Does corporate social responsibility affect corporate tax aggressiveness? This manuscript is a previous version of an article published in the Journal of Cleaner Production (2015), copyright Elsevier, available online at: http://www.sciencedirect.com/science/article/pii/S0959652615006149 (doi: 10.1016/j.jclepro.2015.05.059)
Laguir Issama, Raffaele Staglianòb, Elbaz Jamalc a
Montpellier Business School, Montpellier Research in Management (MRM), Montpellier, 2300 Avenue des Moulins, 34185, Montpellier, France b
Montpellier Business School, Montpellier Research in Management (MRM), 2300 Avenue des Moulins, 34185, Montpellier, France c
Ecole Supérieure de Technologie (EST) d’Agadir, Ibn Zohr University, Rue Oued Ziz, Agadir, Morocco ABSTRACT Recent years have seen a considerable increase in the literatures concerning the separate areas of corporate social responsibility and corporate tax aggressiveness. However, comparatively little scholarly attention has been paid to the link between the two. This paper examines how the different activities of corporate social responsibility affect corporate tax aggressiveness. A structural model was tested using partial least squares regression to determine whether the relationships between corporate social responsibility dimensions and tax aggressiveness are positive or negative. The results indicate that a firm’s tax aggressiveness depends on the nature of its corporate social responsibility activities. Notably, the study demonstrates that greater the activity in the social dimension of corporate social responsibility, the lower the level of corporate tax aggressiveness will be, whereas high activity in the economic dimension is associated with a high level of tax aggressiveness. These results extend the prior literature concerning the relationship between corporate social responsibility and tax aggressiveness and suggest that the nature of the relationship, whether negative or positive, tends to differ across the corporate social responsibility dimensions. Overall, the paper strongly supports the current literature and argues that the dimensions of corporate social responsibility should not be aggregated into a single measure because interesting and explanatory information is lost when such a method is used, especially with regard to an issue like corporate tax aggressiveness.
Corresponding author: Raffaele Staglianò, Montpellier Business School and Montpellier Research in Management, 2300 Avenue des Moulins, 34185 Montpellier Cedex 4, France. Tel.: +33(0)467102695 (Fax: +33(0)467455650). E-mail:
[email protected].
1
Keywords: Corporate social responsibility, Tax aggressiveness, Shareholders, Stakeholders, PLS-SEM
1.
Introduction
In the past decade, academic researchers and professionals alike have given particular attention to both corporate tax aggressiveness and corporate social responsibility (CSR) (e.g., Chen et al., 2010; Desai and Dharmapala, 2006; Lanis and Richardson, 2012). However, despite the growing number of studies in the separate areas of tax aggressiveness and CSR, few studies have examined the link between them and those that have offered contradictory results (e.g., Hanlon and Heitzman, 2010; Landry et al., 2013; Lanis and Richardson, 2015). The goal of tax aggressive policies is to reduce corporate taxes (e.g., Chen et al., 2010; Frank et al., 2009; Lanis and Richardson, 2012). In this paper, and in line with existing empirical tax research (e.g., Lanis and Richardson, 2012), tax aggressiveness is defined as encompassing all tax planning activities, whether legal, illegal, or falling into the gray area. Tax aggressiveness, therefore, does not imply improper activity. Moreover, it should be noted that the terms tax aggressiveness, tax avoidance, and tax management can be used interchangeably (e.g., Chen et al., 2010; Lanis and Richardson, 2012). Studies have shown that tax aggressiveness can reduce corporate costs and increase shareholder wealth (e.g., Hanlon and Heitzman, 2010). Thus, to determine just how aggressive they should be, firms need to trade off the marginal benefits of managing taxes against the marginal costs of doing so (Chen et al., 2010). One of the marginal benefits is greater tax savings, whereas the marginal costs include the potential penalties imposed by tax administrations, implementation costs (time/effort and transaction costs of implementing tax transactions), and the agency costs that inevitably accompany tax aggressive activities. Other studies suggest that firms that use tax shelters are socially irresponsible (Lanis and Richardson, 2012), as the payment of corporate taxes helps ensure the financing of public goods. Thus, a corporation’s tax aggressive policies may have a negative effect on society (Freedman, 2003). Under any of the above conditions, tax decisions are indicative of firm characteristics or management behavior. Previous studies show that CSR policies have an impact on firm decisions (Windsor, 2009) and firm performance (Agudo Valiente et al., 2012). However, despite the substantial contributions to the literature in recent decades, there is no consensual definition of CSR (Van Marrewijk, 2003), one reason being that the concept of CSR has undergone many stages of evolution. The European Commission (2011) nevertheless defined CSR as “actions by companies over and above their legal obligations towards society and the environment. From the traditional agency theory perspective, “engaging in CSR is symptomatic of an agency problem or a conflict between the interests of managers and shareholders” (McWilliams and Siegel, 2001, p 118). Indeed, it is assumed that managers use CSR as a means to further their own social, political, or career agendas at the expense of shareholders. From this perspective, a firm implements CSR activities only if they seem to be a means to desirable profit maximization. Carroll and Joulfaian (2005), Preuss (2010) and Sikka (2010) note that some firms claiming to be socially responsible are also engaged in tax aggressive activities. The balance between societal goals and economic concerns in this case is thus only based on stockholder wealth maximization. The agency theory interpretation has been challenged by other researchers who consider that a firm is more than simply a nexus of contracts, where the firm is managed to maximize shareholder value, and that stakeholders other than shareholders are also important to the 2
firm’s operations (e.g., Hill and Jones, 1992). Thus, according to the corporate social responsibility stream of theories (particularly legitimacy and stakeholder theories), there is an implicit “social contract” between the corporation and society, the terms of which are derived from the expectations of a number of groups. From this perspective, a firm exists above and beyond management, shareholders and any particular stakeholder (e.g., Carroll, 1979; Waddock and Graves, 1997; Wood, 1991). In addition, corporate social responsibility theory proposes that, in seeking to discharge their social responsibilities and gain legitimacy within society, corporations should be less tax aggressive (Lanis and Richardson, 2011). In sum, CSR is a key factor that influences firm performance. Moreover, CSR is likely to have an influence on tax aggressive activities and should, on this basis, be considered a key factor in the success and survival of a corporation. Most previous studies have used a unidimensional (aggregated) measure of CSR (Johnson and Greening, 1999). However, recent research has cast doubt on this use (Hoi et al., 2013; Huseynov and Klamm, 2012; Lanis and Richardson, 2012) because a unidimensional measure may confound the effects of the individual CSR dimensions, which are not equally important or relevant (Johnson and Greening, 1999). This strongly suggests the need to consider the individual dimensions of CSR separately (Hillman and Keim, 2001; Rehbein et al., 2004). This study draws on the work of Johnson and Greening (1999) and expands on it by introducing the multidimensional CSR perceptions of Dahlsrud (2008) and Girerd-Potin et al. (2014). Specifically, social, governance, economic and environmental dimensions are used as proxy measures of the extent to which a corporation engages in CSR activities. Furthermore, effective tax rates (ETRs) (e.g., Gupta and Newberry, 1997; Lanis and Richardson, 2012; Rego, 2003) are used as a proxy measure of corporate tax aggressiveness. Given the sparse research on the effect of CSR components on tax aggressiveness, the authors examine the role of CSR on tax planning activities. Thus, this study addresses the following question: Do the various dimensions of corporate social responsibility affect the tax aggressiveness of a firm? More especially, the study examines whether the relationship between corporate social responsibility and corporate tax aggressiveness depends on the dimensions of CSR. To date, most research on this topic has been conducted in the U.S. context ( e.g., Hoi et al., 2013) and in the Australian market (e.g., Lanis and Richardson, 2013), but the extent to which these findings and their explanations hold in other countries has been largely unexplored. This study contributes to the literature in several ways. First, it provides empirical evidence that firms are more or less likely to engage in tax aggressiveness depending on the CSR dimensions they have developed. To the authors’ knowledge, this paper is one of the very few studies that link CSR dimensions and tax aggressiveness and is the first to address this issue in the French context. Second, it provides insight for policymakers into the conditions under which the risk of corporate tax aggressiveness is higher. Third, it can be of value to economic development specialists, investors, and business consultants seeking to identify the circumstances under which a firm’s CSR activities can be used for tax aggressive purposes. Finally, it provides evidence in support of an emerging research paradigm in the area of CSR and tax aggressiveness, as these two areas have yet to be examined together (Carroll and Joulfaian, 2005; Hanlon and Heitzman, 2010; Moser and Martin, 2012). The remainder of the paper contains four sections: the next section provides an overview of the extant literature on the relationships between CSR and tax aggressiveness and it develops the theoretical model, including a presentation of the hypotheses. The research method, including sample selection and variable measurement, is then presented. This is followed by the presentation of the results and additional tests. The final section discusses the results and presents the paper’s conclusions.
3
2.
Literature review and hypotheses
According to agency theory, a company can be defined as a contract-agent between shareholders and managers, with the firm being guided by a single objective function: shareholder wealth creation (Jensen and Meckling, 1976). From this perspective, social welfare is maximized when all firms in an economy maximize total firm value. Critics have long argued that firm investment in socially responsible but unprofitable ventures will ultimately lead to the demise of the firm, at worst, and to unsustainable support to nonprofit organizations, at best (Murray and Montanari, 1986). Furthermore, “Friedman (1970) asserts that engaging in CSR is symptomatic of an agency problem or a conflict between the interests of managers and shareholders” (McWilliams and Siegel, 2001, p 118). He argues that managers use CSR as a means to promote their own interests at the expense of shareholders. Hence, shareholders can limit divergence from their interests by defining appropriate incentives for the managers (McWilliams and Siegel, 2001). The only corporate objective is, therefore, to maximize shareholder wealth, and social and environmental issues are merely constraints a firm has to integrate into its financial objective function (Friedman, 1970). In line with this objective, a company will implement CSR actions only if they allow profit maximization. In agency theory, the costs of such activities are usually referred to as reputation costs and/or political costs (Chen et al., 2010). The tradeoff between societal goals and economic concerns is, therefore, always driven by the fundamental need to maximize shareholder wealth. Corporate taxes can only be associated with CSR if their payment has implications for the wider society. Indeed, tax aggressiveness might result in significant negative sanctions for and judgments of firms because they are costly to society and firms are likely to engage in opportunistic behavior that runs contrary to societal interests. Thus, CSR could increase shareholder value because it allows shareholders to be protected against the negative effects of tax aggressiveness. The empirical results of Hoi et al. (2013) suggest that firms will increase CSR activities to build up their CSR reputation and thereby lessen the severity of the potential negative sanctions associated with undertaking aggressive tax planning activities. Furthermore, Williams (2007) notes that the most significant issue that arises in attempting to apply CSR principles to firm taxation encompasses those actions that can reduce a corporation’s tax liability through corporate tax aggressiveness. It is nevertheless simplistic to assume that tax aggressive activities always lead to firm value maximization because there are potential costs to being tax aggressive, including nontax costs arising from managers’ concealed actions (Chen et al., 2010). Desai and Dharmapala (2006) and Laguir and Staglianò (2014) highlight the observation that managers may hide rent extraction through tax aggressiveness when the two actions are complementary. The agency theory perspective has been challenged by other researchers who outline a CSR framework. From the stakeholder-oriented perspective, Freeman and Reed (1983) argue that corporations should attempt to satisfy all their stakeholders, even though their primary mandate is to maximize value for shareholders. These scholars were the first to clearly identify the strategic importance of groups and individuals beyond the firm’s stockholders. They pointed to such widely disparate groups as local community organizations, environmentalists, consumer advocates, governments, special interest groups, and even competitors and the media as legitimate stakeholders. Moreover, Freeman and Reed (1983) warned that if these stakeholders were to withdraw their resources, they might endanger the very existence of the firm. The scholars therefore emphasized the importance of efficient stakeholder management as a means to ensure continued support and, ultimately, the achievement of corporate objectives. Shareholder value thus has become one corporate 4
objective among others (Hill and Jones, 1992; Meiseberg and Ehrmann, 2012) and corporate choices are based on both social and economic calculations. Compared with stakeholder theory, legitimacy theory appears to be less tied to the assumption of discrete and identifiable stakeholder actions. According to Suchman (1995), legitimacy is a state in which an organization’s actions are observed to be “...desirable, proper or appropriate within some socially constructed system of norms, values, beliefs and definitions” (1995, p. 574). Thus, firms will try to build and maintain relationships within their social and political environment, seeking the legitimacy they need to survive regardless of how well they perform financially (Gray et al., 1995). Legitimacy theory assumes that an organization is defined in part by its ability to engage in and control the processes of legitimization in order to demonstrate its congruence with societal values (e.g., Guthrie and Parker, 1989). Indeed, with the growth in community awareness and concerns over the past few decades, firms today are expected to take action to ensure that their overall performance is acceptable to the community (Wilmshurst and Frost, 2000). Moreover, legitimacy can be observed as an operational resource (Suchman, 1995) whose value must be maintained to ensure continued support from society. The latter is expressed, for example, in terms of increased capital inflows, customer and supplier appreciation, labor participation, government “blessing,” and community (and media) acceptance when the company acts as a good and environmentally friendly “corporate citizen.” Avi-Yonah (2008) states that the implication of viewing a corporation as a “real world” entity is that CSR may be regarded as a legitimate business activity and not merely a cost on the road to maximizing shareholder wealth. However, any perception of a mismatch between organizational activities and societal values will lead to a legitimacy gap (Haniffa and Cooke, 2005), which in turn may well threaten the organization’s status within the broader social system. For example, if a firm implements a strategy with the sole or dominant aim of avoiding taxation, the general consensus is that it is not paying its “fair share” of tax to the government to ensure the financing of public goods (Freedman, 2003; Williams, 2007). The loss in corporate income tax revenue is likely to arouse hostility, cause reputational damage and, in the worst case scenario, result in the cessation of the firm’s business operations (Lanis and Richardson, 2011; Williams, 2007). Moreover, aggressive tax avoidance practices can be viewed as opportunistic behaviors whereby the firm exploits the implicit contract between the firm and society at the expense of the latter. It follows that aggressive tax avoidance should be inconsistent with CSR (Hoi et al., 2013). The measurement and assessment of CSR builds upon the axiom that “what gets measured, gets managed” (Asif et al., 2013, p.10). As noted earlier, a unidimensional measure of CSR may confound the effects of the individual CSR dimensions. In fact, some dimensions of CSR may be more important than others in explaining corporate decisions about tax aggressive activities (e.g., Hoi et al., 2013; Lanis and Richardson, 2012). Indeed, Johnson and Greening (1999) suggest that “combining all of the corporate social performance dimensions into one construct is inappropriate as there appear to be at least two, conceptually distinct, dimensions; (1) a community, women, minorities, and employee relations dimension and (2) a product quality and environment dimension” (1999, p.565). These authors emphasize that the community, women, minorities, and employee relations group (social dimension) is related to the contributions firms make to communities, to their hiring of women and minorities, and to their treatment of employees, whereas the product quality and environmental dimension (product quality dimension) is related to product and service quality and to a firm's stance toward the natural environment. For Johnson and Greening (1999), product quality and environmentally sound manufacturing are, in effect, two attributes of producing a product. This emphasis on environmentally conscious manufacturing effectively links these two types of corporate social performance. Additionally, Johnson and Greening’s (1999) typology is 5
consistent with the environmental, social, and corporate governance (ESG) perspective, where all concerns related to mankind belong to the “social” dimension because it implies “people”; all concerns related to the natural environment and product and service quality belong to the “environmental” dimension because it implies “product quality”; and all societal concerns that can portray the sustainable performance of a company belong to the “governance” dimension (Kocmanova and Simberova, 2012). In their recent research, Girerd-Potin et al. (2014) highlighted the broad consensus about the ESG classification. However, recent studies (e.g., Dahlsrud, 2008) have extended Johnson and Greening’s (1999) typology by arguing that the natural environment and product and service quality attributes should be analyzed separately as two distinct dimensions of CSR, i.e., the environmental dimension and an economic dimension. Indeed, Van Marrewijk (2003) pointed out that CSR can be considered as an intermediate stage where companies try to balance the triple bottom line (economic, environmental, and social aspects) to reach the ultimate goal of corporate sustainability (CS), further arguing that the three aspects of CS can be translated into a CSR approach that companies need to take into account, as the CSR approach often advocates ethical behavior with respect to them (Hutchins and Sutherland, 2008). Lozano (2011; 2015) and Lozano and Huisingh (2011) explained that CS should be understood as those firm activities that proactively aim to reach sustainability equilibria within the economic, environmental, and social dimensions of today, as well as their inter-relations within and throughout the time dimension (Lozano, 2008), while also taking into account the company’s systems and stakeholders. From this perspective, Dahlsrud (2008) explains that CSR should be understood through the social, environmental and economic dimensions, which are “merely different categories of impacts from business” (2008, p.6), and argues that such a distinction is useful since different sets of tools have to be used when analyzing and managing the social, environmental and economic impacts from business. Thus, the economic dimension of CSR is the means by which firms deal with concerns that might arise in their interactions with customers, suppliers and shareholders in the marketplace (European Commission, 2003). The business behavior in the marketplace is considered as an indicator of how well they have integrated economic responsibility issues into their mainstream organizational structure and decision-making process. The aim of such integration is seen as going beyond short-term profit maximizing to encompass long-term economic performance and contributions to the well-being of all of society (e.g., Bansal 2005). Furthermore, the economic dimension of CSR takes into account shared value creation through the development of innovative products, services and business models that lead to higher product quality and more productive jobs. The environmental dimension of CSR draws on the idea that environmental degradation should be de-coupled from economic growth (Commission of the European Communities, 2001). This dimension often focuses on the adoption of management systems to systematically manage the environmental impacts of a business along the entire product life cycle, thus building a firm’s credibility with external stakeholders and ensuring that the principle of environmental integrity and protection is embraced by internal stakeholders (Kovacs, 2008; Walley and Whitehead, 1994). The social dimension of CSR acknowledges ‘‘the health, safety and general well-being of employees; motivate[s] the workforce by offering training and development opportunities; and enable[s] firms to act as good citizens in the local community’’ (European Commission 2003, p.5). This dimension also involves creating a formal social dialogue to take into account stakeholders’ interests in the decision-making process (Bansal, 2005). Finally, the governance dimension of CSR draws on the assumption that the CSR perspective on business performance is best expressed by considering the voice of multiple stakeholders (Lozano, 2005). Thus, CSR and corporate governance belong to the same 6
corporate accountability continuum promoting sustainability, growth, and stakeholder wellbeing, with the governance dimension reflecting the top management commitment to CSR issues by balancing shareholder value creation with stakeholder value protection (Law, 2011). Indeed, Tudway and Pascal (2006) argue that maximizing shareholder value may incite firm directors to pursue a wider range of social and economic objectives that are consistent with CSR. The theoretical model presented in this study is based on the premise that the different dimensions of CSR have different effects on tax aggressiveness. Thus, tax aggressive dimensions are used as the dependent variable and the four CSR dimensions as independent variables: namely, the social, governance, economic, and environmental dimensions (Dahlsrud, 2008; Johnson and Greening, 1999; Laguir and Elbaz, 2014). Moreover, several control variables from the CSR and tax aggressiveness literature are included in the model to control for other effects. These include corporation size, financial performance, leverage, capital intensity, intangibles and activity sector. Taking into account the agency and corporate social responsibility theories, the study tests the following hypotheses: H1: The level of the CSR social dimension of a firm significantly influences the level of its tax aggressiveness. H2: The level of the CSR governance dimension of a firm significantly influences the level of its tax aggressiveness. H3: The level of the CSR economic dimension of a firm significantly influences the level of its tax aggressiveness. H4: The level of the CSR environmental dimension of a firm significantly influences the level of its tax aggressiveness.
The theoretical model shown in Figure 1 summarizes these hypotheses.
7
Industry
CSR social dimension
H1 Capital
intensity
Intangibles
Financial performance CSR governance dimension
Tax aggressiveness
H2
Leverage
Size
H3 CSR economic dimension
H4
CSR environmental dimension
Fig. 1. Theoretical framework and model.
8
3.
Research method
This section provides information regarding the sample, the variable measurement, and the research method, which will allow an examination of the link between tax aggressiveness and CSR dimensions. 3.1. Data The sample was drawn from the Vigeo database for French firms for the 2003-2011 period, which was the latest and most complete financial period available for data collection at the time this study was carried out. Vigeo has established itself as the leading European expert in the assessment of companies and organizations with regard to their practices and performance on environmental, social, and governance (ESG) issues. It assesses the degree to which companies and public corporations take into account environmental, social, and corporate governance objectives, all of which constitute both risk factors and business opportunities in the definition and implementation of their strategies and policies. The Vigeo data were supplemented with firm-level operational and performance data from the Diane financial database. Diane is one of the Bureau van Dijk Electronic Publishing’s databases. The final sample comprised 83 firm-year observations from 24 listed firms after excluding all firms that could be classified as follows (e.g., Gupta and Newberry, 1997; Lanis and Richardson, 2012): (1) Financial corporations, as their ETRs are likely to differ from those of other corporations because of government regulations; (2) Foreign corporations, as they may be subject to resident country tax laws that differ from French tax laws; (3) Corporations with missing financial data and/or CSR data; (4) Corporations with negative net income or tax refunds, as their ETRs are likely to be distorted (Zimmerman, 1983); (5) Corporations with ETRs exceeding one, as this is likely to cause problems with model estimation (Stickney and McGee, 1982). 3.2. Variable measurement The literature identifies a number of firm-specific determinants that are important in explaining tax aggressiveness. Appendix A details the definitions of these variables. Below is a brief description of the key variables. Tax aggressiveness was measured as a latent construct, using items based on ETRs. ETRs are most often measured from the information given in financial statements, such as tax liability divided by income. However, appropriately defining the numerator and denominator of this equation continues to be a topic of debate (e.g., Hanlon and Heitzman, 2010; Plesko, 2003). The choice to use ETRs in this study was based on the following. First, recent empirical tax research has found that ETRs encapsulate tax aggressiveness (e.g., Chen et al., 2010; Hoi et al., 2013). Second, ETRs are currently the most frequently used proxy measure of tax aggressiveness in academic research (e.g., Dyreng et al., 2008; Lanis and Richardson, 2012; Rego, 2003). Moreover, there are two major issues in the selection of ETR measures 9
related to the nature of the taxes to consider and the methods used to measure profit. First, given that ETRs compare the current tax liability generated by taxable income with pre-tax income based on generally accepted accounting principles (GAAP), they measure the adeptness of a corporation at reducing its current tax liability relative to its pre-tax accounting income. Thus, ETRs indicate the relative tax burden across corporations. Second, given the difference between accounting (book) income and tax income, accounting profit might not represent the actual chargeable income of the firms (Rego, 2003). Generally, firms using tax planning activities to reduce their taxable income while maintaining their financial accounting income have lower ETRs, making ETRs an appropriate measure of tax aggressiveness. To measure the tax aggressiveness (TAG) construct, two ETR measures (e.g., Gupta and Newberry, 1997; Lanis and Richardson, 2012) were used: ETR1 as income tax expense currently payable divided by book income and ETR2 as income tax expense currently payable divided by the operating cash flow. Finally, the data used to construct the ETRs were collected from the DIANE financial database. Researchers generally accept the notion that CSR is multidimensional, but they usually combine the various dimensions used to measure the construct into one aggregate measure. The work of Johnson and Greening (1999) was expanded on by introducing the multidimensional CSR perceptions of Dahlsrud (2008) and Girerd-Potin et al. (2014). CSR was measured with four latent constructs using items from the Vigeo database, which has 38 generic criteria divided into six distinct domains. These domains are human resources, human rights in workplaces, community involvement, corporate governance, environment, and business behavior. Each of these domain ratings ranges from 0 for less socially responsible firms to +4 for more socially responsible firms. The different criteria evaluated by Vigeo for establishing the social ratings are explained in Appendix B. As many researchers have noted (e.g., Dyer and Whetten, 2006), using specialized agency ratings as measures of CSR has several advantages: the firms are rated with an objective set of screening criteria, the agencies apply the ratings consistently across companies, and they have a staff of knowledgeable individuals who are not affiliated with any of the rated. Four dimensions were hypothesized on the basis of the Vigeo database components. The first dimension was labeled the CSR social dimension (CSR_S) and included the human resources (HR score), human rights in workplaces (HRts score), and community involvement (CIN score) ratings to measure it. The second dimension was called the CSR governance dimension (CSR_CG) and it was measured using the corporate governance (CG score) rating. The third dimension was labeled the economic dimension (CSR_ECO) and the business behavior (C&S score) ratings measured it. Finally, the environmental dimension (CSR_ENV) was used, and the environment (ENV score) rating measured it. These four constructs served as the endogenous variables. Several control variables were included, namely, size (SIZE), financial performance (FinPerf), industry, capital intensity (CINT), intangibles (INTG), leverage (LEV), and industry dummies, to investigate how they might trigger TAG or CSR. Previous research (e.g., Johnson and Greening, 1999) has shown that corporation size is positively associated with corporate social performance. Specifically, due to their higher visibility, larger corporations are likely to provide more extensive corporate social performance information in the annual report than smaller corporations (Cho et al., 2010). Furthermore, SIZE may have an impact on tax-reducing activities. Some studies suggest that larger corporations are likely to be more tax aggressive than smaller corporations because they possess greater economic and political power relative to smaller corporations and are able to reduce their tax burdens accordingly (e.g., Gupta and Newberry, 1997). Other studies emphasize that large firms are subject to greater public scrutiny and, as a result, incur a “political cost” in the form of higher
10
ETR (Zimmerman, 1983). In accordance with previous studies (e.g., Lanis and Richardson, 2012), SIZE was measured as the natural logarithm of total assets. Several recent studies have analyzed the link between CSR and FinPerf to define the relationship, along with its meaning and valence (Weisheng et al., 2014). Some academic studies have found a rather positive relationship between social performance and FinPerf (e.g., Lu et al., 2014; Margolis and Walsh, 2003,), while others have found a negative relationship (Kashif et al., 2011). Furthermore, various studies highlight that FinPerf is positively associated with the ETRs because an increase in FinPerf leads to an increase in them (e.g., Lanis and Richardson, 2012). Two items were chosen to measure the FinPerf construct. The first item, return on assets (ROA), is defined as pre-tax income divided by total assets, and the second is defined as pre-tax income divided by shareholders’ equity (ROE). The industry sector (Industry) construct, defined by the French Classification of Activities (nomenclature d'activités françaises: NAF), is also included as a control variable because TAG and CSR may fluctuate across industry sectors (e.g., Lanis and Richardson, 2012). The six Industry dummy items in this construct are as follows: manufacturing (C), energy (D), wholesale and retail trade (G), transportation and storage (H), accommodation and food service activities (I), and professional, scientific and technical activities (M). Last, LEV is long-term debt divided by total assets; CINT is net property, plant and equipment, divided by total assets; and INTG is intangible expenditures divided by total assets. Previous research (Gupta and Newberry, 1997; Hoi et al., 2013) has found that high LEV, CINT, and INTG correspond to low firm ETRs and a high tax aggressiveness level. Leverage is negatively associated with TAG due to tax-deductible interest payments, CINT is negatively associated with TAG because of accelerated depreciation charges corresponding to asset lives, and INTG is negatively associated with TAG as a result of tax-deductible intangible expenditures. 3.3. Data analysis – partial least squares (PLS) The hypotheses were tested using a partial least squares (PLS) method for structural equation model estimation (SEM), as has been used in a number of other studies (e.g., Fornell, 1982; Gallardo-Vázquez and Sanchez-Hernandez, 2014; Zeng et al., 2010). PLS was particularly suitable for this study because it makes minimal data assumptions regarding the distribution of regression residuals and requires a relatively small sample size (GallardoVázquez and Sanchez-Hernandez, 2014). Because PLS is a regression based technique, it requires between five and ten cases for the most complex regression (Chin, 1998; Goodhue et al., 2012; Hair et al., 2011). In this study, the most complex regression is that with TAG as the dependent variable and nine independent variables, suggesting a minimum sample size between 45 and 90 cases. Moreover, a construct is assumed to be reflective if the manifest variables reflect the latent variable and they are its consequence. In contrast, a construct is assumed to be formative if the latent variable is represented by all the manifest variables and is their consequence. Based on the conditions listed by Crié (2005), all the constructs used in this study were assumed to be reflective. SmartPLS software, version 2.00, was used. PLS comprises a measurement model and a structural model. The measurement model specifies the relationships between observed items and latent variables. The component score estimate for each latent variable is obtained using a weighted aggregate of its own indicators (Chin and Newsted, 1999). The structural model specifies the relationships between latent constructs. Although the measurement and structural models are estimated simultaneously in PLS (Barclay et al., 1995), the PLS model is interpreted in two stages. First, the reliability and validity of the measurement model is assessed. Second, the structural model is assessed (Barclay et al., 1995; Hair et al., 1998; Hulland, 1999). Hence, before assessing the structural model, the quality of the measurement model was assessed, thus addressing individual item 11
loadings, construct reliability, and convergent and discriminant validity for the constructs (Bagozzi, 1994). By examining the individual item loadings, it is possible to determine which items can be included in the final model and which items may need to be considered for removal. Items may be removed to avoid bias in the parameter estimates in performing the structural model analysis (Hulland, 1999). The minimum acceptable loading is generally 0.50 (Hair et al., 1998). In the present case, Fornell and Larcker’s (1981) measure of composite reliability is used. An adequate level of composite reliability should be above 0.7, as recommended by Nunnally (1978). The convergent and discriminant validity of the constructs was verified by the average variance extracted (AVE), which represents the average variance shared between a construct and its indicators (Chin, 1998; Fornell and Larcker, 1981). For convergent validity, AVE should be greater than 0.50, which is the minimum acceptable (Fornell and Larcker, 1981). The square root of AVE for each construct should be higher than its correlation with all other constructs. This means that the constructs, even though they are correlated, remain independent (Chin, 1998; Hulland, 1999). In the structural model, the path standardized coefficients are interpreted as in OLS estimation. Because PLS makes no distributional assumptions, the statistical significance of the parameter estimates was assessed using a bootstrap procedure with 1,000 replacements (Chin, 1998). Moreover, the objective of PLS is to maximize the variance explained rather than the fit. Therefore, prediction-orientated measures, such as R2, are used to evaluate PLS models (Chin, 1998). An R² greater than 0.1 demonstrates the significance of the PLS models. 4.
PLS - SEM Results
This section presents the statistical characteristics of the analysis variables and the main results of the adopted methodology along with some additional tests. 4.1. Descriptive statistics Table 1 shows descriptive statistics for all variables used in the baseline model. For the dependent variable ETR1 (ETR 2), the firms in the sample have a mean of 0.1435 (0.3251). The median value is equal to 0.1523 (0.2975). With regard to the dependent variables, the firms in the sample have a mean (median) HR score of 43.9638 (48). Moreover, the mean (median) ENV score is 38.9156 (42). The mean (median) CS score is 41.3855 (46). The mean (median) CG score is 40.3975 (40), and the mean (median) CIN score is 45.0120 (46). The HRts score has a mean (median) of 46.0120 (47). Finally, for the control variables, some variables, such as ROA ROE, and SIZE, seem symmetrically distributed, while others, such as CINT, INTG and LEV, are quite asymmetrically distributed.
Table 1 12
Descriptive statistics. Variable ETR1 ETR2 HR score ENV score CS score CG score CIN score HRts score ROA ROE SIZE CINT INTG LEV
N 83 83 83 83 83 83 83 83 83 83 83 83 83 83
Mean 0.1435 0.3251 43.9638 38.9156 41.3855 40.3975 45.0120 46.0120 .07834 0.1309 6.79456 0.0683 0.1489 0.4701
Std.dev 0.0854 0.2413 17.6320 17.6553 12.2522 12.0809 18.5528 15.2623 0.0503 0.0736 0.8245 0.1663 0.2964 1.0663
p25 0.0659 0.1164 32 29 35 36 34 32 0.0381 0.0712 6.1095 0.0027 0.0073 0.0166
Median 0.1523 0.2975 48 42 46 40 46 47 0.0661 0.1371 6.5286 0.00974 0.0211 0.0536
p75 0.2180 0.4628 57 52 50 48 59 59 0.1007 0.1665 7.7517 0.0164 0.0736 0.4092
min 0.0008 0.0020 4 0 9 6 3 18 0.0016 0.0044 5.1488 0.0000 0.0000 0.0000
max 0.3333 0.9650 75 68 65 65 77 76 0.2344 0.3016 8.1791 0.81022 1.2173 6.0648
4.2. PLS-SEM results The results of the measurement model for the full sample are summarized in Tables 2-4. As Table 2 shows, the industry construct was removed from the model and not used in further analysis because of its low item loadings and lack of composite reliability and convergent validity. All the items of final model loaded on their respective constructs and were greater than 0.5. The factor loadings from the final PLS measurement model are presented in Table 3. Convergent validity appeared adequate for all constructs, as AVE exceeded 0.5 in all cases. The discriminant validity of the constructs is presented in Table 4, which shows that the diagonal AVE values exceed all other scores, suggesting sufficient discriminant validity.
Table 2 Estimation of the measurement model parameters (full sample, n = 83). Loading original sample
TAG
Initial model Composite reliability
0.8346 ETR1 ETR2
0.8949 0.8852 0.8624
CSR_CG CG score
1.0000
CSR_ECO C&S score
1.0000
Loading original sample
0.7181
0.7487 0.9328
CSR_S HR score HRts score CIN score
CSR_ENV
Average variance extracted (AVE)
Final model Composite reliability
Average variance extracted (AVE)
0.8382
0.7227
0.9123
0.7762
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.7829 0.9124 0.9122
0.7761 0.8960 0.8858 0.8610
1.0000
1.0000 1.0000
1.0000
1.0000 1.0000
1.0000
1.0000
13
ENV score
1.0000
FinPerf ROA ROE
0.9824 0.8462
Industry a C D G H M
-0.6623 0.5597 -0.4561 0.0466 0.2679
1.0000 0.9130
1.0000
CINT
1.0000
INTG INTG
1.0000
CINT
1.0000
1.0000
--
--
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000 1.0000 1.0000 1.0000
LEV
0.2068
1.0000 1.0000
LEV
0.8403
-----1.0000
SIZE
0.9128 0.9826 0.8457
0.0148
SIZE
0.8406
1.0000 1.0000
a
This construct was removed from the final analysis because of its low item loadings and lack of composite reliability and convergent validity.
14
Table 3 Cross-loadings (full sample, n = 83). TAG
CSR_S
CSR_CG
CSR_ECO
CSR_ENV
FinPerf
Size
CINT
INTG
LEV
0.1316 0.1146
0.0691 -0.0004
0.6789 0.5926
0.0331 0.0504
0.1676 0.2627
0.0876 0.3682
0.1888 -0.1600
0.0792 -0.2098
0.0144 -0.1905
HR score HRts score CIN score CG score C&S score ENV score ROA ROE
0.7829 0.9124 0.0854 0.1022 0.1773 0.0316 -0.0522 0.0506 0.2811 0.1510
0.8960 0.8858 0.8610 0.4560 0.7026 0.8065 -0.1887 -0.0291
0.4991 0.3704 0.3416 1.0000 0.4958 0.4489 -0.2826 -0.1396
0.6789 0.5926 0.5872 0.4958 1.0000 0.7405 -0.2200 0.0066
0.7769 0.6114 0.7361 0.4489 0.7405 1.0000 -0.1013 0.0306
0.3774 0.3492 0.4148 0.2951 0.4104 0.4078 -0.2416 -0.1801
0.0233 0.1009 0.2792 0.0209 0.0355 0.1907 -0.2167 -0.2082
-0.0905 0.0010 0.0242 -0.0600 -0.1235 0.0072 -0.1392 -0.1775
-0.2231 -0.1165 -0.0825 -0.3217 -0.2333 -0.0731 0.0340 -0.0263
SIZE CINT INTG LEV
0.2983 -0.0250 -0.1104 -0.1268
0.4341 0.1598 -0.0227 -0.1573
0.2951 0.0209 -0.0600 -0.3217
0.4104 0.0355 -0.1235 -0.2333
0.4078 0.1907 0.0072 -0.0731
-0.1358 -0.2571 -0.0334 -0.2594 -0.1705 -0.0709 0.9826 0.8457 -0.2383 -0.2265 -0.1574 0.0195
1.0000 0.2156 0.1355 -0.0175
0.2156
0.1355 0.8976
-0.0175 0.7355 0.8811
ETR1 ETR2
1.0000 0.8976 0.7355
1.0000 0.8811
1.0000
15
Table 4 Discriminant validity coefficients (full sample, n = 83). Diagonal elements (bold) are the square root of the variance shared between the constructs and their indicators (AVE). Offdiagonal elements are the correlations between constructs. For discriminant validity, diagonal elements should be larger than off-diagonal elements.
TAG CSR_S CSR_CG CSR_ECO CSR_ENV FinPerf Size CINT INTG LEV
TAG
CSR_S
CSR_CG
CSR_ECO
CSR_ENV
FinPerf
Size
CINT
INTG
LEV
0.8501 0.1410 0.0316 -0.0522 0.0506 0.2613 0.2983 -0.0250 -0.1104 -0.1268
0.8810 0.4560 0.7026 0.8065 -0.1557 0.4341 0.1598 -0.0227 -0.1573
1 0.4958 0.4489 -0.2594 0.2951 0.0209 -0.0600 -0.3217
1 0.7405 -0.1705 0.4104 0.0355 -0.1235 -0.2333
1 -0.0709 0.4078 0.1907 0.0072 -0.0731
0.9166 -0.2383 -0.2265 -0.1574 0.0195
1 0.2156 0.1355 -0.0175
1 0.8976 0.7355
1 0.8811
1
The results from the structural model, presented in Table 5, indicate how the different dimensions of corporate social responsibility affect tax aggressiveness. The findings reveal a positive and significant relationship between CSR_S and TAG (0.2899, t = 1.8566, p < 0.1), providing support for hypothesis H1. Indeed, the more the firms engaged in human resources activities, human rights in workplaces, and community involvement, the higher their ETRs were and, thus, the lower the likelihood that they would be tax aggressive in nature. The results also show a negative and significant relationship between CSR_ECO and TAG (0.3554, t = 2.5659, p < 0.05), suggesting that the more firms engaged in business behavior efforts, the lower their ETRs were and, thus, the higher the likelihood that they would be tax aggressive in nature. Hence, hypothesis H3 is supported. Moreover, the findings show no significant relationship between CSR_CG, and TAG, thus rejecting hypothesis H2. Finally, the results also reveal no significant relationship between CSR_ENV and TAG, rejecting hypothesis H4. In terms of control variables, financial performance was positively and significantly associated with TAG (0.3781, t = 4.1670, p < 0.01), which is consistent with the expectations because tax rates are progressive according to income. Another alternative explanation for this finding might be that the availability of funds in the case of a financially successful company allows investment in CSR activities. Financial performance was negatively and significantly associated with CSR governance (-0.2004, t = 2.0403, p < 0.05). Moreover, SIZE was positively and significantly associated with TAG (0.4379, t = 4.1933, p < 0.01), CSR_S (0.4209, t = 4.9656, p < 0.01), CSR_CG (0.2474, t = 3.1549, p < 0.01) CSR_ECO (0.3920, t = 5.0793, p < 0.01) and CSR_ENV (0.4145, t = 5.3893, p < 0.01), which is consistent with the literature showing that larger corporations are likely to provide higher CSR efforts (e.g., Johnson and Greening, 1999) and lower tax aggressiveness (e.g., Zimmerman, 1983). The results reveal that CINT was positively and significantly associated with TAG (0.5124, t = 2.329, p < 0.05), which is inconsistent with expectations, although this result is consistent with the findings of Armstrong et al. (2011) and Huseynov and Klamm (2012), who demonstrated that the positive effect may be related to long-term timing issues. This finding can also be interpreted in the light of the observation that higher operating Leverage (% of 16
fixed costs in the cost structure) increases the perceived risk and leads to defensive policies, including tax manipulation and, eventually, lower CSR engagement. Furthermore, the variable INTG is negatively and significantly associated with TAG (-0.5953, t = 1.7978, p < 0.1), which is consistent with expectations because ETRs decrease as a result of taxdeductible intangible expenditures. Finally, the results show no significant relationship between LEV and TAG, which is inconsistent with the conventional hypothesis that an increase in debt is associated with lower tax rates. Nevertheless, the results are similar to those of Huseynov and Klamm (2012) and Minnick and Noga (2010) who found no significant relationship between LEV and TAG in their studies.
17
Table 5 PLS structural model: path coefficients, t-statistics and R² (full sample, n = 83). Each cell reports the path coefficient (t-value). Blank cells indicate that the path was not hypothesized within the model; *** Significance at the 0.01 level; ** Significance at the 0.05 level; * Significance at the 0.1 level. Dependent variables
Independent variables TAG
CSR_S
CSR_CG
CSR_ECO
CSR_ENV
FinPerf
Size
CINT
INTG
LEV
R²
TAG
--
0.2899 (1.8566)*
0.0934 (0.6779)
-0.3554 (2.5659)**
-0.2029 (1.0509)
0.3781 (4.1670)***
0.4379 (4.1933)***
0.4948 (2.1827)**
-0.5953 (1.7978)*
0.0120 (0.0519)
0.3334
CSR_S
--
--
--
--
--
-0.0519 (0.4593)
0.4209 (4.9656)***
--
--
--
0.1914
CSR_CG
--
--
--
--
--
-0.2004 (2.0403)**
0.2474 (3.1549)***
--
--
--
0.1250
CSR_ECO
--
--
--
--
--
-0.0771 (0.6527)
0.3920 (5.0793)***
--
--
--
0.1740
--
--
--
--
0.0278 (0.2494)
0.4145 (5.3893)***
CSR_ENV
0.1671
18
Although no overall index of model validation is given with Smart PLS 2.0 (as is the case for structural equation methods based on covariance), an index of overall model quality was developed (Amato et al., 2005) to remedy this. The index is obtained by calculating the geometric mean of R² and the communality scores. This index is called the goodness-of-fit (GoF), and it ranges from 0 (invalidation of the model) to 1 (perfect model validation). The formula is written as follows (Tenenhaus et al., 2005) in equation (1): √
The GoF was computed, yielding GoF = 0.4223, indicating acceptable validity for the model. The results of the PLS structural model analysis are depicted in Figure 2.
CSR social dimension
0.2899*
0.4209***
Capital
intensity Financial performance
-0.2004***
0.4948**
Intangibles
-0.5953*
0.3781*** CSR governance dimension
Tax
0.4379***
aggressiveness
0.2474***
Size
0.4145***
-0.3554** 0.3920*** CSR economic dimension
CSR environmental dimensionn
Fig. 2. PLS structural model with significant path coefficients. ***p < 0.01, **p < 0.05, *p < 0.1. 19
4.3. Additional analysis The study's dependent variable of interest is corporate tax aggressiveness (TAG). Thus, in this section, several proxy measures for TAG are used, including the book-tax gap (BTG), to improve the robustness of results. BTG is considered an effective measure of tax aggressiveness because large differences between accounting (book) income and taxable income are common among firms that show significant tax aggressive behavior (e.g., Desai and Dharmapala, 2006; Frank et al., 2009; Wilson, 2009). Firms can structure transactions to generate large temporary or permanent differences between accounting and taxable income. Thus, the first measure of tax aggressiveness (BTG1) involves the assessment of the raw BTG, which captures tax strategies that lead to both temporary and permanent differences. Following the method in Manzon and Plesko (2002), BTG1 is computed as pre-tax accounting income less taxable income scaled by total assets, with taxable income calculated as income tax expense divided by the statutory maximum corporate tax rate of 33.1/3%. The second measure of tax aggressiveness (BTG2) is computed as the BTG residual, following Desai and Dharmapala (2006). BTG is adjusted in the same way they do, in order to control for the earnings management strategies (the smoothing of reported income over time in order to reach bonus targets and to achieve other aims) that may be responsible for it. Specifically, the BTG component attributable to earnings management (via income-changing discretionary accruals) is removed to leave a residual value that is inferred to capture tax aggressiveness. As noted earlier, CSR is measured with four latent constructs, using items from the Vigeo database, which has 38 generic criteria divided into six distinct domains. In this section, an additional PLS-SEM analysis is performed to determine which of the six individual CSR categories are significantly associated with tax aggressiveness. Appendix C, Panel A, presents the results. Specifically, the findings reveal a positive and significant relationship between the CIN score and TAG (0.3644, t=2.4834, p
