Rev Account Stud (2012) 17:649–687

DOI 10.1007/s11142-012-9194-7

Where do firms manage earnings?

Scott D. Dyreng • Michelle Hanlon •

Edward L. Maydew

Published online: 24 June 2012

Ó Springer Science+Business Media, LLC 2012

Abstract Despite decades of research on how, why, and when companies manage
earnings, there is a paucity of evidence about the geographic location of earnings
management within multinational firms. In this study, we examine where companies
manage earnings using a sample of 2,067 U.S. multinational firms from 1994 to
2009. We predict and find that firms with extensive foreign operations in weak rule
of law countries have more foreign earnings management than companies with
subsidiaries in locations where the rule of law is strong. We also find some evidence
that profitable firms with extensive tax haven subsidiaries manage earnings more
than other firms and that the earnings management is concentrated in foreign
income. Apart from these results, we find that most earnings management takes
place in domestic income, not foreign income.

Keywords Earnings management  Rule of law  Tax havens  Multinational firms

JEL Classification D22  M40  M41  M42

S. D. Dyreng
Duke University, Durham, NC 27708, USA
e-mail: [Link]@[Link]

M. Hanlon
Massachusetts Institute Technology, Cambridge, MA 02142, USA
e-mail: mhanlon@[Link]

E. L. Maydew (&)
University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA
e-mail: edward_maydew@[Link]

123
650 S. D. Dyreng et al.

1 Introduction

The study of earnings management dates back to at least Healy (1985).1 In

subsequent decades, researchers have conducted hundreds of studies of earnings
management. Among other things, these studies have provided insights into when
firms manage earnings, what types of accounts they manage, why they manage
earnings, and how they manage earnings.2 Surprisingly, however, there is a paucity
of evidence about where firms manage earnings. Does earnings management
generally take place in a firm’s foreign operations, far from headquarters and
perhaps the scrutiny of auditors, or does it happen closer to home? This study
provides initial evidence on the location of earnings management.
Prior research has examined whether the location of a firm’s headquarters or
location of incorporation is associated with measures of earnings management (for
example, Leuz et al. 2003). In these studies, the broad research question is whether
the institutions, customs, laws, and accounting standards in the firm’s parent
location affect financial reporting. In our study, we take the next step to investigate
this broad question, but we go inside the firm to examine whether the institutions
and laws, including tax laws, in the locations of the firm’s subsidiaries (proxied by
the World Bank ‘rule of law’ and tax haven status), are associated with differences
in earnings management in reported financial statements of the entire consolidated
company. Thus, the question is whether the institutional features of the subsidiaries
affect financial reporting, despite the firm as a whole being subject to U.S. Generally
Accepted Accounting Principles (GAAP) and Securities and Exchange Commission
(SEC) regulations.
We examine a sample of U.S.-based multinational firms from the years 1994 to
2009, constituting 11,077 firm-years. Our study focuses on multinational firms; we
exclude purely domestic firms to avoid the trivial result that domestic firms by
definition manage only domestic earnings. An important aspect of our sample is that
GAAP and the regulatory environment is presumptively held constant at the firm
level, in the sense that a U.S. listed firm is subject to U.S. GAAP and U.S. securities
laws across its entire operations, whether domestic or foreign. Our first set of tests
examines whether companies with a higher concentration of material subsidiaries
(disclosed in Exhibit 21 of the 10-K) in low rule of law locations or tax havens have
more discretionary accruals. Our second set of tests investigates whether earnings
management occurs in foreign earnings or in domestic earnings by examining the
mapping of foreign and domestic earnings into discretionary accruals. We also test
the frequency of earnings restatements across the partitions of firms as an alternative
measure of earnings management.
We report three main findings. First, the data show that overall earnings are
managed less when the firm has a high concentration of subsidiaries in foreign
countries with a strong rule of law. This differential in earnings management for
1
Indeed, research on the related phenomenon of accounting method choice dates back to studies such as
Watts and Zimmerman (1978), Hagerman and Zmijewski (1979), Bowen et al. (1981), and Holthausen
(1981).
2
For reviews see Healy and Wahlen (1999), Schipper (1989), Dechow and Skinner (2000), Fields et al.
(2001), and Dechow et al. (2010).

123
Where do firms manage earnings? 651

firms with subsidiaries in high rule of law countries is more pronounced in foreign
income, consistent with the earnings management occurring in the foreign
jurisdiction. Again, we are not comparing earnings management of foreign firms
to U.S. firms—all the firms in the sample are U.S. domiciled and subject to U.S.
laws and accounting standards. Thus, despite the firm being subject to U.S. GAAP
and SEC rules, the data are consistent with the geographic location of a U.S. firm’s
foreign subsidiaries being associated with significant differences in earnings
management.
Second, we find that profitable firms with extensive tax haven subsidiaries engage
in more earnings management, which is also concentrated in foreign earnings. This
is not simply a manifestation of tax havens reducing the tax expense; we measure
earnings management before taxes. Rather, we posit that tax havens are more
desirable locations for earnings management because there is little or no local tax
cost as a result of managing pre-tax income (and likely no U.S. tax cost because
cash repatriations are not affected). For example, Erickson et al. (2004) find that
firms engaging in fraudulent accounting incur actual tax costs when inflating their
accounting earnings. They conjecture that firms can minimize the tax cost of such
activities by locating them in a tax haven. In addition, Desai and Dharmapala (2006)
observe that the same types of activities and structures that facilitate tax avoidance
via tax havens can be used to facilitate managerial misappropriation. Our
predictions are based on implications from these prior papers—earnings manage-
ment can generate tax costs that can be avoided in tax havens, and the structures
employed in tax haven subsidiaries provide obfuscation that aids earnings
management activities.
Finally, apart from the above results, earnings management appears to be more
prevalent in domestic income than in foreign income. This is not a simple result of
domestic operations being larger in scale than foreign operations; on average,
foreign and domestic operations are of the same order of magnitude for the firms in
our sample.
We conduct a number of additional tests to examine the robustness of the results,
including examining the absolute value of discretionary accruals, controlling for
performance differences across firms, using different measures of rule of law and
tax haven status, and using restatements as the indication of earnings management
instead of discretionary accruals. Across these tests, the results are generally
consistent and in line with our main findings. One exception is that when examining
restatements we do not find evidence that firms operating in tax havens are more
likely to restate their financial statements. The evidence on rule of law continues to
hold in the restatement sample as it did in the discretionary accruals tests, that is,
firms with proportionately more subsidiaries in low rule of law countries restate
earnings more often.
We contribute to the literature by examining a new aspect of earnings
management—the role of the geographic footprint of a U.S. multinational’s
subsidiaries. Specifically, we show that the previously documented effect that rule
of law has on earnings management is actually much more pervasive. Not only does
the rule of law in the parent company’s jurisdiction matter, but so does the rule of
law in the jurisdictions of the firm’s subsidiaries. In addition, we show that

123
652 S. D. Dyreng et al.

subsidiary structures involving tax havens are associated with higher discretionary
accruals for profitable firms. Finally, policymakers will find the results of this study
interesting, particularly to the extent they have expressed concern about auditing
and accounting in the foreign operations of U.S. multinationals and as they consider
U.S. policies toward tax haven nations.
The paper proceeds as follows. In the next section, we review the prior literature
with particular attention to research that provides a basis for our predictions about
the geographical incidence of earnings management. In Sect. 3, we develop our
hypotheses. In Sect. 4, we discuss our sample, variable measurement, and empirical
tests. In Sect. 5, we present our results, and in Sect. 6, we conclude.

2 Background and prior literature

2.1 The regulation of multinational companies

Our inquiry focuses on U.S. multinationals and the management of earnings within
the firm, that is, where do multinationals manage earnings. Because our sample
consists entirely of U.S. SEC registrants, all the companies are required to apply
U.S. GAAP to their financial statements regardless of where the underlying
operations are located.3 In our sample, the top-level regulatory environment is held
constant across our firms. What varies, however, are the local legal and regulatory
environments of the countries in which foreign subsidiaries are located.
The ability of the U.S. to effectively exert regulatory control over the accounting
in the foreign operations of U.S. multinationals has been a longstanding issue. For
example, the Foreign Corrupt Practices Act (FCPA) of 1977 aims to prevent U.S.
companies and their agents from bribing foreign officials. The FCPA also contains
important accounting provisions that require U.S. registrants to maintain a system of
internal controls that provide reasonable assurance that, among other things, the
transactions are properly recorded to permit preparation of financial statements in
conformity with GAAP (Golden et al. 2006). In addition, the SEC enforcement
manual states that it may be more difficult to obtain evidence to bring a case forward
with respect to activities in certain foreign jurisdictions (SEC 2011).
Two examples in which the SEC did become aware of and bring action against
firms for earnings management in their foreign subsidiaries are as follows. The SEC,
in Accounting and Auditing Enforcement Release (AAER) No. 2727, accuses
Bristow Group Inc., a Delaware corporation with headquarters in Houston, Texas, of
improprieties at a foreign subsidiary. Specifically, Bristow’s wholly owned U.S.
subsidiary, AirLog International, Ltd., through its Nigerian affiliate, Pan African
Airlines Nigeria Ltd. (PAAN) made improper payments totaling approximately
$423,000 to employees of the governments of two Nigerian states. These improper
payments, however, were not properly recorded in AirLog’s books and were never
3
The subsidiary may also be required to prepare financial statements in accordance with local GAAP for
a variety of reasons. Our point here is that when the earnings are consolidated and reported for the entire
company, all the earnings are reported using U.S. GAAP and are subject to U.S. securities laws.

123
Where do firms manage earnings? 653

included in Bristow’s consolidated statements. Bristow’s internal controls failed to

detect and prevent the improper payments. In another example, the SEC concluded
that a Japanese subsidiary of Boston Scientific recorded false sales and materially
overstated its earnings for several years in the 1990s, which then caused material
misstatements in Boston Scientifics’ consolidated financial statements (AAER
1295).
The Public Company Accounting Oversight Board (PCAOB) has also expressed
concern with quality control in audits across borders. The then-acting chairman,
Daniel L. Goelzer, in a recent speech stated the following concerns: (1) U.S.
engagement partners sometimes do not have a sufficient basis on which to assess
whether the non-U.S. audit personnel are qualified and familiar with U.S. GAAP,
PCAOB standards, and SEC requirements; (2) internal inspection information about
foreign-affiliate firms and personnel are sometimes not made available to the audit
engagement partners and, at times, engagement partners fail to even ask for it; (3)
audit firms may not have minimum levels of training for U.S. GAAP, PCAOB
standards, or SEC requirements for foreign-affiliate personnel; and (4) U.S.
engagement teams sometimes fail to appropriately evaluate the results of a foreign
affiliate’s work or fail to adequately supervise and control the affiliate’s work.4
Thus, while subsidiaries need to report their earnings following U.S. GAAP, the
local institutions will vary across subsidiaries and the knowledge of U.S. GAAP and
securities laws may vary across personnel at the companies and at the audit firms.
In sum, there is concern by accounting regulators about foreign operations of
U.S. multinationals and there exist examples of earnings management in both
domestic and foreign operations. What is lacking is large sample evidence about the
geographic location of earnings management within firms and the factors that drive
the location. There is, however, prior research that informs our thinking of how the
geographic incidence of earnings management within the firm might play out, and
we turn to this research in the next section.

2.2 Prior literature

There is little prior research on where, within a multinational company, earnings are
managed. Leuz et al. (2003), in an influential paper, examine earnings management
around the world in a sample of over 8,000 firms from 31 countries and provide
evidence of the importance of institutional features on accounting. For example,
they provide evidence that earnings management is more likely to occur at
companies that are located in countries where investor protection is weaker,
reasoning that such protection prevents the extraction of private benefits by insiders
and thus reduces the incentives to obfuscate information. However, they do not
examine earnings management within the firm. In other words, they compare a U.S.
domiciled company with a firm domiciled in Singapore, for example, but they do
not compare the domestic earnings of a U.S. multinational with the foreign earnings
of that same multinational company. Like Leuz et al. (2003), we are interested in the
4
December 7, 2009, speech before the American Institute of Certified Public Accountants (AICPA)
National Conference on Current SEC and PCAOB Developments, Washington, DC (Goelzer 2009).

123
654 S. D. Dyreng et al.

effect of rule of law on earnings management, but our focus is on whether the rule of
law at the subsidiary level matters, holding constant the parent rule of law. Thus, our
entire sample is U.S.-based multinationals, and all of the earnings are subject to U.S.
securities laws and regulations and are reported under U.S. GAAP. We examine
whether variation in the local rule of law applying to the foreign subsidiaries affects
earnings management.
Kedia and Rajgopal (2011) examine the geography of financial misreporting, but
their study is focused on the location of corporate headquarters relative to an SEC
office. If corporate headquarters is in a county far from an SEC office and the local
auditor office in the county is lax, companies in that county report a higher
incidence of earnings decreasing restatements. Dyreng et al. (2010), Grullon et al.
(2010), and McGuire et al. (2012) examine the effect of social norms on earnings
management. These studies generally find that firms located in counties with high
levels of religious adherence are less likely to engage in earnings management.
While related and interesting, these papers examine differences in earnings
management across firms and do not look at the location of the earnings
management within the firm.
Other papers do examine foreign earnings relative to U.S. earnings of U.S.
multinational corporations, but these papers do not examine the extent of
manipulation of those earnings. More importantly, the studies do not investigate
the effect of institutional features in the location of the earnings on earnings
properties. These papers are primarily concerned with the relative market valuation
of the earnings. For example, Bodnar and Weintrop (1997) examine the relation
between returns and domestic and foreign incomes and report that foreign earnings
have a greater association with returns, consistent with the foreign earnings being
related to growth opportunities reflected in returns. Another example is Thomas
(1999), who, in a Mishkin (1983) framework, examines whether foreign earnings
have different persistence than domestic earnings and whether trading profits can be
earned via trading on the knowledge of such a difference. Thomas reports that
foreign earnings have greater persistence than domestic earnings but that the market
underestimates the persistence of foreign earnings, leading to predictable future
returns. Hope et al. (2008) extend Thomas (1999) and report evidence consistent
with investor mispricing of foreign earnings decreasing after the adoption of SFAS
131 Disclosures about Segments on Enterprise and Related Information. While
these papers consider foreign versus domestic earnings, their focus is on mispricing,
not where multinationals undertake earnings management.
Perhaps most closely related to our paper are three working papers: Durnev, et al.
(2011), Beuselinck et al. (2010), and Fan (2008). Durnev et al. (2011) examine
governance at offshore financial centers and the effects on earnings management.
The authors classify companies incorporated in an offshore financial center and U.S.
domiciled firms that have affiliates in offshore financial centers both as ‘‘offshore
firms,’’ to which they compare other U.S. firms without offshore affiliates. The
authors report evidence consistent with offshore firms engaging in more earnings
management than non-offshore firms. They further document that companies with a
higher Offshore Attitude Index, intended to measure the firm’s institutional and
legal environment, manage earnings more with accruals than with real earnings

123
Where do firms manage earnings? 655

management. Beuselinck et al. (2010) examine earnings management in European

subsidiaries of EU-based multinational corporations. The authors find that earnings
management is higher at the subsidiary level when the subsidiary-country
institutional quality is weak. They also document evidence consistent with the
governance characteristics of the multinational company parent affecting the
magnitude of subsidiary earnings management over and above subsidiary-specific
characteristics. The authors conclude that the parent-level characteristics have
contagion effects for their subsidiary-level financial reporting quality.
Finally, Fan (2008) examines whether earnings management is evident in foreign
earnings of U.S. multinationals and whether the earnings management pattern
changed after the adoption of SFAS 131. Fan uses the Burgstahler and Dichev
(1997) discontinuity in the distribution of earnings levels and earnings changes as
the measure of earnings management. Fan reports evidence consistent with foreign
earnings being managed to avoid a loss but not to avoid an earnings decrease. She
also reports that domestic earnings are not managed to avoid a loss. She does not
find a significant difference, on average, between the pre-and post-SFAS 131
periods in loss avoidance behavior.
In sum, a few studies have addressed aspects of our research question. However,
the evidence, even when pieced together, is inconclusive. For example, Thomas
(1999) documents that foreign earnings are more persistent than domestic earnings,
while Fan (2008) documents that foreign earnings are used more often than
domestic earnings to avoid a loss for U.S. multinationals. Neither of these studies
examines the institutional forces at work in the various locations (for example, in a
tax haven or not, or in a location with a strong rule of law or not). In contrast, Leuz
et al. (2003) focus on the institutional forces but do not study earnings management
within a multinational firm. Rather, they study the earnings quality of firms located
in different countries operating under different standards, legal environments, and
governance/shareholder protection regimes. Durnev et al. (2011) focus exclusively
on offshore financial centers, combining multinational firms with affiliates in these
centers with firms domiciled in these centers. Despite the advances made in
understanding earnings management, it is not known whether U.S. multinational
firms have more earnings management in domestic earnings, foreign earnings, or
both, and whether such earnings management is affected by the local institutional
regimes in which its foreign subsidiaries operate.5

5
There are other important papers that are somewhat related to our research but less directly. For
example, Duru and Reeb (2002) report evidence that analysts have lower forecast accuracy when firms
have greater international diversification. Looking across firms—thus, more similar in spirit to Leuz et al.
(2003), Pincus et al. (2007) document the occurrence of the accrual anomaly in foreign countries, and
DeFond et al. (2007) provide evidence that earnings announcements are more informative in countries
with strong investor protections. In addition, there are other papers beyond Erickson et al. (2004) that
study the relation between taxes and earnings management. For example, Badertscher et al. (2009)
(discussed below), Frank et al. (2009), and Rego (2003). Frank et al. (2009) find a positive relation
between aggressive tax reporting and aggressive financial reporting, and Rego (2003) finds economies of
scale in tax planning, such that multinational firms are better able to avoid tax. We do not study tax
planning in our paper.

123
656 S. D. Dyreng et al.

3 Hypothesis development

Earnings management can occur at the direction of central management or via the
conduct of a division or subsidiary manager.6 In our setting, the predictions are the
same in either case. For example, if central managers decide to engage in earnings
management, they must decide whether to manage domestic income, foreign
income, or both. Managers are expected to consider the expected costs of earnings
management, including the likelihood of being challenged or discovered. In
addition, managers will consider the potential benefits of earnings management such
as higher compensation (Healy 1985) or avoiding the violation of a debt covenant
(DeFond and Jiambalvo 1994; Sweeney 1994; Dichev and Skinner 2002), for
example. The expected likelihood of being caught may be lower for domestic
earnings management, if directing earnings management in the foreign jurisdiction
would involve more people or would involve people who do not work closely with
top management. In other words, to the extent that top management wants to limit
the number of people involved and keep the misreporting close to the parent
company, earnings management will occur in domestic income. However, it is
possible that foreign income will be managed more in order to avoid getting caught.
For example, if the Kedia and Rajgopal (2011) result—that firms farther from SEC
offices manage earnings more than firms close to SEC offices—can be generalized
to U.S. subsidiaries relative to foreign subsidiaries, then more earnings management
will take place in foreign jurisdictions. Because there are reasons for earnings
management to be located either more or less in foreign earnings than domestic
earnings, we make no directional prediction with regard to the overall geographic
location of earnings management.
Our first directional prediction (Hypothesis 1) focuses on the rule of law in the
foreign countries where the firm’s subsidiaries are located. We predict that the local
legal systems of the firm’s subsidiaries affect the firm’s propensity to engage in
earnings management, such that having subsidiaries in weak rule of law countries
will be associated with more earnings management. The effect of local institutions
on accounting quality has been shown in prior literature, as discussed above (for
example, Leuz et al. 2003). Earnings management in foreign operations may be less
likely to be challenged, and, if so, the consequences are likely to be less stringent
when there is a weak rule of law. Audits of foreign operations are typically
performed by foreign affiliates of the auditor or, in some cases, by an unrelated
foreign audit firm. If those audits are lacking the rigor of the audit of the U.S.
operations, as suggested is often the case by the PCAOB chair as discussed above,
then the managers might face less chance of being challenged if they manage

6
In an analysis of fraudulent financial reporting over the period 1998–2007, the Committee of
Sponsoring Organizations of the Treadway Commission (COSO) states: ‘‘The SEC named the CEO and/
or CFO for some level of involvement in 89 % of the fraud cases, up from 83 % of cases in 1987–1997.
Within 2 years of the completion of the SEC’s investigation, about 20 % of CEOs/CFOs had been
indicted and over 60 % of those indicted were convicted’’ (Beasley et al. 2010). Thus, many cases of
financial manipulation name central managers as the guilty party. However, some cases are done by
‘‘rogue’’ managers, as the two SEC AAER cases in Sect. 2 indicate. Top management at Boston Scientific
and Bristow did not appear to be, and were not accused of being, involved in the fraud.

123
Where do firms manage earnings? 657

foreign income, particularly if the rigor of the local audit is affected by the local rule
of law. The effect of local rule of law on earnings management, if any, is interesting
when examining the earnings of subsidiaries of U.S. multinationals because the
earnings of subsidiaries are required to be reported under U.S. GAAP and U.S.
securities laws. The U.S. is considered to have a strong rule of law, which ideally
should act as a constraint on earnings management in both domestic and foreign
operations, in which case weaker constraints based on local law would not matter.
Our second directional prediction (Hypothesis 2) focuses on the role of tax
havens in earnings management. We predict that having subsidiaries in tax havens is
associated with more earnings management. This prediction comes in part from
Erickson et al. (2004), which examines whether firms engaging in accounting fraud
pay taxes on the fraudulent earnings. Erickson et al. (2004: p. 391) do not study
where firms manage the earnings, or whether the taxes are foreign or domestic, but
the authors conjecture that ‘‘the firm might overstate the income of a foreign
subsidiary located in a low-tax country.’’ Managing earnings in a tax haven avoids
cash tax payments as well as the accounting income tax expense in most cases.7
In a follow-up study to Erickson et al. (2004), Badertscher et al. (2009) examine
restatement firms and characteristics of restatement firms that manage earnings
using methods where book income conforms to taxable income or using methods
that are unconfirmed. One of the premises in that study is that nonconforming
earnings management is less costly than conforming earnings management because
there is no tax cost if the item managed is not conformed between book and taxable
incomes. In our study, we look within a firm and hypothesize another way managers
can avoid paying tax on the income managed—by managing the earnings in a low-
tax jurisdiction (that is, a tax haven).8 Thus, our study complements Badertscher
et al. (2009) because we both hypothesize that earnings management is more likely
when tax costs are lowest.
Another reason we expect to see more earnings management for firms with tax
haven subsidiaries comes from Desai et al. (2007) and Desai and Dharmapala
(2006). Those studies posit that the same structures and activities conducive to tax
avoidance (that is, complex structures and secrecy) are also conducive to private
diversion of the firm’s resources by managers. Indeed, the authors argue that the
primary intent of many tax shelters is to increase accounting earnings. Furthermore,
in a report by the Joint Committee of Taxation (JCT) prepared for the Senate
Finance Committee, the JCT noted that Enron created many entities in jurisdictions
that did not impose taxes on such entities (for example, tax havens). In particular, as
of December 31, 2001, the Enron ownership structure included 441 entities formed
in the Cayman Islands, a country that has never imposed a corporate income tax.
These entities were used for transactions that increased reported earnings (for
example, LJM Cayman) but did not generate additional taxes.

7
Many tax havens have a corporate tax rate of zero, but conceptually for this paper we only need havens
to impose lower taxes on income than does the U.S.
8
Our premise assumes that the earnings from the subsidiary are not repatriated and the U.S. tax is not
incurred. The Badertscher et al. (2009) study assumes all the income is taxable in the U.S. and that
managers lessen the tax via nonconforming earnings management.

123
658 S. D. Dyreng et al.

Of note is that our hypothesized prediction for tax havens is most applicable to
profitable, tax-paying firms. Our prediction is that tax havens allow earnings
management in a tax free manner, which implies that taxes are costly for the firm.
As a result, in the tests that follow, we estimate our regressions over the full sample
and separately over a subsample of profitable firms (that is, positive pre-tax
domestic and foreign earnings).
In addition, we recognize that haven locations could be used by the corporations
in our study for tax planning purposes. The income and accruals measures that we
examine in our tests are pre-tax and thus are not affected by reductions in tax
expense as a result of any tax planning via tax havens (or otherwise). For example,
the earnings management that we observe in our measures is not simply a
manifestation of transfer pricing for tax purposes. Such tax-motivated income
management involving tax havens typically involves shifting income across
jurisdictions (for example, U.S. to foreign) within the current year. While this
can increase pre-tax foreign income and decrease pre-tax domestic income, it will
not affect total consolidated pre-tax earnings. In contrast, accounting-motivated
earnings management involves shifting income over time (for example, accruals
that increase current year income and decrease next year’s income or real
transactions accelerated into the current period).9

4 Variables, sample, and descriptive statistics

4.1 Variables

We examine two measures of discretionary accruals in our analyses, and later in the
paper we examine restatements as well. The first is signed pre-tax discretionary
accruals, PTDACC. The second is the absolute value of pre-tax discretionary
accruals |PTDACC|. The absolute value of discretionary accruals is useful as a
dependent variable to capture both income-increasing and income-decreasing
earnings management when there is reason to expect both types of earnings
management in a sample.10
Note that data on foreign operations is extremely limited from public sources,
especially following the implementation of SFAS 131, and thus direct tests of
earnings management in specific foreign locations are difficult (Berger and Hann
2003). We employ company-wide measures of discretionary accruals and conduct
an array of cross-sectional tests using the rule of law and tax haven location to
investigate our research question.
9
Similarly, companies can tax plan for a variety of other reasons, for example, to maximize foreign tax
credits. While firms in our sample may be doing such tax planning, to our knowledge such tax planning
will not affect our tests. We have constructed the tests using pre-tax measures to avoid incorrect
inferences due to tax planning.
10
Studies examining the absolute value of discretionary accruals include Dechow and Dichev (2002),
Frankel et al. (2002), Klein (2002), Chung and Kallapur (2003), Myers et al. (2003), Leuz et al. (2003),
and Bergstresser and Philippon (2006). As discussed later, we control for operating volatility in our
analyses, following the recommendations of Hribar and Nichols (2007).

123
Where do firms manage earnings? 659

We measure discretionary accruals according to the modified Jones model,

before taxes (Jones 1991; Dechow et al. 1995). We expect tax haven usage to
facilitate the management of pre-tax accounting income, because the earnings
management will be tax free. Thus, we exclude discretionary accruals that may arise
from managing tax expense. Specifically, we start with pre-tax accruals, PTACC,
defined as pre-tax income (PI), less pre-tax cash flow, which is cash flow from
operations (CFO), plus cash tax paid, less cash from extraordinary items and
discontinued operations:
PTACCit ¼ PIit  ðCFOit þ TXPDit  XIDOCit Þ: ð1Þ
Then, PTDACC (pre-tax discretionary accruals) is the estimated residual (^it ) from
the following model (estimated by 2-digit SIC and year):
PTACCit 1 ðDSALESit  DARit Þ ðPPEit Þ
¼ a0 þ a1 þ a2 þ a3 þ it ð2Þ
Ait1 Ait1 Ait1 Ait1
where DSALES is the change in sales from the prior year to the current year; DAR is
the change in accounts receivable from the prior year to the current year; and PPE is
gross property, plant, and equipment at the end of the year.
The explanatory variables of most interest are RULE OF LAW and HAVEN
INTENSITY. RULE OF LAW is measured as the average rule of law of countries in
which the firm discloses subsidiaries in Exhibit 21 of Form 10-K. The rule of law
scores at the country-year level are from the World Bank Governance Indicators
dataset. Thus, in our data RULE OF LAW is a firm-year measure based on the
combination of subsidiaries the firm has and the local rule of law applying to those
subsidiaries. RULE OF LAW will be high when the firm’s material subsidiaries are
predominantly located in countries with a strong rule of law. RULE OF LAW
captures perceptions of the extent to which agents have confidence in and abide by
the rules of society and, in particular, the quality of contract enforcement, property
rights, the police, and the courts, as well as the likelihood of crime and violence.11
We predict that, despite our sample firms being subject to U.S. GAAP and U.S.
securities laws on their worldwide earnings, the local legal systems of their
subsidiaries will affect their propensity to manage earnings, such that more earnings
management happens when the subsidiaries are located in countries with weak rule
of law.
HAVEN INTENSITY is measured as the number of subsidiaries located in tax
haven countries, divided by the total number of subsidiaries. Data on subsidiaries
are from Exhibit 21 of the Form 10-K, where firms are required to list their material
subsidiaries. Tax havens are identified as countries that are on at least two of the
four commonly used tax haven lists.12 While there is not an official definition of a
tax haven, the Organization for Economic Cooperation and Development (OECD)
lists criteria to being labeled a tax haven, including (1) imposing no or only nominal
taxes, (2) a lack of transparency, (3) laws or administrative practices that prevent the

11
See [Link] for a detailed discussion and listing of
factors.
12
We use the lists in Miedema (2008) as per Dyreng and Lindsey (2009).

123
660 S. D. Dyreng et al.

effective exchange of information for tax purposes with other governments on

taxpayers benefiting from the no or nominal taxation, and (4) an absence of a
requirement that the business activity be substantial.13 We expect a positive relation
between HAVEN INTENSITY and the magnitude of discretionary accruals.
We include variables to control for the general complexity of the firm, the scope
of its foreign operations, and the macroeconomic conditions of the countries in
which it has subsidiaries. We measure the extent to which the firm has special
purpose entities, SPE INTENSITY, as the number of limited liability companies,
limited partnerships, and trusts divided by the total number of subsidiaries on
Exhibit 21. Prior research has used this approach to identify special purpose entities
and shown that their use is associated with an increased propensity to disclose
material weaknesses (Doyle et al. 2007).14 We include two measures of the scope of
foreign operations. The first, FOREIGNNESS, is foreign sales divided by total sales
of the entire company. The second is the natural log of the number of countries in
which the company has material subsidiaries, NCOUNTRIES. We include a measure
of firm size, defined as the natural log of the firm’s assets, SIZE. We also include
variables that reflect time-varying macroeconomic characteristics of the countries in
which the firm operates. The first, CPI CHANGE SUB COUNTRIES, reflects the
average of the change in the consumer price index of the countries in which the firm
operates. The second, GDP CHANGE SUB COUNTRIES, reflects the average of the
change in gross domestic product of the countries in which the firm operates. These
two variables are gathered from the World Economic Outlook Database, April 2010
edition, available from the International Monetary Fund.
We also include two variables to reflect the operating volatility of the firm.
Hribar and Nichols (2007) suggest including these variables when using the absolute
value of discretionary accruals as a dependent variable. We include the variables in
our signed accruals tests as well for ease of comparability. SALES VOLATILITY is
the rolling 5-year standard deviation of sales, from year t to t - 4. CASH FLOW
VOLATILITY is the rolling 5-year standard deviation of cash flow from operations
(from the statement of cash flows), going from t to t - 4.

4.2 Sample and descriptive statistics

Our sample selection criteria are described in Table 1. We begin with all U.S.
incorporated firm-years listed on Compustat during the period 1994–2009,
excluding only the smallest of firms (assets less than $1 million or lagged assets
less than $1 million), firms in regulated industries (SIC codes 4900–4999 and
6000–6999), and those with missing values for pre-tax income, operating cash flow,
cash taxes, change in sales, change in receivables, and gross property, plant, and
equipment. This gives us an initial sample of 78,448 firm-years. From these
observations, we exclude firm-years that belong to industry-years with less than 10
13
See [Link] for fur-
ther details.
14
This definition captures noncorporate entities, which may not all meet the technical definition of an
SPE. We follow prior convention for labeling purposes but recognize that there is measurement error in
this variable.

123
Where do firms manage earnings? 661

Table 1 Sample selection

Criteria Firms Firm-
years

U.S. Corporations covered by Compustat between 1994 and 2009 with total assets 10,710 78,448
greater than $1 million and nonmissing values of the following variables: OANCF,
TXPD, SALE, RECCH, PPEGT, and PI
With data necessary to compute Jones Model Accruals (including lagged assets for the 10,412 69,819
scalar, and 10 industry-year observations to estimate the equation)
With nonmissing values of PIDOM and PIFO 6,832 32,734
With nonmissing values of RULE OF LAW, HAVEN INTENSITY, SPE INTENSITY, 2,067 11,077
CPI CHANGE SUB COUNTRIES, GDP CHANGE SUB COUNTRIES,
FOREIGNNESS, CASH FLOW VOLATILITY, and SALES VOLATILITY

This table explains the sample selection criteria used in the study. Variables referred to above are defined
in Table 3 and in Sect. 4

observations. We calculate our modified measure of Jones model discretionary

accruals using this slightly reduced sample of 69,819 firm-years. To investigate the
incidence of earnings management in domestic and foreign income, we need
companies to have foreign income. When we require firm-years to have nonmissing
values for pre-tax domestic and foreign income, we obtain 32,734 firm-years.
Finally, we require nonmissing data to compute the independent variables in the
study. Many of the independent variables, described below, are based on data from
Exhibit 21 of the 10-K, where firms are required to list their material subsidiaries.
As stated above, we use these data to compute measures of tax haven intensity, rule
of law, and special purpose entity intensity. After applying these screens, the main
sample used in our tests has 11,077 firm-years from 2,067 firms.
Table 2 presents the breakdown of the sample across the Fama–French 30
industries.15 The sample is well spread over industries, with no industry accounting
for more than 22 % of the firms. Most industries account for 1–5 % of the firms.
The industry representation in our sample is similar to the composition of
multinational firms listed in Compustat (where we define multinational firms to be
those with nonzero pre-tax foreign income or nonzero foreign tax expense).
Descriptive statistics for the main variables (income based item are scaled by
assets) in the study are presented in Table 3. Pre-tax discretionary accruals have a
mean (median) of 0.003 (0.011) in the sample.16 The absolute value of pre-tax
discretionary accruals has a mean (median) of 0.064 (0.041). This is smaller than the
0.101 mean absolute value of discretionary accruals in Hribar and Nichols (2007),
but this is to be expected since our sample includes only multinational firms, and
Hribar and Nichols (2007) show that larger firms have smaller absolute value of
discretionary accruals. As the sample consists of multinationals, foreign operations
are a substantial part of the firms in the sample. The mean pre-tax domestic income
15
The financials and utilities industries are dropped because we eliminate regulated industries from the
sample. Of the 14 firms in Compustat in the tobacco industry, none fulfill all of our sample criteria. Thus,
we include a breakdown of 27 (not 30) industries in Table 2.
16
The mean is not zero because the model is estimated over the larger 69,819 firm-year sample, which is
prior to imposing other data requirements to arrive at the final sample.

123
662 S. D. Dyreng et al.

Table 2 Sample frequency by industry

Industry Sample Sample firms Compustat MNE Compustat MNE
firms (%) firms firms (%)

Automobiles and trucks 58 2.81 99 2.04

Beer and liquor 5 0.24 14 0.29
Printing and publishing 30 1.45 61 1.26
Business equipment 438 21.19 915 18.88
Aircraft, ships, and railroad 17 0.82 38 0.78
equipment
Chemicals 80 3.87 139 2.87
Apparel 41 1.98 83 1.71
Construction 56 2.71 134 2.76
Coal 1 0.05 7 0.14
Electrical equipment 55 2.66 101 2.08
Fabricated products 160 7.74 258 5.32
Food products 37 1.79 92 1.90
Recreation 42 2.03 154 3.18
Health care 160 7.74 457 9.43
Consumer goods 53 2.56 112 2.31
Restaurants, hotels, motels 14 0.68 53 1.09
Mining 14 0.68 39 0.80
Oil 76 3.68 165 3.40
Other 64 3.10 186 3.84
Paper products 57 2.76 87 1.79
Retail 40 1.94 148 3.05
Services 375 18.14 968 19.97
Steel 40 1.94 80 1.65
Telecommunications 36 1.74 139 2.87
Transportation 31 1.50 87 1.79
Textiles 17 0.82 35 0.72
Wholesale 70 3.39 196 4.04
Total 2,067 100.00 4,847 100.00

This table presents the Fama–French 30 industry composition of the sample, compared to all US
incorporated multinational firms in Compustat during the sample period with at least 1 year of non-
missing assets. Multinational firms are defined as firms that have at least one nonzero value of pre-tax
foreign income (PIFO), foreign tax expense (TXFO), or foreign deferred tax expense (TXDFO). The
financial and utility industries are dropped because we eliminate regulated industries from the sample. Of
the 14 firms in Compustat in the tobacco industry, none fulfill all our sample selection criteria, so the
industry is dropped from the table. Thus, the table has data on 27 industries

is 0.026 in the sample, which is slightly smaller than the mean pre-tax foreign
income of 0.028. The median firm still has more domestic income, however, with
median pre-tax domestic income of 0.036 compared to median foreign pre-tax
income of 0.019.

123
Where do firms manage earnings? 663

Table 3 Descriptive statistics

Name N Mean STD P25 P50 P75

PTDACC 11,077 0.003 0.099 -0.030 0.011 0.050

|PTDACC| 11,077 0.064 0.076 0.019 0.041 0.078
PIDOM 11,077 0.026 0.147 -0.015 0.036 0.090
PIFO 11,077 0.028 0.062 0.003 0.019 0.048
RULE OF LAW 11,077 1.065 0.487 0.780 1.087 1.406
HAVEN INTENSITY 11,077 0.177 0.179 0.038 0.146 0.250
SPE INTENSITY 11,077 0.058 0.221 0.000 0.000 0.050
FOREIGNNESS 11,077 0.381 0.224 0.208 0.358 0.519
CASH FLOW VOLATILITY 11,077 0.062 0.068 0.028 0.045 0.073
SALES VOLATILITY 11,077 0.232 0.267 0.096 0.165 0.279
NCOUNTRIES 11,077 2.348 0.919 1.609 2.398 3.045
SIZE 11,077 6.716 1.805 5.523 6.676 7.872
CPI CHANGE SUB COUNTRIES 11,077 5.205 18.087 2.113 3.003 4.586
GDP CHANGE SUB COUNTRIES 11,077 3.444 1.672 2.300 3.494 4.570

This table contains descriptive statistics for variables used in our study. The sample consists of 11,077
observations, selected as outlined in Table 1. PTDACC is pre-tax discretionary accruals, calculated using
the modified Jones model using pre-tax accruals instead of total accruals (see Eq. 2). The model is
estimated separately for each two-digit SIC code and year for which there are at least 10 observations.
|PTDACC| is the absolute value of pre-tax discretionary accruals. PIDOM is pre-tax domestic income
from Compustat. PIFO is pre-tax foreign income from Compustat. PIDOM and PIFO are scaled by total
assets at the beginning of the year. RULE OF LAW is the average rule of law of the countries in which the
firm operates. Rule of law scores for each country are obtained from the World Bank Governance
Indicators. HAVEN INTENSITY is the number of countries in which the firm discloses a subsidiary that
are tax havens divided by the total number of countries in which the firm discloses a subsidiary. A country
is designated as tax haven if it appears on at least two of the four lists of tax havens in Miedema (2008) as
per Dyreng and Lindsey (2009). SPE INTENSITY is the number of subsidiaries that are limited liability
companies, limited partnerships, or trusts divided by the total number of subsidiaries. FOREIGNNESS is
foreign sales from Compustat geographic segment data divided by total SALES from Compustat. CASH
FLOW VOLATILITY is the moving 5-year standard deviation of pre-tax cash flow from operations (i.e.
CSHO ? TXPD - XIDOC), each year running from year t - 4 to year t. SALES VOLATILITY is the
moving 5-year standard deviation of SALES from Compustat each year running from year t - 4 to year
t. NCOUNTRIES is the log of the total number of countries in which the firm discloses a subsidiary. SIZE
is the log of total assets. CPI CHANGE SUB COUNTRIES is the average change in consumer price index
for the countries in which the firm discloses material subsidiaries. GDP CHANGE SUB COUNTRIES is
the average change in GDP for the countries in which the firm discloses material subsidiaries. Data on
CPI and GDP are gathered from the World Economic Outlook Database, April 2010 edition, available
from the International Monetary Fund

RULE OF LAW has a mean (median) value of 1.065 (1.087). The raw Rule of
Law measure from the World Bank is designed so that it is mean zero, with a
standard deviation of one across all countries in a given year. Thus, the firms in the
sample tend to have most of their subsidiaries in high rule of law countries. There is
a substantial use of tax haven subsidiaries in the sample, consistent with widespread
use of tax haven subsidiaries among multinational firms in general. The variable
HAVEN INTENSITY has a mean (median) of 0.177 (0.146), indicating that 17.7 %
of the average firm’s material foreign subsidiaries are located in tax havens. Most

123
664 S. D. Dyreng et al.

firms in the sample do not report special purpose entities (at least among their
material subsidiaries)–our sample has a median SPE INTENSITY of zero and a mean
of 0.058. The mean (median) value of 0.381 (0.358) for FOREIGNNESS (foreign
sales from Compustat geographic segment data divided by total sales from
Compustat) indicates that a large percentage of sales of the firms in the sample are
earned in foreign jurisdictions. CASH FLOW VOLATILITY and SALES VOLATIL-
ITY have mean values of 0.062 and 0.232, respectively. NCOUNTRIES has a mean
(median) value of 2.348 (2.398), and, since it is the natural log of the number of
countries in which the firm reports material subsidiaries, indicates that the mean
firm has material subsidiaries in approximately 10 countries. The natural log of the
firms’ assets, SIZE, has a mean (median) value of 6.716 (6.676). The variable CPI
CHANGE SUB COUNTRIES, the measure of the average rate of inflation in
countries in which the sample firms have subsidiaries, has a mean (median) of 5.205
(3.003). The variable GDP CHANGE SUB COUNTRIES, a measure of the average
change in gross domestic product in the countries in which the sample firms have
subsidiaries, has a mean (median) of 3.444 (3.494).
Table 4 reports the correlations among these variables. The Pearson correlations
are in the upper right and the Spearman correlations are in the lower left. Consistent
with Hribar and Nichols (2007), we find positive correlations between the absolute
value of (in our case pre-tax) discretionary accruals and the volatility of cash flows
and sales and a negative correlation with firm size. The number of countries in
which the firm lists material subsidiaries is also negatively correlated with the
absolute value of pre-tax discretionary accruals. We next turn to the multivariate
tests.

5 Empirical tests and results

5.1 Foreign subsidiaries and discretionary accruals

Our first set of analyses examine whether foreign subsidiary characteristics are
associated with discretionary accruals. Three models with different combinations of
explanatory variables are estimated, with the full model as follows:
PTDACCit ¼ a0 X
þ a1 RULE OF LAWit þ a2 HAVEN INTENSITYit
þ ak CONTROLkit þ it : ð3Þ
The results are presented in Table 5, Panel A.17 In the first column, we present
the results from estimating the regression including RULE OF LAW and control
variables. In the second column, we present results from estimating the regression
including HAVEN INTENSITY and control variables. The full model with all of the
17
We use robust regression to control for outliers. In the regressions, all continuous variables are mean
centered at zero for ease of interpretation of the interaction effects (Aiken and West 1991). We multiply
the dependent variable by 100 to facilitate interpretation of the coefficients as percentages. The standard
errors in all regressions are computed after clustering observations by firm and year to mitigate the effects
of cross-sectional and intra-firm correlation in the residuals (Petersen 2009). For all regressions we
present one-tailed p values for t statistics where we have a prediction and two-tailed p values otherwise.

123
 Table 4 Correlations
1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 PTDACC -0.29* 0.51* 0.22* -0.06* -0.02 0.02* 0.02 -0.06* -0.04* 0.04* 0.06* 0.01 0.00
2 |PTDACC| 0.13* -0.33* -0.14* 0.06* 0.04* -0.01 0.01 0.29* 0.22* -0.14* -0.21* -0.03* -0.07*
3 PIDOM 0.31* -0.12* 0.14* -0.03* -0.01 0.01 -0.13* -0.14* -0.04* 0.10* 0.21* 0.02* 0.07*
Where do firms manage earnings?

4 PIFO 0.17* -0.11* 0.23* -0.11* 0.06* -0.04* 0.24* -0.01 0.02 0.20* 0.20* 0.01 0.09*
5 RULE OF LAW -0.05* 0.07* -0.00 -0.15* -0.04* -0.01 -0.11* 0.10* 0.06* -0.33* -0.27* -0.11* -0.28*
6 HAVEN INTENSITY -0.01 0.01 -0.01 0.11* -0.10* -0.04* 0.08* 0.08* 0.05* 0.04* 0.05* -0.03* 0.13*
7 SPE INTENSITY 0.02* -0.05* -0.00 0.01 -0.19* 0.04* -0.10* -0.04* -0.02* -0.13* 0.05* -0.02* -0.01
8 FOREIGNNESS 0.01 -0.00 -0.21* 0.36* -0.14* 0.13* -0.08* 0.03* -0.04* 0.28* 0.05* -0.02* 0.06*
9 CASH FLOW -0.07* 0.25* 0.01 -0.05* 0.16* 0.07* -0.15* 0.01 0.43* -0.20* -0.37* -0.02* 0.00
VOLATILITY
10 SALES VOLATILITY -0.02* 0.19* 0.09* -0.01 0.11* 0.01 -0.09* -0.09* 0.44* -0.11* -0.23* -0.02 0.01
11 NCOUNTRIES 0.02* -0.13* 0.08* 0.33* -0.41* 0.25* 0.14* 0.35* -0.24* -0.14* 0.55* 0.06* 0.08*
12 SIZE 0.02* -0.20* 0.17* 0.27* -0.33* 0.14* 0.31* 0.09* -0.39* -0.27* 0.56* 0.07* 0.07*
13 CPI CHANGE SUB 0.04* -0.05* 0.06* 0.15* -0.64* 0.01 0.10* 0.04* -0.20* -0.10* 0.41* 0.35* 0.02*
COUNTRIES
14 GDP CHANGE SUB -0.03* -0.09* 0.07* 0.11* -0.28* 0.15* 0.04* 0.06* -0.02 -0.00 0.12* 0.10* 0.06*
COUNTRIES

This table presents the bivariate Pearson (above the diagonal) and Spearman (below the diagonal) correlations. All variables are defined in Table 3. Sample selection
criteria are in Table 1. * Significance at the 5 % level or better
665

123
666 S. D. Dyreng et al.

Table 5 Discretionary accruals as a function of rule of law, tax haven intensity, and controls

Panel A: Full sample-profit and loss firm-yearsa

Prediction Model 1 Model 2 Model 3

INTERCEPT 4.024*** 2.760*** 4.027***

(7.97) (7.96) (7.99)
RULE OF LAW (-) -0.687*** -0.688***
(-3.89) (-3.90)
HAVEN INTENSITY (?) -0.104 -0.127
(-0.34) (-0.40)
SPE INTENSITY 0.699*** 0.745*** 0.697***
(3.03) (3.00) (3.02)
FOREIGNNESS 0.287 0.300 0.295
(0.77) (0.79) (0.79)
NCOUNTRIES 0.061 0.151 0.060
(0.72) (1.69) (0.72)
SIZE -0.131*** -0.111** -0.130**
(-3.00) (-2.50) (-2.95)
CPI CHANGE SUB COUNTRIES -0.002 -0.001 -0.002
(-1.48) (-1.06) (-1.49)
GDP CHANGE SUB COUNTRIES -0.247** -0.191* -0.246**
(-2.54) (-2.09) (-2.52)
SALES VOLATILITY -0.253 -0.261 -0.250
(-0.75) (-0.76) (-0.74)
CASH FLOW VOLATILITY -10.659*** -10.916*** -10.603***
(-6.86) (-7.09) (-6.87)
N 11,077 11,077 11,077
ADJRSQ 0.019 0.016 0.019

Panel B: Sub-sample of firm-years with positive pre-tax domestic income and positive pre-tax foreign
incomeb

Prediction Model 1 Model 2 Model 3

INTERCEPT 7.168*** 5.329*** 7.170***

(7.59) (9.20) (7.65)
RULE OF LAW (-) -0.980*** -0.981***
(-3.62) (-3.61)
HAVEN INTENSITY (?) 0.828** 0.836**
(1.87) (1.89)
SPE INTENSITY 1.371*** 1.608*** 1.373***
(4.17) (4.16) (4.17)
FOREIGNNESS 2.314*** 2.157*** 2.259***
(5.52) (5.18) (5.36)
NCOUNTRIES -0.261** -0.089 -0.255**
(-2.59) (-0.96) (-2.53)

123
Where do firms manage earnings? 667

Table 5 continued

Panel B: Sub-sample of firm-years with positive pre-tax domestic income and positive pre-tax foreign
incomeb

Prediction Model 1 Model 2 Model 3

SIZE -0.311*** -0.301*** -0.322***

(-5.88) (-5.74) (-6.09)
CPI CHANGE SUB COUNTRIES -0.006** -0.004** -0.006**
(-2.46) (-2.78) (-2.39)
GDP CHANGE SUB COUNTRIES -0.362** -0.292** -0.372**
(-2.66) (-2.35) (-2.75)
SALES VOLATILITY 0.543 0.442 0.495
(1.14) (0.93) (1.03)
CASH FLOW VOLATILITY -19.339*** -19.798*** -19.647***
(-7.86) (-8.18) (-8.12)
N 6,529 6,529 6,529
ADJRSQ 0.060 0.053 0.061
a
This table presents estimates from Eq. (3): PTDACCit ¼ a0 þ a1 RULE OF LAWit þ a2 HAVEN
P
INTENSITYit þ ak CONTROLkit þ it . Each of the variables shown in the table is defined in Table 3.
The dependent variable, PTDACCit, has been multiplied by 100 to ease interpretation of the coefficients.
T statistics, shown in parentheses below the coefficient estimates, are based on standard errors that are
clustered by firm and year. One tailed tests of significance are used where a signed prediction has been
made. ***, **, and * represent statistical significance at the 1, 5, and 10 % levels, respectively
b
This table presents estimates from Eq. (3): PTDACCit ¼ a0 þ a1 RULE OF LAWit þ a2 HAVEN
P
INTENSITYit þ ak CONTROLkit þ it . Only observations with positive domestic and foreign pre-tax
incomes are included. Each of the variables shown in the table is defined in Table 3. The dependent
variable, PTDACCit, has been multiplied by 100 to ease interpretation of the coefficients. T statistics,
shown in parentheses below the coefficient estimates, are based on standard errors that are clustered by
firm and year. One tailed tests of significance are used where a signed prediction has been made. ***, **,
and * represent statistical significance at the 1, 5, and 10 % levels, respectively

explanatory variables is presented in the rightmost column labeled ‘‘Model 3.’’

RULE OF LAW is negative and significant in both specifications where it appears
(-0.687 in Model 1; -0.688 in Model 3). This is consistent with the prediction that
firms with subsidiaries located in strong rule of law countries manage their earnings
less than firms with subsidiaries located in weak rule of law countries. Econom-
ically, a one standard deviation increase in RULE OF LAW lowers the level of
discretionary accruals by about 0.33 % of beginning total assets. HAVEN
INTENSITY is not significantly different from zero in Model 2 or in Model 3. A
positive coefficient would be consistent with the prediction that firms with more
subsidiaries located in tax havens have more earnings management, as proxied by
discretionary accruals.
The analysis in Panel A includes all observations regardless of whether they are
profitable firms or loss firms. However, there is evidence that discretionary accruals
can be asymmetric with respect to losses (Ball and Shivakumar 2006). Moreover, as
stated above, our hypothesis about earnings management being concentrated in tax
havens presumes that firms are profitable, tax-paying firms. Loss firms are unlikely

123
668 S. D. Dyreng et al.

HAVEN INTENSITY

Incremental Effect on Earnings Management

RULE OF LAW

Quintile of RULE OF LAW or HAVEN INTENSITY

Fig. 1 Plot of the effect of RULE OF LAW and HAVEN INTENSITY on Discretionary Accruals
(PTDACC). This figure presents estimates of the effect of RULE OF LAW and HAVEN INTENSITY on
discretionary accruals (PTDACC) when both foreign and domestic pre-tax earnings are positive. The
estimates are obtained by estimating Eq. (3), except replacing the continuous measures of RULE OF LAW
(HAVEN INTENSITY) with 5 indicator variables based on quintiles of RULE OF LAW (HAVEN
INTENSITY). To show the incremental effect, the baseline is centered on zero

to face significant tax costs from earnings management in the first place and thus are
less likely to focus their earnings management in tax havens. Accordingly, in Panel
B of Table 5, we re-estimate Eq. (3) over the set of firm-years that have positive
pre-tax foreign and positive pre-tax domestic income. The results in Panel B
indicate that, among profitable firms, earnings management is increasing in the
extent to which the firms have tax haven subsidiaries. The coefficient on HAVEN
INTENSITY is positive and significant in both specifications where it is included,
ranging from 0.828 in column (2) to 0.836 in column (3). Economically, a one
standard deviation increase in HAVEN INTENSITY increases the average level of
discretionary accruals by about 0.15 % of beginning assets. As in the broad sample
of firms, RULE OF LAW is negative and significant, as predicted, in each
specification where it is included, with a one standard deviation increase in RULE
OF LAW translating to a decrease in discretionary accruals of about 0.5 % of
beginning assets.
In terms of the control variables, the results in Table 5, Panel A, and Table 5,
Panel B, show a consistently positive coefficient on FOREIGNNESS, though
insignificant in Panel A, and a consistently negative coefficient on SIZE.
Coefficients on CASH FLOW VOLATILITY are significantly negative across all
specifications. These findings are consistent with the modified Jones model better
explaining accruals, and thus leaving less to be labeled discretionary, for large firms
and firms with low operating volatility (Hribar and Nichols 2007). We also find
consistently negative coefficients on CPI CHANGE SUB COUNTRIES, though

123
Where do firms manage earnings? 669

Table 6 Discretionary accruals on domestic pre-tax income and pre-tax foreign income and interacted
effects

Panel A: Full sample–profit and loss firm-yearsa

Prediction Model 1 Model 2 Model 3 Model 4 Model 5

INTERCEPT -0.275 -0.699*** -0.635*** -0.717*** -0.617***

(-1.72) (-4.33) (-3.92) (-4.33) (-3.79)
PIDOM (?) 22.031*** 29.050*** 29.090*** 29.000*** 28.950***
(27.33) (27.56) (27.49) (26.71) (27.25)
PIFO (?) 14.817*** 18.428*** 18.844*** 19.237*** 18.560***
(6.24) (7.73) (8.46) (7.60) (8.75)
PIDOM*RULE -6.186*** -6.465***
OF LAW (-5.35) (-5.68)
PIFO*RULE (-) -18.633*** -19.480***
OF LAW (-6.43) (-6.63)
PIDOM*HAVEN 10.605*** 9.202***
INTENSITY (3.55) (3.75)
PIFO*HAVEN (?) -0.522 7.126*
INTENSITY (-0.10) (1.39)
CONTROLS NO YES YES YES YES
N 11,077 11,077 11,077 11,077 11,077
ADJRSQ 0.214 0.333 0.343 0.338 0.346
PIDOM-PIFO 7.215*** 10.622*** 10.246*** 9.763*** 10.390***
(3.43) (5.11) (5.08) (4.43) (5.44)

Panel B: Sub-sample of firm-years with positive pre-tax domestic income and positive pre-tax foreign
incomeb

Prediction Model 1 Model 2 Model 3 Model 4 Model 5

INTERCEPT 1.420*** 0.472** 0.456** 0.479** 0.466**

(6.68) (2.48) (2.41) (2.53) (2.48)
PIDOM (?) 5.517*** 14.731*** 14.360*** 14.726*** 14.350***
(4.86) (11.54) (10.95) (11.68) (11.16)
PIFO (?) 4.984** 8.720*** 9.497*** 8.470*** 9.172***
(2.51) (3.30) (3.59) (3.19) (3.51)
PIDOM*RULE 3.122 3.022
OF LAW (1.48) (1.45)
PIFO*RULE (-) -5.672* -6.417**
OF LAW (-1.68) (-1.93)
PIDOM*HAVEN -6.174 -11.020**
INTENSITY (-1.45) (-2.33)
PIFO*HAVEN (?) 11.747* 12.650*
INTENSITY (1.41) (1.52)
CONTROLS NO YES YES YES YES

123
670 S. D. Dyreng et al.

Table 6 continued
Panel B: Sub-sample of firm-years with positive pre-tax domestic income and positive pre-tax foreign
incomeb

Prediction Model 1 Model 2 Model 3 Model 4 Model 5

N 6,529 6,529 6,529 6,529 6,529

ADJRSQ 0.010 0.163 0.143 0.162 0.144
PIDOM-PIFO 0.533 6.011* 4.864 6.256* 5.179*
(0.24) (1.99) (1.62) (2.09) (1.77)
a
This table presents estimates from Eq. (11): PTDACCit ¼ c þ p0 PIDOMit þ p1 PIDOMit  RULE OF
P
LAWit þ p2 PIDOMit  HAVEN INTENSITYit þ k pk PIDOMit  CONTROLKit þ x0 PIFOit þ x1 PIFOit 
P P
RULE OF LAWit þ x2 PIFOit  HAVEN INTENSITYit þ k xk PIFOit  CONTROLkit þ j cj CONTROLitj þ
eit . Each of the variables shown in the table is defined in Table 3. The dependent variable, PTDACCit, has
been multiplied by 100 to ease interpretation of the coefficients. Each of the variables has been mean-
centered to facilitate interpretation of the interaction terms. T statistics, shown in parentheses below the
coefficient estimates, are based on standard errors that are clustered by firm and year. One tailed tests of
significance are used where a signed prediction has been made. ***, **, and * represent statistical
significance at the 1, 5, and 10 % levels, respectively
b
This table presents estimates from Eq. (11): PTDACCit ¼ c þ p0 PIDOMit þ p1 PIDOMit  RULE OF
P
LAWit þ p2 PIDOMit  HAVEN INTENSITYit þ k pk PIDOMit  CONTROLKit þ x0 PIFOit þ x1 PIFOit 
P P
RULE OF LAWit þ x2 PIFOit  HAVEN INTENSITYit þ k xk PIFOit  CONTROLkit þ j cj CONTROLitj þ
eit . Only observations with positive foreign and domestic pre-tax incomes are included in the tests. Each
of the variables shown in the table is defined in Table 3. The dependent variable, PTDACCit, has been
multiplied by 100 to ease interpretation of the coefficients. Each of the variables has been mean-centered
to facilitate interpretation of the interaction terms. T statistics, shown in parentheses below the coefficient
estimates, are based on standard errors that are clustered by firm and year. One tailed tests of significance
are used where a signed prediction has been made. ***, **, and * represent statistical significance at the 1,
5, and 10 % levels, respectively

insignificant in Panel A, and GDP CHANGE SUB COUNTRIES. We have no

directional prediction for these variables but include them to control for
macroeconomic conditions in the countries in which the firm operates.
To graphically illustrate the effect of RULE OF LAW and HAVEN INTENSITY on
PTDACC, we replace the continuous variables RULE OF LAW and HAVEN
INTENSITY in Panel B of Table 5 with indicator variables that capture quintiles of
RULE OF LAW and HAVEN INTENSITY. The coefficient values for each of the
quintiles are then plotted in Fig. 1. As the figure shows, the effect of RULE OF LAW
on PTDACC is greatest when moving from the first to second quintile and then
continues to strengthen through the fifth quintile. The effect of HAVEN INTENSITY
on PTDACC appears to be greatest when moving from the second to third quintile of
HAVEN INTENSITY, and again from the fourth to fifth quintiles, with the largest
effect concentrated in the quintile containing the most haven-intense firms.

5.2 Discretionary accruals on foreign and domestic pre-tax income

In Table 6 we examine the mapping of foreign and domestic pre-tax income into
discretionary accruals. The idea is to examine how earnings management can be

123
Where do firms manage earnings? 671

attributed to domestic versus foreign activity by directly including measures of such

activity. Since U.S. firms are required to disclose the breakdown of their pre-tax
income (PI) into pre-tax income from domestic sources (pre-tax income—domestic;
PIDOM) and pre-tax income from foreign sources (pre-tax income—foreign;
PIFO), those data are available for a broad sample.18 First, we define the rate of
company-wide discretionary accruals as:
PTDACCit
PTDACC RATEit ¼ : ð4Þ
PIit
This ratio captures the fraction of total pre-tax income that is estimated to be
discretionary accrual income. Re-arranging the terms, and letting dit represent the
rate of pre-tax discretionary accruals, we obtain:
PTDACCit ¼ dit PIit : ð5Þ
In this study, we ask where earnings are managed. One way to provide evidence
on this question is to test whether the discretionary accrual rate on domestic income
is different from the discretionary accrual rate on foreign income. That is, we would
like to know if dit is the same when pre-tax income is derived from domestic sources
(PIDOM) as when it is derived from foreign sources (PIFO). To test this, we can
modify Eq. (5) as follows:
PTDACCit ¼ ait PIDOMit þ bit PIFOit ; ð6Þ
where ait represents the rate at which domestic pre-tax income maps into discre-
tionary accruals, and bit represents the rate at which foreign pre-tax income maps
into discretionary accruals. Empirical implementation requires dropping the sub-
scripts i and t on the coefficients a and b and the inclusion of an error term. We also
include an intercept for completeness. Thus, we estimate an OLS regression of the
following form:
PTDACCit ¼ c þ aPIDOMit þ bPIFOit þ it : ð7Þ
Our first test, for which we have no directional prediction, is whether a = b, or in
words, whether there is a difference between the rate at which PIDOM and PIFO
map into PTDACC. We are also interested in the cross sectional determinants of a
and b. In particular, we hypothesize that b is a function of the rule of law of the
foreign countries in which the firm operates material subsidiaries and that b may
also be a function of whether the firm operates in tax havens. We can extend Eq. (6)
to facilitate these hypotheses as follows:
ait ¼ p0 þ p1 RULE OF LAWit þ p2 HAVEN INTENSITYit ð8Þ
and
bit ¼ x0 þ x1 RULE OF LAWit þ x2 HAVEN INTENSITYit ð9Þ
Substituting Eqs. (8) and (9) into Eq. (6) gives the following:

18
The breakdown of pre-tax income into pre-tax domestic income and pre-tax foreign income is required
by the SEC to be included in the tax footnote of firm’s financial statements to correspond with the
breakdown of tax expense into domestic and foreign components.

123
672 S. D. Dyreng et al.

PTDACCit ¼ p0 PIDOMit þ p1 PIDOMit  RULE OF LAWit þ p2 PIDOMit

 HAVEN INTENSITYit þ x0 PIFOit þ x1 PIFOit ð10Þ
 RULE OF LAWit þ x2 PIFOit  HAVEN INTENSITYit :
To empirically estimate Eq. (10), we also add an intercept, the variables from
Eq. (3), and an error term:19
PTDACCit ¼ c þ p0 PIDOMit þ p1 PIDOMit  RULE OF LAWit
þ p2 PIDOMit  HAVEN INTENSITYit
X
þ pk PIDOMit  CONTROLKit þ x0 PIFOit þ x1 PIFOit
k ð11Þ
 RULE OF LAWit þ x2 PIFOit  HAVEN INTENSITYit
X X
þ xk PIFOit  CONTROLkit þ cj CONTROLitj þ eit :
k j

Hypothesis 1 predicts that firms with subsidiaries in countries with a high rule of
law will have less earnings management in foreign earnings. That is, Hypothesis 1
predicts that x1 \ 0. Hypothesis 2 predicts that firms with subsidiaries in tax haven
countries will have more earnings management in foreign earnings. That is,
Hypothesis 2 predicts that x2 [ 0.
Table 6 presents the results of the estimations of the above equations with Panel
A estimated over all firm-years and Panel B estimated over firm-years with positive
pre-tax foreign and positive pre-tax domestic income. In Panel A of Table 6, Model
1 examines how discretionary accruals are explained by pre-tax domestic income
and pre-tax foreign income without control variables (that is, Eq. (7) above). The
coefficients on both PIDOM and PIFO are positive and significant, with values of
22.031 and 14.817, respectively. The coefficient on PIDOM is significantly greater
than the coefficient on PIFO, as indicated by the test in the bottom row of the table.
The interpretation is that, at the mean, a dollar of pre-tax domestic income results in
approximately 22 cents of pre-tax discretionary accruals, whereas a dollar of pre-tax
foreign income results in approximately 15 cents of pre-tax discretionary accruals.
Dollar-for-dollar, this is consistent with domestic income being subject to more
earnings management than foreign income.
We expand this in Model 2 by first adding HAVEN INTENSITY and RULE OF
LAW and the control variables from Table 5 both as main effects and interacted
separately with PIDOM and PIFO: SPE INTENSITY, FOREIGNNESS, NCOUN-
TRIES, SIZE, CPI CHANGE SUB COUNTRIES, GDP CHANGE SUB COUN-
TRIES, SALES VOLATILITY, and CASH FLOW VOLATILITY. For the sake of
brevity, the coefficients on those variables are not reported in Table 6, and their
presence is instead noted by a ‘‘YES’’ in the rows labeled ‘‘CONTROLS.’’ Notice
that the control variables result in a large increase in explanatory power from an R2
of approximately 21 % in Model 1 to approximately 33 % in Model 2. With the
controls included, both PIDOM and PIFO remain positive and significant (29.050
19
We use the superscript j to indicate a vector of controls that includes all of the controls in the vector k
and also includes RULE OF LAW and HAVEN INTENSITY.

123
Where do firms manage earnings? 673

and 18.428, respectively), and PIDOM continues to be significantly greater than

PIFO.
In Models 3 and 4, we interact PIDOM and PIFO with RULE OF LAW and
HAVEN INTENSITY. We expect the effects of rule of law to be concentrated in pre-
tax foreign income. Accordingly, we predict a negative coefficient on the interaction
of PIFO and RULE OF LAW. We also allow RULE OF LAW and HAVEN
INTENSITY to interact with PIDOM for the sake of completeness, but we make no
prediction on the sign for these interactions. Including these interactions allows for
the possibility that firms that go to low rule of law countries or tax havens are also
those that manage earnings generally, both in domestic and foreign income. Model 3
reveals that the coefficient on PIFO  RULE OF LAW is negative and significant, as
predicted (-18.633). This is consistent with pre-tax foreign income being subject to
more (less) earnings management when the firm’s foreign operations are in
countries with relatively weak (strong) rule of law.
To facilitate interpretation of the interaction terms, all variables have been mean-
centered. Thus, for the mean firm in Model 3, 18.8 cents of each foreign pre-tax
dollar maps into a dollar of pre-tax discretionary accruals. A one standard deviation
increase in RULE OF LAW is associated with a 9.1 cent per dollar lower rate at
which pre-tax foreign income maps into discretionary accruals.20 The coefficient on
the interaction of RULE OF LAW and PIDOM is also significant, but the effect is
much smaller economically. A one standard deviation increase in RULE OF LAW is
associated with a three cent per dollar lower rate at which pre-tax domestic income
maps into discretionary accruals.
Model 4 includes the interaction of PIDOM and PIFO with HAVEN INTENSITY.
As with RULE OF LAW, we expect that the effects of HAVEN INTENSITY will be
concentrated in foreign income. We make no prediction for the interaction of
HAVEN INTENSITY and domestic income. In this specification, the coefficient on
PIFO  HAVEN INTENSITY is negative but insignificant.
In Model 5 we include both RULE OF LAW and HAVEN INTENSITY. Consistent
with the results in Models 1–4, the coefficient on PIFO  RULE OF LAW is negative
and significant (-19.480) in Model 5. The coefficient on PIFO  HAVEN
INTENSITY becomes positive and significant in Model 5 when we include all of
the interactions at the same time. The main effects of both PIDOM and PIFO
remain positive and significant, with values 28.950 and 18.560, respectively.
In Table 6, Panel B, we re-estimate the regressions in Panel A, except we use
only the subsample of firms that have positive pre-tax domestic income and positive
pre-tax foreign income. As in the Table 5 analysis, we include this test because of
evidence that discretionary accruals can be asymmetric with respect to losses (Ball
and Shivakumar 2006) and because our prediction for HAVEN INTENSITY is most
applicable for profitable (tax-paying) firms. Model 1 examines how discretionary
accruals are explained by pre-tax domestic income and pre-tax foreign income
without control variables. The coefficients on both PIDOM and PIFO are positive
and significant, with values of 5.517 and 4.984, respectively. In Model 2 we again

20
Calculated as -18.633 9 0.487 = -9.1. Note that 0.487 is the standard deviation of RULE OF LAW
from Table 3.

123
674 S. D. Dyreng et al.

include each of the control variables SPE INTENSITY, FOREIGNNESS, NCOUN-

TRIES, SIZE, CPI CHANGE SUB COUNTRIES, GDP CHANGE SUB COUN-
TRIES, SALES VOLATILITY, and CASH FLOW VOLATILITY interacted separately
for PIDOM and PIFO. Model 2 also includes the main effect of each of these
variables and the main effects of RULE OF LAW and HAVEN INTENSITY. The
controls increase the explanatory power of the model and cause the coefficients on
both PIDOM and PIFO to increase to 14.731 and 8.720.
Models 3 and 4 consider the effects of interacting PIDOM and PIFO with RULE
OF LAW and HAVEN INTENSITY, respectively. We continue to find a negative
coefficient on the interaction of PIFO with RULE OF LAW. Consistent with Model
5 in Table 6, Panel A, the coefficient on PIFO  HAVEN INTENSITY is positive and
weakly significant as predicted in both Models 4 and 5. This indicates that, among
profitable firms, earnings management of foreign income is increasing in the extent
of subsidiaries in tax havens.
To graphically illustrate the effect of RULE OF LAW and HAVEN INTENSITY on
the mapping of PIFO into PTDACC, we re-estimate the model but replace the
continuous variables RULE OF LAW and HAVEN INTENSITY with indicator
variables that capture quintiles of RULE OF LAW and HAVEN INTENSITY. The
coefficient value for each of the quintiles is then plotted in Fig. 2. As the figure
shows, the effect of RULE OF LAW is greatest when moving from the first to second
quintile. Similarly, the effect of HAVEN INTENSITY is greatest when moving from
the first to second quintile of HAVEN INTENSITY, and increasing thereafter, with
the largest effect concentrated in the quintile containing the most haven-intense
firms.

5.3 Additional tests

5.3.1 Absolute value of discretionary accruals

To examine the robustness of the results, we perform a number of additional tests. In

this subsection, we change the dependent variable to the absolute value of
discretionary accruals (|PTDACC|), to capture both income-increasing and income-
decreasing earnings management.21 By reflecting both positive and negative
accruals, the absolute value of discretionary accruals can capture activities such as
the building up and drawing down of cookie jar reserves, without needing a
specification of the exact periods of management. Table 7 is the analog to Panel A
of Table 5 but with the absolute value of discretionary accruals as the dependent
variable. As in the earlier analysis, RULE OF LAW is negatively associated with
earnings management, with coefficients of -0.306 in both Model 1 and Model 3.
This indicates that firms with subsidiaries in high rule of law countries engage in
less earnings management than firms with subsidiaries in low rule of law countries.
Unlike Table 5, Panel A, the coefficient on HAVEN INTENSITY is positive and
21
Many studies use the absolute value of discretionary accruals, including Dechow and Dichev (2002),
Frankel et al. (2002), Klein (2002), Chung and Kallapur (2003), Myers et al. (2003), Leuz et al. (2003),
and Bergstresser and Philippon (2006).

123
Where do firms manage earnings? 675

HAVEN INTENSITY

Incremental Effect on Earnings Management

RULE OF LAW

Quintile of RULE OF LAW or HAVEN INTENSITY

Fig. 2 Plot of the effect of RULE OF LAW and HAVEN INTENSITY on the Mapping of Pre-tax Income
from Foreign Operations (PIFO) into Discretionary Accruals (PTDACC). This figure presents estimates
of the effect of RULE OF LAW and HAVEN INTENSITY on the mapping of pre-tax income from foreign
operations PIFO into discretionary accruals (PTDACC) when both foreign and domestic pre-tax earnings
are positive. The estimates are obtained by estimating Eq. (11), except replacing the continuous measures
of RULE OF LAW (HAVEN INTENSITY) with 5 indicator variables based on quintiles of RULE OF LAW
(HAVEN INTENSITY). To show the incremental effect, the baseline is centered on zero

significant, as predicted, with a value of 0.649 in Model 2 and 0.649 in Model 3. We

cannot fully explain the economic drivers of a positive relation with the absolute
value of discretionary accruals but no relation with signed discretionary accruals.
The coefficients on the control variables are generally consistent between Tables 5,
Panel A, and 7 in terms of sign and significance. The exceptions are the volatility
controls. SALES VOLATILITY was insignificant in Table 5 but is positive and
significant in Table 7. CASH FLOW VOLATILITY was negative in Table 5 but is
positive in the Table 7 analysis of absolute value of discretionary accruals. Based on
Hribar and Nichols (2007), we expect a positive coefficient on both volatility
controls in Table 7, where we are explaining the absolute value of discretionary
accruals.
Table 8 is the analog to Panel A of Table 6 but with the absolute value of
discretionary accruals as the dependent variable. The analysis examines how RULE
OF LAW and HAVEN INTENSITY affect the mapping of the absolute value of
foreign pre-tax income into the absolute value of discretionary accruals. Model 1 of
Table 8 reveals that the coefficients on both |PIDOM| and |PIFO| are positive and
significant, with values of 8.913 and 4.215, respectively. The coefficient on
|PIDOM| is significantly greater than the coefficient on |PIFO|, as indicated by the
bottom row of the table. Model 2 includes the control variables from Table 6. As in
Table 6, the R2 of the models increase substantially with the addition of the control
variables. With the controls included, both PIDOM and PIFO remain positive and
significant.

123
676 S. D. Dyreng et al.

Table 7 Absolute value of discretionary accruals as a function of rule of law, tax haven intensity, and
controls
Prediction Model 1 Model 2 Model 3

INTERCEPT 5.664*** 5.098*** 5.657***

(11.90) (17.18) (11.97)
RULE OF LAW (-) -0.306** -0.306**
(-2.20) (-2.14)
HAVEN INTENSITY (?) 0.649*** 0.649***
(3.83) (3.61)
SPE INTENSITY 0.223* 0.255* 0.236**
(1.69) (1.86) (1.82)
FOREIGNNESS 0.419*** 0.383** 0.380**
(3.22) (2.92) (2.92)
NCOUNTRIES -0.079* -0.038 -0.078*
(-2.12) (-1.06) (-2.11)
SIZE -0.190*** -0.188*** -0.197***
(-8.83) (-8.55) (-8.98)
CPI CHANGE SUB COUNTRIES -0.002** -0.001 -0.001*
(-2.37) (-0.83) (-1.97)
GDP CHANGE SUB COUNTRIES -0.190** -0.173** -0.198**
(-2.82) (-2.84) (-2.95)
SALES VOLATILITY 1.406*** 1.388*** 1.399***
(6.85) (6.67) (6.70)
CASH FLOW VOLATILITY 11.276*** 10.845*** 10.948***
(11.26) (10.64) (10.77)
N 11,077 11,077 11,077
ADJRSQ 0.097 0.096 0.098

This table presents estimates from the following model: jPTDACCit j ¼ a0 þ a1 RULE OF LAWit þ
P
a2 HAVEN INTENSITYit þ ak CONTROLkit þ it . Each of the variables shown in the table is defined in
Table 3. The dependent variable, jPTDACCit j, has been multiplied by 100 to ease interpretation of the
coefficients. T statistics, shown in parentheses below the coefficient estimates, are based on standard
errors that are clustered by firm and year. One tailed tests of significance are used where a signed
prediction has been made. ***, **, and * represent statistical significance at the 1, 5, and 10 % levels,
respectively

In Models 3 and 4, we interact |PIDOM| and |PIFO| with RULE OF LAW,

HAVEN INTENSITY, and all control variables. As indicated earlier, we expect the
effects of foreign rule of law to be concentrated in pre-tax foreign income.
Accordingly, we predict a negative coefficient on the interaction of |PIFO| and
RULE OF LAW. Model 3 reveals that the coefficient on jPIFOj  RULE OF LAW is
negative and significant, as predicted, at -5.872. This is consistent with pre-tax
foreign income being subject to more (less) earnings management when the firm’s
foreign operations are in countries with relatively weak (strong) rule of law. A one
standard deviation increase in RULE OF LAW is associated with a 2.9 cent per
dollar lower rate at which pre-tax foreign income maps into discretionary accruals.

123
Where do firms manage earnings? 677

Table 8 Absolute value of discretionary accruals on pre-tax domestic income and pre-tax foreign
income and interacted effects
Prediction Model 1 Model 2 Model 3 Model 4 Model 5

INTERCEPT 3.407*** 3.785*** 3.779*** 3.784*** 3.789***

(38.35) (39.84) (39.41) (40.02) (40.79)
|PIDOM| (?) 8.913*** 6.697*** 6.778*** 6.768*** 7.374***
(12.90) (8.00) (7.92) (8.26) (9.17)
|PIFO| (?) 4.215*** 4.147*** 4.338*** 3.935*** 3.979***
(4.51) (3.56) (3.72) (3.53) (3.25)
|PIDOM|*RULE OF LAW -0.539 -2.547***
(-0.64) (-3.84)
|PIFO|*RULE OF LAW (-) -5.872*** -6.715***
(-5.00) (-5.06)
|PIDOM|*HAVEN 2.754 1.537
INTENSITY (1.37) (0.66)
|PIFO|*HAVEN (?) 6.113** 7.776**
INTENSITY (2.03) (2.58)
CONTROLS NO YES YES YES YES
N 11,077 11,077 11,077 11,077 11,077
ADJRSQ 0.057 0.158 0.159 0.162 0.217
PIDOM-PIFO 4.699*** 2.620 3.053* 3.543** 3.545**
(3.92) (1.68) (1.97) (2.47) (2.35)

This table presents estimates from the following model: jPTDACCit j ¼ c þ p0 jPIDOMit j þ p1 jPIDOMit j
P K
 RULE OF LAWit þ p2 jPIDOMit j  HAVEN INTENSITYit þ k pk jPIDOMit j  CONTROL P it þ x0
jPIFOit j þ x1 jPIFOit j  RULE OF LAWit þ x2 jPIFOit j  HAVEN INTENSITYit þ k xk jPIFOit j 
P
CONTROLkit þ j cj CONTROLitj þ eit . Each of the variables shown in the table is defined in Table 3. The
dependent variable, jPTDACCit j, has been multiplied by 100 to ease interpretation of the coefficients.
Each of the variables has been mean-centered to facilitate interpretation of the interaction terms.
T statistics, shown in parentheses below the coefficient estimates, are based on standard errors that are
clustered by firm and year. One tailed tests of significance are used where a signed prediction has been
made. ***, **, and * represent statistical significance at the 1, 5, and 10 % levels, respectively

In contrast to the effect of RULE OF LAW on management of foreign income, the

coefficient on the interaction of RULE OF LAW and |PIDOM| is insignificant,
consistent with the rule of law of the firm’s foreign operations having little or no
effect on the firm’s management of domestic income.
Model 4 includes the interaction of |PIDOM| and |PIFO| with HAVEN
INTENSITY. As predicted, the coefficient on jPIFOj  HAVEN INTENSITY is
positive and significant (6.113), consistent with firms with extensive tax haven
subsidiaries engaging in more earnings management of their foreign pre-tax income.
A one standard deviation increase in HAVEN INTENSITY is associated with
additional discretionary accruals of 1.1 cents per dollar of pre-tax foreign income.
In Model 5 we include all of the interactions at the same time. Consistent with the
results in Models 1–4, the coefficient on jPIFOj  RULE OF LAW is negative and

123
678 S. D. Dyreng et al.

significant (-6.715) and the coefficient on jPIFOj  HAVENINTENSITY is positive

and significant (7.776). The main effects of both |PIDOM| and |PIFO| remain
positive and significant with values 7.374 and 3.979, respectively. Thus, across the
tests with the absolute value of discretionary accruals as the dependent variable, the
results suggest that earnings management of foreign income is decreasing in the rule
of law of the firm’s foreign subsidiaries and increasing in the tax haven intensity of
its foreign subsidiaries. Apart from those results, domestic income appears to be
managed more than foreign income.

5.3.2 Additional control variables

In this subsection, we include additional control variables in the regressions

specified in Eqs. (3) and (11). Specifically, in untabulated results, we include the
following control variables: firm age, leverage, the market to book ratio, the market
value of equity, a proxy for capital intensity, a proxy for intangible intensity, and a
variable that captures the fraction of loss years over the firm’s past 10 fiscal years.
These variables are drawn from prior research that uses similar dependent variables
(for example, Hribar and Nichols 2007; Francis et al. 2005). Results remain
statistically and economically similar with the inclusion of these variables.
We also re-estimate all our tests including a control for firm-level governance, as
captured by the G-Score from Gompers et al. (2003). Because the inclusion of this
variable reduces our sample size by over 20 %, we do not include the variable in our
main analyses. However, in untabulated results, we find the inclusion of the variable
on the subsample with available data does not lead to material changes in our
conclusions, with all coefficients of interest in Tables 5, 6, 7, and 8 remaining
statistically significant with similar magnitudes.
In addition, in untabulated tests, we include pretax income as a control variable in
Table 5, Panels A and B, and in Table 7. Inclusion of pre-tax income in these
regressions slightly increases the coefficient and t statistic on RULE OF LAW and
slightly decreases the coefficient and t statistics on HAVEN INTENSITY. However,
the conclusions remain unchanged.
Finally, we include two variables to control for determinants of book-tax
conforming earnings management. For example, per Badertscher et al. (2009) firms
with high quality auditors are more likely to manage earnings in a book-tax
conforming manner-the same type one would expect in a tax haven. Thus, we
include an indicator variable set equal to one when the company has a non-Big N
auditor. Also based on Badertscher et al. (2009), we control for net operating loss
(NOL) carryforwards. The main findings stay the same with these additional
controls.22
Finally, we find that the inclusion of two-digit SIC code fixed effects does not
alter our findings, which is not surprising because the dependent variable was
constructed in regressions estimated by two-digit SIC code.

22
We thank an anonymous referee for this suggestion.

123
Where do firms manage earnings? 679

5.3.3 Controlling for performance

Kothari et al. (2005) describe how performance matching can improve modified
discretionary accruals models by controlling for extreme performance that makes
estimation of discretionary accruals difficult. Dechow et al. (2011) present a new
method of improving the modified Jones model based on researcher-predicted
timing of accrual reversals. Both of these adjustments are only improvements to the
model under certain conditions. For performance matching, if performance is
correlated with earnings management, the power of the model is actually reduced as
it throws the proverbial baby out with the bathwater. For the Dechow et al. (2011)
model, the modified Jones model’s ability to detect discretionary accruals is only
improved if the researcher has a prediction of when the managed accruals will
reverse.
We control for performance by adding return-on-assets to our discretionary
accrual regression models in untabulated tests. The results are qualitatively the same
for our test variables except for the coefficient on HAVEN INTENSITY in Table 5,
Panel B, which becomes insignificant. We do not have predictions about when
managed accruals will reverse in our setting and thus do not test for reversals in our
analyses.

5.3.4 Alternative measures of rule of law and haven intensity

We test the robustness of our results to different definitions of our main variables.
First, we replace HAVEN INTENSITY throughout the study with the natural log of
the number of tax haven countries in which the firm operates and find similar
results. Second, we replace RULE OF LAW throughout the study with a variable that
counts the number of ‘‘corrupt’’ countries in which the firm operates and we find
that firms in more corrupt countries engage in more foreign earnings management.23
In sum, our results do not appear to be highly dependent on specific definitions of
tax haven intensity or rule of law.

5.3.5 Clustering standard errors by industry and year

To mitigate concerns that there could be industry clustering in our data, in

untabulated tests we re-estimate all results but cluster the standard errors by industry
and year rather than by firm and year. We use the Fama–French 30 industry
classification used in Table 2, when implementing this procedure. While this limits
the sample to just 27 clusters on the industry dimension, and some standard errors
become larger as might be expected, our conclusions are unchanged both
qualitatively and quantitatively.

23
A corrupt country was defined to be any country in the most corrupt quartile of the World Bank’s
Corruption Index.

123
680 S. D. Dyreng et al.

5.3.6 Restatements

In further robustness tests, we replace pre-tax discretionary accruals from Eq. (3) with
a binary variable that equals one if the firm restated its financial statements for that
year. Compared with discretionary accruals, restatements have the advantage that
they are independent of the researcher, as they are not based on a researcher-generated
model of accruals. However, restatements have two main disadvantages. First,
restatements will not capture within-GAAP earnings management. Second, and most
importantly, for a restatement to occur the GAAP violation must be discovered. Thus,
the most successful earnings management may never result in a restatement, either
because it was within GAAP to begin with or because it was well hidden and never
discovered. Since variation in the risk of detection across foreign operations is at the
heart of this paper’s hypotheses (that is, firms manage earnings where they expect the
likelihood of detection to be the lowest), restatements are less than ideal for this paper
(Hennes et al. 2008). (The same would hold true for SEC AAERs.) Indeed, a recent
study by Srinivasan et al. (2011) reports that foreign firms from countries with weak
rule of law are less likely to restate earnings than those companies from strong rule of
law locations, even though the companies from the weak rule of law countries have
more earnings management. The authors conclude that home country enforcement
affects the likelihood of firms reporting an irregularity, and thus less frequent
restatements do not indicate high quality earnings, at least for their sample.
Because the dependent variable is an indicator variable, we estimate the model
with logistic regression. We include a number of additional controls that have been
used in prior research examining accounting restatements and fraud.24 Restatements
are gathered from the Government Accountability Office (GAO) and cover the years
1995–2005. The new data requirements (the restatement variable plus the additional
controls) reduce the sample size to 6,699 firm-years, of which 269 are restatement
firm-years.
In Table 9, we report results from the logistic regression. Consistent with our
predictions, we find that increases in RULE OF LAW significantly reduce the
likelihood of a restatement. The unconditional probability of restatement in our
sample given the explanatory variables is about 4 %. A one standard deviation
improvement in RULE OF LAW decreases the probability of a restatement by
0.9–3.1 %. On the other hand, we find no evidence that HAVEN INTENSITY is
associated with restatements.

5.3.7 Propensity score matching

Current research suggests that propensity score matching can improve the reliability
of results in observational studies (Armstrong et al. 2010). Accordingly, we re-
estimate the results in Tables 5, Panel A, 7, and 9 using a propensity score matching
24
The additional control variables include change in receivables, change in inventory, change in cash
sales, change in return-on-assets, change in the number of employees, the level of ‘‘soft’’ assets, an
indicator for whether the firm issued debt or equity in the period, an indicator for whether the firm has
outstanding leases, a measure of ex ante financing needs, Altman’s Z, and industry fixed effects. See
Dechow et al. (2011) for detailed definitions of these variables.

123
Where do firms manage earnings? 681

Table 9 Restatement firm-

Prediction Restatement year
years as a function of rule of
law, tax haven intensity, and
INTERCEPT -3.251***
controls
(17.78)
RULE OF LAW (-) -0.591***
(6.29)
HAVEN INTENSITY (?) -0.094
(0.02)
SPE INTENSITY 0.923***
(13.59)
FOREIGNNESS 0.715*
(3.76)
NCOUNTRIES 0.156
(1.28)
This table presents results from
SIZE 0.075
a logistic regression where the
dependent variable is one for (1.22)
firm-years where financial CPI CHANGE SUB COUNTRIES -0.123***
statements were restated. The (6.98)
independent variables shown in
the table are described in GDP CHANGE SUB COUNTRIES -0.104
Table 3. Additional control (1.57)
variables are included but not SALES VOLATILITY 0.769***
presented (see footnote 24).
(7.68)
Chi-square statistics, shown in
parentheses below the CASH FLOW VOLATILITY -1.830
coefficient estimates, are based (1.45)
on standard errors that are
Additional controls Yes
clustered by firm and year. ***,
**, and * represent statistical N 6,699
significance at the 1, 5, and Pseudo RSQ 0.041
10 % levels, respectively

approach, as follows. First, we keep only those observations in the lowest quintile of
RULE OF LAW or the highest two quintiles of RULE OF LAW. Firm-years that are
in the lowest quintile of RULE OF LAW are assigned to the ‘‘treatment’’ group, and
firm-years that are in the highest two quintiles of RULE OF LAW are assigned to the
potential ‘‘control’’ group. We then fit a logistic regression predicting the treatment
as a function of the control variables from Table 5. We use the predicted probability
of receiving the treatment to match treatment firm-years to control firm-years,
requiring matches to have predicted probabilities within 10 % of each other and
come from the same industry. Once treatment and control firm-years are matched,
we compare mean values for PTDACC and |PTDACC| across the two groups, and
we compare the frequency of observed restatements across the two groups.25 Results
25
Not every treatment firm will match with a control firm, and the propensity score approach involves a
trade-off. If the matching process is relaxed so that more firms match, then the resulting matches will be
less precise. Conversely, if the matching is required to be very precise, then there will be fewer successful
matches.

123
682 S. D. Dyreng et al.

Table 10 Propensity score test of rule of law on discretionary accruals, absolute value of discretionary
accruals, and restatements

Panel A: Comparison of mean discretionary accruals and mean absolute value of discretionary accruals
across strong and weak RULE OF LAWa

Variable N Strong RULE OF LAW Weak RULE OF LAW

Mean Mean

PTDACC 1,300 0.064*** 1.033

|PTDACC| 1,300 6.205** 6.802

Panel B: Frequency of restatement firm-years across strong and weak RULE OF LAWb

Strong RULE OF LAW Weak RULE OF LAW

Did not restate 845 758

Restated 25 41
Total 870 799
Chi-square 5.590
Prob Chi-square = 0 0.009
a
This table shows the mean of discretionary accruals and signed discretionary accruals of firm-years in
weak RULE OF LAW countries relative to a matched sample of firm-years in strong RULE OF LAW
countries. Firm-years are matched based on their propensity to have weak RULE OF LAW—defined as
being in the lowest quintile of RULE OF LAW. Control firm-years are drawn from the highest two
quintiles of RULE OF LAW. Matches are required to be in the same industry and have propensity scores
(ranging from 0 to 1) within 0.1 of each other. The number of observations varies between Panel A and
Panel B because not all firm-years have restatement data
b
This table shows the frequency of restatement of firm-years in weak RULE OF LAW countries relative
to a matched sample of firm-years in strong RULE OF LAW countries. Firm-years are matched based on
their propensity to have weak RULE OF LAW—defined as being in the lowest quintile of RULE OF LAW.
Control firm-years are drawn from the highest two quintiles of RULE OF LAW. Matches are required to
be in the same industry and have propensity scores (ranging from 0 to 1) within 0.1 of each other. The
number of observations varies between Panel A and Panel B because not all firm-years have restatement
data

are presented in Table 10. Panel A of Table 10 shows that 1,300 firms with weak
RULE OF LAW matched to 1,300 firms with strong RULE OF LAW. The table
shows that firms in the weak RULE OF LAW sample have statistically higher mean
values of PTDACC and |PTDACC| than firms in the strong RULE OF LAW sample,
consistent with the results in Tables 5 and 7. Panel B shows the frequency of
restatements across the two groups. In this panel, the number of observations drops
because matching firms were not required to have restatement data. The panel
shows that of 799 firm-years with weak RULE OF LAW, 41 were restated, while
only 25 of 870 were restated in the strong RULE OF LAW group. The difference is
statistically significant with a p value of 0.009. This finding is consistent with the
results presented in Table 9 that firms with subsidiaries in countries with a weak
rule of law are more likely to restate earnings.
In Table 11, we repeat the exercise of Table 10 but for HAVEN INTENSITY. We
find no evidence that firms with greater tax haven intensity have higher mean values
of PTDACC and |PTDACC|, although the differences are in the direction that we

123
Where do firms manage earnings? 683

Table 11 Propensity score test of haven intensity on discretionary accruals, absolute value of discre-
tionary accruals, and restatements

Panel A: Comparison of mean discretionary accruals and mean absolute value of discretionary accruals
across strong and weak HAVEN INTENSITYa

Name N Strong HAVEN INTENSITY Weak HAVEN INTENSITY

Mean Mean

PTDACC 2,050 -0.073 0.117

|PTDACC| 2,050 7.201 6.855

Panel B: Frequency of restatement firm-years across strong and weak HAVEN INTENSITYb

Strong HAVEN INTENSITY Weak HAVEN INTENSITY

Did not restate 1,315 1,259

Restated 43 43
Total 1,358 1,302
Chi-square 0.039
Prob Chi-square = 0 0.421
a
This table shows the mean of discretionary accruals and signed discretionary accruals of firm-years with
strong HAVEN INTENSITY relative to a matched sample of firm-years with weak HAVEN INTENSITY.
Firm-years are matched based on their propensity to have strong HAVEN INTENSITY—defined as being
in the highest quintile of HAVEN INTENSITY. Control firm-years are drawn from the lowest two quintiles
of HAVEN INTENSITY. Matches are required to be in the same industry and have propensity scores
(ranging from 0 to 1) within 0.1 of each other. The number of observations varies between Panel A and
Panel B because not all firm-years have restatement data
b
This table shows the frequency of restatement of firm-years with strong HAVEN INTENSITY relative to
a matched sample of firm-years with weak HAVEN INTENSITY. Firm-years are matched based on their
propensity to have strong HAVEN INTENSITY—defined as being in the highest quintile of HAVEN
INTENSITY. Control firm-years are drawn from the lowest two quintiles of HAVEN INTENSITY. Matches
are required to be in the same industry and have propensity scores (ranging from 0 to 1) within 0.1 of each
other. The number of observations varies between Panel A and Panel B because not all firm-years have
restatement data

would predict when examining |PTDACC|. Likewise, we find no evidence that firms
with greater tax haven intensity are more likely to restate financial statements.
Again, this is consistent with the logistic regression results in Table 9. Overall, the
tests suggest that having subsidiaries in weak rule of law countries is associated with
more earnings management and a higher likelihood of restatement, whereas having
subsidiaries in tax havens does not appear to be associated with statistically higher
mean values of PTDACC and |PTDACC|, nor does it appear to be associated with
the likelihood of restatement.

6 Conclusions

We examine the location of earnings management across domestic and foreign

income for a sample of 2,067 U.S. multinational firms (11,077 firm-years) over the

123
684 S. D. Dyreng et al.

years 1994–2009. We report three main findings. First, we find that, on average,
domestic income tends to be managed more than foreign income. Second, we
predict and find that firms with extensive foreign subsidiaries in countries with a low
rule of law engage in more earnings management than other firms and that the
earnings management is concentrated in foreign income. Third, we find that
profitable firms with a relatively high proportion of subsidiaries in tax havens
manage earnings more than other firms, as measured by discretionary accruals, and
that the incremental effect is also concentrated in foreign income. Together, these
results provide initial evidence about where firms manage earnings.
All studies are subject to caveats, and ours is no exception. First, there is
disagreement in the literature about how best to measure earnings management. We
employ three measures of earnings management and obtain generally similar
results. Second, some of the tests assume a linear relation between earnings and
discretionary accruals, which is a simplification. Third, in an ideal experiment, we
could assign firms randomly to do business in different locals, with varying levels of
rule of law and taxation, and then observe how their earnings management changed
with the footprint of their operations around the world. Like most researchers using
archival data, we do not have the luxury of random assignment, and thus we have
the limitations that come with observing data as they naturally occur, including
endogeneity and self-selection concerns. We attempt to control for these issues,
introducing a number of controls for both firm-specific and macro-factors, as well as
conducting a propensity matching analysis. However, to the extent these controls
are not sufficient, readers should interpret the results with reasonable caution.
This paper contributes to the long line of research on earnings management by
providing initial evidence on where firms manage earnings, a question that is for the
most part unexplored. In addition, our study is an early step in the literature focused
on looking within multinational firms rather than only across firms. We look forward
to further inquiries in this spirit in the future.

Acknowledgments We appreciate helpful comments from Patricia Dechow (editor), Annalisa Prencipe
(discussant), two anonymous referees, Dirk Black, Alex Edwards, Jürgen Ernstberger, Jeff Hoopes, Chad
Larson, Alina Lerman, K. Ramesh, Tjomme Rusticus, Terry Shevlin, Nemit Shroff, Shyam Sunder, Jake
Thomas, Jake Thornock, Alex Young, Frank Zhang, and workshop participants at the 2011 European
Accounting Association Annual Congress, the 2011 London Business School Accounting Symposium,
the 2011 Review of Accounting Studies Conference, Florida State University, University of Chicago,
University of Notre Dame, University of Southern California, University of Toronto, Texas A&M
University, and the Yale 2010 Accounting Research Conference. Maydew acknowledges financial
support from the Arthur Andersen Faculty Fund.

References

Aiken, L., & West, S. (1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks,
CA: Sage.
Armstrong, C., Jagolinzer, A., & Larcker, D. (2010). Chief executive officer equity incentives and
accounting irregularities. Journal of Accounting Research, 48, 225–271.
Badertscher, B., Phillips, J., Pincus, M., & Rego, S. (2009). Earnings management strategies and the
trade-off between tax benefits and detection risk: To conform or not to conform? The Accounting
Review, 84, 63–97.

123
Where do firms manage earnings? 685

Ball, R., & Shivakumar, L. (2006). The role of accruals in asymmetrically timely gain and loss
recognition. Journal of Accounting Research, 44, 207–242.
Beasley, M. S., Dana, H. H., & Terry, L. N. (2010). Fraudulent financial reporting: An analysis of
U.S. public companies, 1998–2007. Committee of Sponsoring Organizations of the Treadway
Commission (COSO).
Berger, P., & Hann, R. (2003). The impact of SFAS No. 131 on information and monitoring. Journal of
Accounting Research, 41, 163–197.
Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. Journal of Financial
Economics, 80, 511–529.
Beuselinck, C., Deloof, M., & Vanstraelen, A. (2010). Earnings management contagion in multinational
corporations. Tilburg University and Antwerp University working paper.
Bodnar, G., & Weintrop, J. (1997). The valuation of the foreign income of US multinational firms:
A growth opportunities perspective. Journal of Accounting and Economics, 24, 69–97.
Bowen, R., Noreen, E., & Lacey, J. M. (1981). Determinants of the corporate decision to capitalize
interest. Journal of Accounting and Economics, 3, 151–179.
Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and losses.
Journal of Accounting and Economics, 24, 99–126.
Chung, H., & Kallapur, S. (2003). Client importance, nonaudit services, and abnormal accruals. The
Accounting Review, 78, 931–955.
Dechow, P., & Dichev, I. (2002). The quality of accruals and earnings: the role of accrual estimation
errors. The Accounting Review, 77, 35–59.
Dechow, P., Ge, W., Larson, C., & Sloan, R. (2011). Predicting material accounting restatements.
Contemporary Accounting Research, 28, 17–82.
Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of their proxies,
their determinants and their consequences. Journal of Accounting and Economics, 50, 344–401.
Dechow, P., & Skinner, D. J. (2000). Earnings management: Reconciling the views of accounting
academics, practitioners, and regulators. Accounting Horizons, 14, 235–250.
Dechow, P., Sloan, R., & Sweeney, A. (1995). Detecting earnings management. The Accounting Review,
70, 193–225.
DeFond, M., Hung, M., & Trezevant, R. (2007). Investor protection and the information content of annual
earnings announcements: International evidence. Journal of Accounting and Economics, 43, 37–67.
DeFond, M., & Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals. Journal of
Accounting Research, 17, 145–176.
Desai, M., & Dharmapala, D. (2006). Corporate tax avoidance and high-powered incentives. Journal of
Financial Economics, 79, 145–179.
Desai, M., Dyck, A., & Zingales, L. (2007). Theft and taxes. Journal of Financial Economics, 84,
591–623.
Dichev, I., & Skinner, D. J. (2002). Large-sample evidence on the debt covenant hypothesis. Journal of
Accounting Research, 40, 1091–1123.
Doyle, J., Ge, W., & McVay, S. (2007). Accruals quality and internal control over financial reporting. The
Accounting Review, 82, 1141–1170.
Durnev, A., Li, T., & Magnan, M. (2011). Tax avoidance, institutional environment, and financial
reporting: evidence from offshore companies. McGill University, University of Windsor, and
Concordia University working paper.
Duru, A., & Reeb, D. (2002). International diversification and analysts’ forecast accuracy and bias. The
Accounting Review, 77, 415–433.
Dyreng, S., & Lindsey, B. (2009). Using financial accounting data to examine the effect of foreign
operations located in tax havens and other countries on U.S. multinational firms’ tax rates. Journal
of Accounting Research, 47, 1283–1316.
Dyreng, S., Mayew, W., & Willams, C. (2010). Religious social norms and corporate financial reporting.
Duke University and University of Michigan working paper.
Erickson, M., Hanlon, M., & Maydew, E. (2004). How much will firms pay for earnings that do not exist?
Evidence of taxes paid on allegedly fraudulent earnings. The Accounting Review, 79, 387–408.
Fan, N. (2008). A study of foreign earnings management using an empirical distribution approach.
Unpublished dissertation, University of Texas at Arlington.
Fields, T., Lys, T., & Vincent, L. (2001). Empirical research on accounting choice. Journal of Accounting
and Economics, 31, 255–307.

123
686 S. D. Dyreng et al.

Francis, J., LaFond, R., Olsson, P., & Schipper, K. (2005). The market pricing of accruals quality. Journal
of Accounting and Economics, 39, 295–327.
Frank, M., Lynch, L., & Rego, S. (2009). Tax reporting aggressiveness and its relation to aggressive
financial reporting. The Accounting Review, 84, 467–496.
Frankel, R., Johnson, M., & Nelson, K. (2002). The relation between auditors’ fees for nonaudit services
and earnings management. The Accounting Review, 77, 71–105.
Goelzer, D. (2009). Seven years of the public company accounting oversight board: What has been
accomplished and what remains to be done? Speech before the AICPA conference, Washington,
DC, December 7, 2009.
Golden, T., Skalak, S., & Clayton, M. (2006). A guide to forensic accounting. New York: Wiley.
Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices. Quarterly Journal
of Economics, 118, 107–155.
Grullon, G., Kanatas, G., & Weston, J. (2010). Religion and corporate (mis)behavior. Rice University
working paper.
Hagerman, R., & Zmijewski, M. (1979). Some economic determinants of accounting policy choice.
Journal of Accounting and Economics, 1, 141–161.
Healy, P. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and
Economics, 7, 85–107.
Healy, P., & Wahlen, J. (1999). A review of the earnings management literature and its implications for
standard setting. Accounting Horizons, 13, 365–383.
Hennes, K., Leone, A., & Miller, B. (2008). The importance of distinguishing errors from irregularities in
restatement research: The case of restatements and CEO/CFO turnover. The Accounting Review, 83,
1487–1520.
Holthausen, R. W. (1981). Evidence on the effect of bond covenants and management compensation
contracts on the choice of accounting techniques: The case of the depreciation switch-back. Journal
of Accounting and Economics, 3, 73–109.
Hope, O.-K., Kang, T., Thomas, W., & Vasvari, F. (2008). Pricing and mispricing effects of SFAS No.
131. Journal of Business, Finance and Accounting, 35, 281–306.
Hribar, P., & Nichols, D. (2007). The use of unsigned earnings quality measures in tests of earnings
management. Journal of Accounting Research, 45, 1017–1053.
Jones, J. (1991). Earnings management during import relief investigations. Journal of Accounting
Research, 29, 193–228.
Kedia, S., & Rajgopal, S. (2011). Do the SEC’s enforcement preferences affect corporate misconduct?
Journal of Accounting and Economics, 51, 259–273.
Klein, A. (2002). Audit committee, board of director characteristics, and earnings management. Journal
of Accounting and Economics, 33, 375–400.
Kothari, S., Leone, A., & Wasley, C. (2005). Performance matched discretionary accrual measures.
Journal of Accounting and Economics, 39, 163–197.
Leuz, C., Nanda, D., & Wysocki, P. (2003). Earnings management and investor protection: An
international comparison. Journal of Financial Economics, 69, 505–527.
McGuire, S., Omer, T., & Sharp, N. (2012). The impact of religion on financial reporting irregularities.
The Accounting Review, 87, 645–673.
Miedema, D. (March 4, 2008). FACTBOX—tax havens of the world. Reuters.
Mishkin, F. (1983). A rational expectations approach to macroeconometrics: Testing policy effectiveness
and efficient markets models. Chicago, IL: University of Chicago Press for the National Bureau of
Economic Research.
Myers, J., Myers, L., & Omer, T. (2003). Exploring the term of the auditor–client relationship and the
quality of earnings: A case for mandatory auditor rotation? The Accounting Review, 78, 779–799.
Petersen, M. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review
of Financial Studies, 22, 435–480.
Pincus, M., Rajgopal, S., & Venkatachalam, M. (2007). The accrual anomaly: International evidence. The
Accounting Review, 82, 169–203.
Rego, S. (2003). Tax-avoidance activities of U.S. multinational companies. Contemporary Accounting
Research, 20, 805–833.
Schipper, K. (1989). Commentary: Earnings management. Accounting Horizons, 3, 91–102.
Securities and Exchange Commission, Division of Enforcement. (2011). Enforcement manual. Office of
Chief Counsel, Washington, DC, August 2.

123
Where do firms manage earnings? 687

Srinivasan, S., Wahid, A., & Yu, G. (2011). Admitting mistakes: An analysis of restatements by foreign
firms listed in the U.S. Working paper, Harvard Business School.
Sweeney, A. (1994). Debt-covenant violations and managers’ accounting responses. Journal of
Accounting and Economics, 17, 281–308.
Thomas, W. (1999). A test of the market’s mispricing of domestic and foreign earnings. Journal of
Accounting and Economics, 28, 243–268.
Watts, R., & Zimmerman, J. (1978). Towards a positive theory of the determination of accounting
standards. The Accounting Review, 53, 112–134.

123