The Effect of ESG Performance on Tax Avoidance—Evidence from Korea

We analyzed whether a firm’s engagement in socially responsible activities, as measured by environmental, social, and corporate governance (ESG) scores, influences their tendency to avoid tax in the Korean financial market. We found a negative relationship between Korean firms’ ESG scores and tax avoidance in terms of book-tax income difference during the sample period between 2011 and 2017. This result implies that firms with good CSR performance would tend not to manipulate taxable profits, which is in line with corporate culture theory. More interestingly, this trend has become more apparent for chaebol-affiliated firms, a special type of Korean conglomerate, than non-chaebol firms.


Introduction
Environmental, social, and corporate governance (ESG) initiatives are usually motivated by a laudable desire to raise the profile of ethical considerations in business. In most cases, these initiatives are driven by corporate social responsibility (CSR), a management concept that integrates social and environmental concerns in business operations and interactions with stakeholders [1]. This concept was first defined and developed by the United Nations. Subsequently, the ESG concept was proposed in the United Nations Principles of Responsible Investment report, which recommends that investors consider ESG scores as a key factor in investment decisions.
In practice, a wide variety of stakeholders, such as management consulting firms and investors, adopt ESG scores as a major index to understand a firm's overall CSR performance. ESG essentially evaluates a firm's environmental, social, and corporate governance practices and combines the performance of these practices. The ESG scores represent a firm's involvement with environmental, social, and governance activities beyond the classical measures of financial performance [2].
The literature on CSR activities has related firms' economic goals to their social responsibilities [3][4][5]. Because firm managers' main objective is to generate profit while reducing costs, it is natural for firm managers to act in accordance with the needs of shareholders [6,7]. In this profit generation process, the reduction of the tax burden is an important economic goal of managerial activities. In fact, previous studies confirm the effect of taxes on financial choices, organizational form, and restructuring decisions, as well as payout policies, compensation policies, and risk management decisions [8,9].
It may be quite natural for managers to manipulate profits to reduce the tax burden to maximize profits. However, corporate culture theory [10,11] suggests that this may not be the case for socially responsible firms. In the culture theory context, CSR is the belief about the "right" course of actions, where a socially responsible firm considers economic, social, environmental, and other externalized effects of corporate decisions. Because the public views aggressive tax avoidance practices as costly to society and these practices are widely accepted as unethical and irresponsible behavior; thus, socially responsible firms are less likely to avoid tax responsibility by manipulating profits. Specifically, the practice of tax avoidance may arouse negative sentiment, such as the loss of firm management personnel, political pressure, potential fines, and consumer boycott [12,13]. Therefore, a firm with higher ESG scores is less likely to participate in aggressive tax avoidance, which places them at public risk.
By contrast, some studies in the areas of CSR suggest a positive association between ESG scores and tax avoidance. The theory highlights the management of tax avoidance as a tool to enhance a firm's reputation for CSR practices. Specifically, if a firm has already conducted socially irresponsible activities, the firm will attempt to avoid manipulating taxable income as tax avoidance will exacerbate the damage to the firm's CSR reputation [14]. Accordingly, a firm with a lower ESG score is not inclined to use tax avoidance, which potentially implies a positive association between the two variables. This theory highlights tax avoidance as a risk management tool with regard to CSR reputation.
This study empirically examines how a firm's CSR performance affects tax avoidance in a sample of the Korean financial market. We adopt the ESG score, widely used in the literature, published by the Korean Corporate Governance Services to capture the degree of engagement in socially responsible activities. The total ESG score is mainly used to capture a firm's overall CSR performance, whereas each category of environmental, social, and governance scores is also employed. The tendency to avoid tax is measured by the book-tax income difference (BTD), in line with the extant literature [15][16][17].
We deliberately chose the Korean financial market to examine these two contrasting hypotheses for several reasons. First, well-constructed ESG scores are available for a wide range of corporations, covering almost all stocks listed in the KOSPI market. More interestingly, the Korean financial market has a unique set of family-owned business conglomerations, the so-called chaebols. These chaebols have been the main vehicle leading the nation's dramatic economic growth with the support of the government. Over the last decades the critics' demand for their reformation has grown because of their association with political scandals, which causes owner risk. In this social atmosphere, socially responsible management practice has become one of the major agendas in these demands for both chaebol and non-chaebol firms [18,19]. While chaebols have come under wide criticism and have undergone structural changes, regardless of their state, they have faced and continue to meet new "social environmental" demands, which we verify and discuss in our analysis.
This consideration of chaebol affiliates is interesting from the recent development of the literature with regard to corporate governance and CSR activities. A branch of corporate governance literature highlights that the valuation effect of tax avoidance depends on a firm's corporate governance structure. For instance, firms with poor governance may not benefit from tax sheltering, a type of tax avoidance, because it may also indicate a higher likelihood of managerial diversions. A stream of CSR literature highlights more significant public pressure on CSR activities for chaebol affiliates. This pressure results in significantly different relationships between CSR activities and valuation or disclosure quality across chaebol affiliates and non-chaebol affiliates [20,21].
Our main findings are as follows. First, we found a statistically significant negative relationship between a firm's ESG score and BTDs. This negative relationship implies a lower tax avoidance tendency for more socially responsible firms, consistent with corporate culture theory. This finding is also in line with recent empirical studies that support negative relationships [22][23][24].
We also found that this negative relationship is driven mainly by firms with higher social scores. The environment and governance scores did not show significant negative relationships with the measure of tax avoidance. This finding can be naturally interpreted using the construction method of the ESG score. The potential pitfalls of tax avoidance, for instance, loss of firm management personnel, political pressure, potential fines, and consumer boycott, are closely associated with the social reputation of corporations. A firm with a higher social reputation as measured in the social score component of the ESG score is more reluctant to participate in tax avoidance, which seriously damages its social status.
Finally, we found that the negative relationship is more pronounced for non-chaebol affiliates than for chaebol affiliates. In particular, the estimated coefficient on the social score in the group of non-chaebol affiliates was four times larger than that of chaebol affiliates, which indicates a stronger tax avoidance for socially responsible non-chaebol affiliates. This might be because non-chaebol affiliates' social reputation can be more significantly affected by tax avoidance. Chaebol affiliates are already under strict regulations from the Korea Fair Trade Commission (KFTC) with regard to crossholdings, dividend policy, capital budgeting, and other governance structures. Moreover, a wide range of chaebol affiliates' corporate policies have received extensive media attention because of their massive effect on the overall Korean economy. Chaebol affiliates may already have limited incentives to engage in tax avoidance because of strict regulations and media attention, and thus, the tax avoidance tendency is less sensitive to the variation in CSR performance within the group.
Our contribution is twofold. First, this study provides empirical evidence supporting the culture theory of corporations with respect to the decision to avoid tax. We found a robust negative relationship between the ESG score and the tax avoidance tendency [10]. We further highlight the significance of other factors in shaping such negative relationships. Specifically, our results emphasize the importance of the social score in shaping the relationship and highlight a more strongly negative relationship for non-chaebol affiliates. This finding is in line with recent studies that focus on the heterogeneous effects of CSR practices on various corporate decisions [20,21].
The remainder of this paper is organized as follows: Section 2 illustrates previous CSR literature leading to our research. Section 3 introduces the hypothesis development, and Section 4 explains the data and methodology, followed by an analysis of the results in Section 5. Finally, Section 6 concludes our research.

ESG Overview
In CSR research, it is important to use standardized and uniformly comparable CSR information for analysis. As such, the ESG score measures a firm's performance in environmental, social, and corporate governance practices. More precisely, a firm's environmental performance indicates its effort to reduce resource consumption and emissions. A firm's social performance represents respecting human rights, the quality of employment, the responsibility of the product, and community relations. Lastly, a firm's corporate governance performance refers to the rights and responsibilities of the management of a firm.
The ESG scores published by external rating agencies such as KLD (Kinder, Lydenberg, and Domini), Bloomberg, and Thomson Reuters Eikon, are widely employed in the literature. However, most of these data focus on the U.S. market primarily and do not provide financial and firm data for a specific country. Thus, our dataset employs ESG performance scores published by the Korea Corporate Governance Services (KCGS), which includes all firms listed on the KOSPI market. Similar to our work, several recent studies on Korean firms' CSR performance use the ESG data published by the KCGS because of its consistent measurement of ESG performance and wide coverage of listed companies [20,21].

Tax Avoidance and CSR
Previous research suggests that firms' tendency to avoid tax, as best represented by the BTD, is motivated by a wide variety of factors. Among them, Desai and Dharmapala [22] are a representative example. They argue that increases in incentive compensation tend to reduce the level of tax sheltering, suggesting a complementary relationship between diversion and sheltering, leading to the conclusion that tax avoidance activities are increasingly central to corporate financial decision making.
In the context of culture theory and tax avoidance, greater engagement in CSR practices implies a lower tax avoidance tendency. CSR is the belief about the "right" course of actions and, thus, a more socially responsible firm attempts to avoid tax manipulation, which incurs costs to their societies. For example, using measures of both permanent and abnormal book-tax differences, as well as the cash effective tax rate (CASH ETR), Hoi et al. [16] successfully demonstrated that firms with excessively irresponsible CSR activities have a higher likelihood of engaging in tax-sheltering activities and greater discretionary/permanent book-tax differences.
Interestingly, studies on international firms strengthen the argument on negative relationship between CSR and tax avoidance. Based on an international sample of 6442 firm-year observations between 2006 and 2014, López-González et al. [23] found that firms' performance in social and environmental areas is negatively related with tax avoidance, meaning that firms with socially responsible performance demonstrate a lower tendency of tax avoidance. The control variables were the natural log of the ratio of total debt to total equity, Tobin's Q, the natural log of total assets, and the ratio of R&D. A binary variable was also used; the variable was coded 1 if income before extraordinary items was less than 0 and 1 otherwise. Using panel-data regression models, they concluded that firms with a great commitment to CSR activities demonstrate greater focus on the reduction of activities of tax avoidance.
As well, controlling for firm size, industry, region, and other factors, Jones et al. [24] also found a negative relationship between CSR levels and tax aggressiveness for international firms. However, when their samples were analyzed by region, the CSR-tax avoidance relationship was significant in the Asian sub-sample, but largely insignificant in the North American, European, and UK sub-samples. In their study, the control variables were the following: the natural log of the market value of equity, proportion of shares of insiders, return on assets, leverage, cash and other marketable securities, plant, property and equipment, total sales, the market-to-book ratio, operating cash flows, and robust distress proxy. Using a Tobit regression, they concluded that firms manage reputation through CSR reporting or activities, and this, in turn, can prevent the risk of negative corporate events, including tax avoidance.
However, another branch of corporate theory predicts a lower tax avoidance tendency for firms with poorer CSR practices [14]. This literature argues that socially irresponsible firms hesitate to engage in tax avoidance. These additional tax management activities deteriorate firms' social reputation significantly. Therefore, the theory expects a positive relationship between the degree of CSR activities and the tax avoidance tendency.
A recent study focusing on particular regions further supports this positive CSR-tax avoidance relationship. Mao [25] suggests that CSR firms have a higher book-tax difference, indicating that they are more aggressive in tax avoidance, engaging in CSR activities as a risk management strategy. In his study, Mao used the Probit model and regression analysis against around 20,000 Chinese listed firms from 2009 to 2016. The control variables were firm size, leverage, return on assets (ROA), sales revenue changes, intangible asset intensity, and absolute value of performance-adjusted abnormal accruals. He concluded that as firm managers may strategically use CSR as a means to conceal their tax avoidance tendency, the Chinese government should implement CSR audits to supervise the CSR activities of the firms. As well, Lanis et al. [26] presented a consistent positive relationship between CSR disclosure levels and tax aggressiveness, which indicates that high CSR disclosure relates to tax aggressiveness. They used a choice-based sample of 40 corporations between 2001 and 2006, which was composed of 20 tax aggressive firms and 20 non-tax aggressive firms, using various statistical methods such as paired sample statistics, Pearson correlation analysis, and ordinary least squares (OLS) regression analysis. The main control variables for matching samples were the natural log of total assets, long-term debt divided by total assets, net property, plant and equipment divided by total assets, the market value of equity divided by the book value of equity, and pre-tax income divided by total assets. As well, a dummy variable was also employed to check if the corporation was accused of tax aggressiveness by the Australian Tax Office: 1 if the corporation was accused, and 0 otherwise. They concluded that, considering the level of association between corporate tax aggressiveness and CSR disclosure, firms' legitimacy on society through CSR is confirmed.
The main characteristics of the aforementioned studies regarding the relationship between CSR and tax avoidance, on both the international and regional level, are summarized in Table 1 below. Table 1. Previous research on relationship between CSR and tax avoidance (by region).

Lanis et al. (2012) Australia
Natural log of total assets, long-term debt divided by total assets, net property, plant and equipment divided by total assets, the market value of equity divided by the book value of equity, pre-tax income divided by total assets, and a dummy variable; 1 accused by ATO, 0 otherwise High CSR disclosure related to tax aggressiveness (40 firms, 2001 to 2006)/Firms' legitimacy on society in the context of corporate tax aggressiveness is confirmed

Jones et al. (2017) International
Natural log of the market value of equity, proportion of shares of insiders, ROA, leverage, cash and other marketable securities, plant, property and equipment, total sales, market-to-book ratio, operating cash flows, and robust distress proxy.
Negative relationship between CSR levels and tax aggressiveness (2551 public firms as of September 2014)/firms manage reputation through CSR, which prevents the risk of tax avoidance.

López-González et al. (2019) International
Natural log of the ratio of total debt to total equity, Tobin's Q, natural log of total assets, ratio of R&D. An indicator variable: coded 1 if income before extraordinary items less than 0 and 0 otherwise. In terms of tax avoidance measurements, the extant literature has examined the effect of CSR on a firm's tax avoidance by constructing a variety of proxy variables based on financial statements. As mentioned above, Hoi et al. [16] used the book-tax difference as a measure of tax avoidance in their CSR study. To analyze the relationship between CSR, tax management fees, and tax avoidance, Huseynov and Klamm [15] used the following two effective tax rates (ETRs) as tax avoidance measures: the GAAP ETR (Tax Expense/Pre-tax Income) and Cash ETR (Cash Taxes Paid/(Pre-tax Income-Special Items)).

CSR Effect in Korean Firms
Before attempting to embark upon our main analysis, it is worthwhile to consider the South Korean corporate culture. South Korea is widely known for its family-centered business model. This is a mechanism of traditional familism developed on the basis of Confucianism, where loyalty to your family is associated with the interest of the whole family, namely, a corporation. This familism deepens in the case of chaebols, family-owned multi-industry conglomerates in which inheritance relationships based on blood play an essential role. Usually, large-sized chaebols contain 40-50 member firms that are owned and controlled by the original founder. According to the official KFTC website, the number of corporate groups and affiliated firms has grown from 30 to 64 groups, containing 224 firms and 173 mother firms, as of July 2020 [27].
However, whereas there has been a great deal of research on the interaction between CSR activities and firms' tendency for tax avoidance, the relationship is under-researched in developing countries, including South Korea, probably because of the lack of reliable and comprehensive measurement of CSR practices. [28,29] Korea is also an ideal place to examine whether firm characteristics play a significant role in shaping the effect of CSR practices on various corporate policies; financial structure, governance structure, and social reputation differ significantly across chaebol and non-chaebol affiliates [21]. In general, chaebol-affiliated firms have excelled in profit performance. Chang and Choi [30] proved that profits for group-affiliated Korean firms were substantially higher than independent firms, based on the finding that transaction costs are higher for independent firms because of market imperfections. Regarding this finding, Choi and Cowing [31] criticized the results as inconsistent because of Korean market imperfections that favor group-affiliated firms, in the sense that they take substantial advantage of increased intragroup transactions and reduced costs, thereby increasing profits.
Research on chaebol-affiliated firms' tendency to avoid tax suggests that they are more inclined not to manipulate taxable profits. For example, Koh et al. [32] found that, whereas the relationship between chaebols and BTD is negative, the relationship is positive for non-chaebols because the latter has more incentive to avoid tax in Korea. A striking point in their research is that the difference between net income and taxable income appears to be negligible for chaebols, which may result from different accounting standard or tax law interpretations between Korean firms.
This discussion naturally leads to the assumption that Korean firms, especially chaebols, with high CSR performance would tend not to affiliate with activities involving tax avoidance. In fact, recent studies have presented empirical results that support this assumption. For example, Park and Kim [28] argued that Korean firms' CSR activities, as measured by ESG scores, have a statistically meaningful negative relationship with tax avoidance. In addition, Kim et al. [29] suggested that, based on the sample of 600 listed firms, firms that actively promote CSR performance restrain their tendency to avoid tax. Park and Kim [28] reached similar conclusions. However, both studies adopted a very limited sample of fiscal years; the former used the sample period from 2005 to 2007, and the latter adopted the period from 2010 to 2012. Furthermore, their research only focuses on Korean firms in general, not differentiating chaebol-affiliated groups from non-chaebol affiliated groups for their study, and not examining the effect of environmental, social, and governance performance on the tendency to avoid tax separately. We extended the sample period significantly to a total of 7 years, from 2011 to 2017, in order to obtain concrete results and highlight the importance of firm characteristics in deriving the relationship between CSR and tax avoidance.

Hypothesis Development
We now develop the empirical hypotheses to be tested. First, we expect a negative relationship between ESG scores and tax avoidance in Korean firms. Korean firms are under increasing scrutiny from shareholders, such as national pensions and a variety of ESG funds that demand a higher level of social responsibility and transparent disclosure of CSR practices. A more socially responsible firm now has a stronger incentive to avoid tax manipulation, which will damage its social reputation if the manipulation is publicly noticed.

Hypothesis 1 (H1).
There is negative relationship between Korean firms' ESG scores and the measure of tax avoidance.
Next, it is highly likely that the negative relationship between CSR performance and tax avoidance would be more apparent for non-chaebol-affiliated firms than for chaebolaffiliated firms. As argued, chaebol affiliates are under stricter restrictions in terms of financing, dividends, and crossholding policies. These firms are also known for their vast media attention and have less information asymmetry in their financial statements [18]. Chaebol affiliates already have a strong incentive to avoid tax manipulation and, hence, a smaller sensitivity to tax avoidance with regard to CSR performance.

Hypothesis 2 (H2).
The relationship between ESG scores and the tendency to avoid tax is stronger for non-chaebol affiliates than for chaebols.
Finally, we predict that social scores are most closely related to the measure of tax avoidance within the Korean corporation. The total ESG score consists of the environmental, social, and governance scores. Because tax avoidance reduces the amount of tax revenue for society, the verification of tax avoidance more seriously damages a firm's social reputation than its reputation for environmental and governance activities.
Of course, self-interested managers have incentives to reduce the likelihood of tax avoidance with good governance structure, especially in terms of effective auditing system. However, if a firm's good governance indicates a stronger monitoring process that prevents managerial diversions, it is possible that a good governance structure is positively related to the tax avoidance. This is because the tax avoidance could increase operating profits of a corporation, which may enhance the shareholder value of firm. As long as managers could not divert the firm's profits into their pockets, tax avoidance could have positive valuation effects and could become a corporate policy in line with the interests of shareholders. Thus, a good governance structure from the perspective of effective monitoring process could be positively correlated with the tendency of tax avoidance.
The Korean market strengthened a firm's internal monitoring process after the financial crisis of 2008 and the Asian crisis of 1997, mostly by expanding the role of the independent director in monitoring managerial diversions. This stronger monitoring system may provide additional incentives of tax manipulation for the CEOs, which may enhance the shareholder value of corporations. Accordingly, we expect the most significant role to be of the social score in deciding tax avoidance for the Korean corporations compared to that of the environmental or governance score.

Hypothesis 3 (H3).
The negative relationship between the ESG score and tax avoidance is most apparent if the social score is adopted in the sample of Korean firms.

Data Construction and Empirical Model
We used the BTD as the dependent variable that estimates the degree of tax avoidance, which is the most frequently used measure. There are other measures for tax avoidance such as the GAAP ETR (Tax Expense/Pre-tax Income) and Cash ETR (Cash Taxes Paid/(Pre-tax Income-Special Items)) [15,16]. Among them, BTD is frequently chosen because it directly measures the tendency of tax avoidance by just comparing the amount of net income and taxable income in an intuitive way.
In calculating the BTD, scholars have used the variables of net income before income tax expense and taxable income estimate, dividing the former by the latter. Whereas it is possible to extract the former, there has been a wide array of discussions on defining and using appropriate proxies for the BTD [15][16][17]22].
In our study, we employed Cho and Yoon's [33] method in which the BTD is calculated by subtracting the taxable income estimate from the net income before income tax expense. The calculation for the BTD is as follows: BTD = Y s −Ŷ T BTD: Difference between Net Income and Taxable Income Y s : Net Income (Before Income Tax Expense) Y T : Taxable Income Estimate As for the taxable income estimate (Ŷ T ), it was difficult to obtain data directly; therefore, we used the proxy introduced by Park et al. [34], as presented in Table 2. As described in their study, the adjusted tax expense is multiplied by different tax rates determined by the government in Korea. To be specific, taxable income is estimated on the basis of the firm's tax expense and its tax base amount, as well as the progressive tax rate accordingly. The calculation method is the following: Y T = (Tax Expense/1.1-tax base amount (in Won) × tax rate)/progressive tax rate + tax base amount A lower BTD value implies a smaller likelihood of tax avoidance because it means that a firm's taxable income is greater than its net income in a given fiscal year. If a firm wants to avoid tax, it reduces taxable income as much as possible, which eventually induces a positive BTD value.
To control for the effect of other firm-specific factors on tax avoidance, we incorporated Huseynov and Klamm's [15] model. In line with their model for the analysis of tax avoidance, we employed the following financial and firm characteristic variables as dependent variables: LogTA (natural logarithm of total assets), leverage (total debt/total equity), DivDummy (1 if firm pays out dividends, 0 otherwise), PB (price per share/book value of equity per share), ROA (earnings/total assets), CapExp (total capital expenditures/total assets), and advertising expenses (advertising expenses/total assets). The definitions of the variables used in our empirical model are summarized in Table 3. For our research, the individual data that composed each financial and firm variable, such as ROA and CapEXP, were collected from the FnGuide, a data-providing company. We also separated chaebol and non-chaebol affiliates by using the category item published in the FnGuide. Finally, ESG scores, which were used as CSR variables, were obtained from the KCGS. We also obtained each category of ESG scores separately. Accordingly, we adopted four categories of CSR variables in our estimation, namely, the sum of ESG scores (ESG_SUM), environment scores (ESG_ENV), social scores (ESG_SOC), and corporate governance scores (ESG_GOV).
Before conducting the main analysis of our model, it is worthwhile to discuss the economic rationale behind the choice of control variables. First, capital structure and profitability affect taxes significantly. The tax deductibility of interest payments is one reason firms raise capital by issuing debt [7]. We controlled for this effect by introducing a leverage variable, Leverage, measured by total debt divided by the book value of total equity. Furthermore, a higher profitability increases the marginal tax rate faced by a corporation, which provides an incentive for tax avoidance. We controlled for this profitability effect by considering the return on asset (ROA) variable.
Next, we controlled for firms' investment and operating cost structures to test our hypotheses. Firms in capital-intensive industries could have a lower tax rate because of the substantial depreciation from capital expenditure; therefore, we used capital expenditure divided by total assets, namely CapExp, to control for this effect. The price-to-book (PB) ratio was also introduced to capture a firm's investment opportunities as well as industry heterogeneity. A firm's higher PB ratio indicates a greater investment opportunity in the future or the firm's inclusion in a capital-intensive industry. Moreover, we used advertising expenses divided by total assets, AdvExp, as a control variable. Fry et al. [35] report that firms' contributions and advertising expenditure are complementary; thus, it is reasonable to argue that firms with high advertising expenses are more likely to abide by social rules, including tax responsibility, hoping to avoid public criticism. Dyreng et al. [36] also demonstrated that long-run tax avoiders spend less on advertising, which is consistent with the idea that firms that are aware of public criticism may not avoid taxes to such a large degree because they fear a public backlash for poor corporate citizenship.
Finally, we controlled for a firm's life cycle. The natural logarithms of total assets and dummy variables for dividend payments were included in our empirical model. A large firm with dividend payments is probably an older company, which may have already established a long range of public records for its financial performance and determination of financial income. These firms might have different incentives for tax avoidance than smaller and younger firms.
Our empirical model can be described as follows: First, in Equation (1), we tested the relationships between CSR performance and Korean firms' tax avoidance by regressing the difference between net income and taxable income (or BTD) on ESG variables, controlling for other firm-specific variables. Here, the ESG variables encompass the sum of ESG (ESG_SUM), environmental (ESG_ENV), social (ESG_SOC), and governance (ESG_GOV) scores. We adopted the ordinary least squares method (OLS) to estimate our empirical models in line with Huseynov and Klamm [15] and many others [31,33]. The ordinary least square method is widely used in the literature because of its robustness from the fact that the method relies on a limited set of assumptions to obtain consistent estimators. Because a firm's financial variables for a fiscal year is substantially affected by firm-specific events, such as mergers and acquisitions, managerial turnovers, and the outcome of R&D projects, the ordinary least square method is known to be one of the best methods to test empirical hypotheses with financial variables. BTD = β 0 + β 1 LogTA + β 2 Leverage + β 3 DivDummy + β 4 PB + β 5 ROA + β 6 Capex + β 7 AdvExp + β 8 ESG variables + ε it . . .
Finally, we conducted a subsample analysis based on the categories of chaebol and nonchaebol affiliates. Specifically, from the FnGuide, we downloaded all data for Korean listed firms between 2011 and 2017, and selectively chose those affiliated with chaebols, as defined by the KFTC, and assigned them to a chaebol group. Because the ESG score provided by KCGS covers all the sample of firms listed in the KOSPI market, the biggest stock exchange in Korea, the subgroup of chaebol affiliates covered almost all of the chaebol affiliates listed in the KOSPI market except a small set of companies with impaired observations. The remaining firms were assigned to the non-chaebol group. We estimated the above two empirical models separately for chaebol and non-chaebol affiliates. For our analysis totaling 5612 firm-year observations, 1472 firm-year observations were assigned to the chaebol group and 5196 firm-year observations were assigned to the non-chaebol group. Table 4 presents the summary statistics of the ESG scores, along with descriptive statistics for the corporate financial information variables and each ESG category score. The table includes the mean, standard deviation, minimum, first quartile, third quartile, and minimum and maximum values. Note that ESG scores for environment, social, and governance scores, as well as their sum (ESG) were scaled down to 1:100 for calculation convenience. Note: Original environment, social, and governance scores and their sum, ESG, were divided by 100 for convenience. Additionally, the DivDummy variable was not included because it did not convey anything meaningful in our descriptive statistics, as it only takes a value of 1 or 0.
As shown in Table 4, the average ESG scores of the Korean firms in our sample were 0.96 for environmental, 0.86 for social, 0.89 and governance, respectively, for 2011-2017. These three scores are equally important in shaping the total ESG score of 2.71. Specifically, Korean firms had the highest score in environmental performance at an average of 0.96, whereas the standard deviation of the environmental score was also the highest at 0.66, which implies a substantial variation in environmental practices. The social score had the lowest average value of 0.86, and its standard deviation was 0.50. Finally, the corporate governance score had a mean value of 0.89, and its standard deviation was the lowest at 0.30. Table 5 presents the correlation coefficients of the variables in the analysis. The table shows that the three ESG scores were not highly correlated. One exception was the correlation between the environmental and social scores, which was approximately 0.67. This significant correlation indicates that a firm with good environmental practices is likely to be a firm with good social performance.
The table also shows that all other variables of interest had small correlation coefficients. The correlation coefficient between the social score and the natural logarithm of total assets was slightly high at 0.62, which indicates good social performance practice within large firms. However, the absolute value of the correlation coefficient was still smaller than 0.7, and thus, is not likely to cause multicollinearity problems in the estimations. The absolute values of all the other correlation coefficients were also smaller than 0.7.  Table 6 reports the estimation results for the entire sample of Korean firms including both the chaebol and non-chaebol affiliates. To test the effect of CSR performance on tax avoidance in all Korean firms, we applied our first and second empirical specifications to the estimation. Specifically, the first model incorporated the sum of ESG score as the independent variable. The third to fifth model included the environmental, social, and governance scores exclusively as the main independent variable. The second empirical model included all three ESG score components into the regression models. The estimated coefficients and corresponding p-values (in parenthesis) are also reported. Table 6. Results for the entire sample of corporations.  The numbers refer to the estimated coefficients, β. The numbers in parentheses refer to the corresponding t-values. ***, **, and * denote the 1%, 5%, and 10% significance levels, respectively.

ESG_SUM
The first column of Table 6 shows that there was a negative relationship between the sum of ESG scores and BTD in the Korean financial market. The estimated coefficient was −0.81, which was statistically significant at the 99% level in the first empirical specification. This significantly negative coefficient indicates a smaller tendency to avoid tax for firms with higher ESG scores, which supports our first empirical hypothesis (H1) built on corporate culture theory. The R 2 was 0.37, which is low in the absolute term. However, this value is consistent with the extant studies. For instance, Huseynov and Klamm [15] conducted a similar analysis and their R 2 value ranged from 0.044 to 0.157. The range of Hoi et al. [16] was from 0.119 to 0.643 depending on the sample selection, which is also in line with our estimates.
The second to fifth empirical specifications in Table 5 indicate the importance of the social score in deriving the results of the first column. As indicated in the second and fourth columns, the social score had significantly negative effects on the tendency to avoid tax, as measured by BTD. It is also noteworthy that the estimated effect was higher, at −3.54, in the second model controlling for both the environmental and governance scores than the coefficient (−3.09) in the fourth specification.
These robust results contrast with the results of empirical models that incorporate environmental and governance scores. Without controlling for the social and governance scores in the third column, the individual component of the environmental score had a significantly negative effect on the tendency to manipulate tax. However, the coefficient became positive at 0.09 after controlling for the environmental and governance scores in the second column. This finding suggests that the environmental score has limited power in explaining the degree of tax avoidance consistent to its purpose to measure the environmental performance of a corporation. The significant correlation between the environmental and social scores may drive the negative relation. The environmental score itself does not seem to provide additional dimension of information related to tax avoidance if the social score is controlled, as shown in the second column.
The governance score was positively related to the tendency of tax avoidance. This finding seems to be closely associated with the value enhancing aspect of tax avoidance with limited managerial diversions. A greater governance structure indicates a lower level of managerial diversion. A manger in this type of good governance firm is prone to manipulate tax because tax manipulation could enhance the financial performance and the value of a corporation. Significantly positive coefficients on the governance structure are in line with this value-enhancing aspect of tax manipulations.
This finding is consistent with our third hypothesis (H3) predicting the more significant explanatory power of social performance in the determination of tax avoidance propensity. As hypothesized, the decision to avoid tax appears more closely related to the social reputation of a corporation. The R 2 from the second to fifth empirical specifications ranged from 0.37 to 0.38. Table 7 presents the relationship between CSR performance and tax avoidance for chaebols. The results for the individual ESG scores were similar to those of all firms; however, the strength of the ESG sum score and all individual ESG scores, namely environmental, social, and governance scores, weakened, implying that there is less difference between net income and taxable income estimates. The most striking feature of this analysis is that the relationship between the total ESG score and BTD was negative at −0.02, but statistically insignificant. This result suggests that chaebol-affiliated firms with high ESG performance would have an ignorable tendency to manipulate their taxable profits. As we argued, this might be because chaebol affiliates already have a higher standard with regard to financial statement management [18], including the decision to avoid tax. These firms are under strict restrictions under the financing and governance policies of the KFTC and receive intensive media attention, which provides them with little room to manage taxable income. For the analysis on chaebols, the R 2 turned out to be around 0.33, which is low in the absolute term, but is in the range of R 2 value obtained in prior studies, as illustrated above.
In the fourth column, the coefficient of the social score was smaller than the corresponding value of the entire sample analysis but was still significantly negative. The coefficient on the social score, however, turned out to be positive in the second model, controlling for the environmental and governance scores. Such mixed results partly support our third empirical hypothesis (H3), emphasizing the more significant role of social reputation in deciding the tendency to avoid tax. In chaebol affiliates with severe monitoring from a wide range of agencies from government to media, the social reputation may not be an important factor in the determination of tax avoidance.
It is also interesting that the governance structure itself becomes a less significant factor in deciding the tendency of tax manipulation. Such a lower tendency for the chaebol affiliates with good governance firms is in line with the explanation based on the managerial diversion as argued above. The positive association between a better governance structure and the tendency of tax avoidance could be more significant when the managerial diversion is well controlled by the monitoring procedure. Chaebol affiliates are widely known to experience more severe managerial diversion because of their ownership structure concentrated on specific families. Such innated managerial diversion problems may weaken the linkage between good governance structure and the likelihood of tax avoidance. Table 8 examines the relationship between ESG scores and tax avoidance within non-chaebol affiliates. The table indicates that the social score had a significantly negative relationship with the BTD variable, even after controlling for environmental and social scores. The coefficients of the social score were quantitatively much larger in the sample of non-chaebol affiliates than the corresponding values in the chaebol affiliate sample. For instance, the coefficient was −2.33 in the fourth specification of Table 8, which is approximately 40 times larger than −0.06 in the same empirical model in Table 7. This finding supports both the second and third hypotheses. As predicted in H2, the CSR performance of non-chaebol affiliates was more negatively related to the tendency to avoid tax than for chaebol affiliates. Consistent with H3, the social score had the strongest explanatory power in the decision to avoid tax. For the analysis on non-chaebols, the R 2 values were from 0.41 to 0.42, which is also in the range of R 2 value among the previous studies.

Discussions
The above empirical analysis generally supports the three main empirical hypotheses of this work: (1) a negative relationship between ESG scores and tax avoidance, (2) a more significant negative relationship between ESG scores and tax avoidance in the non-chaebol affiliate sample, and (3) the importance of social score in deciding the tendency of tax avoidance. Table 6 supports the first and third hypotheses. Tables 7 and 8 provide evidence for the second hypothesis.
This negative relationship between the ESG score and the tendency of tax avoidance is in line with the extant literature, such as Park et al. [28] and Jones et al. [24]. In other words, our results provide additional evidence to the culture theory of corporations [10,11] and are generally consistent to the majority of existing studies.
The emphasis of our analysis on the heterogeneity across chaebol and non-chaebol affiliates is in line with the recent development of analysis in the Korean financial market studied, for instance, by Yoon et al. [20] and Yoon et al. [21]. Due to the significant controlling power of a specific family and its large size, chaebol affiliates are under stricter monitoring from the government and receive intensive media attentions for a variety of corporate policies. Manipulating tax liabilities is difficult and quite value destructive if it is known to public; therefore, the ESG performance may not be strongly connected with the likelihood of tax avoidance.
The emphasis on the social score in the determination of tax avoidance is closely as-sociated with the significance of social reputation in shaping corporate policies. If the behavior of tax avoidance is revealed to the public, the firm's social reputation could be seriously damaged by the loss of firm management personnel, political pressure, potential fines, and consumer boycott. The environmental performance is, in nature, not closely associated with the determination of policy. The effect of governance structure on tax avoidance depends on the degree of managerial diversion problems. This aspect of social reputation is not widely examined, particularly for the Korean financial market. This is because the evaluation of ESG performance is only conducted based on the letter grade system, which does not provide detailed information about the social performance exclusively. Tables 6-8 highlight this exclusive and significant role of social reputation in the decision of corporate policy.

Conclusions
This study developed three empirical hypotheses and tested them using a sample of Korean corporations. We hypothesized (1) a negative relationship between ESG scores and tax avoidance, (2) a more strongly negative relationship between ESG scores and tax avoidance in the non-chaebol affiliate sample, and (3) a more significant relationship between the social score and the tendency to avoid tax.
Our empirical analysis confirmed all three hypotheses using BTD as the proxy variable for the tendency to avoid tax. Specifically, the sum of ESG scores was shown to have a significantly negative relationship with BTD in the entire sample analysis. The coefficients on the ESG variables turned out to be significantly more negative in the non-chaebol affiliate sample than in the chaebol affiliate sample, indicating that the role of CSR practice in the reduction of tax avoidance is stronger within the non-chaebol affiliates. The explanatory power of the social score in the reduction of the tax avoidance tendency was the most substantial, regardless of our sample choices.
The main contributions of our study are summarized as follows. We found new empirical evidence supporting the culture theory of corporations in terms of CSR performance and tax avoidance, whereas another branch of theory expects a positive association between ESG scores and the tendency to avoid tax. Furthermore, our analysis highlighted that firm characteristics play critical roles in shaping the relationship between CSR performance and corporate policies. In line with the findings of Yoon et al. [20] and Yoon et al. [21], these relationships differ significantly across chaebol and non-chaebol affiliates. Finally, we identified the source of the negative association between the ESG scores and the tendency to avoid tax; social reputation was shown to be the most important factor in shaping the negative relationship.
In this study, we addressed the scarcity of research that expects the same effect in South Korea's developing market by investigating the association between CSR and BTD for firms listed on the Korean financial market. However, although the relatively positive relationship between governance factors and BTD was verified in our study, it has not been fully examined. This positive relationship implies that it would be more effort for Korean firms to restructure their governance structure while restraining their inclination to adjust taxable profits. These issues need to be addressed in future studies.