Tax Evasion, Corporate Social Responsibility and National Governance: A Country-Level Study

Abstract

This study uses a sample of 25 OECD countries to examine the association between CSR, national governance and tax evasion at the country level. The interaction between country-level governance and CSR relative to tax avoidance is also explored. The findings suggest that neither the ESG dimensions nor the overall CSR measure are significant determinants of tax evasion at the country level. In contrast, national governance quality is significantly and negatively related to tax evasion. Significant support is also found for the mediating effect of national governance on the association between CSR and tax evasion: in countries with weak national governance, CSR and country-level governance are substitutes; in countries with strong national governance, CSR reporting (in particular, environmental disclosures) seems to be used as a cosmetic and compensatory tool for firms to mitigate the reputational risk and public concern arising from tax evasion activities. The findings are theoretically and practically relevant as they underscore not only the importance of national governance in mitigating tax evasion but also the relevance of the mediating effect of national governance on the relationship between CSR and tax evasion. The evidence highlights the need for policymakers in countries with strong national governance to design new/strong anti-tax avoidance regulations.

1. Introduction

Even though tax aggressiveness is a controversial topic, it has received increasing attention in recent years from academics and tax authorities because governments worldwide are facing shortfalls in tax revenues along with increased social problems [1]. Corporate aggressive actions designed to minimise taxes are an ongoing challenge for tax administrations worldwide [2,3,4,5]. Noncompliance reduces tax system efficiency by widening administrative costs and by intensifying the tax burden on those who comply [5]. Opportunistic tax minimisations decrease state resources and the funds that are needed to support government programmes to improve social welfare for the whole community [4,6]. They are a matter of significance to legal and tax authorities, and corporations and a key subject for shareholders and stakeholders in general [7,8].

Several international organisations have developed programmes and actions that are specifically designed to combat aggressive tax actions. Initiatives such as the Base Erosion and Profit Shifting Project (BEPS Project) promoted by the Organisation for Economic Cooperation and Development (OECD) G20 highlight the crucial need to fight against and reduce aggressive tax avoidance practices by strengthening national regulations and increasing international tax cooperation. As argued by Schneider and Buehn [9], fighting against the shadow economy and tax evasion have been central policy objectives in OECD countries for decades. The Global Reporting Initiative Standard on Tax (GRI 207) further stresses the relevance of public reporting on taxes, as it increases transparency and promotes trust and credibility in firm tax practices and in the tax systems and enables stakeholders to make informed judgments about the entities’ tax positions. Tax aggressiveness “is not just a financial problem for tax authorities, but one that erodes critical common spaces necessary for the smooth functioning of regulatory compliance, organizational integrity, and society” [6] (p. 1009).

Given the multidisciplinary nature of the subject, the literature on opportunistic tax actions has been increasing as new facets are progressively added [8]. There is a wide and rising interest and concern over the level, the determining factors and the consequences of aggressive tax practices [8].

Under the presumption that organisations exist within society by continuously managing risk and reputation, a growing number of studies have investigated the relationship between corporate tax avoidance and corporate social responsibility (CSR) [3,8,10,11,12,13,14,15]. These studies have collected contradictory evidence. Some studies have found that firms that are more socially responsible are more likely to be less tax aggressive in nature, while others have concluded that there is a negative association between CSR disclosures and tax avoidance and show how firms can use CSR disclosures to meet social expectations. All of these studies have conducted analyses at the firm level, and only a few are cross-country studies. Moreover, they capture aggressive corporate tax actions using a variety of measures (effective tax rates, cash effective tax rates, book-tax differences, tax differentials measures and tax sheltering activities), and several questions have been raised in prior research about the accuracy of these measures [10,12,16,17,18].

This study seeks to bring new contributions to the literature by using a sample of 25 OECD countries to explore the association between CSR, national governance and tax evasion at the country level. The moderator effect of national governance on the relationship between CSR and tax evasion is also explored.

Tax management (planning), tax aggressiveness, tax avoidance and tax evasion are terms used interchangeably in the literature, and the absence of a common definition presents a problem as these expressions might encompass different meanings [8]. In this paper, I use a broad concept of tax evasion that covers a grey area related to questionable ethical practices [15,19,20,21]. While I conceptually locate tax evasion in a grey area related to questionable ethical practices, the expressions “tax aggressiveness”, “tax avoidance” and “tax evasion” are indistinctly used in this paper. The Buehn and Schneider [22] comprehensive measure of tax evasion that I use in this study fits this broad definition, as it is computed at the national level and is the result of a model estimation in which indirect tax burden (taxation on goods and services) and self-employment are the driving forces.

Overall, the findings of this study suggest that neither the environmental, social and governance (ESG) dimensions nor the overall CSR measure prove to be a significant direct determinant of tax evasion at the country level. In contrast, national governance quality does prove to be a significant determinant of tax evasion at the country level (national governance includes not only government efficiency, regulatory quality, rule of law and corruption control but also the national quality of public company auditors’ working environments). The higher the levels of country-level governance, the lower the levels of tax evasion. I also found support for the mediator effect of national governance on the relationship between CSR and tax evasion. In countries with weak national governance, substantive overall CSR activities are slightly associated with significantly lower levels of tax evasion, implying that country-level governance and CSR are substitutes. In countries with stronger national governance, higher overall CSR reporting (especially environmental disclosures) are significantly related to higher levels of tax evasion, suggesting that CSR disclosures can be used as a means to mitigate public concerns arising from tax evasion activities. Additionally, the results highlight the role of national culture as a determining factor of tax evasion and are robust in several additional tests.

This study contributes to the literature on the relationship between CSR and tax avoidance in several ways. First, while the association between CSR and tax avoidance has been widely explored at a firm-level basis in single country contexts, cross-country studies are scarce [23]. This study fills this gap in the literature by exploring such a relationship on a country-level basis in a sample of 25 OECD countries, which, to the best of my knowledge, has not been investigated in any previous research attempt. Second, studies that explore how a country’s legal and institutional context impacts the relationship between CSR and tax evasion are scarce [8,23,24,25]. This study expands past research by going beyond an examination of the association between national governance quality and tax evasion to explore the effects of the interplay between national governance and CSR on tax evasion. It further innovates by introducing the role of the quality of public auditors’ work environments as well as the activities of independent accounting enforcement bodies into the national governance dimension.

The results are also practically relevant for academics, policymakers, tax managers and the community as a whole. They highlight the importance of promoting the quality of national governance (government efficiency, regulatory quality, rule of law and corruption control), including the quality of public auditors’ working environments, as national rules on CSR and national governance cannot ensure an equivalent corporate disclosure style and behaviour across countries [7]. The results further emphasise the need for policymakers in countries with strong national governance to design strong anti-tax avoidance regulations as businesses seem to use CSR activities to moderate the reputational risk and public concern arising from tax evasion activities.

The remainder of this paper is structured as follows. Section 2 reviews current literature and describes the research hypotheses. Section 3 presents the research design. Section 4 presents and discusses the results. Finally, the conclusions and avenues for future research are presented in Section 5.

2. Theory and Hypotheses Development

A prolific body of literature has explored the social aspects of corporate taxation. Researchers have also investigated the relationship between CSR and tax aggressiveness.

Tax aggressiveness is generally referred to as being unethical and immoral [5,15,20,26], while CRS is referred to as a form of soft law, with academics arguing that corporate social responsibilities prescribe that companies should responsibly observe the tax law [6].

As aggressive tax actions can be considered socially irresponsible, socially responsible businesses by their nature are predictably less tax avoidant [10,27]. This argument assumes that a company takes an ethical stance regarding its business and that CSR provides corporations with a set of guiding principles that may be used to contribute to their own wealth and that of the wider community [11,13,28,29]. Several theories developed in the business and society literature take this view. Under the corporate culture theory, corporate decisions should mirror a shared belief of ‘‘right behaviour’’ and corporate culture reflects a set of shared values within an organisation about the ‘‘right’’ corporate behaviour [30,31]. Based on a complementary view, the stakeholder theory principles advocate that corporations cannot maximise their long-term value without balancing conflicting demands from various stakeholders and without maintaining a good relationship with all stakeholders [32,33,34].

Under this view, several studies have been performed and have reported a negative relation between CSR and tax avoidance: Lanis and Richardson [3] in a sample of Australian corporations, Gulzar et al. [7] in the Chinese context, and Salhi et al. [8] within a sample of United Kingdom and French firms. In the United States, Lanis and Richardson [10] found similar evidence and further concluded that CSR community relations and diversity categories are the elements with the most significant impact in reducing tax avoidance. In the French context, the results of Laguir et al. [35] revealed that the greater the corporate performance in the social dimension of CSR, the lower the level of tax avoidance. In terms of cross-country studies, López-González et al. [14] used data from 28 countries and found evidence indicating that firms which are committed to CSR engage in less tax avoidance. Finally, a study by Ortas and Gallego-Alvarez [36] used data from 30 countries and revealed that companies with higher performance in CSR, social, environmental and governance categories are less likely to engage in tax avoidance.

From a distinct perspective, a significant research branch highlights that CSR disclosures may turn into a corporate cosmetic tool to achieve private goals and gains, to improve the company’s image, and to alleviate the reputational risk arising from socially irresponsible activities [24,29]. This view follows traditional economic and legitimacy-based theories as well as the notions of organisational façades and organised hypocrisy and suggests a positive relationship between CSR and aggressive tax actions [13,37,38].

Traditional economic theories argue that companies dedicate resources to CSR activities only to the extent that these activities are aligned with the main goal of maximising profits and shareholder wealth [13,37]. In a similar vein, risk management theory claims that firms increase their CSR performance in order to maximise shareholder interests and reduce the reputational risk arising from negative corporate events [39,40]. In legitimacy-based theories, companies are characterised by the use of strategical CSR disclosures for “window-dressing and impression management purposes” [41] (p. 80). Consistently, under the theoretical lens of organisational façades and organised hypocrisy, conflicting social and institutional pressures require corporations to develop façades and to become involved in hypocrisy, reducing the prospects that CSR reporting may evolve into more substantive disclosures [38,41].

Taking these last perspectives, a prolific stream of research has revealed that corporations tend to intensify their CSR activity to reduce undesirable reputational effects related to aggressive tax practices. Using data from US companies, Hoi et al. [11] and Watson [12] reported that companies with extremely irresponsible CSR activities were involved in significantly more aggressive tax evasion activity, while the results of the study by Davis et al. [13] suggested that CSR and tax payments act as substitutes. In the context of Canadian firms, Landry et al. [42] observed that tax behaviours and CSR are not necessarily aligned. Abdelfattah and Aboud [15] used a sample of Egyptian firms and obtained similar results, leading to the conclusion of a positive relation between CSR and tax aggressiveness. Zeng [23] and Alsaadi [25] performed cross-country studies (using firm data from 36 countries and 15 European countries, respectively) and provided significant evidence that CSR and tax avoidance are positively related.

In summary, research on the association between firm-level CSR and tax avoidance has received increased attention. Quite surprisingly, however, the direction and significance of the linkage between tax evasion and CSR remains an intriguing question [6]. In addition, the majority of the studies have focussed on a specific country, with a very limited number of cross-country studies. “Cross-country studies of CSR and tax avoidance in various legal and institutional environments are rare” [23] (p. 245), and to the best of my knowledge, the link between tax evasion and CSR at the country level has not yet been explored.

In light of the mixed results on the association between CSR and tax avoidance, I postulate the following nondirectional hypothesis:

Hypothesis 1 (H1).

Country-level CSR is associated with tax evasion.

Recent research has discussed the role of national governance in the association between CSR and tax aggressiveness [8,23,24,25].

As Bird and Davis-Nozemack [6] argued, tax evasion has the potential to deplete three common resources that provide value to society: (1) erodes the social commons by depriving public system of the necessary financial support to operate; (2) erodes the regulatory commons, a set of mutual expectations, norms and understandings that are shared between firms and the authorities who regulate them; and (3) erodes shared spaces within organisations that rely upon a foundation of mutually shared social capital upon which trust, honesty and integrity are based. Thus, exploring the national context of tax evasion is a matter of great importance [5].

National governance, which reflects the legal, political and institutional context as well as the financing and taxation systems of a country, has been shown to be related to opportunistic actions, such as earnings management and tax avoidance [8,14,43,44,45,46,47,48,49].

For Zeng [23], the association between CSR and aggressive tax actions varies across the legal and institutional contexts of the countries, and Lin et al. [24] further argued that the relationship between CSR and tax avoidance is mediated by the robustness and quality of the institutional settings in which the companies operate. Stephenson and Vracheva [50] also claimed that it is of fundamental importance when exploring the link between CSR and tax evasion to account for the disparities in national governance across countries.

In fact, in the Chinese context, Lin et al. [24] found significant evidence of a divergence between the quality of the CSR reporting and the corporate tax payment level in regions with poor institutions; in contrast, firms in other regions were more socially responsible in terms of tax compliance. Zeng [23] and Alsaadi [25] in their cross-country studies with data from European countries also explored the mediating effect of the institutional environment on the relationship between CSR and tax aggressiveness. Contrary to Lin et al. [24], Zeng [23] observed that, in countries with weaker national governance, businesses with greater CSR disclosure present lower tax avoidance, suggesting that national governance and CSR act as substitutes. Furthermore, the Alsaadi [25] results showed that, in countries with low financial-tax reporting conformity, firms with higher CSR performance are more engaged in aggressive tax actions compared with firms in countries with a high level of conformity between financial and tax reporting systems.

As Zeng [23] argued, it is not clear whether national governance and CSR are complements or are substitutes in terms of their impact on tax evasion. While it is expected that stronger national governance implies less discretion and higher transparency in CSR reporting, weaker national governance systems are expected to be less demanding in terms of transparency of CSR and tax reporting [23]. The results of Cahan et al. [51], who examined CSR disclosures in 21 countries, showed that CSR disclosures are more robust and transparent in institutionally strong countries. Nevertheless, in weak institutional contexts, corporations may voluntarily adopt more transparent and high-quality CSR reporting systems to positively differentiate themselves from other companies, in which case, CSR acts as a substitute for weak national governance [23].

Given the association between country-level governance and tax evasion reported in the literature and the different perspectives on the mediating effect of national governance on the association between CSR and tax evasion, I state the following nondirectional hypotheses:

Hypothesis 2 (H2).

Country-level governance is associated with tax evasion.

Hypothesis 3 (H3).

Country-level governance mediates the relationship between country-level CSR and tax evasion.

3. Research Design

3.1. Data

The sample used in this study covered 25 OECD countries for the period between 2004 and 2010. These countries were Australia, Austria, Belgium, Canada, Chile, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, South Korea, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States.

The initial sample was based on the 38 OECD countries over the period of 1999 to 2010 for which the tax evasion measure of Buehn and Schneider [22] was available.

However, I excluded 13 countries for which there was incomplete data on CSR available in the Refinitiv–ESG–ASSET4 database (for six countries, there was no CSR data available in the ASSET4 database, and seven countries were excluded for having less than 15 companies with valid CSR data in the ASSET4 database). Furthermore, even though the ASSET4 contains data on CSR categories collected since 2002, for most countries, there is a lack of valid CSR data for the initial years of 2002 and 2003.

Therefore, the final sample consisted of 165 country–year observations from 25 OECD countries for the period from 2004 to 2010.

3.2. Measuring Country-Level Tax Evasion

I used the Buehn and Schneider [22] tax evasion measure (TE), which is computed at the national level.

Buehn and Schneider [22] computed time series estimates of tax evasion (in % of official GDP) across 38 OECD countries over the period 1999 to 2010 based on a Multiple Indicators Multiple Causes (MIMIC) model grounded on the results of Feld and Schneider [52], and Schneider and Buehn [9].

In their estimates, Buehn and Schneider [22] considered the indirect tax burden (taxes on goods and services) and self-employment as driving forces of tax evasion. The average scores for the 38 OECD countries of the Buehn and Schneider [22] measure over the period 1999 to 2010 range from 2.5 to 10.3. The higher the scores, the higher the country-level tax evasion.

3.3. Measuring Country-Level Corporate Social Responsibility

I computed CSR measures using data provided by the Refinitiv–ESG–ASSET4 database, which has collected data based on the environmental, social and governance (ESG) dimensions of worldwide public companies (publicly traded firms) since 2002. For each firm, this database collects more than 400 evaluation points based on publicly available data, including company annual financial/CSR/sustainability reports, websites and news, among other sources.

The ASSET4 database organises data into ten categories within three pillars:

(1) Environment Pillar Score (a company’s score resulting from three environmental categories—resource use, emissions and innovation);

(2) Social Pillar Score (a company’s score resulting from four social categories—workforce, human rights, community and product responsibility);

(3) Governance Pillar Score (a company’s score resulting from three governance categories—management, shareholders and CSR strategy).

A combination of the three pillars (environmental, social and corporate governance) formulates the final ESG score, which is a reflection of the company’s ESG performance (i.e., ASSET4 database provides an overall ESG score for each firm).

The company’s scores for the environmental, social and corporate governance pillars as well as the final company’s ESG score as reported by the ASSET 4 database range from 0 to 100.

The CSR data provided by the ASSET4 database have been widely used in prior research [11,12,13,25,53].

Based on the company’s scores on the environmental, social and corporate governance pillars as well as the final company’s ESG score reported by the ASSET4 database, I computed four country-level measures of CSR:

(1) The country–year CSR ESG measure (CSREGS), which corresponds to the median of the final ESG score of all companies included in the database for a given country and a given year (values range from 0 to 100);

(2) The country–year CSR Environmental measure (CSREnv), which corresponds to the median of the environmental scores of all companies included in the database for a given country and a given year (values range from 0 to 100);

(3) The country–year CSR Social measure (CSRSocial), which corresponds to the median of the social scores of all companies included in the database for a given country and a given year (values range from 0 to 100);

(4) The country–year CSR Governance measure (CSRGov), which corresponds to the median of the governance scores of all companies included in the database for a given country and a given year (values range from 0 to 100).

For each of the four measures of CSR, a country must have a minimum of 15 companies/year included in the ASSET4 database with valid scores to enter in the analysis. The average number of companies per country year used to compute CSR variables between 2004 and 2010 is as follows: Australia (396); Austria (32); Belgium (51); Canada (413); Chile (43); Denmark (48); Finland (39); France (162); Germany (197); Greece (26); Ireland (20); Italy (104); Mexico (55); the Netherlands (57); New Zealand (58); Norway (72); Poland (44); Portugal (16); South Korea (125); Spain (78); Sweden (161); Switzerland (128); Turkey (55); the United Kingdom (484) and the United States (3071). Furthermore, of the 25 OECD countries in the sample, 22 present CSR variables with valid scores for each of the seven years of the panel data; Chile and Poland present CSR variables with missing values for 2004, 2005 and 2006 (and valid scores for 2007, 2008, 2009 and 2010) and Turkey has CSR variables with missing values for 2004, 2005, 2006 and 2007 (and valid scores for 2008, 2009 and 2010).

The higher the scores of the four CSR variables, the higher the country level of CSR disclosure of information and the higher the CSR.

3.4. Measuring Country-Level Governance

The main country-level governance measure was computed using data taken from the Worldwide Governance Indicators (WGI) published by the World Bank. WGI reports aggregate/individual governance indicators for over 200 countries over the period 1996–2019 for six dimensions of governance (voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law and control of corruption). For each dimension of governance, WGI reports a minimum estimated value of −2.5 (weakest governance) and a maximum of 2.5 (stronger governance). Similar to Zeng [23], I used four WGI indicators to compute an aggregate measure of country-level governance: government effectiveness, regulatory quality, rule of law and control of corruption. These four indicators have been shown to be important to a country’s institutional and legal context [23,54].

As these four WGI indicators are strongly correlated with each other, a principal-component analysis was used to compute the aggregate measure of country-level governance. All of the correlation coefficients among the four WGI indicators are very high and statistically significant at the 1% level, with the lowest correlation coefficient being 0.888.

Using factor analysis to combine several dimensions highly correlated into one is quite common in the literature [55,56]. The principal-component analysis revealed one robust interpretable factor with an eigenvalue of 3.81, which included the four WGI indicators. This factor explained 95.4% of the total variance, and the pattern matrix showed the importance of each WGI indicator in the factor: the government effectiveness indicator has a loading of 0.988, the regulatory quality presents a loading of 0.957, the rule of law indicator has a loading of 0.985 and the control of corruption a loading of 0.978. The value of the Kaiser–Meyer–Olkin test was 0.8205, indicating that factor analysis yielded distinct and reliable factors. The aggregate measure of country-level governance corresponded to the scores of the prediction of the retained factor; the higher the value, the stronger the country-level governance (values ranged from −3.60 to 1.06). Negative (positive) values were considered weak (strong) country-level governance. Thus, I created a dummy variable, WGI, equal to one if the country had strong governance and zero otherwise.

While WGI controls overall institutional quality at the country level, it does not capture differences between countries specifically related to the institutional setting of financial reporting [57]. As Leuz [58] noted, financial reporting regulations should be considered within the elements of the institutional infrastructure.

Thus, I included two additional variables designed to capture that specific aspect. In particular, I included country-level measures as computed by Brown et al. [57] of the quality of public company auditors’ working contexts (AUD) and the degree of accounting enforcement activity (ENF) by independent enforcement bodies. The AUD variable is an index resulting from nine items related to the audit function, including requirements related to auditor training, the professional quality assurance programme, requirements of independence and monitoring mechanisms; the ENF variable is an index resulting from six items related to the quality of monitoring and the promotion of compliance activity with accounting standards developed by independent enforcement bodies (for details, see Brown et al. [57]). Brown et al. [57] only provided values for AUD and ENF for each of the years 2002, 2005 and 2008, while the sample in this study goes from 2004 to 2010. To overcome this issue, values of 2002 were used for 2004, values of 2005 were also used for 2006 and 2007, and values of 2008 were also used for 2009 and 2010. In fact, the data provided by Brown et al. [57] shows low variation in the AUD and ENF values between 2005 and 2008; thus, no flag is raised that using those values for the two following years might cause a problem in my findings. Values for the 25 OECD countries for the AUD index of Brown et al. [57] for 2002 range from 2 to 18; for 2005, between 4 to 32; and for 2008, from 4 to 32. Values of the ENF index for 2002 range from 2 to 21; for 2005, from 5 to 22; and for 2008, from 5 to 22. The higher the values of AUD and ENF, the higher the quality of the auditors’ working environment and the higher the degree of accounting enforcement activity.

3.5. Control Variables

I included a set of country-level variables that, according to prior literature, might be associated with tax evasion.

Consistent with prior studies, I controlled for the natural logarithm of gross domestic product (GDP) to control for countries’ economic development and growth (GDP data were retrieved from the Word Bank Database) and used the worldwide tax system (WW) computed as a dummy variable equal to one if the country had a worldwide tax system and zero if the country had a territorial tax system [21,23,25,59,60]. The statutory tax rate of each country was not included as a control variable because the measure of tax evasion I used already accounts for the effects of the tax burden.

I further included the original four indexes of national culture developed by Hofstede [61]: power distance (PD), individualism versus collectivism (Ind), masculinity versus femininity (Mas) and uncertainty avoidance (UA). Consistent with Yoo and Lee [62], I only used the four original Hofstede’s cultural dimensions due to the supplementary nature of the two more recently introduced dimensions (indulgence and long-term orientation). I also used the most recent scores reported by Hofstede in 2010 (for each dimension, scores range from 0 to 100) in order to preserve consistency with the sample data used in this study (i.e., 2004 to 2010). The Hofstede cultural dimensions scores are available at https://www.hofstede-insights.com (accessed on 10 May 2021). According to prior studies, cultural dimensions are determinants of tax evasion [2,5,62,63,64,65]. All of the models also included year dummies in order to capture year fixed effects.

3.6. Regression Models

To determine the country-level relationships between CSR reporting, governance and tax evasion and to test Hypotheses 1 and 2, I designed the following regression models:

TE = β₀ + β₁ CSREGS + β₂ WGI + β₃ AUD + β₄ ENF + β₅ GDP + β₆ WW + β₇ PD + β₈ Ind + β₉ Mas + β₁₀ UA + Year Dummies + ε

(1)

TE = β₀ + β₁ CSREnv + β₂ CSRSocial + β₃ CSRGov + β₄ WGI + β₅ AUD + β₆ ENF + β₇ GDP + β₈ WW + β₉ PD + β₁₀ Ind + β₁₁ Mas + β₁₂ UA + Year Dummies + ε

(2)

where TE is a country-level measure of tax evasion; CSREGS, CSREnv, CSRSocial and CSRGov are country-level measures of CSR reporting; WGI, AUD and ENF are measures of national governance quality; and GDP, WW, PD, Ind, Mas and UA are control variables.

In order to test Hypothesis 3, I estimated the following regression models:

TE = β₀ + β₁ CSREGS + β₂ WGI + β₃ CSREGS*WGI + β₄ AUD + β₅ ENF + β₆ GDP + β₇ WW + β₈ PD + β₉ Ind + β₁₀ Mas + β₁₁ UA + Year Dummies + ε

(3)

TE = β₀ + β₁ CSREnv + β₂ CSRSocial + β₃ CSRGov + β₄ WGI + β₅ CSREnv*WGI + β₆ CSRSocial*WGI + β₇ CSRGov*WGI +β₈ AUD + β₉ ENF + β₁₀ GDP + β₁₁ WW + β₁₂ PD + β₁₃ Ind + β₁₄ Mas + β₁₅ UA + Year Dummies + ε

(4)

The variables used in models (3) and (4) are the same as those of models (1) and (2); however, I added the interaction between the CSR variables and WGI to explore the extent to which the slope of CSR variables varies for countries with strong versus weak national governance. A table summarising all of the variables is presented in Appendix A.

In model (3), my interest is in the coefficients β₁ and (β₁ + β₃). β₁ evaluates the relationship between CSREGS and TE in countries with weak national governance (WGI = 0). If higher CSR scores are related to higher tax evasion in these countries, I expect β₁ to be positive. β₃ measures the incremental association between CSREGS and TE (or a change in the direction of this association) in countries with strong national governance (WGI = 1). If higher CSREGS scores are related to incremental decreases in TE in these countries, I expect β₃ to be negative. When linearly combined, β₁ and β₃ measures the total effect of CSREGS on TE in countries with strong national governance. The logic for model (4) is exactly the same as just described, and the coefficients of interest are β₁, β₂, β₃, (β₁ + β₅), (β₂ + β₆) and (β₃ + β₇).

4. Results and Discussion

4.1. Descriptive Statistics and Correlation Analysis

Table 1. Descriptive statistics.

Variable	N	Mean	Median	SD	Min		Max
Panel A: Descriptive statistics of the full sample
TE	165	2.61	2.1	1.42	0.5		7.2
CSREGS	165	40.43	37.68	11.11	18.13		69.81
CSREnv	158	29.98	27.49	20.07	0		79.32
CSRSocial	158	39.60	37.31	13.73	9.41		89.69
CSRGov	158	47.68	48	4.69	32.78		65.24
WGI	165	0.05	0.42	0.99	−3.60		1.05
AUD	165	21.33	22	7.27	4		32
ENF	165	14.93	16	6.25	2		24
GDP	165	27.27	27.01	1.09	25.36		30.34
WW	165	0.43	0	0.50	0		1
PD	165	43.95	38	18.15	11		81
Ind	165	63.52	70	21.08	18		91
Mas	165	47.74	54	21.49	5		79
UA	165	64.23	59	22.83	23		100
Variable	N	Mean	Median	SD	N	Mean	Median	SD
Panel B: Descriptive statistics of strong country-level governance vs. weak country-level governance
	Strong country-level governance				Weak country-level governance
TE	118	1.88	1.9	0.61	47	4.45	4.3	1.21
CSREGS	118	38.66	37.09	9.55	47	44.88	41.78	13.41
CSREnv	111	27.59	26.1	18.46	47	35.98	31.78	22.77
CSRSocial	111	37.56	35.76	10.94	47	44.41	41.29	17.99
CSRGov	111	47.92	48.19	4.13	47	47.09	47.69	5.89
AUD	118	23.07	24	7.03	47	16.96	17	5.97
ENF	118	16.25	19	6.51	47	11.64	10	4.04
GDP	118	27.27	26.91	1.20	47	27.27	27.49	0.78
WW	118	0.395	0	0.49	47	0.53	1	0.50
PD	118	36.43	35	15.11	47	62.83	60	9.19
Ind	118	72.65	71	13.59	47	40.60	35	19.07
Mas	118	45.88	54	23.46	47	52.40	57	14.71
UA	118	54.69	51	19.47	47	88.17	85	8.72

Notes: The WGI variable values reported in Panel A are based on the original scores of the aggregate measure of country-level governance resulting from the factor analysis process; GDP is the natural logarithm of gross domestic product; data on CSREnv, CSRSocia, CSRGov are only available for 158 country–year observations; the sample period is 2004–2010.

, Panel A, provides summary statistics for the full sample. Panel B reports the statistics for variables grouped by strong (weak) country-level governance subsamples.

The tax evasion scores (TE) range between 0.5 and 7.2, with a mean of 2.61. In Panel B, we can observe that the mean score of tax evasion for the strong country-level governance subsample is 1.88, which is remarkably lower than the score for the weak country-level governance subsample (4.45). A t-test (=18.06***) confirmed that the difference in the tax evasion score between the strong (weak) country-level governance samples is statistically significant. The global ESG scores (CSREGS) range between 18.13 and 69.81 with a mean of 40.43 for the full sample, while for the strong and weak country-level governance subsamples, the mean scores are 38.66 and 44.88, respectively. These results show that the weak country-level governance subsample has higher CSR disclosures than the strong country-level governance subsample. To deepen this evidence, I performed a t-test in order to explore the significance of the differences in the CSR variables between strong (weak) country-level governance subsamples (CSREGS t-test = 3.34 ***, CSREnv t-test = 2.46 **, CSRSocial t-test = 2.93 *** and CSRGov t-test = −1.02). The results clearly indicate that for other than the CSR governance dimension, the weak country-level governance subsample reports significantly higher global CSR information, as is also the case in the environmental and social dimensions, than the strong country-level governance subsample. Cahan et al. [51] found contradictory evidence that revealed that CSR reporting activity was enhanced in countries with strong institutions. In terms of the country-level governance indicators, the WGI index (based on the factor analysis original scores of the aggregate measure of country-level governance) ranged from −3.60 to 1.05; the AUD scores were located between a minimum of 4 and a maximum of 32, and the ENF scores ranged between 2 and 24.

Table 2. Pearson correlation matrix.

	TE	CSREGS	CSREnv	CSRSocial	CSRGov	WGI	AUD	ENF	GDP	WW	PD	Ind	Mas	UA
TE	1
CSREGS	0.11	1
CSREnv	0.08	0.90	1
CSRSocial	0.06	0.91	0.78	1
CSRGov	−0.08	0.31	0.26	0.31	1
WGI	−0.82	−0.25	−0.19	−0.23	0.08	1
AUD	−0.62	0.10	0.08	0.11	0.07	0.38	1
ENF	−0.46	−0.06	−0.04	−0.06	−0.03	0.33	0.7	1
GDP	−0.33	−0.01	−0.11	0.03	0.05	0.13	0.36	0.35	1
WW	0.21	−0.02	−0.02	−0.04	−0.12	−0.13	0.01	0.15	0.34	1
PD	0.66	0.23	0.20	0.19	−0.01	−0.60	−0.38	−0.17	0.01	0.25	1
Ind	−0.79	−0.22	−0.24	−0.13	0.05	0.60	0.61	0.6	0.33	−0.10	−0.56	1
Mas	−0.00	−0.2	−0.21	−0.19	−0.12	−0.14	0.11	0.17	0.23	0.18	0.06	0.1	1
UA	0.55	0.21	0.23	0.15	−0.1	−0.66	−0.49	−0.31	−0.16	0.09	0.61	−0.60	0.19	1

Note: bold numbers indicate statistical significance at the 0.01 level.

presents the Pearson correlation matrix for the study’s variables. The correlation between tax evasion and the CSR variables is not statistically significant. Regarding the country-level governance variables (WGI, AUD and ENF), they are negative and statistically significantly associated with tax evasion. Furthermore, WGI is statistically significantly associated with all of the CSR variables (CSREGS, CSREnv, CSRSocial and CSRGov). These findings underscore the importance of exploring the institutional context and governance mechanisms to explain tax evasion at a country level as well as a mediating factor between CSR and tax evasion [5,6,8,23]. The correlation values between the independent variables of interest for models (1) and (2) do not suggest multicollinearity problems in my findings.

4.2. Multivariate Results

Table 3. Regression results.

	Model (1)	Model (2)	Model (3)	Model (4)
CSREGS	0.000271		−0.00859 *
	(0.00339)		(0.00447)
CSREnv		0.00263		−0.00469
		(0.00293)		(0.00423)
CSRSocial		−0.00539		−0.00181
		(0.00409)		(0.00539)
CSRGov		−0.00146		−0.0125
		(0.00742)		(0.0101)
WGI	−1.669 ***	−1.732 ***	−2.410 ***	−2.963 ***
	(0.111)	(0.115)	(0.274)	(0.660)
CSREGS*WGI			0.0177 ***
			(0.00604)
CSREnv*WGI				0.0117 **
				(0.00529)
CSRSoc*WGI				−0.00482
				(0.00801)
CSRGov*WGI				0.0223
				(0.0144)
AUD	−0.0602 ***	−0.0640 ***	−0.0595 ***	−0.0620 ***
	(0.00867)	(0.00878)	(0.00846)	(0.00850)
ENF	0.0128 *	0.0185 **	0.0171 **	0.0219 ***
	(0.00746)	(0.00799)	(0.00743)	(0.00815)
GDP	−0.0001 ***	−0.0001 ***	−0.0001 ***	−0.0001 ***
	(0.0001)	(0.0001)	(0.0001)	(0.0001)
WW	0.285 ***	0.201 **	0.273 ***	0.202 ***
	(0.0713)	(0.0787)	(0.0696)	(0.0770)
PD	0.0356 ***	0.0361 ***	0.0346 ***	0.0359 ***
	(0.00336)	(0.00340)	(0.00329)	(0.00335)
Ind	−0.0210 ***	−0.0198 ***	−0.0234 ***	−0.0220 ***
	(0.00314)	(0.00317)	(0.00317)	(0.00320)
Mas	0.00594 ***	0.00403 **	0.00636 ***	0.00445 **
	(0.00173)	(0.00186)	(0.00169)	(0.00180)
UA	−0.0349 ***	−0.0334 ***	−0.0354 ***	−0.0343 ***
	(0.00289)	(0.00303)	(0.00282)	(0.00296)
Constant	6.539 ***	6.691 ***	7.121 ***	7.460 ***
	(0.286)	(0.432)	(0.342)	(0.544)
Year Fixed Effects	✓	✓	✓	✓
Observations	165	158	165	158
R-squared	0.933	0.937	0.936	0.943

Notes: Data on CSR components are only available for 158 country–year observations. The variables definitions are in Appendix A. *, ** and *** indicate statistical significance at the 0.10, 0.05 and 0.01 levels, respectively.

reports the regression (OLS) results for the study’s models. Columns 1 and 2 relate to tests for Hypotheses 1 and 2, and columns 3 and 4 relate to tests for Hypothesis 3.

For all of the models reported in Table 3, the variance inflation factors (VIF) are lower than 10; thus, they do not suggest multicollinearity problems in the main findings [66].

Looking at the association between the CSR variables (CSREGS, CSREnv, CSRSocial and CSRGov) and tax evasion (columns 1 and 2), we observe that none of the coefficients is statistically significant. Therefore, country-level CSR is not associated with tax evasion, leading to the rejection of H1.

As discussed in Section 2, the findings of past research exploring the link between tax avoidance and CSR (overall measure) have failed to produce consistent results, and the same happens for studies examining the EGS dimensions. For instance, while Lanis and Richardson [3] found no significant impact on the environment, social and governance dimensions, Ortas and Gallego-Alvarez [36] provided evidence that all three of the ESG dimensions have a significant effect on tax avoidance; other studies have reported no significant impact on the environmental dimension [35,67] or on the governance dimension [35,68]. As Bird and Davis-Nozemack [6] argued, the direction and significance of the CSR and tax evasion relationship remain an intriguing question.

Regarding the national governance mechanisms, the overall results clearly indicate that countries with a higher quality of country-level governance present significantly lower levels of tax evasion. Except for the ENF variable, the WGI and AUD coefficients are negative and statistically significant at the 1% level, thus providing support for H2 (i.e., country-level governance is associated with tax evasion).

Consistent with Brown et al. [57], Bird and Davis-Nozemack [6], Salhi et al. [8] and Ermasova et al. [5], I conclude that the higher the country levels of government effectiveness, regulatory quality, rule of law and control of corruption (all components of the WGI variable) and the higher the quality of the public auditors’ working context (including requirements related to auditor training and independence, the professional quality assurance programme and monitoring mechanisms, and all components of the AUD variable), the lower the country levels of tax evasion.

Turning to the ENF variable, which is related to country-level quality of monitoring and promotion of compliance activity with accounting standards developed by independent enforcement bodies, it presents a slight positive significant effect on tax evasion. Kucera and Roncolato [69], and Schneider and Buehn [9] claimed that countries that were more heavily regulated, especially with respect to enforcement rather than the overall extent of regulation, tended to have a higher share of the unofficial economy.

Columns 3 and 4 of Table 3 show the results of regressing TE on CSR variables, WGI and their interaction.

In column 3, the CSREGS coefficient (β₁) is negative and significant at the 10% level, suggesting that, in countries with weak national governance, substantive overall CSR activities are slightly associated with significantly lower levels of tax evasion. In weak institutional contexts, businesses might voluntarily adopt high-quality CSR reporting systems, taking a differentiated position in the market; in such conditions, CSR and national governance are substitutes [23]. In fact, in the descriptive statistics, I already reported that the subsample of weak country-level governance presents significantly higher overall CSR activities than does the subsample of strong governance countries.

The coefficient of the interaction term (CSREGS*WGI) shows that the association between CSREGS and TE changes from negative (β₁ = −0.00859) to positive (β₁ + β₃ = 0.00911), with an incremental variation of 0.0177 (statistically significant at the 1% level). Contrary to my prediction, this evidence suggests that, in countries with stronger governance, substantive overall CSR activities are associated with statistically significantly higher levels of tax evasion. Moreover, the interaction terms of the ESG dimensions and WGI, which are reported in column 4, show that only CSREnv*WGI is positive and statistically significant at the 1% level. This further indicates that, in countries with strong national governance and in which businesses engage in higher CSR environmental disclosures, the levels of tax evasion are higher. Under the risk management theory [39,40] and the theoretical lens of organisational façades and organised hypocrisy [38,41], the evidence collected is consistent with the idea that, in countries with strong national governance, CSR reporting, and in particular, environmental disclosures, can become corporate “camouflaging” tools, enabling firms to achieve self-interest goals, to improve their image and to mitigate public alarm arising from tax evasion.

Overall, the results indicate that national governance mediates the relationship between country-level CSR and tax evasion, thus providing support for H3. Additionally, they underscore the importance of considering CSR dimensions separately when investigating the relationship between tax evasion and CSR [42].

Regarding the control variables, Table 3 shows that they are statistically significant. The coefficients on GDP and WW are negative (positive) and significant at the 1% level. Consistent with Zeng [23], countries with higher economic development and growth present lower levels of tax evasion. In contrast with Zeng [23] and Atwood et al. [46], I found significant evidence that firms in countries with tax systems based on a worldwide approach engage in more tax evasion activities than firms in countries with territorial tax systems. A possible explanation is that multinational companies located in countries with worldwide tax systems may be able to use international tax management techniques that allow them to achieve levels of tax evasion equivalent to that of multinational companies located in countries with territorial tax systems [46].

The coefficients of the four original Hofstede’s cultural dimensions reveal that, (1) the higher the country-levels of power distance (PD) and masculinity (Mas), the higher the tax evasion and, (2) the higher the country levels of individualism (Ind) and uncertainty avoidance (UA), the lower the tax evasion.

Consistent with Ortas and Gallego-Alvarez [36], and Tsakumis et al. [70], the results suggest that firms that reach greater power in high-power distance countries may use it to improve their financial wealth and market position through tax evasion practices. Furthermore, high-masculinity societies are less committed to common well-being [71]. Therefore, it is not surprising that high-masculinity countries are positively connected with tax evasion practices [36,62].

Hofstede [72] argued that highly individualistic societies assume that legal norms (including tax laws) are universal. Uncertainty avoidance refers to the extent to which members of a society feel uncomfortable with uncertainty and ambiguity [36,62] and the tendency they have to eliminate these types of occurrences by adopting strict behavioural codes, laws and regulations and by disapproving of these deviant attitudes [62]. Thus, in high individualist and high uncertainty avoidance countries, tax evasion is lower, mainly because people are less tolerant of aggressive tax practices [36,62,73].

4.3. Robustness Tests

4.3.1. Cross-Sectional and Time-Series Dependence

Since panel data were used in this study, the baseline results may suffer from both cross-sectional and time-series dependence. Thus, I re-estimated Equations (1) through (4) using a random effects approach. As several independent variables present low within-country variation (e.g., WGI, Aud, Enf and GDP) or even are time invariant (WW, PD, Ind, Mas and UA), the fixed-effects approach is not appropriate for this analysis. As Wooldrige [74] argued in cases in which key independent variables do not vary much (or are constant) over time or have low within-variation, the fixed-effect method can lead to imprecise estimates and can eliminate the estimation of variables that are constant over time.

The results of the re-estimation using random-effects are reported in

Table 4. Random-effects GLS regression results.

	Model (1)	Model (2)	Model (3)	Model (4)
CSREGS	0.00041		−0.00483 *
	(0.00270)		(0.00284)
CSREnv		0.00203		−0.00158
		(0.00217)		(0.00315)
CSRSocial		−0.00173		−0.00243
		(0.00304)		(0.00418)
CSRGov		−0.00659		−0.01447
		(0.00501)		(0.00758)
WGI	−0.73715 ***	−0.76313 ***	−1.48687 ***	−2.21293 ***
	(0.14478)	(0.14843)	(0.24574)	(0.54432)
CSREGS*WGI			0.01731 ***
			(0.00429)
CSREnv*WGI				0.00738 **
				(0.00406)
CSRSoc*WGI				0.00383
				(0.00639)
CSRGov*WGI				0.01659
				(0.01059)
AUD	−0.01055 ***	−0.01623 ***	−0.01373 ***	−0.01799 ***
	(0.00905)	(0.00975)	(0.00851)	(0.00941)
ENF	0.00413	0.00539	0.00198	0.00366
	(0.00711)	(0.00742)	(0.00666)	(0.00752)
Observations	165	158	165	158
R-squared	0.889	0.889	0.888	0.919

Notes: Data on CSR components are only available for 158 country–year observations. The variables definitions are in Appendix A. *, ** and *** indicate statistical significance at the 0.10, 0.05 and 0.01 levels, respectively.

(to save space, I have not reported the coefficients of the control variables). In summary, regarding the explanatory variables in Equations (1) through (4), the random-effects estimation produces evidence qualitatively similar to the main results (except for the ENF variable that remains positive but is now statistically insignificant). The coefficients of control variables are also consistent with that reported in Table 3.

In additional analysis, models (1) through (4) of Table 3 and Table 4 were re-estimated with robust standard errors, and the results (untabulated) remained qualitatively similar. Unit-root tests for panel data were also performed. For instance, the Hadri Lagrange multiplier test revealed that all of the variables contained unit roots at the panel level (are not stationary).

I thus conclude that the main results of this study seem not to be affected by the issue of cross-sectional and time-series dependence.

4.3.2. Alternative Measures of Tax Evasion, CSR and WGI, and the Impact of Financial Crisis on Results

I re-estimated regression models (1) through (4) of Table 3 with alternative measures of the tax evasion, CSR and WGI variables.

In the main analysis, I used the Buehn and Schneider [22] country-level measure of tax evasion computed considering indirect tax burden (taxes on goods and services) and self-employment as driving forces of tax evasion. However, Buehn and Schneider [22] presented an alternative measure of tax evasion in which they only considered the indirect tax burden (taxes on goods and services) as a driving force of tax evasion. The untabulated re-estimated results of models (1) to (4) using the alternative measure of TE are consistent with those presented in Table 3.

The main results of Table 3 are based on the CSR variables (CSREGS, CSREnv, CSRSocial and CSRGov) computed at the country level using the median of the annual ESG scores of all companies included in the ASSET4 database (for details, see Section 3.3). Using alternative measures of the CSR variables computed based on the mean annual ESG scores (rather than the median) to re-estimate models (1) to (4) produces qualitatively similar evidence to that reported in Table 3 (unreported results).

For the baseline results reported in Table 3, I used a main country-level governance variable (WGI). The WGI results from using factor analysis to combine the four WGI dimensions of governance into an aggregate measure with values ranging from −3.60 to 1.06. The negative (positive) values are then considered weak (strong) country-level governance levels, and a dummy variable is created, equal to one if the country has strong governance and zero otherwise (details are in Section 3.4). Following Zeng [23], as a robustness check, I replaced the WGI variable by considering a dummy variable equal to one if the country value of the aggregate measure of governance falls in the top 20% (i.e., strong country-level governance) and equal to zero if the country value of the aggregate measure of governance falls in the bottom 20% (i.e., weak country-level governance). Despite the smaller sample size, the untabulated results of the re-estimations of models (1) to (4) are qualitatively similar to that observed in Table 3.

Finally, to account for the impact of the international financial crisis on the main results, I re-estimated models (1) through (4) with a short panel data of three years (2008, 2009 and 2010). The untabulated results are qualitatively similar to that presented in Table 3.

5. Conclusions

This study was motivated by the multidisciplinary nature of the subject of tax evasion and its determinants as well as the lack of country-level research on the relationships between CSR, national governance and tax evasion.

Using a sample of 165 country–year observations from 25 OECD countries for the period from 2004 to 2010, this paper examined the associations between CSR, national governance and tax evasion at the country level. The moderator effect of national governance on the relationship between CSR and tax evasion was also explored.

The results failed to support whether CSR disclosures at the country level significantly explain the national variations in tax evasion. Neither the ESG dimensions nor the overall CSR measure proved to be a significant direct determinant of tax evasion at the country level.

In contrast, I find strong evidence that national governance is a significant determinant of tax evasion at the country level. The higher the country levels of government effectiveness, rule of law, regulatory quality and control of corruption, the lower the tax evasion. Additionally, the higher the national quality of the public auditors’ working context (including requirements related to auditor training and independence, the professional quality assurance programme and monitoring mechanisms), the lower the tax evasion levels.

Furthermore, I found robust support for the mediator effect of national governance on the relationship between CSR and tax evasion. On the one hand, the findings suggest that, in countries with weak national governance, substantive overall CSR activities are slightly associated with significantly lower levels of tax evasion, implying that CSR and country-level governance act as substitutes. On the other hand, I found significant evidence that, in countries with stronger national governance, substantive overall CSR activities (especially environmental disclosures) are associated with significantly higher levels of tax evasion. In consonance with the concepts of organisational façades and organised hypocrisy [41], the results suggest that, in countries with strong national governance, CSR reporting, especially for environmental disclosures, can become a cosmetic and compensatory tool for businesses to achieve self-interested goals, to polish their image or to mitigate public concern resulting from their socially irresponsible actions. Overall, the results indicate that country-level governance mediates the relationship between country-level CSR and tax evasion and underscores the importance of separately considering CSR dimensions when investigating the relationship between tax evasion and CSR [42].

The results further highlight the role of national culture as a determinant of tax evasion and are robust with several additional tests.

This study contributes to accounting and business ethics literature. The findings are theoretically relevant as they underscore the importance of national governance in preventing tax-aggressive actions and the relevance of the mediating effect of national governance on the link between CSR and tax evasion. Given the lack of country-level research on CSR, national governance and tax evasion, this paper fills this gap in the literature.

The results also have significant implications for policymakers. Overall, they emphasise the importance of national governance in mitigating/preventing tax evasion. The results document the importance of not only promoting national government effectiveness, regulatory quality, the rule of law and the control of corruption but also the quality of the public company auditors’ working environment (requirements for auditors’ training and independence, the professional quality assurance programme and monitoring mechanisms). The evidence collected also highlights the need for policymakers in countries with strong national governance to design strong anti-tax avoidance regulations as businesses seem to use CSR activities to mitigate public concern resulting from tax evasion activities.

Notwithstanding its contributions, this study has inherent limitations. First, the research design used in this study only allows me to claim an association rather than a causal effect, between country-level CSR, national governance and tax evasion. Second, the regression models used in this study may suffer from endogeneity related to omitted variables and measurement error (since all variables are computed at the country level). Though I have considered the relevant determinants of tax evasion used in previous studies, I have used fixed-year effects and added several robustness tests with alternative specifications, I recognise that it is very difficult to properly deal with endogeneity issues. Therefore, a careful interpretation is required. Third, the sample data (2004–2010) can be considered outdated but the relevance of the Buehn and Schneider [22] tax evasion measure (which is only available for this period) makes the analysis performed interesting.

Future research could investigate the association between CSR, national governance and tax evasion using data from a larger sample of countries and from emerging economies.