The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/0737-4607.htm

Journal of
The impact of superstition Accounting
Literature
on corporate tax avoidance:
how do CEOs trade off risks
associated with tax avoidance?
Guanming He and Dongxiao Shen Received 3 February 2024
Revised 2 June 2024
Department of Accounting, Durham University Business School, Durham, UK 27 July 2024
Accepted 28 August 2024

Abstract
Purpose – We examine how superstition shapes corporate tax avoidance and do so by taking a risk
perspective and focusing on the zodiac-year belief prevalent in China.
Design/methodology/approach – We adopt a difference-in-differences research design to compare the
degree of corporate tax avoidance in the CEOs’ zodiac year with that in the adjacent years. We do propensity-
score matching to form a sample of Chinese listed firms for the regression analysis.
Findings – We find causal evidence that firms exhibit a greater magnitude of tax avoidance in the CEOs’
zodiac years, a result attributable to relatively weak tax enforcement in the Chinese context. We also find that
the zodiac-year effect on corporate tax avoidance is more pronounced for firms with tight financial constraints,
firms with high business risk, firms headquartered in regions with a high degree of superstition and non-state-
owned firms.
Originality/value – This study is the first to show that superstition is a determinant factor of tax avoidance
and contributes to the tax literature by shedding light on the behavioral risk factors that shape corporate tax
avoidance. We take the perspective of CEOs’ risk appetite to analyze how tax avoidance is influenced by the
CEOs’ trade-off between the costs and benefits of avoiding taxes. Our results suggest that, when CEOs are
more risk-averse, they attach more importance to financial risk than the risk of reputational losses and
litigation associated with corporate tax avoidance. The findings imply that tax avoidance can be curbed by
increasing (or decreasing) the tax (financial) risk confronting the CEOs.
Keywords Superstition, Risk averseness, Risk trade-off, Corporate tax avoidance
Paper type Research paper

1. Introduction
Corporate tax avoidance refers to firm activities that lead to any reduction in explicit taxes,
including the adoption of various legal or illegal tax strategies (Dyreng et al., 2008; Hanlon
and Heitzman, 2010). Despite voluminous research on the economic determinants of
corporate tax avoidance, such as profitability, research and development investments, and
the intensity of foreign operations (e.g. Graham and Tucker, 2006; Rego, 2003), the role of
behavioral factors associated with managers has been underexplored (Hanlon et al., 2022).
Managerial decision-making is likely shaped by behavioral factors such as individual
preferences, cognitive biases, and culture (Rodgers and Gago, 2001; Schwenk, 1986).
The objective of our research is to investigate whether and how superstition, which can

JEL Classification — H26, M41, D81

We thank Tom Smith and Bruce Vanstone for their support in the review process. We are also
grateful to an anonymous reviewer, Scott Dyreng, Christopher Williams, Edward Owens, Michael Guo,
Yeqin Zeng, and participants at the 2023 Global Emerging Scholars Research Workshop, the tax
concurrent session of 2023 American Accounting Association annual meeting, and the 2023 Northern Journal of Accounting Literature
Advanced Research Training Initiative Annual Conference for their helpful comments and suggestions © Emerald Publishing Limited
0737-4607
on the paper. We are accountable for errors, if any, in the paper. DOI 10.1108/JAL-02-2024-0020
JAL impact various human behaviors (e.g. Block and Kramer, 2009) and decision-making
(Rice, 1985; Tsang, 2004a, b), impacts corporate tax avoidance.
Superstition is defined by Merriam-Webster Dictionary as a belief or practice resulting
from ignorance, fear of the unknown, trust in magic or chance, or a false conception of
causation. Superstition prevails in human societies, even in modern times. For example, the
opening ceremony of Beijing Summer Olympics began at 8 p.m. on 8 August 2008 because
Chinese people believe that the digit 8 is associated with prosperity. In this study, we
examine whether and how zodiac-year belief, a superstition upheld by a great deal of Chinese
people, impacts corporate tax avoidance. In Chinese society, every individual would
encounter her/his zodiac year once every 12 years. It is believed that in the zodiac year, the
likelihood and degree of a person experiencing misfortune will be higher. Thus, individuals
will be more risk-averse in their decision-making and behaviors.
We focus on zodiac-year belief for two reasons. First, zodiac-year belief provides a nice
scenario in which to investigate how managers weigh different risks associated with
corporate tax avoidance when they are risk-averse. A decision on whether and to what degree
to engage in tax avoidance results from a trade-off between the marginal benefits and costs of
avoiding taxes (e.g. Hanlon and Heitzman, 2010) [1]. The key benefit of tax avoidance to a
firm is an increase in internal funds, which reduce potential financial constraints or financial
distress (e.g. Desai and Dharmapala, 2009). However, tax avoidance, if detected, will subject
firms and managers to reputational losses and legal penalties (e.g. Hanlon and Slemrod,
2009). The cost-benefit trade-off depends crucially on the risk appetites of chief executive
officers (CEOs) who make the decision on tax avoidance. If CEOs are more (less) averse to the
risk associated with financial constraints or distress than the reputational and legal risks
arising from the revelation of tax avoidance, they will be more (less) likely to avoid taxes. We
thus propose two competing hypotheses. On the one hand, CEOs with zodiac-year belief
might have an incentive to avoid taxes to increase internal funds and thereby mitigate the
potential financial risk. On the other hand, the CEOs might have a propensity to refrain from
tax avoidance for fear of the possible reputational losses and threat of litigation.
Second, zodiac-year belief provides a reasonable setting to examine the causal
relationship between superstition and tax avoidance. Some superstitious beliefs arise
when an individual lacks control over an outcome (e.g. Case et al., 2004; Felson and Gmelch,
1979; Jahoda, 1969; Keinan, 2002; Malinowski, 1925; Rice, 2003; Schippers and Van Lange,
2006), and such beliefs are therefore endogenous. For instance, Case et al. (2004) discover that
people are more likely to utilize superstitious explanations for a failed outcome if the failure is
more salient. By contrast, zodiac-year belief is typically imprinted in a person’s mindset;
furthermore, zodiac years take place on a 12-year cycle for all individuals and are presumably
distributed in a randomly staggered manner that is independent of the events for, and
characteristics of, CEOs and their firms. Therefore, zodiac-year belief should be exogenous to
the CEOs’ decision-making for their firms, facilitating us to draw causal inferences from our
empirical analysis.
Our empirical tests are based on a sample of publicly listed Chinese firms for the period
2009–2019. Our main measure of tax avoidance is the residual book-tax difference, which is
estimated from the firm-fixed-effect regression of the total book-tax difference on the total
accruals (Desai and Dharmapala, 2006). We test the competing hypotheses through
difference-in-differences (DID) research design coupled with propensity-score matching. The
treatment sample used in the DID regression analysis comprises firms whose CEOs are in
their zodiac years for the current year but not for the previous year and following year. Each
treatment firm is matched with a control firm that has similar characteristics in the same year
and industry. The distribution of the CEOs’ zodiac years is presumably orthogonal to
corporate events and firm/CEO characteristics. That said, to ensure a clean comparison with
the treatment sample (Baker et al., 2022; Roth et al., 2023), we require each control firm to have
its CEO that is not in her/his zodiac year for the three-year period centered in the zodiac year Journal of
of the CEO of the matched treatment firm. We find that the degree of tax avoidance for the Accounting
treatment firms is higher in their CEOs’ zodiac years, compared to the year before and after Literature
the zodiac years and relative to that of the control firms. This finding is robust to controlling
for firm-fixed effects and using alternative measures of tax avoidance. We also draw the
same inference after performing a placebo test to enhance the validity of the treatment effect.
We further examine whether the firm’s financial constraints, business risk, the degree of
superstition in different regions, and state ownership moderate the impact of zodiac-year
belief on corporate tax avoidance. We expect the zodiac-year effect to be more prominent
when firms face severe financial constraints, as tax avoidance is a means of generating
internal cash flows pivotal for the financially constrained firms. We employ the SA index
(Hadlock and Pierce, 2010) and cash dividends (e.g. Denis and Sibilkov, 2010; Fazzari et al.,
1988) as the proxies for financial constraints, and find results consistent with the prediction.
We also expect that the relationship between the CEOs’ zodiac years and tax avoidance is
more pronounced for firms with high business risk, based on two grounds. First, the
psychology and sociology literature documents that individuals are more likely to rely on
superstition to make judgments and decisions when the risks and uncertainty of a future
outcome are high. In such a case, superstition could provide individuals with a sense of
control over the outcome and thus help relieve their stress. Second, superstition could provide
managers with supernatural information which can help compensate for their cognitive
limitations when they confront the high risk (Tsang, 2004a). Using the volatility of return on
sales and that of return on assets as the proxies for the firm’s business risk, we find that the
zodiac-year effect on tax avoidance is evident only in cases when firms have high
business risk.
Furthermore, we postulate that the zodiac-year effect is more salient for firms located in
the highly superstitious regions. In these regions, superstition is spread easily through the
inheritance from early generations, or via the interactions with peer groups, and is thereby
embedded within the social norm of the local people. Given that individuals are likely to
habitually follow the shared norms without a rational calculation (e.g. Bicchieri, 1990;
Bicchieri et al., 2018), we argue that CEOs have a greater tendency to engage in superstitious
practices if they work in the areas with strong superstitious norms. As for the empirical tests,
we employ the Baidu search index to construct the measure of the degree of zodiac-year belief
for each provincial area [2]. We find evidence that the positive relationship between the
CEOs’ zodiac-year belief and tax avoidance is evident only for the subsample of firms
headquartered in the highly superstitious provinces. In addition, we expect that CEOs in
state-owned enterprises (SOEs) are unlikely to be superstitious because they are typically
disciplined by the Chinese Communist Party, of which the rules oppose the superstitious
beliefs and practices. Consistent with this supposition, we find that the zodiac-year effect on
tax avoidance takes place only in non-state-owned enterprises.
Predominant literature documents that corporate tax avoidance is determined by
economic motives associated with various firm characteristics (e.g. Graham and Tucker,
2006; Rego, 2003; Wilson, 2009), the internal and external governance (e.g. Cheng et al., 2012;
Desai and Dharmapala, 2006; McGuire et al., 2012a, b), the macroeconomic factors (e.g. Hong
et al., 2019; Katz and Owen, 2013), and the power of tax enforcement (Desai et al., 2007),
among others. Some studies on tax avoidance examine the cultural determinants relating to
religiosity (Boone et al., 2013) and cultural diversity (Lei et al., 2022). Yet, much remain to
learn regarding the role of social and behavioral factors in shaping tax avoidance.
To the best of our knowledge, we are the first to show that superstition is a determinant
factor of tax avoidance. To this end, and unlike the related literature, we take the perspective
of CEOs’ risk appetite to analyze how tax avoidance is influenced by the CEOs’ trade-off
between the costs and benefits of avoiding taxes, and provide causal evidence that
JAL superstition increases tax avoidance. From this evidence, we infer that in cases when CEOs
are risk-averse, they are apt to weigh the risk of financial constraints and financial distress
more than the reputational and legal risks arising from tax avoidance. As such, our paper
contributes to the tax literature by shedding light on the behavioral risk factors that shape
corporate tax avoidance. It also responds to the call of Hanlon et al. (2022) for more research
on behaviorial economics of accounting.
Our paper is related to two prior studies on corporate tax avoidance. Badertscher et al.
(2013) find that managers in the firms with greater concentrations of ownership and control
are likely to be more risk averse and avoid less income taxes. Boone et al. (2013) show that
firms headquartered in more religious counties are more risk averse and less aggressive in
tax avoidance. Both studies imply a negative relationship between corporate tax avoidance
and managerial risk aversion. Yet, their research is based on the US setting where tax
enforcement is stronger, and the legal and reputational risks are higher for tax-avoiding
firms (e.g. Graham et al., 2014; Hanlon and Slemrod, 2009; Lee et al., 2021), compared to the
Chinese context (Brondolo and Zhang, 2016). Our paper complements the two studies by
accounting for the risk-related benefits of tax avoidance and showing that the tax
misconduct could be more intensive for risk-averse managers in a country where tax risk is
lower for firms due to relatively weaker tax enforcement.
In addition, our study contributes to the scarce literature on the effect of superstition on
managerial decision-making and corporate business activities, such as recruiting staff
(Tsang, 2004a), pursuing research and development, and engaging in mergers and
acquisitions (Fisman et al., 2023). We focus on tax avoidance and find that it increases with
the degree of superstition. This finding reconciles with the previous studies which show the
dark side of superstition (e.g. Bai et al., 2020; Bhattacharya et al., 2018; Li et al., 2021).
The remainder of the paper is organized as follows. Section 2 develops hypotheses.
Section 3 discusses the sample and variable measurements. Section 4 explains the research
design and discusses the results. Section 5 concludes the study.

2. Literature review and hypotheses development

2.1 Zodiac-year superstition
The Chinese zodiac is a traditional astrological scheme that assigns a distinct sign of animal
to each lunar year in a recurring 12-year cycle. The cycle involves 12 animals in the sequence
of Rat, Ox, Tiger, Rabbit, Dragon, Snake, Horse, Goat, Monkey, Rooster, Dog, and Pig. The
lunar year with the same animal sign as an individual’s birth year is known as her/his zodiac
year. In traditional Chinese culture, the zodiac year is regarded as a year in which a person is
likely to encounter bad luck and calamities. As such, Chinese people tend to be risk averse in
their decision-making and behaviors. Recent studies (e.g. Dou et al., 2024; Fisman et al., 2023;
Zeng et al., 2022) provide evidence consistent with this notion. For instance, Fisman et al.
(2023) document that managers reduce risky acquisition and innovation activities in their
zodiac years. Dou et al. (2024) show that auditors feeling higher audit risk in their zodiac
years would provide higher-quality audits for their clients.

2.2 Zodiac-year superstitious belief and tax avoidance

The literature in psychology documents that the superstitious behavior will be aroused
in situations in which the individuals lack control over an important outcome (e.g. Case et al.,
2004; Keinan, 2002; Rice, 2003; Schippers and Van Lange, 2006). No matter whether they are
convinced by the superstition, individuals may still take superstitious actions as a means to
preserve emotional stability and boost confidence (e.g. Bleak and Frederick, 1998; Case et al.,
2004). Given that firm managers often face uncertainty and risks in their business decisions,
they tend to adhere to superstitious beliefs when making choices and taking actions (Liu Journal of
et al., 2023; Tsang, 2004a, b). One such business decision that carries inherent risks is whether Accounting
to pursue corporate tax avoidance. In light of this, we posit that managers’ superstitious Literature
beliefs might play a role in shaping tax avoidance.
Whether to engage in tax avoidance results from a trade-off between the benefits and
costs of doing so (e.g. Gallemore et al., 2014; Hanlon and Heitzman, 2010). If CEOs expect the
benefits outweigh the costs, they will have a tendency of going after opportunities to avoid
taxes (Hanlon and Heitzman, 2010). The main benefit of tax avoidance to a firm is to reduce
its tax liabilities, generating additional internal cash flows that may reduce financial risk (e.g.
Cen et al., 2017a, b; Chen et al., 2010; Desai and Dharmapala, 2009; Graham et al., 2014; Rego
and Wilson, 2012). Managing financial risk is a crucial part of risk management for Chinese
listed firms. China is recognized as an investment-driven economy, where there is a
considerable demand for investments, particularly among listed firms, leading to substantial
capital requirements and thus the great need for financing. However, China’s capital markets
do not function as efficiently as those of developed countries, making it relative more difficult
to meet the financing needs by the listed firms. As a result, there is a certain degree of
financial risk that managers need to navigate and address. Since the zodiac-year belief
foretells that there is a high risk of unfortunate events happening in the zodiac year, CEOs
may foresee that the uncertainties of investments and operation are higher, and thus
financial risk is higher, in their zodiac years. As such, CEOs may intend to avoid taxes to
mitigate the financial risk.
On the other hand, there are risks for conducting tax-avoidance activities. Tax avoidance,
if discovered by tax authorities, will subject managers and their firms to reputational losses
and legal penalties (e.g. Chen et al., 2010; Desai and Dharmapala, 2009; Hanlon and Slemrod,
2009; Kim et al., 2011). Thus, CEOs in their zodiac years might be averse to the risks and costs
associated with tax avoidance and therefore abstain from engaging in it. Yet, provided that
such risks and costs are low for the CEOs, not least in the Chinese setting in which tax
enforcement is relatively weak (Brondolo and Zhang, 2016), they might be prone to pursue
tax avoidance. By and large, if the CEOs are more (less) averse to the foregoing financial risk
than the tax risks, they will be more (less) likely to avoid taxes in their zodiac year. Therefore,
we formulate the following competing hypotheses for empirical analysis.
H1a. CEOs are more likely to engage in tax avoidance in their zodiac years.
H1b. CEOs are less likely to engage in tax avoidance in their zodiac years.

2.3 The moderating effect of financial constraints

Financially constrained firms face greater difficulty in external funding. Consequently, they
might have to rely on internal funds to make necessary investments and avoid debt defaults.
One potential means of generating internal cash flows is to avoid taxes. In so doing, their
financial constraints would be relieved. Consistent with this notion, a number of studies (e.g.
Dyreng and Markle, 2016; Edwards et al., 2016; Law and Mills, 2015) provide evidence on a
positive association between financial constraints and corporate tax avoidance. Since firms
that are in financial constraints tend to have higher distress risk (e.g. He and Ren, 2023), the
benefits of their engagement in tax avoidance to contain the default risk would be higher.
Hence, we expect that CEOs are more prone to avoid taxes in their zodiac years when their
firm is in financial constraint.
H2. The positive (negative) impact of zodiac-year superstitious belief on tax avoidance,
as hypothesized in H1a (H1b), is stronger (weaker) when firms face tight financial
constraints.
JAL 2.4 The moderating effect of business risk
The literature relating to the psychology aspect of superstition (e.g. Keinan, 2002; Kramer
and Block, 2008) documents that individuals are more likely to be superstitious when they
confront greater uncertainty, and that superstitions can relieve the stress associated with the
uncertainty. On this basis, we posit that the degree of business risk would moderate the effect
of zodiac-year belief on corporate tax avoidance. In particular, in cases when a firm faces high
business risk (i.e. when the firm’s business outcome is highly uncertain), managers are under
pressure to contain the risk so as to avoid increased costs of capital (e.g. Francis et al., 2004),
reduced business support from stakeholders (Hannan and Freeman, 1984), and heightened
distress risk (e.g. Amit and Wernerfelt, 1990). Thus, we expect that CEOs are prone to resort
to the superstition to obtain psychological benefits when the firm’s business risk is
relatively high.
Furthermore, to manage the firm’s risk well, CEOs need to seek additional information
(Beckman et al., 2004; Lipshitz and Strauss, 1997), and may interpret ambiguity as a
threat to their decision-making process (Budner, 1962). When CEOs are in bounded
rationality due to the high uncertainty of the decision process, they are prone to refer to
multiple sources of information and create more diverse viewpoints to help overcome
their cognitive limitations (e.g. Eisenhardt and Zbaracki, 1992; Payne et al., 1988; Wang
et al., 2012). As such, superstition could help CEOs to deal with the uncertainty and risks
by providing an additional source of information for their reference (Tsang, 2004a, b). Put
differently, the zodiac-year belief could provide complementary information by
foretelling bad luck. As the zodiac CEOs account for the misfortune, they are apt to
contain the high business risk; to this end, they might avoid taxes more aggressively to
save cash. Based on the above two sets of arguments, we propose our second hypothesis
as follows:
H3. The positive (negative) impact of zodiac-year superstitious belief on tax avoidance,
as hypothesized in H1a (H1b), is stronger (weaker) when firms face high
business risk.

2.5 The moderating effect of the region-level superstition

The degree of superstitious norms differs across regions owing to their variation in history,
culture, and customs of the locals. In the highly superstitious areas, a considerable number of
individuals perceive that superstition is appropriate, such that those refraining from the
superstitious practice would incur higher social or economic costs than those following it;
hence, the superstitious practices or beliefs become entrenched and prevail over time (Smith,
2003). According to the sociology literature (e.g. Akerlof and Kranton, 2005), individuals will
seek conformity and follow social norms under the peer pressure and influences. As such, a
person tends to engage in superstitious practices in the region with strong superstitious
norms, no matter whether s/he is convinced. In this sense, CEOs might be affected by the
social environment in which their firms are headquartered (e.g. Dyreng et al., 2012; Hilary
and Hui, 2009; McGuire et al., 2012a, b). Put differently, for firms located in a region that
features strong superstitious norms, CEOs are likely to act in ways that conform with the
norms. On this basis, we expect that the zodiac-year effect on corporate tax avoidance is more
prominent in regions where zodiac-year belief is more pervasive, and propose the fourth
hypothesis as follows:
H4. The positive (negative) impact of zodiac-year superstitious belief on tax avoidance,
as hypothesized in H1a (H1b), is stronger (weaker) for firms that are headquartered
in regions with a high degree of superstition.
2.6 The moderating effect of state ownership Journal of
There is substantial difference in the organizational culture between state-owned enterprises Accounting
(SOEs) and non-state-owned enterprises (non-SOEs). For SOEs, their corporate culture is Literature
deeply shaped by the Chinese Communist Party, since each SOE establishes a communist-
party committee to take the responsibility for the managerial appointments and promotions,
the implementation of the government’s propositions, and the party disciplines for the firm
(Hu and Xu, 2022; Yu, 2019). The CEOs of SOEs are typically the Communist Party members
and/or government officials. Their work performance, ideology, and even lifestyles are
regularly inspected by the Party Committee. Since superstition is repressed under the
communist rules (Tsang, 2004b), we expect that the CEOs of SOEs are less likely to pursue
superstitious practices. On the contrary, the corporate culture of non-SOEs is more inclusive,
and their CEOs are more open-minded and should be more likely to embrace the superstition.
We therefore expect the zodiac-year effect on corporate tax avoidance to be more pronounced
for non-SOEs than SOEs, and accordingly, put forward the fifth hypothesis as follows:
H5. The positive (negative) impact of zodiac-year superstitious belief on tax avoidance,
as hypothesized in H1a (H1b), is stronger (weaker) for non-state-owned firms than
state-owned firms.

3. Variable measurements and samples

3.1 Variable measurements
3.1.1 Measurement of zodiac-year belief. We focus on chief executive officers (CEOs) for our
hypothesis tests because CEO is the ultimate person who is in charge of corporate tax
planning and responsible for tax avoidance. CEOs typically have a broader and more
encompassing authority over corporate decision-making, including tax planning/reporting.
Their position at the helm of organization empowers them to navigate the risks and potential
rewards associated with tax avoidance. They can affect tax avoidance by setting the “tone at
the top” with regard to the firm’s tax activities (Dyreng et al., 2010). Much literature has
provided evidence that tax avoidance is associated with CEO compensation (e.g. Armstrong
et al., 2015; Gaertner, 2014; Powers et al., 2016; Rego and Wilson, 2012), CEO turnover (Chyz
and Gaertner, 2018), and various personal characteristics of CEOs (e.g. Duan et al., 2018; Law
and Mills, 2017; Olsen and Stekelberg, 2016). In the specific context of our study, CEOs have
an incentive to avoid taxes for more internal funds to mitigate financial risk associated with
perceived uncertainties of investments and operation. But chief financial officers (CFOs) do
not have such an incentive to do so, as they hold responsibility for making corporate tax
payments comply with tax laws. Therefore, we exclude the observations of CFOs’ zodiac
years in the sample selection process [3]. We do not examine the effect of the board
chairmen’s zodiac-year beliefs either, because the chairmen tend to be concerned mainly with
the strategic decision-making on their firm’s business activities rather than with corporate
tax planning/reporting.
The zodiac year starts from a day in January or February and ends in the next January or
February of the Gregorian calendar, resulting in one or two months of a calendar year not
being in the zodiac year. Therefore, to accurately capture the zodiac-year effect on
superstition for CEOs, we create an indicator variable, Zodiac_ceo, based on two conditions,
respectively. First, if the CEO was born on a day between the start date of the Chinese Lunar
new year and 31 December, Zodiac_ceo equals 1 for cases in which a CEO is at the age of a
multiple of 12, and 0 otherwise. Second, if the CEO was born on a day between 1 January and
the start date of the Chinese Lunar new year, Zodiac_ceo equals 1 for cases in which a CEO is
at the age of one year less than a multiple of 12, and 0 otherwise. Nevertheless, only a small
subset of our sample includes the CEOs’ birthdates, since such information is typically
JAL considered private and would not be publicly disclosed unless agreed by the CEOs. For the
sample that just has the information of CEOs’ birth months and birth years, we construct
Zodiac_ceo in the following way. In the case in which the CEO was born in a month between
March and December, or in February that is later than the start date of Chinese lunar new
year in January, we assign a value of 1 to Zodiac_ceo if the CEO’s age is a multiple of 12, and
0 otherwise. In the case where the CEO was born in January that precedes the start date of
Chinese lunar new year in February, Zodiac_ceo is assigned a value of 1 if the CEO is at the
age of one year less than a multiple of 12, and 0 otherwise. For the rest of cases in which we
only know the birth year of the CEO, or the CEO’s birth month overlaps with the start month
of Chinese lunar new year, we are unable to identify whether the CEO is in the zodiac year
when s/he is aged a multiple of 12 or one less than the multiple of 12. Accordingly, the
observations for these CEOs are excluded from our sample.
3.1.2 Measurement of corporate tax avoidance. To estimate the degree of corporate tax
avoidance, we run the following firm-fixed-effects regression model (Desai and
Dharmapala, 2006):

BTDi;t ¼ α1 TACC i;t þ μi þ εi;t (1)

where BTDi;t is the total book-tax difference for firm i in year t, computed as pre-tax financial
income minus taxable income; the latter equals income tax expense minus deferred tax
expense and divided by the nominal tax rate; TACC i;t is the total accruals for firm i in year t,
calculated as operating income minus operating cash flows. Both variables are scaled by the
lagged total assets and winsorized at the 1st and 99th percentiles. Desai and Dharmapala
(2006) propose that the book-tax difference (BTD) is partly attributable to earnings
management rather than tax avoidance. To remove the confounding effect of earnings
management, they use the residual (DD_BTD) estimated from Model (1) to proxy for the
degree of corporate tax avoidance. A higher DD_BTD represents a higher level of tax
avoidance.
3.1.3 Measurements of moderating variables. The moderating variable for the test of H2 is
financial constraints. It is measured by the SA index (Hadlock and Pierce, 2010) and defined
as follows:

SA index ¼ −0:737 3 size þ 0:043 3 size2 � 0:040 3 age (2)

where size is the natural logarithm of the book value of a firm’s total assets, and age is the
number of years for which the firm has been listed. A higher SA index indicates more severe
financial constraints. Though SA index is arguably more advantageous than KZ index and
WW index (Hadlock and Pierce, 2010), it is constructed based on US market, which may not
be applicable to the Chinese market. Thus, we also use cash dividends (Dividend) to proxy for
financial constraints (e.g. Denis and Sibilkov, 2010; Fazzari et al., 1988), given that financially
constrained firms tend to keep the funds they generate instead of paying dividends. Lower
cash dividends suggest tighter financial constraints.
The moderating variable involving the test of H3 is business risk. We measure it by the
volatility of return on sales (Std_ros) and that of return on assets (Std_roa). The moderating
variable used to test H4 relates to the region-level zodiac-year superstition. We utilize the
Baidu search index (http://index.baidu.com/) to measure the degree of zodiac-year belief in
the regions where firms are headquartered. The Baidu Index provides the search volume by
keywords put in the Baidu’s search engine at different regions and over different time
periods. The variable for the region-level superstition (SUP_Province) is calculated as the
moving average of daily volume of search for the keyword “zodiac year” for a province in a
year, which is scaled by the natural logarithm of the province’s family households [4]. We
select the top ten provincial regions that have the highest degree of zodiac-year belief Journal of
(SUP_Province) for each year and define them as the regions of a relatively high degree of Accounting
superstition. Other provincial regions are classified as having a relatively smaller extent of Literature
zodiac-year superstition. Lastly, the moderating variable for the test of H5 is the indicator
variable, SOE, which equals 1 if the firm’s ultimate controller is a central or local government
or a government-controlled enterprise, and 0 otherwise.
3.1.4 Control variables. We control for firm size (SIZE), financial leverage (LEV), capital
intensity (PPE and Intangible), research and development activities (RD), growth
opportunities (Asset_growth and MB), cash holdings (Cash), and foreign incomes
(Foreign), since previous studies document that these firm characteristics are associated
with tax avoidance (e.g. Chen et al., 2010; Guenther et al., 2019; Rego, 2003). We also control
for profitability (ROA) and net operating loss carry-forward (NOL) to account for the firm’s
need to avoid income taxes (e.g. Bradshaw et al., 2019; Chen et al., 2010; Rego, 2003). In
addition, we control for abnormal accruals (DA) since prior research finds a positive
relationship between financial reporting aggressiveness and tax avoidance (e.g. Frank et al.,
2009; McGuire et al., 2012a, b). Finally, we control for corporate governance variables and
other factors that prior studies find to be correlated with corporate tax avoidance (e.g.
Armstrong et al., 2015; Bauer, 2016; McGuire et al., 2012a, b); these variables include
managerial stock ownership (Managerial_shareholding), the largest shareholders’ stock
ownership (Top_shareholding), institutional shareholding (Institution), board size
(Boardsize), the duality of chairman and CEO (Duality), board independence (Indp), state
ownership (SOE), and big-4 audit (BIG4). The detailed definitions of all the control variables
are provided in Appendix 1.

3.2 Data sources and samples

Our main data sources are the China Stock Market and Accounting Research (CSMAR)
database and the Wind database. Our initial sample covers all the companies listed on the
Shanghai Stock Exchange or Shenzhen Stock Exchange for the years 2009–2019. To identify
the CEOs’ zodiac years, we need the data on their birthdays. To this end, we collect the names
of CEOs from the CSMAR database. We then manually collect the information about the
birthdays of CEOs via the Baidu search engine. Our initial sample consists of 7,531 unique
CEOs. We manage to obtain the data on the dates of birth for 397 CEOs, the birth years and
birth months, absent the dates of the months, for 5,139 CEOs, and the birth years without
birth months and birth dates for 1,995 CEOs for our sample firms. There are 343 CEOs whose
birth months coincide with the start month of Chinese lunar new year in their respective birth
years. With these data, we use the strategies, discussed previously in Section 3.1.1, to identify
the CEOs’ zodiac years for each firm-year observation.
We refine the sample through the following steps: (1) we exclude financial firms as
their financial characteristics are not comparable to those of non-financial firms; (2) we
remove firm-year observations with missing data on tax expense, with negative pre-tax
income, or with missing values in our measure of corporate tax avoidance; (3) we delete
observations without the information of CFOs’ zodiac years as well as those that are in
the CFOs’ zodiac years; (4) we eliminate observations in the year of CEO turnover; (5) we
rule out observations for which we fail to identify the CEOs’ zodiac years; (6) we delete
observations that have missing values in any of the variables covered in Section 3.1.4.
Appendix 2 summarizes our sample selection procedure. We end up with a sample
comprising 10,450 firm-year observations for 2,406 unique firms. We winsorize all
continuous variables at the top and bottom 1% points to mitigate the influence of outliers
on our multivariate analysis.
JAL 4. Research design and empirical results
4.1 Tests of the hypothesis H1
We use the following difference-in-differences (DID) ordinary least squares (OLS) regression
model to test our first hypothesis regarding the effect of zodiac-year belief on corporate tax
avoidance:

DD BTDi;t ¼ α0 þ α1 Treatedi þ α2 Treatedi 3 Tzodiact þ Controls þ ε (3)

The dependent variable, DD BTDi;t, is the residual book-tax difference estimated from
Model (1). The treatment indicator, Treatedi, is set equal to one for the treated firms and zero
for the control firms. The treated firms are those whose CEOs are in their zodiac years for the
current year but not for the year before or after the zodiac year. The control firms are those
whose CEOs are not in the zodiac year for the consecutive three years [5]. Tzodiact is the time
indicator variable, which equals 1 for the year when the treated firm’s CEO is in the zodiac
year, and equals 0 for the years before and after the CEO’s zodiac year. Inclusion of this time
variable in the interaction term, Treatedi 3 Tzodiact, helps control for the time-varying
factors that are common to both the treated firms and control firms. The coefficient of interest
to our hypothesis test is α2. It captures the change in corporate tax avoidance by the same
treated firms between the zodiac year and the year before or after the zodiac year, relative to
that by the control firms. If the hypothesis H1a (H1b) is tenable, we expect corporate tax
avoidance in the CEOs’ zodiac years to be of a higher (lower) degree than that in the adjacent
years. Thus, a significantly positive (negative) α2 would be consistent with our hypothesis
H1a (H1b). On top of the control variables mentioned in Section 3.1.4, year dummies and
industry dummies are also included in the regression, since the degree of corporate tax
avoidance varies substantially across years and industries (e.g. Becker et al., 1998; He et al.,
2020; Jones, 1991; Teoh et al., 1998). We do not include Tzodiac in the regression because this
time indicator variable is multicollinear with year dummies.
There are a disproportionately large sample of observations that are not in the zodiac
years, and thus we employ sample-matching technique to select the most comparable
observations for our regression estimation. We use the propensity-score matching (PSM) to
simulate the condition of random assignment of observations into the treatment and control
groups, thereby reducing potential sample selection bias and eliminating any potential
systematic differences between the two groups. To this end, we adopt the one-to-one nearest
neighborhood propensity-score matching approach. We match firms, which have CEOs in
their zodiac years (i.e. treated firms), with firms whose CEOs are not in their zodiac years (i.e.
control firms), based on the closest propensity scores derived from observable firm
characteristics. Since we have a panel of firm-year observations with staggered distributions
of zodiac years over the sample period, we do the propensity-score matching year by year,
and then pool all the yearly-matched sample to form the final matched sample for the period
2009–2019. In specific, for each year from 2010 to 2018, we first restrict the treatment group to
firms whose CEOs are in their zodiac years for the current year but not for the previous year
and following year, and the control group to firms whose CEOs are not in the zodiac year for
the consecutive three years. Then each treated firm is matched, without replacement, with a
control firm which has the propensity score closest to that of the treated firm. The propensity
scores are estimated using a logit regression, in which the binary variable (Zodiac_ceo),
indicating whether the firm’s CEO is in the zodiac year, is regressed on a set of covariates
which reflect the firm’s fundamental characteristics, including firm size (SIZE), return on
assets (ROA), assets growth (Asset_growth), financial leverage (LEV), firm risk (Std_return),
and board independence (Indp). Appendix 1 details the definitions of these variables. Panel A
of Table 1 reports the descriptive statistics of variables used for the propensity-score
matching. We also include industry dummies in the matching regression. After the
Panel A: Descriptive statistics of variables used for propensity-score matching
Variables n Mean Min. 25% Median 75% Max Std

DD_BTD 10,450 �0.000562 �0.141 �0.0156 �0.00227 0.0119 0.104 0.0314

Zodiac_ceo 10,450 0.0600 0 0 0 0 1 0.237
SIZE 10,450 22.40 19.87 21.53 22.23 22.40 26.14 1.252
ROA 10,450 0.0642 �0.000544 0.0334 0.0533 0.0824 0.222 0.0444
Asset_growth 10,450 0.451 0.0545 0.300 0.451 0.601 0.851 0.196
LEV 10,450 0.184 �0.248 0.0277 0.109 0.231 1.617 0.314
Std_return 10,450 0.431 0.113 0.297 0.389 0.513 1.152 0.203
Indp 10,450 0.372 0.250 0.333 0.333 0.429 0.571 0.0535
Note(s): This table presents descriptive statistics of the variables used for the propensity-score matching. The pre-matched sample contains 10,450 firm-years from
2009 to 2019. All the variables are defined in Appendix 1. All the continuous variables are winsorized at the 1 and 99% levels, respectively

Panel B: Univariate tests of covariate balance

No. of firm- No. of Mean for treatment Mean for control Standardized
Variables Matching statuses years firms firms firms bias t-stat

SIZE Unmatched sample in zodiac 2,789 1,219 22.35 22.444 �7.9 �1.15
year t
Matched sample in zodiac year t 470 415 22.35 22.325 2 0.22
Matched sample from t�1 to tþ1 1,410 415 22.349 22.323 2.2 0.24
ROA Unmatched sample in zodiac 2,789 1,219 0.0714 0.0698 3.3 0.5
year t
Matched sample in zodiac year t 470 415 0.0714 0.0721 �1.5 �0.16
Matched sample from t�1 to tþ1 1,410 415 0.713 0.721 �1.6 �0.18
Asset_ Unmatched sample in zodiac 2,789 1,219 0.189 0.18 3.7 0.54
growth year t
Matched sample in zodiac year t 470 415 0.189 0.177 5 0.53
Matched sample from t�1 to tþ1 1,410 415 0.189 0.175 5.6 0.61

(continued )
Propensity-score

Accounting
Literature

Journal of
Table 1.
matching
 Table 1.

JAL
Panel B: Univariate tests of covariate balance
No. of firm- No. of Mean for treatment Mean for control Standardized
Variables Matching statuses years firms firms firms bias t-stat

LEV Unmatched sample in zodiac 2,789 1,219 0.451 0.44 6.1 0.87
year t
Matched sample in zodiac year t 470 415 0.451 0.448 1.5 0.16
Matched sample from t�1 to tþ1 1,410 415 0.451 0.448 1.4 0.16
Std_return Unmatched sample in zodiac 2,789 1,219 0.408 0.419 �6.1 �0.85
year t
Matched sample in zodiac year t 470 415 0.408 0.412 �2.4 �0.26
Matched sample from t�1 to tþ1 1,410 415 0.408 0.412 �2.4 �0.26
Indp Unmatched sample in zodiac 2,789 1,219 0.371 0.371 0 �0.01
year t
Matched sample in zodiac year t 470 415 0.371 0.37 1.5 0.16
Matched sample from t�1 to tþ1 1,410 415 0.371 0.37 1.7 0.19
Note(s): This table reports the descriptive statistics of the matching covariates for the sample of treatment firms (i.e. firms that have CEOs in their zodiac years) and the
sample of control firms (i.e. firms whose CEOs are not in their zodiac years). We do the propensity-score matching year by year, and pool all the yearly-matched sample to
form the sample for covariate-balance tests. For each year from 2010 to 2018, we drop the control firms that have missing observations in the consecutive three-year
period centered in the zodiac year of the matched treatment firms. The results of the two-sample tests of mean differences, and the results of the standardized bias, for the
covariates are provided for the pre-matched and post-matched samples, respectively, in the zodiac year, and for the post-matched sample used in the difference-in-
differences regression analysis which covers the consecutive three-year period centered in the zodiac year of the matched control firm. All the covariates are defined in
Appendix 1. ***, **, * denote the two-tailed statistical significance at the 1%, 5%, and 10% levels, respectively

Panel C: Multivariate tests of covariate balance

Zodiac_ceo
Pre-matched sample in zodiac year t Post-matched sample in zodiac year t Post-matched sample from year t�1 to tþ1
Variables (1) (2) (3)

SIZE �0.056 �0.004 �0.003

(�0.74) (�0.04) (�0.04)
ROA 0.001 0.063 0.074
(0.00) (0.09) (0.16)

(continued )
Panel C: Multivariate tests of covariate balance
Zodiac_ceo
Pre-matched sample in zodiac year t Post-matched sample in zodiac year t Post-matched sample from year t�1 to tþ1
Variables (1) (2) (3)

Asset_growth 0.074 �0.606 �0.306

(0.04) (�0.25) (�0.18)
LEV 0.092 0.171 0.111
(0.31) (0.42) (0.39)
Std_return �0.293 �0.277 �0.459
(�0.57) (�0.40) (�0.79)
Indp 0.187 0.187 0.008
(0.14) (0.10) (0.01)
Constant �0.783 0.131 �1.538
(�0.44) (0.05) (�0.99)
Observations 2,770 470 1,314
Pseudo R2 0.032 0.003 0.013
Year-fixed effects included included included
Industry-fixed effects included included included
Note(s): This table reports the logistic regression result for comparing firm characteristics between the treatment sample (composed of firms whose CEOs are in their
zodiac years) and control sample (composed of firms whose CEOs are not in their zodiac years). The sample period for Zodiac_ceo in Columns (1) and (2) spans the years
2010–2018, while the sample period for Zodiac_ceo in Column (3) covers the years 2009–2019. Columns (1), (2), and (3) show the regression results based on the pre-
matched sample in the zodiac year, the post-matched sample in the zodiac year, and the post-matched sample used for the difference-in-differences (DID) regression,
respectively. The dependent variable, Zodiac_ceo, equals 1 if a firm’s CEO is in the zodiac year, and 0 otherwise. All the determinant variables for Zodiac_ceo (i.e.
matching covariates) are defined in Appendix 1. Year dummies and industry dummies are included in the regressions, but their results are not reported for the sake of
simplicity. t-statistics in parentheses are based on the robust standard errors clustered by firm. ***, **, * represent the two-tailed statistical significance at the 1%, 5%, and
10% levels, respectively
Source(s): Table by authors

Accounting
Literature

Journal of
Table 1.
JAL matching, we obtain a sample comprising 470 firm-year observations associated with 415
unique firms for the zodiac year t [6].
We next check whether the covariates are balanced between the treated firms and control
firms for the post-matched sample. To this end, we use the two-sample t-tests and absolute
standardized differences for the means of covariates to check whether observations with the
closest propensity scores have similar distribution of firm characteristics (Rosenbaum and
Rubin, 1983). Provided that the absolute standardized differences of the covariate means are
less than 5% or that t-statistics for the mean differences in the covariates are nonsignificant
post matching (D’Agostino, 1998), we may rest assured that the preexisting observed
differences between the treated firms and control firms would be eliminated substantially.
Panel B of Table 1 shows the results for the univariate check of covariate balance for the post-
matched sample. All the mean differences in the covariates are not statistically significant,
and the standardized bias is less than 5% for all the covariates. Furthermore, we run the logit
regression based on the pre-matched and post-matched samples, respectively, to further
check the covariate balance. Panel C of Table 1 reports the results. It is shown in Columns
(2) and (3) that the coefficients for all the covariates are not statistically significant,
suggesting that our post-matched sample achieves a covariate balance. Panel A of Table 2
reports the mean, standard deviation, minimum values, maximum values, and quartiles of all
variables, which are based on the post-matched sample used for the difference-in-differences
regression analysis. Panel B presents the Spearman correlation matrix among the variables.
The parallel trends assumption required of the difference-in-differences regression
analysis is that, in the absence of the treatment event, both the treated firms and control firms
exhibit similar trends in the outcome variable (Roberts and Whited, 2013). Before running the
difference-in-differences regression, it is necessary to test the parallel trends assumption to
avoid the possible confounding effects of other concurrent events on corporate tax avoidance
in the zodiac years. To this end, we define Pre as a time indicator equal to 1 (0) for the year
before (after) the zodiac year. We then use our post-matched sample along with the DID
regression to compare the change in tax avoidance during the pre- and post-zodiac years. In
Panel A of Table 3, the coefficient on Treated*Pre is not statistically significant. This null
result supports the assumption that the treated firms and control firms would have
experienced common trends in corporate tax avoidance in a counterfactual without the
zodiac year, thereby suggesting that the changes in tax avoidance, as indicated by our
baseline results, are likely attributed to the zodiac-year belief rather than other potential
omitted factors.
Panel B of Table 3 shows the regression results for the hypothesis H1. Column (1) reports
the OLS regression result. The coefficient on the interaction term, Treated*Tzodiac, is
positive (0.004) and statistically significant at the 5% level (t-stat. 5 2.01). It is possible that
some unobserved firm-specific factors happen by chance in the CEOs’ zodiac years to affect
corporate tax avoidance. To deal with this issue, we run firm-fixed-effects regression for
Model (3). We omit the treatment indicator variable and the industry dummies to avoid their
potential multicollinearity with the included firm-fixed effects. Instead, we interact industry
dummies with year dummies to control for potential industry shocks that can affect the
magnitude of tax avoidance across different industries and years. The firm-fixed-effects
regression result is shown in Column (2) under Panel B of Table 3. The coefficient for the
interaction term, Treated*Tzodiac, is significantly positive at the 5% level (0.004 with
t-stat. 5 2.20). The degree of corporate tax avoidance is higher by 12.74% of its standard
deviation in the CEOs’ zodiac years. Collectively, our results are both statistically and
economically significant in supporting the hypothesis H1a that CEOs are more likely to avoid
taxes in their zodiac years. CEOs in their zodiac years appear more averse to their firms’
financial risk than the legal and reputational risks associated with tax avoidance. Such
finding and inference are in line with some prior research which finds no evidence of
Panel A: Descriptive statistics of variables
Variables n Mean Min 25% Median 75% Max Std

DD_BTD 1,410 �0.001 �0.117 �0.017 �0.002 0.012 0.105 0.033

Adj_ETR 1,410 0.030 0 0 0 0.039 0.499 0.055
Ind_adj_ETR 1,410 �0.101 �1 �1 �1 0.125 8.914 1.559
Treated 1,410 0.500 0 0 0.5 1 1 0.500
Tzodiac 1,410 0.333 0 0 0 1 1 0.472
SIZE 1,410 22.330 19.890 21.490 22.200 22.980 25.790 1.200
ROA 1,410 0.071 �0.179 0.038 0.060 0.093 0.226 0.050
Asset_growth 1,410 0.179 �0.248 0.036 0.120 0.243 1.023 0.251
LEV 1,410 0.449 0.056 0.300 0.449 0.600 0.828 0.190
NOL 1,410 �0.021 �1.956 0 0 0 0 0.123
PPE 1,410 0.224 0.003 0.101 0.193 0.323 0.615 0.154
Intangible 1,410 0.044 0 0.016 0.031 0.054 0.362 0.055
RD 1,410 0.004 0 0 0 0 0.063 0.017
Capital_exp 1,410 0.050 0.000 0.017 0.039 0.070 0.194 0.044
MB 1,410 0.544 0.087 0.334 0.517 0.720 1.114 0.251
DA 1,410 0.012 �0.250 �0.029 0.009 0.049 0.207 0.073
Cash 1,410 0.162 0.010 0.091 0.139 0.210 0.498 0.099
Foreign 1,410 0.104 0 0 0.000 0.127 0.779 0.180
Managerial_shareholding 1,410 0.045 0 0 0.000 0.014 0.534 0.110
Top_shareholding 1,410 34.200 9.030 23.340 32.830 43.760 71.150 14.060
Boardsize 1,410 2.168 1.609 2.079 2.197 2.197 2.639 0.185
Duality 1,410 0.247 0 0 0 0 1 0.431
Institution 1,410 47.590 0.108 34.960 49.700 63.780 87.980 21.490
Indp 1,410 0.370 0.250 0.333 0.333 0.400 0.571 0.053
BIG4 1,410 0.055 0 0 0 0 1 0.229
SOE 1,410 0.441 0 0 0 1 1 0.497
Note(s): This table tabulates descriptive statistics of the variables used for the difference-in-differences regression tests. The sample consists of a post-matched sample of 1,410 firm-years and covers the years
2009–2019. All the variables are defined in Appendix 1. All the continuous variables are winsorized at the 1 and 99% levels, respectively
(continued )
Univariate statistics

Accounting
Literature

Journal of
Table 2.
 Table 2.

JAL
Panel B: Correlation matrix
Variables DD_BTD SIZE ROA Asset_growth LEV NOL PPE Intangible RD Capital_exp

DD_BTD 1
SIZE �0.036 1
ROA 0.153*** 0.078*** 1
Asset_growth �0.007 0.149*** 0.122*** 1
LEV �0.177*** 0.471*** �0.322*** 0.088*** 1
NOL 0.01 �0.098*** 0.071*** 0.004 �0.065** 1
PPE 0.038 �0.073*** �0.033 �0.159*** �0.049* �0.079*** 1
Intangible �0.085*** �0.044* 0.031 �0.003 �0.038 0.016 0.094*** 1
RD 0.126*** 0.004 0.016 �0.058** �0.108*** 0.009 �0.070*** �0.026 1
Capital_exp 0.008 0 0.139*** 0.152*** �0.045* 0.014 0.386*** 0.187*** �0.042 1
MB �0.091*** 0.559*** �0.285*** �0.097*** 0.550*** �0.065** 0.107*** �0.021 0.014 �0.019
DA 0.154*** 0.146*** 0.220*** 0.149*** �0.056** 0.061** �0.145*** �0.121*** 0.038 �0.076***
Cash �0.041 �0.121*** 0.212*** 0.100*** �0.273*** 0.013 �0.272*** �0.084*** �0.029 �0.089***
Foreign 0.062** �0.008 0.055** 0.003 �0.064** �0.019 0.032 �0.006 0.013 0.117***
Managerial_shareholding 0.012 �0.191*** 0.024 0.04 �0.242*** 0.048* �0.099*** �0.012 0.107*** 0.016
Top_shareholding �0.026 0.207*** 0.026 0.056** 0.213*** �0.028 �0.049* �0.035 �0.110*** �0.033
Boardsize �0.051* 0.170*** 0.090*** �0.036 0.071*** �0.054** 0.208*** 0.02 �0.031 0.097***
Duality 0.075*** �0.072*** �0.006 0.029 �0.117*** 0.027 �0.110*** �0.04 0.082*** �0.04
Institution 0.032 0.364*** 0.198*** 0.133*** 0.226*** �0.047* 0.046* 0.046* �0.132*** 0.076***
Indp 0.012 0.120*** �0.028 0.022 0.062** �0.035 �0.057** �0.062** �0.036 0.006
BIG4 �0.050* 0.394*** 0.015 �0.002 0.147*** �0.080*** �0.024 0.006 �0.016 0.046*
SOE 0.000 0.182*** �0.158*** �0.095*** 0.264*** �0.068** 0.103*** 0.04 �0.115*** �0.023

Variables MB DA Cash Foreign Managerial_shareholding Top_shareholding Boardsize Duality Institution Indp BIG4 SOE

MB 1
DA �0.02 1
Cash �0.243*** �0.063** 1
*
Foreign �0.046 �0.015 0.142*** 1
***
Managerial_shareholding �0.236 0.032 0.003 0.072*** 1
***
Top_shareholding 0.088 0.038 �0.026 �0.029 �0.063** 1
*** *
Boardsize 0.109 �0.007 �0.046 0.028 �0.200*** �0.062** 1
*** *** ***
Duality �0.158 0.031 �0.002 0.143 0.481 �0.039 �0.315*** 1
*** ** *** ***
Institution 0.124 0.036 0.057 �0.009 �0.559 0.530 0.170*** �0.260*** 1
Indp 0.056** 0.001 �0.041 �0.020 0.063** 0.145*** �0.411*** 0.137*** 0.037 1
BIG4 0.191*** �0.030 �0.073*** �0.038 �0.095*** 0.177*** 0.099*** �0.081*** 0.221*** 0.073*** 1
*** * *** *** *** *** *** ***
SOE 0.264 �0.010 �0.043 �0.069 �0.345 0.178 0.234 �0.329 0.306 0.003 0.154*** 1
Note(s): This table presents the results for the Spearman correlations. The correlation matrix involves the variables used for the difference-in-differences regression tests. The sample consists of a post-
matched sample of 1,410 firm-years and covers the years 2009–2019. All the variables are defined in Appendix 1. ***, **, * represent the two-tailed statistical significance at the 1%, 5%, and 10% levels,
respectively, for the correlation coefficients
Source(s): Table by authors
 Journal of
Panel A: Tests of parallel trends assumption Accounting
DD_BTD
Variables (1)
Literature

Treated*Pre 0.002
(0.87)
Treated �0.001
(�0.20)
SIZE �0.000
(�0.20)
ROA 0.094**
(2.35)
Asset_growth �0.001
(�0.20)
LEV �0.028**
(�2.54)
NOL �0.006
(�0.98)
PPE 0.004
(0.41)
Intangible �0.079***
(�2.99)
RD 0.214**
(2.41)
Capital_exp �0.022
(�0.83)
MB 0.004
(0.40)
DA 0.050***
(2.62)
Cash �0.048***
(�3.08)
Foreign 0.011
(1.57)
Managerial_shareholding �0.006
(�0.35)
Top_shareholding �0.000
(�0.72)
Boardsize �0.017*
(�1.78)
Duality 0.007**
(1.98)
Institution 0.000*
(1.83)
Indp �0.011
(�0.42)
BIG4 �0.005
(�0.82)
SOE 0.004
(1.52)
Constant 0.099**
(2.44)
Observations 940 Table 3.
Adj R2 0.190 Baseline regression
analysis for the
(continued ) hypothesis H1
JAL Panel A: Tests of parallel trends assumption
DD_BTD
Variables (1)

Year-fixed effects included

Industry-fixed effects included
Note(s): This table presents the results for the multivariate test of the parallel trends assumption. The sample
period spans the years 2009–2019. All the variables are defined in Appendix 1. The dependent variable is the
residual book-tax difference (DD_BTD) which are estimated from a regression of the total book-tax difference
on the total accruals. The treatment variable, Treated, equals 1 for the treatment firm, and 0 for the control
firm. The treatment (control) firms are firms whose CEOs are (not) in their zodiac years. Pre equals 1 (0) for the
year before (after) the zodiac year. Year dummies and industry dummies are included in the regression, but
their results are not reported for simplicity. t-statistics in parentheses are based on robust standard errors
clustered by firm. *, **, and *** indicate the two-tailed statistical significance at the 10%, 5%, and 1% levels,
respectively

Panel B: Baseline difference-in-differences regression on the zodiac-year effect

DD_BTD
OLS regression Firm-fixed-effects regression
Variables (1) (2)

Treated*Tzodiac 0.004** 0.004**

(2.01) (2.20)
Treated 0.001
(0.52)
SIZE �0.000 0.006
(�0.03) (0.87)
ROA 0.080** 0.007
(2.13) (0.19)
Asset_growth �0.002 �0.004
(�0.29) (�0.64)
LEV �0.035*** �0.046**
(�3.48) (�2.38)
NOL �0.004 0.012**
(�0.59) (2.18)
PPE 0.002 �0.000
(0.27) (�0.01)
Intangible �0.080*** 0.037
(�2.94) (0.71)
RD 0.208*** 0.074
(2.96) (0.82)
Capital_exp �0.009 �0.035
(�0.38) (�1.03)
MB 0.007 �0.014
(0.82) (�1.18)
DA 0.039** 0.044***
(2.34) (2.70)
Cash �0.049*** 0.027
(�3.46) (1.45)
Foreign 0.014** 0.005
(2.00) (0.26)
Managerial_shareholding �0.004 �0.006
(�0.29) (�0.24)
Top_shareholding �0.000 0.000
(�0.93) (0.19)

Table 3. (continued )
 Journal of
Panel B: Baseline difference-in-differences regression on the zodiac-year effect Accounting
DD_BTD Literature
OLS regression Firm-fixed-effects regression
Variables (1) (2)

Boardsize �0.017** 0.013

(�1.98) (0.81)
Duality 0.006* �0.010
(1.74) (�1.54)
Institution 0.000* �0.000
(1.68) (�0.79)
Indp �0.012 0.027
(�0.47) (0.55)
BIG4 �0.006 0.004
(�1.11) (0.68)
SOE 0.005* �0.004
(1.75) (�0.45)
Constant 0.094** �0.131
(2.27) (�0.88)
Observations 1,410 1,410
Adj R2 0.168 0.157
Year-fixed effects included
Industry-fixed effects included
Firm-fixed effects included
Industry-year fixed effects included
Note(s): This table reports the results of difference-in-differences regression analysis of the effect of CEOs’
zodiac years on corporate tax avoidance. The sample period ranges from 2009 to 2019. All the variables are
defined in Appendix 1. The dependent variable is the residual book-tax difference (DD_BTD) which is
estimated from a regression of the total book-tax difference on the total accruals. The treatment variable,
Treated, equals 1 for a treatment firm, and 0 for a control firm. The treatment (control) firms are firms whose
CEOs are (not) in their zodiac years. The time indicator variable, Tzodiac, equals 1 for each year from 2010 to
2018 when the treatment firm’s CEO is in her/his zodiac year, and 0 for the preceding year and the year after the
zodiac year. The interaction term, Treated*Tzodiac, is the variable of interest for the main hypothesis. All the
variables are defined in Appendix 1. Column (1) reports the OLS regression result. Industry dummies and year
dummies are included in the regression, but their results are not reported for brevity. Column (2) reports the
firm-fixed-effects regression result. Firm-fixed effects and industry-year interacted dummies are included in
the regression, but their results are not reported for simplicity. t-statistics in parentheses are based on robust
standard errors clustered by firm. *, **, and *** indicate the two-tailed statistical significance at the 10%, 5%,
and 1% levels, respectively
Source(s): Table by authors Table 3.

significant reputational costs to firms for pursuing tax avoidance (Gallemore et al., 2014), and
reconcile with the view that the tax enforcement by Chinese tax authorities is inadequate
(Brondolo and Zhang, 2016). In addition, we conduct a test of variance inflation factors (VIF).
The un-tabulated results show that the VIF values for all independent variables are less than
5, suggesting that multicollinearity does not pose a threat to our regression analysis.

4.2 Further analyses for the hypothesis H1

We conduct further analyses for the hypothesis H1 in the following. First, to reinforce the
treatment effect of zodiac-year belief on corporate tax avoidance, we conduct a placebo test,
where we re-do our baseline regression with a pseudo-treatment sample. Specifically, we
randomly select a number of firms, which equals the number of treatment firms, from the
unmatched control group to create a pseudo-treatment group for each year. These selected
JAL samples are merged with the previously matched control group. We then use this combined
sample to run the DID regression model (3). We repeat this process for 1,000 times. As shown
in Figure 1, the coefficients for the placebo DID estimator are normally distributed and
concentrated around zero. The majority of them have p-values greater than the significance
level of 0.1 and are located to the left of the baseline DID coefficient (0.004 as depicted by the
dotted vertical line). These results indicate that the effect of zodiac superstition on corporate
tax avoidance vanishes after the randomization and placebo processes, which in reverse,
corroborates the treatment effect implied by our baseline results.
Second, the standard errors of the coefficients in our baseline results are clustered by firm.
We also cluster the standard errors by industry to account for potential cross-correlations of
regression residuals within each industry, and report the results in Panel A of Table 4. They
are qualitatively the same as those reported in Panel B of Table 3.
Lastly, we check the robustness of our baseline results to using an alternative measure of
corporate tax avoidance that involves the effective tax rate (e.g. Bradshaw et al., 2019; Chen
et al., 2010; Dyreng et al., 2010). The China’s local governments generally employ tax
incentives to encourage investments, leading to effective tax rates lower than the statutory
tax rate (e.g. He, 2016; Shevlin et al., 2012; Wu et al., 2007). Therefore, we measure corporate
tax avoidance by taking the difference between the nominal tax rate and the effective tax rate

Placebo test
1

250
0.8

200
Kernel density estimation
0.6

150
P value
0.4

100
0.2

50
0

–0.005 –0.004 –0.003 –0.002 –0.001 0 0.001 0.002 0.003 0.004 0.005
Coefficient

Note(s): The X-axis indicates the coefficients for the interaction term that are estimated
based on the pseudo sample. The left Y-axis indicates the p-values of the coefficients. The
right Y-axis indicates the kernel density of the estimated coefficients. The red dots
Figure 1.
(solid curve) represent(s) the p-values (kernel density) corresponding to the estimated
Distribution of the coefficients. The vertical dashed line represents the estimated coefficient of the interaction
1,000 coefficient term in the baseline DID regression analysis. The horizontal dashed line represents the
estimates in a significance level of p = 0.1 for a DID estimator
placebo test
Source(s): Figure by authors
 Journal of
Panel A: Regression results based on standard errors clustered by industry Accounting
DD_BTD
OLS regression Firm-fixed-effects regression
Literature
Variables (1) (2)

Treated*Tzodiac 0.004** 0.004***

(2.29) (4.19)
Treated 0.001
(0.63)
SIZE �0.000 0.006*
(�0.04) (1.90)
ROA 0.080*** 0.007
(3.75) (0.41)
Asset_growth �0.002 �0.004
(�0.42) (�0.88)
LEV �0.035*** �0.046***
(�8.93) (�3.74)
NOL �0.004* 0.012***
(�1.92) (5.04)
PPE 0.002 �0.000
(0.48) (�0.02)
Intangible �0.080* 0.037
(�2.16) (1.20)
RD 0.208*** 0.074*
(5.45) (1.84)
Capital_exp �0.009 �0.035*
(�0.27) (�2.14)
MB 0.007 �0.014
(1.06) (�1.72)
DA 0.039** 0.044*
(2.50) (1.96)
Cash �0.049*** 0.027
(�4.42) (1.19)
Foreign 0.014** 0.005
(2.56) (0.59)
Managerial_shareholding �0.004 �0.006
(�0.83) (�0.24)
Top_shareholding �0.000 0.000
(�1.34) (0.25)
Boardsize �0.017* 0.013
(�1.82) (0.83)
Duality 0.006*** �0.010
(3.97) (�1.61)
Institution 0.000*** �0.000
(5.95) (�0.71)
Indp �0.012 0.027
(�0.50) (0.47)
BIG4 �0.006*** 0.004**
(�4.06) (3.03)
SOE 0.005** �0.004
(2.97) (�0.26)
Constant 0.094*** �0.132
(3.96) (�1.47)
Observations 1,410 1,410
Adj R2 0.141 0.089
Year-fixed effects included Table 4.
Further analyses for
(continued ) the hypothesis H1
JAL Panel A: Regression results based on standard errors clustered by industry
DD_BTD
OLS regression Firm-fixed-effects regression
Variables (1) (2)

Industry-fixed effects included

Firm-fixed effects included
Industry-year fixed effects included
Note(s): This table reports the results of the DID regression analysis of the effect of CEOs’ zodiac years on
corporate tax avoidance clustering, with the coefficients’ standard errors clustered by industry. The sample
period ranges from 2009 to 2019. All the variables are defined in Appendix 1. The dependent variable is the
residual book-tax difference (DD_BTD) which is estimated from a regression of the total book-tax difference
on the total accruals. The treatment variable, Treated, equals 1 for a treatment firm, and 0 for a control firm.
The treatment (control) firms are firms whose CEOs are (not) in their zodiac years. The time indicator variable,
Tzodiac, equals 1 for each year from 2010 to 2018 when the treatment firm’s CEO is in her/his zodiac year, and
0 for the preceding year and the year after the zodiac year. The interaction term, Treated*Tzodiac, is the
variable of interest for the main hypothesis. All the variables are defined in Appendix 1. Column (1) reports the
OLS regression result. Industry dummies and year dummies are included in the regression, but their results
are not reported for brevity. Column (2) reports the firm-fixed-effects regression result. Firm-fixed effects and
industry-year interacted dummies are included in the regression, but their results are not reported for
simplicity. t-statistics in parentheses are based on robust standard errors clustered by industry. *, **, and ***
indicate the two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively

Panel B: Regression results based on alternative measures of corporate tax avoidance

Adj_ETR Ind_adj_ETR
OLS Firm-fixed-effects OLS Firm-fixed-effects
regression regression regression regression
Variables (1) (2) (3) (4)

Treated*Tzodiac 0.008** 0.008** 0.248*** 0.247**

(2.22) (1.98) (2.61) (2.53)
Treated 0.000 0.068
(0.02) (0.62)
SIZE 0.000 0.001 0.004 0.119
(0.12) (0.13) (0.05) (0.50)
ROA 0.029 0.085 0.877 1.099
(0.61) (1.37) (0.67) (0.78)
Asset_growth 0.002 0.001 �0.016 0.122
(0.30) (0.19) (�0.10) (0.78)
LEV �0.040** �0.005 �0.994* �0.479
(�2.22) (�0.17) (�1.89) (�0.56)
NOL �0.016 0.026 �0.442 0.371
(�1.21) (1.56) (�1.18) (1.31)
PPE �0.003 �0.049 �0.408 �0.978
(�0.20) (�1.01) (�0.88) (�0.79)
Intangible �0.085** �0.031 �3.304*** �1.070
(�1.99) (�0.52) (�2.68) (�0.60)
RD 0.061 �0.149 �0.431 �7.395
(0.58) (�0.76) (�0.18) (�1.50)
Capital_exp �0.042 �0.083 �0.895 �1.619
(�1.01) (�1.49) (�0.79) (�1.03)
MB 0.015 �0.015 0.302 �0.373
(1.24) (�0.77) (0.82) (�0.59)
DA 0.026 0.034 0.984 1.139*
(1.13) (1.59) (1.41) (1.72)

Table 4. (continued )
 Journal of
Panel B: Regression results based on alternative measures of corporate tax avoidance Accounting
Adj_ETR Ind_adj_ETR Literature
OLS Firm-fixed-effects OLS Firm-fixed-effects
regression regression regression regression
Variables (1) (2) (3) (4)

Cash �0.082*** 0.018 �2.578*** 0.300

(�4.06) (0.83) (�4.06) (0.42)
Foreign 0.019 0.003 0.549* 0.161
(1.46) (0.16) (1.66) (0.29)
Managerial_ �0.038 0.016 �1.176* 0.892
shareholding (�1.55) (0.40) (�1.72) (0.70)
Top_shareholding 0.000 �0.000 0.001 �0.015
(0.48) (�0.92) (0.16) (�1.25)
Boardsize �0.038*** �0.029 �1.179*** �1.210*
(�3.25) (�1.34) (�3.35) (�1.79)
Duality 0.009 �0.034 0.324* �0.616
(1.58) (�1.62) (1.83) (�1.33)
Institution �0.000 �0.000 0.001 �0.013
(�0.06) (�0.81) (0.31) (�1.52)
Indp �0.029 �0.031 �1.162 �1.689
(�0.71) (�0.58) (�1.05) (�0.89)
BIG4 �0.008 0.011 �0.158 0.297
(�1.16) (0.53) (�0.66) (0.50)
SOE 0.009* �0.022 0.328** �0.431
(1.94) (�1.15) (2.32) (�1.02)
Constant 0.261*** 0.167 3.577* 3.028
(3.24) (0.79) (1.85) (0.53)
Observations 1,410 1,410 1,410 1,410
Adj R2 0.104 0.056 0.063 0.055
Year-fixed effects included included
Industry-fixed effects included included
Firm-fixed effects included included
Industry-year fixed included included
effects
Note(s): This table reports the results of difference-in-differences regression analysis of the effect of the
CEOs’ zodiac-year belief on corporate tax avoidance, which are based on alternative measures of tax
avoidance. The sample period spans the years 2009–2019. All the variables are defined in Appendix 1. In
Columns (1) and (2), the dependent variable, Adj_ETR, is the difference between nominal tax rate and effective
tax rate (ETR). ETR is calculated as tax expense minus deferred tax expense, deflated by profit before tax. In
Columns (3) and (4), the dependent variable, Ind_adj_ETR, is the difference between a firm’s Adj_ETR and the
industry average of Adj_ETR for the same year, divided by this industry average. The treatment variable,
Treated, equals 1 for a treatment firm, and 0 for a control firm. The treatment (control) firms are firms whose
CEOs are (not) in their zodiac years. The time indicator variable, Tzodiac, equals 1 for each year from 2010 to
2018 when the treatment firm’s CEO is in her/his zodiac year, and 0 for the preceding year and the year after the
zodiac year. The interaction term, Treated*Tzodiac, is the variable of interest. Columns (1) and (3) reports the
OLS regression result. Industry dummies and year dummies are included in the regression, but their results
are not reported for brevity. Columns (2) and (4) reports the firm-fixed-effects regression result. Firm-fixed
effects and industry-year interacted dummies are included in the regression, but their results are not reported
for simplicity. t-statistics in parentheses are based on robust standard errors clustered by firm. *, **, and ***
indicate the two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively
Source(s): Table by authors Table 4.
JAL (Adj_ETR) (Amiram et al., 2019; Cen et al., 2017a, b; Chan et al., 2016; Tang, 2020). We also
follow Balakrishnan et al. (2019) to derive an industry-adjusted measure of effective tax rate
(Ind_adj_ETR). The higher Adj_ETR and Ind_adj_ETR, the greater the tax avoidance. We
substitute DD_BTD for Adj_ETR or Ind_adj_ETR in Model (3) and re-estimate the
regression. Panel B of Table 4 shows the results, which elicit the same inferences as do our
baseline results and are both statistically and economically significant.

4.3 Tests of the hypotheses H2-H5

To test the hypothesis H2, we partition the pre-matched sample into subsamples with tighter
financial constraints and those with lower financial constraints, based on the medians of SA_
index and Dividend, respectively. We then redo the propensity-score matching and the DID
regression for each subsample. Columns (1) and (3) of Table 5 display the results for the
subsample firms that confront relatively severe financial constraints, as indicated by higher
SA_index and lower Dividend. The coefficients on the interaction term amount to 0.006
(t-stat. 5 1.97) and 0.005 (t-stat. 5 2.11), indicating that the degree of tax avoidance is higher
in the CEOs’ zodiac years in cases when the firms are in the financial constraints. But
regarding the results in Columns (2) and (4) for the low-financial-constraints subsample,
Treated*Tzodiac does not take on a statistically significant coefficient. The hypothesis H2 is
thus supported.
To test the hypothesis H3, we divide our pre-matched sample into two subsamples
based on the medians of Std_ros and Std_roa, respectively. Then for each subsample, we
match each treatment firms, without replacement, with a control firm utilizing the same
propensity-score-matching approach as described previously, and re-run the difference-
in-differences regression. Table 6 presents the results. The coefficients of the interaction
term, Treated*Tzodiac, are significantly positive at the 5% level for the high-risk firms,
whereas the coefficients are not statistically significant for the low-risk firms. These
results are consistent with H3, indicating that the positive relationship between zodiac-
year belief and tax avoidance is evident only among firms with relatively higher
business risk.
To test the hypothesis H4, we re-run our matching process and DID regression based
on the subsample of firms that are headquartered in the more-superstitious regions,
which are defined in Section 3.1.3, and those in the other regions, respectively. Table 7
provides the regression results. For the subsample of firms headquartered in the highly
superstitious provinces, the interaction term, Treated*Tzodiac, is positive and
significantly associated with tax avoidance (0.005 with t-stat. 5 2.06), implying that
for these provinces, the magnitude of corporate tax avoidance is greater in the CEOs’
zodiac years compared to the years before and after the zodiac years. However, for the
low-regional-superstition subsample, the coefficient of the interaction term is not
statistically significant (0.005 with t-stat. 5 1.18). These subsample regression results
are therefore consistent with H4 - that the positive link between tax avoidance and the
CEOs’ zodiac years is more pronounced for firms headquartered in highly superstitious
regions.
To test the hypothesis H5, we split the pre-matched sample into the SOE subsample
and non-SOE subsample, re-do the matching for each subsample, and then run
subsample regressions based on Model (3). As shown in Table 8, the coefficient of
Treated*Tzodiac is significantly positive at the 1% level for non-SOEs (0.008 with
t-stat. 5 �2.97) whereas the coefficient is not statistically significant for SOEs (�0.001
with t-stat. 5 �0.43). These results lend support to H5 and reconcile with Fisman et al.
(2023) which find that the effect of zodiac-year belief on corporate investments exists only
in non-state-owned firms.
 Journal of
DD_BTD Accounting
Tighter financial Lower financial Tighter financial Lower financial
constraints constraints constraints constraints
Literature
(high SA_index) (low SA_index) (low Dividend) (high Dividend)
Variables (1) (2) (3) (4)

Treated*Tzodiac 0.006* 0.001 0.005** �0.001

(1.97) (0.42) (2.11) (�0.33)
SIZE �0.001 0.009 0.004 �0.024*
(�0.05) (0.65) (0.26) (�1.66)
ROA 0.013 �0.051 �0.046 �0.028
(0.25) (�1.04) (�0.72) (�0.36)
Asset_growth �0.001 0.003 0.004 0.003
(�0.18) (0.51) (0.46) (0.51)
LEV �0.010 �0.086** 0.004 0.012
(�0.43) (�2.33) (0.13) (0.31)
NOL 0.020* �0.013** �0.002 0.017
(1.92) (�2.42) (�0.46) (0.48)
PPE �0.038 �0.015 �0.008 0.034
(�1.04) (�0.46) (�0.18) (0.78)
Intangible 0.004 0.018 �0.022 0.028
(0.03) (0.15) (�0.31) (0.28)
RD 0.060 0.202 0.069 0.249
(0.67) (1.44) (0.67) (1.49)
Capital_exp �0.044 �0.034 �0.060 �0.102
(�0.93) (�0.51) (�1.14) (�1.21)
MB 0.021 �0.034** �0.014 �0.017
(1.29) (�2.10) (�0.61) (�0.92)
DA 0.020 0.038 �0.030 0.058
(0.88) (1.57) (�1.42) (1.62)
Cash 0.042* �0.000 �0.013 0.066**
(1.89) (�0.10) (�0.55) (2.19)
Foreign 0.011 0.027 0.021 �0.011
(0.47) (1.27) (0.90) (�0.45)
Managerial_ 0.029 0.037 �0.023 �0.002
shareholding (0.89) (0.34) (�0.78) (�0.02)
Top_shareholding 0.000* 0.000 0.000 �0.000
(1.72) (0.44) (0.47) (�0.17)
Boardsize 0.021 �0.033 �0.022 0.058
(0.93) (�0.66) (�0.98) (1.45)
Duality �0.015* �0.009 0.007 0.009
(�1.88) (�0.72) (0.94) (0.55)
Institution �0.000 �0.000 �0.000 �0.000
(�0.71) (�0.32) (�0.54) (�0.05)
Indp 0.013 �0.072 �0.020 0.067
(0.20) (�0.62) (�0.37) (0.61)
BIG4 0.007 �0.007 �0.002 0.008
(1.24) (�0.56) (�0.22) (0.50)
SOE �0.004 0.002 0.005 0.020**
(�0.81) (0.16) (0.62) (2.26)
Constant �0.046 �0.023 �0.028 0.401
(�0.21) (�0.08) (�0.09) (1.34) Table 5.
Observations 623 509 382 492 Tests of the
moderating effect of
(continued ) financial constraints
JAL DD_BTD
Tighter financial Lower financial Tighter financial Lower financial
constraints constraints constraints constraints
(high SA_index) (low SA_index) (low Dividend) (high Dividend)
Variables (1) (2) (3) (4)

Adj R2 0.231 0.269 0.190 0.261

Firm-fixed effects included included included included
Industry-year fixed included included included included
effects
Note(s): This table reports the results from testing the moderating effect of financial constraints on the
baseline regression results. The sample period covers the years 2009–2019. All the variables are defined in
Appendix 1. The dependent variable is the residual book-tax difference (DD_BTD) which is estimated from a
regression of the total book-tax difference on the total accruals. The moderating variables are the SA index
(SA_index) and cash dividend (Dividend). A higher (lower) value of SA_index (Dividend) indicates that the firm
faces tighter financial constraints. The difference-in-differences regressions are run separately in the high-SA_
index (low-Dividend) subsample and the low-SA_index (high-Dividend) subsample, which are split based on
the median of SA_index (Dividend) in the pre-matched sample. Treated equals 1 for the treatment firm (i.e. a
firm whose CEO is in her/his zodiac year), and 0 for the control firm (i.e. a firm whose CEO is not in her/his
zodiac year). The time indicator variable, Tzodiac, equals 1 for each year from 2010 to 2018 when the treatment
firm’s CEO is in her/his zodiac year, and 0 for the preceding year and the year after the zodiac year. The
interaction term, Treated*Tzodiac, is the variable of interest. Firm-fixed effects and industry-year interacted
dummies are included in the regression, but their results are not reported for simplicity. t-statistics in
parentheses are based on robust standard errors clustered by firm. *, **, and *** indicate the two-tailed
statistical significance at the 10%, 5%, and 1% levels, respectively
Table 5. Source(s): Table by authors

5. Conclusion
The literature on psychology and sociology documents that superstition prevails
astonishingly in the modern societies, influencing the attitudes and decisions of people in
their daily lives. In this paper, we examine the role played by zodiac-year belief in shaping
corporate tax avoidance. We postulate that CEOs trade-off between the benefits and costs of
pursuing tax avoidance. The benefits (costs) are associated with the decreased (increased)
risk of financial constraints and/or distress (legal penalties and/or reputational losses). The
zodiac-year belief provides a nice setting to investigate whether CEOs perceive the benefits
as exceeding the costs when they are risk-averse. Under the zodiac-year belief, individuals
are likely to confront misfortune in their zodiac years and tend to be more cautious and risk-
averse in their decision-making and behaviors. Our empirical results indicate that CEOs are
more likely to avoid taxes in their zodiac years, suggesting that they attach more importance
to containing financial risk than the exposure of legal and reputational risks as a result of tax
avoidance. We also find that the effect of zodiac-year belief on corporate tax avoidance is
stronger in cases in which (1) the firm faces tight financial constraints; (2) the firm’s business
risk is high; (3) the firm is headquartered in the highly superstitious areas; and (4) the firm is
non-state-owned. It is worth noting that our findings are not generalizable to other types of
superstition, which is a limitation of the paper. Yet, as mentioned in Section 1, focusing on
zodiac-year superstitious belief for the study facilitates us to draw casual inferences and
enhance the internal validity of results.
Our paper has important implications for practitioners. From our baseline results, it can
be inferred that, to curb corporate tax avoidance, we ought to increase the legal and
reputational risks to firms for their tax avoidance. This emphasizes the need for
policymakers to improve tax regulations, for tax authorities to enhance tax law
enforcements, and for the media and market participants to increase coverage and
 Journal of
DD_BTD
High risk Low risk High risk Low risk Accounting
(Std_ros) (Std_ros) (Std_roa) (Std_roa) Literature
Variables (1) (2) (3) (4)

Treated*Tzodiac 0.010** 0.001 0.008* 0.000

(2.36) (0.27) (1.78) (0.02)
SIZE 0.003 �0.006 0.010 �0.007
(0.26) (�0.68) (0.74) (�0.75)
ROA 0.031 0.074 0.065 �0.202
(0.93) (0.97) (1.48) (�1.55)
Asset_growth 0.005 0.010 �0.014 0.016**
(0.62) (1.39) (�1.15) (2.51)
LEV �0.003 �0.002 �0.011 �0.008
(�0.12) (�0.12) (�0.37) (�0.36)
NOL 0.001 �0.128 0.001 2.210
(0.09) (�1.34) (0.13) (1.25)
**
PPE �0.024 0.073 0.010 0.030
(�0.68) (2.31) (0.23) (1.04)
*
Intangible 0.131 0.104 �0.016 �0.037
(1.69) (1.33) (�0.21) (�0.72)
RD �0.212* 0.301*** �0.074 0.231
(�1.69) (2.71) (�0.66) (1.62)
Capital_exp �0.034 0.065 0.066 0.025
(�0.62) (1.16) (1.17) (0.74)
* *
MB �0.040 �0.021 0.007 �0.034**
(�1.94) (�1.76) (0.32) (�2.34)
DA 0.040 0.013 0.045 �0.005
(1.57) (0.85) (1.50) (�0.31)
*
Cash 0.013 0.046 0.012 0.000
(0.39) (1.73) (0.32) (1.37)
Foreign 0.049 0.013 0.037 �0.011
(0.92) (0.56) (1.05) (�0.56)
Managerial_shareholding 0.028 0.008 �0.007 0.009
(0.25) (0.30) (�0.04) (0.26)
Top_shareholding �0.000 �0.000 0.000 �0.000*
(�0.63) (�0.05) (1.01) (�1.89)
Boardsize 0.028 �0.009 �0.048*** �0.002
(0.96) (�0.65) (�2.83) (�0.12)
Duality �0.006 0.003 0.002 �0.003
(�0.43) (0.47) (0.11) (�0.23)
*
Institution 0.001 �0.000 0.000 �0.000
(1.92) (�1.05) (0.74) (�0.88)
Indp 0.105 �0.015 �0.185*** 0.028
(1.43) (�0.49) (�3.04) (0.64)
BIG4 0.007 0.006 �0.004
(0.82) (1.19) (�0.33)
*** ***
SOE �0.086 �0.001 �0.091 �0.000
(�4.87) (�0.39) (�6.10) (�0.01)
Constant �0.124 0.138 �0.061 0.210
(�0.43) (0.70) (�0.21) (0.95)
Observations 473 624 490 510
AdjR2 0.431 0.200 0.195 0.0802
Firm-fixed effects included included included included
Industry-year fixed effects included included included included
Note(s): This table reports the results from testing the moderating effect of business risk on the baseline regression results. The sample
period covers the years 2009–2019. All the variables are defined in Appendix 1. The dependent variable is the residual book-tax
difference (DD_BTD) which is estimated from a regression of the total book-tax difference on the total accruals. The moderating
variables are the standard deviation of a firm’s return on sales (Std_ros) and return on assets (Std_roa) for the most recent five years. A
higher value of Std_ros (Std_roa) indicates higher business risk for the firm. The difference-in-differences regressions are run separately
in the high-Std_ros (high-Std_roa) subsample and the low-Std_ros (low-Std_roa) subsample, which are split based on the median of
Std_ros (Std_roa) in the pre-matched sample. Treated equals 1 for the treatment firm (i.e. a firm whose CEO is in her/his zodiac year),
and 0 for the control firm (i.e. a firm whose CEO is not in her/his zodiac year). The time indicator variable, Tzodiac, equals 1 for each year
from 2010 to 2018 when the treatment firm’s CEO is in her/his zodiac year, and 0 for the preceding year and the year after the zodiac year.
The interaction term, Treated*Tzodiac, is the variable of interest. Firm-fixed effects and industry-year interacted dummies are included Table 6.
in the regression, but their results are not reported for simplicity. t-statistics in parentheses are based on robust standard errors Tests of the
clustered by firm. *, **, and *** indicate the two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively moderating effect of
Source(s): Table by authors business risk
JAL DD_BTD
High-superstition regions Low-superstition regions
Variables (1) (2)

Treated*Tzodiac 0.005** 0.005

(2.06) (1.18)
SIZE �0.010 0.030**
(�1.37) (2.47)
ROA �0.005 �0.099**
(�0.12) (�2.23)
Asset_growth 0.006 �0.014
(0.85) (�1.55)
LEV �0.041** �0.004
(�2.13) (�0.16)
NOL 0.007 �0.010**
(0.70) (�2.41)
PPE 0.029 �0.029
(0.95) (�1.03)
Intangible �0.007 0.033
(�0.11) (0.45)
RD 0.029 �0.537
(0.33) (�0.90)
Capital_exp �0.015 0.037
(�0.41) (0.70)
MB �0.016 �0.019
(�1.27) (�0.71)
DA 0.026 �0.013
(1.32) (�0.36)
Cash 0.030 0.060*
(1.48) (1.69)
Foreign 0.005 0.019
(0.23) (0.21)
Managerial_shareholding 0.003 �0.028
(0.12) (�0.14)
Top_shareholding �0.000 0.000
(�0.39) (0.81)
Boardsize �0.008 0.014
(�0.35) (0.53)
Duality 0.004 �0.009*
(0.32) (�1.67)
Institution �0.000 �0.000
(�1.23) (�0.65)
Indp 0.055 0.019
(0.66) (0.33)
BIG4 0.008 �0.016
(0.86) (�1.29)
SOE 0.001 �0.000
(0.14) (�0.04)
*
Constant 0.257 �0.676**
(1.67) (�2.49)
Observations 960 341
Adj R2 0.175 0.364
Firm-fixed effects included included
Industry-year fixed effects included included
Note(s): This table reports the results from testing how the degree of superstition in different regions, where the sample firms are
headquartered, moderates the baseline regression results. The sample period covers the years 2009–2019. All the variables are defined
in Appendix 1. The dependent variable is the residual book-tax difference (DD_BTD) which is estimated from a regression of the total
book-tax difference on the total accruals. The region-level superstition is measured by the provincial moving average of daily volume of
search for the keyword “zodiac year” scaled by the natural logarithm of family households for each province. For each year, we select the
top ten provincial regions that have the relatively highest degree of zodiac-year belief to form the high-regional-superstition subsample,
which comprises the firms that are headquartered in the high-superstition provinces. Other provincial regions are classified as those of
relatively lower zodiac-year belief to form the low-regional-superstition subsample, which consists of the firms that are headquartered
Table 7. in the low-superstition provinces. Treated equals 1 for the treatment firm (i.e. a firm whose CEO is in her/his zodiac year), and 0 for the
Tests of the control firm (i.e. a firm whose CEO is not in her/his zodiac year). The time indicator variable, Tzodiac, equals 1 for each year from 2010 to
2018 when the treatment firm’s CEO is in her/his zodiac year, and 0 for the preceding year and the year after the zodiac year. The
moderating effect of interaction term, Treated*Tzodiac, is the variable of interest. Firm-fixed effects and industry-year interacted dummies are included in
the superstition in the regression, but their results are not reported for simplicity. t-statistics in parentheses are based on robust standard errors clustered
regions where firms by firm. *, **, and *** indicate the two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively
are headquartered Source(s): Table by authors
 Journal of
DD_BTD
SOEs Non-SOEs
Accounting
Variables (1) (2) Literature
Treated*Tzodiac �0.001 0.008***
(�0.43) (2.97)
SIZE �0.016* 0.011
(�1.77) (1.43)
ROA �0.030 0.063
(�0.80) (0.92)
Asset_growth 0.004 0.003
(0.56) (0.35)
LEV �0.031 �0.038**
(�1.04) (�1.98)
NOL 0.010 �0.010*
(0.57) (�1.86)
PPE �0.010 0.003
(�0.27) (0.12)
**
Intangible �0.152 �0.019
(�2.00) (�0.27)
*
RD 0.396 0.074
(1.70) (0.87)
Capital_exp �0.007 0.033
(�0.13) (0.70)
MB �0.012 �0.016
(�0.67) (�1.16)
DA 0.018 0.035
(0.73) (1.32)
*
Cash 0.056 0.028
(1.76) (1.38)
Foreign �0.009 0.006
(�0.36) (0.21)
***
Managerial_shareholding �1.189 0.032
(�3.36) (1.14)
Top_shareholding 0.000 �0.000
(1.16) (�0.29)
Boardsize 0.004 �0.016
(0.29) (�0.94)
Duality �0.007 �0.002
(�0.65) (�0.35)
Institution �0.000 �0.000
(�0.29) (�0.82)
Indp 0.062 �0.109
(1.45) (�1.61)
BIG4 �0.006 0.015
(�0.93) (0.73)
*
Constant 0.336 �0.163
(1.84) (�0.90)
Observations 580 774
Adj R2 0.249 0.232
Firm-fixed effects included included
Industry-year fixed effects included included
Note(s): This table reports the results from testing the moderating effect of state ownership on the baseline regression
results. The sample period covers the years 2009–2019. All the variables are defined in Appendix 1. The dependent variable is
the residual book-tax difference (DD_BTD) which is estimated from a regression of the total book-tax difference on the total
accruals. The moderating variable is the dummy variable, SOE, which equals 1 if a firm’s largest shareholder is a central or
local government or a government-controlled enterprise, and 0 otherwise. Treated equals 1 for the treatment firm (i.e. a firm
whose CEO is in her/his zodiac year), and 0 for the control firm (i.e. a firm whose CEO is not in her/his zodiac year). The time
indicator variable, Tzodiac, equals 1 for each year from 2010 to 2018 when the treatment firm’s CEO is in her/his zodiac year,
and 0 for the preceding year and the year after the zodiac year. The interaction term, Treated*Tzodiac, is the variable of
interest. Firm-fixed effects and industry-year interacted dummies are included in the regression, but their results are not Table 8.
reported for simplicity. t-statistics in parentheses are based on robust standard errors clustered by firm. *, **, and *** indicate Tests of the
the two-tailed statistical significance at the 10%, 5%, and 1% levels, respectively moderating effect of
Source(s): Table by authors state ownership
JAL oversight on tax avoidance. It is also important to provide appropriate training and
education to CEOs in order to increase their understanding of tax laws and ethical codes and
raise their awareness of potential negative consequences of tax avoidance. In so doing, CEOs
are more likely to abstain from avoiding taxes, thereby contributing to more equitable
market and society. Meanwhile, it is imperative to enhance the efficiency of capital markets,
thereby reducing the financial risk for listed firms and lessening their incentives to
avoid taxes.

Notes
1. According to the Enterprise Income Tax Law of the People’s Republic of China, firms are required to
submit a prepayment of income taxes to tax authorities within fifteen days following the end of the
month or quarter, which implies that, if a firm avoids income taxes, any such activity may come to
light within the same fiscal year. Therefore, the benefits and costs of avoiding taxes have to be
traded-off in the same year by managers and their firms.
2. Baidu is the search engine most widely used in China. Google is not available for search in
mainland China.
3. In an unpresented analysis, we test the effects of CEOs’ and CFOs’ zodiac years jointly, and obtain
qualitatively the same results consistent with the stated hypotheses.
4. The Baidu search index begins to provide the moving average of daily search volumes from 2013.
We therefore define the region-level superstition between 2009 and 2012 to be in line with that
in 2013.
5. Each treated firm involves the CEO’s zodiac year alongside two adjacent years for a time-series
comparison, and is matched with a control firm that involves the three firm-year observations for a
cross-sectional comparison in the DID regression analysis.
6. In our matching process, some treated firms for a year could be the control firms for another year
when the CEOs are not in their zodiac years. Thus, the number of firm-year observations is higher
than that of unique firms after the matching.

References
Akerlof, G.A. and Kranton, R.E. (2005), “Identity and the economics of organizations”, The Journal of
Economic Perspectives, Vol. 19 No. 1, pp. 9-32, doi: 10.1257/0895330053147930.
Amiram, D., Bauer, A.M. and Frank, M.M. (2019), “Tax avoidance at public corporations driven by
shareholder taxes: evidence from changes in dividend tax policy”, The Accounting Review,
Vol. 94 No. 5, pp. 27-55, doi: 10.2308/accr-52315.
Amit, R. and Wernerfelt, B. (1990), “Why do firms reduce business risk?”, Academy of Management
Journal, Vol. 33 No. 3, pp. 520-533, doi: 10.5465/256579.
Armstrong, C.S., Blouin, J.L., Jagolinzer, A.D. and Larcker, D.F. (2015), “Corporate governance,
incentives, and tax avoidance”, Journal of Accounting and Economics, Vol. 60 No. 1, pp. 1-17,
doi: 10.1016/j.jacceco.2015.02.003.
Badertscher, B.A., Katz, S.P. and Rego, S.O. (2013), “The separation of ownership and control and
corporate tax avoidance”, Journal of Accounting and Economics, Vol. 56 No. 2, pp. 228-250, doi:
10.1016/j.jacceco.2013.08.005.
Bai, M., Xu, L., Yu, C.F. and Zurbruegg, R. (2020), “Superstition and stock price crash risk”, Pacific-
Basin Finance Journal, Vol. 60, 101287, doi: 10.1016/j.pacfin.2020.101287.
Baker, A.C., Larcker, D.F. and Wang, C.C.Y. (2022), “How much should we trust staggered difference-
in-differences estimates?”, Journal of Financial Economics, Vol. 144 No. 2, pp. 370-395, doi: 10.
1016/j.jfineco.2022.01.004.
Balakrishnan, K., Blouin, J.L. and Guay, W.R. (2019), “Tax aggressiveness and corporate Journal of
transparency”, The Accounting Review, Vol. 94 No. 1, pp. 45-69, doi: 10.2308/accr-52130. Accounting
Bauer, A.M. (2016), “Tax avoidance and the implications of weak internal controls”, Contemporary Literature
Accounting Research, Vol. 33 No. 2, pp. 449-486, doi: 10.1111/1911-3846.12151.
Becker, C.L., DeFond, M.L., Jiambalvo, J. and Subramanyam, K.R. (1998), “The effect of audit quality
on earnings management”, Contemporary Accounting Research, Vol. 15 No. 1, pp. 1-24, doi: 10.
1111/j.1911-3846.1998.tb00547.x.
Beckman, C.M., Haunschild, P.R. and Phillips, D.J. (2004), “Friends or strangers? Firm-specific
uncertainty, market uncertainty, and network partner selection”, Organization Science, Vol. 15
No. 3, pp. 259-275, doi: 10.1287/orsc.1040.0065.
Bhattacharya, U., Kuo, W.Y., Lin, T.C. and Zhao, J. (2018), “Do superstitious traders lose money?”,
Management Science, Vol. 64 No. 8, pp. 3772-3791, doi: 10.1287/mnsc.2016.2701.
Bicchieri, C. (1990), “Norms of cooperation”, Ethics, Vol. 100 No. 4, pp. 838-861, doi: 10.1086/293237.
Bicchieri, C., Muldoon, R. and Sontuoso, A. (2018), “Social norms”, in Zalta, E.N. (Ed.), The Stanford
Encyclopedia of Philosophy (Winter 2018), Metaphysics Research Lab, Stanford University.
Bleak, J.L. and Frederick, C.M. (1998), “Superstitious behavior in sport: levels of effectiveness and
determinants of use in three collegiate sports”, Journal of Sport Behavior, Vol. 21
No. 1, pp. 1-15.
Block, L. and Kramer, T. (2009), “The effect of superstitious beliefs on performance expectations”,
Journal of the Academy of Marketing Science, Vol. 37 No. 2, pp. 161-169, doi: 10.1007/s11747-
008-0116-y.
Boone, J.P., Khurana, I.K. and Raman, K.K. (2013), “Religiosity and tax avoidance”, Journal of
American Taxation Association, Vol. 35 No. 1, pp. 53-84, doi: 10.2308/atax-50341.
Bradshaw, M., Liao, G. and Ma, M.S. (2019), “Agency costs and tax planning when the government is
a major Shareholder”, Journal of Accounting and Economics, Vol. 67 No. 2, pp. 255-277, doi: 10.
1016/j.jacceco.2018.10.002.
Brondolo, J.D. and Zhang, Z. (2016), “Tax administration reform in China: achievements, challenges,
and reform priorities”, IMF Working Papers, 2016(068), A001, doi: 10.5089/9781475523614.
001.A001.
Budner, S. (1962), “Intolerance of ambiguity as a personality variable”, Journal of Personality, Vol. 30
No. 1, pp. 29-50, doi: 10.1111/j.1467-6494.1962.tb02303.x.
Case, T.I., Fitness, J., Cairns, D.R. and Stevenson, R.J. (2004), “Coping with uncertainty: superstitious
strategies and secondary control”, Journal of Applied Social Psychology, Vol. 34 No. 4,
pp. 848-871, doi: 10.1111/j.1559-1816.2004.tb02574.x.
Cen, L., Maydew, E.L., Zhang, L. and Zuo, L. (2017a), “Customer–supplier relationships and corporate
tax avoidance”, Journal of Financial Economics, Vol. 123 No. 2, pp. 377-394, doi: 10.1016/j.
jfineco.2016.09.009.
Cen, W., Tong, N. and Sun, Y. (2017b), “Tax avoidance and cost of debt: evidence from a natural
experiment in China”, Accounting and Finance, Vol. 57 No. 5, pp. 1517-1556, doi: 10.1111/
acfi.12328.
Chan, K.H., Mo, P.L.L. and Tang, T. (2016), “Tax avoidance and tunneling: empirical analysis from an
agency perspective”, Journal of International Accounting Research, Vol. 15 No. 3, pp. 49-66, doi:
10.2308/jiar-51345.
Chen, S., Chen, X., Cheng, Q. and Shevlin, T. (2010), “Are family firms more tax aggressive than non-
family firms?”, Journal of Financial Economics, Vol. 95 No. 1, pp. 41-61, doi: 10.1016/j.jfineco.
2009.02.003.
Cheng, C.S.A., Huang, H.H., Li, Y. and Stanfield, J. (2012), “The effect of hedge fund activism on
corporate tax avoidance”, The Accounting Review, Vol. 87 No. 5, pp. 1493-1526, doi: 10.2308/
accr-50195.
JAL Chyz, J.A. and Gaertner, F.B. (2018), “Can paying ‘too much’ or ‘too little’ tax contribute to forced CEO
turnover?”, The Accounting Review, Vol. 93 No. 1, pp. 103-130, doi: 10.2308/accr-51767.
Denis, D.J. and Sibilkov, V. (2010), “Financial constraints, investment, and the value of cash holdings”,
The Review of Financial Studies, Vol. 23 No. 1, pp. 247-269, doi: 10.1093/rfs/hhp031.
Desai, M.A. and Dharmapala, D. (2006), “Corporate tax avoidance and high-powered incentives”,
Journal of Financial Economics, Vol. 79 No. 1, pp. 145-179, doi: 10.1016/j.jfineco.2005.02.002.
Desai, M.A. and Dharmapala, D. (2009), “Corporate tax avoidance and firm value”, The Review of
Economics and Statistics, Vol. 91 No. 3, pp. 537-546, doi: 10.1162/rest.91.3.537.
Desai, M.A., Dyck, A. and Zingales, L. (2007), “Theft and taxes”, Journal of Financial Economics,
Vol. 84 No. 3, pp. 591-623, doi: 10.1016/j.jfineco.2006.05.005.
Dou, H., Khoo, E.S., Tan, W. and Zhang, J.J. (2024), “Superstition, risk aversion, and audit quality:
evidence from China”, Auditing: A Journal of Practice and Theory, pp. 1-35, doi: 10.2308/AJPT-
2022-140.
Duan, T., Ding, R., Hou, W. and Zhang, J.Z. (2018), “The burden of attention: CEO publicity and tax
avoidance”, Journal of Business Research, Vol. 87, pp. 90-101, doi: 10.1016/j.jbusres.2018.02.010.
Dyreng, S.D. and Markle, K.S. (2016), “The effect of financial constraints on income shifting by US
multinationals”, The Accounting Review, Vol. 91 No. 6, pp. 1601-1627, doi: 10.2308/accr-51420.
Dyreng, S.D., Hanlon, M. and Maydew, E.L. (2008), “Long-run corporate tax avoidance”, The
Accounting Review, Vol. 83 No. 1, pp. 61-82, doi: 10.2308/accr.2008.83.1.61.
Dyreng, S.D., Hanlon, M. and Maydew, E.L. (2010), “The effects of executives on corporate tax
avoidance”, The Accounting Review, Vol. 85 No. 4, pp. 1163-1189, doi: 10.2308/accr.2010.85.
4.1163.
Dyreng, S.D., Mayew, W.J. and Williams, C.D. (2012), “Religious social norms and corporate financial
reporting”, Journal of Business Finance and Accounting, Vol. 39 Nos 7-8, pp. 845-875, doi: 10.
1111/j.1468-5957.2012.02295.x.
D’Agostino, R.B. Jr (1998), “Propensity score methods for bias reduction in the comparison of a
treatment to a non-randomized control group”, Statistics in Medicine, Vol. 17 No. 19,
pp. 2265-2281, doi: 10.1002/(sici)1097-0258(19981015)17:19<2265::aid-sim918>3.0.co;2-b.
Edwards, A., Schwab, C. and Shevlin, T. (2016), “Financial constraints and cash tax savings”, The
Accounting Review, Vol. 91 No. 3, pp. 859-881, doi: 10.2308/accr-51282.
Eisenhardt, K.M. and Zbaracki, M.J. (1992), “Strategic decision making”, Strategic Management
Journal, Vol. 13 No. S2, pp. 17-37, doi: 10.1002/smj.4250130904.
Fazzari, S., Hubbard, R.G. and Petersen, B. (1988), “Investment, financing decisions, and tax policy”,
The American Economic Review, Vol. 78 No. 2, pp. 200-205.
Felson, R.B. and Gmelch, G. (1979), “Uncertainty and the use of magic”, Current Anthropology, Vol. 20
No. 3, pp. 587-589, doi: 10.1086/202333.
Fisman, R., Huang, W., Ning, B., Pan, Y., Qiu, J. and Wang, Y. (2023), “Superstition and risk taking:
evidence from ‘zodiac year’ beliefs in China”, Management Science, Vol. 69 No. 9, pp. 5174-5188,
doi: 10.1287/mnsc.2022.4594.
Francis, J., LaFond, R., Olsson, P.M. and Schipper, K. (2004), “Costs of equity and earnings attributes”,
The Accounting Review, Vol. 79 No. 4, pp. 967-1010, doi: 10.2308/accr.2004.79.4.967.
Frank, M.M., Lynch, L.J. and Rego, S.O. (2009), “Tax reporting aggressiveness and its relation to
aggressive financial reporting”, The Accounting Review, Vol. 84 No. 2, pp. 467-496, doi: 10.2308/
accr.2009.84.2.467.
Gaertner, F.B. (2014), “CEO after-tax compensation incentives and corporate tax avoidance”,
Contemporary Accounting Research, Vol. 31 No. 4, pp. 1077-1102, doi: 10.1111/1911-3846.12058.
Gallemore, J., Maydew, E.L. and Thornock, J.R. (2014), “The reputational costs of tax avoidance”,
Contemporary Accounting Research, Vol. 31 No. 4, pp. 1103-1133, doi: 10.1111/1911-3846.12055.
Graham, J.R. and Tucker, A.L. (2006), “Tax shelters and corporate debt policy”, Journal of Financial Journal of
Economics, Vol. 81 No. 3, pp. 563-594, doi: 10.1016/j.jfineco.2005.09.002. Accounting
Graham, J.R., Hanlon, M., Shevlin, T. and Shroff, N. (2014), “Incentives for tax planning and Literature
avoidance: evidence from the field”, The Accounting Review, Vol. 89 No. 3, pp. 991-1023, doi: 10.
2308/accr-50678.
Guenther, D.A., Wilson, R.J. and Wu, K. (2019), “Tax uncertainty and incremental tax avoidance”, The
Accounting Review, Vol. 94 No. 2, pp. 229-247, doi: 10.2308/accr-52194.
Hadlock, C.J. and Pierce, J.R. (2010), “New evidence on measuring financial constraints: moving
beyond the KZ index”, The Review of Financial Studies, Vol. 23 No. 5, pp. 1909-1940, doi: 10.
1093/rfs/hhq009.
Hanlon, M. and Heitzman, S. (2010), “A review of tax research”, Journal of Accounting and Economics,
Vol. 50 Nos 2-3, pp. 127-178, doi: 10.1016/j.jacceco.2010.09.002.
Hanlon, M. and Slemrod, J. (2009), “What does tax aggressiveness signal? Evidence from stock price
reactions to news about tax shelter involvement”, Journal of Public Economics, Vol. 93 Nos 1-2,
pp. 126-141, doi: 10.1016/j.jpubeco.2008.09.004.
Hanlon, M., Yeung, K. and Zuo, L. (2022), “Behavioral economics of accounting: a review of archival
research on individual decision makers”, Contemporary Accounting Research, Vol. 39 No. 2,
pp. 1150-1214, doi: 10.1111/1911-3846.12739.
Hannan, M.T. and Freeman, J. (1984), “Structural inertia and organizational change”, American
Sociological Review, Vol. 49 No. 2, pp. 149-164, doi: 10.2307/2095567.
He, G. (2016), “Fiscal support and earnings management”, The International Journal of Accounting,
Vol. 51 No. 1, pp. 57-84, doi: 10.1016/j.intacc.2016.01.009.
He, G. and Ren, H.M. (2023), “Are financially constrained firms susceptible to a stock price crash?”,
The European Journal of Finance, Vol. 29 No. 6, pp. 612-637, doi: 10.1080/1351847X.2022.
2075280.
He, G., Ren, H.M. and Taffler, R. (2020), “The impact of corporate tax avoidance on analyst coverage
and forecasts”, Review of Quantitative Finance and Accounting, Vol. 54 No. 2, pp. 447-477, doi:
10.1007/s11156-019-00795-7.
Hilary, G. and Hui, K.W. (2009), “Does religion matter in corporate decision making in America?”,
Journal of Financial Economics, Vol. 93 No. 3, pp. 455-473, doi: 10.1016/j.jfineco.2008.10.001.
Hong, H.A., Lobo, G.J. and Ryou, J.W. (2019), “Financial market development and firm investment in
tax avoidance: evidence from credit default swap market”, Journal of Banking and Finance,
Vol. 107, 105608, doi: 10.1016/j.jbankfin.2019.105608.
Hu, H.W. and Xu, D. (2022), “Manager or politician? Effects of CEO pay on the performance of state-
controlled Chinese listed firms”, Journal of Management, Vol. 48 No. 5, pp. 1160-1187, doi: 10.
1177/01492063211015301.
Jahoda, G. (1969), The Psychology of Superstition, Allen Lane, London.
Jones, J.J. (1991), “Earnings management during import relief investigations”, Journal of Accounting
Research, Vol. 29 No. 2, pp. 193-228, doi: 10.2307/2491047.
Katz, B.G. and Owen, J. (2013), “Exploring tax evasion in the context of political uncertainty”,
Economic Systems, Vol. 37 No. 2, pp. 141-154, doi: 10.1016/j.ecosys.2012.12.005.
Keinan, G. (2002), “The effects of stress and desire for control on superstitious behavior”, Personality
and Social Psychology Bulletin, Vol. 28 No. 1, pp. 102-108, doi: 10.1177/0146167202281009.
Kim, J.B., Li, Y. and Zhang, L. (2011), “Corporate tax avoidance and stock price crash risk: firm-level
analysis”, Journal of Financial Economics, Vol. 100 No. 3, pp. 639-662, doi: 10.1016/j.jfineco.
2010.07.007.
Kramer, T. and Block, L. (2008), “Conscious and nonconscious components of superstitious beliefs in
judgment and decision making”, Journal of Consumer Research, Vol. 34 No. 6, pp. 783-793, doi:
10.1086/523288.
JAL Law, K.K.F. and Mills, L.F. (2015), “Taxes and financial constraints: evidence from linguistic cues”,
Journal of Accounting Research, Vol. 53 No. 4, pp. 777-819, doi: 10.1111/1475-679X.12081.
Law, K.K.F. and Mills, L.F. (2017), “Military experience and corporate tax avoidance”, Review of
Accounting Studies, Vol. 22 No. 1, pp. 141-184, doi: 10.1007/s11142-016-9373-z.
Lee, Y., Ng, S., Shevlin, T. and Venkat, A. (2021), “The effects of tax avoidance news on employee
perceptions of managers and firms: evidence from Glassdoor.com ratings”, The Accounting
Review, Vol. 96 No. 3, pp. 343-372, doi: 10.2308/TAR-2019-0148.
Lei, G., Wang, W., Yu, J. and Chan, K.C. (2022), “Cultural diversity and corporate tax avoidance:
evidence from Chinese private enterprises”, Journal of Business Ethics, Vol. 176 No. 2,
pp. 357-379, doi: 10.1007/s10551-020-04683-2.
Li, J., Guo, J.M., Hu, N. and Tang, K. (2021), “Do corporate managers believe in luck? Evidence of the
Chinese zodiac effect”, International Review of Financial Analysis, Vol. 77, 101861, doi: 10.1016/
j.irfa.2021.101861.
Lipshitz, R. and Strauss, O. (1997), “Coping with uncertainty: a naturalistic decision-making analysis”,
Organizational Behavior and Human Decision Processes, Vol. 69 No. 2, pp. 149-163, doi: 10.
1006/obhd.1997.2679.
Liu, J., Tsang, E.W.K. and Shi, W. (2023), “The superstitious heuristic in strategic decision making”,
Journal of Management. doi: 10.1177/01492063231198191.
Malinowski, B. (1925), “Magic, science and religion”, in Needham, J. (Ed.), Science, Religion and
Reality, The Macmillan, New York, NY, pp. 19-84.
McGuire, S.T., Omer, T.C. and Sharp, N.Y. (2012a), “The impact of religion on financial reporting
irregularities”, The Accounting Review, Vol. 87 No. 2, pp. 645-673, doi: 10.2308/accr-10206.
McGuire, S.T., Omer, T.C. and Wang, D. (2012b), “Tax avoidance: does tax-specific industry expertise
make a difference?”, The Accounting Review, Vol. 87 No. 3, pp. 975-1003, doi: 10.2308/
accr-10215.
Olsen, K.J. and Stekelberg, J. (2016), “CEO narcissism and corporate tax sheltering”, Journal of the
American Taxation Association, Vol. 38 No. 1, pp. 1-22, doi: 10.2308/atax-51251.
Payne, J.W., Bettman, J.R. and Johnson, E.J. (1988), “Adaptive strategy selection in decision making”,
Journal of Experimental Psychology: Learning, Memory, and Cognition, Vol. 14 No. 3,
pp. 534-552, doi: 10.1037/0278-7393.14.3.534.
Powers, K., Robinson, J.R. and Stomberg, B. (2016), “How do CEO incentives affect corporate tax
planning and financial reporting of income taxes?”, Review of Accounting Studies, Vol. 21 No. 2,
pp. 672-710, doi: 10.1007/s11142-016-9350-6.
Rego, S.O. (2003), “Tax-avoidance activities of US multinational corporations”, Contemporary
Accounting Research, Vol. 20 No. 4, pp. 805-833, doi: 10.1506/VANN-B7UB-GMFA-9E6W.
Rego, S.O. and Wilson, R. (2012), “Equity risk incentives and corporate tax aggressiveness”, Journal
of Accounting Research, Vol. 50 No. 3, pp. 775-810, doi: 10.1111/j.1475-679X.2012.00438.x.
Rice, T.W. (2003), “Believe it or not: religious and other paranormal beliefs in the United States”,
Journal for the Scientific Study of Religion, Vol. 42 No. 1, pp. 95-106, doi: 10.1111/1468-
5906.00163.
Rice, G.H. Jr (1985), “Available information and superstitious decision making”, Journal of General
Management, Vol. 11 No. 2, pp. 35-44, doi: 10.1177/030630708501100204.
Roberts, M.R. and Whited, T.M. (2013), “Endogeneity in empirical corporate finance”, in
Constantinides, G.M., Harris, M. and Stulz, R.M. (Eds), Handbook of the Economics of
Finance, Vol. 2, pp. 493-572, doi: 10.1016/b978-0-44-453594-8.00007-0.
Rodgers, W. and Gago, S. (2001), “Cultural and ethical effects on managerial decisions: examined in a
throughput model”, Journal of Business Ethics, Vol. 31 No. 4, pp. 355-367, doi: 10.1023/A:
1010777917540.
Rosenbaum, P.R. and Rubin, D.B. (1983), “The central role of the propensity score in observational Journal of
studies for causal effects”, Biometrika, Vol. 70 No. 1, pp. 41-55, doi: 10.1093/biomet/70.1.41. Accounting
Roth, J., Sant’Anna, P.H., Bilinski, A. and Poe, J. (2023), “What’s trending in difference-in-differences? Literature
A synthesis of the recent econometrics literature”, Journal of Econometrics, Vol. 235 No. 2,
pp. 2218-2244, doi: 10.1016/j.jeconom.2023.03.008.
Schippers, M. and Van Lange, P.A. (2006), “The psychological benefits of superstitious rituals in top
sport”, Journal of Applied Social Psychology, Vol. 36 No. 10, pp. 2532-2553, doi: 10.1111/j.0021-
9029.2006.00116.x.
Schwenk, C.H. (1986), “Information, cognitive biases, and commitment to a course of action”,
Academy of Management Review, Vol. 11 No. 2, pp. 298-310, doi: 10.2307/258461.
Shevlin, T., Tang, T.Y.H. and Wilson, R.J. (2012), “Domestic income shifting by Chinese listed firms”,
Journal of the American Taxation Association, Vol. 34 No. 1, pp. 1-29, doi: 10.2308/atax-10150.
Smith, V.L. (2003), “Constructivist and ecological rationality in economics”, The American Economic
Review, Vol. 93 No. 3, pp. 465-508, doi: 10.1257/000282803322156954.
Tang, T.Y.H. (2020), “A review of tax avoidance in China”, China Journal of Accounting Research,
Vol. 13 No. 4, pp. 327-338, doi: 10.1016/j.cjar.2020.10.001.
Teoh, S.H., Welch, I. and Wong, T.J. (1998), “Earnings management and the underperformance of
seasoned equity offerings”, Journal of Financial Economics, Vol. 50 No. 1, pp. 63-99, doi: 10.
1016/s0304-405x(98)00032-4.
Tsang, E.W.K. (2004a), “Superstition and decision-making: contradiction or complement?”, Academy
of Management Journal, Vol. 18 No. 4, pp. 92-104, doi: 10.5465/ame.2004.15268696.
Tsang, E.W.K. (2004b), “Toward a scientific inquiry into superstitious business decision making”,
Organization Studies, Vol. 25 No. 6, pp. 923-946, doi: 10.1177/0170840604042405.
Wang, S., Chen, D. and Dong, J. (2012), “Superstition in strategic decision making: a two-level study of
Chinese executives”, Academy of Management Proceedings, Vol. 2012 No. 1, 13099, doi: 10.
5465/AMBPP.2012.13099abstract.
Wilson, R.J. (2009), “An examination of corporate tax shelter participants”, The Accounting Review,
Vol. 84 No. 3, pp. 969-999, doi: 10.2308/accr.2009.84.3.969.
Wu, L., Wang, Y., Lin, B.X., Li, C. and Chen, S. (2007), “Local tax rebates, corporate tax burdens, and
firm migration: evidence from China”, Journal of Accounting and Public Policy, Vol. 26 No. 5,
pp. 555-583, doi: 10.1016/j.jaccpubpol.2007.08.003.
Yu, H. (2019), “Reform of state-owned enterprises in China: the Chinese Communist Party strikes
back”, Asian Studies Review, Vol. 43 No. 2, pp. 332-351, doi: 10.1080/10357823.2019.1590313.
Zeng, H., Zhang, X., Zhou, Z., Zhang, T. and Zhou, Q. (2022), “Does the effect of the annual year taboo
exist? Empirical evidence from senior managers’ zodiac year and corporate inefficient
investment”, The British Accounting Review, Vol. 54 No. 6, 101114, doi: 10.1016/j.bar.2022.
101114.

(The Appendix follows overleaf)

JAL Appendix 1

Dependent variables in the baseline regression analysis

DD_BTD The residual book-tax difference estimated from the following regression
BTDi;t ¼ α1 TACC i;t þ μi þ εi;t
where BTD is the total book-tax difference, computed as pre-tax financial income
minus taxable income scaled by the lagged total assets. Taxable income is calculated
as income tax expense minus deferred tax expense, divided by the nominal tax rate.
TACC is total accruals scaled by the lagged total assets. The total accruals are
calculated as the operating income minus operating cash flows. A higher DD_BTD
represents a higher level of tax avoidance.
Alternative measure of tax avoidance
Adj_ETR The difference between nominal tax rate and effective tax rate (ETR). ETR is
calculated as the total tax expense minus deferred tax expense, scaled by pre-tax
income. Nominal tax rate is the corporate income tax rate disclosed by the firm. Adj_
ETR equals 0 if ETR is higher than nominal tax rate. A higher Adj_ETR indicates a
greater level of tax avoidance
Ind_adj_ETR The industry-adjusted measure of Adj_ETR, calculated as the difference between a
firm’s Adj_ETR and the industry average of Adj_ETR for the same year, divided by
this industry average. The higher Ind_adj_ETR, the greater the tax avoidance
Variable as to whether a CEO is in her/his zodiac year
Zodiac_ceo If the CEO was born on a day between the start date of the Chinese Lunar new year and
31 December, Zodiac_ceo equals 1 for cases in which a CEO is at the age of a multiple of
12, and 0 otherwise. Provided that the CEO was born on a day between 1 January and
the start date of the Chinese lunar new year, Zodiac_ceo equals 1 for cases in which a
CEO is at the age of one year less than a multiple of 12, and 0 otherwise
Key variables that compose the DID estimator
Treated 1 if a firm is a treatment firm, and 0 if a firm is a control firm. Treatment (control) firms
are those whose CEOs are (not) in their zodiac years
Tzodiac 1 for the zodiac year t, and 0 for the year before, and the year after, the zodiac year (i.e.
the years t�1 and tþ1)
Moderating variables
Std_ros The standard deviation of a firm’s return on sales for the most recent five years
Std_roa The standard deviation of a firm’s return on assets for the most recent five years
SA_index Following Hadlock and Pierce (2010), SA_index is defined as follows
SA index ¼ −0:737 3 size þ 0:043 3 size2 − 0:040 3 age where size is the natural
logarithm of the firm’s book value of the total assets, and age is the number of years for
which the firm has been listed.
Dividend The natural logarithm of 1 plus cash dividends paid to common shareholders in a fiscal
year. Lower cash dividends indicate tighter financial constraints
SUP_Province The degree of zodiac-year belief for a province in a year. SUP_Province is calculated as
the average daily volume of search for the keyword “zodiac year” in the Baidu search
engine for a province over a fiscal year, which is scaled by the natural logarithm of the
family households of the province in that year
SOE 1 if a firm’s largest shareholder is a central or local government or a government-
controlled enterprise, and 0 otherwise
Matching covariates
SIZE The natural logarithm of a firm’s book value of total assets at the end of a fiscal year
Asset_growth The change in the total assets from the previous year to the current year, deflated by
the average total assets for the previous year
Table A1.
Variable definitions (continued )
 Journal of
ROA Earnings before interests and taxes over a fiscal year, divided by the total assets, at the Accounting
end of the year Literature
LEV The sum of short- and long-term debt divided by the total assets at the end of a fiscal
year
Std_return The annualized standard deviation of a firm’s monthly stock returns over a fiscal year
Indp The number of independent directors as a fraction of the total directors on the board of
a firm at the end of a year
Control variables
SIZE The natural logarithm of a firm’s book value of total assets at the end of a fiscal year
ROA Earnings before interests and taxes over a fiscal year, divided by the total assets, at the
beginning of the year
Asset_growth The change in the total assets from the previous year to the current year, deflated by
the total assets for the previous year
LEV The sum of short- and long-term debt divided by the total assets at the end of a fiscal
year
NOL The net operating losses. Since Chinese firms do not report the tax benefits from net
operating losses in the balance sheet, we use a continuous variable, equal to
accumulated pre-tax losses (in billions) reported in the last five years, to proxy for the
net operating losses (Bradshaw et al., 2019). NOL equals 0 if the accumulated pre-tax
losses are positive
PPE The fixed assets divided by the total assets at the end of a fiscal year
Intangible The intangible assets divided by the total assets at the end of a fiscal year
RD R&D expenses divided by net sales for a fiscal year
Capital_exp The capital expenditures in a fiscal year, deflated by the total assets at the end of the
year
MB The market value of equity, divided by the book value of equity, at the beginning of a
fiscal year
DA Abnormal accruals of a firm in a fiscal year, which is estimated using the cross-
sectional version of modified Jones model with at least 20 firm-year observations for
each industry-year
Cash The cash and cash equivalents divided by the total assets at the end of a fiscal year
Foreign The foreign sales scaled by the total sales over a fiscal year
Managerial_ The shares held by a firm’s executives, divided by the firm’s total shares outstanding,
shareholding at the end of a fiscal year
Top_shareholding The shares held by a firm’s largest shareholder, divided by the firm’s total shares
outstanding, at the end of a fiscal year
Boardsize The natural logarithm of the number of directors on the board of a firm at the end of a
fiscal year
Duality 1 if CEO and the chairman of board of directors are the same person, and 0 otherwise
Institution The percentage of institutional shareholding at the end of a fiscal year
Indp The number of independent directors as a fraction of the total directors on the board of
a firm at the end of a fiscal year
BIG4 1 if a firm’s financial statement is audited by one of the big-4 auditors, and 0 otherwise
SOE 1 if a firm’s largest shareholder is a central or local government or a government-
controlled enterprise, and 0 otherwise
Source(s): Table by authors Table A1.
JAL Appendix 2

Initial firm-year observations which cover companies listed on the Shenzhen or Shanghai 29,686
Stock Exchange for the period 2009–2019
Less: observations in the financial industry (902)
Less: observations with missing data on tax expense or with negative pre-tax income (2,776)
Less: observations with missing values in the measure of corporate tax avoidance (DD_BTD) (2,143)
Less: observations without information required to identify the CFOs’ zodiac years (3,496)
Less: observations which are in the CFOs’ zodiac years (1,137)
Less: observations that have CEO turnover (2,998)
Less: observations for which we fail to identify the CEOs’ zodiac years (750)
Less: observations that are missing in the values of control variables used in the multivariate (4,910)
tests of hypotheses
Table A2. Firm-year observations (unique firms) available for the propensity-score matching 10,450
Sample selection (2,406)
procedure Source(s): Table by authors

About the authors

Guanming He’s research appears in varied prestigious international journals (FT 50/ABS 4–3 ranked)
and has been featured in various mainstream media and press releases such as the Columbia Law
School’s Blue Sky Blog on corporations and capital markets, the Canadian HR Newswire, the European
Climate Change Review, the Institute of Chartered Accountants in England and Wales (ICAEW)
insights, and Bloomberg media, etc. He is currently serving on the editorial advisory board of
International Journal of Contemporary Hospitality Management, and on the editorial board of Corporate
Governance: An International Review. He also serves as the ad hoc reviewer for a range of
internationally reputable accounting, finance, management, and economics journals (FT 50/ABS 4–3
ranked) and books (by Cengage/Pearson/ SAGE). Thus far, he has performed hundreds of paper reviews
as fairly and objectively as possible for various top-tier and second-tier international journals. He won
best paper awards at the 3rd Australian Financial Market and Corporate Governance conference, 7th
Indonesian Finance Association International Conference, 24th and 26th International Business
Research conference, obtained a top paper award at the 28th annual Global Finance conference, as well
as a Best-in-Track paper award at the 35th annual meeting of the Academy of Finance, and was
nominated as a finalist for the best conference paper award of European Financial Management
Association. He was the director of accounting PhD program, and on the university’s research degree
committee, from 2019 to 2023. Guanming He is the corresponding author and can be contacted at:
[email protected]
Dongxiao Shen’s research interests broadly include financial reporting, tax avoidance, insider
trading, digitalization, and corporate social responsibilities. She has presented her work at a range of
prestigious international conferences, including the American Accounting Association Annual
Meeting. Dongxiao has also contributed to the academic field through publications in peer-reviewed
journals and book chapters. She serves as the ad hoc referee for various esteemed journals and
conferences. In addition, Dongxiao holds the qualification of Association of Chartered Certified
Accountants (ACCA).

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]