Is there a relationship between firm performance and Social CSR? An investigation for the United States of America.

(1)

Is there a relationship between firm performance and Social CSR? An

investigation for the United States of America.

Grace Chong1

Abstract

In the US, firms are losing the trust of customers and are now faced with the increasing pressure to partake in CSR activities. Studies show that CSR have no uniform conclusion whereas studies on the social aspect of CSR (Social CSR) and firm performance show a positive relationship. The Social CSR specifically affects the reputation of the firm thereby creating value for the firm. Reputation is fickle by nature so it might not be consistent. Findings could differ over time or between places. Thus I find it necessary to investigate whether the current relationship between Social CSR and firm performance is indeed positive. This study finds that the social aspect of CSR insignificant even with a quadratic model.

JEL classification: G12; G30

Keywords:

CSR; Financial performance; Social aspect of CSR

1_{Student number: s2350041}

Study program: MSc Finance Faculty of Economics and Business University of Groningen, The Netherlands First supervisor: dr. J.H. (Henk) von Eije

(2)

Page 2 of 31 1. Introduction

In 2008, Google joined the Global Network Initiative which is a non-governmental organization that protects the internet privacy rights of users (Google, 2008). Furthermore, Google (2008) stated in its annual report that it planned to make donations to enterprises whose goals were helping the poor, improve the environment, and other socially responsible acts. Microsoft (2017) disclosed of the donations it made to non-profit organizations and its plans to cater towards the less privileged. The purpose of these socially responsible acts can be summed up concisely by Johansson (2018), who argues in her article that this is the competitive edge companies should incorporate into their strategy to win over the trust of the Americans.

If we look at the literature on corporate social responsibility (henceforth CSR) and firm performance as an indication on whether the above is beneficial for the firm, we see that there are conflicting views on the relationship between CSR and firm performance. For some studies, the abnormal returns of socially responsible firms (or financial performance) are often statistically insignificant and such investments are usually associated with large firms (Mollet and Ziegler, 2014; Aras et al, 2010). Then there are studies which find statistically significant results but conflicting views as to the direction of the relationship between CSR and performance. For example, the study by Becchetti and Ciciretti (2009) shows socially responsible firms underperform on average relative to a control sample. Others, however, find that (in the long run) CSR firms tend to perform better relative to their counterparts (Carini et al, 2017).

(3)

Page 3 of 31 meta-analysis2 that the Social aspect of CSR (henceforth Social CSR) generally has a positive effect on firm performance, mostly due to the corporate reputation associated with Social CSR.

Barnett and Salomon (2012) find that the relationship between Social CSR and firm performance is firm specific, meaning that it could be negative for some but positive for others. Though they found a positive effect, Busch and Friede (2018) found a diminishing relationship over time with respect to the meta-analyses on the relationship between Social CSR and firm performance. They did not use actual data to come to this finding but instead used findings from previous papers. Therefore it is interesting to see if with real data, I can look at the current situation and still find a significant relationship between Social CSR and firm performance. Based on these arguments, I present my research question as follows:

How does Social CSR impact firm performance nowadays?

In this paper, I tackle my research question by using a cross section fixed effects and time fixed effects panel regression on a sample of US firms for the fiscal years 2010-2017.

My research contributes to the existing literature by offering a slightly different model approach in measuring the relationship between Social CSR and firm performance. Managers should consider following in the footsteps of socially responsible firms like Google and Microsoft because Social CSR activities are generally not detrimental to firm health. As for investors, they should consider adding CSR firms to their portfolios of investments, especially the firms who are more active in Social CSR to benefit from diversification if not improved returns.

The rest of the paper is organized as follows. Section 2 covers the relevant literature regarding CSR, Social SCR, and firm performance. Section 3 describes the approach used to evaluate the relationship between Social CSR and firm performance. Section 4 gives details as to where the data is obtained and discusses the descriptive statistics. Section 5 provides the results and an interpretation of these results. Lastly, section 6 provides concluding remarks and ends with recommendations for further study.

2

(4)

Page 4 of 31 2. Literature review and hypothesis development

2.1 An overview of CSR

In the available literature on CSR, many definitions exist for this concept. The lack of consensus over the definition of CSR implies that there is also no consensus of how to measure it (Aras et al., 2010). Diverse definitions and concomitantly different measurement approaches may therefore also naturally lead to different findings on the matter at hand (as already indicated in the Introduction).

Proponents against CSR tend to cite Friedman’s (1970) work, stating that firms have no such obligation to society. On the contrary, Schwartz and Saiia (2012) and Barnett and Salomon (2012) argued that Friedman (1970) never said that businesses have no social responsibilities. For the sake of simplicity, I will use the definition mentioned in Choi et al. (p. 292, 2010): “CSR can be defined as actions that appear to further some social good, beyond the interests of the firm and that which is required by law”. If we relax the definition of CSR to exclude the “beyond the interests of the firm” part and apply it to the context of what Friedman (1970) stated regarding how businesses should operate and its fiduciary duties towards shareholders, then it can be shown that Friedman’s (1970) work represents a constrained perspective of CSR. An example given by Schwartz and Saiia (2012) is that if a firm can prove that charitable activities have positive consequences on the firm’s bottom line then this action would be supported by Friedman (1970) as well provided that the firm does not lie about it being an act of altruism to the public.

If we agree with Schwartz and Saiia (2012), then the comparison of the stakeholder view (Freeman, 1984) versus the shareholder view (Friedman, 1970) could break down into being the two sides of the same coin. Both views state that firms have social responsibilities to take the relevant party or parties into consideration when planning corporate strategies. The stakeholder view regards the relevant parties to be all who have a stake in the firm whereas the shareholder view regards the relevant party to be the shareholders.

(5)

Page 5 of 31 that CSR investments increase costs, puts firms into economically disadvantageous positions, and lead to lower market value (Miralles-Quirós et al., 2018). Whether or not CSR conforms to the value enhancing theory or shareholder expense theory depends on what existing literature reveals.

Apparently, there is a lack of consensus in the literature regarding the relationship of CSR and firm performance from as early as the 1970s. One of the earlier studies conducted regarding CSR was by Alexander and Buchholz (1978) who used survey data to determine the relationship between CSR and stock performance. The researchers used the methodology introduced by earlier studies in which students and businessmen were asked to rank the top 40 leading firms from most socially responsible to the least socially responsible. Alexander and Buchholz (1978) found that the relationship was insignificant which contradicted the findings of existing literature that the relationship could be positive or negative. Alexander and Buchholz (1978) argued that a potential explanation for their finding was that stock markets were efficient and therefore any effects induced by higher perceived levels of CSR would immediately be reflected in the stock price. Approximately four decades after the finding by Alexander and Buchholz (1978), the conclusion of the meta-analysis by Wang et al. (2016) on CSR and firm performance indicated that the relationship was positive. As such, it appears that the value enhancing theory holds.

2.2 Social CSR

Even in the literature concerning Social CSR and firm performance, there has been no uniform consensus as to the direction of the relationship (Busch and Friede, 2018). One explanation for this discrepancy by Marom (2006) is that the investment is an upfront cost in one moment in time whereas the benefits are reaped over a longer time period. Thus studies using short time frames may only find a negative if not insignificant relationship (Marom, 2006). Preston and O’Bannon (1997) summed up the cause for such disparity in findings among research in their work but I will only mention what is relevant for my study.

(6)

Page 6 of 31 (Preston and O’Bannon, 1997) between Social CSR and firm performance in which good firm reputation must precede before any financial consequence can be observed/realized. The trade-off hypothesis (akin to the shareholder expense theory) posits that social investments lead to higher costs and puts firms at a competitive disadvantage relative to their peers who are less invested in Social CSR.

Van der Laan et al. (2008) and Busch and Friede (2018) both agree that the relationship between Social CSR and firm performance is positive. Contrary to this view, I do not expect such a relationship. Even if previous research showed that the Social CSR impacts firm performance positively like the meta-analysis by Orlitzsky et al. (2003), it is not certain that such effects remain. This is because Social CSR mostly impacts firm performance through corporate reputation which is fickle. Reputation is a volatile construct that is intangible. Negative news on firm misconduct could potentially have a stronger (and perhaps longer lasting) impact on corporate reputation than Social CSR would thereby result in a net loss. This line of reasoning follows from what prospect theory tells us (Van der Laan et al., 2008). Therefore, my null hypothesis and alternative hypothesis are as follows, respectively:

H0: There is no relationship between Social CSR and firm performance nowadays.

H1: There is a positive relationship observed nowadays.

(7)

Page 7 of 31 Bajic and Yurtoglu (2016) addressed these problems by using a global sample of firms over the period 2003 to 2012. They used a broad measure of CSR (namely ASSET43), which is created to be sufficiently diverse such that it can be generalized across countries. Furthermore, they analyzed whether their measure of CSR affects firm value and operating performance. They found a significant and positive relationship between firm performance and CSR but the impact of CSR on operating performance was insignificant. Moreover, using a single measure of CSR and/or omitting firm characteristics can lead to omitted variable bias.

Admittedly these matters mentioned above are related to studies regarding CSR and not Social CSR but I find it relevant to heed these warnings and take the necessary steps to avoid or mitigate these problems.

3. Methodology

In the studies conducted thus far, the majority focused on the linear relationship between CSR and firm performance. This linear relationship serves as my point of departure, where I look at this linear relationship as my baseline model. In my case, I am particularly interested in the social aspect of CSR therefore I modeled the impact of the social aspect of CSR on firm performance. I used RoA as a measure of firm performance and used RoE as a test of robustness. Moreover, Scholtens (2008) commented that the relationship between CSR and firm performance could be complex, perhaps U-shaped. Taking this perspective into account in the form of an additional robustness test, I essentially look at two types of models: a linear one and a quadratic one.

The first model is a linear model (also baseline model) which is a cross section fixed effects and time fixed effects panel regression as follows:

. (1)

where is the accounting performance measure (RoE and RoA) for firm in year ;

is the intercept; is the Social score that measures the social aspect of CSR; is the firm’s log of total revenue, is the leverage ratio of the firm which is defined as the

3

(8)

Page 8 of 31 Total Debt divided by the market value of equity, is the Market/Book ratio, is firm’s cash at time , is the market value of equity of the previous year (or at time

) is the cross section fixed effects, is the time fixed effects, and is the error term.

According to Fama and French (1993) and Modigliani and Miller (1958), firm performance is affected by firm size, market to book ratio, and capital structure. Consequently, I control for these three factors using , , and , respectively.

Bajic and Yurtoglu (2016) stated that if you were to conduct a study on one aspect of CSR and you do not include sufficient control variables then your model could suffer from omitted variable bias. Specifically, the error term would capture the other aspects of CSR that was left out in the explicit part of the model. Therefore, I included the marginal value of cash variable introduced in Faulkender and Wang (2006) and Zhang et al. (2015). Given the financial situation of a firm and its level of corporate governance, the value shareholders place on an additional dollar of cash will vary. Though this variable captures how cash affects performance, one can make inferences about the corporate governance practice (see Faulkender and Wang, 2006; Zhang et al., 2015) which is one component that makes up the general measure of CSR.

The second model, a quadratic cross section fixed effects and time fixed effects model (henceforth quadratic model), is an extension of the linear model with the addition of the square of the Social score, , as shown below:

. (2)

Scholtens (2008) stated that the relationship between CSR and firm performance was U-shaped. At the time of writing, not much research has explored this possibility with the social aspect of CSR hence I take the initiative to do so.

The correlation matrix in Table A1 in Appendix A indicates that and were highly correlated and the VIF4 values from Table A3 (in the Appendix) are above 10 which suggest that the quadratic model could suffer from multi-collinearity. This

4

(9)

Page 9 of 31 necessitated an additional model to be used which will serve the same purpose as the quadratic model.

The third model, a pseudo quadratic cross section fixed effects and time fixed effects model (henceforth pseudo quadratic model), was used to test for the presence of a complex relationship between firm performance and the social aspect of CSR. The pseudo quadratic model is specified exactly as in the linear model (equation 1). The result of this model is shown in Table 4. Its logTR had a high VIF value as shown in Table A6 in Appendix A which indicates that it introduces multi-collinearity in my model thus I had to remove the variable. Table 6 shows the results of the pseudo quadratic model without the logTR variable as a robustness test to the pseudo quadratic model with logTR. The difference between the pseudo quadratic model and the quadratic model lies in the sample used because for the pseudo quadratic regression, the sample will be divided into two groups: one with a large Social score (a Social score that is greater than or equal to 50) and the other with a small Social score (a Social score that is less than 50). Thus I estimated equation 1 twice for the sample with small Social scores since there are two performance measures and the same goes for the sample of large Social scores.

I created a dummy variable that splits the sample based on whether the Social score is less than 50 or not. The purpose of using 50 as the cutoff point between the large and small Social scores is due to the assumption that the scores fall in the range of 0 to 100. There is no available information as to what the actual range of the Social scores is but based on the distribution of the Social scores, it seems likely that the upper bound for these scores is 100 while the lower bound is 0 (for firms which do not practice CSR).

To test the robustness of the third model, I created a second dummy variable for the Social score with a different cutoff point. This cutoff point is calculated by taking the derivative of equation 2 with respect to then setting it equal to zero which results in the following:

. If we plug in the coefficients from Column 1 in

(10)

Page 10 of 31 case as well (refer to Table A7 in Appendix A) thus it was dropped in the second test of the model (results shown in Table 7).

4. Data and Descriptive statistics

I downloaded a list of all firms traded on the U.S. stock exchanges (i.e. NYSE, AMEX, and NASDAQ) via the NASDAQ website5. On the website, each stock exchange has its own listing of firms traded on their exchange. I therefore have 3414 firms from NASDAQ, 3160 firms from NYSE, and 334 firms from AMEX. After removing all financial and utility related companies from my sample of firms, I am then left with 2047 firms from NASDAQ, 1335 firms from NYSE, and 192 firms from AMEX. After pooling the firms into one sample and removing duplicates6, there were 3399 unique firms remaining. A second check reveals that there were more financials and utilities left in the sample. Once they were removed, the final sample consists of 1827 firms. These lists provide information on the firm’s industry and ticker symbol.

I used Thomson Reuters to gather yearly data on the following variables: Financial measures (RoE and RoA), Social score7, Book Value of Equity, Market Value of Equity, Cash, Total Revenue, Total Debt, and Total Assets. Currency variables in Thomsen Reuters are in local currency by default so special care was taken to ensure that all currency variables are expressed in U.S. Dollars.

To avoid having large coefficients resulting from substantial differences in the scale or distributions of the variables, a few changes had to be made. After inspecting the summary statistics of the Social score and reviewing the Thomsen Reuters methodology8 on how the score was calculated, I decided that there was no need to winsorize it. All variables, excluding Social score and the two dependent variables, are winsorized at the 5% and 95% tails to tackle the presence of outliers then transformed into log variables. I could have chosen to winzorize all variables excluding the Social score at the 5% and 95% tails but decided that the resulting winsorized sample for RoE and RoA could still appear ridiculous. Therefore I

5

Accessed on October 4, 2018 https://www.nasdaq.com/screening/companies-by-industry.aspx?exchange=NASDAQ

6

Duplicates exist since firms either cross list on the exchanges and/or have multiple listings (e.g. Google is listed twice since it has two different types of shares).

7_{What the social score is comprised of is explained in Appendix C} 8

The pdf version was downloaded on 29 January 2019 and can be accessed at:

(11)

Page 11 of 31 replaced all observations below zero by 1% and all observations above 100% by 99% for RoE and RoA.

Table 1 presents the descriptive statistics (number of observations, min, max, median, mean, and standard deviation) for the variables used in my models. The marginal value of cash is on average $0.03 but surprisingly the extreme values are bigger than the values in Zhang et al. (2015) and Faulkender and Wang (2006). Leverage ratio, despite having a mean and median of less than 1, has a relatively larger max value of 317. Market-to-book ratio also has rather big extreme values. This could indicate that I still have outliers in my leverage ratio and market-to-book ratio but based on the values of the mean and median, I mitigated the presence of outliers after winsorizing my data. Comparing the Social score dummy variable in Table 1 (cut-off point is 50) with the one in Table A8 in Appendix A (cut-off point is 43.55), we see that the mean is higher in Table 1 which means that the subsample of low Social scores is bigger in Table 1 than in Table A8.

5. Results

Table 2 displays the results from my baseline model (equation 1, section 3). In Column 1, the dependent variable is RoA whereas in Column 2 the dependent variable is RoE. The coefficients of the Social score are positive in both columns, indicating that the relationship between firm performance and Social CSR is positive albeit insignificant. In this case, my model fails to reject the null hypothesis of there being no relationship between Social CSR and firm performance. The fact that the adjusted R-squared for both specifications of the model is between 0.0214 and 0.0234 means that my linear model has practically no explanatory power. Moreover, the VIF test9 for my baseline model has a value greater than 10, indicating the presence of multi-collinearity.

Table 3 displays the results from my quadratic model. Just as in Table 2, the dependent variables for Column 1 and Column 2 are RoA and RoE, respectively. This time, the Social score in both columns contradict the result obtained in Table 2, indicating that the relationship between Social CSR and firm performance is negative. The sign of the Social score squared is positive in both columns, indicating that the relationship between Social CSR and firm performance could be U-shaped. Once again, the null hypothesis of no relationship is not rejected. The adjusted R-squareds for this model are slightly higher than the previous model but not substantially so thus my model still lacks explanatory power. The

9

(12)

Page 12 of 31 values of the VIF test (reported in Table A3 of Appendix A) are much higher for this model than the previous model with more than one variable exceeding 10. Despite having higher explanatory power, the presence of multi-collinearity is higher therefore the model is less reliable.

Table 4 shows the results of the pseudo quadratic model which is the robustness test of the quadratic model. Small (large) in Column 1 and Column 2 is an indicator for the sample of low (high) Social scores for the dependent variable RoA and RoE. Small represents the Social scores that are less than 50 whereas large represents the Social scores that are at least equal to 50. Thus “small” and “large” has nothing to do with the size of the sample but the magnitude of the observations in each sub-sample. Small represents the Social scores that are less than 50 whereas large represents the Social scores that are at least equal to 50. Even with the pseudo quadratic model, the null hypothesis is not rejected. The sign for the sample of small (high) Social score is negative (positive) which corroborates my suspicion mentioned above that the relationship between Social CSR and firm performance is U-shaped. The R-squareds may still be low but they are higher than the previous two models. Unfortunately, this model also suffers from multi-collinearity as shown by the VIF values which are all greater than 10. In other words, this model has slightly better explanatory power but is not reliable due to the multi-collinearity issue.

(13)

Page 13 of 31 Table 6 shows the results of the pseudo quadratic model without logTR. Small (large) in Column 1 and Column 2 is an indicator for the sample of low (high) Social scores for the dependent variable RoA and RoE. Small represents the Social scores that are less than 50 whereas large represents the Social scores that are at least equal to 50. This pseudo quadratic model is a robustness test of the pseudo quadratic model with logTR (refer to Table 4) which also fails to reject the null hypothesis of there being no relationship between Social CSR and firm performance. The R-squareds from this model are generally better than the ones in Table 4 with the exception of the sample of small Social scores in Column 1 so one can say that the explanatory power of this model is slightly better. Moreover, the VIF values are below 10 meaning that this model is much more reliable compared to the pseudo quadratic model with logTR (Table 4).

Lastly, Table 7 is a robustness test of the robust pseudo quadratic model with logTR (refer to Table 5). As in Table 6, the small (large) sample in Column 1 and Column 2 is the sample of low (high) Social scores for the dependent variable RoA and RoE. The difference between Table 6 and Table 7 is that Social scores in this model are split at a cut-off point of 43.55 whereas in Table 6 the cut-off point is at 50. Ultimately, the null hypothesis is not rejected. The signs of almost all variables are consistent with those found in Table 6 with the exception of the market-to-book ratio coefficient in the small sample of Column 1. Table 7 has slightly higher adjusted R-squareds than Table 6 with the exception of the large column in Column 1. The VIF values are also below 10, meaning that this model is more reliable than the robust pseudo quadratic model with logTR (Table 5). In general, this model is better than the rest of the models presented based on reliability and explanatory power.

According to Fama and French (1993), firms characterized as small and/or has a high book-to-market ratio tends to outperform their counterparts. My proxies for those two factors are logTR and MtB, respectively. MtB is the inverse of the book-to-market ratio thus I would expect a negative relationship between MtB and the independent variables. logTR should also exhibit a negative relationship. Clearly, my expectations are not met at all for logTR (positive in Tables 1, 2, 3 and in the small columns for Table 4 and 5) whereas MtB has inconsistent signs for all model specifications.

Based on financial theory, the signs for the leverage ratio for RoE should be positive10 but negative for RoA. The sign of the leverage ratio is correct in all situations where RoA is

10

(14)

Page 14 of 31 the dependent variable. In the case where RoE is the dependent variable, the sign of the leverage ratio is never correct.

According to the correlation matrix in Table A1 in Appendix A, the correlation between marginal value of cash and the financial measures are is low and negative. On the contrary, we see that the signs of marginal value of cash are inconsistent. I believe this inconsistency has to do with the way the dependent variables are defined relative to marginal value of cash. Based on the way RoA is defined, marginal value of cash should have a negative relationship with it11. As for RoE, it does not explicitly account for cash effects but it resembles marginal value of cash to some extent thus the relationship should be positive. Tables 2 and 3 coincide with my line of reasoning but Tables 4 to 7 refute my line of reasoning. For Tables 5 to 7, it appears in general that firms with small Social have positive marginal value of cash relative to their counterparts. Perhaps this sample of firms with small Social have better corporate governance mechanisms thereby they have access to cheaper financing relative to their counterparts thus allowing them to have better firm performance (Zhang et al., 2015).Table 4 seems to be an exceptional case, in which the sign for marginal value of negative only for large Social scores in Column 1.

6. Conclusion

I investigated the impact of Social CSR on firm performance using a cross section fixed effects and time fixed effects panel regressions method on a panel data of US firms for the period 2010-2017. Regardless of the model I used to analyze the relationship, the null hypothesis of no relationship was not rejected even once. Contrary to Busch and Friede (2018), my findings are insignificant which leads me to conclude that there is no relationship between Social CSR and firm performance nowadays. A closer inspection of the relationship between Social CSR and firm performance reveals that low (high) Social scores correlate with negatively (positively) with firm performance. This implies that both the social impact hypothesis and trade-off hypothesis are present but my study suggests that they cancel each other out.

11_{The cash term is in the denominator of RoA but in the numerator of marginal value of cash. The equity term}

(15)

Page 15 of 31 There are a few limitations present in my study. My baseline model is not robust to different specifications which is not surprising due to the predominantly and extremely low explanatory power of my models. VIF tests conducted on each model show that there is varying degrees of multi-collinearity present. In my study, I do not address the issue of endogeneity but it is present (Busch and Friede, 2018) which could be the reason why the relationship between Social CSR and firm performance is insignificant. I also did not address the issue of limited data in the sense that I focus on only one country (Bajic and Yurtoglu, 2016).

(16)

Page 16 of 31 Table 1: Descriptive statistics

This table shows the summary statistics for the variables used in this study. RoA and RoE were winsorized at the 1% and 99% tails and are expressed in percentage. All monetary variables are in U.S. Dollars. logTR is the natural log of Total Revenues. Marginal value of cash is defined as

L is total debt divided by market value of equity. MtB is the market-to-book ratio. Social is the Social CSR measure. Social_sq is

the square of the Social score. Social_S is a dummy variable that takes the value of 1 when Social score is less than 50 but takes the value of 0 otherwise.

Firm Year Obs. Min Median Max Mean Std. Dev.

RoA 11992 0 0.1272 99.9900 4.7654 8.8186

RoE 10994 0 1.8442 99.9900 9.4955 16.3312

logTR 11489 4.7875 18.9699 22.4292 18.6459 2.5463

Marginal value of cash 3199 -9.5528 0.0009 8.1625 0.0272 0.4185

L 12625 0 0.1060 317.0482 0.4544 5.6697

MtB 5033 -836.8625 3.0182 758.6584 3.8611 36.8543

Social 2886 4.1692 42.9357 98.7740 45.3861 19.5249

Social _sq 2886 17.3824 1843.4710 9756.2970 2440.9870 2016.7560

(17)

Page 17 of 31 Table 2: Linear cross section fixed effects and time fixed effects model

RoA is the dependent variable in column 1.RoE is the dependent return in Column 2. All monetary variables are in U.S. Dollars. Social is the Social CSR measure. logTR is the natural log of Total Revenues. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name suggests, the

maximum VIF value from the VIF test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

Variable Column 1 Column 2

Social score 0.0086 (0.0283) 0.0549 (0.0594) logTR 0.8925 (0.7982) 1.0790 (1.1173) L -0.0009*** (0.0003) -0.0096** (0.0045) MtB 0.0003 (0.0023) -0.0056 (0.0105)

Marginal value of cash -0.4473

(18)

Page 18 of 31 Table 3: Quadratic cross section fixed effects and time fixed effects model

The dependent variable in columns 1 and 2 is RoA and RoE, respectively. All monetary variables are in U.S. Dollars. Social is the Social CSR measure. Social_sq is the square of the Social score. logTR is the natural log of Total Revenues. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name

suggests, the maximum VIF value from the VIF test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

Social -0.0821 (0.0871) -0.1925 (0.2039) Social_sq 0.0009 (0.0010) 0.0025 (0.0023) logTR 0.8917 (0.8031) 1.0744 (1.1308) L -0.0009*** (0.0003) -0.0097** (0.0045) MtB 0.0000 (0.0024) -0.0058 (0.0.04)

Marginal value of cash -0.3773

(19)

Page 19 of 31 Table 4: Pseudo quadratic model

The dependent variable in columns 1 and 2 is RoA and RoE, respectively. The sample is split into two sub samples based on whether the value of Social score is small (less than 50) or large (greater than or equal to 50). All monetary variables are in U.S. Dollars. Social is the Social CSR measure. logTR is the natural log of Total Revenues. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name suggests,

the maximum VIF value from the VIF test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

(small) (large) (small) (large)

(20)

Page 20 of 31 Table 5: Robust pseudo quadratic model

The dependent variable in columns 1 and 2 is RoA and RoE, respectively. The sample is split into two sub samples based on whether the value of Social score is small (less than 43.55) or large (greater than or equal to 43.55). All monetary variables are in U.S. Dollars. Social is the Social CSR measure. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name suggests, the maximum VIF value from the VIF

test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

(21)

Page 21 of 31 Table 6: Pseudo quadratic model (without logTR)

This table presents the robustness test results of the pseudo quadratic model in Table A8 in Appendix A. In this model, the firm size variable, logTR, is dropped. The dependent variable in columns 1 and 2 is RoA and RoE, respectively. The sample is split into two sub samples based on whether the value of Social score is small (less than 50) or large (greater than or equal to 50). All monetary variables are in U.S. Dollars. Social is the Social CSR measure. logTR is the natural log of Total Revenues. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name suggests, the maximum VIF value from the VIF

test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

(22)

Page 22 of 31 Table 7: Robust pseudo quadratic model (without logTR)

This table presents the robustness test results of the robust pseudo quadratic model in Table A9 in

Appendix A. In this model, the firm size variable, logTR, is dropped. The dependent variable in columns 1 and 2 is RoA and RoE, respectively. The sample is split into two sub samples based on whether the value of Social score is small (less than 43.55) or large (greater than or equal to 43.55). All monetary variables are in U.S. Dollars. Social is the Social CSR measure. L is total debt divided by market value of equity. MtB is the market-to-book ratio. Marginal value of cash is defined as

Maximum VIF is, as the name suggests, the maximum VIF value from the VIF test (excluding the VIF value of the Social score variable). The “a)” indicates that the VIF value for Social is above 10. Standard errors are reported in parentheses. P-values smaller than 0.01, 0.05, and 0.10 are indicated by ***, **, and *, respectively.

(23)

Page 23 of 31 References

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

Amini, C., & Dal Bianco, S. (2017). Corporate social responsibility and Latin American firm performance. Corporate Governance: The international journal of business in society, 17(3), 403-445.

Aras, G., Aybars, A., & Kutlu, O. (2010). Managing corporate performance: Investigating the relationship between corporate social responsibility and financial performance in emerging markets. International Journal of productivity and Performance management, 59(3), 229-254.

Bajic, S., & Yurtoglu, B. B. (2016). CSR, Market Value, and Profitability: International Evidence. Available at SSRN: https://ssrn.com/abstract=2848099 or

http://dx.doi.org/10.2139/ssrn.2848099

Barnett, M. L., & Salomon, R. M. (2012). Does it pay to be really good? Addressing the shape of the relationship between social and financial performance. Strategic Management Journal, 33(11), 1304-1320.

Bauer, R., Guenster, N., & Otten, R. (2004). Empirical evidence on corporate governance in Europe: The effect on stock returns, firm value and performance. Journal of Asset

management, 5(2), 91-104.

Becchetti, L., & Ciciretti, R. (2009). Corporate social responsibility and stock market performance. Applied Financial Economics, 19(16), 1283-1293.

doi:10.1080/09603100802584854

Berk, J. B., DeMarzo, P. M., & Harford, J. V. (2012). Fundamentals of corporate finance: Global edition. Boston: Pearson.

Bhagat, S., & Bolton, B. (2008). Corporate governance and firm performance. Journal of corporate finance, 14(3), 257-273.

(24)

Page 24 of 31 Carini, C., Comincioli, N., Poddi, L., & Vergalli, S. (2017). Measure the performance with the market value added: Evidence from CSR companies. Sustainability, 9(12), 2171. Carter, C. R., Kale, R., & Grimm, C. M. (2000). Environmental purchasing and firm performance: an empirical investigation. Transportation Research Part E: Logistics and Transportation Review, 36(3), 219-228.

Choi, J. S., Kwak, Y. M., & Choe, C. (2010). Corporate social responsibility and corporate financial performance: Evidence from Korea. Australian journal of management, 35(3), 291-311.

Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of financial economics, 33(1), 3-56.

Faulkender, M., & Wang, R. (2006). Corporate financial policy and the value of cash. The Journal of Finance, 61(4), 1957-1990.

Freeman, R. E. (1984). Strategic Management: A Stakeholder Approach. Boston: Pitman. Friedman, M. (1970). The social responsibility of the corporation is to increase its profits. New York Times Magazine, 13.

Google. (2008). 2008 annual report of Google. Retrieved from

http://www.annualreports.com/HostedData/AnnualReportArchive/g/NASDAQ_GOOG_2008 .pdf

Johansson, A. (2018). Corporate social responsibility can actually be a competitive advantage, so where’s your CSR program? Retrieved from

https://www.entrepreneur.com/article/311648

Klassen, R. D., & McLaughlin, C. P. (1996). The impact of environmental management on firm performance. Management science, 42(8), 1199-1214.

Microsoft. (2017). 2017 annual report of Microsoft. Retrieved from

(25)

Page 25 of 31 Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. The American economic review, 48(3), 261-297.

Mollet, J., & Ziegler, A. (2014). Socially responsible investing and stock performance: New empirical evidence for the US and European stock markets. Review of Financial Economics, 23(4), 208-216. doi:10.1016/j.rfe.2014.08.003

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

Preston, L. E., & O'bannon, D. P. (1997). The corporate social-financial performance relationship: A typology and analysis. Business & Society, 36(4), 419-429.

Scholtens, B. (2008). A note on the interaction between corporate social responsibility and financial performance. Ecological economics, 68(1-2), 46-55.

Schwartz, M. S., & Saiia, D. (2012). Should Firms Go “Beyond Profits”? Milton Friedman versus Broad CSR 1. Business and Society Review, 117(1), 1-31.

(26)

Page 26 of 31 Appendix A: Tables

Table A1: Correlation Matrix

This table shows the correlation matrix between all the variables. RoA and RoE are the dependent variables. All monetary variables are in U.S. Dollars. logTR is the natural log of Total Revenues. Marginal value of cash is defined as

L is total debt divided by market value of equity. MtB is the

market-to-book ratio. Social is the Social CSR measure. Social_sq is the square of the Social score. Social_S is a dummy variable that takes the value of 1 when Social score is less than 50 but takes the value of 0 otherwise.

RoA RoE Social Social_q Social_S logTR L MtB

(27)

Page 27 of 31 Table A2: VIF test of the linear model in Table 2

The following table is the output from performing the VIF test and corresponds to the columns in Table 2. logTR is the natural log of Total Revenues. Marginal value of cash is defined as

L is total debt divided by market value of equity. MtB is the market-to-book

ratio. Social is the Social CSR measure.

Social 6.7800 6.9500

logTR 29.4000 29.1300

L 1.1900 1.3900

MtB 1.0100 1.0200

Marginal value of cash 1.0100 1.0000

Table A3: VIF test for Quadratic model in Table 3

ratio. Social is the Social CSR measure. Social_sq is the square of the Social score.

Social 95.5000 96.5700

Social_sq 28.7500 36.4300

logTR 38.3700 38.5900

L 1.1900 1.4000

MtB 1.0100 1.0200

(28)

Page 28 of 31 Table A4: VIF test of the pseudo quadratic model in Table 4

ratio. Social is the Social CSR measure. The sample is split into two sub samples based on whether the value of Social score is small (less than 50) or large (greater than or equal to 50).

Variable Column 1 Column 2

Social 10.8000 31.0100 10.8700 31.2200

logTR 31.5300 47.6500 30.8700 47.5200

L 1.4100 1.2300 1.4100 1.3600

MtB 1.0100 1.0300 1.0200 1.0700

Marginal value of cash 1.0000 1.1600 1.0000 1.0300

Table A5: VIF test of the robust pseudo quadratic model in Table 5

ratio. Social is the Social CSR measure. The sample is split into two sub samples based on whether the value of Social score is small (less than 43.55) or large (greater than or equal to 43.55).

Social 12.0300 24.2100 12.1100 24.2500

logTR 32.1500 44.0000 31.4100 43.6500

L 1.4300 1.2000 1.4300 1.3500

MtB 1.0100 1.0200 1.0300 1.0200

(29)

Page 29 of 31 Table A6: VIF test of the pseudo quadratic model in Table 6

The following table is the output from performing the VIF test and corresponds to the columns in Table 4. Marginal value of cash is defined as

L is total debt divided by market value of

equity. MtB is the market-to-book ratio. Social is the Social CSR measure. The sample is split into two sub samples based on whether the value of Social score is small (less than 50) or large (greater than or equal to 50).

Social 8.4500 14.8000 8.3600 14.6200

L 1.3300 1.1700 1.3200 1.3200

MtB 1.0100 1.0300 1.0200 1.0700

Marginal value of cash 1.0000 1.1200 1.0000 1.0200

Table A7: VIF test of the robust pseudo quadratic model in Table 7

The following table is the output from performing the VIF test and corresponds to the columns in Table 5. Marginal value of cash is defined as

L is total debt divided by market value of

equity. MtB is the market-to-book ratio. Social is the Social CSR measure. The sample is split into two sub samples based on whether the value of Social score is small (less than 43.55) or large (greater than or equal to 43.55).

Social 8.8900 13.2400 8.7800 13.0800

L 1.3400 1.1500 1.3400 1.3100

MtB 1.0100 1.0200 1.0300 1.0200

(30)

Page 30 of 31 Table A8: Descriptive statistics (cut-off=43.55)

This table shows the summary statistics for the variables used in this study. RoA and RoE were winsorized at the 1% and 99% tails and are expressed in percentage. All monetary variables are in U.S. Dollars. logTR is the natural log of Total Revenues. Marginal value of cash is defined as

L is total debt divided by market value of equity. MtB is the market-to-book ratio. Social is the Social CSR measure. Social_sq is

the square of the Social score. Social_S is a dummy variable that takes the value of 1 when Social score is less than 43.55 but takes the value of 0 otherwise.

Firm Year Obs. Min Median Max Mean Std. Dev.

RoA 11992 0 0.1272 99.9900 4.7654 8.8186

RoE 10994 0 1.8442 99.9900 9.4955 16.3312

logTR 11489 4.7875 18.9699 22.4292 18.6459 2.5463

Marginal value of cash 3199 -9.5528 0.0009 8.1625 0.0272 0.4185

L 12625 0 0.1060 317.0482 0.4544 5.6697

MtB 5033 -836.8625 3.0182 758.6584 3.8611 36.8543

Social 2886 4.1692 42.9357 98.7740 45.3861 19.5249

Social _sq 2886 17.3824 1843.4710 9756.2970 2440.9870 2016.7560

(31)

Page 31 of 31 Appendix B: Derivations

According to Berk et al. (2012), RoA and RoE can be defined as follows:

(equation B1) (equation B2)

where NI is net income, TA is total assets, S is sales, E is equity. Let D denote debt and rewrite the above to the following:

(equation B1.1) (equation B2.1)

Appendix C: The composition of the Social score

According to the Thomsen Reuters ESG score manual12, the Social score is a compilation of four other scores/categories: Workforce, Human rights, Community and Product

responsibility. The following table summarizes what each category measures.

Category What it measures

Workforce  Job satisfaction

 Healthy and safe workplace

 Maintain diversity and equal opportunities  Development opportunities for employees

Human rights  Respect of fundamental human rights conventions Community  Good citizen

 Protecting public health  Respect business ethics

Product responsibility  Capacity to produce quality goods and services  Product should account for customer’s health, safety,

integrity and data privacy

12

Downloaded on 6 January 2019 via