Does prior corporate financial performance affect corporate social performance?

(1)

Does prior corporate financial performance affect corporate social

performance?

A study concerned with the relationship between financial firm characteristics and corporate social responsibility within the slack resources theoretical framework

Author: M.P. (Martin) Knipper1 Student number: S2211084

Date: 8th of June, 2017 Supervisor: Dr. J.J. (Jakob) Bosma

Word count: 12,869

Abstract

This paper assesses the effect of prior corporate financial performance (FP) on corporate social performance (CSP). The findings complement the work of Waddock and Graves (1997) by assessing the FP-CSP relationship from the perspective of the slack resources theory. The dataset includes 11 years of observations for 1122 U.S. firms, in which the CSP measure is disaggregated in an environmental, social, and corporate governance dimension. The results indicate partial evidence in favour of the slack resources theory. Environmental performance is not significantly affected by prior FP, whereas social performance and corporate governance performance are affected by prior FP. Additional tests for Granger causality supports the findings of a preceding effect of FP on CSP, whereas no consistent evidence was found with respect to the contrary direction.

Keywords: Corporate financial performance, corporate social performance, corporate social

responsibility, good management theory, ESG, slack resources theory

JEL Codes: G30, M14, M21

1_{Correspondence information: University of Groningen, Faculty of Economics and Business, P.O. Box 800, 9700}

(2)

2 Introduction

The increasing significance of corporate social performance (CSP) in portfolio selections and investment strategies amplifies the necessity of understanding CSP and its relationship with corporate financial performance (FP). However, the majority of academic literature is concerned with the one way effect of CSP on FP (Ziegler and Schröder, 2010). In order to fill the aforementioned literature gap, this study attempts to assess whether CSP is depended on lagged FP by examining corporate social and corporate financial data for a sample of 1122 U.S. firms. The findings provide partial evidence in favour of the slack resources theory, as expressed by Waddock and Graves in 1997. The results indicate that prior FP positively affects social and corporate governance performance, whereas no effect is found for environmental performance. The results contribute to the literature concerned with corporate social responsibility (CSR) by strengthening the understanding of the FP-CSP relationship with regard to the sign and the direction of causality. Additionally, the results are relevant for the increasing social responsible investment (SRI) trend. According to the U.S. based Forum for Sustainable and Responsible Investment (US-SIF), a total of 8.72 trillion dollars was invested in the U.S. based on a SRI strategy in 2016, which is one fifth of total U.S. investments (US-SIF, 2016). Moreover, the US-SIF states in its annual report that most U.S. money managers use some combination of negative screening (socially themed exclusion) and positive screening (social responsible performance ranking) when integrating environmental, social, or corporate governance (ESG) factors in asset management. Hence, strengthening the knowledge about CSP versus FP dynamics yields value for investors as SRI is increasing to substantial levels. This paper first addresses the theoretical background concerned with CSR and the relation between CSP and FP. Secondly, hypothesises development is described, succeeded by further clarification of the adopted dataset and the methodology construction. Subsequently, univariate and multivariate analysis are executed, complemented with additional robustness checks. Lastly, the aggregated results are presented accompanied by closing remarks and recommendations.

Theoretical background

Corporate responsibility

(3)

3 (Freeman, 1984; Friedman, 1970), in which a firm’s responsibility is either limited to the shareholder, or extents to the stakeholder as well. The sole accountability of a firm is to its shareholder, according to Friedman (1970). In 1997, Carrol published an article which considered the concept of CSR as an addition to a firm’s financial performance, which is in line with the stakeholder perspective. Subsequently, academics and business increasingly started to study and integrate non-financial corporate factors, such as sustainability, gender equality, and employee satisfaction within firms. Currently, these factors are unified in the three ESG pillars. This paper adopts the allocation methodology of the Thomson Reuters ASSET4 Database (Thomson Reuters, 2012) for identifying the three ESG dimensions and their underlying factors, which are presented in table 1.

Definitions of CSR

In general, the substantial strand of literature discussing corporate social behaviour is concerned with a wide range of expressions to define this non-financial spectrum, such as CSR and CSP. A clear definition of a firm’s management engaging in CSR is given by: “Following the law and integrating social, environmental, ethical, consumer, and human rights concerns into business strategy and operations” (EU Commission, 2017). Furthermore, as mentioned by Kitzmueller and Shimshack (2012), this behaviour is summarized as: “Corporate social or environmental behaviour that goes beyond the legal or regulatory requirements of the relevant market(s) and/ or economy(s)”. Combined, CSR embodies a law transcending element which is concerned with a more expansive spectrum of responsibility than solely a firm’s core financial business.

Theoretical foundations of CSR

The literature generally allocates CSR engagement to two fundamental views, which are closely related to the shareholder and stakeholder perspective. First, CSR is considered as an agency problem, of which the costs rest on the shoulders of the shareholders. In the agency view, the

Table 1. ESG dimensions and sub dimension categories

Environmental Social Corporate governance

Emission reduction Community Board function

Product innovation Diversity Board structure

Resource reduction Employment quality Compensation policy

Health and safety Shareholder policy

Human rights Vision and strategy

(4)

4 firm’s adoption of CSR policies is related with management’s pursuit of private benefits (Krüger, 2015) and over prioritization of CSR objectives at the expense of a firm’s core business (Jensen, 2001). Additionally, as Chang et al. (2008) argue, CSR investments are long term projects which have a stronger association with direct private benefits of management, instead of generating shareholder value. In general, from the perspective of the agency view, CSR engagement is not beneficial for the stockholder (Ferrell et al., 2016). The second view is based on the stakeholder perspective, which embraces the view that pursuing shareholder value maximization can be complementary to social responsible engagement, which benefits the society as a whole (Ferrell et al., 2016). This is in line with the good management theory, which was first described in the study of Waddock and Graves (1997). The good management theory supports the view that CSP causes FP improvement. For example, due to more efficient usage of resources, better commitment of the workforce, or an improved relation with all stakeholders (Scholtens, 2008). Ferrell et al. (2016) argue that most empirical findings “lend support to the good governance view and suggest that CSR in general is not inconsistent with shareholder wealth maximization”. However, the reverse causal relationship between CSP and FP is embodied in the slack resources theory. According to Waddock and Graves (1997), slack financial resources availability as a result of improved financial performance, positively affects CSR investments. The slack resources theory is in line with the study of Ullmann (1985), who argues that inadequate financial performance prioritizes economic demands, while positive economic performance provides the capability of pursuing social investments. Xu et al. (2015) argue that slack resources can be divided in absorbed slack and unabsorbed slack. Absorbed slack is concerned with abundant resources already allocated to a specific objective, such as excess costs or expenses greater than what are needed by a firm. Examples of unabsorbed slack are uncommitted abundant resources such as financial slack and customer relational slack. Xu et al. (2015) conclude that unabsorbed slack is positively related with increased incentives and capacity to directly invest in CSR. In order to summarize, an overview of the most dominant theoretical frameworks surrounding CSR is provided in table 2.

Table 2. CSR theoretical framework

Perspectives CSR occurs as Implications of CSR CSP-FP relationship in the stakeholder view Stakeholder

view

Value adding policy for both shareholders and society as a whole ➢ Beneficial for shareholder wealth ➢ Generates stakeholder value CSP FP

Good management theory 

(5)

5

Determinants of CSP

In the academic area, a vast amount of CSP literature is concerned with portfolio returns of SRI related investments (e.g. Kempf and Osthoff, 2007), abnormal returns (e.g. Schröder, 2014; Dimson et al., 2015), or the cost of capital (e.g. Cheng, 2014; El Ghoul et al., 2011; Oikonomou et al., 2014). However, as noted by Scholtens (2008) with regard to the meta study of Margolis and Walsh (2001), the majority of studies that analyse the FP-CSP relation treat CSP as the explanatory variable, of which 50% identifies a positive relationship. Additionally, Margolis and Walsh (2001) summarized that two thirds of the minor strand of literature treating CSP as dependent variable found a positive effect. Additionally, as Ziegler and Schröder (2010) sharply report, only a relative small amount of research is related to the determinants of CSP. Moreover, Ziegler and Schröder (2010) note in their study concerned with the financial determinants of sustainable stock index inclusion, that the available literature affiliated with the FP-CSP relationship mainly focuses on portfolio analysis, event studies or longer term econometric approaches which either target the effects of CSP on FP, or adopt non-financial factors as explanatory variables, which are generally solely concerned with environmental variables. Hence, the interrelationship between CSP and FP, with CSP as depended variable, is a theoretical territory yet to be further analysed. The question in this context considers which firm specific financial characteristics affect CSP. In particular, do financial slack resources influence a firm’s management to increase CSR investments, such as reducing the carbon footprint or improving employment quality? Furthermore, the direction of causality between the FP-CSP relationship needs further clarification. Understanding these dynamics could yield an improved view on various firm related CSR theories, such as the slack resources theory and the good management theory.

The FP-CSP relationship

The first comprehensive paper concerned with the two way relationship between CSP and FP was published by Waddock and Graves (1997). Their study addressed the “virtuous circle” of CSP and FP and hypothesized a simultaneous relationship based on the slack resources theory

Shareholder view

Agency Problem

(6)

6 and the good management theory. Complementary to the study of Waddock and Graves (1997), the studies of Ziegler and Schröder (2010), Artiach et al. (2010), and Lourenco and Branco (2013) continue to contribute to the strand of literature concerned with the effects of FP on CSP. However, even though the aforementioned studies implement financial firm specific explanatory variables in their regression analysis, the majority examines sustainability stock index inclusion as dependent variable. Hence, their methodology has a significant dependence of the ESG assessment and – in particular – the aggregation procedures of the index developer. Furthermore, adopting index inclusion as dependent variable does not yield additional insights in the underlying ESG dimensions, since the three dimensions are not comprehensively distinguished. However, the study of Garcia et al. (2017), is the first study which individually assesses each ESG dimension as dependent variable, while studying the effects of FP on CSP in emerging markets. Moreover, Zhao and Murrell (2016) revisit the model of Waddock and Graves (1997) on a more substantial dataset, in which they confirm the findings of the original model with regard to the preceding effect of FP in relation to CSP, whereas their study could not support the initial argument that prior CSP positively affects future FP. Furthermore, Bird et al. (2006) clarified the causality within the FP-CSP relation and found evidence in favour of the slack resources theory using Granger causality models on a limited set of CSR data. Scholtens (2008) continued the methodology of Bird et al. (2006) and found preliminary evidence in favour of the slack resources theory. However, a comprehensive panel data study concerned with the effects of FP indicators on separately defined non-binary ESG dimensions, complemented with a proper identification of the direction of causality, is absent in the current strand of literature. This study attempts to fill this gap in the literature by examining relevant FP indicators while controlling for other firm specific financial measures which could affect CSR investments of firms.

Formulation of hypothesises

This paper studies the effects of firm specific financial performance on CSP from the slack resources perspective as proposed by Waddock and Graves (1997). Based on the aforementioned literature, various financial firm characteristics are adopted as drivers for CSP.

Financial performance

(7)

7 find a positive relationship between CSP and FP, as measured by Return of Equity (ROE). However, profitability as measured by Return on Assets (ROA) does not yield a significant result, which suggests that CSP occurs solely via the return which is available to equity holders. Furthermore, the study of Waddock and Graves (1997) shows that CSP is positively affected by different indicators of FP such as Return on Sales, Return on Equity, and Return on Assets. Moreover, Lourenco and Branco (2013) find similar results regarding Return on Equity for Brazilian listed firms. Additionally, Lourenco and Branco (2013) argue that higher returns permit firms to engage in CSR without neglecting shareholder and analyst expectations. Therefore, higher profitability is associated with more substantial investments in CSR. Moreover, in line with Guenster et al. (2006), Ziegler and Schröder (2010) adopt the Tobins Q measure for financial performance and find a positive association with sustainable index inclusion. Tobins Q differs from ROE, since it is a forward looking FP measure (Ziegler and Schröder, 2010) and is an adequate proxy for investment opportunities as well (Ferrell et al., 2016). Overall, a positive relationship between CSP and FP is expected for all three ESG dimensions. Furthermore, the empirical model of this study separately adopts the accounting performance measures ROA and ROE, and the financial market performance measure Tobins Q as proxies for FP in order to adequately review potential effects on CSP. For each of the aforementioned measures the following hypothesises are tested:

H1a: Prior FP positively affects environmental performance H1b: Prior FP positively affects social performance

H1c: Prior FP positively affects corporate governance performance

Sin industries

(8)

8

Size

Size is a relevant factor as well, since larger companies might receive an increased amount of attention from various stakeholders (Ullman, 1985). Artiach et al. (2010) find that large firms are likely to be industry leaders regarding CSP and maintain a more suitable position to realize economies of scale when implementing CSR schemes. Additionally, the significant visibility of large firms to the social public and legislative stakeholders discourages these firms to pursue a passive CSR strategy. This is in line with Ziegler and Schröder (2010), who argue that size may be considered as an indicator for the capacity of a firm to engage in environmental and social activities, which lead to fixed costs that are relatively less significant for larger companies. Therefore, size is adopted as control variable with the expectation it is positively associated with CSP.

Leverage and dividends

Ferrell et al. (2016) argue that debt and dividends could provide a restriction for a firm’s management to invest excess funds in unprofitable projects in order to pursue private benefits. The notion is complementary to Chang et al. (2008), who argue that investments in social responsible projects generally focus on the long run and do not yield shareholder value. Additionally, Artiach et al. (2010) argue that a firm’s management will prioritize obligations to debt holders over society oriented stakeholders, since debtholders are more dominant stakeholders. Moreover, Ziegler and Schröder (2010) state that firms with modest levels of leverage could be more adaptive to CSR related investments. Hence, leverage and dividend yield could affect the effect of FP on CSP. The association between debt and CSP is expected to be negative, similar to the expected relation between dividend yield and CSP.

Growth opportunities

(9)

9 related with CSP. Therefore, growth opportunities of a firm could affect CSR engagement and need to be taken into account in the FP-CSP relationship. In line with Lourenco and Branco (2013), growth opportunities will be measured via the price to book ratio, which is a firm’s market capitalization divided by common equity. Based on the literature, a positive association between the price to book ratio and CSP is expected.

Data

In order to study the FP-CSP relationship, a comprehensive dataset is retrieved from Thomson Reuters DataStream. The dataset includes end of the year CSP and financial panel data for all constituents of the S&P 1500 index, ranging from the year 2005 to (and including) 2015. The dataset embodies an exhaustive time horizon for which adequate CSP data was available.

Asset4 ESG scores

The depended variable ESG performance is based on the aggregated ASSET4 ESG scores as provided by the ASSET4 unit of Thomson Reuters. As noted by Schäfer et al. (2006), the ASSET4 database intents to provide an integrated view of corporate performance by combining financial and non-financial information. The ASSET4 ESG scores are based on 500 separate data points collected from multiple sources, such as company reports, CSR reports, and company filings (Thomson Reuters, 2013). The data points are subsequently converted to 226 key performance indicators (KPIs), which are allocated into the three pillars of ESG. The three pillars are defined as the environmental pillar, social pillar, and corporate governance pillar, resulting in three firm specific scores. The environmental score includes resource usage and reduction, emissions and emissions reductions, environmental activism and initiative, and product or process innovation. The social score includes employment quality, health and safety issues, training, diversity, human rights, community involvement, and product responsibility. The corporate governance score includes board structure, compensation policy, board functions, financial and operational transparency, shareholder rights, and vision and strategy. Appendix A provides additional information concerned with the scoring aggregation methodology of the ASSET4 ESG scores.

Financial data

(10)

10 ratio, dividend yield, and the price to book ratio are control variables. Tobins Q is defined as the market to book ratio of assets and is calculated as the market value of equity plus the total debt outstanding, divided by the book value of total assets (Maury and Pajuste, 2005). Additionally, firms which operate within a sin industry are identified via a dummy variable with the value of 1, and 0 otherwise. The industry affiliation is based on the industry group segregation indicator as provided by Thomson Reuters DataStream, of which an overview can be found in Appendix B. In line with Ziegler and Schröder (2010) and Ferrell et al. (2016), the total assets variable is converted to its logarithmic value.

Methodology

This paper studies the effect of FP on CSP, as measured by the aforementioned financial indictors and a threefold ESG measures. The advantage of panel data is to evade spurious correlation as a result of sole cross sectional analysis and unobserved factors of influence, such as well performing management which could affect both FP as CSP (Ziegler and Schröder, 2010). The depended variable is CSP (𝐶𝑆𝑃_𝑡,𝑖), which is estimated by a threefold CSP measures being environmental performance (𝐸𝑁𝑉 𝑆𝑐𝑜𝑟𝑒_𝑡,𝑖), social performance (𝑆𝑂𝐶 𝑆𝑐𝑜𝑟𝑒_𝑡,𝑖), and corporate governance performance (𝐶𝐺𝑉 𝑆𝑐𝑜𝑟𝑒_𝑡,𝑖), as measured by the Thomson Reuters ASSET4 scores. The varying regressors of the model are corporate financial performance (𝐹𝑃𝑡,𝑖), the logarithm of total assets (𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑡,𝑖), the debt to assets ratio (𝐷𝑒𝑏𝑡 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡𝑠𝑡,𝑖), dividend yield (𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑌𝑖𝑒𝑙𝑑𝑡,𝑖), and the price to book ratio (𝑃𝑟𝑖𝑐𝑒 𝑡𝑜 𝐵𝑜𝑜𝑘𝑡,𝑖). One time invariant regressor is adopted, being the control variable for a sin industry (𝑆𝑖𝑛𝑖). Each model will be estimated with ROA (𝑅𝑂𝐴𝑡,𝑖), ROE (𝑅𝑂𝐸𝑡,𝑖) and Tobins Q (𝑇𝑜𝑏𝑖𝑛𝑠 𝑄𝑡,𝑖) separately plugged in as 𝐹𝑃𝑡,𝑖 variable. All time varying explanatory variables are lagged with one year, in line with Ferrell et al. (2016). Moreover, three linear methods are used to estimate the regression model.

Pooled OLS model

(11)

11 𝐶𝑆𝑃_𝑡,𝑖 = 𝛼 + 𝛽₁𝐹𝑃_𝑡-1,𝑖+ 𝛽₂𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖+ 𝛽₃𝐷𝑒𝑏𝑡 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖

+ 𝛽₄𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑌𝑖𝑒𝑙𝑑_𝑡-1,𝑖+ 𝛽₅𝑃𝑟𝑖𝑐𝑒 𝑡𝑜 𝐵𝑜𝑜𝑘_𝑡-1,𝑖 + 𝛽₆𝑆𝑖𝑛_𝑖+ 𝑢_𝑡,𝑖 (1)

Fixed effects model

Secondly, the fixed effects model will be adopted, which allows the individual specific intercept (𝛼) to vary cross sectional, holding all slope coefficients fixed both cross sectional and over time. Hence, the constant 𝛼 is allowed to be correlated with the regressors. Moreover, the fixed effects model provides the leftover variation in the depended variable that cannot be explained by the regressors. For example, firm culture, firm strategy, or the CEO’s personal values could affect CSP as well (Garcia-Castro et al., 2010). The fixed effects model is estimated by disaggregating 𝑢_𝑡,𝑖 of equation (1) into an individual specific effect 𝑢_𝑖 and the leftover disturbance 𝑣_𝑖,𝑡, which varies over time and cross sectional. The fixed effects model is provided in equation (2).

𝐶𝑆𝑃_𝑡,𝑖 = 𝛼 + 𝛽₁𝐹𝑃_𝑡-1,𝑖+ 𝛽₂𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖+ 𝛽₃𝐷𝑒𝑏𝑡 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖 + 𝛽₄𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑌𝑖𝑒𝑙𝑑_𝑡-1,𝑖+ 𝛽₅𝑃𝑟𝑖𝑐𝑒 𝑡𝑜 𝐵𝑜𝑜𝑘_𝑡-1,𝑖 + 𝑢_𝑖 + 𝑣_𝑡,𝑖 (2)

Random effects model

Thirdly, a random effects model will be adopted, which allows – likewise to the fixed effects model – for cross sectional varying intercepts, while maintaining the intercepts constant over time. However, the random effects model assumes that the cross sectional intercept arises from both a universal intercept – which does not vary cross sectional or over time – and a time fixed random term 𝜖𝑖, which measures the deviation of each cross section’s intercept with regard to the universal intercept. The random effects model is provided in equation (3).

𝐶𝑆𝑃_𝑡,𝑖 = 𝛼 + 𝛽₁𝐹𝑃_𝑡-1,𝑖+ 𝛽₂𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖+ 𝛽₃𝐷𝑒𝑏𝑡 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖 + 𝛽₄𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑌𝑖𝑒𝑙𝑑_𝑡-1,𝑖+ 𝛽₅𝑃𝑟𝑖𝑐𝑒 𝑡𝑜 𝐵𝑜𝑜𝑘_𝑡-1,𝑖 + 𝛽₆𝑆𝑖𝑛_𝑖+ 𝜖_𝑖 + 𝑣_𝑡,𝑖 (3)

Model validation

(12)

12 models to identify which model is the most appropriate model to adopt. The likelihood test for redundant fixed effects is used to verify whether the fixed effects of the fixed effects model are supported by the pooled OLS model. Hence, if the fixed effects are not supported, the fixed effects model is preferred over the pooled OLS model. Additionally, the Hausman test is employed to verify whether the random effects term of the random effects model is uncorrelated with the explanatory variables. Rejection of the null hypothesis, which tests for no correlation between the random effects and the explanatory variables, indicates that the fixed effects model is preferred over de random effects model.

Univariate analysis

The dataset is based on the S&P 1500 stock index, which consists of the largest U.S. companies as measured by market capitalization (S&P U.S. Indices Methodology, 2017). However, the dataset of ASSET4 does not provide sufficient ESG data for all S&P 1500 constituents. Hence, the firms of which ASSET4 does not provide data are excluded from the analysis. Additionally, as a result of the research scope of ASSET4, ESG data prior to 2005 is not available for a fair amount of firms in the remaining dataset, and ESG data succeeding 2015 is not yet available. Summarizing, the non-lagged panel data consists of 7932 observations for 1122 U.S. firms over the period 2005 to 2015. An amount of 996 firms are considered as being in a non-sin industry, whereas 126 firms are classified as participants of a sin industry. Additionally, the dataset yields an unbalanced panel since a portion of the financial data is unavailable. Furthermore, all quantitative results of this paper are available upon request from the author.

Descriptive statistics

(13)

13 the FP of non-sin industry firms is on average higher than firms which are active in a sin industry. Hence, based on the descriptive statistics, firms active in a non-sin industry appear to have a higher CSP and FP on average.

Correlation matrix

Table 4 provides the correlation coefficients of all variables. Size, as measured by the logarithm of total assets, and the dividend yield ratio appear to have the highest positive correlation with the ESG factors. Additionally, firms which are active in a sin industry are negatively associated with the ENV score and SOC score, which corresponds with the expectations of the aforementioned literature. However, a positive association between CGV score and sin industry is suggested, which could indicate that the corporate governance dimension shows divergent dynamics with regard to the ENV score and SOC score. The discordant relation is illustrated by the correlation coefficients, which indicate that the ENV score and the SOC score are more

Table 3. Descriptive statistics for non-sin industries and sin industries

Mean Median Maximum Minimum Std. Dev. Skewness Kurtosis Observations

Non-lagged variables ENV score 45.39 (42.96) 35.86 (31.39) 97.47 (97.23) 8.28 (9.02) 32.15 (30.33) 0.33 (0.45) 1.45 (1.57) 6957 (975) SOC score 48.42 (47.32) 46.28 (43.33) 98.93 (97.56) 3.55 (5.14) 28.65 (28.37) 0.14 (0.24) 1.67 (1.68) 6957 (975) CGV score 73.48 (75.48) 76.85 (78.99) 97.52 (97.44) 1.42 (9.05) 16.37 (15.33) -1.11 (-1.26) 4.4 (4.95) 6957 (975)

Lagged variables (1 year lag)

ROA 0.07 (0.07) 0.06 (0.07) 0.89 (0.34) -0.85 (-0.33) 0.08 (0.07) -0.67 (-0.55) 17.82 (6.83) 5905 (844) ROE 0.17 (0.15) 0.14 (0.14) 8.72 (2.32) -2.53 (-1.39) 0.34 (0.21) 7.8 (2.34) 153.98 (26.29) 5863 (841) Tobins Q 1.65 (1.31) 1.3 (1.08) 15.97 (6.84) 0.02 (0.27) 1.39 (0.76) 2.93 (2.42) 17.6 (11.85) 5915 (848) Log total assets 6.99 (7.04) 6.91 (7) 9.41 (8.54) 5.35 (5.78) 0.64 (0.49) 0.68 (0.37) 3.63 (3.01) 5949 (848) Debt to assets 0.24 (0.24) 0.22 (0.23) 1.47 (0.76) 0 (0) 0.18 (0.12) 0.83 (0.58) 4.16 (3.49) 5949 (848) Dividend yield 0.0 (0.02) 0.01 (0.01) 0.21 (0.18) 0 (0) 0.02 (0.02) 2.24 (2.53) 15.55 (18.53) 5919 (848) Price to book 3.62 (2.6) 2.53 (1.93) 87.67 (32.23) -9.93 (-5.06) 5.02 (2.72) 7.92 (5.31) 96.61 (44.65) 5903 (845)

Time invariant variable

Sin industry

(14)

14 harmonious with each other than CGV score, since the correlation between ENV score and SOC score is 0.78, whereas there individual correlation with CGV score is approximately 0.54. Furthermore, the profitability measures ROA and ROE show a positive association with the three ESG measures, whereas Tobins Q is negatively associated with the ESG measures. Hence, univariate analysis does not provide consistent evidence in favour or opposing the slack resources theory.

Table 4. Correlation matrix ENV score SOC score CGV score

ROA ROE Tobins Q Log total assets Debt to assets Dividend yield Price to book Sin industry ENV score 1 SOC score 0.78 1 CGV score 0.54 0.53 1 ROA 0.06 0.09 0.05 1 ROE 0.06 0.09 0.06 0.55 1 Tobins Q -0.05 -0.02 -0.05 0.53 0.22 1

Log total assets 0.39 0.36 0.20 -0.22 -0.06 -0.42 1

Debt to assets 0.04 -0.03 0.02 -0.21 0.00 -0.18 0.07 1

Dividend yield 0.16 0.11 0.08 -0.11 -0.02 -0.23 0.23 0.24 1

Price to book 0.02 0.04 0.01 0.24 0.44 0.46 -0.18 0.11 -0.09 1

Sin industry -0.04 -0.02 0.03 0.00 -0.02 -0.08 0.03 0.00 -0.04 -0.07 1 The correlation matrix is based on a balanced sample with listwise missing value deletion. All time varying explanatory variables are lagged with 1 year.

Multivariate analysis

The multivariate analysis consists of three econometric panel data models, applied over three different FP measures for each ESG dimension. Hence, the multivariate analysis yields 27 models in total in order to adequately analyse the effects of lagged financial performance on CSP. This study aims to comprehensively assess evidence in favour of the slack resources theory for each of the three ESG dimensions. First, the ENV score model will be discussed, succeeded by the SOC score model and the CGV score model, respectively.

Environmental score model

(15)

15 However, OLS does not distinguish individuality or heterogeneity that may exist among the different companies of the dataset. Therefore, a random effects and fixed effects model are provided as well. First, the most appropriate model must be selected in order to analyse the regression outcomes. Hence, by analysing the Hausman test and likelihood F-statistic, the fixed effects model is the most adequate model. The F-statistic suggests inconsistency in the coefficients as provided by the pooled OLS model. Moreover, the Hausman test rejects the hypothesis that the parameters, as controlled by the random effects model, are consistent. Since both statistics are significant on the 0.05 confidence level for all of the three sets of models, the fixed effects model proves to be the most appropriate estimation. The size variable, as measured by log total assets, and the proxy for growth opportunities, the price to book ratio, have a significant positive effect on the ENV score within the fixed effects model for every FP

Table 5. Regression analysis estimations with environmental score as depended variable

ROA ROE Tobins Q

T = 2006-2015. Cross sections: 755 Obs. (unbalanced): 6706 T = 2006-2015. Cross sections: 751 Obs. (unbalanced): 6671 T = 2006-2015. Cross sections: 755 Obs. (unbalanced): 6748

OLS Random Fixed OLS Random Fixed OLS Random Fixed

Constant -111.907*** (4.351) -149.487*** (6.983) -190.086*** (9.087) -97.376*** (4.246) -151.601*** (7.059) -194.664*** (9.151) -117.885*** (4.733) -153.987*** (7.429) -194.461*** (9.547) Financial performance 59.547*** (5.031) 6.496* (3.494) 2.358 (3.568) 5.389*** (1.222) -0.35 (0.782) -0.614 (0.822) 3.13*** (0.337) 0.509* (0.281) 0.39 (0.295) Log total assets 21.693*** (0.608) 27.8*** (1) 33.694*** (1.307) 20.2*** (0.605) 28.25*** (1.016) 34.455*** (1.321) 22.466*** (0.642) 28.417*** (1.046) 34.25*** (1.354) Debt to assets 2.438 (2.204) 6.251** (2.545) 7.112** (2.852) -1.148 (2.27) 3.3 (2.668) 3.206 (2.977) 1.391 (2.2) 5.493** (2.521) 6.369** (2.809) Dividend yield 144.209*** (20.57) 36.498** (16.825) 22.007 (17.412) 138.308*** (20.963) 34.831** (16.85) 23.399 (17.392) 162.63*** (20.753) 36.468** (16.83) 23.296 (17.383) Price to book 0.444*** (0.079) 0.136*** (0.053) 0.108** (0.055) 0.469*** (0.091) 0.195*** (0.061) 0.195*** (0.064) 0.284*** (0.085) 0.129** (0.054) 0.115** (0.056) Sin industry -4.116*** (1.079) -6.254** (2.766) -4.078*** (1.093) -6.362** (2.834) -3.166*** (1.084) -6.076** (2.808) R-squared 0.186 0.11 0.81 0.169 0.111 0.809 0.178 0.11 0.81 Adjusted R-squared 0.185 0.109 0.785 0.168 0.11 0.785 0.177 0.109 0.785 Log likelihood -32129.19 -27256.26 -32030.62 -27124.74 -32364.29 -27429.48 F-statistic 255.184*** 137.512*** 33.332*** 226.251*** 138.042*** 33.214*** 243.284*** 138.607*** 33.546*** F-test for cross section fixed

effects

25.917*** 26.514*** 26.387***

Chi-square test for cross section fixed effects

9760.412*** 9825.698*** 9878.165***

Hausman Chi-square test for cross section random effects

100.593*** 67.945*** 81.421***

(16)

16 measure. The positive association between CSP and growth opportunities is in line with Artiach et al. (2010) and Padgett and Galan (2010). Firms seem to adopt the availability of growth opportunities by allocating investments in an environmentally friendly manner. Additionally, the positive association with the size variable in is line with Ziegler & Schröder (2010), who argue that environmentally friendly investments induces substantial fixed costs which are less difficult accommodated by large firms. However, the debt to assets ratio yields a statistically significant positive effect when ROA and Tobins Q are adopted as FP measure, whereas the literature suggests a negative association. Leverage seems to positively affect corporate social behaviour within the environmental dimension. Moreover, none of the FP measures appear to have a significant effect on environmental performance. Hence, the ENV score model yields no evidence in favour of the slack resources theory. The hypothesis that increased financial performance does not affect investments in environmental performance cannot be rejected. Additionally, it could be argued that environmental friendly investments are generally caused by non-financial drivers – instead of direct financial performance – of which firm size has a significant positive influence.

Social score model

(17)

17 the positive association between the dividend yield and social performance could suggest that the ability to generate dividends is a signal indicating the existence of sufficient available funds which are partly assigned to shareholders as dividend and partly allocated to socially favourable investments. In line with Artiach et al. (2010), one could argue that if a firm’s management prioritises payments to debtholders and shareholders before CSR investments, the generation of a firm’s dividend payments complements CSR investment. Hence, the significant positive effect of dividend yield could be in favour of the slack resources theory as well.

Table 6. Regression analysis estimations with social score as depended variable

ROA ROE Tobins Q

Constant -81.075*** (3.805) -105.391*** (6.12) -132.39*** (8) -68.389*** (3.713) -107.506*** (6.183) -136.43*** (8.023) -85.945*** (4.14) -119.055*** (6.516) -148.714*** (8.388) Financial performance 57.345*** (4.399) 12.602*** (3.074) 9.342*** (3.141) 5.833*** (1.068) 1.1 (0.685) 1.047 (0.72) 2.828*** (0.294) 1.678*** (0.247) 1.681*** (0.259) Log total assets 18.235*** (0.532) 21.787*** (0.877) 25.672*** (1.151) 17.009*** (0.529) 22. 293*** (0.89) 26.44*** (1.158) 18.901*** (0.561) 23.501*** (0.917) 27.728*** (1.189) Debt to assets -7.875*** (1.927) 5.36** (2.236) 9.591*** (2.511) -12.901*** (1.985) 0.595 (2.338) 3.877 (2.61) -9.022*** (1.925) 5.542** (2.213) 9.469*** (2.468) Dividend yield 77.861*** (17.987) 59.753*** (14.801) 55.648*** (15.329) 73.988*** (18.331) 56.157*** (14.771) 53.826*** (15.249) 94.821*** (18.155) 63.901*** (14.782) 60.76*** (15.272) Price to book 0.507*** (0.069) 0.144*** (0.046) 0.113** (0.048) 0.543*** (0.079) 0.258*** (0.054) 0.254*** (0.056) 0.373*** (0.074) 0.064 (0.048) 0.043 (0.049) Sin industry -2.341** (0.943) -3.318 (2.409) -2.113** (0.955) -3.23 (2.479) -1.563* (0.948) -2.897 (2.457) R-squared 0.171 0.093 0.804 0.154 0.093 0.805 0.161 0.096 0.804 Adjusted R-squared 0.171 0.092 0.779 0.154 0.093 0.78 0.16 0.096 0.779 Log likelihood -31229.21 -26401.55 -31135.54 -26247.25 -31461.84 -26555.71 F-statistic 231.046*** 113.843*** 32.065*** 202.597*** 114.503*** 32.275*** 215.75*** 119.881*** 32.367*** F-test for cross section fixed

effects

25.422*** 26.287*** 26.067***

9661.483*** 9781.472*** 9814.977***

Hausman chi-Square test for cross section random effects

98.328*** 54.38*** 71.645***

Standard errors are given in parentheses. The significance levels are provided for each coefficient on the 1%, 5%, and 10% level via ***, **, and *, respectively. All time varying independent variables are lagged with one year.

Corporate governance score model

(18)

18 in a sin industry have a more superior corporate governance structure than their non-sin industry opposites. One possible solution could be that firms which are active in sin industries have

improved board structures and compensation packages, or their industry provides more upside potential related to corporate governance performance than firms active in non-sin industries. Furthermore, equivalently to the former two estimation methods, the fixed effects model proves to be the most appropriate estimator. However, contrary to the ENV score model and the SOC score model, all three FP measures suggest a positive significant effect of prior FP on corporate governance performance. Hence, the CGV score model provides evidence in favour of the slack resources theory. Additionally, ROE solely shows a significant positive effect within the CGV score model, whereas no significant effect is found with respect to the environmental or social dimension. Therefore, the significant results regarding ROE as provided by the studies of

Table 7. Regression analysis estimations with corporate governance score as depended variable.

ROA ROE Tobins Q

Constant 37.483*** (2.107) 32.209*** (3.669) 25.697*** (5.628) 41.554*** (2.029) 33.603*** (3.646) 24.55*** (5.651) 39.133*** (2.289) 26.947*** (3.913) 14.468** (5.903) Financial performance 16.984*** (2.436) 8.415*** (2.124) 7.423*** (2.21) 2.535*** (0.584) 1.046** (0.467) 1.011** (0.507) 0.383** (0.163) 0.778*** (0.167) 1.037*** (0.182) Log total assets 5.116*** (0.294) 5.878*** (0.525) 6.923*** (0.81) 4.692*** (0.289) 5.78*** (0.526) 7.193*** (0.815) 4.983*** (0.31) 6.559*** (0.55) 8.394*** (0.837) Debt to assets 0.346 (1.067) 1.033 (1.448) 2.923* (1.766) -0.372 (1.085) -0.03 (1.506) 0.908 (1.838) -0.472 (1.064) 1.108 (1.435) 2.8 (1.737) Dividend yield 34.184*** (9.96) 11.981 (10.107) 2.427 (10.784) 31.285*** (10.017) 9.631 (10.088) 0.872 (10.74) 35.757*** (10.038) 12.977 (10.102) 5.126 (10.747) Price to book 0.117*** (0.038) 0.096*** (0.032) 0.096*** (0.034) 0.085** (0.043) 0.105*** (0.037) 0.128*** (0.039) 0.11*** (0.041) 0.035 (0.033) 0.022 (0.034) Sin industry 1.329** (0.522) 1.53 (1.226) 1.296** (0.522) 1.49 (1.223) 1.479*** (0.524) 1.741 (1.235) R-squared 0.053 0.021 0.638 0.049 0.02 0.635 0.046 0.022 0.639 Adjusted R-squared 0.052 0.02 0.592 0.048 0.019 0.588 0.046 0.021 0.593 Log likelihood -27265.47 -24043.23 -27104.38 -23908.96 -27463.17 -24184.65 F-statistic 62.807*** 24.242*** 13.799*** 56.813*** 22.795*** 13.63*** 54.679*** 25.453*** 13.972*** F-test for cross section fixed

effects

12.75*** 12.689*** 13.068***

6450.953*** 6397*** 6564.987***

16.466*** 15.45*** 28.807***

(19)

19 Artiach et al. (2010), Lourenco and Branco (2013), or Waddock and Graves (1997) could be particularly attributed to the corporate governance dimension, since no evidence was found that prior ROE significantly affected CSP as measured by either environmental performance or social performance. Overall, evidence in favour of the slack resources theory is suggested in the social and corporate governance dimensions, whereas no evidence is found within the environmental dimension.

Granger causality

However, association does not confirm causality and the model could be misleading. For example, if the “virtuous circle” as proposed by Waddock and Graves (1997) holds, then CSP does not only depend on FP, but reciprocally affects FP as well. Hence, the FP measures could be affected by the problem of endogeneity. Therefore, in order to verify the direction of causality between CSP and FP, the relationship between the indicators must be further clarified. Hence, a test for Granger causality is adopted. Granger causality reviews the relation between two variables while attempting to answer the question: do changes in variable I cause changes in variable II (Brooks, 2008)? The methodology of Granger causality assesses whether lagged values of variable I precede changes in variable II, which would signal a unidirectional relationship. Additionally, if Granger causation appears for both directions, a bidirectional relationship exists. Furthermore, if no Granger causation can be identified for both directions, the two variables are exogenous with respect to each other. In line with Scholtens (2008), the Granger causality test is adopted to review the direction of causality between the financial performance variables and the ESG scores. However, as noted by Brooks (2008), the identification of Granger causality does not indicate that changes in variable I directly cause changes in variable II. Granger causality merely identifies evidence for preceding or succeeding movements between variables and assesses the correlation between past values of variable I and future values of variable II. Summarizing, the Granger causality test is employed in order to strengthen the multivariate findings of the previous paragraph by assessing whether there is evidence for Granger causality between CSP and FP, and to further clarify the direction of causality. Equation (4) and equation (5) are adopted to run the regression for each combination of CSP measure 𝑌 and explanatory variable 𝑋 for 𝑙 amount of lags.

𝑌_𝑡= 𝛼₀+ 𝛼₁𝑌_𝑡-1+ . .. +𝛼_𝑙𝑌_𝑡-𝑙+ 𝛽₁𝑋_𝑡-1+ . . . +𝛽_𝑙𝑋_𝑡-𝑙 + 𝜖_𝑡 (4)

(20)

20 Table 8 provides the probability values of the F-statistics for two periods of lags, while testing for the null hypothesis of no Granger causation between the variables via a Wald test. In line with Scholtens (2008), three and five lags are used for every ESG dimension. The results suggest that FP generally precedes CSP, since the null hypothesis of no Granger causality is

rejected for all three FP measures as lagged variable in relation with all three lead CSP variables, whereas the opposite of CSP preceding FP in Granger sense is not precedent consistently. These findings are in line with the results of Scholtens (2008), who finds an overall preceding effect of FP over CSP. Additionally, the results show a strong bidirectional relationship between ROE and CSP, which suggests that reciprocal endogeneity, with respect to ROE and CSP, cannot be ruled out. Hence, profitability as measured by the return available to shareholders could be both a consequence and a cause of CSP. In general, the test for Granger causality provides evidence

Table 8. Granger causality test

Lags: 3 F-values per CSP Dimension Lags: 5 F-values per CSP Dimension Null Hypothesis: Obs. ENV score SOC score CGV score Obs. ENV score SOC score CGV score

ROA does not Granger cause CSP 5177 5.942*** 8.467*** 5.815*** 3708 4.899*** 6.867*** 4.038*** CSP does not Granger cause ROA 2.004 3.085** 0.528 0.299 2.762** 0.402

ROE does not Granger cause CSP 5129 10.656*** 11.17*** 7.002*** 3675 5.329*** 7.947*** 4.339*** CSP does not Granger cause ROE 6.83*** 7.79*** 2.289* 1.869* 4.556*** 0.649 Tobins Q does not Granger cause

CSP

5261 4.957*** 7.133*** 4.024*** 3798 5.311*** 11.246*** 5.941*** CSP does not Granger cause

Tobins Q

2.492* 2.112* 0.918 0.774 0.885 1.126

LOGTA does not Granger cause CSP

5289 3.524** 0.969 10.385*** 3818 2.136* 4.635*** 1.251 CSP does not Granger cause log

total assets

1.565 1.062 0.484 1.098 1.249 0.125

Debt to assets does not Granger cause CSP

5289 1.887 0.774 4.423*** 3818 0.489 1.392 2.739** CSP does not Granger cause debt

to assets

3.739** 1.023 5.971*** 2.713** 0.954 3.07***

Dividend yield does not Granger cause CSP

5265 0.015 1.172 0.344 3802 0.631 4.139*** 3.585*** CSP does not Granger cause

dividend yield

3.458** 2.542* 3.379** 6.851*** 4.079*** 3.771***

Price to book ratio does not Granger cause CSP

5248 4.23*** 7.447*** 3.792*** 3787 1.764 5.98*** 2.494** CSP does not Granger cause price

to book ratio

(21)

21 in favour of the slack resources theory. Moreover, on the short term, a unidirectional relationship in Granger sense is suggested between all CSP indicators and the dividend yield ratio. However, a bidirectional relationship is shown for the five year lag test regarding social and corporate governance performance. The effects of a firm’s environmental performance seems to follow a structural unidirectional association in Granger sense, which suggests that increased environmental performance precedes a higher dividend yield.

Robustness tests

Winsorized model

(22)

22

Non-linearity validation

Furthermore, as found by Barnett and Salomon (2012), there is evidence that the FP-CSP relationship is curvilinear and could be described by a U-shaped form. Barnett and Salomon (2012) find that relative low and relatively high CSP generally results in high FP, whereas moderate CSP levels do not. However, it is unclear whether this effect exists for the reverse causality when CSP is depended on prior FP. Hence, in order to assess the model’s validity with respect to non-linear effects or cross dependence between the variables, an auxiliary regression model is estimated. Since the fixed effects model proves the most viable for all prior estimations, the auxiliary regression is estimated directly under fixed effects specifications. The squared value of the variable of interest, FP, is added to the model to verify the existence of a non-linear association between CSP and FP. Additionally, the cross product of FP and size, as measured the logarithm of total assets, is included in the auxiliary regression. However, since size and FP are relatively highly correlated, the cross product might not yield appropriate estimations and the interaction effect cannot be measured correctly. Therefore, additional justification is present for adding the squared term of size to the model as well. The auxiliary resulting regression model is provided in equation (6).

𝐶𝑆𝑃_𝑡,𝑖 = 𝛼 + 𝛽₁𝐹𝑃_𝑡-1,𝑖 + 𝛽₂𝐹𝑃_𝑡-1,𝑖2 + 𝛽₃(𝐹𝑃_𝑡-1,𝑖× 𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖) + 𝛽₄𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖+ 𝛽₅𝐿𝑜𝑔 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖2 + 𝛽₆𝐷𝑒𝑏𝑡 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡𝑠_𝑡-1,𝑖

+ 𝛽7𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑌𝑖𝑒𝑙𝑑𝑡-1,𝑖 + 𝛽8𝑃𝑟𝑖𝑐𝑒 𝑡𝑜 𝐵𝑜𝑜𝑘𝑡-1,𝑖+ 𝑢𝑖 + 𝑣𝑖,𝑡 (6)

(23)

23 Table 9. Auxiliary regressions estimations for non-linearity and cross dependence

T = 2006-2015. Cross-sections included: 755 Obs. (unbalanced): 6706

ENV score model (fixed effects) SOC score model (fixed effects) CGV score model (fixed effects)

ROA ROE Tobins Q ROA ROE Tobins Q ROA ROE Tobins Q

Constant -215.831*** (57.272) -227.104*** (56.501) -213.396*** (67.104) -503.4*** (50.211) -489.677*** (49.304) -584.165*** (58.679) 44.496 (35.456) 52.083 (34.887) 70.093* (41.46) Financial performance -93.033** (40.707) -4.426 (8.383) 1.252 (3.863) 41.657 (35.688) 0.928 (7.315) 10.005*** (3.378) -30.421 (25.2) -3.573 (5.176) -3.145 (2.387) FP sq. 14.646 (9.763) -0.59** (0.265) -0.139** (0.059) 17.684** (8.559) -0.621*** (0.231) -0.123** (0.051) 19.513*** (6.044) -0.369** (0.163) -0.07* (0.036) FP × log total assets 14.393** (6.143) 0.72 (1.181) 0.081 (0.553) -5.014 (5.386) 0.169 (1.03) -1.087** (0.484) 5.671 (3.803) 0.769 (0.729) 0.757** (0.342) Log total assets 41.842** (16.47) 43.913*** (16.305) 39.157** (18.835) 133.26*** (14.439) 129.395*** (14.228) 152.101*** (16.47) 1.402 (10.196) -0.799 (10.068) -6.937 (11.637) Log total assets sq. -0.643 (1.182) -0.689 (1.174) -0.348 (1.321) -7.733*** (1.036) -7.441*** (1.024) -8.829*** (1.155) 0.396 (0.732) 0.571 (0.725) 1.02 (0.816) Debt to assets 7.246** (2.86) 3.729 (3) 6.401** (2.814) 8.267*** (2.507) 3.129 (2.618) 8.383*** (2.461) 3.097* (1.77) 1.428 (1.852) 2.876* (1.739) Dividend yield 23.163 (17.438) 23.906 (17.401) 31.07* (17.605) 57.213*** (15.288) 57.506*** (15.184) 68.385*** (15.395) 1.692 (10.795) 0.651 (10.744) 10.87 (10.877) Price to book 0.109** (0.055) 0.206*** (0.064) 0.11** (0.056) 0.111** (0.048) 0.27*** (0.056) 0.042 (0.049) 0.094*** (0.034) 0.135*** (0.039) 0.018 (0.034) R-squared 0.81 0.809 0.81 0.806 0.807 0.806 0.639 0.635 0.64 Adjusted R-squared 0.786 0.785 0.786 0.781 0.782 0.781 0.592 0.589 0.594 Log likelihood -27251.47 -27121.22 -27425.2 -26369.06 -26212.29 -26519.82 -24035.72 -23904.88 -24175.92 F-statistic 33.243*** 33.109*** 33.449*** 32.31*** 32.552*** 32.652*** 13.786*** 13.596*** 13.967*** The regression models are based on the fixed defects model estimation methodology. Standard errors are given in parentheses. The significance levels are provided for each coefficient on the 1%, 5%, and 10% level via ***, **, and *, respectively. All time varying independent variables are lagged with one year.

Industry peer performance ratio

(24)

24 the corporate governance score. The industry disaggregation is based on the industry group segregation indicator as provided by Thomson Reuters DataStream, of which an overview can be found in Appendix B. The resulting ratio resembles a peer performance score, with a score of 1 indicating industry neutral peer CSP, and a score higher or lower than 1 indicating superior or inferior CSP with regard to the corresponding industry, respectively. Subsequently, the scores are compared with the size variable, the logarithm of total assets, of which the corresponding scatterplots for each CSP measure can be found in figure 1. The graphical results

Figure 1. Scatterplots of peer performance ratio

The y-axis resembles the industry peer ratio, which is calculated as a firm’s individual annual ASSET4 score, divided by the average ASSET4 score within the corresponding industry of the same year. The x-axis resembles the size indicator, measured as the logarithm of total assets. The fitted regression line is based on OLS and indicates the general relationship between the peer performance ratio and size of a firm. 0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5 5,5 6 6,5 7 7,5 8 8,5 9 9,5 En vi ro n m en ta l p ee r ra ti o

Log total assets Environmental Peer Performance

Lineair (Environmental Peer Performance)

0 0,5 1 1,5 2 2,5 3 3,5 4 5 5,5 6 6,5 7 7,5 8 8,5 9 9,5 S o ci a l p ee r ra ti o

Log total assets Social Peer Performance

Lineair (Social Peer Performance)

0 0,5 1 1,5 2 5 5,5 6 6,5 7 7,5 8 8,5 9 9,5 C o rp o ra te g o ve rn a n ce p ee r ra ti o

Log total assets _{Corporate Governance Peer Performance}

(25)

25 indicate that firms with the best CSP relatively to their industry peers do not necessarily need to be the largest firms. Interestingly, environmental performance and social performance show that that the highest peer group performance appears within the 7 to 7.5 threshold level of the logarithm of total assets, corresponding with a total asset value ranging from approximately 10 to 30 billion dollars. Furthermore, a similar effect is found within the corporate governance peer performance, albeit that the distribution suggests a flat pattern between the peer performance ratio and size for the corporate governance dimension. Overall, the peer score ratio does yield supporting evidence for the positive association between CSP and size. Hence, on an industry level, the relation between size an CSP is present as well.

Conclusion

(26)

26 Discussion and recommendations

(27)

(28)

28 References

Appel, I.R., Gormley, T.A. and Keim, D.B. (2016). Passive Investors, Not Passive Owners. Journal of Financial Economics, 121(1), pp.111–141.

Artiach, T., Lee, D., Nelson, D. and Walker, J. (2010). The Determinants of Corporate Sustainability Performance. Accounting & Finance, 50(1), pp.31–51.

Barnett, M.L. and 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), pp.1304–1320.

Baron, D.P., Harjoto, M.A. and Jo, H. (2009). The Economics and Politics of Corporate Social Performance. Working Paper No. 45. Stanford University Graduate School of Business, Stanford.

Bénabou, R. and Tirole, J. (2010). Individual and Corporate Social Responsibility, Economica, 77, pp.1-19.

Bird, R., Casavecchia, L. and Reggiani, F. (2006). Corporate social responsibility and corporate performance: where to begin? Working Paper. University of Technology, Sydney & Bocconi University, Milan.

Brooks, C. (2008). Introductory econometrics for finance, Second Edition. Cambridge University Press, Cambridge.

Carrol, A.B. (1979). A three-dimensional conceptual model of corporate performance. Academic Management Review. 4(4), pp.497-505.

Chang, D. and Kuo, L.R. (2008). The Effects of Sustainable Development on Firms' Financial Performance - an Empirical Approach. Sustainable Development, 16(6), pp.365–380. Cheng, B., Ioannou, I. and Serafeim, G. (2014). Corporate Social Responsibility and Access to

Finance. Strategic Management Journal, 35(1), pp.1–23.

Dimson, E. Karakaş, O. and Li, X. (2015). Active Ownership. Review of Financial Studies, 28(12), pp.3225–3268.

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

European Commission (2017). Corporate Social Responsibility (CSR). Accessed on 21 April 2017, http://ec.europa.eu/growth/industry/corporate-social-responsibility_nl

(29)

29 Freeman, R.E. (1984). Strategic management: a stakeholder approach. Boston, Mass.:

Pitman/Ballinger.

Friedman, M. (1970). The social responsibility of business is to increase its profits. New York Times Magazine, 13 September: 32–3.

Garcia, A.S., Mendes-Da-Silva, W. and Orsato, R.J. (2017). Sensitive Industries Produce Better ESG Performance: Evidence from Emerging Markets. Journal of Cleaner Production, 150(7), pp.135–147.

Garcia-Castro, R., Arin_{̃o, M.A. and Canela, M.A. (2010). Does Social Performance Really Lead} to Financial Performance? Accounting for Endogeneity. Journal of Business Ethics, 92(1), pp.107–126.

Guenster, N., Bauer, R., Derwall, J. and Koedijk, K. (2011). The Economic Value of Corporate Eco-Efficiency. European Financial Management, 17(4), pp.679–704.

Jensen, M.C. (2010). Value Maximization, Stakeholder Theory, and the Corporate Objective Function. Journal of Applied Corporate Finance, 22(1), pp.32–42.

Kempf, A. and Osthoff, P. (2007). The Effect of Socially Responsible Investing on Portfolio Performance. European Financial Management, 13(5), pp.908–922.

Kitzmueller, M. and Shimshack, J. (2012). Economic Perspectives on Corporate Social Responsibility. Journal of Economic Literature, 50(1), pp.51–84.

Krüger, P. (2015). Corporate Goodness and Shareholder Wealth. Journal of Financial Economics, 115(2), pp.304–329.

Lourenço, I.C. and Branco, M.C. (2013). Determinants of Corporate Sustainability Performance in Emerging Markets: The Brazilian Case. Journal of Cleaner Production, 57, pp.134–141.

Matten, D., and Moon, J. (2008). “Implicit” and “explicit” CSR: A conceptual framework for a comparative understanding of corporate social responsibility. Academy of management Review, 33(2), 404-424.

Margolis, J.D. and Walsh, J.P. (2001). People and profits? : The search for a link between a company's social and financial performance. LEA's organization and management series. Mahwah, N.J.: Lawrence Erlbaum Associates.

(30)

30 McWilliams, A. and Siegel, D. (2000). Corporate Social Responsibility and Financial Performance: Correlation or Misspecification? Strategic Management Journal, 21(5), pp.603–609.

McWilliams, A. and Siegel, D. (2001). Corporate social responsibility: A theory of the firm perspective. Academy of management review, 26(1), pp.117-127.

Oikonomou, I., Brooks, C. and Pavelin, S. (2014). The Effects of Corporate Social Performance on the Cost of Corporate Debt and Credit Ratings. Financial Review, 49(1), pp.49–75. Padgett, R.C. and Galan, J.I. (2010). The Effect of R&D Intensity on Corporate Social

Responsibility. Journal of Business Ethics, 93(3), pp.407–418.

Richardson, A.J. and Welker, M. (2001). Social Disclosure, Financial Disclosure and the Cost of Equity Capital. Accounting, Organizations and Society, 26(7), pp.597–616.

S&P Dow Jones Indices (2017). U.S. Indices Methodology. Accessed on 29 April 2017,

https://us.spindices.com/documents/methodologies/methodology-sp-us-indices.pdf?force_download=true

Schäfer, H., Beer, J., Zenker, J., and Fernandes, P. (2006). Who is who in corporate social responsibility rating? A survey of internationally established rating systems that measure corporate responsibility. Bertelsmann Foundation, pp.120.

Scholtens, B. (2008). A Note on the Interaction between Corporate Social Responsibility and Financial Performance. Ecological economics: The journal of the International Society for Ecological Economics, 68(1), p.46.

Schröder, M. (2014). Financial Effects of Corporate Social Responsibility: A Literature Review. Journal of Sustainable Finance & Investment, 4(4), pp.337–350.

Thomson Reuters (2013). Thomson Reuters Corporate Responsibility Ratings (TRCRR).

Accessed on 12 April 2017,

http://financial.thomsonreuters.com/content/dam/openweb/documents/pdf/tr-com-financial/methodology/corporate-responsibility-ratings.pdf

Thomson Reuters (2012). ASSET4 ESG DATA. Accessed on 5 May 2017,

https://www.thomsonreuters.com/content/dam/openweb/documents/pdf/tr-com-financial/fact-sheet/esg-data-fact-sheet.pdf

Tirole, J. (2001). Corporate Governance. Econometrica, 69(1), pp.1–35.

(31)

31 US-SIF Foundation (2016). US Sustainable, Responsible and Impact Investing Trends 2016.

Accessed on 4 May 2017,

http://www.ussif.org/files/SIF_Trends_16_Executive_Summary(1).pdf

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

Xu, E., Yang, H., Quan, J.M. and Lu, Y. (2015). Organizational Slack and Corporate Social Performance: Empirical Evidence from China’s Public Firms. Asia Pacific Journal of Management, 32(1), pp.181–198.

Zhao, X. and Murrell, A.J. (2016). Revisiting the Corporate Social Performance-Financial Performance Link: A Replication of Waddock and Graves. Strategic Management Journal, 37(11), pp.2378–2388.

(32)

32 Appendix A

Table 10. Sin industry selection

Non-sin Sin

Aerospace Financial Admin. Prop. & Casualty Ins. Aluminium

Airlines Fixed Line Telecom. Publishing Brewers

Alternative Fuels Food Products Railroads Coal

Apparel Retailers Food Retail, Wholesale Real Estate Hold, Dev Defence

Asset Managers Footwear Real Estate Services Distillers & Vintners Auto Parts Full Line Insurance Recreational Products Exploration & Prod. Automobiles Furnishings Recreational Services Farm Fish Plantation

Banks Healthcare Providers Reinsurance Forestry

Biotechnology Heavy Construction Renewable Energy Eq. Gambling Broadcast & Entertain Home Construction Residential REITs Gas Distribution Broadline Retailers Home Improvement Ret. Restaurants & Bars General Mining Building Mat.& Fix. Hotel & Lodging REITs Retail REITs Gold Mining Bus.Train & Employment Hotels Semiconductors Integrated Oil & Gas Business Support Svs. Ind. & Office REITs Soft Drinks Iron & Steel Clothing & Accessory Industrial Machinery Software Nonferrous Metals Comm. Vehicles, Trucks Industrial Suppliers Spec. Consumer Service Oil Equip. & Services Commodity Chemicals Insurance Brokers Specialty Chemicals Paper

Computer Hardware Internet Specialty Finance Pipelines

Computer Services Investment Services Specialty REITs Plat.& Precious Metal Con. Electricity Life Insurance Specialty Retailers Tobacco

Consumer Electronics Marine Transportation Telecom. Equipment Consumer Finance Media Agencies Tires

Containers & Package Medical Equipment Toys

Delivery Services Medical Supplies Transport Services Divers. Industrials Mobile Telecom. Travel & Tourism Diversified REITs Mortgage Finance Trucking

Drug Retailers Mortgage REITs Waste, Disposal Svs. Dur. Household Prod. Multiutilities Water

(33)

33 Appendix B

Table 11. Thomson Reuters DataStream ASSET4 scoring aggregation methodology

Pillar Hierarchy Name Description Scaling Units Sector Relevancy Rule Corporate Governance PILLAR SCORE Corporate Governance Positive Percent (100=100%)

Relevant for all companies The corporate governance pillar measures a company's systems and processes, which ensure that its board members and executives act in the best interests of its long term shareholders. It reflects a company's capacity, through its use of best management practices, to direct and control its rights and responsibilities through the creation of incentives, as well as checks and balances in order to generate long term shareholder value.

Environmental PILLAR SCORE

Environmental Positive Percent (100=100%)

Relevant for all companies The environmental pillar measures a company's impact on living and non-living natural systems, including the air, land and water, as well as complete ecosystems. It reflects how well a company uses best management practices to avoid environmental risks and capitalize on environmental opportunities in order to generate long term shareholder value.

Social PILLAR SCORE

Social Positive Percent

(100=100%)

Relevant for all companies The social pillar measures a company's capacity to generate trust and loyalty with its workforce, customers and society, through its use of best management practices. It is a reflection of the company's reputation and the health of its license to operate, which are key factors in determining its ability to generate long term shareholder value.

Scores Calculation Methodology.

The ASSET4 ESG framework allows to rate and compare companies against approximately 700 individual data points, which are combined into over 250 key performance indicators (KPIs). These KPI scores are aggregated into a framework of 18 categories grouped within 4 pillars that are integrated into a single overall score. The ASSET4 ESG framework allows to rate and compare companies against approximately 700 individual data points, which are combined into over 250 key performance indicators (KPIs). These KPI scores are aggregated into a framework of 18 categories grouped within 4 pillars that are integrated into a single overall score.

What is the calculation method of your ratings?

Indicators, Categories, Pillars and Overall Score are calculated by equally weighting and z-scoring all underlying data points and comparing them against all companies in the ASSET4 universe. The resulting percentage is therefore a relative measure of performance, z-scored and normalized to better distinguish values and position the score between 0 and 100%. What is a Z-Score?

A Z Score, or "standard score" is a relative measure comparing one company with a given benchmark. It expresses the value in units of standard deviation of that value from the mean value of all companies. Among other things, this allows to create more distinction between values that otherwise might be very close together. To read more:

http://en.wikipedia.org/wiki/Standard_score. How are “Yes/No” values translated into a score?

Yes/No = 1 or 0 – these are converted into % using z-scoring. The % will depend on the number of companies that share the same value. Example: If having “Yes” is positive, it is worth a very high score if only few companies have a “Yes”. If most companies have “Yes”, it will only provide an average score.

(34)

34 Appendix C

Table 12. Environmental score as depended variable with winsorized explanatory variables

ROA ROE Tobins Q

Constant -146.383*** (4.977) -167.243*** (7.713) -204.563*** (9.972) -122.488*** (4.85) -166.94*** (7.764) -208.148*** (10.055) -154.924*** (5.37) -174.811*** (7.984) -209.227*** (10.09) Financial performance 125.883*** (8.066) 21.38*** (6.033) 9.677 (6.207) 32.993*** (3.533) 1.924 (2.303) -0.067 (2.344) 7.659*** (0.599) 2.51*** (0.534) 1.605*** (0.561) Log total assets 25.361*** (0.686) 29.893*** (1.098) 35.519*** (1.428) 22.576*** (0.683) 30.076*** (1.111) 36.153*** (1.443) 26.535*** (0.726) 30.802*** (1.123) 36.018*** (1.434) Debt to assets 9.583*** (2.366) 8.318*** (2.786) 7.4** (3.129) 2.975 (2.366) 4.53 (2.807) 3.77 (3.148) 7.291*** (2.357) 8.25*** (2.77) 7.308** (3.093) Dividend yield 256.694*** (24.893) 118.303*** (24.232) 89.205*** (25.723) 251.916*** (25.512) 117.859*** (24.43) 92.081*** (25.851) 300.179*** (25.063) 131.295*** (24.47) 98.587*** (25.942) Price to book 0.804*** (0.189) 0.382** (0.156) 0.294* (0.164) 1.096*** (0.21) 0.543*** (0.169) 0.471*** (0.177) -0.074 (0.245) -0.048 (0.195) 0.024 (0.202) Sin industry -4.449*** (1.065) -6.409** (2.688) -3.637*** (1.078) -6.379** (2.741) -2.965*** (1.067) -6.046** (2.748) R-squared 0.219 0.115 0.81 0.199 0.114 0.809 0.208 0.115 0.81 Adjusted R-squared 0.218 0.114 0.786 0.198 0.113 0.785 0.207 0.115 0.786 Log likelihood -31990.82 -27253.27 -31908.48 -27125 -32238.14 -27425.75 F-statistic 312.977*** 144.432*** 33.369*** 276.116*** 142.752*** 33.21*** 295.356*** 146.522*** 33.591*** F-test for cross section fixed

effects

24.594*** 25.262*** 25.16***

9492.538*** 9578.331*** 9632.518***

117.428*** 94.397*** 77.06***