The determinants of corporate sustainability performance
The authors thank Philip Gray, two anonymous referees and participants at the 2007 AFAANZ Conference for their helpful comments and suggestions. We also thank the Dow Jones Sustainability Index for provision of data.
Abstract
This paper investigates the factors that drive high levels of corporate sustainability performance (CSP), as proxied by membership of the Dow Jones Sustainability World Index. Using a stakeholder framework, we examine the incentives for US firms to invest in sustainability principles and develop a number of hypotheses that relate CSP to firm-specific characteristics. Our results indicate that leading CSP firms are significantly larger, have higher levels of growth and a higher return on equity than conventional firms. Contrary to our predictions, leading CSP firms do not have greater free cash flows or lower leverage than other firms.
1. Introduction
Corporate sustainability is considered to be a business and investment strategy that seeks to use the best business practices to meet and balance the needs of current and future stakeholders (Report of the United Nations World Commission on Environment and Development, 1987). This entails the complex task of providing competitive outcomes in the short-term while seeking to protect, maintain and augment the human and natural resources required in the future. Accordingly, corporate sustainability performance (CSP) measures the extent to which a firm embraces economic, environmental, social and governance factors into its operations, and ultimately the impact they exert on the firm and society.
The focus of this paper is to explore the factors that influence firms’ decisions to invest in CSP. Using a stakeholder framework, a number of hypotheses are developed that relate firm-specific characteristics to CSP. We then compare a group of high-ranking CSP firms, as proxied by their membership of the Dow Jones Sustainability World Index (DJSI, 2006), with a control sample of non-DJSI firms. Our results indicate that DJSI firms are significantly different from other firms on the dimensions of size, profitability and level of growth options. However, our finding that leading CSP firms are more profitable than conventional firms is sensitive to the specification of proxy for profitability. Only firms with high levels of profit available to shareholders (ROE) rather than high profit levels per se (ROA) appear to achieve high CSP. Contrary to our predictions, neither leverage nor the level of cash resources available to the firm is an important factor in determining CSP.
Overall, our findings indicate that size, profitability and potential for growth are associated with high levels of investment in CSP, consistent with the stakeholder management framework used in this paper. These findings support the conclusion that there are incentives for some types of firms to invest more heavily in corporate sustainability programmes because such investments help to maintain the firm’s competitive position.
The remainder of this paper is structured as follows. Section 2 describes the contribution of the paper and positions it in the relevant literature. The section concludes with a discussion of stakeholder theory and the hypothesis development. Section 3 describes the research design and results are presented in Section 4. Section 5 summarizes and concludes the paper.
2. Theoretical background and hypothesis
2.1. Literature review and contribution
Much of the extant research literature dealing with CSP has focussed on the relationship between three key factors; the level of corporate sustainability performance, corporate financial performance and the level/quality of corporate sustainability disclosure (Al-Tuwaiji et al., 2004). Of particular relevance to this paper is the stream of research examining the relation between financial performance and sustainability performance. Researchers have hypothesized that there is a negative, positive or neutral association between a firm’s CSP and its financial performance.
One perspective argues that there is a negative association between CSP and financial performance because investment in CSP is costly (Alexander and Buchholz, 1978; Becchetti et al., 2005). Consequently, those firms that invest in CSP incur additional costs such as improved employee conditions, the adoption of environmentally friendly practices, charitable donations, the promotion of community development, the development of economically depressed areas and opportunity costs from forgoing socially irresponsible investment. This perspective implies investment in CSP is contrary to the best interests of investors as it represents a re-allocation of scarce resources away from a firm’s investors to its external stakeholders (Aupperle et al., 1985; McGuire et al., 1988; Barnett, 2005).
A second perspective suggests that there is no direct association between CSP and financial performance. Ullmann (1985) argues that the relationship between CSP and financial performance is complex, and there are likely to be many intervening influences. Given the difficulty of controlling for these intervening factors, Ullmann (1985) suggests that there is insufficient theoretical support for anyone to expect a direct relationship between CSP and financial performance.
Finally, a third perspective claims that CSP is positively related to financial performance. There are three predominant views consistent with this perspective. First, researchers suggest that the financial benefits from investing in CSP exceed its costs (McGuire et al., 1988; Barnett, 2005). It is argued that CSP investment produces benefits such as enhanced employee morale, goodwill, improved relationships with bankers, investors and government and better access to capital; each of which is expected to lead to greater financial performance. Second, there is a compatible view supported by stakeholder theory which contends that CSP investment generates positive financial benefits by managing stakeholders. Third, the resource view suggests that firms that invest in CSP have superior resources (Alexander and Buchholz, 1978; Waddock and Graves, 1997; Clarkson et al., 2006). The resource view proposes that only firms with sufficient resources have the capacity to invest in CSP, and suggests that CSP is positively associated with financial performance because the types of firms that invest in CSP have greater underlying resources which produce higher financial performance.
Prior empirical research examining the association between CSP and financial performance has produced relatively mixed results. The studies by Lee (2006), Becchetti et al. (2005) and McWilliams and Siegel (2000) conclude that there is no evidence of a significant association between CSP and financial performance. Alternatively, Pava and Krausz (1996) and Cochran and Wood (1984) find weak evidence to suggest that a positive association exists. Finally, the studies by Waddock and Graves (1997), Herremans et al. (1993) and McGuire et al. (1988) conclude that there is significant evidence to suggest that CSP is positively associated with financial performance. It is likely that the variation in research methodologies contributes to the apparent ambiguity of previous findings and propagated research in the area (Cochran and Wood, 1984; Aupperle et al., 1985; Ullmann, 1985; Pava and Krausz, 1996; Barnett, 2005). Notably, across the literature there is considerable variation in the measurement of CSP and financial performance, the controls included in the research design, the specific tests employed and time periods examined.
Rather than examining whether improved financial performance is a consequence of high CSP, we focus on the motivations for managers to invest in sustainability programmes. In doing so, this paper contributes to the ongoing research debate about the benefits of CSP investment by providing insights into management incentives to achieve high levels of CSP. A clearer appreciation of the types of firms most likely to be CSP leaders will facilitate a better understanding of the likely benefits of the CSP investment.
This paper also contributes to the research literature by using a strong proxy for leading CSP; consistent membership of the DJSI across the study period. Much of the prior literature has used alternative proxies for sustainability performance (see Waddock and Graves, 1997; McWilliams and Siegel, 2000; Statman, 2000; Becchetti et al., 2005). The DJSI is advantageous because it incorporates a best-in-class methodology to recognize the leading CSP firms from each industry sector. Our high CSP group includes only those US firms which have been members of the DJSI for every year of the 2002–2006 sample period. Hence, this study has a sharper focus by looking only at firms that have made a significant and ongoing investment in sustainability activities.
2.2. Hypothesis development
Prior research suggests diverse incentives for investment in corporate social responsibility activities. However, more recent literature has posited a link between corporate social responsibility and competitive advantage (Porter and Kramer, 2006). The prominence of global issues such as climate change has brought greater public awareness of the impact of business on society at large with a consequent trend towards the broader concept of the ‘stakeholder society’ (Tirole, 2001). A firm’s stakeholders can be defined broadly as ‘any group or individual who can affect or is affected by the achievement of the firm’s objectives’ (Freeman, 1984). Stakeholder theory proposes that stakeholders ultimately control a firm’s access to scarce resources and firms must manage their relationship with key stakeholders to ensure that such access to resources is maintained (Roberts, 1992).
While there appears to be little formal analysis of the economics of the stakeholder society, there is tangible evidence of greater acceptance of this view. In Europe, for example, German companies with greater than 500 employees must have a two-tiered board structure, with the higher board consisting of representatives from major stakeholder groups such as banks, customers, suppliers and workers (Tirole, 2001). This trend towards considering the needs of the firm’s wider stakeholders is also evidenced by the emergence of the socially responsible investor, who typically applies a social responsibility screen or evaluation process as part of their investment decision-making process.1 There is also mounting evidence of direct pressures on corporations to act in a more socially responsible fashion. CSP issues have invaded the corporate boardroom, with Porter and Kramer (2006) reporting that in the USA in 2005, 360 different CSP-related shareholder resolutions were filed on issues ranging from labour conditions to global warming.
In Australia, regulators and financial markets are also beginning to recognize the importance of corporate sustainability issues. For example, Principle 3 of the ASX Corporate Governance Council Guidelines recommends that companies ‘Promote ethical and responsible decision-making’, while Principle 7 requires companies to disclose ‘a summary of the company’s policies on risk oversight and management of material business risk. These risks may include but are not limited to: operational, environmental, sustainability, compliance, strategic, ethical conduct, reputation or brand, technological, product or service quality, human capital, financial reporting and market-related risks’ (ASX Corporate Governance Council, 2007).
The discussion above suggests that investment in CSP can be value adding. Corporate sustainability spending can maintain/enhance the reputation of the firm, forestall stakeholder and regulatory action and make the firm a more appealing investment to the socially responsible investor. However, not all firms invest in corporate sustainability programmes and those that do invest do so at various levels.
If a firm’s investment in sustainability is determined by its stakeholder environment, then differences in CSP across firms imply differences in firm-specific characteristics related to the range and importance of the firm’s stakeholder demands. Consistent with Ullmann (1985), we expect that stakeholder power will be positively related to sustainability performance. If stakeholders control resources critical to the firm, then the firm is more likely to respond to those stakeholder demands. Conversely, stakeholders with little power over resources are likely to receive little attention from the firm. The nature of the firm’s response to external stakeholder pressures will depend on its strategic posture as well as its financial performance (Bansal and Roth, 2000). In times of low profitability and high debt, financial stakeholders will receive priority over social demands. In addition, economic performance will influence the extent to which the firm has the financial resources to participate in costly sustainability programmes.
The discussion so far suggests that firm-specific characteristics related to stakeholder power, strategic posture and financial performance may explain variations in the levels of investment in corporate sustainability. In particular, firms with high levels of CSP are expected to be larger, more profitable, have lower gearing, have higher levels of growth and have greater cash resources at their disposal.
2.2.1.Size
Firm size is likely to be an important determinant of CSP. Large firms are more visible politically and so draw greater attention from the general public, government and other stakeholders. Large firms are more likely to create correspondingly larger social problems because of the sheer scale and prominence of their activities. For example, pollution emissions to some extent will be a function of the size of operations. Size is also likely to influence the firm’s strategic response to stakeholder demands. A passive or even negative response is unlikely to be a successful strategy for big firms which face greater public scrutiny and external pressures. Furthermore, large firms are more likely to realize economies of scale in corporate sustainability activities. For example, the fixed cost element of reducing waste emissions from manufacturing activities can be spread across a larger production volume for large firms. It follows that:
H1: Firm size and corporate sustainability performance are positively related.
2.2.2.Leverage
The level of debt in the firm’s capital structure provides a measure of the relative importance of the firm’s financial stakeholders. The stakeholder view of the firm proposes that the firm has numerous claimants, both financial and non-financial. However, these stakeholder groups have varying degrees of power over the resources required by the organization (Ullmann, 1985). As suppliers of capital to the firm, debtholders are a powerful stakeholder group and management is more likely to address their concerns than those of less powerful stakeholders, such as employees or the community at large. Hence, we expect that as the firm’s leverage increases, so will its emphasis on the claims of the debtholders over those of less powerful claimants. This leads to the second hypothesis:
H2: Leverage and corporate sustainability performance are negatively related.
2.2.3. Financial capacity
Free cash flow
The earlier discussion described the incentives that can exist for firms to invest in corporate sustainability programmes, consistent with the stakeholder view of the firm. Stakeholder theory also recognizes that faced with scarce resources, management may have to prioritize the needs of its various stakeholders. In periods of low profitability, economic demands will have priority over social claimants (Ullmann, 1985). Hence, the firm’s economic performance influences its capability to undertake corporate sustainability programmes to fulfil social demands. Free cash flow is a measure of the firm’s liquidity and financial slack. High levels of free cash flow indicate that the firm has sufficient financial capacity to invest in sustainability programmes without sacrificing the demands of economic claimants. It follows that:
H3: Free cash flow and corporate sustainability performance are positively related.
Profitability
Just as investment in CSP requires some financial ‘slack’ in the form of free cash flow, levels of profitability will likewise influence the investment decision. When economic performance is high, the firm faces less pressing demands from its financial stakeholders and so has the financial capacity to invest in programmes with social, as well as economic, merit. High levels of profitability allow the firm to meet analyst and shareholder expectations and still retain the ability to meet social stakeholder demands through expenditure on CSP. By contrast, during times of low profitability, the pressure will be on management to reduce costs and maximize economic returns to financial stakeholders. Thus we expect that:
H4: Profitability and corporate sustainability performance are positively related.
2.2.4. Innovation and product differentiation
The level of growth options in the firm’s investment opportunity set is also an indicator of CSP. Firms with a high level of investment in tangible assets have fewer incentives to pursue innovation or product differentiation strategies that incorporate sustainability principles because of their sunk cost investment in existing production technologies. Conversely, the higher the level of growth options in the firm’s asset mix, the more likely it is that the firm is able to incorporate sustainability principles into its competitive strategy. A measure of growth options will also capture the extent of each firm’s investment in R&D, which is also expected to be associated with CSP (McWilliams and Siegel, 2000).2 Hence, we expect that:
H5: Growth options and corporate sustainability performance are positively related.
3. Empirical tests
3.1. Measurement of variables
Leading CSP firms are US firms that are included in the DJSI every year for the 5 year sample period 2002–2006. Firms which are persistently included in the DJSI have made a more substantial financial and strategic investment in CSP than firms that are only occasionally included. Consequently, persistent inclusion in the DJSI represents a material CSP investment and is therefore a strong proxy to identify leading CSP firms.
Firms included in the DJSI consist of the top 10 per cent of leading sustainability firms for each industry as derived from the 2500 largest companies in the Dow Jones Global Index. The DJSI is frequently argued as being one of the world’s best sustainability indices (Lee, 2006). Firm sustainability is evaluated by the Sustainable Asset Management (SAM) Group, an independent asset management company involved in sustainability research. The SAM group assesses the opportunities and risks associated with each firm’s economic, environmental and social development. The evaluation is supported by a thorough examination of each firm’s investment in CSP, and the firm ranking and selection is independently verified by PricewaterhouseCoopers. As a best-in-class index, the DJSI is preferred because the impact of CSP investment can be examined for firms from all industries. The integrity of the DJSI as a proxy for CSP is highlighted in the report by Beleo et al. (2004), which recommends the SAM Group research as best practice in corporate social responsibility research.
Our sample of conventional firms are those firms comprising the S&P 500 index firms and represent the same GICS industry group as the leading CSP firms. Conventional firms are required not to have been included in the DJSI during the entire sample period, thereby representing an ongoing lack of investment in CSP. The S&P 500 is an appropriate index for the conventional firms as all of the DJSI firms, being large firms, are also included in the S&P 500 index. A dichotomous dependent variable (CSP) is used to indicate leading CSP firms. The dichotomous variable is coded one for firms that have been persistently included in the DJSI for the full sample period (i.e. leading CSP firms), and zero otherwise.
3.2. Sample description
The DJSI annual reviews provided by the SAM group demonstrate that 107 unique US firms represented the DJSI during the sample period 2002 to 2006. Of these, only 26 firms are included in the index every year for the sample period, whilst 81 firms are occasionally included. This leads to a maximum possible sample of 130 firm-year observations for the leading CSP firms. For the sample of conventional firms, there are 1381 firm-year observations, comprising all firms in the S&P 500 for each year that are not in the DJSI in any year of the study period.
Table 1 provides a break-down of the final sample of CSP and conventional firms by industry. Table 1 shows that 15 GICS industry groups are represented in the sample, with Capital Goods the most dominant at 10.6 per cent. The smallest representation at 1.1 per cent is the Household and Personal Products industry group. It is apparent from Table 1 that a number of industries are not represented in the sample.3 The DJSI selects the leading sustainability firms for each industry from a global population. Consequently, with the exception of industries that are specifically excluded (Banking and Diversified Financials), non-represented industry groups arise because the sustainability industry leader is not a US firm or if the US firm was not included in the DJSI for the entire 5 year period.
| GICS industry group | CSP | CSP% | Conv. | Conv. % | Total | Total % | |
|---|---|---|---|---|---|---|---|
| 1010 | Energy | 5 | 3.8 | 135 | 9.8 | 140 | 9.3 |
| 1510 | Materials | 5 | 3.8 | 105 | 7.6 | 110 | 7.3 |
| 2010 | Capital Goods | 20 | 15.4 | 140 | 10.1 | 160 | 10.6 |
| 2030 | Transportation | 5 | 3.8 | 40 | 2.9 | 45 | 3.0 |
| 2520 | Consumer Durables and Apparel | 5 | 3.8 | 90 | 6.5 | 95 | 6.3 |
| 2530 | Consumer Services | 10 | 7.7 | 46 | 3.3 | 56 | 3.7 |
| 2540 | Media | 10 | 7.7 | 55 | 4.0 | 65 | 4.3 |
| 2550 | Retailing | 5 | 3.8 | 125 | 9.1 | 130 | 8.6 |
| 3030 | Household & Personal Products | 5 | 3.8 | 12 | 0.9 | 17 | 1.1 |
| 3510 | Health Care Equipment & Services | 10 | 7.7 | 115 | 8.3 | 125 | 8.3 |
| 3520 | Pharmaceuticals, Biotechnology | 10 | 7.7 | 89 | 6.4 | 99 | 6.6 |
| 4510 | Software & Services | 10 | 7.7 | 102 | 7.4 | 112 | 7.4 |
| 4520 | Technology, Hardware & Equipment | 10 | 7.7 | 109 | 7.9 | 119 | 7.9 |
| 4530 | Semiconductors & Semi-Equipment | 15 | 11.5 | 74 | 5.4 | 89 | 5.9 |
| 5510 | Utilities | 5 | 3.8 | 144 | 10.4 | 149 | 9.9 |
| 130 | 100 | 1381 | 100 | 1511 | 100 | ||
- The sample breakdown by GICS industry code for the pooled sample period 2002–2006 is presented. CSP Firms are US-listed firms that have been included in the DJSI for each year during the sample period. Conventional firms are US-listed firms drawn from the S&P 500 index that are not in the DJSI throughout the examination period.
3.3. Test methodology
()
is measured as a 0/1 dummy variable where 1 if firm i represents a leading CSP firm from the DJSI index or 0 if representing a conventional firm from the S&P 500; 
is size measured as the natural log of total assets for firm i in year t; 
is leverage measured as total debt divided by total assets for firm i in year t; 
is financial capacity measured as free cash flow divided by net sales for firm i in year t; 
(the base model) representing profitability is return on assets measured as EBIT divided by total assets for firm i in year t. As a sensitivity check, ROA is replaced with Return on Equity (ROE), which is measured as net income before extraordinary items divided by common equity. 
is growth options measured as the price-to-book ratio for firm i in year t. Industry and Year are dummy variables controlling for industry and temporal effects. Accounting data are obtained from the Research Insight database.44.Results
4.1. Univariate tests
Table 2 presents descriptive statistics for the sub-samples of the 130 leading CSP firm-year and 1381 conventional firm-year observations as provided in panels A and B, respectively. ROA, ROE, PB and FCF display considerable dispersion and skewness with the influence of some extreme values. The financial characteristics are particularly dispersed and skewed for the conventional firms. Given that the skewness and kurtosis coefficients for a normal distribution are zero and three, respectively, the variables shown in Table 2 suggest that ROA and PB in particular are unlikely to conform to the distributional assumptions of parametric tests. As winsorized regressors are generally more robust to outliers, each of the independent variables is winsorized to within three standard deviations from the mean to approximate a normal distribution (Dixon, 1960).5 The variables are winsorized based on the mean and standard deviation of the pooled sample of 1511 firm-year observations. The winsorized variables do little to help the distributional properties of ROE in either the CSP firms or conventional firms.6 However, the distributions of the other variables, particularly PB, are markedly less skewed after winsorizing.
| Variable | Mean | Median | SD | Min | Max | Skew | Kurt |
|---|---|---|---|---|---|---|---|
| Panel A: CSP firms (n = 130) | |||||||
| ln TAw | 9.668 | 9.790 | 1.138 | 7.096 | 11.818 | −0.156 | −0.559 |
| ROA w | 5.810 | 6.812 | 8.927 | −32.760 | 19.805 | −1.988 | 6.209 |
| ROE w | 18.089 | 18.577 | 25.664 | −84.393 | 139.780 | 0.643 | 8.774 |
| LEV w | 21.664 | 17.710 | 14.314 | 0.000 | 60.232 | 0.829 | 0.253 |
| PB w | 5.541 | 3.681 | 6.994 | 0.000 | 53.181 | 4.547 | 25.279 |
| FCF w | 0.066 | 0.063 | 0.096 | −0.294 | 0.398 | 0.491 | 4.663 |
| Panel B: Conventional firms (n = 1381) | |||||||
| ln TAw | 8.834 | 8.704 | 1.096 | 5.535 | 12.277 | 0.259 | −0.367 |
| ROA w | 5.904 | 5.686 | 7.430 | −32.760 | 34.133 | −1.445 | 7.442 |
| ROE w | 14.032 | 14.133 | 22.361 | −109.860 | 139.780 | 0.065 | 13.924 |
| LEV w | 21.178 | 20.740 | 14.659 | 0.000 | 65.110 | 0.328 | −0.473 |
| PB w | 4.360 | 2.885 | 10.958 | 0.000 | 263.480 | 19.892 | 456.499 |
| FCF w | 0.063 | 0.053 | 0.108 | −0.305 | 0.426 | 0.035 | 2.324 |
| Panel C: Difference in mean values for CSP firms and conventional firms | ||||||
|---|---|---|---|---|---|---|
| t-test | Wilcoxon signed ranks test | |||||
| Mean diff | t-stat | p-value | Mean diff | t-stat | p-value | |
| ln TAw | 0.834 | 8.263 | <0.001 | 1.086 | 7.646 | <0.001 |
| ROA w | −0.094 | −0.134 | 0.893 | 1.126 | 0.984 | 0.325 |
| ROE w | 4.057 | 1.938 | 0.053 | 4.444 | 3.051 | 0.002 |
| LEV w | 0.486 | 0.362 | 0.717 | −3.030 | −0.068 | 0.946 |
| PB w | 1.181 | 1.205 | 0.228 | 0.796 | 3.766 | <0.001 |
| FCF w | 0.003 | 0.297 | 0.767 | 0.001 | 1.184 | 0.236 |
- Panel A presents the descriptive statistics for the winsorized variables using the pooled sample of 130 CSP firm-year. Panel B presents the descriptive statistics for the pooled sample of 1381 firm-year conventional firms. Panel C presents the t-test and Wilcoxon signed ranks test for differences in mean/median winsorized variables for the CSP and conventional firms. Variables defined and measured as follows: ln TA is the natural log of total assets; ROA is return on assets measured as EBIT divided by total assets multiplied by 100; ROE is return on equity measured as net income before extraordinary items divided by common equity multiplied by 100; LEV is leverage measured as total debt divided by total assets multiplied by 100; PB is the price-to-book ratio; FCF is free cash flow divided by net sales. All independent variables have been winsorized to within three standard deviations from their pooled sample (1511 firm-year) mean. The subscript ‘w’ denotes that the variable has been winsorized. Significance at p ≤ 0.05 is highlighted in bold.
Firm characteristics of the leading CSP and conventional firm sub-samples are statistically compared by tests for differences in mean values with results presented in Panel C of Table 2. The t-tests indicate that leading CSP firms are significantly larger (ln TAw) and are more profitable (ROEw) than conventional firms, thereby providing initial support for hypotheses 1 and 4, respectively. However, the t-tests indicate that the difference in leverage (LEVw), growth opportunities (PBw), level of cash resources (FCFw) and ROAw is not significant.
The nonparametric Wilcoxon signed ranks tests also shown in Table 2, Panel C, produce similar results, with CSP firms once again differing from conventional firms on the dimensions of size and profitability, where profitability is measured by ROEw. Levels of growth opportunities (PBw) also reaches statistical significance in the Wilcoxon signed ranks test, in support of hypothesis 5. Untabulated results for the t-tests and Wilcoxon Signed Ranks tests conducted on the raw variables are qualitatively consistent with the winsorized variables. While at first glance the significant result for ROE but not ROA appears anomalous, it is consistent with stakeholder explanations of CSP. ROA is calculated on the basis of earnings before considering capital structure and it therefore represents the return available to all the financial stakeholders of the firm.
By contrast, ROE represents the return available to shareholders after considering debt and tax claimants. Stakeholder theory suggests that firms will prioritize stakeholder claims such that financial stakeholders will take priority over social stakeholders. It follows that it is the level of resources available after paying the highest ranking financial stakeholders that will have the greatest influence on the firm’s CSP and this is captured more precisely by ROE than by ROA. Thus, while CSP and conventional firms do not appear to differ significantly with respect to the extent of debt in their capital structure, leading CSP firms generate significantly higher levels of profit after tax and leverage adjustments when compared with conventional firms.
Table 3 presents the Pearson Correlation Coefficients and Spearman Rho’s correlations. For brevity, and in view of the earlier discussion about the variable distributions, only results for the winsorized variables are reported.7 For the purpose of the regressions reported below, the independent regressors, whilst showing some indications of collinearity, have no pairwise correlation coefficients in excess of 80 per cent, indicating that the threat of multicollinearity is limited.8 The highest coefficient (Spearman’s ρ) at 0.589 is that between PBw and ROAw.
| Variable | ln TAw | ROA w | ROE w | LEV w | PB w | FCF w |
|---|---|---|---|---|---|---|
| ln TAw | 1 | −0.264 (<0.001) | −0.079 (0.002) | 0.315 (<0.001) | −0.318 (<0.001) | −0.233 (<0.001) |
| ROA w | −0.145 (<0.001) | 1 | 0.831 (<0.001) | −0.432 (<0.001) | 0.589 (<0.001) | 0.409 (<0.001) |
| ROE w | −0.052 (0.043) | 0.755 (<0.001) | 1 | −0.122 (<0.001) | 0.582 (<0.001) | 0.211 (<0.001) |
| LEV w | 0.290 (<0.001) | −0.312 (<0.001) | −0.019 (0.467) | 1 | −0.295 (<0.001) | −0.446 (<0.001) |
| PB w | −0.128 (<0.001) | 0.135 (<0.001) | 0.385 (<0.001) | 0.012 (0.632) | 1 | 0.395 (<0.001) |
| FCF w | −0.190 (<0.001) | 0.348 (<0.001) | 0.182 (<0.001) | −0.435 (<0.001) | 0.075 (0.003) | 1 |
- The Pearson’s correlation coefficients and Spearman rho pairwise correlations are presented. Spearman rho values are reported above the diagonal and Person correlation coefficients are reported below. Variables defined and measured as follows: ln TA is the natural log of total assets; ROA is return on assets measured as EBIT divided by total assets; ROE is return on equity measured as net income before extraordinary items divided by common equity; LEV is leverage measured as total debt divided by total assets; PB is the price-to-book ratio and FCF is free cash flow divided by net sales. All independent variables have been winsorized to within three standard deviations from their pooled sample mean. The subscript ‘w’ denotes that the variable has been winsorized. Correlation coefficients are reported with the p-value using a two-tailed test of significance shown in parentheses. Significance at p ≤ 0.05 is highlighted.
To summarize the univariate test results, firm size appears to be the most important factor associated with high CSP, in support of hypothesis 1. Profitability as proxied by ROEw is also significantly related to CSP, consistent with hypothesis 4. Weaker support was found for hypothesis 5 in that the level of growth options, as proxied by PBw, is also significantly associated with CSP. However, this is only in the nonparametric testing. On a univariate basis, CSP firms do not significantly differ from conventional firms on the basis of return on assets, free cash resources or leverage.
The univariate tests permit observation of differences in individual variables between CSP and conventional firms. The fixed effects probit model described in the next section reports the influence of the explanatory variables on CSP in a multivariate setting.
4.2. Multivariate tests
The hypotheses are tested in a multivariate analysis by estimating the fixed effects probit model shown in equation (1) on the sample of leading CSP and conventional firms. The impact of factors correlated with industry and the year is investigated by estimating the model with industry and year fixed effects. The fixed effects model controls for the influence of omitted correlated variables that are associated with industry or the time period. Equation (1) is estimated using both the raw variables and winsorized variables. The results are presented in Table 4 below.
| Variable | Pred sign | ROA with fixed industry(1) | ROAw with fixed industry(2) | ROAw with fixed industry & Year(3) | ROE with fixed industry(4) | ROEw with fixed industry(5) | ROEw with fixed industry & year(6) |
|---|---|---|---|---|---|---|---|
| Intercept | −6.558 (<0.001) | −6.733 (<0.001) | −6.713 (<0.001) | −6.873 (<0.001) | −7.018 (<0.001) | −6.998 (<0.001) | |
| ln TA | + | 0.469 (<0.001) | 0.486 (<0.001) | 0.495 (<0.001) | 0.504 (<0.001) | 0.511 (<0.001) | 0.525 (<0.001) |
| LEV | − | 0.264 (0.553) | 0.188 (0.678) | 0.154 (0.736) | 0.080 (0.852) | 0.114 (0.792) | 0.009 (0.983) |
| FCF | + | −0.471 (0.361) | −0.747 (0.218) | −0.797 (0.190) | −0.337 (0.518) | −0.681 (0.261) | −0.708 (0.243) |
| ROA | + | 0.530 (0.442) | 0.531 (0.506) | 0.748 (0.360) | |||
| ROE | + | 0.300 (0.002) | 0.675 (0.010) | 0.758 (0.004) | |||
| PB | + | 0.015 (0.818) | 0.802 (0.017) | 0.800 (0.017) | 0.002 (0.983) | 0.4069 (0.305) | 0.364 (0.366) |
| McFadden R2 | 0.158 | 0.164 | 0.166 | 0.171 | 0.175 | 0.179 | |
| LR-ratio (p-value) | 140 (<0.001) | 145 (<0.001) | 147 (<0.001) | 150 (<0.001) | 153 (<0.001) | 157 (<0.001) |
- The estimates obtained from probit model (1) using a pooled sample of 1511 firm-year observations are presented. CSP is a dummy variable assigned 1 if the firm is a DJSI firm and 0 if a conventional (S&P 500) firm. ln TA is the natural log of total assets; LEV is leverage measured as total debt divided by total assets; FCF is free cash flow divided by net sales; ROA is return on assets measured as EBIT divided by total assets; ROE is return on equity measured as net income before extraordinary items divided by common equity; P/B is the price-to-book ratio. Winsorized regressors are the raw variables winsorized to within three standard deviations from their pooled sample mean. The subscript ‘w’ denotes that the variable has been winsorized. Parameter estimates are reported with their p-value using a two-tailed test of significance in parentheses. Significance at p ≤ 0.05 is highlighted. Unreported results for the industry dummy and year dummy are available from the authors.
The base model (ROA) is first estimated using raw variable regressors and fixed industry effects, with the results presented in column 1. To examine the impact of outliers on the results, equation (1) is then re-estimated using the winsorized regressors, with the results shown in column 2. Column 3 shows the results of the ROA model using winsorized regressors and controlling for both industry and temporal fixed effects. Columns 4–6 of Table 4 repeat this analysis but employ our alternative proxy for profitability, ROE, rather than ROA.9 The following discussion focuses on the results for the winsorized regressors as in each case these models yield the highest McFadden R2.
The results summarized in Table 4 indicate that the coefficient for size (ln TAw) is positive and significant in all tests. This provides strong support for hypothesis 1 which predicts that firm size and CSP are positively related. Firm size and CSP are positively and significantly associated with CSP for both the ROA and ROE models. The coefficients for leverage (LEVw) and free cash flows (FCFw) are statistically insignificant. On this basis, the results of the fixed effects model provide no support for hypotheses 2 or 3.
To test hypothesis 4, we use two alternative proxies for profitability, ROA and ROE.Table 4 shows that the coefficient for β4 is positive and significant at p < 0.05 only for the ROE and ROEw models and these results are consistent with the univariate test results reported earlier. The multivariate analysis therefore suggests that once industry and temporal effects are controlled for, CSP and profitability are positively related, when profitability is measured after taking leverage and tax effects into account.
Finally, the fixed effects model indicates mixed support for hypothesis 5. Growth (PBw) is positively and significantly associated with high CSP in the ROA model, but this effect is statistically insignificant when ROE is used as the proxy for profitability. Our univariate tests indicate support for hypothesis 5 only in the nonparametric testing. Taken together, these results for hypothesis 5 suggest that our proxy for growth (PBw) is sensitive to both the inclusion of outliers and the choice of proxy for profitability. On this basis, we conclude only limited support for hypothesis 5.
Overall, the fixed effects analysis confirms a positive relationship between CSP and firm size (H1), profitability as proxied by return on equity (H4) and level of growth options (H5); although the latter is only prevalent in the ROA models. There is no support for hypotheses 2 and 3, which predict that leading CSP firms will have lower leverage (H2) and greater cash resources (H3).10 Interestingly, support for hypothesis 4 is sensitive to the specification of the profitability proxy; with statistically significant findings being limited to return on equity measures. This result is consistent with stakeholder theory as it suggests that it is the level of profit available to shareholders after considering higher ranking financial claimants rather than the level of profitability per se that drives CSP.
4.3. Sensitivity analysis
4.3.1.Endogeneity
To ensure that the results reported above are robust across different model specifications, we undertook a number of additional sensitivity analyses. The first of these analyses addressed the potential issue of endogeneity. The problem of endogeneity, that is, where the disturbance term of the econometric model is correlated with one or more of the explanatory variables, can cause a bias in the results produced, especially for OLS estimation (Wooldridge, 2002). The relation between CSP and at least some of the explanatory variables used in this study is likely to be endogenous, thereby making it difficult to interpret our results due to the bias caused by correlated omitted variables or reverse causation. The fixed effects model reported in Table 4 addresses some of these concerns because it controls for unobservable firm heterogeneity (Nikolaev and van Lent, 2005). Nonetheless, the results of the fixed effects model may still be subject to endogeneity arising from other sources.
In an untabulated analysis, we re-estimated equation (1) using a pooled probit model with no controls for industry or temporal effects, with the aim of assessing the extent to which the fixed effects model is correcting for endogeneity. The fixed effects model clearly dominated the standard pooled probit, with universally higher McFadden R2 and LR-Ratio values for the fixed effects model. The results for the standard probit model are also highly consistent with those obtained for the fixed effects model.11 Overall, we conclude that endogeneity does not appear to be a serious concern for our results given the similarities produced by the standard and fixed effect models, respectively.
4.3.2. Cross-sectional estimation
To ensure that our results are not driven simply by cross-sectional variation in CSP, equation (1) is also estimated using the average of the explanatory variables over the sample period.12 The untabulated results when estimating the probit model using the average explanatory variables is qualitatively consistent with the prior analysis with firm size and level of growth options positive and significant at p < 0.01.13 However, profitability, measured by either ROA or ROE, is not a significant determinant of CSP in this cross-sectional analysis. This result suggests that the panel estimation produces insights not apparent in a cross-sectional analysis. The results shown in Table 4 indicate that across time, it is realized growth, in the form of profits available to shareholders, rather than growth options, that seems to be the more important determinant of CSP.
4.3.3. Ordered probit analysis
The analysis thus far has elected to include only those firms that are consistently included/excluded from the DJSI over the full period, thereby allowing us to better capture a firm’s ongoing commitment, or lack thereof, to CSP. We extend our analysis to examine whether our results are impacted when including those firms that are only occasionally included in the DJSI.14 If there is a definitive relation between CSP and our various independent variables as hypothesized, then we would expect that this relation would continue when employing an ordered probit analysis. The interpretation of the output from an ordered probit is similar to that of a standard probit. Therefore, a positive and significant coefficient returned for one of the independent variables (e.g. ROE) is associated with an increase in the likelihood of a similar increase in the dependent variable (e.g. CSP). If the relation between CSP and our independent variables does not follow a particular order; then we would expect that many of our prior findings would not be robust to this additional analysis.
To examine the impact on our earlier results if we include the inconsistent CSP firms, we extended our sample of leading CSP firms to include each of the 81 firms that are only occasionally included in the DJSI. This process increased our sample to 2063 firm-year observations and an additional four industries (GICS 2020, 2510, 3010 and 3020). In an untabulated analysis, we re-examined equation (1) but now re-coded the dependent variable such that there is a natural ordering of CSP whereby a firm that is always, occasionally and never included in the DJSI is coded 2, 1 and 0, respectively. In the light of the CSP ordering of the firm data, we examined the relation between CSP and that of our independent variables employing an ordered logit analysis on the pooled sample of leading CSP, occasional and conventional firms (Greene, 2003).15 We continued to control for possible industry and year fixed effects.
The (untabulated) results of this analysis are generally consistent with those obtained when firms only occasionally included in the DJSI were excluded. A positive relation between firm size and CSP continues to be a dominant factor in our models, providing further support for hypothesis 1. Growth, as proxied by PB is likewise positive and significant for both the ROAw and ROEw models. Profitability (ROAw and ROEw) does not achieve conventional significance levels in the ordered probit analysis. However, the expanded sample introduced high levels of correlation between PB, profitability and free cash flows, with consequent difficulties created in the interpretation of the ordered probit results.
5. Summary and conclusions
This study has investigated the firm-specific factors associated with high corporate sustainability performance. Using the Dow Jones Sustainability Index to identify leading CSP firms, we develop and test a number of explanations for investment in corporate sustainability programmes. Our results indicate that firm size is strongly and consistently associated with high levels of CSP. Capacity for growth also appears to be a significant, although providing a somewhat weaker and less direct, influence on CSP. Contrary to our expectations, neither the level of cash resources available to the firm or its leverage are associated with leading CSP firms. We also find that leading CSP firms are more profitable when compared with conventional firms. However, this result is limited to ROE as the measure of profitability.
The implications of our results are summarized as follows. First, CSP leaders are most likely to be the largest firms in each industry. Consistent with the arguments underlying hypothesis 1, large firms are more visible, thereby drawing the attention of a wider range of external stakeholders. A passive strategic posture on sustainability issues is less feasible for large firms because they create correspondingly larger social problems and are subject to more scrutiny from both financial regulators and social stakeholders when compared with smaller firms. Large firms are also better placed to realize economies of scale in the implementation of sustainability programmes.
Our results also suggest some evidence that capacity for growth is associated with high CSP, especially when all DJSI firms are included in the analysis. This finding supports hypothesis 5 and is consistent with previous research that finds that CSP and product differentiation and innovation are related. High growth firms have more opportunities to implement sustainability principles into their expanding operations. Low growth firms on the other hand have already made a sunk cost investment in existing production technologies and have less scope to pursue product differentiation or innovation strategies that incorporate sustainability principles.
The study finds no support for hypothesis 2, which predicts that high CSP firms will have lower leverage than other firms. Indeed, the direction of the observed relationship was mixed at best and dominated by statistically insignificant levels. The level of cash resources available to the firm is likewise not associated with CSP. The results for hypothesis 3 show that the free cash flow and CSP association is not statistically significant.
Our results also indicate a persistent and significant positive relationship between CSP and profitability, but this finding is sensitive to the measurement of profitability as return on equity, rather than as return on assets. This result is consistent with stakeholder theory as it suggests that it is the level of profit available to shareholders after considering higher ranking financial claimants rather than the level of profitability per se that drives CSP.
The findings of this study are important to the ongoing debate about the benefits of corporate sustainability performance. Much of this debate has centred on the financial consequences of corporate investment in sustainability activities. By contrast, the current study focuses on the incentives for managers to achieve high levels of CSP, by examining the factors associated with leading CSP firms. This shift in focus allows a deeper consideration of the factors that influence the decision to adopt sustainability principles and therefore provides greater insight into the likely financial consequences of investment in CSP.
