International Review of Economics and Finance 67 (2020) 207–224

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier.com/locate/iref

Do multiple large shareholders affect tax avoidance? Evidence

from China
Caiyue Ouyang a, Jiacai Xiong b, Kun Huang c, *
a
School of Economics and Management, Beijing Jiaotong University, Beijing, China
b
School of Accounting, Jiangxi University of Finance and Economics, Nanchang, China
c
School of Management, Huazhong University of Science and Technology, Wuhan, China

A R T I C L E I N F O A B S T R A C T

JEL classification: We examine the collusion versus monitoring effects of multiple large shareholders (MLS) on firms’
G32 tax avoidance (TA). MLS can enhance the monitoring of the top shareholder and executives to
G34 restrain agency problems, or they can collude with the top shareholder to exacerbate the expro-
H26
priation of minority shareholders. Using a sample of Chinese firms, we find that firms with MLS
Keywords: engage in less TA than their non-MLS counterparts do, supporting the monitoring hypothesis. The
Multiple large shareholders
findings are robust after accounting for endogeneity concerns. Further tests show that the impact
Tax avoidance
of MLS on TA is more pronounced when MLS have stronger power and when firms have severe
Agency problems
agency problems. In addition, we find that effective external corporate governance can substitute
for the monitoring effects of MLS. Most importantly, we find that TA activities are more likely to
enhance value with the monitoring of MLS. In sum, our findings are consistent with the literature
indicating that MLS make a positive contribution to firm value by reducing the nontax costs of TA.

1. Introduction

Tax is generally a burden on firms globally, consuming a significant share of a firm’s pretax earnings. Hence, firms should have strong
incentives to engage in tax avoidance (TA) activities to pay as little tax as possible. The argument, however, ignores the nontax costs of
TA (Scholes, Wolfson, Erickson, Maydew, & Shevlin, 2005). These nontax costs drive some firms to forego tax savings from TA. Among
these costs, those due to agency conflict receive increasing attention. In particular, Desai (2004) provides some notable TA cases about
Enron, Tyco, and Xerox, whose motivation for increasing reported earnings through tax planning was related to earnings manipulation
and management malfeasance. Desai and Dharmapala (2006) provide empirical evidence that the equity incentive reduces TA and
interpret the result to mean that the equity incentive prevents managers from extracting rent, with the assumption that complex TA
activities facilitate managers in masking their rent extraction. Subsequently, another study reports supporting evidence from Russia in a
case study of Sibneft to illustrate how TA can serve managers’ rent extraction purposes (Desai, Dyck, & Zingales, 2007). Following this
research, Chen, Chen, Cheng, and Shevlin (2010) and Kim, Li, and Zhang (2011) both cite the agency view of the opacity surrounding
TA. Taking advantage of the agency conflicts between controlling shareholders and minority shareholders, as well as the weak corporate
governance environment in a transitional economy, Chan, Mo, and Tang (2016) provide direct empirical evidence that Chinese listed
companies use TA to shield tunneling activities. Moreover, in an insider trading context, Chung, Goh, Lee, and Shevlin (2019) find that
managers exploit the opacity arising from aggressive tax activities to extract rent from shareholders, particularly those shareholders who

* Corresponding author. School of Management, Huazhong University of Science and Technology, 430074, China.
E-mail addresses: [email protected] (C. Ouyang), [email protected] (J. Xiong), [email protected] (K. Huang).

https://doi.org/10.1016/j.iref.2019.12.009
Received 14 September 2019; Received in revised form 16 December 2019; Accepted 28 December 2019
Available online 30 December 2019
1059-0560/© 2020 Elsevier Inc. All rights reserved.
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

sell their shares to managers, which provides direct evidence that managers extract rents with the shield of TA activity. Atwood and
Lewellen (2019) also provide evidence that TA and manager diversion are complementary when corporate governance is ineffective
using the sample of multinational firms in 28 counties. In sum, these studies document that a severe agency problem causes firms to
engage in more TA, and they focus on the type I agency problem (executives expropriate shareholders) except Chan et al. (2016).
However, in contrast to the United States, highly concentrated ownership structure dominates the global economy (La Porta,
Lopez-de-Silanes, & Shleifer, 1999). In these countries, the agency conflict between controlling and minority shareholders (type II
agency problem) is the main agency problem (Claessens, Djankov, & Lang, 2000; Enriques & Volpin, 2007; Jiang, Lee, & Yue, 2010; He,
Chiu, & Zhang, 2015; La Porta, Lopez-de-Silanes, & Shleifer, 1999). Typically, controlling shareholders retain substantial control in the
firm’s operations, such as the appointment of CEOs and directors (Jiang et al., 2010; Kimber & Lipton, 2005; Ma & Khanna, 2016).
Therefore, the interests of controlling shareholders and senior executives are often bound and aligned to infringe the interests of mi-
nority shareholders (Jiang et al., 2010, 2018). However, whether in western countries or developing countries, the ownership structure
of multiple major shareholders is common. For example, Claessens et al. (2000) point out that 32.2% of East Asian companies have at
least two major shareholders with shareholding ratios in excess of 10%. Maury and Pajuste (2005) find that 48% of Finnish listed
companies have at least two major shareholders. Laeven and Levine (2008) study 1,657 enterprises in 13 western European countries
and find that about 34% of the enterprises have two or more major shareholders. Even in the United States, where ownership structures
are considered highly fragmented, Edmans and Manso (2010) find that 70% of companies have multiple major shareholders who own
more than 5% shares. Cai, Hillier, and Wang (2016) show that more than 49% (30%) of listed companies have at least two large
shareholders when using a 5% (10%) ownership threshold in the Chinese market. Furthermore, major shareholders are not homoge-
neous and contestability exists among them (Hope, 2013; Maury & Pajuste, 2005), which may make the firms’ behavior toward TA
uncertain.
Following the literature (e.g., Attig, El Ghoul, & Guedhami, 2009, 2008; Jiang et al., 2018; Laeven & Levine, 2008; Maury & Pajuste,
2005), we consider large shareholders as those holding at least 10% of total outstanding shares. When a firm has one or more non-top
large shareholders (in addition to the top shareholder), we refer to this case as a firm having multiple large shareholders (MLS). While
previous studies devote some effort to unearthing the link between ownership structure and TA activities (Badertscher, Katz, & Rego,
2013; Chen et al., 2010; Cheng, Huang, Li, & Stanfield, 2012; Khan, Srinivasan, & Tan, 2017; McGuire, Wang, & Wilson, 2014), few
studies explore the impact of MLS on TA. These MLS are not top large shareholders; thus, they cannot expropriate minority shareholders
alone. Nonetheless, MLS have a voice on corporate boards, and they can influence firm activities. Thus, it is unclear whether MLS
attenuate or worsen the type II agency problem, which leads to uncertainty about TA.
We argue that the impact of MLS on TA is unclear for two reasons. On the one hand, MLS may collude with the top shareholder at the
expense of minority shareholders (Ashbaugh-Skaife, Collins, & LaFond, 2006; Cai et al., 2016; Cheng, Lin, & Wei, 2013; Kahn & Winton,
1998; Maury & Pajuste, 2005; Zwiebel, 1995), which exacerbates the type II agency problem. Tax planning activities require sophis-
ticated transactions and arrangements, among other activities, making organization more complicated and financial reporting less
transparent (Balakrishnan, Blouin, & Guay, 2019; Chan et al., 2016; Chuang et al., 2019; Desai, 2004; Desai & Dharmapala, 2006; Desai
et al., 2007; Kim et al., 2011). The top shareholder and MLS may shield their opportunistic behaviors with tax planning activities. Hence,
MLS induce higher TA levels. On the other hand, MLS have a strong incentive to monitor the opportunistic behaviors of the top
shareholders to protect their own vested interests in the firms. Most importantly, MLS have the abilities and resources to monitor the top
shareholder, which alleviates the type II agency problem (Attig, Guedhami, & Mishra, 2008; Ben-Nasr, Boubaker, & Rouatbi, 2015;
Bennedsen & Wolfenzon, 2000; Boubaker & Sami, 2011; Guti
errez & Pombo, 2009; Jiang et al., 2018; Mishra, 2011; Pagano & R€ oell,
1998). Thus, MLS make monitoring effective. In simple terms, with MLS, more eyes are on the firm to scrutinize the top shareholder,
which makes shielding its opportunistic behavior with tax-planning activities more difficult. Therefore, the top shareholder is less likely
to engage in aggressive tax activities, resulting in a lower level of TA. Linking the literature on MLS and the agency theory framework
leads to the question of whether MLS lead to more or less TA.
This study aims to investigate the impact of MLS on TA. Specifically, we study the collusion versus monitoring hypotheses of MLS on
TA using a sample of Chinese firms. The Chinese environment is suitable for our examination for three reasons. First, in terms of tax
enforcement, China’s tax administration lacks the capability or willingness to detect and constrain sophisticated TA arrangements due to
underdeveloped financial markets, inadequate financial resources and human expertise (He, Wong, & Young, 2012; Lin, Mills, Zhang, &
Li, 2018). In addition, considering the preferential tax rates for firms in specific regions and industries, separate tax reporting based on a
legal entity, and the adoption of multilayered pyramid structures that firms can use to reduce the tax burden, TA is more prevalent and
rampant in the Chinese setting (Liu, Shi, & Ferrantino, 2016). Second, weak investor protections impede the Chinese capital market
(Allen, Qian, & Qian, 2005). Chinese firms typically have a severe type II agency problem and weak corporate monitoring. The top
shareholder can easily seek private benefits (Jiang et al., 2010; Lou, Wang, & Yuan, 2014). For example, Chan et al. (2016) provide
empirical evidence that Chinese listed firms use TA to shield tunneling activity, which indicates severe agency problems. Third, China
began the split-share structure reform (the Reform) in 2005, which falls within our sample period. The Reform allows investors to
convert non-tradable shares into tradable shares. After the Reform, the ownership of top shareholders and MLS decreased. Hence, the
Reform is an exogenous quasi-experiment that changed the ownership structure, but is not related to a firm’s TA activities. Therefore, we
can conduct a subsample analysis using the pre- and post-Reform setting to mitigate the endogeneity concern between MLS and TA. In
sum, China offers an excellent institutional background to examine the competing hypotheses.
Our findings suggest that firms with MLS engage in less TA. Specifically, we document that an average firm with MLS pays
approximately 4.65% more in corporate income tax than a firm without MLS, which supports the monitoring hypothesis. The findings
remain robust after using (1) an instrumental variable approach, (2) firm fixed effect model, and (3) a subsample of firms that subject to
the Reform, which represents an exogenous shock on the impact of MLS on TA to account for endogeneity. An alternative explanation of

208
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

the negative relationship between MLS and TA is that MLS alleviate firms’ financial constraints and firms do not need to keep tax
savings. To exclude this concern, we partition the full sample into low- and high-financial constraint subsamples and find that the
coefficients of MLS are basically positive and significant for both subsamples. The additional robustness tests suggest that the negative
impact of MLS on TA remains when using alternative specifications of MLS and TA, more control variables, and the one-period lagged
MLS measure. Moreover, we find that the strength of MLS (in terms of the number of MLS or a high percentage of MLS ownership)
contributes to lower TA. Furthermore, the impact of MLS is more pronounced among non-state-owned firms and firms with high other
receivables or high related-party transactions, namely the entrenchment motivation of whose controlling shareholders is stronger.
Additionally, we find that the impact of MLS on TA magnifies for firms with poor corporate governance in terms of having a non-Big 4
auditor or facing lower product market competition. These additional findings corroborate with the monitoring hypothesis of MLS on
TA. Finally, we examine the value implication of TA. With the monitoring role of MLS, TA activities are more likely to enhance value. In
line with the agency view of TA, our results show that the value-enhancing implication of TA exists only in firms with MLS.
We make two contributions to the literature. First, we provide the first study on the impact of MLS on TA in an agency-problem
setting. Our findings are consistent with the interpretation of TA as increasing agency costs (Atwood & Lewellen, 2019; Chan et al.,
2016; Chen et al., 2010; Cheng et al., 2012; Chung et al., 2019; Desai & Dharmapala, 2006), and MLS can restrain agency problems in the
context of TA. Second, our study contributes to the growing literature on the governance effect of MLS (Attig et al., 2008; Bennedsen &
Wolfenzon, 2000; Jiang et al., 2018; Pagano & R€ oell, 1998) by documenting the negative impact of MLS on TA via a monitoring channel.
Our findings are consistent with the literature indicating that MLS make a positive contribution to firm value by reducing the nontax
costs of TA.

2. Literature review and hypothesis development

While TA allows a firm to garner tax savings, it also incurs nontax costs. The controlling shareholders can use TA as a shield to divert
resources out of the firm (Desai & Dharmapala, 2006; Bauer, Fang, Pittman, Zhang, & Zhao, 2019). For example, by claiming that
related-party transactions between business group are for TA purposes, firms do not report insider transactions to avoid detection by tax
authorities and setting up subsidiaries in tax havens or low-tax regions is in order to keep tax savings, and so on (Lin et al., 2018; Liu
et al., 2016). Consistent with this argument, Xia, Cao and Chan (2017) find that a high social trust environment can reduce agency
conflict so that tax avoidance is less. The controlling shareholder can also take advantage of the increased firm opaqueness induced by
TA to engage in insider trading (Chung et al., 2019). Utilizing the Ninth Circuit’s 1999 ruling, Arena, Wang and Yang (2019) find that
the threat of shareholder litigation plays a disciplinary role in curbing managerial rent extraction from TA activities. In addition, by
examining tunneling through inter-corporate loans in China, Bauer et al. (2019) find that additional cash flow from tax savings and
increased financial opacity from tax planning provide controlling shareholders with both incentives and opportunities for tunneling
activities.Therefore, in addition to the costs associated with potential fines by a tax authority and the execution of TA, the increase in
agency costs (rent extraction) from TA activities, which outweigh the potential benefits, is a major source of firm value destruction.
Based on the findings in the literature, we postulate that TA likely leads to worsening agency problems. We put forth two potential
situations related to MLS in the context of TA with a special focus on the Chinese environment. The first is the monitoring hypothesis of
MLS. Because MLS have significant ownership of total shares, they are typically directors of the board. Hence, MLS have sufficient
incentives and abilities to restrain the type II agency problem with the top shareholder to protect their own interests. Previous studies
suggest that a firm with MLS has lower private benefit extraction from the top shareholder (Guti
errez & Pombo, 2009), fewer
related-party transactions, and fewer agency problems (Laeven & Levine, 2008; Maury & Pajuste, 2005; Pagano & R€ oell, 1998). MLS
make it more difficult for the top shareholder to extract private benefits under the shelter of TA, and MLS also limit the top shareholder’s
ability to tunnel the firm’s resources via TA. For instance, MLS can require the board to set strict rules to judge the rationality of
related-party transactions, require subsidiaries to establish a perfect internal control system, or make financial reports to the parent
company more frequently. Hence, the top shareholder is less likely to direct the firm to engage in TA activities. Overall, MLS means more
eyes and increased scrutiny on a firm. Therefore, firms with MLS, in general, engage in less TA than do firms without MLS.
The second is the collusion hypothesis. We argue that MLS may collude with the top shareholder to expropriate minority share-
holders. In this case, MLS work with the top shareholder to increase the extent of the expropriation of minority shareholders. Thus, the
type II agency problem worsens. Several studies offer evidence to support this possibility (Kahn & Winton, 1998; Zwiebel, 1995). Hence,
MLS may destroy firm value (Cai et al., 2016; Fang, Hu, & Yang, 2018; Laeven & Levine, 2008; Maury & Pajuste, 2005). TA makes firms
opaque and makes it more convenient for MLS and the top shareholder to extract private benefits. For example, MLS and the top
shareholder could engage in insider trading, conduct more related-party transactions for private benefits, and so on. Prior studies argue
that firms engage in TA to obtain tax savings (Slemrod, 2004), while in the Chinese setting, insiders have a strong incentive to
expropriate cash savings from TA (Bauer, Fang, Pittman, Zhang, & Zhao, 2019). Hence, in the collusion hypothesis, the type II agency
problem worsens. Firms with MLS engage in more TA to leverage more opportunities to cover up their rent extraction or expropriate the
tax savings.
Taken together, our testable hypothesis has two alternatives:
H1A. Due to the monitoring effect of MLS, a firm engages in less TA.
H1B. Due to the collusion effect of MLS, a firm engages in more TA.

209
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 1
Variable definition.
Variables Definition

ETR1 tax expense/income before tax (Chen et al., 2010; Porcano, 1986).
ETR2 (tax expense - deferred tax expense)/income before tax (Porcano, 1986).
BTD Total book-tax difference, which is calculated by book income less taxable income scaled by lagged assets (Manzon & Plesko, 2002).
DDBTD Desai and Dharmapala’s (2006) residual book-tax difference, which equals the residual from the following fixed effects regression: BTDit ¼
β1TACCitþμiþεit, where BTD is the total book-tax difference, TACC is total accruals measured using the cash flow method per Hribar and Collins
(2002).
MLS A dummy variable with a value of if the firm has at least one or more non-top large shareholders and zero otherwise. A large shareholder is defined as
one holding at least 10% of total shares.
MLS_5% A dummy variable with a value of if the firm has at least one or more non-top large shareholders and zero otherwise. A large shareholder is defined as
one holding at least 5% of total shares.
MLS_NUM In addition to the top shareholder, the number of other non-top large shareholders.
MLS_RATIO The ratio of the sum of total shares held by non-top large shareholders to the number of shares held by the top shareholder.
TAXRATE The firms’ statutory tax rate.

SIZE Natural logarithm of total assets.

LEV Ratio of total liabilities to total assets.
ROA Ratio of net profits to total assets.
BM Ratio of book value of equity to market value of equity.
PPE Ratio of property, plant and equipment assets to lagged total assets.
INTANG Ratio of intangible assets to total assets.
INVENT Ratio of inventory to total assets.
CASH Ratio of a sum of cash and marketable securities to total assets.
LIQUIDITY Ratio of current assets to current liabilities.
ABSDA The absolute value of discretionary accruals using the Kothari, Leone, and Wasley (2005) model.
LOSS If a firm has a net loss in the previous year, the value is one and zero otherwise.
BIG4 If a firm is audited by a Big 4 auditor, the value is one and zero otherwise.
BSIZE Natural logarithm value of one plus the number of directors.
BIND Ratio of the number of independent directors to board size.
DUAL If the chairman of the board and the CEO are the same individual, the value is 1 and zero otherwise.

SOE An indicator variable that equals one if the government is the ultimate controlling shareholder and zero otherwise.
RPT Ratio of total related party transactions to total assets.
ORECTA Ratio of other receivables to total assets.
CASH_DIV Cash dividend per share, which is the sum of cash dividends issued by a company for every ordinary share outstanding.

TQ The sum of the market value of tradable shares, the book value of nontradable shares, and liabilities scaled by the book value of total assets.
SGROW The growth of annual sales.

3. Methods and data

3.1. Methods

Following Edwards, Schwab, and Shevlin (2016) and Jiang et al. (2018), we use the following multivariate regression model to
examine our competing hypotheses:

ETR1i,t (or ETR2i,t) ¼ α0 þ α1 MLSi,tþ CONTROL þ INDUSTRY þ YEAR þ εi,t (1)

We use effective tax rate (ETR1 and ETR2) to measure firms’ TA (Chen et al., 2010; Li, Liu, & Ni, 2017; Porcano, 1986). Specifically,
we define ETR1 as the total tax expense divided by pretax income, ETR2 as the total tax expense minus deferred tax expense divided by
pretax income. In the robustness check, we employ two alternative measures of TA: the book-tax difference measure (BTD) proposed by
Manzon and Plesko (2002) and the book-tax difference measure (DDBTD) developed by Desai and Dharmapala (2006). MLS is a dummy
variable with a value of one if a firm has MLS and zero otherwise. In section 4.6.1, we use the number of non-top large shareholders
(MLS_NUM) and the ratio of the sum of total shares held by non-top large shareholders to the number of shares held by the top
shareholder (MLS_RATIO) to proxy for the strength of MLS. If MLS monitor the top shareholder effectively, we expect a positive α1. If
MLS collude with the top shareholder, we expect a negative α1.
For control variables, we include determinants that have been shown to explain TA activities in prior literature (e.g., Chen et al.,
2010; Li et al., 2017). We control for firm’s statutory tax rate (TAXRATE) because the statutory tax rates of Chinese listed firms are
different across industry and time. For example, high-tech firms have a statutory tax rate of 15%; firms in the agriculture and fishery
industry enjoy tax exemption. Firm size (SIZE) is defined as the natural logarithm of total assets. The evidence on the relationship
between ETR and firm size in prior studies is mixed. Financial leverage (LEV) is defined as total liabilities divided by total assets, which
can capture the extent of the tax shield of debt. Return on assets (ROA) is defined as the ratio of net profit to total assets as proxies for
current profitability, and more profitable firms have stronger incentives to conduct TA. The level of tangible assets (PPE) is defined as
the percentage of fixed to total assets. A higher level of tangible assets means a larger degree of freedom to choose depreciation methods,

210
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 2
Summary statistics.
N mean SD Median Min Max

ETR1 14757 0.160 0.095 0.148 0 0.502

ETR2 13528 0.172 0.104 0.158 0 0.567
BTD 13579 0.001 0.038 0.002 0.243 0.172
DDBTD 13579 0.004 0.038 0.002 0.240 0.176
MLS 14757 0.441 0.497 0 0 1
MLS_5% 14757 0.672 0.470 0 0 1
MLS_NUM 14757 0.545 0.693 0 0 3
MLS_RATIO 14757 0.332 0.452 0 0 1.917
TAXRATE 14757 0.201 0.068 0.150 0 0.330
SIZE 14757 21.73 1.148 21.61 18.96 25.66
LEV 14757 0.450 0.198 0.453 0.113 0.821
ROA 14757 0.046 0.039 0.037 0.012 0.204
BM 14757 0.528 0.238 0.508 0.152 0.954
PPE 14757 0.242 0.170 0.209 0.002 0.754
INTANG 14757 0.045 0.050 0.032 0 0.308
INVENT 14757 0.164 0.148 0.126 0 0.728
CASH 14757 0.167 0.128 0.131 0.003 0.690
LIQUIDITY 14757 2.230 2.438 1.491 0.194 16.47
ABSDA 14757 0.064 0.061 0.047 0.001 0.323
LOSS 14757 0.078 0.268 0 0 1
BIG4 14757 0.042 0.201 0 0 1
BSIZE 14757 8.991 1.805 9 5 15
BIND 14757 0.364 0.052 0.333 0 0.556
DUAL 14757 0.201 0.401 0 0 1

This table reports the descriptive statistics of the sample firms listed on the Shanghai or Shenzhen Stock Exchanges that are contained in the China
Securities Market and Accounting Research database. All variables are defined in Table 1.

which will lower the effective tax. The level of intangible assets (INTANG) is defined as the percentage of intangible to total assets. When
firms have more intangible assets, they are more likely to have lower effective tax as a result of tax deductions. The level of inventory
(INVENT) is defined as the ratio of inventory to total assets. The level of cash (CASH) is defined as the proportion of cash and marketable
securities to total assets. Firm liquidity (LIQUIDITY) is defined as the ratio of current assets to current liabilities. To some extent, our
sample firms exhibit different financial reporting qualities, for which we need to control. Thus, we account for earnings management
using the absolute value of discretionary accruals (ABSDA). The book-to-market ratio (BM) is defined as the book value of equity divided
by the market value of equity. Growth firms receive more attention from investors and are less likely to engage in TA activities. We also
include an indicator variable for a loss (LOSS), because firms with negative profits are likely to have lower tax rate.
We also account for the impact of corporate governance characteristics on TA. This set of variables includes auditor quality (BIG4),
board size (BSIZE), the ratio of independent directors in the board (BIND), and chairman and CEO duality (DUAL). We present detailed
definitions of all variables in Table 1. We also control for industry and year fixed effects to account for fundamental differences in TA that
may exist across years and industries in Eq. (1). We estimate Eq. (1) using the ordinary least square (OLS) method and cluster standard
error at the firm level.

3.2. Data

Our sample firms are publicly listed on the Shanghai and Shenzhen Stock Exchanges in China during 2003–2016. Our sample period
begins in 2003, given the sparse availability of the top-ten shareholder data from the China Securities Market Accounting Research
database before 2003. We obtain all financial data from the China Stock Market and Accounting Research database. According to the
New Corporate Law of China promulgated in 2006, shareholders can call for interim shareholders’ meetings only when they own more
than 10% of the total shares. Hence, we define a shareholder as a large shareholder if he or she owns at least 10% of total shares, which is
also consistent with the literature (e.g., Attig et al., 2009, 2008; Jiang et al., 2018; Laeven & Levine, 2008; Maury & Pajuste, 2005). In
robustness tests, we use an alternative definition of large shareholders if he or she owns more than 5% of total shares. We also require
that sample firms must have at least one shareholder that owns at least 10% of total shares. We delete firms with top shareholders who
own more than 50% of total shares because he or she, in this case, does not need to collude with other MLS, and other MLS have limited
power and incentive to monitor him or her. Following the previous literature, we further exclude financial firms and observations with
missing values for key variables in our analyses. Our final sample has 14,757 firm-year observations. Following the common practice, we
winsorize all continuous variables at the 1% and 99% levels.

4. Results and discussion

4.1. Summary statistics and univariate analysis

We present the summary statistics in Table 2. In terms of TA, the means of ETR1 and ETR2 are 0.160 and 0.172, respectively,

211
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 3
Univariate analysis.
MLS ¼ 0 MLS ¼ 1 T-test Wilcoxon
Mean Median
N Mean Median N Mean Median

ETR1 8245 0.158 0.145 6512 0.162 0.149 0.004** 7.984***

ETR2 7573 0.169 0.156 5955 0.176 0.162 0.006*** 15.731***
BTD 7596 0.001 0.002 5983 0.001 0.002 0 0.033
DDBTD 7596 0.003 0.002 5983 0.004 0.002 0 0.047
TAXRATE 8245 0.203 0.150 6512 0.199 0.150 0.005*** 37.700***
SIZE 8245 21.76 21.68 6512 21.69 21.52 0.067*** 50.244***
LEV 8245 0.466 0.472 6512 0.429 0.426 0.038*** 85.443***
ROA 8245 0.043 0.034 6512 0.051 0.042 0.008*** 137.943***
BM 8245 0.540 0.521 6512 0.512 0.492 0.028*** 21.174***
PPE 8245 0.241 0.211 6512 0.243 0.207 0.002 1.042
INTANG 8245 0.046 0.032 6512 0.044 0.030 0.003*** 9.060***
INVENT 8245 0.174 0.134 6512 0.152 0.118 0.022*** 48.381***
CASH 8245 0.158 0.128 6512 0.178 0.137 0.020*** 20.400***
LIQUIDITY 8245 1.983 1.430 6512 2.542 1.579 0.559*** 63.442***
ABSDA 8245 0.063 0.046 6512 0.065 0.047 0.001 0.365
LOSS 8245 0.086 0 6512 0.069 0 0.017*** 14.105***
BIG4 8245 0.030 0 6512 0.058 0 0.029*** 74.313***
BSIZE 8245 8.950 9 6512 9.042 9 0.092*** 18.547***
BIND 8245 0.364 0.333 6512 0.365 0.333 0 2.441
DUAL 8245 0.193 0 6512 0.212 0 0.020*** 8.739***

This table reports the univariate statistics to compare the MLS versus non-MLS sub-samples. The last two columns report the difference between MLS
and non-MLS firms in means and medians, respectively. T-tests (Wilcoxon rank tests) are used to test the difference in means (medians). All variables
are defined in Table 1.

suggesting that the mean effective tax rates paid by the firms range from 16% to 17.2%. The means of MLS, MLS_NUM, and MLS_RATIO
are 0.441, 0.545, and 0.332, respectively, indicating that (1) approximately 44.1% of the firms in our sample have at least one top and
one non-top and large shareholders; (2) on average, each firm has 0.545 more non-top large shareholders in addition to the top
shareholder; and (3) the ratio of shares held by the other large shareholders is approximately 33.2% of those held by the top shareholder.
We provide the univariate statistics to compare the MLS versus non-MLS subsamples in Table 3. For ETR1, the means are 0.162 and
0.158 for the MLS and non-MLS subsamples respectively. For ETR2, they are 0.176 and 0.169, respectively. The differences in ETR1 and
ETR2 for the two subsamples are both significant at the 1% level, indicating that the MLS firms have higher effective tax rates than the
non-MLS firms. For control variables, we find that the MLS firms, on average, have smaller size (SIZE), lower leverage (LEV) and higher
financial performance (ROA), hold more cash (CASH), and are less likely to be financially constraint (LIQUIDITY) than the non-MLS
firms. Additionally, we find that the MLS subsamples have better corporate governance such as being more likely to be audited by
BIG4, having a larger board size, which preliminarily supports H1A. If we use medians as the criteria, the results are qualitatively the
same.
We report correlations among key variables in Table 4. Not surprisingly, the two TA measures (ETR1 and ETR2) are positively
correlated. The correlation coefficients between TA and various MLS measures are positive and significant at the 1% level. Thus, when a
firm has a large degree of MLS, it engages in less TA (it pays more tax). The preliminary findings support H1A.

4.2. Main findings

Table 5 presents the main results for the testable hypotheses. We report the results for the alternative specification without the
control variables in columns (1) and (3), and those with the control variables in columns (2) and (4). The coefficients of MLS under all
specification are positive and significant (0.004, t ¼ 1.932 for ETR1 in column (2); 0.008, t ¼ 3.290 for ETR2 in column (4)) at the 1% or
10% level, suggesting that when a firm has MLS, it engages in less TA (pays more tax). The findings support H1A. The results are also
economically significant. Using column (4), the coefficient of MLS is 0.008, that is, when a firm has non-top large shareholders (MLS ¼
1), it pays 0.8 percentage points more tax than does a firm without MLS. With the mean ETR2 of 17.2% (from Table 2), the 0.8

Table 4
Pearson correlation coefficients.
ETR1 ETR2 MLS MLS_NUM MLS_RATIO

ETR1 1.0000
ETR2 0.852*** 1.0000
MLS 0.020** 0.031*** 1.0000
MLS_NUM 0.015* 0.026*** 0.885*** 1.0000
MLS_RATIO 0.0120 0.022*** 0.827*** 0.933*** 1.0000

This table presents Pearson correlations of the key variables used in the analyses. All variables are defined in Table 1. ***, ** and * represent the 1%,
5% and 10% levels of significance, respectively, for two-tailed tests.

212
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 5
Multiple large shareholders and tax avoidance.
ETR1 ETR2

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

MLS 0.007*** 0.004* 0.011*** 0.008***

(3.141) (1.932) (4.007) (3.290)
TAXRATE 0.305*** 0.377*** 0.279*** 0.357***
(14.368) (17.758) (12.051) (15.664)
SIZE 0.002 0.004**
(1.258) (2.399)
LEV 0.104*** 0.109***
(-10.388) (-9.803)
ROA 0.272*** 0.376***
(-7.878) (-9.856)
BM 0.015* 0.026***
(-1.811) (-2.853)
PPE 0.019** 0.033***
(-2.084) (-3.180)
INTANG 0.028 0.045
(1.150) (1.613)
INVENT 0.104*** 0.125***
(8.264) (9.018)
CASH 0.128*** 0.129***
(12.105) (10.628)
LIQUIDITY 0.000 0.001
(0.119) (-1.008)
ABSDA 0.057*** 0.060***
(3.980) (3.805)
LOSS 0.024*** 0.039***
(-6.393) (-9.146)
BIG4 0.001 0.002
(-0.170) (0.350)
BSIZE 0.001 0.001
(-1.147) (-1.130)
BIND 0.005 0.012
(0.222) (-0.484)
DUAL 0.006** 0.001
(-2.438) (-0.419)
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 16077 14757 14907 13879
R2 0.129 0.226 0.110 0.201
F 107.1 55.11 80.87 50.30

This table reports the OLS regression results of the effect of MLS on TA. All models include constant term, industry and year fixed effects. All variables
are defined in Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance, respectively. All reported t-statistics are based on standard
errors adjusted for clustering at the firm level.

percentage points translate into 4.65% more in tax paid for an average firm. These findings support the monitoring hypothesis.
The signs of control variables, if significant, are as expected and generally consistent with prior literature. For instance, the co-
efficients of TAXRATE are positive and significant, indicating that the effective tax rate is positively related to a firm’s statutory tax rate.
The coefficients of LEV are negative and significant, suggesting that the effective tax rate is negatively associated with financial leverage.
The coefficient of ROA is negative and significant, suggesting that more profitable firms engage in more TA activities. BM is significantly
negative, implying that growth firms receive more attention from investors and are less likely to engage in TA activities. The coefficients
of PPE and LOSS are consistently negative and significant at the 1%, or 5% levels. If a firm has more fixed assets or incurs a loss, it has a
stronger incentive to engage in TA activities due to the need to do so (e.g., incur a loss), or the convenience of doing so (e.g., choose
depreciation methods). In contrast, the coefficients of INVENT and CASH are positive and significant. This result suggests that firms with
a high level of inventory or cash have a lower level of TA. The coefficient of ABSDA is positively related to the effective tax rate,
consistent with the notion that firms that make profit-upward adjustments pay more taxes.

4.3. Endogeneity mitigation

The results in Section 4.2 may suffer from endogeneity. Specifically, when a firm engages in less tax planning activities, it may attract
potential investors to hold more shares. Thus, the firm is more likely to have MLS. We mitigate the endogeneity concern using three
approaches.

213
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 6
Instrumental variable regressions.
FIRST STAGE SECOND STAGE

ETR1 ETR2

(1) (2) (3)

AVG_MLS 0.969***
(58.168)
PRED_MLS 0.009* 0.010*
(1.855) (1.939)
TAXRATE 0.078 0.378*** 0.357***
(-0.794) (17.782) (15.683)
SIZE 0.017* 0.002 0.004**
(1.723) (1.223) (2.389)
LEV 0.014 0.104*** 0.109***
(0.289) (-10.398) (-9.798)
ROA 0.131* 0.275*** 0.378***
(1.689) (-7.972) (-9.886)
BM 0.184*** 0.014* 0.025***
(-4.393) (-1.671) (-2.789)
PPE 0.024 0.019** 0.033***
(0.488) (-2.102) (-3.187)
INTANG 0.010 0.029 0.046
(0.086) (1.179) (1.624)
INVENT 0.177*** 0.105*** 0.125***
(-3.019) (8.308) (9.034)
CASH 0.001 0.128*** 0.129***
(-0.012) (12.100) (10.623)
LIQUIDITY 0.016*** 0.000 0.001
(4.755) (-0.056) (-1.049)
ABSDA 0.147** 0.056*** 0.060***
(2.219) (3.901) (3.768)
LOSS 0.007 0.024*** 0.039***
(0.593) (-6.366) (-9.137)
BIG4 0.127*** 0.002 0.002
(3.068) (-0.296) (0.313)
BSIZE 0.009** 0.001 0.001
(2.006) (-1.206) (-1.151)
BIND 0.235* 0.004 0.012
(1.898) (0.184) (-0.494)
DUAL 0.001 0.006** 0.001
(0.060) (-2.462) (-0.432)
INDUSTRY Yes Yes Yes
YEAR Yes Yes Yes
N 14757 14757 13879
R2 0.257 0.143 0.127
F 3386.6 54.52 49.08
Kleibergen-Paap rk 546.3

This table presents estimates of two-stage least square (2SLS) model. We use the average value of MLS (except the firm itself) in the same
industry, province, and year as the instrument for MLS in the first-stage regression. We label the variable as AVG_MLS and use it to
generate a predicted MLS (PRED_MLS) in Eq. (1). Other variables are defined in Table 1. ***, ** and * represent the 1%, 5% and 10%
levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the
firm level.

4.3.1. Instrumental variables

We first use an instrumental variable approach to mitigate the concern of endogeneity between MLS and TA. Specifically, we follow
Jiang et al. (2018, 2019) and use the mean value of MLS in the same industry, province, and year (except the firm itself, AVG_MLS) as the
instrumental variable for MLS. Since firms in the same province likely encounter the same group of wealthy investors and may expe-
rience a “spillover effect” from other firms in the same industry, they may be more likely to have a similar ownership structure (Jiang,
Cai, Jiang, & Nofsinger, 2019), therefore AVG_MLS is a reasonable instrument for MLS. We use AVG_MLS to generate a predicted MLS
(PRED_MLS) in Eq. (1).
Column (1) of Table 6 reports the first stage results. We find that the coefficient of AVG_MLS is positive and significant at 1% level,
indicating that AVG_MLS is a suitable instrument variable. More importantly, the coefficients of PRED_MLS in the second stage are both
positive and significant at the 10% level, suggesting that firms with MLS engage in less TA activities, which is consistent with the main
findings in Table 5.

4.3.2. Firm fixed effect

The results in Table 5 suggest that MLS help mitigate TA. However, the above results may be driven by some omitted unobservable

214
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 7
Fixed effect model.
ETR1 ETR2

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

MLS 0.007*** 0.008*** 0.006* 0.006**

(2.586) (2.810) (1.881) (1.978)
TAXRATE 0.324*** 0.297*** 0.329*** 0.309***
(13.055) (11.909) (12.436) (11.634)
SIZE 0.004 0.004 0.003 0.003
(-1.403) (-1.436) (1.150) (0.901)
LEV 0.095*** 0.089*** 0.095*** 0.091***
(-7.716) (-7.417) (-7.246) (-7.091)
ROA 0.417*** 0.401*** 0.559*** 0.550***
(-10.611) (-10.331) (-12.599) (-12.387)
BM 0.016* 0.007 0.026*** 0.017*
(-1.951) (-0.812) (-2.801) (-1.856)
PPE 0.031** 0.032*** 0.036*** 0.034**
(-2.512) (-2.602) (-2.678) (-2.503)
INTANG 0.063* 0.066* 0.060 0.056
(-1.855) (-1.933) (-1.618) (-1.455)
INVENT 0.072*** 0.071*** 0.092*** 0.089***
(4.359) (4.270) (4.875) (4.756)
CASH 0.068*** 0.065*** 0.068*** 0.065***
(5.731) (5.376) (5.084) (4.770)
LIQUIDITY 0.001 0.001 0.000 0.000
(1.196) (1.531) (-0.100) (0.119)
ABSDA 0.052*** 0.052*** 0.053*** 0.052***
(3.888) (3.929) (3.445) (3.470)
LOSS 0.017*** 0.018*** 0.025*** 0.026***
(-4.528) (-4.818) (-5.865) (-6.089)
BIG4 0.006 0.006 0.006 0.007
(-0.756) (-0.651) (0.648) (0.729)
BSIZE 0.001 0.001 0.001 0.001
(-0.872) (-0.962) (-0.523) (-0.500)
BIND 0.027 0.014 0.031 0.014
(1.036) (0.543) (1.119) (0.508)
DUAL 0.004 0.005 0.002 0.002
(-1.366) (-1.631) (-0.496) (-0.683)
FIRM Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
PROVINCE No Yes No Yes
INDUSTRY-YEAR No Yes No Yes
PROVINCE-YEAR No Yes No Yes
N 14523 14523 13693 13693
R2 0.534 0.545 0.524 0.533
F 27.84 293.2 26.59 265.3

This table reports the OLS regression results of the effect of MLS on TA, controlling for addition firm fixed effect, province fixed effect, industry-by-year
fixed effect, region-by-year fixed effect. All models include constant term, industry and year fixed effects. All variables are defined in Table 1. ***, **
and * represent the 1%, 5% and 10% levels of significance, respectively. All reported t-statistics are based on standard errors adjusted for clustering at
the firm level.

variables that simultaneously affect MLS and TA. Hence, we add firm fixed effect into Eq. (1) to control for the effect of omitted time-
invariant firm characteristics. In addition, following recent research (Gormley & Matsa, 2014, 2017), we further augment Eq. (1) by
adding province fixed effect, industry-by-year fixed effect, and province-by-year fixed effect to alleviate the omitted variable bias due to
confounding factors related to industry and region trends.
Table 7 shows the results of fixed effect regression. Column (1) and (3) include firm fixed effect, and Column (2) and (4) control for
province fixed effect, industry-by-quarter fixed effect, and province-by-quarter fixed effect in addition. Across all specification, the
coefficients on MLS are positive and significant at the 1%, 5%, or 10% levels, suggesting that our results in Table 5 are unlikely to be
driven by omitted variables or industry and region trends.

4.3.3. Difference-in-differences analysis

To further address endogeneity concerns, we rely on a quasi-natural experiment and use a difference-in-differences (DID) approach.
Following Jiang et al. (2018), we employ the split-share reform of 2005 as an exogenous shock to examine how the change in the
structure of the MLS composition before and after the Reform affects TA. In the Reform, the Chinese government encouraged
non-tradable shareholders (top and non-top large shareholders) to convert their non-tradable shares into tradable ones. To do so,
non-tradable shareholders must negotiate a compensation plan with tradable shareholders. Li, Wang, Cheung, and Jiang (2011)

215
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 8
Split share structure reform in 2005 as an exogenous shock.
ETR1 ETR2

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

TREAT*POST 0.030* 0.033*

(-1.732) (-1.835)
POST 0.082*** 0.014
(-2.866) (-0.424)
TREAT*BEFORE1 0.028 0.023
(-1.215) (-1.003)
TREAT*AFTER1 0.047* 0.038
(-1.915) (-1.535)
TREAT*AFTER2 0.035 0.038
(-1.407) (-1.479)
TREAT*AFTER3 0.054** 0.061**
(-2.249) (-2.455)
BEFORE1 0.021 0.000
(0.672) (-0.015)
AFTER1 0.032 0.034
(-0.426) (-0.456)
AFTER2 0.019 0.034
(-0.183) (-0.350)
AFTER3 0.029 0.034
(0.227) (-0.284)
Controls Yes Yes Yes Yes
FIRM Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 652 652 652 652
R2 0.708 0.712 0.688 0.690
F 5.356 4.338 4.227 3.668

This table presents estimates of difference-in-differences regressions in which the dependent variables are ETR1 and ETR2, respectively. POST is an
indicator variable that equals one for firm i in year t as well as afterward three years if firm i has the split share structure reform in year t (and zero
otherwise). BEFORE1 is an indicator variable that equals to one for the year before split share structure reform, and zero otherwise. AFTER1, AFTER2,
AFTER3 is an indicator variable that equals to one for the first, second, and third year after split share structure reform, and zero otherwise. TREAT is an
indicator variable that equals one for firm if it had MLS in the three years before the split share structure reform but it did not have MLS three years after
the split share structure reform and that equals zero for firm if a firm had MLS in the three years before the split share structure reform and it had MLS
three years after the split share structure reform. Other variables are defined in Table 1. ***, ** and * represent the 1%, 5% and 10% levels of sig-
nificance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level.

document that the average compensation is approximately 30% of shares previously held by non-tradable shareholders. Consequently,
some of the top and non-top large shareholders no longer held at least 10% of total shares after the Reform. In addition, the China
Securities and Regulatory Commission required that large shareholders (both top and non-top shareholder) who cannot sell, can sell up
to 5%, or can sell up to 10% of their newly converted shares until 12 months, 24 months, and 36 months, respectively, after the Reform.
The shareholder ownership changes resulted in few MLS in the post-Reform period, and it occured within the lockup period of the
Reform. Since the decrease in MLS firms<LS is exogenous to the impact of MLS on TA, if the monitoring hypothesis is valid, we would
expect TA activities to increase after the Reform.
Specifically, we define an indicator variable of POST with a value of one if a firm is within three years post-Reform (excluding the
reform year) and zero for the three years pre-Reform. Then, we define a firm as a treatment firm (TREAT ¼ 1) if it had MLS in the three
years before the Reform, but had no MLS within three years after the Reform. In contrast, if a firm had MLS in the three years before and
after the Reform, the TREAT takes a value of zero. Thus, we augment Eq. (1) as follows:

ETR1i,t (or ETR2i,t) ¼ β0 þ β1 TREAT*POSTi,t þ β2 POSTi,tþ CONTROL þ YEAR þ Firm þ πi,t (2)

If the results supporting H1A remain intact, we expect the coefficients of β1 to be negative and significant. It means that when a firm
receives “treatment” (had MLS before the Reform but no MLS after the Reform), it pays less tax after the Reform.
The identifying assumption for the DID model Eq. (2) is that the treatment and control groups follow “parallel trends." Following
Bertrand and Mullainathan (2003), we estimate the following regression to evaluate this assumption:

ETR1i;t ETR2i;t ¼ α0 þ α1 BEFORE1t þ α2 AFTER1t þ α3 AFTER2t þ α4 AFTER3t þ α5 TREATi  BEFORE1t
þα6 TREATi  AFTER1t þ α7 TREATi  AFTER2t þ α8 TREATi  AFTER3t (3)
þCONTROLit þ YEAR þ Firm þ μi;t

Specifically, we generate the following four year indicators: BEFORE1, AFTER1, AFTER2, and AFTER3. BEFORE1 is an indicator
variable that equals to one for the year before the split share structure reform and zero otherwise. AFTER1, AFTER2, and AFTER3 are
indicator variables that equal one for the first, second, and third year after Reform, respectivel, and zero otherwise. If the Reform causes

216
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 9
Financial constraints.
Panel A: Size

Low FC High FC

ETR1 ETR2 ETR1 ETR2

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

MLS 0.008** 0.013*** 0.001 0.004

(2.437) (3.783) (0.487) (1.182)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 7150 6794 7607 7085
R2 0.263 0.240 0.213 0.185
F 29.51 28.32 33.81 28.16

Panel B: Firm age

Low FC High FC
ETR1 ETR2 ETR1 ETR2
(1) (2) (3) (4)

MLS 0.006* 0.009** 0.005* 0.007**

(1.777) (2.328) (1.881) (2.542)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 7019 6584 7738 7295
R2 0.207 0.196 0.266 0.223
F 22.48 20.97 40.11 36.86

Panel C: Cash dividend

Low FC High FC
ETR1 ETR2 ETR1 ETR2
(1) (2) (3) (4)

MLS 0.005* 0.007** 0.003 0.008**

(1.934) (2.488) (0.786) (2.052)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 7895 7350 6862 6529
R2 0.312 0.265 0.194 0.177
F 42.41 34.50 26.60 25.60

This table investigates the impact of the financial constraints on the relationship between MLS and TA. We proxy a firm’s financial constraints by firm
size, firm age, and cash dividend. We use the year median to classify high versus low sub-samples of financial constraints. Other variables are defined in
Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on
standard errors adjusted for clustering at the firm level.

a change in firms’ MLS and TA, we would observe insignificant coefficients on TREAT*BEFORE1, but negative and significant co-
efficients on TREAT *AFTER1, TREAT *AFTER2, and TREAT *AFTER3.
We present the results for Eq. (2) in columns (1) and (3) of Table 8. The coefficients of TREAT*POST are negative and significant at
the 10% levels in columns (1) and (3), suggesting that when a firm had no MLS after the Reform, TA increases, as reflected in the lower
tax payment. Columns (2) and (4) display the regression results of Eq. (3). The coefficient estimates for TREAT*BEFORE1 are statistically
insignificant, which suggests that the parallel trends assumption is satisfied. The coefficients on TREAT*AFTER1 and TREAT*AFTER3 are
negative and significant at the 5% or 10% level, indicating that firms engage in more TA in one or more years after the Reform. Overall,
the results in Table 8 confirm the validity of our identification strategy and the causal effect of MLS on TA.

4.4. Excluding alternative explanations: financial constraints and covering collusion by reducing TA

In the hypothesis development section, we argue that TA activities create a shield for the diversion of rents (Chan et al., 2016; Desai
and Dharmapala, 2006, 2009), while the control contestability or monitoring by the MLS alleviates firms’ agency costs and thus re-
straints TA. However, another explanation may be that when corporate insiders and controlling shareholders pursue private benefits
through transfer pricing, related-party transactions, and so on, outside investors (both shareholders and creditors) are less willing to
invest in these firms because they face the risk that the return on investment will never materialize (Claessens et al., 2000; La Porta et al.,
1999; Lin, Ma, & Xuan, 2011). Thus, such firms become more financially constrained due to more costly external finance. The presence
of MLS reduces firms’ agency costs and information asymmetry (Attig et al., 2008; Jiang et al., 2018), and alleviates the financial
constraints. Some argue that a financially unconstrained firm is less likely to engage in TA activities to generate additional internal funds
(Edwards et al., 2016). Taken together, financial constraints may be alternative explanation of the relationship between MLS and TA.

217
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 10
Robustness checks.
MLS_5% Book-tax difference Additional control Lag(MLS)

ETR1 ETR2 BTD DDBTD ETR1 ETR2 ETR1 ETR2

(1) (2) (3) (4) (5) (6) (7) (8)

MLS 0.004* 0.008** 0.006*** 0.011***

(2.731) (2.731) (2.869) (4.092)
MLS_5% 0.003 0.009***
(1.337) (3.189)
MLSi,t-1 0.003** 0.008***
(2.324) (3.170)
TAXRATE 0.377*** 0.357*** 0.135*** 0.135*** 0.374*** 0.345*** 0.377*** 0.349***
(17.760) (15.673) (17.318) (17.302) (16.226) (13.593) (26.510) (14.840)
SIZE 0.002 0.005** 0.005*** 0.005*** 0.001 0.004* 0.003*** 0.006***
(1.309) (2.496) (-7.152) (-7.180) (0.754) (1.842) (2.733) (2.893)
LEV 0.103*** 0.108*** 0.004 0.004 0.101*** 0.104*** 0.101*** 0.106***
(-10.367) (-9.765) (1.195) (1.204) (-9.994) (-9.117) (-14.964) (-9.443)
ROA 0.271*** 0.375*** 0.612*** 0.613*** 0.275*** 0.379*** 0.276*** 0.385***
(-7.861) (-9.837) (66.093) (66.013) (-7.886) (-9.879) (-10.592) (-10.010)
BM 0.015* 0.026*** 0.041*** 0.041*** 0.016* 0.024** 0.019*** 0.030***
(-1.843) (-2.851) (12.626) (12.673) (-1.856) (-2.511) (-3.342) (-3.222)
PPE 0.019** 0.033*** 0.006* 0.006* 0.029*** 0.045*** 0.023*** 0.037***
(-2.074) (-3.162) (-1.864) (-1.847) (-3.088) (-4.202) (-3.886) (-3.482)
INTANG 0.027 0.042 0.037*** 0.037*** 0.048* 0.069** 0.036** 0.052*
(1.105) (1.514) (-4.296) (-4.289) (1.952) (2.405) (2.157) (1.800)
INVENT 0.104*** 0.124*** 0.019*** 0.018*** 0.093*** 0.116*** 0.104*** 0.128***
(8.229) (8.994) (-4.230) (-4.214) (7.135) (8.107) (12.668) (9.134)
CASH 0.128*** 0.128*** 0.040*** 0.040*** 0.120*** 0.123*** 0.125*** 0.129***
(12.094) (10.592) (-8.157) (-8.156) (10.967) (9.674) (15.660) (10.232)
LIQUIDITY 0.000 0.001 0.001** 0.001** 0.000 0.001 0.000 0.001
(0.211) (-0.892) (2.286) (2.283) (0.550) (-0.850) (0.436) (-0.858)
ABSDA 0.057*** 0.060*** 0.049*** 0.049*** 0.045*** 0.050*** 0.061*** 0.068***
(3.982) (3.785) (-6.180) (-6.219) (3.065) (3.036) (4.725) (4.232)
LOSS 0.025*** 0.039*** 0.007*** 0.007*** 0.023*** 0.039*** 0.023*** 0.038***
(-6.386) (-9.098) (5.696) (5.678) (-5.794) (-8.600) (-5.931) (-8.501)
BIG4 0.001 0.003 0.004 0.004 0.005 0.000 0.002 0.001
(-0.100) (0.432) (1.361) (1.362) (-0.782) (-0.006) (-0.517) (0.206)
BSIZE 0.001 0.001 0.000 0.000 0.001 0.001 0.001*** 0.001
(-1.167) (-1.213) (1.460) (1.464) (-1.370) (-1.157) (-2.674) (-1.464)
BIND 0.006 0.010 0.014* 0.014* 0.008 0.011 0.001 0.016
(0.258) (-0.418) (1.753) (1.760) (0.372) (-0.432) (0.081) (-0.631)
DUAL 0.006** 0.001 0.000 0.000 0.006** 0.001 0.007*** 0.002
(-2.474) (-0.525) (0.166) (0.167) (-2.386) (-0.353) (-3.838) (-0.626)
TOP1 0.000** 0.000**
(2.025) (2.081)
SOE 0.003 0.002
(1.168) (0.475)
R&D 0.001*** 0.000
(-3.260) (-1.044) Yes Yes
INDUSTRY Yes Yes Yes Yes Yes Yes Yes Yes
YEAR Yes Yes Yes Yes Yes Yes 13907 13044
N 14757 13879 15880 15880 13055 12180 0.228 0.202
R2 0.225 0.201 0.517 0.518 0.230 0.202 75.84 47.32
F 54.71 50.01 402.9 400.9 44.63 40.54 Yes Yes

This table reports the robustness checks on the MLS and TA. We redefine a shareholder as a large shareholder if it owns more than 5% of total shares,
uses two alternative TA proxy, includes additional control, and considerable delay impact of MLS (MLSi,t-1). Other variables are defined in Table 1. ***,
** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on standard errors
adjusted for clustering at the firm level.

To address this concern, we partition the full sample into low and high financial constraints subsamples. If the financial constraints
channel dominates, then we should observe a more pronounced effect in the high financial constraints subsample; otherwise, the agency
mitigation hypothesis explains the result. Following Fazzari, Hubbard, and Petersen (1988), Almeida, Campello, and Weisbach (2004),
Hadlock and Pierce (2010), we proxy a firm’s financial constraints by firm size, firm age and cash dividend.1 We hypothesize that large,
old, and high dividend paying firms are less financially constrained than their counterparts are. We use the year medians of the

1
We get qualitatively similar results when we use interest coverage ratio and size-age index proposed by Hadlock and Pierce (2010) to proxy for
financial constraints.

218
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 11
The strength of MLS.
ETR1 ETR2 ETR1 ETR2

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

MLS_NUM 0.003* 0.005***

(1.778) (3.018)
MLS_RATIO 0.004* 0.008***
(1.710) (2.869)
TAXRATE 0.377*** 0.357*** 0.377*** 0.357***
(17.755) (15.657) (17.761) (15.667)
SIZE 0.002 0.004** 0.002 0.004**
(1.249) (2.384) (1.265) (2.403)
LEV 0.103*** 0.108*** 0.103*** 0.108***
(-10.367) (-9.772) (-10.374) (-9.783)
ROA 0.271*** 0.375*** 0.271*** 0.374***
(-7.867) (-9.827) (-7.858) (-9.809)
BM 0.015* 0.026*** 0.015* 0.026***
(-1.817) (-2.868) (-1.830) (-2.891)
PPE 0.019** 0.033*** 0.019** 0.033***
(-2.080) (-3.174) (-2.059) (-3.142)
INTANG 0.027 0.044 0.027 0.044
(1.129) (1.578) (1.120) (1.561)
INVENT 0.104*** 0.124*** 0.104*** 0.125***
(8.251) (8.992) (8.263) (9.018)
CASH 0.128*** 0.129*** 0.128*** 0.130***
(12.123) (10.643) (12.133) (10.665)
LIQUIDITY 0.000 0.001 0.000 0.001
(0.108) (-1.030) (0.100) (-1.035)
ABSDA 0.057*** 0.060*** 0.057*** 0.060***
(3.982) (3.808) (3.972) (3.796)
LOSS 0.024*** 0.039*** 0.025*** 0.039***
(-6.393) (-9.141) (-6.397) (-9.147)
BIG4 0.001 0.003 0.001 0.002
(-0.154) (0.367) (-0.172) (0.347)
BSIZE 0.001 0.001 0.001 0.001
(-1.142) (-1.117) (-1.144) (-1.110)
BIND 0.005 0.012 0.005 0.011
(0.231) (-0.469) (0.232) (-0.464)
DUAL 0.006** 0.001 0.006** 0.001
(-2.417) (-0.379) (-2.414) (-0.378)
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 14757 13879 14757 13879
R2 0.226 0.201 0.226 0.201
F 54.93 49.68 54.62 49.49

This table examines the impact on firms’ TA of the strength of MLS. We measure the strength of MLS using MLS_NUM and MLS_RATIO, where MLS_NUM
is the number of other large shareholders and MLS_RATIO is the ratio of the sum of total outstanding shares held by non-top large shareholders to the
number of shares held by the top shareholder. Other variables are defined in Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance,
respectively, for two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level.

respective variable as the cutoff point to classify high versus low financial constraints.
We presents the results are in Table 9. For brevity, we do not present the coefficients of the control variables. As shown in Panel A of
Table 9, the coefficients of MLS are positive and significant at the 1% level for the low financial constraint sub-sample, while the same
coefficients are insignificant for high financial constraint constraints sub-sample, indicating that the impact of MLS is more pronounced
for firms with low financial constraints and contradicting the financial constraint hypothesis. In Panel B and Panel C, we find that the
coefficients on MLS are significant and with similar magnitudes in both subsamples, suggesting that the effect of MLS does not depend on
financial constraints. Combined, the results in Table 9 indicate that financial constraints do not dictate the relationship between MLS
and TA.
Our main results demonstrate that MLS can reduce corporate TA, though there may be an alternative explanation that MLS and the
controlling shareholder conspire to conduct earnings management or other hollowing out activities in order to cover up their bad
behaviors and avoid attracting the attention of regulators, thereby reducing TA. Therefore, to exclude this possible explanation, we
examine the effect of MLS on earnings management and tunneling behavior. The untabulated results show that the presence of MLS can
significantly reduce earnings management and tunneling behavior, indicating that MLS play a supervisory role rather than a collusion
role.

219
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 12
Entrenchment motivation.
Panel A: SOE vs Non-SOE

SOE Non-SOE

ETR1 ETR2 ETR1 ETR2

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

MLS 0.003 0.004 0.005** 0.010***

(0.712) (0.857) (2.063) (3.586)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 3862 3671 10894 10207
R2 0.271 0.254 0.214 0.190
F 23.97 20.62 41.41 39.45

Panel B: Other receivables

Low High
ETR1 ETR2 ETR1 ETR2
(1) (2) (3) (4)

MLS 0.002 0.005 0.009*** 0.014***

(0.879) (1.540) (2.864) (3.983)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 8021 7399 7404 6947
R2 0.233 0.201 0.186 0.171
F 41.01 38.75 22.52 20.38

Panel C: Related party transactions

Low High
ETR1 ETR2 ETR1 ETR2
(1) (2) (3) (4)

MLS 0.001 0.003 0.008*** 0.014***

(0.334) (0.944) (2.798) (4.280)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 7794 7275 6811 6452
R2 0.235 0.204 0.221 0.207
F 36.38 30.02 26.13 28.07

This table investigates the impact of the severity of the agency problem on the relationship between MLS and TA. For brevity, we do not report the
coefficients of the control variables. We proxy a firm’s the severity of agency problem by whether it was owned by the state (SOE) or whether the firm
had more other receivables or related party transactions. We use the year sample median to classify high versus low sub-samples of other receivables or
related party transactions. Other variables are defined in Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance, respectively, for
two-tailed tests. All reported t-statistics are based on standard errors adjusted for clustering at the firm level.

4.5. Robustness checks

We conduct several robustness checks on the main findings in Table 5. (1) We redefine a shareholder as a non-top large shareholder if
he or she owns more than 5% of total outstanding shares (MLS_5%) instead of 10% as a threshold. (2) In the main regression, we use the
effective tax rate to proxy for TA. In this subsection, we also employ two alternative measures of TA: the book-tax difference (BTD)
measure proposed by Manzon and Plesko (2002) and the book-tax difference measure (DDBTD) developed by Desai and Dharmapala
(2006). (3) Since top shareholders’ ownership determines their power and incentive to engage in TA, we control for the percentage of
shares owned by the largest shareholder (TOP1) and whether a firm is a state-owned enterprise (SOE). Additionally, we control for the
ratio of R&D expenses to sale revenues (R&D), as China offers an additional deduction for R&D costs, which influence the tax rate. (4)
We lag MLS by one period to test the delayed TA responses to MLS. Table 10 reports the results of the robustness check.
The results in Table 10 show that the coefficients of MLS, MLS_5%, and lag one period MLS are consistently positive and significant at
the 10% or above level across all columns, suggesting that the impact of MLS on TA is robust.
In addition, considering econometric procedures, corporate governance systems due to B/H share markets, and the Chinese tax
reform in 2008, we also re-estimate Eq. (1) using a double firm and time cluster of errors using a subsample that excludes B- and H-share
firms, and a subsample using only post-2007 firm years. The untabulated results are consistent with those in Table 5.

4.6. Additional analysis

4.6.1. Strength of MLS

Our main and robust findings in Tables 5–10 support the monitoring hypothesis of MLS. If the logic is correct, then the strength of

220
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 13
The role of external monitoring.
Panel A: Big4 vs non-Big4

Big4 Non-Big4

ETR1 ETR2 ETR1 ETR2

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

MLS 0.003 0.005 0.004* 0.008***

(-0.331) (0.489) (1.872) (3.115)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 623 544 14133 13334
R2 0.381 0.363 0.224 0.199
F 6.101 4.513 56.17 51.62

Panel B: Herfindahl-Hirschman index

High competition Low competition
ETR1 ETR2 ETR1 ETR2
(1) (2) (3) (4)

MLS 0.002 0.006 0.005* 0.009***

(0.575) (1.289) (1.925) (3.055)
Controls Yes Yes Yes Yes
INDUSTRY Yes Yes Yes Yes
YEAR Yes Yes Yes Yes
N 623 544 14133 13334
R2 0.381 0.363 0.224 0.199
F 6.101 4.513 56.17 51.62

This table investigates the impact of the external corporate governance on the relationship between MLS and TA. For brevity, we do not report the
coefficients of the control variables. We proxy a firm’s external corporate government by whether it was audited by a “Big4” auditor (BIG4) and
whether it was in the competitive product market environment. We use a Big4 auditor as better audit quality. We use a high Herfindahl-Hirschman
index as less competitive in the industry. We use the full sample median of the index to conduct the classification. Other variables are defined in
Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance, respectively, for two-tailed tests. All reported t-statistics are based on
standard errors adjusted for clustering at the firm level.

MLS should matter. Specifically, we measure the strength of MLS using MLS_NUM and MLS_RATIO, where MLS_NUM is the number of
non-top large shareholders in MLS, and MLS_RATIO is the ratio of the sum of the total shares held by non-top large shareholders to the
number of shares held by the top shareholder. Thus, we use MLS_NUM and MLS_RATIO to replace MLS in Eq. (1). We expect that when
the number of non-top large shareholders is greater (MLS_NUM is larger) or when MLS own more shares (MLS_RATIO is larger), the
monitoring effect is stronger; thus, the impact of MLS on TA magnifies.
We report the results in Table 11. The coefficients of MLS_NUM and MLS_RATIO are positive and significant at the 1% level across
columns (1) to (4), and they are economically significant. For instance, using column (4), the coefficient of MLS_RATIO is 0.008. When
there is one standard deviation increase of MLS_RATIO (0.452), there will be a 3.62% standard deviation increase of TA. The findings in
Table 11 corroborate those in Table 5.

4.6.2. Moderating factors

The core logic of our study is that controlling shareholders shelter their private benefits with TA or directly expropriate tax savings,
and MLS can effectively monitor them. If the logic is correct, then we expect that the monitoring effect of MLS is more significant when
the entrenchment motivation of controlling shareholders have a stronger entrenchment motivation. We believe that state-ownership,
related-party transactions, and other receivables can capture the severity of the entrenchment motivation of controlling shareholders
(Chan et al., 2016; Jian & Wong, 2010; Jiang et al., 2010). Specifically, when a firm is state-owned (SOE), the type II agency problem is
less severe than when it is not (Jiang et al., 2010). Furthermore, SOEs are less likely to engage in TA activities because taxes are actually
dividends rather than costs to their controlling shareholders (Bradshaw, Liao, & Ma, 2018). Hence, SOEs are less likely to engage in tax
planning activities to pursue private benefits. Jian and Wong (2010) and Jiang et al. (2010) find that controlling shareholders often
transfer corporate resources from one firm to another through related-party transactions, which increases listed firms’ receivables and
inter-corporate loans. Therefore, we use the year median of related-party transactions and other receivables to classify high versus low
tunneling subsamples.
The results in Panel A of Table 12 suggest that only the coefficients of MLS for the non-SOE, the high related-party transactions, and
high other receivables subsamples are positive and significant at the 1% or 5% level, while those of SOE or the low related party
transactions, and low other receivables subsamples are not significant. Therefore, the impact of MLS is more pronounced when the
entrenchment motivation of controlling shareholders is stronger.
In addition, in the presence of external corporate governance, MLS should have a different effect on TA. Because of the internal
corporate governance function of MLS, we focus on external corporate moderating factors. Operationally, we consider auditor quality
and product market competition. If auditor quality is weak or product market competition is low, then the external corporate

221
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Table 14
Multiple large shareholders, tax avoidance, and firm value.
Full sample MLS ¼ 1 MLS ¼ 0

(1) (2) (3) (4) (5) (6)

ETR1 0.349** 0.479** 0.240

(2.403) (2.243) (1.258)
ETR2 0.302** 0.441** 0.168
(2.300) (2.311) (0.971)
TAXRATE 0.338* 0.303 0.596** 0.465 0.228 0.262
(-1.673) (-1.496) (-1.982) (-1.538) (-0.891) (-1.031)
SIZE 0.657*** 0.656*** 0.584*** 0.584*** 0.718*** 0.714***
(-33.343) (-34.142) (-20.514) (-20.388) (-28.719) (-29.743)
LEV 0.803*** 0.843*** 1.096*** 1.173*** 0.573*** 0.593***
(-7.860) (-8.522) (-7.916) (-8.139) (-4.367) (-4.796)
ROA 9.783*** 9.953*** 9.985*** 9.999*** 9.214*** 9.501***
(19.172) (19.541) (12.633) (12.519) (15.902) (16.389)
SGOW 0.187*** 0.172*** 0.209*** 0.213*** 0.157** 0.126**
(3.905) (3.525) (2.780) (2.783) (2.566) (2.001)
BIG4 0.424*** 0.422*** 0.399*** 0.383*** 0.316*** 0.335***
(5.395) (5.334) (3.940) (3.823) (2.918) (2.923)
BSIZE 0.010 0.010 0.009 0.012 0.011 0.007
(1.179) (1.127) (0.714) (1.006) (1.002) (0.620)
BIND 1.130*** 1.078*** 1.284*** 1.242*** 1.009*** 0.949***
(4.374) (4.134) (3.391) (3.253) (2.936) (2.748)
DUAL 0.090*** 0.090*** 0.012 0.017 0.145*** 0.142***
(2.685) (2.698) (0.271) (0.371) (3.145) (3.150)
INTERCEPT 15.590*** 15.613*** 14.136*** 14.136*** 16.844*** 16.869***
(37.336) (38.266) (23.723) (23.734) (31.640) (32.532)
INDUSTRY Yes Yes Yes Yes Yes Yes
YEAR Yes Yes Yes Yes Yes Yes
N 13544 13131 5959 5748 7584 7382
R2 0.595 0.598 0.611 0.613 0.589 0.593
F 310.2 321.6 151.0 152.7 188.2 197.0

This table presents the results on the value implication of TA. The dependent variable is Tobin’s Q in year tþ1. All models include constant term,
industry and year fixed effects. All variables are defined in Table 1. ***, ** and * represent the 1%, 5% and 10% levels of significance, respectively. All
reported t-statistics are based on standard errors adjusted for clustering at the firm level.

governance is weak (Lee, Byun, & Park, 2019). Therefore, the impact of MLS on TA magnifies.
We use the presence of a Big 4 auditor to indicate better audit quality, suggesting that it faces one source of strong external corporate
governance. Similarly, we calculate the Herfindahl-Hirschman index in the industry to which a firm belongs. A high index means that
the industry is less competitive. We use the full sample median of the index for the classification. The results in Table 13 indicate that the
coefficients of MLS are positive and significant at the 1% or 10% level for the subsample of non-Big 4 auditors and the low product
market competition subsamples. Hence, in the presence of a weak external corporate governance, MLS have a stronger impact on TA.

4.7. The value implications of TA

TA activities that reduce transfers from shareholders to the government should generally enhance shareholder wealth. However, the
obfuscatory feature of TA can create a shield for insider opportunism and rent diversion (Chan et al., 2016; Desai and Dharmapala, 2006,
2009). If the hidden agency costs substantially outweigh the potential tax saving from TA activities, then outside investors will discount
firm with aggressive TA, leading to a negative association between TA and firm value.
Since MLS either compete for control (Attig et al., 2008, 2009) or monitor the controlling shareholders, reducing information
asymmetry and managerial opportunism, they therefore alleviate outside investors’ concerns about the potential agency problems
underlying TA. Combined, we reason that, as MLS reduce agency conflict, TA activities are more likely to benefit minority shareholders.
Specifically, following Li et al. (2017), we employ the following regression model to test the association between TA and firm value:

TQi,tþ1 ¼ δ0 þ δ1ETR1i,t (or ETR2i,t) þ CONTROL þ INDUSTRY þ YEAR þ εi,t (4)

Consistent with Li et al. (2017), we use Tobin’s Q (TQ) to proxy for firm value. The control variables included in the model are: firms’
statutory tax rate (TAXRATE), firm size (SIZE), financial leverage (LEV), return on assets (ROA), sales revenue growth (SGROW), big 4
auditing (BIG4), board size (BSIZE), the percentage of outside director in boards (BIND), whether the chairman of and the CEO are the
same individual (DUAL). We also include industry and year effect, and cluster standard error at the firm level.
Table 14 displays the regression results for the relationships among TA, MLS, and firm value. Columns (1) and (2) report the results
for the whole sample, and columns (3) to (6) report the results for the subsample with and without MLS. As columns (1) and (2) show,
the coefficients of ETR1 and ETR2 are positive and significant at the 5% level, suggesting that firms with a highertax burden (or lower
TA) have high firm value. In terms of economic significance, a one-standard-deviation increase in ETR1 translates into a 1.68%

222
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

(0.349*0.095/1.975) rise in a firm’s value relative to the sample mean. When we further differentiate between firms with and without
MLS, we find that the coefficients of ETR1 and ETR2 are significant only in the subsample with MLS. The magnitude of the coefficients of
MLS in the subsample with MLS is twice (or more) as large as those without MLS, indicating that with the monitoring by MLS, TA
activities are more likely to be value-enhancing. Overall, the results in Table 14 are in line with the agency view of TA, and good
governance can alleviate outside investors’ concerns about agency conflicts.

5. Summary

We examine the impact of MLS on a firm’s TA. While the TA literature suggests that agency problems between executives and
shareholders drive TA activities, the impact of MLS on TA is underexplored. We develp two competing hypotheses. The first is the
monitoring hypothesis, in which the presence of MLS means that there are more large shareholders to monitor the top shareholder.
Thus, the type II agency problem is less severe. Consequently, the firm engages in less TA. The second is the collusion hypothesis suggests
that MLS may collude with the top shareholder to expropriate minority shareholders. The type II agency problem becomes severe,
resulting in more TA. Our robust evidence is more consistent with the monitoring hypothesis. In addition, when MLS are stronger in
terms of the number of non-top large shareholders and their ownership, the impact of MLS on TA magnifies. Essentially, MLS mean more
sets of eyes from more non-top large shareholders to scrutinize the top shareholder, suggesting that their vested interests in the firm
drive them to protect their investment rather than to collude with the top shareholder to extract rent from minority shareholders.
Additional analyses on the moderating role of the agency problem environment and external corporate governance suggest that the
negative impact of MLS on TA is more pronounced when the entrenchment motivation is severe or when external corporate governance
is weak. The additional findings corroborate the monitoring hypothesis. MLS play a corporate governance role in mitigating the adverse
effect of the agency problem such that TA is less severe. In line with the agency view of TA, the value-enhancing implications of TA exist
only in firms with MLS. Therefore, with the monitoring role of MLS, TA activities are more likely to be value-enhancing.

CRediT authorship contribution statement

Caiyue Ouyang: Conceptualization, Validation, Resources, Writing - original draft, Supervision, Funding acquisition. Jiacai Xiong:
Software, Data curation, Validation, Writing - original draft. Kun Huang: Methodology, Formal analysis, Validation, Resources, Writing
- original draft, Writing - review & editing.

Acknowledgements

Caiyue Ouyang acknowledges the financial support from the Fundamental Research Funds for the Central Universities
(2018RCW014).

References

Allen, F., Qian, J., & Qian, M. (2005). Law, finance, and economic growth in China. Journal of Financial Economics, 77(1), 57–116.
Almeida, H., Campello, M., & Weisbach, M. S. (2004). The cash flow sensitivity of cash. The Journal of Finance, 59, 1777–1804.
Arena, M., P., Wang, B., & Yang, R. (2019). Securities litigation and corporate tax avoidance. Journal of Corporate Finance. https://doi.org/10.1016/
j.jcorpfin.2019.101546. In press.
Ashbaugh-Skaife, H., Collins, D. W., & LaFond, R. (2006). The effects of corporate governance on firms’ credit ratings. Journal of Accounting and Economics, 42(1–2),
203–243.
Attig, N., El Ghoul, S., & Guedhami, O. (2009). Do multiple large shareholders play a corporate governance role? Evidence from East Asia. Journal of Financial Research,
32(4), 395–422.
Attig, N., Guedhami, O., & Mishra, D. (2008). Multiple large shareholders, control contests, and implied cost of equity. Journal of Corporate Finance, 14(5), 721–737.
Atwood, T. J., & Lewellen, C. (2019). The complementarity between tax avoidance and manager diversion: Evidence from tax haven firms. Contemporary Accounting
Research, 36(1), 259–294.
Badertscher, B. A., Katz, S. P., & Rego, S. O. (2013). The separation of ownership and control and corporate tax avoidance. Journal of Accounting and Economics, 56(2),
228–250.
Balakrishnan, K., Blouin, J. L., & Guay, W. R. (2019). Tax aggressiveness and corporate transparency. The Accounting Review, 94(1), 45–69.
Bauer, A., M., Fang, J., Pittman, J., Zhang, Y., & Zhao, Y. (2019). How aggressive tax planning facilitates the diversion of corporate resources: Evidence from path
analysis. Contemporary Accounting Research. https://doi.org/10.1111/1911-3846.12563. In press.
Ben-Nasr, H., Boubaker, S., & Rouatbi, W. (2015). Ownership structure, control contestability, and corporate debt maturity. Journal of Corporate Finance, 35, 265–285.
Bennedsen, M., & Wolfenzon, D. (2000). The balance of power in closely held corporations. Journal of Financial Economics, 58(1–2), 113–139.
Bertrand, M., & Mullainathan, S. (2003). Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy, 111(5), 1043–1075.
Boubaker, S., & Sami, H. (2011). Multiple large shareholders and earnings informativeness. Review of Accounting and Finance, 10(3), 246–266.
Bradshaw, M., Liao, G., & Ma, M. S. (2018). Agency costs and tax planning when the government is a major Shareholder. Journal of Accounting and Economics, 67(2–3),
255–277.
Cai, C. X., Hillier, D., & Wang, J. (2016). The cost of multiple large shareholders. Financial Management, 45(2), 401–430.
Chan, K. H., Mo, P. L. L., & Tang, T. (2016). Tax avoidance and tunneling: Empirical analysis from an agency perspective. Journal of International Accounting Research,
15(3), 49–66.
Chen, S., Chen, X., Cheng, Q., & Shevlin, T. (2010). Are family firms more tax aggressive than non-family firms? Journal of Financial Economics, 95(1), 41–61.
Cheng, C. A., Huang, H. H., Li, Y., & Stanfield, J. (2012). The effect of hedge fund activism on corporate tax avoidance. The Accounting Review, 87(5), 1493–1526.
Cheng, M., Lin, B., & Wei, M. (2013). How does the relationship between multiple large shareholders affect corporate valuations? Evidence from China. Journal of
Economics and Business, 70, 43–70.
Chung, S. G., Goh, B. W., Lee, J., & Shevlin, T. (2019). Corporate tax aggressiveness and insider trading. Contemporary Accounting Research, 36(1), 230–258.
Claessens, S., Djankov, S., & Lang, L. H. (2000). The separation of ownership and control in East Asian corporations. Journal of Financial Economics, 58(1–2), 81–112.
Desai, M. (2004). The degradation of corporate profits. Harvard University. Working paper.

223
C. Ouyang et al. International Review of Economics and Finance 67 (2020) 207–224

Desai, M. A., & Dharmapala, D. (2006). Corporate tax avoidance and high-powered incentives. Journal of Financial Economics, 79(1), 145–179.
Desai, M. A., & Dharmapala, D. (2009). Corporate tax avoidance and firm value. The Review of Economics and Statistics, 91(3), 537–546.
Desai, M. A., Dyck, A., & Zingales, L. (2007). Theft and taxes. Journal of Financial Economics, 84(3), 591–623.
Edmans, A., & Manso, G. (2010). Governance through trading and intervention: A theory of multiple blockholders. Review of Financial Studies, 24(7), 2395–2428.
Edwards, A., Schwab, C., & Shevlin, T. (2016). Financial constraints and cash tax savings. The Accounting Review, 91(3), 859–881.
Enriques, L., & Volpin, P. (2007). Corporate governance reforms in continental Europe. The Journal of Economic Perspectives, 21(1), 117–140.
Fang, Y., Hu, M., & Yang, Q. (2018). Do executives benefit from shareholder disputes? Evidence from multiple large shareholders in Chinese listed firms. Journal of
Corporate Finance, 51, 275–315.
Fazzari, S., Hubbard, G. R., & Petersen, B. (1988). Financing constraints and corporate investment. Brookings Papers on Economic Activity, 1, 141–195.
Gormley, T. A., & Matsa, D. A. (2014). Common errors: How to (and not to) control for unobserved heterogeneity. Review of Financial Studies, 27, 617–661.
Gormley, T. A., & Matsa, D. A. (2017). Playing it safe? Managerial preferences, risk, and agency conflicts. Journal of Financial Economics, 122, 431–455.
Guti
errez, L. H., & Pombo, C. (2009). Corporate ownership and control contestability in emerging markets: The case of Colombia. Journal of Economics and Business,
61(2), 112–139.
Hadlock, C., & Pierce, J. (2010). New evidence on measuring financial constraints: Moving beyond the KZ Index. Review of Financial Studies, 23, 1909–1940.
He, X., Wong, T. J., & Young, D. (2012). Challenges for implementation of fair value accounting in emerging markets: Evidence from China. Contemporary Accounting
Research, 29(2), 538–562.
He, Y., Chiu, Y., & Zhang, B. (2015). The impact of corporate governance on state-owned and non-state-owned firms efficiency in China. The North American Journal of
Economics and Finance, 33, 252–277.
Hope, O. K. (2013). Large shareholders and accounting research. China Journal of Accounting Research, 6(1), 3–20.
Hribar, P., & Collins, D. W. (2002). Errors in estimating accruals: Implications for empirical research. Journal of Accounting Research, 40, 105–134.
Jiang, F., Cai, X., Jiang, Z., & Nofsinger, J. (2019). Multiple large shareholders and dividends: Evidence from China. Pacific-Basin Finance Journal, 57, 101201.
Jiang, F., Cai, W., Wang, X., & Zhu, B. (2018). Multiple large shareholders and corporate investment: Evidence from China. Journal of Corporate Finance, 50, 66–83.
Jiang, G., Lee, C. M., & Yue, H. (2010). Tunneling through intercorporate loans: The China experience. Journal of Financial Economics, 98(1), 1–20.
Jian, M., & Wong, T. Y. (2010). Propping through related party transactions. Review of Accounting Studies, 15, 70–105.
Kahn, C., & Winton, A. (1998). Ownership structure, speculation, and shareholder intervention. The Journal of Finance, 53(1), 99–129.
Khan, M., Srinivasan, S., & Tan, L. (2017). Institutional ownership and corporate tax avoidance: New evidence. The Accounting Review, 92(2), 101–122.
Kimber, D., & Lipton, P. (2005). Corporate governance and business ethics in the Asia-Pacific region. Business & Society, 44(2), 178–210.
Kim, J. B., Li, Y., & Zhang, L. (2011). Corporate tax avoidance and stock price crash risk: Firm-level analysis. Journal of Financial Economics, 100(3), 639–662.
Kothari, S. P., Leone, A. J., & Wasley, C. E. (2005). Performance matched discretionary accrual measures. Journal of Accounting and Economics, 39(1), 163–197.
La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (1999). Corporate ownership around the world. The Journal of Finance, 54(2), 471–517.
Laeven, L., & Levine, R. (2008). Complex ownership structures and corporate valuations. Review of Financial Studies, 21(2), 579–604.
Lee, J. H., Byun, H. S., & Park, K. S. (2019). How does product market competition affect corporate takeover in an emerging economy. International Review of Economics
& Finance, 60, 26–45.
Li, O. Z., Liu, H., & Ni, C. (2017). Controlling shareholders’ incentive and corporate tax avoidance: A natural experiment in China. Journal of Business Finance &
Accounting, 44(5–6), 697–727.
Lin, C., Ma, Y., & Xuan, Y. (2011). Ownership structure and financial constraints: Evidence from a structural estimation. Journal of Financial Economics, 102, 416–431.
Lin, K. Z., Mills, L. F., Zhang, F., & Li, Y. (2018). Do political connections weaken tax enforcement effectiveness? Contemporary Accounting Research, 35(4), 1941–1972.
Liu, X., Shi, H., & Ferrantino, M. (2016). Tax evasion through trade intermediation: Evidence from Chinese exporters. International Review of Economics & Finance, 42,
518–535.
Li, K., Wang, T., Cheung, Y. L., & Jiang, P. (2011). Privatization and risk sharing: Evidence from the split share structure reform in China. Review of Financial Studies,
24(7), 2499–2525.
Lou, F., Wang, J., & Yuan, H. (2014). Causes and consequences of corporate asset exchanges by listed companies in China. International Review of Economics & Finance,
31, 205–217.
Ma, J., & Khanna, T. (2016). Independent directors’ dissent on boards: Evidence from listed companies in China. Strategic Management Journal, 37(8), 1547–1557.
Manzon, G., & Plesko, G. (2002). The relation between financial and tax reporting measures of income tax. Tax Law Review, 55, 175–214.
Maury, B., & Pajuste, A. (2005). Multiple large shareholders and firm value. Journal of Banking & Finance, 29(7), 1813–1834.
McGuire, S. T., Wang, D., & Wilson, R. J. (2014). Dual class ownership and tax avoidance. The Accounting Review, 89(4), 1487–1516.
Mishra, D. R. (2011). Multiple large shareholders and corporate risk taking: Evidence from East Asia. Corporate Governance: An International Review, 19(6), 507–528.
Pagano, M., & R€ oell, A. (1998). The choice of stock ownership structure: Agency costs, monitoring, and the decision to go public. Quarterly Journal of Economics, 113(1),
187–225.
Porcano, T. (1986). Corporate tax rates: Progressive, proportional, or regressive. Journal of American Taxation Association, 7(2), 17–31.
Scholes, M., Wolfson, M., Erickson, M., Maydew, E., & Shevlin, T. (2005). Taxes and business strategy: A planning approach (3rd ed.). Upper Saddle River, NJ: Pearson
Prentice Hall.
Slemrod, J. (2004). Are corporate tax rates, or countries, converging? Journal of Public Economics, 88(6), 1169–1186.
Xia, C., Cao, C., & Chan, K., C. (2017). Social trust environment and firm tax avoidance: Evidence from China. The North American Journal of Economics and Finance, 42,
374–392.
Zwiebel, J. (1995). Block investment and partial benefits of corporate control. The Review of Economic Studies, 62(2), 161–185.

224