Asia-Pacific Journal of Accounting & Economics

ISSN: 1608-1625 (Print) 2164-2257 (Online) Journal homepage: http://www.tandfonline.com/loi/raae20

Moderated mediation effects of corporate social

responsibility performance on tax avoidance:
evidence from China

Chih-Wen Mao & Wen-Chieh Wu

To cite this article: Chih-Wen Mao & Wen-Chieh Wu (2018): Moderated mediation effects of
corporate social responsibility performance on tax avoidance: evidence from China, Asia-Pacific
Journal of Accounting & Economics

To link to this article: https://doi.org/10.1080/16081625.2019.1546157

Published online: 16 Nov 2018.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=raae20
ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS
https://doi.org/10.1080/16081625.2019.1546157

Moderated mediation effects of corporate social responsibility

performance on tax avoidance: evidence from China
Chih-Wen Maoa and Wen-Chieh Wub
a
Department of Public Finance and Tax Administration, National Taipei University of Business, Taipei City, Taiwan
(R.O.C.); bDepartment of Public Finance, National Chengchi University, Taipei City, Taiwan (R.O.C.)

ABSTRACT KEYWORDS
Using data of publicly listed Chinese companies practicing corporate Corporate social
social responsibility (CSR) activities between 2009 and 2016, this study responsibility; tax avoidance;
empirically examines whether CSR performance has a conditional indir- profitability; moderated
mediation effect. JEL: M40,
ect effect (or moderated mediation effect) on the level of corporate tax
M48, M14, H26
avoidance. We find that corporate profitability serves as a full mediator in
the association between CSR performance and corporate tax avoidance.
CSR performance reduces corporate profitability, and corporate profit-
ability in turn increases the level of corporate tax avoidance. In other
words, CSR performance first reduces corporate profitability, and there-
fore results in lower corporate tax avoidance. Moreover, CSR perfor-
mance does not significantly moderate the effect of corporate
profitability on corporate tax avoidance. Therefore, our results suggest
that CSR performance has an indirect, but not a conditional, effect on the
level of corporate tax avoidance.

1. Introduction
Tax avoidance activities are often witnessed in the real world. Therefore, research on tax avoidance
frequently attracts attention from scholars in a variety of academic fields. Hanlon and Heitzman
(2010) conduct a comprehensive review of both the theoretical and empirical literature on tax
avoidance. The researchers broadly define tax avoidance as legally or illegally reducing explicit
taxes. In addition, they claim that tax avoidance represents a continuum of tax-planning strategies.
The legal reduction of explicit tax is at one end of the continuum, whereas tax noncompliance, tax
evasion, tax aggressiveness, and tax sheltering are situated closer to the other end.
A variety of determinants of the decision to engage in tax avoidance engagement are discussed
in the literature. Public finance economists (e.g., Slemrod and Yitzhaki 2002) suggest that tax
rates, the probability of detection and punishment, penalties, and risk aversion are fundamental
factors influencing the decision to engage in tax avoidance or noncompliance. One branch of the
literature (e.g., Gupta and Newberry 1997; Wilson 2009; Kubick et al. 2015) adds a few firm-level
characteristics such as the firm size, leverage, and profitability as crucial determinants of the
decision to engage in corporate tax avoidance. Since the early 2010s, a group of researchers have
begun to hypothesize that corporate social responsibility (CSR) activities are associated with the
decision to engage in corporate tax avoidance (e.g., Huseynov and Klamm 2012; Hoi, Wu, and
Zhang 2013; Lanis and Richardson 2012, 2015). According to corporate culture theory proposed
by Kreps (1990), the causal effect of CSR on the level of corporate tax avoidance is argued to be

CONTACT Wen-Chieh Wu [email protected] Department of Public Finance, National Chengchi University, No. 64,
Sec. 2, ZhiNan Rd., Wenshan Dist, Taipei City 116, Taiwan (R.O.C.)
© 2018 City University of Hong Kong and National Taiwan University
2 C.-W. MAO AND W.-C. WU

negative. Engaging in CSR activities is normally considered an aspect of corporate culture

involving a shared belief within a corporation about a correct course of action. However,
aggressive tax avoidance activity is commonly regarded as irresponsible (e.g., Landolf 2006;
Hasseldine and Morris 2013). Therefore, CSR activities and aggressive tax avoidance activities
are in conflict with each other, suggesting that the level of corporate tax avoidance is negatively
associated with the level of CSR activity performance.
The negative association between CSR and corporate tax avoidance may not be always true if the
strategic risk management effect of CSR activities is considered. According to strategic risk manage-
ment theory proposed in the literature (e.g., Godfrey 2005; Godfrey, Merrill, and Hansen 2009), we
can argue that a corporation can enhance its positive reputation by increasing the frequency of
engaging in responsible CSR activities. Such an enhanced reputation can provide the corporation
with a certain degree of insurance protection against a variety of sanctions associated with engaging
in aggressive tax avoidance. Consequently, it is likely that a corporation with strong CSR activity
performance may expect negligible sanction costs associated with aggressive tax avoidance practices.
Therefore, the level of corporate tax avoidance can be positively or at least non-negatively associated
with CSR activity performance.
Because of the coexistence of corporate culture and strategic risk management effects, the
causal effect of CSR activity performance on the level of corporate tax avoidance is ambiguous.
The causal effect is expected to be negative if the corporate culture effect dominates the strategic
risk management effect, and vice versa. Empirical evidence from recent studies employing a
variety of data mainly shows that the level of corporate tax avoidance is negatively associated
with responsible CSR activities. However, such empirical evidence is mainly from developed
countries in the West. Therefore, finding empirical evidence from developing economies, parti-
cularly those in the East, is essential. In the present study, we choose China, the largest emerging
economy in the East, as the study sample because China has witnessed significant growth in the
frequency of corporate tax avoidance activities and CSR activities in recent years. Using a sample
of publicly listed companies in China, our estimation results show that CSR activity performance
has a negative and significant effect on the level of corporate tax avoidance.
In contrast to previous empirical studies that focus only on the direct effect of CSR activity
performance on the corporate tax avoidance level, this study investigates whether indirect effects
also occur. We propose that CSR activity performance indirectly affects the level of corporate tax
avoidance because CSR performance initially affects corporate profitability (mediator), and cor-
porate profitability (mediator) in turn affects the level of corporate tax avoidance. The literature
provides contrasting evidence to verify the effects of CSR activities on corporate profitability. One
branch of the literature claims that CSR activities can provide positive feedback to corporate
business opportunities, thereby enhancing corporate profitability (e.g., Tsoutsoura 2004; Scholtens
2008; Callan and Thomas 2009; Aras and Crowther 2009; Schadewitz and Niskala 2010; Garcia-
Castro, Ariño, and Canela 2010; Gras-Gil, Manzano, and Fernández 2016). However, another
group of studies argue that CSR activities do not enhance corporate profitability because they
increase operating costs, lower the market competitiveness of products, and make corporations to
exert less effort in profit-seeking activities (e.g., Cardebat and Sirven 2010; Shen and Chang 2008).
Based on these two opposite effects, we hypothesize that CSR performance negatively influences
corporate profitability if a variety of negative effects dominate the positive feedback effect, and
vice versa. Furthermore, prior research (e.g., Frank, Lynch, and Rego 2009; Wilson 2009; Kubick
et al. 2015) finds that the effect of corporate profitability (mediator) on the level of corporate tax
avoidance is positive; therefore, we hypothesize that corporate profitability positively affects the
level of corporate tax avoidance.
In addition to the indirect effect (mediating effect) of CSR activities on the decision to engage
in corporate tax avoidance, this study assumes the existence of a conditional indirect effect (or
moderated mediation effect) of CSR activities on the level of corporate tax avoidance (Figure 1).
We assume that the strength of the mediating effect (or indirect effect) may be conditional on the
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 3

CSR Performance Profitability Tax Avoidance

Figure 1. Conceptual moderated mediation model.

value of CSR activity performance (moderator). To explain this assumption, we argue that a
corporation with a higher CSR performance score (value) is likely to have a strong corporate
culture. Therefore, as corporate profitability increases, the level of tax avoidance may decrease, or
at least would not increase significantly. However, we argue that a corporation with high CSR
performance has obtained a high degree of insurance protection against the risk of sanctions for
practicing tax avoidance. Consequently, a corporation with high CSR performance may raise its
level of tax avoidance as corporate profitability increases. Therefore, we hypothesize that the
moderating effect of CSR performance on the mediating effect may be nonsignificant if the
corporate culture and risk management effects offset each other.
To conduct tests on our hypotheses, we employ the following models. First, the simple
mediation model is used to test whether corporate profitability is a mediator in the association
between corporate tax avoidance and CSR performance. Second, the moderated mediation effect
model is used to verify whether CSR performance has a conditional indirect effect on the level of
corporate tax avoidance. Our estimation results show that corporate profitability fully mediates
the association between CSR performance and corporate tax avoidance. In other words, the reason
for the decline in corporate tax avoidance is that CSR performance reduces corporate profitability.
Moreover, our results reveal that the moderating effect of CSR performance on the mediating
effect is statistically nonsignificant. According to these results, we conclude that CSR performance
has an indirect effect, but not a conditional indirect effect, on the level of corporate tax avoidance.
This study contributes to the empirical literature by being the first to determine the moderated
mediation effect of CSR on tax avoidance; therefore, we employed the moderated mediation effect
model, which is a structural equation modeling (SEM) model comprising two simultaneous
equations, namely a profitability equation and tax avoidance equation. In other words, two
regressions are simultaneously estimated in this model. The remainder of this paper is organized
as follows. Section 2 reviews related studies, and Section 3 discusses econometric strategies.
Section 4 describes the variable selection process. Data and samples are described in Section 5,
and the estimation results are presented in Section 6. Section 7 checks the robustness. Concluding
remarks are put in the last section.

2. Literature review
In contrast to other firm-level determinants of the decision to engage in corporate tax avoidance
that are addressed in previous studies, CSR activities were not considered as such a determinant
until the early 2010s. Huseynov and Klamm (2012) and Lanis and Richardson (2012) are
considered the first to investigate the effect of CSR on corporate tax avoidance. To test the effects
of a firm’s negative and positive social actions on tax avoidance, Huseynov and Klamm (2012)
investigate American firms using auditor-provided tax services. The researchers use the following
three measures of CSR obtained from the Kinder, Lydenberg, and Domini (KLD) database:
corporate governance, community, and diversity. Their estimation results reveal that the CSR
level affects tax avoidance. Instead of employing American firms, Lanis and Richardson (2012) use
a small sample of Australian firms during the 2008 and 2009 fiscal years to estimate the
association between CSR performance and the likelihood of tax avoidance. They rely on company
4 C.-W. MAO AND W.-C. WU

self-reported CSR disclosure data. They find a negative association between CSR disclosure levels
and effective tax rates (ETRs). Based on these findings, the researchers assert that CSR activities
are related to decisions to engage in tax avoidance. However, they also admit that effective tax
rates do not accurately reflect aggressive avoidance practices and that CSR disclosure does not
necessarily reflect CSR activities.
Lanis and Richardson (2015) extend their study of Australian firms conducted in 2012 to
further investigate the effect of CSR performance on the likelihood of corporate tax avoidance,
this time among American firms. Instead of using the proxy measures of tax avoidance used in
previous studies, they use a direct measure of tax avoidance based on tax disputes. Employing a
matched sample of 434 American companies during 2003–2009, their logit estimation results
reveal that the higher the level of CSR performance, the lower the likelihood of tax avoidance. Hoi,
Wu, and Zhang (2013) further examine the association between irresponsible CSR activities and
aggressive tax avoidance.1 The researchers employ a large sample of American firms during the
period 2003–2009. Their results show that only irresponsible CSR activities are significantly
positively associated with tax avoidance, whereas responsible CSR activities are not significantly
related to tax avoidance. Watson (2015) extends the empirical model of Hoi, Wu, and Zhang
(2013) by adding an interaction term, namely CSR performance multiplied by the firm profit
dummy variable, as a covariate. The author uses the firm profit dummy variable to divide the
sample into high-profit and low-profit firms.2 Evidence indicates that in low-profit firms, CSR is
positively associated with tax avoidance; however, this effect is diminished in high-profit firms.
In summary, previous studies (e.g., Lanis and Richardson 2015; Hoi, Wu, and Zhang 2013;
Watson 2015) generally assume that CSR directly affects tax avoidance, and thus traditional OLS
is commonly used to analyze this aspect. Some of these studies (e.g., Watson 2015) treat profit-
ability as a covariate. However, we argue that profitability may serve as a mediator. In other
words, CSR may indirectly affect tax avoidance through its effect on profitability. SEM is thus a
more appropriate method to examine the mediation effect.

3. Empirical strategies
3.1. Traditional linear regression specification
The traditional linear regression specification for corporate tax avoidance in the tax literature
considering CSR activities is expressed as follows:
btdit ¼ α1 þ α2 Xit þ α3 CSRit þ εit (1)
where a subscript it denotes the firm i in year t. X is a vector of firm-level variables influencing the
level of corporate tax avoidance (btd). The CSR activity variable CSR is the independent variable
of central interest. The coefficient of central interest α3 measures the strength of the direct effect of
CSR performance on the level of corporate tax avoidance.

3.2. Simple mediation effect model specification

We first hypothesize that corporate profitability serves as the only mediator in the association
between CSR performance and corporate tax avoidance. Considering only the mediation effect of
profitability, we then employed a SEM model to express the simple mediation effect model. This
simple mediation effect model comprises of both profitability equation and tax avoidance equa-
tion. The simple mediation effect model is expressed as the following SEM model:
Profit ¼ β1 þ β2 CSRit þ β3 Zit þ ϕit ; (2)

btdit ¼ α1 þ α2 Xit þ α03 CSRit þ γ1 Profit þ ωit : (3)

 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 5

Equation (2) is the specification for the mediator, corporate profitability. Corporate profitability
(Prof ) is regressed on CSR performance as well as a vector of confounding variables (Z) influen-
cing corporate profitability. Equation (3) is the extended specification for the level of corporate tax
avoidance, and it simultaneously includes CSR performance and corporate profitability Prof (the
mediator) as the regressors. The estimated direct effect of CSR performance on corporate tax
avoidance is thus α ^0 3 , and the estimated indirect effect (or mediating effect) of CSR performance
on the level of corporate tax avoidance is ^β2 ^γ1 . Equations (2) and (3) are simultaneously
estimated.
Corporate profitability can be confirmed as the mediator if the estimated coefficients α ^,
^3 , β2
and ^γ1 in Equations (1–3), respectively, are statistically significant (Baron and Kenny 1986).
Moreover, Baron and Kenny recommend that the statistical significance of the estimated mediat-
ing effect (or estimated indirect effect) of CSR on the level of corporate tax avoidance (β ^ ^γ ) can
2 1
be assessed by the Sobel z-test (Sobel 1982), which is expressed as follows:
^  ^γ
β
z ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1
ffi (4)
^γ1 s2^ þ β ^ s2
β 2 ^
γ1
2

^ and ^γ , respectively. However, because the

where s2β^ and s^2γ1 are the squared standard errors of β2 1
2
estimated mediating effect (or estimated indirect effect) is the product of two parameters, the
sampling distribution of the Sobel z-test is not normal. The bootstrap test, a resampling strategy
for estimation and hypothesis testing discussed in Preacher, Rucker, and Hayes (2007), can be
used to solve this problem. The sampling distribution of such an indirect effect is estimated
through bootstrapping by resampling for n times. For each resample, β ^þ ^γþ is computed.
2 1
Subsequently, we obtain the sampling distribution of such an indirect effect by sorting the n
bootstrap values of β^þ ^γþ in ascending order. The lower bound of the 100ð1  αÞ% confidence
2 2
^ ^γ is defined as the ðα=2Þnth value in the sorted distribution, and the upper bound is
level for β2 2
defined as the ½1 þ ð1  ðα=2ÞÞnth value. If the confidence interval does not contain a zero, then
CSR performance is considered to have a significant indirect effect (mediating effect) on the level
of corporate tax avoidance.

3.3. Moderated mediation effect model specification

In addition to serving as the independent variable in the corporate tax avoidance equation, CSR
performance is hypothesized to function as a moderator of the mediating effect. We extend
Equation (3) from the simple mediation effect model to the following equation representing the
moderated mediation effect model:
btdit ¼ α1 þ α2 Xit þ α03 CSRit þ ðγ2 þ γ3 CSRit Þ Profit þ νit : (5)
Equation (5) differs from Equation (3) by including the cross term CSR  Prof as the regressor.
The moderated mediation effect (or conditional indirect effect) of CSR on btd is quantified as
^ ð^γ þ ^γ CSRÞ, which is conditional on the value of the moderator CSR. The SEM model for
β2 2 3
estimating the moderated mediation effects is expressed as follows:
Profit ¼ β1 þ β2 CSRit þ β3 Zit þ ϕit ;
btdit ¼ α1 þ α2 Xit þ α0 3 CSRit þ ðγ2 þ γ3 CSRit ÞProfit þ νit
If the effect of the interaction between CSR performance and corporate profitability is close to
0, then ^γ3 should also be close to 0. Therefore, CSR performance has little influence on the
mediating effect (or indirect effect). In this case, the conditional indirect effect of CSR
6 C.-W. MAO AND W.-C. WU

performance on corporate tax avoidance decreases to β ^ ^γ for all moderator (CSR performance)
2 2
values. In other words, the effect of CSR performance on the level of corporate tax avoidance is
indirect, but not conditionally indirect. By contrast, if ^γ3 is not close to 0, CSR performance
influences the mediating effect (or indirect effect). We can further use the bootstrap approach to
test the statistical significance of the conditional indirect effect (or moderated mediation effect) of
CSR performance on corporate tax avoidance.

4. Variable selection
This study contains three main variables. The level of corporate tax avoidance is the dependent
variable, and CSR performance is the independent variable of central interest as well as the
moderator of the mediating effect. Corporate profitability is the mediator in the association
between corporate tax avoidance and CSR performance. We use various measures to capture
the level of corporate tax avoidance because each measure has its own limitations (Hanlon and
Heitzman 2010; Lisowsky, Robinson, and Schmidt 2013). Two adjusted book-tax difference
measures are used to capture tax avoidance practices. These measures are the Khurana and
Moser (2013) permanent book-tax difference (pbtd), and the Desai and Dharmapala (2006)
discretionary book-tax difference (ddbtd). In the robustness check, we also use the long-run
book effective tax rate (letr) to capture consequences of tax avoidance practices. The overall
score of CSR evaluated by a neutral organization serves as the proxy variable of CSR performance.
In addition, we employ the ROA as the proxy variable for corporate profitability. We present
detailed definitions of all variables in Appendix.
The equation for the level of corporate tax avoidance includes control variables such as firm
size, leverage, fixed asset intensity (property, plant, and equipment), inventory density, a dummy
variable indicating loss carried forward, and firm growth potential. Firm size (size) measured by
the natural logarithm of total assets is expected to influence the level of corporate tax avoidance
because a larger firm has a higher incentive and more power to engage in tax avoidance activities
(Wilson 2009; Lanis and Richardson 2012). Because a corporation with a high debt ratio is under
pressure to engage in aggressive tax avoidance to pay off its matured debt, we adopt firm leverage
(lev), measured as a long-term debt divided by total assets as covariance (Hoi, Wu, and Zhang
2013；Kubick et al. 2015). Moreover, due to tax-deductible interest payments, leverage is asso-
ciated with corporate tax avoidance (Gupta and Newberry 1997; Lanis and Richardson 2015). We
expect that the fixed asset intensity variable (ppe), measured as fixed assets (property, plant,
equipment) scaled by total assets, can influence the level of corporate tax avoidance because the
accelerated depreciation charges related to fixed assets cause tax shield effects, thus reducing the
tax burden on the firm (Lanis and Richardson 2012; Kubick et al. 2015; Hoi, Wu, and Zhang 2013;
Gupta and Newberry 1997; Lanis and Richardson 2015). The inventory density variable (inv),
measured as the inventory scaled by total assets, is expected to have an influence on the level of
tax avoidance, suggesting that inventory-intensive firms are less able to engage in tax avoidance
than are capital-intensive firms (Lanis and Richardson 2012, 2015). A firm’s annual loss can be
carried forward to be tax deductible in the following few years. The dummy variable indicating
loss carried forward (nol) is expected to have an influence on the level of tax avoidance (Kubick et
al. 2015; Watson 2015). Furthermore, sales revenue changes (dsale) are expected to influence the
level of tax avoidance because the growth in future profitability increases the future tax burden
and consequently increases the incentive for tax avoidance (Hoi, Wu, and Zhang 2013; Watson
2015). Absolute value of performance-adjusted abnormal accruals (absda) are related to tax
avoidance (Frank, Lynch, and Rego 2009). Therefore, we use absda as a control variable in the
regression models. Last, we include cash holdings (cash) to control for liquidity and other firm
attributes that could potentially affect our tax avoidance measures, such as intangible assets
(intan) and equity income in earnings (eqinc) (Hoi, Wu, and Zhang 2013).
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 7

In the equation for corporate profitability, we control two confounding variables: size and lev.
The sign of the estimated effect of CSR performance on corporate profitability is expected to be
ambiguous depending on whether the positive business feedback effect or negative operation cost
effect dominates. Firm size is expected to have a positive effect on corporate profitability (Shen
and Chang 2008). Moreover, a corporation can use long-term debt to finance investment and
adjust fund flexibility. Therefore, debt ratio is considered to have a positive effect on corporate
profitability (Tsoutsoura 2004).

5. Data source and sample

Our CSR performance data source is the official report of enterprise CSR activities released by the
CSR research center at the Chinese Academy of Social Sciences (CASS-CSR). The research center
designs overall evaluation systems of an enterprise’s CSR activities. Such a development index
refers to a group of CSR indices such as International Organization for Standardization 26,000,
Global Reporting Initiative, Dow Jones Sustainability Index, Wealth Magazine CSR index, and
Financial Times Stock Exchange 4. In addition, the design of such a development index considers
China’s current CSR situation.
The procedure for evaluating CSR performance scores is described as follows. (1) A company’s
four CSR items (responsibility management, market responsibility, social responsibility, and
environmental responsibility) are scored. (2) The scores of the four items are weighted according
to the appropriate weights3 and then summed to obtain the original score.4 (3) The original score
is added to the score of the adjusted items5 to obtain the CSR performance score.6 The CSR
performance score produces a range from 0 to 100.
Data of corporate profitability, tax avoidance level, firm size, debt ratio, fixed assets, inventory,
loss deductible, changes in sales revenue, cash holdings, intangible assets, and equity income in
earnings are obtained by the Taiwan Economic Journal. The sample enterprises in this study are
publicly listed companies in the A-Shares market that engaged in CSR activities during the sample
period of 2009–2016. Publicly listed companies that did not engage in any such activity during
this period are excluded because some corporations that did engage in such activities receive a
score of 0 for performance.7 To distinguish these two types of corporation from each other, we
exclude the nonpracticing type. The total sample consists of 94 firms. Moreover, all continuous
independent variables are winsorized at the 1st and 99th percentiles.
As shown in Table 1, the average pbtd and ddbtd are 1.75% and 1.18%, respectively. Moreover,
the average CSR score is 33.48, the highest is 91, and the lowest is 0. The average ROA is 5.48%.

Table 1. Descriptive statistics.

N Mean Std. Dev. Min. Max.

pbtd 495 1.75 2.69 −3.46 12.07
ddbtd 271 1.18 2.08 −2.73 9.14
CSR 495 33.48 25.95 0.00 91.00
ROA 491 5.48 4.07 1.00 24.00
size 495 17.53 1.79 13.47 21.50
lev 495 10.41 10.80 0.00 51.72
inv 495 19.02 15.84 0.60 71.87
ppe 495 31.17 23.73 0.58 85.50
nol 495 0.12 0.32 0.00 1.00
dsale 491 19.57 50.83 −33.95 428.34
cash 495 16.25 11.13 0.87 52.43
intan 495 3.93 3.28 0.06 17.56
absda 233 0.14 0.37 0.00 2.90
eqinc 495 0.45 1.17 −0.35 6.95
All continuous independent variables are winsorized at the 1st and 99th percentiles.
8 C.-W. MAO AND W.-C. WU

Table 2 presents the correlation matrix of the explanatory variables. As shown, the highest
correlation coefficient between CSR and size is 0.63. The remaining correlation coefficients are
around −0.58 to 0.01, all of which are acceptable when it comes to avoiding the problem of multi-
collinearity.

6. Estimation results
6.1. Traditional regression results
The results of traditional tax avoidance models are summarized in Table 3. We use pbtd and ddbtd
as the proxy measures of tax avoidance, respectively. No matter we use pbtd or ddbtd as the
dependent variables, the CSR performance score has a significant negative effect on the level of
corporate tax avoidance. This result is consistent with the findings in most literature and suggests
that the corporate culture effect of CSR activities still dominates the decisions of listed Chinese
corporations to engage in tax avoidance. Moreover, the level of corporate tax avoidance is
significantly and negatively associated with the inventory density, and the intangible assets, but
positively associated with equity income in earnings.

6.2. Moderated mediation effect

The first column of Table 4 shows that the CSR performance score has a significant negative effect
on the ROA, a proxy measure of corporate profitability. This finding suggests that a variety of
negative effects of CSR activities dominate the positive feedback effect of CSR activities on
corporate profitability, which is consistent with the findings in the literature (e.g., Aupperle,
Carroll, and Hatfield 1985; Ullmann 1985; Becchetti, Ciciretti, and Hasan 2007; Shen and Chang
2008; Cardebat and Sirven 2010). Moreover, corporate profitability is positively associated with
firm size, but negatively associated with leverage.
The second column of Table 4 shows the estimation results of the corporate tax avoidance
equation in the simple mediation effect model and pbtd is used as a proxy variable of tax
avoidance. The effect of the CSR performance score on the level of corporate tax avoidance is
statistically nonsignificant when the mediator corporate profitability variable (ROA) is simulta-
neously included as the regressor. Moreover, the ROA variable significantly increases the level of
corporate tax avoidance, thereby supporting the theoretical argument proposed in the literature
(e.g., Frank, Lynch, and Rego 2009; Wilson 2009; Kubick et al. 2015).

Table 2. Correlations.

pbtd ddbtd CSR ROA size lev inv ppe nol dsale cash intan absda eqinc
pbtd 1.00
ddbtd 0.93 1.00
CSR −0.17 −0.17 1.00
ROA 0.66 0.68 −0.20 1.00
size −0.17 −0.22 0.63 −0.14 1.00
lev −0.18 −0.24 0.31 −0.22 0.41 1.00
inv −0.21 −0.32 −0.06 −0.16 0.06 −0.03 1.00
ppe −0.08 −0.01 0.19 −0.05 0.16 0.39 −0.58 1.00
nol −0.09 0.00 0.05 −0.13 0.01 0.08 −0.13 0.20 1.00
dsale 0.09 0.09 −0.12 0.28 −0.05 −0.05 0.14 −0.11 0.15 1.00
cash 0.21 0.19 −0.27 0.24 −0.25 −0.36 −0.02 −0.49 −0.10 0.05 1.00
intan 0.03 0.00 0.01 0.00 0.08 0.03 −0.27 0.02 −0.02 −0.06 0.03 1.00
absda −0.08 −0.07 0.06 0.04 0.06 0.02 0.09 −0.08 0.07 −0.05 −0.09 0.02 1.00
eqinc 0.34 0.32 0.06 0.15 0.06 −0.03 −0.18 −0.15 0.00 0.05 0.11 0.01 −0.06 1.00
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 9

Table 3. Linear regression model: dependent variable: tax avoidance.

pbtd ddbtd
size 0.0368 0.0421
(0.100) (0.100)
lev −0.0177 −0.0181
(0.016) (0.016)
inv −0.0534*** −0.0532***
(0.012) (0.012)
ppe −0.0098 −0.0091
(0.010) (0.010)
nol −0.3532 −0.3428
(0.521) (0.521)
dsale 0.0062 0.0062
(0.004) (0.004)
cash 0.0172 0.0182
(0.017) (0.017)
intan −0.0772* −0.0762*
(0.045) (0.045)
absda −0.0142 −0.0124
(0.294) (0.294)
eqinc 0.3971*** 0.4004***
(0.118) (0.118)
CSR −0.0114* −0.0116*
(0.007) (0.007)
_cons 2.3819 2.2651
(1.721) (1.718)
N 233 233
adj. R2 0.243 0.245
F 7.757 7.861
p-value [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard
error is included in the parentheses.

Our findings match the following three conditions proposed in the literature (e.g. Baron and
Kenny 1986; Edwards and Lambert 2007; Hayes 2013).8 First, the independent variable (CSR
performance) has a significant effect on the dependent variable (level of tax avoidance) when the
mediator is excluded. Second, the independent variable (CSR performance) has a significant effect
on the mediator (corporate profitability). Third, the independent variable (CSR performance) has
a nonsignificant effect on the dependent variable (level of tax avoidance) when the mediator is
included. Therefore, corporate profitability (ROA) is a full mediator in the association between
CSR performance and the level of corporate tax avoidance. In summary, our results reveal that
CSR performance significantly reduces corporate profitability, and corporate profitability in turn
significantly increases the level of corporate tax avoidance.
The third column of Table 4 shows the results of the equation of corporate tax avoidance in the
moderated mediation effect model. The estimated coefficient of CSR  ROA is positive and
statistically nonsignificant. This finding implies that the moderating effect of CSR performance
on the mediating effect is positive, but nonsignificant.
We demonstrate the mediation and moderated mediation effect through the aforementioned
Sobel z-test and bootstrap test. The results are shown in Table 5. Through the Sobel z-test, the
estimated coefficient of the mediator is −0.01456 and is statistically significant at the 5% sig-
nificance level. Moreover, through the bootstrap method, the estimated coefficient is −0.00945 and
lies in the 95% bias-corrected confidence interval (−0.0167, −0.0034) obtained, suggesting
10 C.-W. MAO AND W.-C. WU

Table 4. Moderated mediation model: tax avoidance (pbtd).

(1) (2) (3)

ROA pbtd pbtd
Control Var.
size 0.0933 −0.1245* −0.1279*
(0.141) (0.069) (0.069)
lev −0.0672*** 0.0098 0.0105
(0.019) (0.011) (0.011)
inv −0.0254*** −0.0258***
(0.009) (0.009)
ppe −0.0067 −0.0070
(0.007) (0.007)
nol 0.1462 0.1648
(0.356) (0.359)
dsale −0.0138*** −0.0136***
(0.003) (0.003)
cash 0.0117 0.0117
(0.012) (0.012)
intan −0.0275 −0.0277
(0.031) (0.031)
absda −0.3855* −0.3870*
(0.202) (0.202)
eqinc 0.3403*** 0.3336***
(0.080) (0.082)
Explanatory Var.
CSR −0.0274*** 0.0014 −0.0006
(0.009) (0.005) (0.006)
Mediator
ROA 0.3963*** 0.3856***
(0.025) (0.034)
Moderator
CSR×ROA 0.0005
(0.001)
_cons 5.4847** 1.9484* 2.0680*
(2.257) (1.171) (1.200)
N 491 233 233
adj. R2 0.058 0.649 0.648
F 10.99 36.79 33.85
p-value [0.00] [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard error
is included in the parentheses.

statistical significance. Consequently, the results of both tests confirm that the mediating effect
exists. Regarding the moderated mediation effect, the estimated coefficient is −0.00943 when the
CSR score is low (i.e., mean minus standard deviation) and −0.00929 when the CSR score is high
(i.e. mean plus standard deviation). However, according to the results in Table 4, the coefficient of
CSR  ROA is statistically nonsignificant, suggesting that the moderating effect of CSR perfor-
mance is not significant.
Table 6 presents the empirical results of using ddbtd as a proxy variable of tax avoidance.
Likewise, the empirical result shows that ROA serves as a full mediator in the association between
CSR performance and the level of corporate tax avoidance. The estimated coefficient of the
interaction term between CSR and ROA is positive but statistically nonsignificant. Therefore,
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 11

Table 5. Test results of mediation effect and moderated mediation effect: tax avoidance (pbtd).

Test results of mediation effect

Value Std. Err. z LL 95% CI UL 95% CI
Indirect effect and significance using normal distribution
Sobel −0.01456 0.0055 −2.6407 −0.0254 −0.0038
Bootstrap results for indirect effect
Bootstrap −0.00945 0.0035 −2.6832 −0.0167 −0.0034
Test results of moderated mediation effect

Std. Err. z LL 95% CI UL 95% CI

−1 SD (7.5239) −0.00943 0.0034 −2.7331 −0.0160 −0.0023

M (33.4788) −0.00936 0.0033 −2.8500 −0.0176 −0.0041
+1 SD (59.4337) −0.00929 0.0033 −2.8519 −0.0181 −0.0047
Bootstrap sample size = 1,000. LL = lower limit; CI = confidence interval; UL = upper limit.

the CSR performance score does not moderate the mediating effect. This mediating effect is
verified by a Sobel z-test and bootstrap test in Table 7. The test results obtained from a Sobel z-test
and bootstrap method suggest that the CSR performance score has an indirect effect on corporate
tax avoidance.
A summary of the results from the moderated mediation model reveals that CSR performance
has an indirect effect but not a conditional indirect effect on the level of corporate tax avoidance.

7. Robustness checks
In this section, we perform a series of additional tests to ensure the robustness of the results. To
save space, we include only the table of the empirical results of Equations (2) and (3) in the
manuscript.

7.1. Adoption alternative tax avoidance measurement

To verify our empirical results, we use an alternative tax avoidance proxy in our robustness check.
Lower values of the effective tax rates suggest greater tax avoidance. Therefore, we use the 4-year
average book effective tax rate to capture consequences of tax avoidance practices. In Table 8, the
CSR performance score has a significant positive effect on the level of corporate tax burden. In
other words, CSR performance has a negative effect on tax avoidance. Using ddbtd as the proxy of
tax avoidance, we derive the results summarized in Table 8, showing that ROA still serves as a full
mediator in the association between CSR performance and the level of corporate tax burden. The
bootstrap test results suggest that the CSR performance score has an indirect effect on corporate
tax burden.

7.2 Considering the endogenous issue

The decision of CSR activities is taken as exogenous in the empirical model. However, the
decision of CSR activities is part of the set of policies established by a company. Thus, the
endogenous nature of CSR should be considered. Therefore, we use the lag of CSR performance
(CSRit1 ) as an instrument to circumvent the bias of endogeneity.
The empirical results in Table 9 demonstrate that the coefficient of ROA is significantly
positive, whereas that of CSR is nonsignificant. The bootstrap results reveal that the indirect effect
12 C.-W. MAO AND W.-C. WU

Table 6. Moderated mediation model: tax avoidance (ddbtd).

(1) (2) (3)

ROA ddbtd ddbtd
Control Var.
size 0.0933 −0.1186* −0.1219*
(0.141) (0.069) (0.069)
lev −0.0672*** 0.0092 0.0100
(0.019) (0.011) (0.011)
inv −0.0254*** −0.0257***
(0.009) (0.009)
ppe −0.0060 −0.0063
(0.007) (0.007)
nol 0.1545 0.1725
(0.357) (0.360)
dsale −0.0138*** −0.0136***
(0.003) (0.003)
cash 0.0127 0.0127
(0.012) (0.012)
intan −0.0267 −0.0268
(0.031) (0.031)
absda −0.3823* −0.3837*
(0.202) (0.202)
eqinc 0.3438*** 0.3374***
(0.081) (0.082)
Explanatory Var.
CSR −0.0274*** 0.0012 −0.0007
(0.009) (0.005) (0.006)
Mediator
ROA 0.3947*** 0.3842***
(0.025) (0.034)
Moderator
CSR×ROA 0.0004
(0.001)
_cons 5.4847** 1.8335 1.9495
(2.257) (1.173) (1.202)
N 491 233 233
adj. R2 0.058 0.648 0.647
F 10.99 36.65 33.73
p-value [0.00] [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard error
is included in the parentheses.

reaches the 10% level of significance. Consequently, after consideration of the endogenous issue,
the conclusion remains unchanged; ROA serves as a full mediator.

7.3 Adoption of long-term ROA as the proxy variable of corporate performance

Most related studies (e.g., Garcia-Castro, Ariño, and Canela 2010; Shen and Chang 2008) regard
the relationship between corporate performance and CSR as a short-term aspect, and thus these
studies employ current corporate performance and current CSR as variables for analysis.
However, this relationship may not be a short-term aspect. In other words, CSR may affect a
company’s long-term profitability. Consequently, we use the average return on assets of the
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 13

Table 7. Test results of mediation effect and moderated mediation effect: tax avoidance (ddbtd).

Test results of mediation effect

Value Std. Err. z LL 95% CI UL 95% CI
Indirect effect and significance using normal distribution
Sobel −0.01450 0.0055 −2.6404 −0.0253 −0.0037
Bootstrap results for indirect effect
Bootstrap −0.01412 0.0060 −2.3508 −0.0288 −0.0046
Test results of moderated mediation effect

Std. Err. z LL 95% CI UL 95% CI

−1 SD (7.5239) −0.01424 0.0059 −2.3990 −0.0287 −0.0048

M (33.4788) −0.01466 0.0058 −2.5199 −0.0281 −0.0042
+1 SD (59.4337) −0.01509 0.0059 −2.5575 −0.0276 −0.0049
Bootstrap sample size = 1,000. LL = lower limit; CI = confidence interval; UL = upper limit.

Table 8. Robustness checks: tax avoidance (letr).

(1) (2)
ROA letr
Control Var.
size 0.0933 0.2587
(0.141) (0.520)
lev −0.0672*** −0.0461
(0.019) (0.081)
inv 0.0692
(0.066)
ppe −0.0620
(0.052)
nol −1.8606
(2.693)
dsale −0.0275
(0.022)
cash −0.1849**
(0.087)
intan 0.1616
(0.231)
absda 1.6282
(1.523)
eqinc −1.5450**
(0.608)
Explanatory Var.
CSR −0.0274*** 0.0557
(0.009) (0.036)
Mediator
ROA −0.3107*
(0.187)
_cons 5.4847** 14.3930
(2.257) (8.853)
N 491 233
adj. R2 0.058 0.143
F 10.99 4.225
p-value [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard error
is included in the parentheses.
14 C.-W. MAO AND W.-C. WU

Table 9. Robustness checks: considering the endogenous issue.

(1) (2) (3) (4)

ROA pbtd ROA ddbtd
Control Var.
size 0.1936 −0.0918 0.1936 −0.1337**
(0.145) (0.100) (0.145) (0.067)
lev −0.0669*** 0.0153 −0.0669*** 0.0089
(0.020) (0.016) (0.020) (0.011)
inv −0.0269** −0.0249***
(0.013) (0.009)
ppe −0.0104 −0.0061
(0.010) (0.007)
nol −0.1163 0.1603
(0.526) (0.353)
dsale −0.0145*** −0.0139***
(0.004) (0.003)
cash 0.0326* 0.0129
(0.017) (0.011)
intan −0.0210 −0.0246
(0.046) (0.031)
absda −0.6358** −0.3856*
(0.300) (0.202)
eqinc 0.4202*** 0.3458***
(0.120) (0.080)
Explanatory Var.
CSR −0.0287*** 0.0052 −0.0287*** 0.0030
(0.010) (0.007) (0.010) (0.005)
Mediator
ROA 0.4366*** 0.3966***
(0.037) (0.025)
_cons 3.4914 1.2961 3.4914 2.0224*
(2.340) (1.732) (2.340) (1.163)
N 392 233 392 233
adj. R2 0.056 0.512 0.056 0.649
F 8.680 21.25 8.680 36.74
p-value [0.00] [0.00] [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard error
is included in the parentheses.

current and subsequent 3 years (LROA) as the proxy variable of corporate performance. As
indicated in Table 10, adopting another dependent variable does not qualitatively change the
previous results. In addition, the test results obtained from a bootstrap method suggest an indirect
effect of CSR performance on tax avoidance.

8. Concluding remarks
This paper first contributes to the literature in several manners. First, we employ data for publicly
listed companies in China, the largest emerging economy in the East. Similar to evidence obtained
regarding the West, in China, the level of corporate tax avoidance is negatively associated with
CSR performance. Second, we provide evidence of the indirect effect of CSR performance on the
level of corporate tax avoidance. We find that corporate profitability serves as a full mediator in
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 15

Table 10. Robustness checks: adoption of long-run ROA as the proxy variable of corporate
performance.

(1) (2) (3) (4)

LROA pbtd LROA ddbtd
Control Var.
size 0.1231 0.0183 0.1231 −0.0446
(0.102) (0.112) (0.102) (0.085)
lev −0.0392*** 0.0128 −0.0392*** 0.0038
(0.014) (0.018) (0.014) (0.013)
inv −0.0483*** −0.0449***
(0.014) (0.011)
ppe −0.0222* −0.0161*
(0.011) (0.009)
nol −0.4353 −0.1640
(0.583) (0.442)
dsale −0.0087* −0.0070**
(0.005) (0.004)
cash 0.0323* 0.0135
(0.019) (0.014)
intan −0.0557 −0.0565
(0.050) (0.038)
absda −0.5324 −0.2642
(0.331) (0.251)
eqinc 0.3484*** 0.2927***
(0.133) (0.101)
Explanatory Var.
CSR −0.0269*** −0.0035 −0.0269*** −0.0033
(0.007) (0.008) (0.007) (0.006)
Mediator
LROA 0.5540*** 0.4524***
(0.064) (0.049)
_cons 2.7889* 0.9633 2.7889* 2.0146
(1.626) (1.923) (1.626) (1.458)
N 495 233 495 233
adj. R2 0.060 0.407 0.060 0.457
F 11.45 14.27 11.45 17.25
p-value [0.00] [0.00] [0.00] [0.00]
*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Standard error
is included in the parentheses.

the association between corporate tax avoidance and CSR performance. Initially, CSR perfor-
mance reduces corporate profitability, and corporate profitability in turn increases the level of
corporate tax avoidance. Moreover, we observe that CSR performance does not moderate the
mediating effect. Therefore, we conclude that CSR performance has an indirect effect (or mediat-
ing effect), but not a conditional indirect effect (or moderated mediation effect), on the level of
corporate tax avoidance.
This study investigates only the simple mediation effect and moderated mediation effect.
However, other mediators and moderators may affect the association between CSR performance
and the level of corporate tax avoidance. Therefore, future research can further investigate
additional mediators and moderators.
16 C.-W. MAO AND W.-C. WU

Notes
1. Irresponsible CSR activities include corporate actions widely regarded as damaging to aspects such as
corporate governance, employee relations, communities, public health, human rights, diversity, and the
environment (Hoi, Wu, and Zhang 2013).
2. The firm profit dummy variable equals 1 if a firm’s ROA is at least 10% (high-profit firms) and 0 otherwise
(low-profit firms).
3. The weights are determined based on the relative importance of various social responsibility items. In
addition, the weights vary according to P the industry to which the company in question belongs.
4. The formula for the original score is j¼1;2;3;4 Aj Wj , where j is the social responsibility item, A is the score
of each social responsibility item, and W is the weight.
5. The adjusted items include CSR awards and major CSR failure events, etc.
6. In contrast to Western countries, whose CSR evaluation institutions are numerous and possess compre-
hensive data, China’s CSR development is in its preliminary stage. Therefore, China has few CSR evaluation
institutions and their data are limited. CASS-CSR research center data are currently China’s most crucial
CSR assessment data. Thus, although CASS-CSR research center data are imperfect because of problems
such as the short time span between making an application and receiving the rating report, we still use this
information for analysis.
7. Besides, we exclude financial, insurance, securities, agriculture, forestry, fishing, animal husbandry, educa-
tion, comprehensive, and non-industry companies.
8. The mediating effect exists when the following conditions are satisfied: First, the independent variable has a
significant effect on the mediator. Second, the independent variable has a significant effect on the dependent
variable. Third, when the mediator is controlled, the effect of the independent variable on the dependent
variable becomes nonsignificant; Under such conditions, the mediator is called the ‘full mediator.’ However,
if the effect of the full mediator is significant but its size is smaller than that of the uncontrolled mediator, it
is called the ‘partial mediator.’

Acknowledgments
We thank Yu-Hui Su, Jenn-Shyong Kuo, Wen-Sheng Shieh, Li-Peng Hsiao, and the workshop participants at the
2017 Taiwan Accounting Association Annual Conference and Asian Accounting Association Conference for their
helpful comments. We also appreciate the insightful comments and suggestions made by the anonymous reviewers
and the editor.

Disclosure statement
No potential conflict of interest was reported by the authors.

References
Aras, G., and D. Crowther. 2009. “Corporate Sustainability Reporting: A Study in Disingenuity?” Journal of
Business Ethics 87 (1): 279–288. doi:10.1007/s10551-008-9806-0.
Aupperle, K., A. Carroll, and J. Hatfield. 1985. “An Empirical Examination of the Relationship between Corporate
Social Responsibility and Profitability.” Academy of Management Journal 28 (2): 446–463.
Baron, R. M., and D. A. Kenny. 1986. “The Moderator–Mediator Variable Distinction in Social Psychological
Research: Conceptual, Strategic, and Statistical Considerations.” Journal of Personality and Social Psychology 51
(6): 1173–1182. doi:10.1037/0022-3514.51.6.1173.
Becchetti, L., R. Ciciretti, and I. Hasan. 2007. “Corporate Social Responsibility and Shareholder’s Value: An Event
Study Analysis.” Working Paper No. 2007-6, Federal Reserve Bank of Atlanta.
Callan, S. J., and J. M. Thomas. 2009. “Corporate Financial Performance and Corporate Social Performance: An
Update and Reinvestigation.” Corporate Social Responsibility and Environmental Management 16 (2): 61–78.
doi:10.1002/csr.v16:2.
Cardebat, J. M., and N. Sirven. 2010. “What Corporate Social Responsibility Reporting Adds to Financial Return?”
Journal of Economics and International Finance 2 (2): 20–27.
Dechow, P. M., R. G. Sloan, and A. P. Sweeney. 1995. “Detecting Earnings Management.” Accounting Review 70 (2):
193-225.
Desai, M. A., and D. Dharmapala. 2006. “Corporate Tax Avoidance and High-Powered Incentives.” Journal of
Financial Economics 79 (1): 145–179. doi:10.1016/j.jfineco.2005.02.002.
 ASIA-PACIFIC JOURNAL OF ACCOUNTING & ECONOMICS 17

Edwards, J. R., and L. S. Lambert. 2007. “Methods for Integrating Moderation and Mediation: A General Analytical
Framework Using Moderated Path Analysis.” Psychological Methods 12 (1): 1–22. doi:10.1037/1082-989X.12.1.1.
Frank, M., L. Lynch, and S. Rego. 2009. “Tax Reporting Aggressiveness and Its Relation to Aggressive Financial
Reporting.” The Accounting Review 84 (2): 467–496. doi:10.2308/accr.2009.84.2.467.
Garcia-Castro, R., M. A. Ariño, and M. A. Canela. 2010. “Does Social Performance Really Lead to Financial Performance?
Accounting for Endogeneity.” Journal of Business Ethics 92 (1): 107–126. doi:10.1007/s10551-009-0143-8.
Godfrey, P. C. 2005. “The Relationship between Corporate Philanthropy and Shareholder Wealth: A Risk
Management Perspective.” Academy of Management Review 30 (4): 777–798. doi:10.5465/amr.2005.18378878.
Godfrey, P. C., C. B. Merrill, and J. M. Hansen. 2009. “The Relationship between Corporate Social Responsibility
and Shareholder Value: An Empirical Test of the Risk Management Hypothesis.” Strategic Management Journal
30 (4): 425–445. doi:10.1002/smj.v30:4.
Gras-Gil, E., M. P. Manzano, and J. H. Fernández. 2016. “Investigating the Relationship between Corporate Social
Responsibility and Earnings Management: Evidence from Spain.” BRQ Business Research Quarterly 19 (4): 289–
299. doi:10.1016/j.brq.2016.02.002.
Gupta, S., and K. Newberry. 1997. “Determinants of the Variability in Corporate Effective Tax Rates: Evidence from
Longitudinal Study.” Journal of Accounting and Public Policy 16 (1): 1–34. doi:10.1016/S0278-4254(96)00055-5.
Hanlon, M., and S. Heitzman. 2010. “A Review of Tax Research.” Journal of Accounting and Economics 50 (2): 127–
178. doi:10.1016/j.jacceco.2010.09.002.
Hasseldine, J., and G. Morris. 2013. “Corporate Social Responsibility and Tax Avoidance: A Comment and
Reflection.” Accounting Forum 37 (1): 1–14. doi:10.1016/j.accfor.2012.05.001.
Hayes, A. 2013. Introduction to Mediation, Moderation, and Conditional Process: A Regression-Based Approach.
New York, NY: The Guilford Press.
Hoi, C. K., Q. Wu, and H. Zhang. 2013. “Is Corporate Social Responsibility (CSR) Associated with Tax Avoidance?
Evidence from Irresponsible CSR Activities.” The Accounting Review 88 (6): 2025–2059. doi:10.2308/accr-50544.
Huseynov, F., and B. K. Klamm. 2012. “Tax Avoidance, Tax Management and Corporate Social Responsibility.”
Journal of Corporate Finance 18 (4): 804–827. doi:10.1016/j.jcorpfin.2012.06.005.
Khurana, I. K., and W. J. Moser. 2013. “Institutional Shareholders’ Investment Horizons and Tax Avoidance.” The
Journal of the American Taxation Association 35 (1): 111–134. doi:10.2308/atax-50315.
Kreps, D. M. 1990. “Corporate Culture and Economic Theory.” In Perspectives on Positive Political Economy, edited
by J. E. Alt and K. A. Shepsle. Cambridge, U.K.: Cambridge University Press: 90-143.
Kubick, T. R., D. P. Lynch, M. A. Mayberry, and T. C. Omer. 2015. “Product Market Power and Tax Avoidance: Market
Leaders, Mimicking Strategies, and Stock Returns.” The Accounting Review 90 (2): 675–702. doi:10.2308/accr-50883.
Landolf, U. 2006. “Tax and Corporate Responsibility.” International Tax Review 29 (July): 6–9.
Lanis, R., and G. Richardson. 2012. “Corporate Social Responsibility and Tax Aggressiveness: An Empirical
Analysis.” Journal of Accounting and Public Policy 31 (1): 86–108. doi:10.1016/j.jaccpubpol.2011.10.006.
Lanis, R., and G. Richardson. 2015. “Is the Corporate Social Responsibility Performance Associated with Tax
Avoidance?” Journal of Business Ethnics 127 (2): 439–457. doi:10.1007/s10551-014-2052-8.
Lisowsky, P., L. A. Robinson, and A. P. Schmidt. 2013. “Do Publicly Disclosed Tax Reserves Tell Us about Privately
Disclosed Tax Shelter Activity?” Journal of Accounting Research 51 (3): 583–629. doi:10.1111/joar.2013.51.issue-3.
Preacher, K., D. Rucker, and A. Hayes. 2007. “Addressing Moderated Mediation Hypotheses: Theory, Methods, and
Prescriptions.” Multivariate Behavior Research 42 (1): 185–227. doi:10.1080/00273170701341316.
Schadewitz, H., and M. Niskala. 2010. “Communication via Responsibility Reporting and Its Effect on Firm Value
in Finland.” Corporate Social Responsibility and Environmental Management 17 (2): 96–106.
Scholtens, B. 2008. “A Note on the Interaction between Corporate Social Responsibility and Financial
Performance.” Ecological Economics 68 (1): 46–55. doi:10.1016/j.ecolecon.2008.01.024.
Shen, C. H., and Y. Chang. 2008. “Does Corporate Social Responsibility Improve Financial Performance? Evidence
from FTSE4GOOD UK Index.” Academia Economic Papers 36 (3): 339–385.
Slemrod, J., and S. Yitzhaki. 2002. “Tax Avoidance, Evasion, and Administration.” Handbook of Public Economics 3:
1423–1470.
Sobel, M. E. 1982. “Asymptotic Confidence Intervals for Indirect Effects in Structural Equation Models.”
Sociological Methodology 13: 290–312. doi:10.2307/270723.
Tsoutsoura, M. 2004. Corporate Social Responsibility and Financial Performance. Center for Responsible Business,
Haas School of Business, USA: University of California, Berkely.
Ullmann, A. 1985. “Data in Search of A Theory: A Critical Examination of the Relationships among Social
Performance, Social Disclosure, and Economic Performance of US Firms.” Academy of Management Review
10 (3): 540–557.
Watson, L. 2015. “Corporate Social Responsibility, Tax Avoidance, and Earnings Performance.” The Journal of the
American Taxation Association 37 (2): 1–21. doi:10.2308/atax-51022.
Wilson, R. 2009. “An Examination of Corporate Tax Shelter Participants.” The Accounting Review 84 (3): 969–999.
doi:10.2308/accr.2009.84.3.969.
18 C.-W. MAO AND W.-C. WU

Appendix

Variable Definitions

Variable Definition
pbtd (%) Khurana and Moser (2013) permanent book-tax difference.
pbtd is calculated as follows:
bi taxexp taxexpdef
pbtd ¼ taxrateta taxrate
where bi is pre-tax book income; taxexp is tax expense; taxrate is corporate tax rate; taxexpdef is
deferred tax expense; and ta is total assets.
ddbtd Desai and Dharmapala (2006) discretionary book-tax difference.
ddbtd is equal to μiνit from the following regression:
BTit ¼ θtacc þ μi þ νit
where BT is the pretax income less taxable income, divided by total assets; tacc is Dechow et al. (1995)
total accruals measure, scaled by total assets;νit is the average value of the residual for firm i over the
sample period; and νit is the deviation of the residual in year t from firm i’s average residual.
letr (%) 4-year (the current and past 3 years) average effective tax rate. Effective tax rate is defined as book tax
expense divided by pre-tax book income. This study truncates letr to the range [0, 1].
ROA (%) Net income scaled by total assets;
LROA (%) 4-year (the current and subsequent 3 years) average return on assets;
CSR Overall score of CSR activities is reported by the CSR research center at the Chinese Academy of Social
Sciences;
Size Natural logarithm of the firm’s total assets;
lev (%) Debt scaled by total assets;
ppe (%) Property, plant, and equipment divided by total assets;
inv (%) Ending inventory scaled by total assets;
nol Indicator variable equal to 1 if a tax loss is carried forward and 0 otherwise;
dsale (%) Changes in sales scaled by lagged sales;
absda Absolute value of discretionary accruals, where discretionary accruals are computed using the modified
Jones model including lagged ROA as an additional regressor;
intan (%) Intangible assets scaled by total assets;
cash (%) Cash holding divided by total assets;
eqinc (%) Equity method income in earnings scaled by total assets.