSR on stock return differs across industries. The results in all three industries show a positive effect of CSR on stock return. The effect is largest in the healthcare industry and smallest in the oil industry.
JEL classification: G17; M14
Keywords: Corporate social responsibility; Stock return; Comparative industry analysis
Student number: S1898183
Name: Sander Marissen
2
1. Introduction
This paper examines the effect of CSR performance on stock returns in the different industries of the US market. As starting point I try to define CSR; however, there is not one single definition. As argued by Malik (2014), pp 423: “The term social responsibility is a brilliant one; it means something, but not always the same thing to everybody”. The review study by Malik (2014) adopts the stakeholders-based definition of CSR; I will also follow this definition. The stakeholders-based definition of CSR is explained by Malik (2014), pp 425: as “CSR is a firm’s various voluntary initiatives toward its different stakeholders, such as customers, suppliers, regulators, employees, investors and communities”. Basically, the company does not only maximize their financial profit, it additionally takes the community, customers etc. into account.
3
strongly over time”. Guenster et al. (2011) argue that this occurs because in the beginning eco-efficient companies were undervalued, but they underwent an upward price correction. Additionally, Arx and Ziegler (2008) provide an explanation for this increase in importance of CSR activities. They compare firms engaging is CSR with inactive CSR firms in the same industry and conclude that financial markets value the activities in CSR. Thus, firms engaging in CSR do not only avoid the punishment from scandals, they even generate a financial profit from the activity according to Arx and Ziegler (2008).
Malik (2014) studies the existing literature on the value-enhancing capabilities of CSR. The following was argued by Malik (2014), pp 433: “CSR is used as a strategic tool to maximize value, and firms with better CSR performance have greater potential to increase shareholder value”. In addition, Malik (2014) argues that several investigated relationships in the field of CSR have led to inconsistent results. Furthermore, Malik (2014) recommends future researchers to conduct a comparative industry analysis, in order to determine the different effects of CSR on firm value for specific industries.
This paper will follow the recommendation of Malik (2014) and aims to examine the effect of CSR on stock return in different industries. The focus will be on the US healthcare industry, mining industry and oil industry. The main research question is:
What is the effect of CSR on stock return, and does the effect of CSR on stock returns differ between industries?
4
To examine the research question I develop a model relating CSR to stock return. The model incorporates interaction terms between industry and CSR to analyze the comparative effect of the industries. The data sample contains firms in the healthcare, mining and oil industry for the period 2002-2013.
I will start with a literature review, which will discuss the relevant literature and the development of a theoretical framework, followed by an explanation of the model and methodology. Thereafter the data will be described. Then, the results will be presented and discussed. Finally a short summary and the conclusions will be given.
2. Literature review and conceptual framework
5
have greater potential to increase shareholder value as well as the value of other stakeholders”. According to Malik (2014) CSR impacts on the stock market return in three distinctive ways; the cost of equity, the cost of debt and analysts’ forecasts. Malik (2014) finds that most studies on this subject report a strong negative relationship between CSR and cost of equity. In addition Malik (2014) argues that both cost of equity and cost of debt decrease, since high CSR firms can access the capital markets better. According to Malik (2014) the social responsible firms present their financial reports in a responsible way. Therefore, analysts trust the financial reports of social responsible firms more, and they can provide better forecasts. Furthermore, Malik (2014) suggest future researchers to conduct a comparative industry analysis to investigate if the effect of CSR on stock returns differs across industries. Arx and Ziegler (2008) argue that financial markets value the activities in CSR if firms engaging in CSR are compared with non-CSR firms within the same industry. These propositions lead to the idea that the relationship between CSR and stock return could differ across industries.
6
Bénabou and Tirole (2009) argue that if more people participate in CSR, the honor gained from it decreases; however, the stigma of those who do not participate is increased. Figure 2 illustrates this relationship.
This relationship is at the individual level, with the stakeholder theory I will argue that it could also apply at the firm level. Since the management of the firm takes all the stakeholders into account, the firm could be seen as the sum of all the stakeholders’ interest. In my opinion the theory of Bénabou and Tirole (2009) could hold for a lot of the stakeholders of a firm, and therefore it could impact the decisions of management significantly. This would mean that industries with high CSR scores, gain from participating in CSR by avoiding stigma. While in industries with low CSR scores, the gain from participating in CSR is the increase of honor. This could be a reason for the differences across industries for the relationship between CSR and stock returns.
7
Again, this theory explains the relationship at the individual level. With the stakeholder theory I try to change this relationship from the individual level to the firm level. The stakeholder theory argues that not only shareholders influence the firm’s management, but also other stakeholders like suppliers, employees and customers. In my opinion the theory of Bénabou and Tirole (2009) could hold for a lot of the stakeholders of a firm, thereby the management of a firm could be influenced. When I look at Figure 3 at the firm level, I can derive that engaging in CSR might be most beneficial in industries with high CSR levels. In addition also in industries with low CSR levels the gain from engaging in CSR might be large. However, in industries with mediocre CSR levels, the gain from engaging could be minimal according to the theory of Bénabou and Tirole (2009). So, the theory developed by Bénabou and Tirole (2009) could explain the differences across industries in the relationship between CSR and stock return, if it is perceived at the firm level.
8
hardly show any similarities with the oil and mining industry. With two similar and one completely different industry, the regression results could be very insightful. The theoretical framework leads to the following two hypotheses.
Hypothesis 1: I expect to find a positive relationship between CSR and stock performance.
This is in line with the results of Nuryaman (2013), Margolis et al (2009) and Malik (2014).
Hypothesis 2: I expect to find that the effects of CSR on stock return will differ across industries.
In the following paragraph a model will be developed to examine the hypothesized relationships between CSR and stock return.
3. Methodology
9
Firm size; Fama and French (1992), pp 423: find a “strong relationship between average return and size”. In addition, Nuryaman (2013) also includes a control variable for firm size in his model. I use the logarithm of total assets as proxy for firm size, which is in line with the models by Gregory et al. (2014) and Lioui and Sharma (2012). In line with Lioui and Sharma (2012), I expect firm size to have a significant negative estimator. This means that an increase in firm size will lead to a decrease in stock return, since firm size has a negative effect on growth according to Lioui and Sharma (2012).
Market to book ratio; El Ghoul et al. (2011) include a control variable for the book to market value in the analysis of the effect of CSR on cost of capital. In addition, Fama and French (1992) find a positive relationship between the book to market ratio and the stock returns. Therefore, I include the market to book ratio as a control variable, which I expect to have a significant positive estimator.
10
This results in the following model:
In this equation the dependent variable is stock return. is the intercept, which I expect to be significantly different from zero, due to for instance inflation. is the estimator for the effect of CSR score on the stock return in the industry not mentioned in the interaction terms. estimate the effects of the control variables on stock return; firm size, leverage and market to book ratio. estimate the effects of the interaction terms on stock return. In order to accept the aforementioned hypothesis 2, , and should be significantly different from zero. For instance, in the case of , the effect of CSR on stock return in industry i would be significantly different from the effect of CSR on stock return in the not interacted industry. To test the effect of CSR on stock return in all industries combined I omit the interaction terms. In order to confirm hypothesis 1, that CSR has a positive effect on stock return, should be significantly larger than zero.
11
4.
Data
The data used to examine the developed model I gathered by using DataStream. Thomson Reuters provides a large dataset, called DataStream, with all kinds of economic data on countries and companies across the world. Stock return is manually computed from historical stock prices and dividends provided by DataStream. Furthermore, DataStream has data on CSR scores from the Asset4 Thomson Reuters database, which contains CSR information on 3,000 global firms from 2002 onward. The Asset4 Thomson Reuters database combines the scores of 4 pillars; corporate governance, economic, environmental and social. The final CSR score is constructed by giving each pillar an equal weighting. The data on total assets was also provided by DataStream. Additionally the leverage was determined by calculating the debt-to-equity ratio. Data on debt and equity of the firms was again provided by DataStream. Finally, the market to book ratio is the market value of the company divided by its book value. Also the data on the market and book value was gathered through DataStream.
12
Furthermore, all firms which ceased to exist during the specified time period are deleted from the sample. Below I present the descriptive statistics in Table 1.
Table 1: Descriptive statistics Stock
return (%)
CSR score Firm size (*$1mln) Leverage (%) Market to book ratio Mean 15.140 47.062 11.416 117.021 2.900 St dev 52.683 24.881 13.054 232.261 3.118 Median 13.720 41.745 6.822 66.395 2.175 Min -90.174 7.800 0.479 -728.646 -6.917 Max 388.705 97.510 69.443 1593.878 25.797 Obs 394
The average yearly stock return of all the companies in the sample is 15.140%, which can be seen in the first column of Table 1. The stock returns show a lot of variation, varying from -90.174% to 388.705% and a standard deviation of 52.683%. Furthermore the average CSR score is 47.062, with a standard deviation of 24.881. Table 1 also presents the descriptive statistics of the control variables. The total number of observations is 394.
13
Table 2: Descriptive statistics on CSR, separated industries CSR Score mining industry CSR Score oil industry CSR Score healthcare industry Mean 56.997 41.535 40.767 St Dev 26.616 23.585 22.144 Median 56.330 36.170 34.990 Min 12.56 7.800 4.280 Max 96.570 79.750 97.510 Obs 122 211 92
Table 2 shows the descriptive statistics of CSR in each of the three separate industries. The characteristics of the CSR scores for the healthcare and oil industry differ from the characteristics of the CSR scores for the mining industry. So, while I argued that the oil and mining industry are in the same field, their CSR score characteristics show no similarities. This could be an indicator that the effect of CSR on stock return differs in each industry.
Furthermore, table 16-18 in the appendix present descriptive statistics for each specific industry. From these tables two observations can be made. The firms in the oil industry are larger than the firms in the healthcare and mining industry. Besides that, the firms in the healthcare industry have higher leverage than firms in the oil and mining industry.
5. Results
14
Table 3: Regression results without interaction terms
Stock Return T-statistic C 282.89* 1.858 CSR 0.423*** 2.605 Firm size -18.823* -1.933 Leverage -0.066** -2.118 MtB ratio 5.781** 2.320 Obs 394 0.436 D-W statistic 2.259
(*, **, *** indicates significant at the 10%, 5%, 1% level)
15
shareholder wealth in comparison to non-CSR investments.
The Durbin-Watson statistic of 2.259 indicates that the errors of the model are not autocorrelated (Hill et al. 2008). In addition, the Dickey-Fuller test on the residuals of the estimation shows that the residuals are stationary (appendix, table 7). Therefore the residuals are cointegrated and the regression relationship is not spurious (Hill et al. 2008). To test the appropriateness of fixed effects for the model, I conduct a Hausman test. The results (appendix, table 5) support my decision to use fixed effects for both time and firm.
So, the results validate hypothesis 1, that CSR has a positive and significant effect on stock return. This supports the findings of previous research by Margolis et al. (2009) and Arx and Ziegler (2008).
16
Table 4: Regression results with interaction terms
Stock Return T-statistic
C 218.358 1.413 CSR 1.059*** 3.195 Firm size -14.592 -1.481 Leverage -0.069** -2.333 MtB ratio 6.206** 2.495 CSR*Mining -0.713* -1.802 CSR*Oil -0.965** -2.525 Obs 394 0.445 D-W statistic 2.343
(*, **, *** indicates significant at the 10%, 5%, 1% level)
17
interaction term between CSR and the mining industry ( ) shows a significant negative effect at the 10% level. This indicates that, compared to the healthcare industry, the effect of CSR on stock return in the mining industry is 0.713% per 1% increase less than in the healthcare industry. So, a 1% increase in CSR in the mining industry only generates a 0.346% (1.059-0.713) increase in stock return, compared to a 1.059% increase in the healthcare industry. The estimator for the interaction term between CSR and the oil industry ( ) shows a significant negative effect at the 5% level. This indicates that, compared to the healthcare industry, the effect of CSR on stock return in the oil industry is 0.965% per 1% increase less than in the healthcare industry. So, a 1% increase in CSR in the oil industry only generates a 0.094% (1.059-0.965) increase in stock return, compared to a 1.059% increase in the healthcare industry. All control variables show the expected signs; however, the estimator for firm size is not significant. Thus, the positive relationship between CSR and stock return is largest in the healthcare industry, followed by the mining industry and the oil industry. So, Investments in CSR generate a larger stock return in the healthcare industry. Therefore, investors should take CSR more into account in their investment decision when they invest in firms in the healthcare industry in comparison to firms in the oil and mining industry.
The Durbin-Watson statistic of 2.343 indicates that the model errors are not auto correlated (Hill et al. 2008). In addition, the Dickey-Fuller test on the residuals of the estimation shows that the residuals are stationary (appendix, table 8). Therefore the residuals are cointegrated and the regression relationship is not spurious (Hill et al. 2008). To test the appropriateness of fixed effects for the model, I conduct a Hausman test. The results (appendix, table 6) support my decision to use fixed effects for both time and firm.
18
19
To validate the results of both regression analyses I performed several robustness checks, by splitting the sample in two time frames and by taking different industries in the interaction terms. For the first robustness check, I decided to divide the sample in pre-crisis (2002-2007) and post-crisis (2008-2013). The results of this robustness checks can be found in the appendix (Table 9-12). For the model without interaction terms, all variables lose their significance for the pre-crisis sample. This could be caused by the small number of observations remaining (111). For the post-crisis sample, the control variables for leverage and market to book ratio remain significant. The control variable for firm size does not remain significant. The effect of CSR on stock return increases, but drops to the 10% significance level. For the model with interaction terms, again all variables lose their significance for the pre-crisis sample. This could also be due to the small number of observations remaining (111). The control variables for leverage and market to book ratio remain significant at the 5% level in the post-crisis period. However, all other variables, including CSR and the interaction terms lose their significance. So, unfortunately the estimators for CSR and the interaction terms are not robust, however, a larger dataset might solve this shortcoming.
20
the mining industry at a 10% significance level. However, the estimator for the effect of CSR on stock return in the mining industry is insignificant. Furthermore, the estimator for the interaction term between CSR and the oil industry is insignificant. So, this robustness check indicates that the effects for the health industry remain significant. The effect of CSR on stock return in the health industry is indeed larger than in the mining and oil industry. However, the results for the mining and oil industry in the initial regression could be misleading, since CSR in the mining and oil industry by itself do not have a significant effect on the stock returns. Therefore, the interaction terms in the initial model should be interpreted carefully.
6. Conclusion
In this paper, I follow the recommendation by Malik (2014). I conduct a comparative industry analysis to examine whether the effect of CSR on stock return across three different industries. The sample contains 54 firms from the oil, mining and healthcare industry, for the period 2002-2013. First, I examine the relationship between CSR and stock returns for all 54 firms, by adding control variables for size, leverage and market to book ratio. The results show a significant positive effect of CSR on stock returns, in line with the hypothesis. By including two interaction terms between industry and CSR, I analyzed the difference in relationships between CSR and stock return. The results show that the effect of CSR on stock return is the highest in the healthcare industry, followed by the mining industry, and the effect is the smallest in the oil industry. All industries show positive and significant effects. Therefore, the second hypothesis is also accepted, the effect of CSR on stock return differs across industries.
21
comparative industry analysis to see if CSR has a larger effect in certain industries can be confirmed by the results of this paper. The effect of CSR on stock return is largest in the healthcare industry and smallest in the oil industry. When trying to relate this result to the existing literature, the observation of the higher debt to equity ratio in this industry in the data section is crucial. Malik (2014) argues that companies with high CSR levels have better access to the capital markets. The descriptive statistics of the healthcare industry show the largest debt to equity ratio of all industries, this could indicate that the healthcare industry indeed has easier access to at least the debt markets. However, the average CSR level in the oil industry is larger than the average CSR level in the healthcare industry. So, a reason for the higher effect of CSR on stock return in the healthcare industry could be that it is not about the actual CSR level, it is also about the perception of the CSR of the total industry. The oil and mining industry are associated with pollution, while the healthcare industry is not associated with those kinds of factors.
This paper has some data limitations and methodological issues. The data has limitation regarding the availability of CSR data; the Asset4 Thomson Reuters dataset only contains data for CSR from 2002 onward. Furthermore, I have some concerns regarding causality; table 15 in the appendix presents the results of the Granger causality test. The results show that stock return does not affect CSR; however, CSR does also not affect stock return according to the test. In order to deal with this I tried to take the lag of CSR; however, the Granger causality test then suggest reversed causality.
22
influence government policies. Governments might change their subsidies when they become aware of differences across industries in the relationship between CSR and stock return. Industries with a small effect might need a larger incentive to invest in CSR activities.
Future research could examine this relationship for a broader sample by adding industries and/or countries. This could give insight in whether the effects differ across countries or continents. By adding more appropriate control variables this research could be improved. In addition, other models, like the Fama French four factor model, could be used in examining these hypothesis. Also, future research could examine whether the effect is robust for different measures of CSR, by applying; for instance, the Msci KLD 400 Social Index.
7. References
Arx von, U., & Ziegler, A. (2008). The effect of CSR on stock performance: New evidence for the USA and Europe. Working Paper 08/85.
Bénabou, R., & Tirole, J. (2009). Individual and Corporate Social Responsibility. Economica, 2010, 77, 1-19.
Derwall, J., Koedijk, K., & Ter Horst, J. (2010). A tale of values-driven and profit-seeking social investors. Journal of Banking & Finance, 35, 2137-2147.
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.
Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikström, C., & Wold, S. (2006). Multi- and megavariate data analysis, part 2, advanced applications and method extensions. Umetrics Academy.
23
Freeman, E. (2004). The stakeholder approach revisited. Zeitschrift für Wirtschafts- und unternehmensethik, 5(3), 220-241.
Gregory, A., Tharyan, R., & Whittaker, J. (2014). Corporate social responsibility and firm value: Disaggregating the effects on cash flow, risk and growth. Journal of Business Ethics, 124, 633-657.
Guenster, N., Bauer, R., Derwall, J., & Koedijk, K. (2011). The economic value of corporate eco-efficiency. European Financial Management, 17, 679-704.
Hill, C., Griffiths, W., & Lim, G. (2008). Principles of econometrics third edition. John Wiley & Sons, Inc.
Lioui, A., & Sharma, Z. (2012). Environmental corporate social responsibility and financial performance: Disentangling direct and indirect effects. Ecological Economics, 78, 100-111.
Malik, M. (2014). Value-enhancing capabilities of CSR: a brief review of contemporary literature. Journal of Business Ethics, 127(2), 419-438.
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.
Moore, L., Silva, I., & Hartmann, S. (2012). An investigation into the financial return on corporate social responsibility in the apparel industry. Journal of Corporate Citizenship, 105-123.
24
8. Appendix
Table 5: Redundant fixed effects test model without interaction terms
Statistic Prob.
Cross-section F 1.477 0.024
Period F 17.333 0.000
Cross-Section/Period F 4.569 0.000
Reject the null hypothesis of not appropriately applying fixed effects
Table 6: Redundant fixed effects test model with interaction terms
Statistic Prob.
Cross-section F 1.504 0.019
Period F 19.598 0.000
Cross-Section/Period F 4.652 0.000
Reject the null hypothesis of not appropriately applying fixed effects
Table 7: Cointegration test model without interaction terms (dep. variable residuals)
Coefficient Prob.
Errors(-1) -1.221 0.000
D(Errors(-1)) 0.086 0.0953
Reject the null hypothesis of nonstationarity of the residuals
Table 8: Cointegration test model with interaction terms (dep. variable residuals)
Coefficient Prob.
Errors(-1) -1.307 0.000
D(Errors(-1)) 0.127 0.013
25
Table 9: Regression results without interaction terms (2002-2007)
Stock Return T-statistic C 92.634 0.583 CSR 0.107 0.626 Firm size -4.345 -0.429 Leverage -0.013 -0.579 MtB ratio 1.009 0.537 Obs 111 0.549 D-W statistic 2.222
(*, **, *** indicates significant at the 10%, 5%, 1% level)
Table 10: Regression results without interaction terms (2008-2013)
Stock Return T-statistic C 453.796 1.358 CSR 0.538* 1.794 Firm size -30.576 -1.450 Leverage -0.090*** -2.690 MtB ratio 7.996** 2.380 Obs 283 0.452 D-W statistic 2.251
26
Table 11: Regression results with interaction terms (2002-2007)
Stock Return T-statistic C 104.357 0.625 CSR -0.209 -0.599 Firm size -5.135 -0.484 Leverage -0.014 -0.620 MtB ratio 1.030 0.544 CSR*Mining 0.230 0.414 CSR*Oil 0.500 1.041 Obs 111 0.543 D-W statistic 2.257
(*, **, *** indicates significant at the 10%, 5%, 1% level)
Table 12: Regression results with interaction terms (2008-2013)
Stock Return T-statistic C 434.025 1.192 CSR 1.086 1.031 Firm size -29.391 -1.280 Leverage -0.086** -2.410 MtB ratio 7.991** 2.400 CSR*Mining -0.645 -0.556 CSR*Oil -0.674 -0.567 Obs 283 0.449 D-W statistic 2.267
27
Table 13: Regression results with interaction terms for Mining and Health
Stock Return T-statistic C 218.358 1.412 CSR 0.094 0.484 Firm size -14.592 -1.481 Leverage -0.069** -2.333 MtB ratio 6.206** 2.495 CSR*Mining 0.252 0.808 CSR*Health 0.965** 2.525 Obs 394 0.445 D-W statistic 2.343
(*, **, *** indicates significant at the 10%, 5%, 1% level)
Table 14: Regression results with interaction terms for Oil and Health
Stock Return T-statistic C 218.358 1.413 CSR 0.346 1.288 Firm size -14.592 -1.481 Leverage -0.069** -2.333 MtB ratio 6.206** 2.495 CSR*Health 0.713* 1.802 CSR*Oil -0.252 -0.808 Obs 394 0.445 D-W statistic 2.343
28
Table 15: Granger causality test CSR and stock return
F-Statistic Prob. CSR does not Granger
cause stock return
0.662 0.517
Stock return does not Granger cause CSR
2.075 0.127
Obs. 289
Table 16: Descriptive statistics mining industry Stock
return (%)
CSR score Firm size (*$1mln) Leverage (%) Market to book ratio Mean 12.559 57.624 8.659 86.331 3.525 St dev 61.254 26.572 9.346 151.039 3.587 Median 9.260 57.765 5.470 67.805 2.435 Min -79.297 12.560 0.479 0.000 0.300 Max 388.705 96.570 63.471 1593.878 25.797 Obs 118
Table 17: Descriptive statistics oil industry Stock
return (%)
29
Table 18: Descriptive statistics healthcare industry Stock
return (%)
