Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Corporate Social Reporting

694 views

Published on

Published in: Business, Economy & Finance
  • Be the first to comment

  • Be the first to like this

Corporate Social Reporting

  1. 1. Rev Quant Finan Acc DOI 10.1007/s11156-008-0090-y ORIGINAL RESEARCH Corporate social responsibility and financial performance: the ‘‘virtuous circle’’ revisited Edward Nelling Æ Elizabeth Webb Ó Springer Science+Business Media, LLC 2008 Abstract We examine the causal relation between corporate social responsibility (CSR) and financial performance. Consistent with past studies, we find that the two variables appear to be related when we use traditional statistical techniques. However, using a time series fixed effects approach, we find that the relation between CSR and financial per- formance is much weaker than previously thought. We also find little evidence of causality between financial performance and narrower measures of social performance that focus on stakeholder management. Our results suggest that strong stock market performance leads to greater firm investment in aspects of CSR devoted to employee relations, but that CSR activities do not affect financial performance. We conclude that CSR is driven more by unobservable firm characteristics than by financial performance. Corporate governance Á Corporate culture Á Social responsibility Keywords JEL Classification G34 Á M14 1 Introduction The area of corporate social responsibility (CSR) has become a notable topic both in the business and academic press. McWilliams and Siegel (2001) define CSR as ‘‘actions that appear to further some social good, beyond the interests of the firm and that which is required by law,’’ (p. 17). Examples of CSR include support of local businesses or char- ities, developing recycling programs, and promoting minority employment. Though these activities certainly result in societal benefits, opinions differ as to whether a firm’s CSR activity enhances financial performance. For instance, CSR has been described as being E. Nelling (&) LeBow College of Business, Drexel University, Philadelphia, PA 19104, USA e-mail: nelling@drexel.edu E. Webb School of Business, La Salle University, Philadelphia, PA 19141, USA 123
  2. 2. E. Nelling, E. Webb ‘‘little more than a cosmetic treatment,’’ (The Good Company 2005) while an earlier article in the financial press touted the economic advantages for companies practicing CSR (Fittipaldi 2004). CSR has attracted the attention of mutual fund investors, including TIAA CREF, and a number of funds use CSR as a screening device for investment selection. However, any benefits investors derive from including socially responsible firms in their portfolios do not appear to be in form of superior returns. Bello (2005) matches socially responsible mutual funds to conventional funds in the late 1990s and finds no significant difference in long-run investment performance or portfolio diversification between the two groups. According to Statman (2000), the Domini Social Index (DSI) did as well as the S&P 500 between 1990 and 1998. He finds that some socially responsible funds performed worse than the S&P 500 and the DSI, but fared no worse than conventional mutual funds during the time period. Academic research has examined the causal relation between CSR and financial performance—what is sometimes referred to as the ‘‘virtuous circle’’—to determine if ‘‘doing good’’ socially leads to ‘‘doing well’’ financially, and whether firms exhibiting superior financial performance devote more resources to social activities. Why should a ‘‘virtuous circle’’ exist? Waddock and Graves (1997) suggest that positive stakeholder relationships can reduce the likelihood of difficulty when dealing with groups such as employees, customers, and the community. In addition, good social performance and good managerial practice may be related, so this in turn may lead to strong financial performance. For instance, Cheng et al. (2006) find that firms with strong shareholder rights tend to have a lower cost of equity capital than competing firms, which supports the idea that reducing agency problems between stakeholders and management improves financial performance. Ferreira et al. (2008) find that large firms benefit in the form of positive long-run stock performance following certification of quality management. This suggests that in the case of large companies, good managerial practice does indeed improve the bottom line. The empirical evidence suggests that a virtuous circle exists. Waddock and Graves (1997) and Hillman and Keim (2001) find that increased CSR leads to enhanced financial performance and vice versa. McGuire et al. (1988) find that prior year’s stock returns and accounting-based performance measures are related to current measures of CSR, but that a past record of good social performance does not affect the current financial performance of a firm. Despite the extensive body of research on CSR and financial performance, we believe that additional analysis is warranted. In this study, we use alternative statistical techniques to provide new insights on the causal relationship between firm performance and CSR. The studies cited above and others on the topic have typically employed ordinary least squares (OLS) regressions with 1-year lagged values of one variable (e.g., CSR) used as deter- minants of the other (e.g., ROA). Indeed, we also find that CSR and performance are related when we use standard OLS regression models. However, in a time series fixed effects approach with over 2,800 firm-year observations, we find that the relation between CSR and financial performance is much weaker than that documented in past studies. We also find little evidence of causality between financial performance and narrower measures of social performance. Overall, we find weak evidence that strong stock market perfor- mance leads to greater levels of CSR that focus specifically on employee relations. However, CSR activity does not lead to changes in a firm’s financial performance. This study contributes to the literature on CSR in several ways. We use a more com- prehensive time-series of CSR data relative to past studies. Also, previous research analyzing the link between CSR and financial performance does not account for possible 123
  3. 3. Corporate social responsibility and financial performance persistence or unobservable characteristics that may play a role in the level of CSR activity undertaken by a firm. We examine these important considerations by using Granger causality and controlling for firm fixed-effects in our time series. We also use a Tobit specification for models of CSR causality to account for the censored nature of the data. Finally, our specifications also include analysis of narrow measures of social performance such as how the company treats employees, the environment, the community, and diversity issues. The use of appropriate statistical methods and examination of the components of CSR yield added insights regarding the relation between CSR and financial performance. 2 Data and research methods To measure CSR, we use the KLD Socrates Database.1 This database rates companies on the various degrees of social responsibility including community participation, diversity, employee interests, environmental considerations, and shareholder interests. This index is superior to alternative measures of CSR because it is compiled by an independent rating service that focuses on a wide range of firms over a broad spectrum of CSR screens, whereas alternative measures of CSR focus on a small sample of firms or use a narrow CSR screen. Each screen in the KLD database is summarized in terms of strengths (positive values) and concerns (negative values). For instance, the community screen is calculated on the basis of strengths, such as generous giving or support for education in the area, and concerns, such as investment controversies or negative economic impact on the community. We calculate a weighted average of the overall index to represent each firm’s aggregate level of CSR, using the weights developed by Waddock and Graves (1997). We use these weights to facilitate comparison of results, and to highlight the effect of a different esti- mation methodology on statistical inferences. Waddock and Graves calculate weights using expert panel evaluations on the relative importance of each CSR attribute. Weigh- tings are as follows: Employee relations: 0.168; Product: 0.154; Community relations: 0.148; Environment: 0.142; Treatment of women and minorities: 0.136; Nuclear power: 0.089; Military contracts: 0.086; South Africa: 0.076. Our sample extends from 1993 to 2000 and includes data on more than 600 U.S. firms. We assess financial performance using both accounting (return on assets, or ROA) and market-based (common stock returns) measures from Compustat. Table 1 presents summary statistics for our measures of CSR and financial performance and control variables. The variable SCORE, which is the sum of the individual scores for the various CSR strengths and concerns, ranges from -11 (indicating significant concerns) to +11 (indicating significant strengths), with a median value of zero. The minimum WSCORE, which represents the weighted average measure of CSR for each firm, is -1.07 with a maximum value of +1.74. We then create narrower CSR scores by summing the scores in four categories of the KLD index: community, diversity, employee relations, and environment. The median value for each is zero. The lowest minimum score is for total environment score at -5 while the highest maximum score is diversity at 7. Panel B presents data on the values of SCORE by industry. Not surprisingly, significantly negative averages appear in the tobacco industry (average = -2.08), the mining industry (aver- age = -1.54), and the transportation industry (average = -1.43). The largest positive average (1.02) exists in the retail trade industry, possibly because these firms stand to 1 Information about this database is available online at http://www.kld.com/research/socrates/index.html. 123
  4. 4. E. Nelling, E. Webb Table 1 Descriptive statistics and CSR score by industry Mean Median Min Max N Panel A CSR SCORE 0.24 0.0 -11.0 11.0 3943 W_SCORE 0.11 0.12 -1.07 1.74 3940 Return on assets (%) 5.75 4.99 -97.17 57.47 3197 Annual stock return (%) 16.77 11.95 -95.75 503.08 3536 Environmental score -0.13 0.0 -5.0 3.0 3938 Community score 0.35 0.0 -2.0 4.0 3943 Employee score 0.27 0.0 -3.0 4.0 3943 Diversity score 0.49 0.0 -2.0 7.0 3943 Panel B—CSR SCORE by industry Mining -1.54 -1.0 -11.0 3.0 129 Construction -0.67 0.0 -6.0 2.0 12 Tobacco products -2.08 -2.0 -5.0 0.0 24 Textile mill products 0.40 0.0 -3.0 5.0 75 Lumber and wood products 0.04 0.0 -7.0 8.0 169 Chemical and rubber products -0.30 0.0 -8.0 9.0 373 Primary metal industries -0.23 0.0 -6.0 5.0 140 Industrial machinery & equipment 0.67 0.0 -8.0 11.0 477 Transportation equipment -1.43 -1.0 -9.0 6.0 143 Instrumentation products 0.49 1.0 -6.0 5.0 138 Communications 0.39 0.0 -5.0 6.0 69 Wholesale trade -0.57 0.0 -6.0 3.0 83 Retail trade 1.02 1.0 -5.0 7.0 322 Business services 0.50 0.0 -9.0 9.0 147 Other 0.73 1.0 -10.0 9.0 1223 This table presents data on corporate social responsibility (from the KLD Socrates Database) and financial performance over the period 1993–2000. Panel A represents the whole sample, and Panel B provides statistics on the social responsibility by industry. CSR SCORE is the sum of the component scores in the KLD Socrates Database, and WSCORE is the weighted average benefit the most from the positive association their customers may attribute to the firm’s CSR activity. In Table 2, we report the correlation matrix for the sample of firms used in this study. In particular, we are interested in the correlations between WSCORE and measures of financial performance. In the first column of Table 2, we see that WSCORE is positively correlated with contemporaneous ROA and annual stock returns and negatively correlated with leverage (long-term debt divided by total assets). There is no significant correlation between WSCORE and firm size, measured by total assets or sales. This simple correlation analysis does not address issues of causality, but provides initial evidence that CSR and financial performance are directly related. We use traditional regression analysis in our initial examination of the causal relationship between CSR and financial performance. We then argue that ordinary least-squares regressions cannot adequately assess causality issues, and introduce Granger causality models and Tobit regressions to examine the relationship in greater detail. Furthermore, using a panel dataset of a large cross-section of firms over a 123
  5. 5. Corporate social responsibility and financial performance Table 2 Correlations between corporate social responsibility and firm characteristics WSCORE Return on Stock return Sales Assets Financial assets leverage WSCORE 1.000 Return on 0.066 (0.01) 1.000 assets Stock return 0.050 (0.01) 0.201 (0.01) 1.000 Sales -0.012 (0.47) -0.047 (0.01) 0.070 (0.01) 1.000 Assets 0.016 (0.33) -0.200 (0.01) 0.081 (0.01) 0.857 (0.01) 1.000 Financial -0.087 (0.01) -0.293 (0.01) -0.098 (0.01) 0.135 (0.01) 0.178 (0.01) 1.000 leverage This table presents correlation coefficients between a measure of corporate social responsibility and other firm characteristics. WSCORE is the weighted social responsibility score from the KLD Socrates Database. Sales and Assets are the log of total sales and total assets. Financial leverage is long-term debt divided by total assets. P-values are in parentheses number of years requires a fixed effects regression approach to account for correlation in the error term. We explain the details of these statistical approaches in the following section. 3 Results 3.1 Results of OLS models Panel A of Table 3 presents results of OLS regression models similar to those in Waddock and Graves (1997) with the weighted social responsibility score in year t as the dependent variable. The primary independent variable is financial performance in year t - 1, measured either by return on assets or the return on the firm’s common stock. Sales, total assets, and debt-to-assets in year t - 1 are included as controls for firm size and leverage. In general, we find that the relationship between social responsibility and lagged financial performance is positive and significant, similar to what is reported by Waddock and Graves (1997). Leverage is negative and significant in each model, sug- gesting that firms with a higher proportion of debt financing exhibit lower levels of CSR. The coefficients on lagged ROA and lagged stock returns (in Panel A) are all positive and significant at the 1% level. Coefficients on the size control variables (sales and total assets) are positive, suggesting that larger firms may have greater resources available to support CSR efforts. Panel B of Table 3 presents the results of models with financial performance—ROA and annual stock returns—as the dependent variable. In each of the models (5)–(8) in Table 3, the coefficient on the lagged weighted CSR score is positive and significant. This result, consistent with that of Waddock and Graves (1997), implies that a firm’s level of CSR activity contributes to its financial performance. In summary, the OLS regression results presented in Table 1 suggest that a ‘‘virtuous circle’’ exists, linking a firm’s level of CSR with its financial performance. In the sections below, we introduce alternative estimation methods that may be more appropriate than OLS for the analysis. 123
  6. 6. E. Nelling, E. Webb Table 3 OLS regressions of corporate social responsibility and financial performance Panel A WSCORE is dependent variable Model (1) (2) (3) (4) INTERCEPT 0.073 (1.62) -0.028 (-0.71) 0.104 (2.37)** 0.029 (0.79) ROA(t - 1) 0.0035 (3.13)*** 0.004 (3.83)*** RET(t - 1) 0.001 (3.74)*** 0.001 (3.56)*** LEV(t - 1) -0.157 (-2.76)*** -0.187 (-3.32)*** -0.184 (-3.35)*** -0.221 (-4.03)*** SALES(t - 1) 0.007 (1.42) 0.005 (1.02) ASSETS(t - 1) 0.019 (4.36)*** 0.015 (3.39)*** N 2731 2789 2727 2786 Adjusted R2 0.008 0.015 0.010 0.014 F 8.47*** 14.94*** 9.88*** 14.27*** Panel B ROA is dependent variable RET is dependent variable Model (5) (6) (7) (8) INTERCEPT 8.755 (11.15)*** 13.380 (20.64)*** 4.262 (0.90) 2.693 (0.68) WSCORE(t - 1) 0.755 (2.08)** 0.852 (2.43)** 7.193 (3.33)*** 6.860 (3.23)*** LEV(t - 1) -12.274 (-12.51)*** -10.966 (-11.42)*** -20.514 (-3.53)*** -22.533 (-3.91)*** SALES(t - 1) -0.081 (-0.85) 1.9450 (3.39)*** ASSETS(t - 1) -0.669 (-8.65)*** 2.103 (4.51)*** N 2558 2615 2869 2932 Adjusted R2 0.061 0.087 0.011 0.014 F 56.41*** 84.22*** 11.57*** 14.86*** WSCORE is the weighted social responsibility score from the KLD Socrates Database. ROA is return on assets and RET represents annual common stock return. LEV is total debt/total assets. SALES and ASSETS are the log of total sales and total assets. Numbers in parentheses are t-statistics. *, **, *** indicates significance at the .10, .05, .01 levels, respectively 3.2 Results of fixed effects models Previous studies of the relationship between financial performance and social responsibility have failed to control for unobservable characteristics in panel data that differ between firms but are constant over time. For example, variables such as corporate culture or managerial influence may be significant determinants of CSR activity. Using a fixed effects regression model allows us to estimate the effects of the independent variables on the dependent variable while controlling for the effect of unobserved variables. The results of our fixed effects models are presented in Table 4 and they reveal interesting differences compared to the OLS regression results presented in Table 3. A firm’s past financial performance as measured by ROA continues to be a significant determinant of CSR score—i.e., ROA and CSR are positively related. However, as noted in Models (3) and (4), a firm’s past stock return is no longer a significant determinant of its CSR score. Furthermore, the weighted CSR score is no longer a significant predictor of financial performance, as seen in Models (5)–(8) in Panel B of Table 4. These results suggest that by controlling for firm fixed effects, the relationship between CSR and financial performance is not as strong as previously reported in the literature. 123
  7. 7. Corporate social responsibility and financial performance Table 4 Fixed effects regressions of corporate social responsibility and financial performance Panel A WSCORE is dependent variable Model (1) (2) (3) (4) INTERCEPT -0.439 (-4.31)*** -0.387 (-3.91)*** -0.428 (-4.19)*** -0.357 (-3.64)*** ROA(t - 1) 0.002 (2.57)*** 0.002 (2.88)*** RET(t - 1) 0.001 (0.06) 0.001 (0.10) LEV(t - 1) 0.047 (0.76) 0.028 (0.45) 0.014 (0.23) -0.010 (-0.16) SALES(t - 1) 0.067 (5.35)*** 0.068 (5.41)*** ASSETS(t - 1) 0.059 (4.92)*** 0.058 (4.85)*** N 2731 2789 2727 2786 Adjusted R2 0.001 0.005 0.001 0.004 F 12.16*** 10.79*** 9.88*** 8.02*** Panel B ROA is dependent variable RET is dependent variable Model (5) (6) (7) (8) INTERCEPT 14.320 (4.82)*** 24.680 (8.63)*** 109.752 (4.64)*** 116.675 (5.07)*** WSCORE(t - 1) -0.417 (-0.70) -0.004 (-0.01) 0.869 (0.18) 1.270 (0.26) LEV(t - 1) -2.332 (-1.38) -0.793 (-0.48) 0.736 (0.05) 6.480 (0.47) SALES(t - 1) -0.971 (-2.62)*** -11.451 (-3.90)*** ASSETS(t - 1) -2.238 (-6.38)*** -12.096 (-4.31)*** N 2558 2615 2869 2932 Adjusted R2 0.008 0.041 0.003 0.006 F 3.62** 14.94*** 5.19*** 6.32*** WSCORE is the weighted social responsibility score from the KLD Socrates Database. ROA is return on assets and RET represents a firm’s annual common stock return. LEV is total debt/total assets. SALES and ASSETS are the log of total sales and total assets. Numbers in parentheses are t-statistics. *, **, *** indicates significance at the .10, .05, .01 levels, respectively 3.3 Results of Granger causality models The regression results presented above in Table 3 suggest a link between CSR and financial performance. An interesting question arises as to the nature of causality, if any, between doing ‘‘good’’ through greater CSR activity and doing ‘‘well’’ financially. Does good financial performance cause a firm to become more socially responsible, or does a high level of CSR lead to subsequent superior financial performance? Good financial performance may result in more resources being available for the pursuit of CSR goals. A high level of CSR may attract the attention of investors or customers or motivate employees to perform better and reduce costs, boosting financial performance. Another possibility is that a ‘‘virtuous circle’’ exists, and the causality runs in both directions. We address the link between financial performance and CSR in the context of Granger causality. This approach, developed by Granger (1969), involves regression models of the form SCOREt ¼ a0 þ a1 SCOREtÀ1 þ a2 PERFt þ a3 PERFtÀ1 þ et PERFt ¼ b0 þ b1 PERFtÀ1 þ b2 SCOREt þ b3 SCOREtÀ1 þ et where SCOREt, and SCOREt-1 represent a firm’s CSR score in year t and t - 1, PERFt and PERFt-1 represent a firm’s financial performance in year t and t - 1, and et is an error term. 123
  8. 8. E. Nelling, E. Webb If the coefficients a2 or a3 are significantly different from zero, we conclude that financial performance ‘‘Granger causes’’ social responsibility. Similarly, if coefficients b2 or b3 are significant, we infer causality from CSR to financial performance. (For ease in exposition, we will omit the word ‘‘Granger’’ from subsequent discussion, reminding the reader that all uses of the words ‘‘cause’’ or ‘‘causality’’ refer specifically to ‘‘Granger causality.’’) Before discussing our results, a comment on model specification is in order. The typical problem that arises in OLS regression when lagged dependent variables are included as regressors is serial correlation in the error term, rendering the parameter estimates biased and inconsistent. We test for this using the method of Maddala (1971) and find that serial correlation is not a problem in our model. First, we present models of causality using ordinary least squares (without controlling for firm fixed effects). The results of these specifications are presented in Table 5, and differ widely from those reported in Table 3. The results in Panel A indicate a strong relationship between past CSR performance and current CSR, but ROA is not a significant determinant of CSR. The significance of the relationship between stock returns and CSR is weaker than reported in Table 3 (with a t-statistic of 1.93 versus 3.74). Results presented in Panel B indicate that firms do not ‘‘do well’’ by ‘‘doing good’’—CSR activity is not Table 5 Granger causality analysis of CSR and financial performance Panel A WSCORE is dependent variable Intercept 0.014 (2.94)*** -0.012 (-0.65) WSCORE(t - 1) 0.913 (102.99)*** 0.908 (107.52)*** ROA 0.001 (0.79) ROA(t - 1) 0.001 (0.91) RET 0.001 (1.40) RET(t - 1) 0.001 (1.93)* SALES(t - 1) 0.001 (0.82) 0.003 (1.47) N 2912 3218 Adjusted R2 0.786 0.784 F 2675.11*** 2922.92*** Panel B ROA is dependent variable RET is dependent variable INTERCEPT 2.831 (4.95)*** 5.662 (1.31) ROA(t - 1) 0.651 (45.94)*** RET(t - 1) 0.022 (1.19) WSCORE 0.463 (0.84) 5.896 (1.40) WSCORE(t - 1) -0.162 (-0.29) 1.086 (0.25) SALES(t - 1) -0.086 (-1.26) 1.282 (2.43)*** N 2912 3218 Adjusted R2 0.425 0.005 F 538.35*** 5.13*** WSCORE is the weighted social responsibility score from the KLD Socrates Database. ROA is return on assets and RET represents a firm’s annual common stock return. LEV is total debt/total assets. SALES is the log of total sales. Numbers in parentheses are t-statistics. *, **, *** indicates significance at the .10, .05, .01 levels, respectively 123
  9. 9. Corporate social responsibility and financial performance causally related to contemporaneous or subsequent financial performance. The explanatory power of these models with the CSR score (R2 = 0.78) or return on assets (R2 = 0.42) is significantly larger than that of the OLS regressions in Table 3. The increase is mainly due to the inclusion of lagged values of the dependent variable in each model. 3.4 Results of Granger causality models with fixed effects We next incorporate both fixed effects and Granger causality in our models using the CSR score and financial performance as dependent variables. Specifically, the model now includes n intercepts (where n is the number of firms in the panel dataset) that are rep- resented by a set of indicator variables, such that: SCOREt ¼ a0 þ a1 SCOREtÀ1 þ a2 PERFt þ a3 PERFtÀ1 þ ai þ et PERFt ¼ b0 þ b1 PERFtÀ1 þ b2 SCOREt þ b3 SCOREtÀ1 þ ai þ et where ai = b0 + b1Zi, and Zi represents the unobserved variable that varies from one firm to another, but does not change over time. The variables a1,…, an are treated as unknown intercepts to be estimated, one for each firm. In this model, the errors et are assumed to be uncorrelated over time, conditional on the independent variables. The results of the causality specifications using firm fixed effects are presented in Table 6, and corroborate those of our causality models presented earlier. In Panel A, CSR scores are highly persistent over time. We find a positive and significant relationship between CSR score and lagged return on assets, which differs from the causality model presented without fixed effects in Table 5. However, this causal relationship disappears when stock returns are used as the measure of financial performance. The results from the causality fixed effects models using financial performance as the dependent variable are presented in Panel B of Table 6. Both the lagged return on assets and lagged stock returns are significantly related to current return on assets and stock returns, respectively. More importantly, the coefficients on current and lagged WSCORE are not significant. This implies that, using a fixed effects causality model specification, CSR does not contribute to better financial performance. In additional robustness tests, we repeat the analyses in Panel A of Tables 5 and 6 using a Tobit specification when WSCORE is the dependent variable to account for its censored nature. The results, which are not reported here, are very similar to those reported above. 4 Narrowing the definition of social responsibility Harrison and Freeman (1999) note that research on CSR has over time focused more on specific measures of social responsibility, and less on the aggregate social performance measure. For instance, Hillman and Keim (2001) separate social performance into two components: stakeholder management and social issue participation. Primary stakeholders include shareholders, employees, and customers, and they find that this measure is directly related to shareholder value creation. However, social issue participation, which refers to the use of corporate resources for social issues outside firm strategy such as avoidance of nuclear energy, refraining from alcohol, tobacco, and gambling industries, etc., is nega- tively related to changes in shareholder wealth. Fisman et al. (2005) focus specifically on CSR that is targeted towards the community in a study on industry variability in CSR. 123
  10. 10. E. Nelling, E. Webb Table 6 Granger causality analysis of CSR and financial performance using fixed effects Panel A WSCORE is dependent variable INTERCEPT 0.036 (3.96)*** -0.173 (-2.21)** WSCORE(t - 1) 0.459 (25.66)*** 0.466 (27.47)*** ROA 0.001 (1.01) ROA(t - 1) 0.001 (2.16)** RET 0.001 (0.35) RET(t - 1) 0.001 (0.31) SALES(t - 1) 0.001 (3.31)*** 0.030 (3.07)*** N 2912 3218 Adjusted R2 0.742 0.745 F 183.01*** 204.40*** Panel B ROA is dependent variable RET is dependent variable INTERCEPT 0.036 (3.96)*** 105.046 (5.41)*** ROA(t - 1) 0.088 (4.31)*** RET(t - 1) -0.196 (-9.39)*** WSCORE 0.722 (1.24) -1.695 (-0.35) WSCORE(t - 1) -0.256 (-0.45) 3.517 (0.74) SALES(t - 1) -0.729 (-2.41)** -10.460 (-4.34)*** N 2912 3218 Adjusted R2 0.142 0.002 F 6.42*** 26.29*** This table reports coefficients and t-statistics (in parentheses) from Granger causality model specifications with panel fixed effects. WSCORE is the weighted social responsibility score from the KLD Socrates Database. ROA is return on assets and RET represents annual common stock return. LEV is total debt/total assets. SALES is the log of total sales. *, **, *** indicates significance at the .10, .05, .01 levels, respectively They justify the use of this measure as it is the most prominent element of CSR, and suggest that activities directed towards the community can be used as a means of differ- entiation by firms in competitive industries. They find that the community CSR measure is positively related to profitability in advertising-intensive industries. We test the relationship between financial performance and specific aspects of CSR using individual factors in the KLD database rather than the weighted average CSR score. The KLD database rates companies on eight different screens of social responsibility. Each screen focuses on ‘‘strengths’’ and ‘‘concerns’’ of a particular social indicator. We create separate ‘‘stakeholder management’’ scores for community, diversity, employee relations, and environment by adding up the number of strengths listed in each category and sub- tracting out the total number of weaknesses. By doing this, we can analyze the causal relationship between financial performance and specific indicators of CSR rather than the overall CSR score. The results of the causality models with this narrower definition of social performance are presented in Tables 7 and 8. Specifically, we analyze the following regression models substituting score indicators of community, diversity, employee relations, and environment for SRPERF: 123
  11. 11. Corporate social responsibility and financial performance Table 7 Granger causality between community and diversity score and stock returns Dependent variable COMM RET DIV RET INTERCEPT 0.036 (3.81)*** 16.959 (14.40)*** 0.133 (9.64)*** 17.590 (15.29)*** COMM 1.013 (0.55) COMM(t - 1) 0.849 (83.26)*** 1.869 (1.00) DIV 0.455 (0.37) DIV(t - 1) 0.908 (93.34)*** 0.422 (0.33) RET 0.001 (0.70) 0.001 (1.19) RET(t - 1) 0.001 (0.18) -0.072 (-3.91)*** 0.001 (1.05) -0.072 (-3.89)*** N 3155 3153 3155 3153 Wald Chi2 6960.16*** 19.66*** 8747.43*** 16.03*** This table reports the results of fixed effects causality models with Tobit specifications when the dependent variable is a firm’s community (COMM) or diversity (DIV) score. RET is the firm’s annual common stock return. T-statistics are in parentheses. *, **, *** indicates significance at the .10, .05, .01 levels, respectively SRPERFt ¼ a0 þ a1 SRPERFtÀ1 þ a2 SRPERFtÀ2 þ a3 RETt þ a4 RETtÀ1 þ ai þ et RETt ¼ b0 þ b1 RETtÀ1 þ b2 RETtÀ2 þ b3 SRPERFt þ b4 SRPERFtÀ1 þ ai þ et where RET represents the firm’s annual common stock returns. We use Tobit models when each CSR score is used as the dependent variable. Upper and lower limits on the censored variable are the maximum and minimum CSR score.2 We present fixed effects causality models for community (COMM) and diversity (DIV) scores in Table 7. Similar to the broader definition of CSR used earlier, we find strong persistence in both narrow measures of CSR. The coefficients on lagged community and diversity score are positive and significant. However, current and lagged annual stock returns are not significantly related to these CSR measures. Also, when stock returns are used as the dependent variable, the coefficients on current and lagged community and diversity scores are statistically insignificant, suggesting no evidence of causality running from these narrow measures of CSR to financial performance. In Table 8 we repeat the procedure using employee relations (EMP) and environment (ENV) score as the CSR measures. Again, we observe positive and significant coefficients on lagged CSR scores, indicating persistence in the employee and environmental dimen- sions of CSR. We also note that past financial performance is a positive and significant determinant of the employee relation score. That is, as stock returns increase, CSR, as measured by employee relations, also increases. Firms can increase their employee relation score by, for instance, having strong retirement benefits, employee involvement (in the form of widespread employee options or cash profit sharing), and having a strong record of employee safety and general well-being. In contrast, we find no evidence of causality running from stock returns to a firm’s environmental score. In the models with stock returns as the dependent variable, the coefficients on current and lagged employee relations scores are insignificant, and the coefficient on the current environmental score is significant 2 The upper and lower limits for the CSR scores are the following: community (-2, 4), diversity (-2, 7), employee relations (-3, 4), and environment (-5, 3). 123
  12. 12. E. Nelling, E. Webb Table 8 Granger causality between employee and environmental score and stock returns Dependent variable EMP RET ENV RET INTERCEPT 0.051 (4.30)*** 17.808 (15.96)*** -0.015 (-1.50) 18.267 (16.75)*** EMP -1.239 (-0.89) EMP(t - 1) 0.821 (70.71)*** 2.034 (1.42) ENV 2.860 (1.79)* ENV(t - 1) 0.882 (91.97)*** -1.500 (-0.92) RET -0.001 (-0.19) 0.001 (1.45) RET(t - 1) 0.001 (3.26)*** -0.070 (-3.80)*** 0.001 (0.09) -0.073 (-3.95)*** N 572 3153 3155 3150 Wald Chi2 5021.30*** 16.84*** 8483.39*** 19.15*** This table reports the results of fixed effects causality models with Tobit specifications when the dependent variable is a firm’s employee (EMP) or environment (ENV) score. RET is the firm’s annual common stock return. T-statistics are in parentheses. *, **, *** indicates significance at the .10, .05, .01 levels, respectively only at the ten percent level, suggesting little evidence of causality running from these measures of CSR to financial performance. 5 Conclusions A number of studies have examined the relationship between CSR and financial perfor- mance, and most have found that the two characteristics are directly related. In this paper, we analyze the existence and direction of causality between CSR and financial perfor- mance by using a more comprehensive data set and an improved statistical methodology. We use the KLD index as the measure of CSR, and both return on assets and stock returns to assess financial performance. First, we analyze causality using standard OLS regression analysis, similar to past research. Our results are consistent with the existing literature in that past financial performance is significant in explaining variability in CSR, and that CSR is significant in explaining variability in financial performance. However, when we introduce panel fixed effects into the model, the results no longer suggest a strong relationship between CSR and financial performance. We further examine the link between CSR and financial performance using Granger causality models. Results here again demonstrate the weakened evidence of a relationship between CSR and financial performance. However, there is some evidence that stock returns cause CSR performance. We then introduce both panel fixed effects and Tobit models in the causality model specification. We find no causality from CSR to financial performance, and only weak evidence of causality running from stock returns to a firm’s CSR score. Finally, we examine the relation between financial performance and stakeholder-specific measures of CSR. We find that higher stock returns lead to higher employee relations scores, but find no evidence of causality between stock returns and aspects of CSR related to the community, diversity, or environment. In summary, CSR and financial performance appear to be related when using traditional OLS regression models. However, this ‘‘virtuous circle’’ is found to be much weaker when examined using a fixed effects Granger causality approach. There is no evidence that CSR 123
  13. 13. Corporate social responsibility and financial performance affects a firm’s financial performance, and the only aspect of CSR driven by stock market performance is employee relations. If socially responsible activities provide benefits to the firm, they appear to manifest themselves in forms unrelated to financial performance. Acknowledgements We thank Eliezer Fich, Jacqueline Garner, Mukunthan Santhanakrishnan, and sem- inar participants at the 2006 Financial Management Association Annual Meeting for helpful comments. References Bello Z (2005) Socially responsible investing and portfolio diversification. J Financ Res 28:41–57 Cheng A, Collins D, Huang H (2006) Shareholder rights, financial disclosure and the cost of equity capital. Rev Quant Finance Account 27:175–204 Ferreira E, Sinha A, Varble D (2008) Long-run performance following quality management certification. Rev Quant Finance Account 30:93–109 Fisman R, Heal G, Nair V (2005) Doing well by doing good? Working paper. University of Pennsylvania Fittipaldi S (2004) When doing the right thing provides a pay-off. Global Finance 18:18–22 Granger CWJ (1969) Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37:428–438 Harrison J, Freeman R (1999) Stakeholders, social responsibility, and performance: empirical evidence and theoretical perspectives. Acad Manage J 42:479–487 Hillman A, Keim G (2001) Shareholder value, stakeholder management, and social issues: what’s the bottom line? Strategic Manage J 22:125–139 Maddala GS (1971) Generalized least squares with an estimated variance-covariance matrix. Econometrica 39:23–33 McGuire J, Sundgren A, Schneeweis T (1988) Corporate social responsibility and firm financial perfor- mance. Acad Manage J 31:854–872 McWilliams A, Siegel D (2001) Corporate social responsibility: a theory of the firm perspective. Acad Manage Rev 26:117–127 Statman M (2000) Socially responsible mutual funds. Financ Anal J 56:30–38 The Good Company (2005) January 22, The Economist, 374 Waddock S, Graves S (1997) The corporate social performance-financial performance link. Strategic Manage J 18:303–319 123

×