Professional Documents
Culture Documents
Can Book-Tax Differences Capture Earnings Management and Tax Management? Empirical Evidence From China
Can Book-Tax Differences Capture Earnings Management and Tax Management? Empirical Evidence From China
Can Book-Tax Differences Capture Earnings Management and Tax Management? Empirical Evidence From China
; Firth, Michael
Working Paper
Can Book-Tax Differences Capture Earnings
Management and Tax Management? Empirical
Evidence from China
Reference: Tang, Tanya Y. H./Firth, Michael (2011). Can Book-Tax Differences Capture Earnings
Management and Tax Management? Empirical Evidence from China. [S.l.] : SSRN.
https://ssrn.com/abstract=1679190.
Kontakt/Contact
ZBW – Leibniz-Informationszentrum Wirtschaft/Leibniz Information Centre for Economics
Düsternbrooker Weg 120
24105 Kiel (Germany)
E-Mail: rights[at]zbw.eu
https://www.zbw.eu/econis-archiv/
ABSTRACT
This study investigates whether and how book-tax differences (BTDs) are related to
companies. Using unique tax-effect BTDs obtained from Chinese B-share listed firms, we
find that firms with strong incentives for earnings and tax management exhibit high levels
of abnormal BTDs. This suggests that BTDs can be used to capture both accounting and
tax manipulations induced by managerial motivations. Our results indicate that earnings
management explains 7.4% of abnormal BTDs, tax management accounts for 27.8% of
abnormal BTDs, and their interaction explains 3.2% of abnormal BTDs. Tax-effect BTDs
are more powerful than income-effect BTDs in capturing opportunistic reporting at both
An early version of this paper was titled: “Book-tax-differences, a proxy for earnings and
tax management”. We thank A. Rashad Abdel-khalik (the editor), In-Mu Haw (the co-
editor), and an anonymous reviewer for their extensive comments and suggestions. We
are especially grateful to Terry Shevlin for his long-standing support and guidance on this
paper. We also gratefully acknowledge the helpful comments from Millicent Chang, Sharon
Cox, Lillian Mills, John Robinson, Casey Schwab, and workshop participants at the 2006
American Accounting Association Annual Meeting, Simon Fraser University, York
University, and University of British Columbia.
Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1679190
http://ssrn.com/abstract=1679190
Can Book-Tax Differences Capture Earnings Management and Tax Management?
Empirical Evidence from China
1. INTRODUCTION
A large body of literature from the U.S. indicates that managers have strong
incentives to engage in earnings and tax management. 1 For example, they manage
market pricing, and government and stock exchange regulations (Healy and Wahlen,
1999; Fields et al., 2001), and they manage taxes to maximize shareholders’ returns
(Scholes et al., 2005; Swenson, 1999), reduce the risk of tax scrutiny and political cost
(Watts and Zimmerman, 1986; Fields et al., 2001), and satisfy tax-based contract
management (EM) and tax management (TM) is important in assessing the quality of
earnings and studying management behavior. Some studies have demonstrated that
book-tax differences (BTDs) reflect earnings management and earnings quality (Mills and
Newberry, 2001; Phillips et al., 2003; Hanlon, 2005), while other studies show that BTDs
are associated with tax sheltering (Shevlin, 2002; McGill and Outslay, 2004; Wilson, 2009;
Frank et al., 2009). However, very few studies examine the simultaneous impacts of EM,
In this study we examine BTDs, EM and TM in China. In doing so, we extend the
literature in terms of research method, empirical design, and country coverage. We use
tax-effect BTDs that are calculated directly from financial statement disclosures. In
contrast, U.S. research has used income-effect BTDs, which are based on estimations that
involve “grossing up” tax payments. These estimations result in well-know measurement
errors that are impossible to avoid (Hanlon, 2003). Estimation problems are eliminated
when we use tax-effect BTDs and we show they are a much better measure of BTDs and
capture opportunistic reporting in a more powerful way, at least in the case of China. In
their review paper, Graham et al. (2009) conclude that BTDs can be a function of earnings
Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1679190
http://ssrn.com/abstract=1679190
management, tax planning, regulatory differences, or some other factors. They note that,
while numerous studies document BTDs as evidence of EM and TM, almost all of them
use aggregated BTDs. They suggest that future research should study the disaggregated
components of BTDs to better explain the underlying causes. In this paper, we distinguish
decompose BTDs into normal BTDs (NBTDs) and abnormal BTDs (ABTDs). NBTDs are
explained by differences between GAAP and tax law and we include these factors in the
regression model of BTDs. ABTDs are the result of EM and TM and we use the residual
from the BTD regression model as a proxy for ABTDs. We examine whether firms with
strong incentives for EM and TM have higher ABTDs. Consistent with our argument that
disaggregated BTDs not only inform us of the magnitude of the mechanical differences
between book and tax reporting rules, but also EM and TM, we provide evidence that
abnormal BTDs are associated with EM and TM incentives but normal BTDs are not. Our
results suggest that simply using total BTDs as a proxy for EM and TM may lead to an
overstatement of EM and TM. In addition, evidence that abnormal BTDs can be explained
by EM, TM, and their interaction provides a caveat to researchers when using this
institutional settings. For example, China has variable statutory tax rates, separate tax
and rigid capital issuance requirements. These characteristics are very different from those
in the U.S. and they each have an influence on EM and TM. Differences in tax laws
between China and the U.S., and differences in institutional factors, imply that the results
from U.S. studies cannot be automatically imputed to the China setting. This provides a
motivation for our research. We incorporate the special features of China’s corporate
sector and tax law into our model. Examples include the number of different tax rates that
context to our research design and interpretation of results, we make references to the
China is an important country in which to carry out our research as it has the third
largest economy in the world, the second largest market capitalization, and has unique
institutional characteristics. Inward portfolio investment from the U.S. and other countries
is accelerating as international investors search for growth markets with fast expanding
domestic demand. One impediment to this growth is investors’ real concerns over the
quality of accounting information, which they need in order to make good investment
decisions (Haw et al., 2001; Sami and Zhou, 2004). Our study will highlight some aspects
of earnings management and tax management in China and thus will contribute to
Our empirical results are based on data drawn from Chinese listed firms that have
foreign investors (B-share firms) as these companies have detailed tax notes that allow us
to calculate tax-effect BTDs. The results show that abnormal BTDs (ABTDs) are
significantly related to a set of earnings and tax management incentives. We find that EM
explains 7.4% of ABTDs, TM accounts for 27.8% of ABTDs, and the combined EM and TM
factor explains 3.2% of ABTDs. These findings suggest that ABTDs are a useful proxy for
both EM and TM in China. Our results are robust to a series of sensitivity tests, including
losses incentive, the rights issues incentive test, the proportion of temporary and
The results provide important insights into the different EM and TM practices that
have emerged in response to the unique institutional settings in the world’s largest
transitional economy, China. Our findings have implications for policymakers, auditors,
international investors, and users of financial reports of listed Chinese companies. BTDs
can help investors estimate and evaluate the quality of financial statements, and tax
firms. The analysis of BTDs may also assist China’s regulators in their quest to enhance
the credibility of book and tax reporting and the efficiency of China’s capital markets. To
the best of our knowledge, this is the first study to examine whether and how both EM and
The remainder of the paper is organized as follows. The next section discusses prior
Chinese BTDs in terms of its development, measurement, and features. We also discuss
Sections 3 and 4 describe the research design, sample selection, and summary statistics.
The fifth section presents empirical results and the last section concludes.
BTDs refer to the gap between the pre-tax income reported in a company’s published
financial statements (hereafter book income) and the taxable income reported to the tax
authorities. The conflicting objectives of accounting and tax rules lead to divergent
requirements in income reporting, which generate mechanical BTDs (Smith and Butters,
1949; Beresford et al., 1983). However, accounting rules and tax laws cannot specify the
accounting and tax treatment for every business transaction because business activities
are complex and continually changing. This necessarily leaves considerable uncertainty in
applying accounting standards and tax laws. For example, in an effort to achieve a variety
of social, economic, and political goals, tax laws often vary across time, industries, and
even firms.
GAAP, per se, permits considerable discretion in reporting practices (Fields et al.,
2001; Watts and Zimmerman, 1986; Manzon and Plesko, 2002). For example, the
expenses to recognize in any given period. These standards also permit flexibility in the
choice of accounting methods, such as depreciation, cost allocation, and asset valuation.
When managers have incentives to distort a firm’s performance, they may opportunistically
use accounting standards and tax laws, and thereby create abnormal BTDs.
Drawing on U.S. data, Mills and Newberry (2001) present evidence that firms with
EM incentives have larger BTDs. Phillips et al. (2003) evaluate the usefulness of BTDs in
constant, they find that temporary BTDs are incrementally useful beyond total accruals and
earnings declines or are in danger of reporting losses. Hanlon (2005) finds that earnings
and the accrual portion of earnings are less persistent for one-year-ahead earnings for
firms with large temporary BTDs, suggesting that extreme temporary BTDs indicate lower
A growing body of literature (Shevlin, 2002; Desai and Dharmapala, 2006; Plesko,
2004; McGill and Outslay, 2004) suggests a link between BTDs and TM. The underlying
idea is that inconsistent financial accounting and tax reporting should be a characteristic of
tax planning, because the goal of tax planning is to reduce tax payments. If taxable income
is managed without affecting book income, the variation in BTD is reflective of tax
shelters.3 Some U.S. studies (Manzon and Plesko, 2002; Mills et al., 2003; Plesko, 2004)
find that aggregate BTDs have increased over the 1990s, but this growth cannot be
unexplained BTDs may be partly attributable to increased tax sheltering activities. Wilson
(2009) and Frank et al. (2009) find evidence of a positive association between BTDs and
identified tax shelter firms. Mills (1998) finds that proposed audit adjustments increase as
firms’ BTDs widen, suggesting that larger positive BTDs imply high levels of tax non-
simultaneously and interact with each other (Shackelford and Shevlin, 2001). Current
studies indicate that firms engaged in aggressive book reporting may be also involved in
aggressive tax reporting (Frank et al., 2009). The propensity to engage in opportunistic
These strategies can be summarized as follows: (1) managing book and taxable income
(taxes) in an opposite direction, e.g., reporting higher earnings and lower taxable income
(taxes); (2) managing book income while keeping taxable income (taxes) constant, e.g.,
boosting earnings or taking a bath; (3) managing taxable income (taxes) while keeping
book income constant, e.g., reducing (deferring) taxes or smoothing taxes.4 However, no
matter what strategy management chooses, the ultimate purposes of EM and TM are to
influence book income and tax payments. These actions increase the variation in BTDs
because BTDs are a function of tax-effect book income and tax-effect taxable income. The
resulting variations can help quantify the magnitude of opportunistic book and tax
reporting. Therefore, BTDs inform us not only of the magnitude of the mechanical
discrepancy between book and tax reporting rules, but also of management’s strategies or
Unlike the U.S. and many other developed nations, which have a long history of BTDs,
China’s BTDs emerged when the Chinese accounting and tax systems changed from a
in 1985. Before that, the rules for measuring accounting profit were the same as those for
measuring taxable income, so no BTDs existed in China (Tang et al., 2000; Davidson et al.,
1996). In the transition to a modern business enterprise system, China changed its
accounting standards to comply with IFRS and imported elements from international tax
laws, giving rise to a lot of mechanical BTDs (Tang, 2006). Appendix I lists the major
differences between corporate income tax laws and Chinese GAAP. The harmonization
with IFRS has dual implications: (1) it increases the relevance and comparability of
the departure of book income from taxable income provides opportunities for firms to
aggressively report their book income and taxable income simultaneously (Desai, 2005).
Figures 1 and 2 show the patterns of BTDs, permanent differences, and temporary
differences, over time. We identify three major features of Chinese BTDs. First, unlike U.S.
BTDs, aggregate Chinese BTDs are negative during our observation period.5 Tang (2006)
interprets the negative aggregate BTDs as evidence that Chinese income tax laws have
more conservative expense recognition relative to Chinese GAAP and IFRS. Examples of
advertisement fees, entertainment fees, lower tax depreciation vis-a-vis book depreciation,
and numerous non-deductible items such as the provision for impairment of assets and
expenses without authorized supporting invoices. Second, the pattern of BTDs is mainly
differences to total BTDs is 76% for the sample we use in the main tests. Appendix I
details different book and tax treatments as reflected in permanent and temporary
revenues) are non-deductable (recognized) under tax laws, resulting in a high proportion
explained solely by changes in accounting and tax rules during our sample period.
BTDs.
discretion in current accounting standards. For example, Eccher and Healy (2000)
document that Chinese managers are more likely to exercise greater discretion in
reporting accruals, particularly through the write-down of obsolete inventory and accounts
receivable allowance under IFRS. Aharony et al. (2000) conclude that accelerating credit
sales is a widely used method for Chinese B-share firms’ earnings management before
making an IPO. Haw at al. (2005) and Chen and Yuan (2004) provide evidence that
Chinese listed firms manage earnings upward to gain approval to make a rights issue.
Some studies attribute these phenomena to the lack of effective controls that monitor
legal protection for minority investors (DeFond et al., 2000; Abdel-khalik et al., 1999;
Eccher and Healy, 2000; Lin and Chan, 2000; Aharony et al., 2000). The instruments of
earnings management used in China include, but are not limited to, accruals (Lu, 1999),
non-operating earnings or below-the-line items (Chen and Yuan, 2004; Haw et al., 2005),
related party transactions (Jian and Wong, 2010, Lo et al., 2010a, b), and government
and local governments’ interests (Tang, 2006). For example, the Chinese government
offers favorable tax incentives to domestic firms operating in five special economic zones,
32 economic and technology development zones, 13 free trade zones, and 52 high-tech
development zones (Wu et al., 2007). As a result of that, actual tax rate applied to firms
that enjoy preference policies could be 24%, 15%, 7.5% and even 0, much lower than the
nominal statutory tax rate of 33%. The variation in tax burdens within a country and the
lack of efficiency in tax administration induce firms to engage in tax planning (Chan and
registration location to a low tax rate region. Shevlin et al. (2009) demonstrate that
Chinese listed firms save taxes through shifting income from high tax rate subsidiaries to
low tax rate subsidiaries. In particular, Chan et al. (2009) find that tax noncompliance
increases when the level of book and tax conformity decreases. We incorporate these
3. RESEARCH DESIGN
We can measure BTDs in two ways: (1) using book income minus taxable income or the
sum of permanent differences and temporary differences, which is called income-effect BTDs;
and (2) using prima facie income tax expense (i.e., the multiplication of book income by the
statutory tax rate) minus current tax expense (or the sum of the multiplication of the statutory
tax rate by the permanent differences and the multiplication of the statutory tax rate by the
temporary differences), which is called tax-effect BTDs. A common method to estimate BTDs
in the U.S studies is to gross-up current tax expense to estimate taxable income, or to use
the effective tax rate reconciliation to infer total BTDs. The income-effect method from this
estimation introduces measurement error due to credits, tax rate differences, consolidation,
and tax loss carry forwards (see Hanlon, 2003 for a detailed discussion). Appendix II gives
an example of computing tax-effect BTDs and income-effect BTDs. The numbers are quite
different.
Tax-effect BTDs are particularly appropriate in China because firms are subject to
varying tax rates due to different government tax incentives and because they are required
to declare their corporate income tax on an individual firm tax reporting basis. Furthermore,
tax-effect BTDs allow researchers to examine those tax strategies that reduce the overall
tax burden without affecting total book and taxable income (i.e., income shifting).
Conversely, the use of income-effect BTDs may restrict empirical tests as it only can
10
further, we test whether our EM and TM results are affected by the way we calculate BTDs
We are able to use tax-effect BTDs in this study because they are disclosed in B-share
listed firms’ financial statement tax footnotes. China’s B-shares are mainly traded and held
Ministry of Finance (MOF) and the CSRC, B-share firms must prepare financial reports in
two languages and use two sets of accounting standards: the Chinese-version under
Chinese GAAP and the English-version under IFRS. A reconciliation of tax-effect book
income and current tax expense is presented in the IFRS-based reports (see Appendix III
for an example). These data enable us to avoid the measurement errors inherent in having
Most prior BTD studies focus on either temporary differences (e.g., Phillips et al.,
2003; Hanlon, 2005) or permanent differences (Frank et al., 2009) and examine either EM
or TM behavior. However, EM/TM can create both permanent differences and temporary
differences (Wilson, 2009; Hanlon and Heitzman, 2009). Using a sample of firms identified
as engaging in tax shelter activities, Wilson (2009) analyzes nine major types of tax shelter
activities and identifies that three of them create temporary differences. The results in his
Table 4 indicate that both temporary and permanent BTDs are correlated with identified tax
shelter firms. As a result, we use total BTDs that consist of both permanent and temporary
Prior literature has used a residual approach to disentangle the different components
in BTDs. For example, Desai and Dharmapala (2006) use the residual from a regression of
total BTDs on total accruals to estimate a measure of tax planning and assume that BTDs
are a result of EM and TM. Frank et al. (2009) construct a tax aggressiveness measure by
11
differences between accounting and tax rules. In doing so, they restrict their measure to
tax aggressiveness that does not generate temporary differences. In this paper, we follow
their notion and develop ABTD and NBTD measures that take account of the uniqueness
of Chinese accounting and tax systems. As discussed in Section 2.1, BTDs are a function
of accounting-tax misalignment, EM, and TM. To estimate NBTDs, we regress total BTDs
on nondiscretionary items that are known to cause NBTDs but are less likely to relate to
Manzon and Plesko (2002), we use changes in sales, gross property, plant and equipment
(PPE), non-goodwill intangible assets, and NOL to explain NBTDs. We also add tax rate
where
TAX _ DIFFit = the difference between the consolidated company’s applicable tax rate
NBTDs are driven by the mechanical divergence between accounting standards and
tax laws and are subject to investment scale in fixed and intangible assets (ΔINV),
economic growth (ΔREV), tax loss positions (TL and TLU), and tax rate differences
12
method, useful life, and residual values for specific assets. However, for financial reporting
purposes, these items are all subject to management choice based on Chinese GAAP and
IFRS. Given that tax depreciation is slower than book depreciation in China, taxable
income will be larger than book income in the earlier years of the assets’ lives. When the
calculate depreciation and amortization, thereby giving rise to large negative NBTDs. In
addition, investment growth will increase the provision for impairment of fixed and
intangibles assets. Provisions for impairments are shown in the income statement as
required under existing accounting standards but they are non-deductible under Chinese
tax laws; this leads to a higher (lower) taxable income compared to book income in the
earlier (later) years. Therefore, investment growth will lead to large negative NBTDs in
China, which is quite different from the U.S. setting. We use INVit to capture the effect of
the growth in investment scale on NBTDs that relate to mechanical depreciation and
amortization.8
NBTDs.9 Usually, accounting rules tend to prevent companies from overstating revenue or
understating expenses to financial statement users and thus protect third party creditors
and investors. By contrast, tax laws tend to constrain taxpayers from understating revenue
revenue. Thus, growth in revenue will increase NBTDs due to different income and expense
recognition in GAAP and tax laws. As discussed earlier, Chinese income tax laws are more
Wilkie (1992) indicates that BTDs will be understated when a tax loss occurs and
overstated when a tax loss is utilized. For example, when a tax loss occurs, taxable income
13
year's tax losses are offset against the current year's taxable income, thereby resulting in
positive NBTDs. We use TLit and TLUit variables to control for the mechanical effects of
tax policies specific to tax loss carry forwards and the asymmetric treatment of taxable
Finally, we use TAX _ DIFFit to control for the mechanical effect of tax rate
differentials and the different reporting requirements on NBTDs. Unlike the U.S. tax code,
Chinese tax laws require separate tax reporting. Each entity must compute its income tax
on the basis of its independent taxable income and the applicable tax rate. Therefore,
applicable tax rate variation among subsidiaries in a consolidated group affects NBTDs. 11
All variables are scaled by total assets for year t to control for firm size except for
TAX _ DIFFit .
3.3 Testing the Relations between ABTDs and Incentives for EM and TM
NBTDs are the fitted values from equation (1) and the residuals are abnormal book-
tax differences (ABTDs). To test whether ABTDs can capture EM and TM, we develop the
following multiple regression model in which estimated ABTDs are regressed on a set of
variables that proxy for various EM and TM incentives. If ABTDs are indicative of potential
managerial manipulations, firms with stronger incentives and opportunities for EM and TM
where
Dependent variable:
EM and TM can yield negative or positive ABTDs. For our study, signed ABTDs are not
appropriate measures of manipulations since a firm with positive ABTDs does not necessarily
14
TM incentives variables:
ATR = applicable tax rate for the sample listed firm as shown in the tax note;
EM incentives variables:
SEON = dummy variable that equals 1 when a consolidated entity has a rights issue or
LOSS = dummy variable that equals 1 when a consolidated entity has a loss in the
TM and EM variables:
Control variables:
IND = control variable that equals 1 for the manufacturing industry and 0
otherwise;
The tax planning literature suggests that the significant factors that affect TM include
the corporate tax rate (Klassen et al., 1993; Klassen and Shackelford, 1998), non-tax costs
consideration (Scholes et al., 2005), and different tax treatments or jurisdictions (Klassen
A major feature of Chinese corporate income tax is that the income tax rate varies
across firms with different investors and firms established in different industries and
locations. The applicable income tax rate (ATR) for listed firms’ ranges from 0% to 33%.
This encourages firms with a higher tax rate to engage in tax planning. Chan and Mo
(2000) find that Chinese firms with a high tax rate have high tax audit adjustments,
15
Based on non-tax cost considerations, a well-designed tax strategy can reduce taxes
but does not diminish income reported to investors. Consequently, income-shifting through
exploiting different tax treatments in a consolidated group is an ideal tax strategy due to it
having little or no impact on a firm’s reported profit. Some U.S.-based studies demonstrate
that U.S. multinational and multi-state companies lower their taxes by cross-jurisdiction
income shifting (Klassen et al., 1993; Gupta and Mills, 2002). These studies argue that
firms with different tax rates can strategically arrange operations, capital, and product
flows by shifting income from high tax brackets to low tax brackets. In contrast, firms with a
single tax treatment have no opportunity to pursue these potential tax strategies (Mills et
al., 1998).
Chinese companies with subsidiaries in different regions and industries are commonly
subject to multiple tax rates, making tax-induced income shifting among subsidiaries
possible. In particular, under Chinese income tax laws, all corporations must declare their
IFRS and Chinese GAAP. Thus, Chinese listed firms have strong incentives and
opportunities to shift income from subsidiaries with a high tax rate to those with a low tax
rate. Shevlin et al. (2009) find evidence that the magnitude of income shifting in Chinese
listed firms increases with the tax rate range in a consolidated group. The greater the
differences in tax rates among subsidiaries, the more likely a holding company engages in
the time of equity offerings (Teoh et al., 1998; Marquardt and Wiedman, 2004). Many
studies in China (mostly in Chinese) also show that listed firms inflate earnings to reach
16
can be explained by Chinese listed firms’ heavy reliance on equity financing and the high
threshold for making SEOs set by the CSRC (see Aharony et al., 2000). Jian and Wong
(2010) find that China’s listed firms use recurring related party transactions to manage
operating earnings upward to meet the SEO hurdle. Chen et al. (2000) and Haw et al.
(2005) show that firms with ROEs in the range of 10 to 11 percent have unusually high
discretionary items such as abnormal accruals and non-core profits. Following these
studies, we predict that SEO applicants (SEON firms) face an immediate incentive to
improve profits through EM in year t so as to gain SEO eligibility in year t+1 and this will
While numerous studies show that avoiding a loss is a strong incentive for EM (e.g.,
Burgstahler and Dichev, 1997), Chinese listed firms exhibit an even stronger motivation to
report positive earnings because of the CSRC’s delisting and trading restrictions
regulations.13 Much of the literature demonstrates that Chinese firms are keen to report
positive earnings to avoid delisting from the stock market (Cai et al., 2003). Lu (1999)
demonstrates that loss firms manipulate their earnings upward or downward in different
time periods to avoid delisting and trading restrictions. Accordingly, we expect that loss
firms have strong incentives for EM, thereby giving rise to larger ABTDs.
The tax and non-tax trade-offs literature suggests that EM and TM may exist
simultaneously and interact with each other (Shackelford and Shevlin, 2001). We include
and TM incentives. A distinct characteristic of Chinese capital markets is that the share
ownership is highly concentrated in the hands of the central and local governments. The
state is the largest shareholder in approximately 73% of B-share listed firms. The state’s
shares are administered by government bodies such as state asset management agencies
or institutions authorized to hold shares on behalf of the state, such as wholly state-owned
investment companies (CSRC, 2005). The government frequently appoints top managers
17
managers may have more incentives to act in the best interests of the state rather than in
the best interest of public shareholders (Firth et al., 2006). In addition, concentrated
and this can lead to poor corporate governance (Liu and Lu, 2007). By examining firms
that had allegations of accounting manipulation made by the CSRC during 1994 to 2002,
Liu and Du (2003) demonstrate that state-owned enterprises (SOEs) are more likely to
manage earnings.
Some listed firms are controlled by the central government (SOECG) and some are
controlled by local governments (SOELG).14 Local governments play multiple roles such as
providing public services, monitoring listed firms as China’s central government agents,
and being major shareholders of listed firms (Chen et al., 2008). This complicated
relationship and conflicts of interest between local governments and the central
administration in China create a unique incentive for SOELGs to engage in both EM and
TM. On the one hand, local governments want to attract more investment to develop the
local economy, increase employment opportunities, and create regional revenue as the
performance of the local economy affects government officers’ promotions and political
careers (Li, 1998). Chen et al. (2008) find that local governments use fiscal subsidies as a
means to help their listed firms boost earnings to meet the threshold for making a SEO and
to avoid delisting. Their evidence shows that 21% of SEO qualified firms are unqualified
before fiscal subsidies and 96% of firms with a slight loss become profitable after receiving
fiscal subsidies.
On the other hand, local governments have strong incentives to encourage their listed
firms to engage in tax planning for their own interests. As the largest shareholders, local
governments are the largest beneficiaries of high post-tax returns. Further, the tax-sharing
system in China requires that income tax paid by a state controlled listed firm must be
shared with the central government.15 In this context, SOELG firms may have incentives to
18
Recent studies show that local governments provide their listed firms with generous
tax preferences as a means to attract and retain more equity capital into their own
jurisdictions (e.g., Wu et al., 2007; Chen et al., 2008). In an untabulated test, we find that
the means of number of tax rates (NUMBER) and ATR for SOELG firms are 3.11 and 15%,
in contrast to 2.4 and 20% for SOECGs, suggesting that SOELG firms have more tax
incentives and lower applicable tax rates.16 Tax incentives have contrasting effects on the
tax planning decisions of SOELG firms. Tax rate differentials facilitate tax aggressiveness
in a consolidated group. SOELG firms are likely to maximize tax benefits via income
shifting. However, tax aggressiveness is costly and risky. Studies in other countries have
documented that tax aggressiveness is associated with low stock returns (Hanlon and
Slemrod, 2009) and those large BTDs due to tax aggressiveness lead to high tax
adjustments (Mills, 1998; Cho et al., 2006). If SOELG firms can reduce their tax burdens
directly by taking advantage of tax preferences, they may reduce aggressive tax planning
activities as reflected in ABTDs.17 To investigate the tax preference effect on SOELG firms’
We expect that firms that have a number of tax rates are less likely to engage in
aggressive tax reporting due to risk and cost considerations, thereby having lower ABTDs.
The variables IND and SIZE are used to control for industry and size effects on ABTDs.
We collect BTD data from companies’ English-version financial reports available at the
financial reports offered voluntarily by firms upon our request. The initial sample contains
all Chinese B-share firms18 listed on either the Shanghai or Shenzhen Stock Exchanges
from 1999 to 2004, resulting in 664 firm-year observations. We exclude firms without three-
19
accounting rules (e.g., agriculture, mining, and construction). Table 1 shows that out of the
664 firm-years that comprise the population of Chinese B-share listed firms over the period
1999-2004, our sample consists of 525 firm-year observations. The sample by year and by
Table 2 reports the summary statistics and correlation matrix for the variables used in
the BTD and ABTD models. Panel A shows that the mean (median) of BTD is -0.008 (-
0.001), suggesting that aggregate Chinese BTDs are generally negative and opposite to
aggregate BTDs in the U.S. (Plesko, 2004; Hanlon and Shevlin, 2005). The means
(medians) of ∆REV and ∆INV are 8.1% (6.7%) and 1.9% (1.1%) of total assets,
respectively. The univariate correlation analysis shows that BTD has a significantly
negative association with ∆INV and NOL and a positive relation with ∆REV and TLU,
Panel C of Table 2 presents descriptive statistics for variables in the ABTD model.
The mean (median) of ABS (ABTD) is 0.7% (0.3%) of total assets. The mean of ATR is
19.6%, which is far below the statutory rate of 33%. The mean and the maximum of
NUMBER for sample firms are 2.60 and 11, consistent with Chinese listed firms being
subject to multiple tax rates within a consolidated group due to different tax incentives
offered by the government. About 18.7% of observations have a loss and 72% of firm-year
observations make a seasoned equity offering in the following year (the number of SEOs
is quite small because, in part, the CSRC severely limits the number of SEOs in any given
20
We use the residual calculated from cross-sectional model (1) to estimate ABTDs.
Table 3 shows that the overall model is a good fit and explains most of the variation in
BTDs. The adjusted R² is 77% and the F-statistic is 352.90 (significant at the 0.01 level).20
The coefficients are all significantly different from zero. In line with expectations, the signs
on ∆INV, ∆REV, and TL are negative and the signs on TLU and TAX_DIFF are positive.
Table 4 presents the results of the regression of the estimated ABTDs on TM and EM
incentives. In the main results (column 1), all estimated coefficients are significant at the
0.05 level using a one-tailed test (two-tailed when there is no predicted sign).21Consistent
with the prediction that a higher ATR creates incentives for tax planning, resulting in larger
ABTDs, ATR is significantly positive. Holding other variables constant, the coefficient on
ATR suggests that a one-percent increase in ATR contributes a 3.8 percent increase in the
absolute value of ABTDs, which is approximately 0.82 million RMB (about 0.13 million U.S.
The coefficient on NUMBER is significantly positive, suggesting that firms that face
more tax rates engage in tax planning, presumably by shifting income from subsidiaries
with high tax rates to those with low tax rates. This finding is also consistent with the U.S.
literature suggesting that greater differences in tax rates provide more opportunities for tax
avoidance.
As expected, SEON firms and LOSS firms have larger ABTDs than their
counterparts due to their strong EM incentives. The estimated coefficients on both dummy
variables are significantly positive. This finding is consistent with the prior literature
suggesting that meeting the SEO profitability hurdle and avoiding delisting are two major
EM incentives in China. SOELG firms exhibit larger ABTDs than their counterparts as they
21
NUMBER*SOELG suggests that a SOELG that has obtained favorable tax rates is less
Overall, ABTDs are positively associated with all EM and TM incentives. Firms with
more incentives and opportunities for tax and earnings management exhibit larger ABTDs,
evidence that BTDs are a useful proxy for the combined effect of EM and TM after
To test if our tax-effect BTDs are superior to income-effect BTDs in capturing EM and
TM, we calculate income-effect BTDs following the procedures used in U.S. studies. We
dividing by the company’s applicable tax rate (See Appendix II). We then rerun the BTD
Table 4 shows that the results from this estimation are quite different. The coefficients are
insignificant except for NUMBER and SEON. ABTDs estimated from income-effect BTDs
suggesting that tax-effect BTDs are more powerful in identifying EM and TM.
Some studies suggest that in-sample evidence based on pooled cross-section time
series data is likely to lead to spurious conclusions, especially when there is parameter
instability due to unmodelled structural change across time (Poon and Granger, 2003).
Therefore, to validate our results, we use an out-of-sample test where we estimate ABTD
in a different time period to the test period. In particular, we use the parameters estimated
in year t-1 to calculate NBTDs for year t. Because the parameters are re-estimated each
22
Column 2 of Table 4 shows that the out-of-sample tests do not qualitatively change
our inferences. All independent variables are significant with their predicted signs,
suggesting that the prior regression results are robust to an alternative measurement of
ABTDs.
remove underlying determinants driven by different book and tax treatments in GAAP and
income tax laws leaving the portion driven by EM and TM in the residual. A major concern
dependent variable in Equation (2). Our results (untabulated) indicate that none of the
independent variables, except for LOSS, is significantly different from zero. This test
suggests that only opportunistic differences (ABTDs) capture EM and TM, and that our
We use the LOSS dummy variable as a proxy for one of the EM incentives, where the
dummy variable is set to one if firms have a loss in year t and zero otherwise. However,
the loss or profit level in year t might have been an outcome of EM. For example, firms
with losses in a prior year have incentives to boost earnings and so yield a profit in year t.
To address this concern, we examine firms’ incentives to avoid losses in the previous year.
We use “Loss”, “Loss1” and “Loss2” to replace LOSS. Where Loss = 1 if firms have a loss
in year t but not in year t-1, 0 otherwise; Loss1 = 1 if firms have a loss in year t-1 but not in
year t-2, 0 otherwise; Loss2 = 1, if firms have a loss in year t-1 and in year t-2, 0 otherwise.
This classification allows us to test whether ABTDs relate to loss avoidance incentives at a
more robust level. The untabulated results show that Loss and Loss2 are positively
23
previous predictions that loss firms desire to take a big bath in the first loss year to create
reserves for future use while two-consecutive-year-loss firms are keen to boost earnings in
order to avoid delisting and trading restrictions. In a relative sense, the loss avoidance
incentives for firms with a one-year loss in year t-1 (Loss1) are weaker than Loss and
Loss2 firms. This robustness check does not qualitatively change our prior conclusions.
Haw et al. (2005) find that all SEO applicants (successful and unsuccessful
focused sample, namely successful applicants. A potential concern is that successful SEO
applicants whose real economic performance has met ROE requirements may not have
EM motivations. In this case, using the SEON dummy variable as an EM incentive will
affect the power of the test. To mitigate this concern, we analyse the ROE distribution for
the SEON sample. We find that only two SEON firms exhibit high and stable ROE before
and after making SEOs. The remainder either exhibit a dramatic jump (a big drop) in ROE
before (after) making SEOs. We remove these two firms and find that the significance level
of SEON holds (p-value = 0.025), suggesting that the positive association of ABTDs with
Figures 1 and 2 show that the differences between temporary differences and
permanent differences vary across the sample period and are especially high in 2001 and
2002. To assess the impact of this on our results, we exclude Y2001 and Y2002 and rerun
Model (2). Our results are similar to those shown in Table 4, suggesting the prior results
are not driven by the variation in these two years. We also add year variables in Model 2 to
and Appendix I, and this raises a concern that the high proportion of permanent
24
excluding firms with a ratio of permanent differences higher than 60%. Our results do not
Table 5 reports the extent to which different factors affect ABTDs. The results from
the four models indicate that EM explains 7.4% of ABTDs, TM accounts for 27.8% of
ABTDs, and the combined EM and TM factor explains 3.2% of ABTDs. These findings
suggest that TM and EM are dependent and interactive, consistent with the tax and non-
tax cost literature. Evidence from BTDs on the interplay of EM and TM in China extends
Frank et al. (2009) and Wilson (2009), where they find U.S. tax sheltering firms also report
a high level of discretionary accruals, implicitly suggesting that firms manage book income
upwards on the one hand and taxable income downwards on the other hand.
6. CONCLUSIONS
drawn considerable attention from researchers, regulators, and financial information users.
important issue. Evidence from the U.S. shows that BTDs can detect EM in some settings
(Phillips et al., 2003) and BTDs are a good indicator of tax shelters (Frank et al., 2009;
Wilson, 2009). Our paper extends the current literature by demonstrating that BTDs
capture not only EM but also TM in China, a country with different regulatory and market
simultaneously investigate the association of Chinese BTDs with various EM and TM incentives.
We extend the BTD literature by using a different method to calculate BTDs (i.e., tax-effect
25
identify factors that determine BTDs and construct a regression model to explain them.
The fitted values from the regression give NBTDs and the residuals give the ABTDs. This
manipulations. Our evidence indicates that ABTDs capture reporting distortions induced by
management motivations but NBTDs do not. We also find that tax-effect BTDs are
superior to income-effect BTDs in capturing earnings and tax management at both the
conceptual and empirical levels. Evidence that ABTDs are caused by EM, TM, and their
or TM separately.
Our findings provide additional insights into international accounting and EM/TM
research. Provided that NBTDs and ABTDs are reliably measured, the estimated NBTDs
and how it changes over time. Although nonconformity is not a new issue in developed
example, prior international accounting research, such as Alford et al. (1993) and Ali and
Hwang (2000), evaluate the major differences in accounting and tax rules in a specific
TM practices affect the variation in BTDs. ABTDs provide investors with an indicator to
assess the quality of financial reports. Tax authorities and auditing firms can exploit this
indicator to perform efficient and effective audits. The findings of our paper have
implications for Chinese regulators who want to enhance the credibility of book and tax
26
EM and TM occur when management has incentives and discretion, and when
conformity can reduce opportunistic tax and book reporting is an empirical issue for future
research.
27
28
29
30
31
32
33
34
35
5,000,000
4,000,000
3,000,000
2,000,000
1,000,000
0
1999 2000 2001 2002 2003 2004
-1,000,000
-2,000,000
-3,000,000
Tax-effect Book Income Current Tax Expense Tax-effect BTDs
Figure 2
The Comparison of Permanent Differences and Temporary Differences
(With 525 Firm-year Observations, Thousand Yuan)
0
1999 2000 2001 2002 2003 2004
-500,000
-1,000,000
-1,500,000
-2,000,000
-2,500,000
Permanent Differences Temporary Differences
36
37
38
Note:
(1) The figure in parentheses is the p-value using two-tailed tests.
(2) Variable definitions:
BTD = reported book-tax differences in year t, scaled by total assets;
∆REV = changes in revenues from year t-1 to year t and proxies for economic growth,
scaled by total assets;
TL = the value of accounting loss and proxies for tax loss, scaled by total assets;
TLU = reported tax loss utilized for firm in year t, scaled by total assets;
∆INV = changes in investment in the sum of gross property, plant and equipment and
intangible assets in year t, and proxies for investment growth, scaled by total assets;
TAX_DIFF = the gap between the consolidated company’s applicable tax rate and the
average tax rate of its consolidated parties.
ABS (ABTD) = the absolute value of ABTD, calculated as the residual from the cross-
sectional regression (Equation 1);
ATR = applicable tax rate of the sample company;
39
40
Sample 525
Notes:
(1) Asterisks *, **, and *** denote two-tailed (one-tailed when there is a predicted sign) statistical
significance at 10%, 5%, and 1% respectively.
(2) Variable definitions:
BTD = reported book-tax differences in year t;
∆INV = changes in investment in the sum of gross property, plant and equipment and intangible
assets in year t;
∆REV = changes in revenues from year t-1 to year t, and proxies for economic growth;
NOL = the value of accounting losses;
TLU = reported tax loss utilized for firm in year t;
TAX_DIFF = the gap between the consolidated company’s applicable tax rate and the average tax
rate of its consolidated parties.
All variables are scaled by total assets except for TAX_DIFF.
41
Notes:
(1) Asterisks *, **, and *** denote two-tailed (one-tailed when there is a predicted sign) statistical
significance at 10%, 5%, and 1% respectively. P-values are provided in parentheses and are based on robust
panel corrected standard errors (Beck and Katz 1995)
(2) Variable definitions:
ABS (ABTD_full) = the absolute value of ABTD, calculated as the residual from Equation 1;
ABS (ABTD_out) = the absolute value of ABTD, calculated as the residual from an annual cross-sectional
regression in period t-1;
ATR = applicable tax rate;
NUMBER = the number of different applicable tax rates in a consolidated entity;
SEON = dummy variable that equals 1 when a consolidated entity has a rights issuing or public offering in the
next year, and 0 otherwise;
LOSS = dummy variable that equals 1 when a consolidated entity has a loss in the current year t, and 0
otherwise;
SOELG = dummy variable that equals 1 when a consolidated entity is a state-owned enterprise controlled by
local governments, and 0 otherwise;
IND=dummy variable that equals 1 for manufacturing industry and 0 otherwise;
SIZE= the natural logarithm of total assets at the fiscal year-end.
†
Three observations with zero ATR are dropped, leading to 522 firm-years.
42
43
44
45
46
N
Estimated taxable income = (TI i * ti ) / t0 (2)
i 0
Assume that a consolidated entity A has $600 taxable income for itself in year t, and it
has 3 subsidiaries called B, C and D. Their taxable income and tax rates are $400, $500 and
$500, 24%, 33% and 10%, respectively. Assume the statutory tax rate for the consolidated
entity is 15%. Assume the consolidated group’s total book income is $1000. Thus,
Total taxable income = (600+400+500+500) = $2000
Income-effect BTDs = $1000- $2000 = -1000
Tax-effect BTDs = 1000*15% - (600*15% +400*24% +500*33% +500*10%) = -$251
However, if we use equation (2) to estimate taxable income,
Current tax expense = 600*15% + 400*24% + 500*33% + 500*10% =402
Estimated taxable income = 402/15% = $2673
Income-effect BTDs = $1000 – $2673 = $-1673
47
48
49