Accepted Manuscript: Journal of International Accounting, Auditing and Taxation
Accepted Manuscript: Journal of International Accounting, Auditing and Taxation
Accepted Manuscript: Journal of International Accounting, Auditing and Taxation
PII: S1061-9518(18)30211-8
DOI: https://doi.org/10.1016/j.intaccaudtax.2018.09.001
Reference: ACCAUD 250
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
A Reinvestigation into Accounting Quality Following Global IFRS
Adoption: Evidence via Earnings Distributions
May 2018
Madeline Trimble
PT
Illinois State University
RI
Corresponding author:
SC
Telephone: +1 309 438 5184
Email address: mktrimb@ilstu.edu
Postal address: Campus Mailbox 5520
State Farm Hall of Business
U
Illinois State University
Normal, IL, USA 61790-5520
N
A
Abstract
M
Despite over 140 countries adopting IFRS in some form, its effect on
accounting quality remains unclear. Potential explanations for divergent
findings in the literature include the focus on a narrow sample of EU
D
constant sample of firms. Results are consistent for both EU and non-EU
countries and are more pronounced for countries with high enforcement
and where users' demand for high quality reporting is high. Furthermore,
I investigate the level of discretionary accruals and real earnings
CC
its effectiveness in improving accounting quality1 has not yet been reached. Even studies focusing on the
2005 IFRS adoption in the European Union (EU)2 that use the same measure of accounting quality find
both confirming (Chen et al., 2010; Zeghal et al., 2011, 2012) and opposing results (Jeanjean & Stolowy,
PT
2008; Callao & Jarne, 2010; Capkun et al., 2012; Ahmed et al., 2013a,b). These divergent findings could
RI
result from (1) mismeasurement and power concerns of using discretionary accruals, especially in a
global setting (Dechow et al., 2003; Wysocki, 2004), (2) lack of post-adoption observations in early
SC
studies, (3) noise around the 2005 adoption event, including changes in enforcement regulation
(Christensen et al., 2013), or (4) the narrow focus on developed, mostly Western European countries,
U
which does not represent the IASB's current reach.
N
To address these points, I use an alternative, more relative accounting quality measure that works to
A
capture multiple dimensions of manager manipulation: the discontinuities of earnings distributions at the
M
zero earnings threshold. I use a sample of 46 countries that adopted IFRS at different times. I plot and
compare scaled earnings distributions for both a total and a constant sample of firms in the pre- and post-
D
IFRS adoption periods. Findings reveal that, while there are still significantly greater-than-expected
TE
observations in the interval directly above zero and significantly fewer-than-expected observations in the
interval directly below zero, the size of the discontinuity decreases. Additionally, the ratio of small profit
EP
to small loss firms also decreases for a total and constant sample of firms. These results are robust to
CC
different earnings and scaling measures, addressing methodological concerns raised by Beaver et al.
(2007) and Durtschi & Easton (2005, 2009). Using this alternative accounting quality measure, these
A
findings suggest that mandatory IFRS adoption achieves its first-order objective of improving the quality
1
The use of the terms accounting quality, reporting quality, and financial reporting quality are used interchangeably
throughout this work.
2
EU Regulation 1606/2002 required the consolidated accounts of listed companies in regulated EU markets to
adhere to IFRS as issued by the IASB for the reporting period ending on or after December 31, 2005.
2
To investigate differences in reporting environments and assess the validity of the distribution
measure, I conduct a series of cross-sectional analyses. I find a similar increase in accounting quality for
both EU and non-EU samples, although the increase is larger for the latter group. This indicates that the
benefits of IFRS are realized in developing countries and that mandated IFRS adoption, not concurrent
changes to enforcement, was the catalyst for improved accounting quality in the EU. Furthermore, I
PT
revisit the role of reporting environment and IFRS adoption and find that the discontinuity decreases more
RI
in countries with high regulatory enforcement and completely disappears for common law countries.
These findings should aid the IASB in identifying reporting environments that are conductive for
SC
successful future IFRS projects.
In a next step, I investigate the levels of signed accrual and real earnings management for firms
U
around the discontinuity at the zero earnings threshold. This analysis serves two functions: (1) it isolates
N
what is being captured in the earnings distributions, thus, validating the methodology and (2) it
A
investigates the relation between accrual and real earnings management conditional on the effects of IFRS
M
adoption. Similar to Dechow et al. (2003), I do not find a clear pattern for the role of signed accrual and
real earnings management in this setting. However, empirical results indicate that benchmark beating
D
firms generally are more likely to have higher levels of real earnings management. The effect of IFRS
TE
adoption, although negative, does not significantly deter this discretionary behavior.
In a further analysis, I plot operating income (EBIT) pre- and post-IFRS adoption and find that the
EP
discontinuity is insignificant following IFRS adoption whereas the discontinuity persists with bottom-line
income. Importantly, this indicates that IFRS improves the quality of core earnings and managers rely on
CC
continuing operation items (e.g. special items) to meet their reporting goals following IFRS adoption.
This paper contributes to the extant literature in a number of ways. First, this study adds to the debate
A
3
accounting quality that expands on the earlier work of Jeanjean & Stolowy (2008)3. A major advantage
of this study is the use of a worldwide sample with differing IFRS adoption dates. By including rolling
adoption dates from 2005 to 2012, this study alleviates the confounding events around the isolated 2005
adoption date and mitigates the effect of the global financial crisis. Additionally, the political push to
spread IFRS to all countries, no matter the economic size, indicates how relevant IFRS studies focusing
PT
outside of Western Europe are to the next wave of research in this area.
RI
This study also contributes to a stream of literature investigating the interplay between different
earnings management methods conditional on changes to the reporting environment. While studies that
SC
investigate the relationship following the implementation of Sarbanes-Oxley (SOX) in the U.S. find a
substitutive relationship (Cohen et al. 2008; Bartov & Cohen 2009), studies that investigate IFRS
U
adoption in Europe report mixed results (Doukakis 2014). This study expands on this area of research by
N
investigating the interplay between accrual and real earnings management with respect to benchmark
A
beating in a global IFRS adoption setting.
M
Finally, this work extends the emerging literature that uses earnings distributions to gauge the
effectiveness of accounting policy changes. By holding country, industry, and firm characteristics
D
constant, earnings distributions are an ideal methodology to isolate and evaluate changes in firms'
TE
reporting behavior. This method has been used to study national policies like the U.S.’ SOX and Japan's
Financial Instruments and Exchange Law (J-SOX) (Gilliam et al. 2015; Enomoto & Yamaguchi 2016), as
EP
well as industry-specific regulations such as the Federal Deposit Insurance Corporation Improvement Act
(FDICIA) in banking (Altamuro & Beatty, 2010). This study contributes to this literature by investigating
CC
the change in accounting quality following a mandatory switch from local GAAP to IFRS.
The remainder of this paper proceeds as follows: Section 2 highlights the relevant existing literature
A
on accounting quality following IFRS adoption and earnings distributions, and outlines the hypotheses.
3
The early IFRS study by Jeanjean & Stolowy (2008) investigates the effect of IFRS on benchmark beating using
earnings distributions for a small sample of three countries (United Kingdom, Australia, and France) and a limited
time horizon (2004-2006).
4
Section 3 describes the data collection and cleaning process, the IFRS adoption identification strategy,
and the accounting quality measures used. The results of the earnings distributions and empirical analyses
are discussed in Section 4. Section 5 provides robustness analyses and Section 6 concludes.
PT
2.1 Accounting Quality following Mandatory IFRS Adoption
The most common setting for investigating mandatory IFRS adoption is the 2005 adoption of IFRS in
RI
the European Union, Australia, South Africa, and Hong Kong4,5. Using this exogenous shock, Chen et al.
SC
(2010), Zeghal et al. (2011), and Zeghal et al. (2012) generally find an increase in accounting quality
through higher quality accruals, timelier reporting, and lower likelihood of artificially meeting reporting
U
thresholds. It is important to note, however, that their findings are dependent on characteristics of the
N
reporting environment and that the high costs of IFRS adoption leading to noncompliance are often not
A
fully incorporated (Larson & Street, 2004; Jermakowicz & Gornik-Tomaszewski, 2006; Haller &
M
Wehrfritz, 2013).
Using the same 2005 IFRS adoption sample, Ahmed et al. (2013a), Callao & Jarne (2010), and
D
Capkun et al. (2012) document that earnings smoothness, benchmark beating activities, and the level of
TE
discretionary accruals actually increase following IFRS adoption. Jeanjean & Stolowy (2008), the only
IFRS study to my knowledge that uses the earnings distribution methodology, finds no change in the
EP
frequency of benchmark beating in U.K. and Australian firms and finds an increase in benchmark beating
for French firms. Taken together, it remains unclear what effect IFRS had on accounting quality in the
CC
2005 setting.
From a global perspective, Houqe et al. (2012) and Cai et al. (2014) investigate the effect of
A
mandatory IFRS adoption on the level of discretionary accruals in over 30 countries and find that
4
See Soderstrom & Sun (2007) for an excellent and extensive summary of the literature regarding accounting
quality and IFRS adoption in developing countries.
5
Voluntary adoption of IFRS has an inherent self-selection bias regarding firms' own incentives to adopt IFRS
either legitimately or merely as a label (Barth et al. 2008; Daske et al., 2013; Christensen et al., 2015). To
circumvent this endogeneity issue, I explicitly focus on the mandatory IFRS adoption setting.
5
accounting quality as measured by discretionary accruals improves following IFRS adoption, but only in
environments with strong investor protection and enforcement, respectively. A meta-analysis on the
existing IFRS literature finds no substantive change to discretionary accruals, even when taking country-
specific characteristics, such as legal origin and enforcement, into account (Ahmed et al. 2013b)6.
There are a number of potential explanations for the divergence of findings in the literature. First,
PT
since accounting quality cannot be directly observed and collected from commercial databases, there
RI
remains a debate on what exactly constitutes accounting quality and how it should be measured. The
popular residual discretionary accrual models have been criticized for their measurement errors and low
SC
power settings, which limit their ability to capture subtle examples of earnings management (Dechow et
al., 1995; Bernard & Skinner, 1996; McNichols, 2000; Dechow et al., 2003; Burgstahler & Chuk, 2015).
U
Additionally, discretionary accrual models may not be appropriate in the cross-country setting because of
N
data restrictions (Wysocki 2004). This paper uses earnings distributions to circumvent the measurement
A
issues inherent in a global setting.
M
Second, many early IFRS studies only extend a few years before and after the 2005 EU adoption date.
As Zeff (2007) and Barth (2015) note, the application of IFRS, for most countries, has a learning curve
D
involving understanding new accounting procedures, training staff, and updating accounting information
TE
systems, all of which take several years to implement correctly. Therefore, studies that only extend a few
years after the adoption year may not be capturing the true IFRS effect, but merely a transition effect.
EP
This study uses a nearly 20 year time horizon to address this issue.
CC
Third, there were several concurrent changes to reporting infrastructure regarding enforcement
around the IFRS adoption date in the EU in 2005, which makes disentangling the IFRS effect exceedingly
difficult (Christensen et al., 2013). The use of rolling adoption dates across countries in this work
A
alleviates this clustering effect and provides a cleaner setting to measure the effect of IFRS adoption on
reporting quality.
6
This finding casts doubt on the role that reporting environment plays in IFRS adoption, which is addressed in
Section 4.2.
6
Fourth, while the majority of existing IFRS studies focus on developed countries, the IASB’s current
mission is to spread IFRS to the developing countries in the hopes of stabilizing their economies and
requires the adoption of IFRS with sanctions for non-compliance (Lamoreaux et al., 2015). Additionally,
many subsidiaries of multinational corporations, who are required to report using IFRS, are located in
PT
these emerging markets (Ezzamel & Xiao, 2011). As such, IFRS plays a significant role in developing
RI
countries, yet the effect of IFRS on reporting quality in this setting is still relatively unknown. This paper
investigates 46 countries that vary in terms of development, adoption date, and adoption style.
SC
2.2 Earnings Distributions as an Accounting Quality Measure
Using distributions of reporting items around certain thresholds has long been used as a tool for
U
evaluating accounting quality. Seminal work by Burgstahler & Dichev (1997) provides large sample
N
evidence of benchmark beating activities related to positive earnings, while other studies focus on
A
alternative reporting benchmarks such as meeting and beating analysts' forecasts (Degeorge et al., 1999)8.
M
Despite its popularity, this methodology has been criticized by researchers for some supposedly
haphazard research design choices. From a methodology perspective, Durtschi & Easton (2005, 2009)
D
question the validity of many prominent distribution studies stating that mechanical issues could
TE
potentially explain their findings including scaling choices, sample selection choices, or systematic
differences between firms that fall immediately above and below the thresholds.
EP
Nevertheless, a number of studies have solely focused on establishing manager manipulation as the
CC
cause of the discontinuity in earnings distributions. Burgstahler & Chuk (2015) respond to these critiques
arguing that low power settings were selected when making these claims, which exacerbated the supposed
A
issues. The authors use unscaled earnings split into size quantiles and find that the discontinuity is
consistent in each subsample, lessening the scaling concerns. Jacob & Jorgensen (2007) investigate
7
An expansive report of the extent of the IASB's coverage is provided in the following link:
http://www.ifrs.org/Features/Pages/Global-reach-of-IFRS-is-expanding.aspx.
8
The analysts' estimate benchmark has an inherent issue of disentangling expectation versus earnings management.
Disentangling these effects is outside the scope of this work, thus the focus on the bottom-line earnings benchmark.
7
quarterly versus annual earnings and confirm that the discontinuity is only prevalent in the final quarter
where the incentive to manage earnings is highest. Finally, Donelson et al. (2013) identify firms that have
been prosecuted by the SEC for fraudulent reporting and were forced to issue restatements. They find that
the discontinuity disappears in the restatements, which suggests that discretionary behavior by managers
PT
Additionally, there is an emerging trend of using earnings distributions to assess the effectiveness of
RI
changes in accounting regulation. Gilliam et al. (2015) investigate accounting quality following the 2002
adoption of SOX and find that the discontinuity in earnings disappears following the regulatory change.
SC
A similar analysis regarding Japan's J-SOX by Enomoto & Yamaguchi (2016) finds that the discontinuity
of earnings changes decreases following the regulation change. Additionally, Altamuro & Beatty (2010)
U
investigate changes to internal controls for the banking industry after the passing of FDICIA, noting a
N
decrease in the occurrence of benchmark beating observations. Following this trend, my study assesses
A
the effectiveness of global IFRS adoption using the same measure.
M
The lack of consensus regarding changes in accounting quality following mandatory IFRS adoption
D
and applicability issues of the traditional accounting quality measures in international settings motivate
TE
the use of earnings distributions to investigate the effect that global IFRS adoption has on accounting
Hypothesis 1: The global mandatory adoption of IFRS is associated with a decrease in the
discontinuity of earnings distributions around the zero earnings benchmark.
CC
The second hypothesis tests whether the discontinuities in earnings distributions truly captures
managers' manipulation of earnings. I investigate the levels of signed accrual earnings management
A
(AEM) and real earnings management (REM) in the interval directly above the zero earnings threshold,
relative to the interval directly below the zero earnings threshold9. If the earnings distributions are
9
It is interesting to note that the IFRS literature has majorly ignored the role of real earnings management; this is
surprising given the assumed relationship between IFRS adoption and discretionary accruals and the latter’s
supplementary relationship with real earnings management (Zang, 2012).
8
capturing discretionary reporting activities, I would expect to observe higher levels of earnings
management activities above the profit benchmark. This leads to the second hypothesis:
Hypothesis 2: The observations above the zero earnings benchmark have higher observed levels
of earnings management activities relative to other firms.
There are a number of studies that attempt to establish a substitutive and/or complementary
PT
relationship between AEM and REM (Roychowdhury, 2006; Zang, 2012). However, it is not intuitively
clear how this relationship is affected when there is a change in the reporting environment. Cohen et al.
RI
(2008) and Bartov & Cohen (2009) investigate the effect that SOX had on earnings management
SC
activities, finding that the enhanced scrutiny decreased accrual earnings management, yet increased real
earnings management. Doukakis (2014) explores this potential substitution effect following IFRS
U
adoption in European countries and finds mixed results. Therefore, I keep the global investigation
exploratory. N
A
3. Research Design
M
Due to varying dates and degrees of mandated IFRS adoption across the globe, I am careful to
TE
accurately identify timely, mandatory adopting firms to include in the analyses. My procedure is outlined
in Table 1. From the complete Thomson Reuters Datastream universe, I include all countries that have
EP
made a meaningful attempt to adopt IFRS. To identify country-level IFRS adoption efforts, I cross-
reference a variety of established sources including: the IASB jurisdiction profiles10, Deloitte's IAS Plus
CC
“Use of IFRS by Jurisdiction” webpage11, and PwC's “IFRS Adoption by Country” survey12,13. I limit the
A
10
https://www.ifrs.org/use-around-the-world/use-of-ifrs-standards-by-jurisdiction/.
11
http://www.iasplus.com/en/resources/use-of-ifrs.
12
https://www.pwc.com/gx/en/services/audit-assurance/ifrs-reporting/publications.html.
13
To determine the most accurate adoption date per country, I use the date most often cited among the three sources.
In situations where the cross-referenced sources diverge without consensus, the IASB profile date is used.
9
sample to firms in countries that converge with IFRS, require IFRS for subsets of listed firms, or require
IFRS for all listed firms14. Countries that merely permit IFRS are excluded.
I identify a firm's individual adoption date using the Datastream item “Accounting Standards
Followed” (WC07536), as outlined in Appendix A1 of Daske et al. (2013). Those firms with no
identifiable adoption or those that switch before or after the country's official adoption date are excluded.
PT
As a result, the final sample includes only firms that adopt IFRS in the correct year mandated by their
RI
country and have observations in both the pre- and post-IFRS sample, keeping the two samples constant.
SC
To clean the data for analysis, I begin with all firms from IFRS adopting countries covered in
Thomson Reuters Datastream that have net income, total assets, market value of equity, and sales data
U
available in a given year. Any firm that identifies as reporting under U.S. GAAP is excluded. Because the
N
study focuses on industrial firms with similar capital structures and oversight, financial institutions are
A
omitted (i.e., Datastream industry codes (WC06011) between 4300-4600). Next, I only include firms that
M
fit the IFRS identification strategy outlined above. Finally, I exclude the adoption year due to the use of
Table 3 shows that the final sample of IFRS adopting firms distributed across 46 countries is fairly
aligned with other IFRS studies using the Thomson Reuters Datastream universe (e.g., Chen et al. 2010
EP
and Zeghal et al. 2012). To give a more global perspective, the countries included are a mix of developed
and developing economies, with developing countries comprising 36% of the sample15. The benefit of
CC
this sample is that it provides a picture of IFRS adoption spanning from 2005 until the latest wave of
adoptions in 2012 by Argentina, Malaysia, Mexico, Nigeria, Peru, Russia, Sri Lanka, and the Ukraine
A
14
As Zeff (2007) notes, countries that converge with IFRS oftentimes mislabel IFRS as “national standards” in their
annual reports. Countries with IFRS convergence projects show staggered adoptions at the firm level, thus, I choose
to keep firms that adopt IFRS after the convergence date in order to preserve some crucial markets (e.g., China,
Malaysia, Philippines). However, the results remain constant if I omit these firms.
15
Although there is a relatively large portion of Canadian and South Korean firms with relatively recent adoption
dates, omitting them from the sample does not drastically change the shape of the distributions.
10
(34% of sample observations). Finally, I include a range of adoption styles including five convergence
countries and six limited requirement countries, jointly resulting in a third (33%) of the total firm-year
observations. Compared to the existing IFRS literature regarding financial reporting quality, this paper
has substantial global coverage and provides a more dynamic picture of IFRS adoption.
PT
3.2 Measures of Accounting Quality
RI
3.2.1 Earnings Distribution Metrics
To investigate the change in accounting quality, I first plot the distribution of scaled earnings. The
SC
necessary assumption for using earnings distributions as an accounting quality measure is that, in the
aggregate, earnings follow a smooth distribution. However, if low accounting quality is present, one
U
would expect to see a discontinuity, or “kink”, in the distribution around the important zero earnings
threshold16.
N
A
It is imperative in this setting to investigate the following research choices: the numerator (earnings
M
measure), the denominator (size scaler), and the selected bin width. In this setting, the measure of
earnings is net income (NI) and, following prior literature, the deflator is beginning market value of
D
equity (MV). The appropriate bin width is calculated using the Freedman-Diaconis equation:
TE
1⁄ )
2(𝐼𝑄𝑅)𝑁 (− 3 where 𝐼𝑄𝑅 is the interquartile range of scaled earnings and 𝑁 is the number of plotted
observations. The optimal bin size for the total sample is 0.005 where small loss firms are in the interval
EP
range (-0.005, 0] and small profit firms are in the interval range [0, 0.005)17. The distribution plot is
CC
In order to determine the significance of divergence from the assumed smooth distribution of
A
earnings, I use the standardized difference statistic calculated for the interval immediately to the left and
16
By selecting a specific setting with an ex ante expectation in benchmark beating behaviors, such as the mandatory
adoption of IFRS, the strength of the earnings distribution test increases (Dechow & Skinner, 2000).
17
To assure that using too small of bin size does not artificially create the distribution (Burgstahler & Chuk, 2015), I
also use a more conservative bin size of 0.01 and find fundamentally the same results.
11
right of zero earnings (Burgstahler & Dichev 1997). The test statistic is the difference between actual and
expected observations in the interval divided by the variance of the difference between actual and
expected observations18. Furthermore, I calculate the ratio of the number of small profit firms in the
interval above zero to the number of small loss firms in the interval below zero as a measure of the
severity of the discontinuity following Leuz et al. 2003. A higher value indicates a larger discontinuity.
PT
In a second step, I use a logit regression-based approach to identify the probability of a firm engaging
RI
in low quality reporting. By including IFRS as an explanatory variable, I can assess the likelihood of a
firm reporting in the benchmark beating interval in the post-IFRS adoption period relative to the pre-IFRS
SC
adoption period. The model is as follows:
U
+ 𝛽6 𝑆𝑎𝑙𝑒𝑠 𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡 + 𝛽7 𝑂𝐶𝐹𝑖𝑡 + 𝛽8 𝐷𝑒𝑏𝑡 𝐼𝑠𝑠𝑢𝑒𝑖𝑡 + 𝛽9 𝐸𝑞𝑢𝑖𝑡𝑦 𝐼𝑠𝑠𝑢𝑒𝑖𝑡 (1)
N
+ 𝛽10 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟𝑖𝑡 + 𝛽11 𝐶𝑙𝑜𝑠𝑒𝑙𝑦 𝐻𝑒𝑙𝑑𝑖𝑡 + 𝛽12 𝐵𝑖𝑔4𝑖 + 𝛽13 𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝐶𝑜𝑢𝑛𝑡𝑖
A
+ 𝛽14 𝑈. 𝑆. 𝐶𝑟𝑜𝑠𝑠 − 𝑙𝑖𝑠𝑡𝑒𝑑𝑖 + 𝑓𝑖𝑥𝑒𝑑 𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜖𝑖𝑡
M
The dependent variable Profit is an indicator variable that equals one if firms report scaled earnings
D
between 0 and 0.005, which are those firms that are identified as “benchmark beaters”. Following Barth
TE
et al. (2008), Chen et al. (2010), and Zeghal et al. (2012), I would expect to see a significant negative
coefficient for 𝛽1 , indicating that, ceteris paribus, the propensity of firms to engage in benchmark beating
EP
activities decreases following mandatory IFRS adoption in cases accompanying low levels of managerial
discretion19. 𝐸𝑀 is a placeholder variable for the aggregate discretionary accrual (Agg. AEM) or real
CC
earnings management measures (Agg. REM). I do not make a prediction on the interplay of the two
measures of accounting quality in relation to IFRS. However, I would expect that the interaction effect (
A
𝛽3 ) would be significantly negative indicating that firms with more discretionary reporting practices
18
The expected frequency of observations is calculated as the average of the two adjacent intervals. The variance is
measured as follows: θ = Npi(1 - pi)+(1/4)N(pi + pi+1)(1 - pi-1 - pi+1) where p is the probability of an observation
falling in interval i.
19
The overall effect of IFRS adoption on decreasing benchmark beating activities is captured in 𝛽1 + 𝛽3 .
12
would, on average, have a smaller decrease in benchmark beating behaviors following mandatory IFRS
I control for classic accounting quality-related firm characteristics that might influence the probability
of a firm managing earnings including 𝑆𝑖𝑧𝑒, 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒, and 𝑆𝑎𝑙𝑒𝑠 𝐺𝑟𝑜𝑤𝑡ℎ20. I control for the
established negative relationship between operating cash flow levels (𝑂𝐶𝐹) and accruals. Changes in
PT
debt issuances (𝐷𝑒𝑏𝑡 𝐼𝑠𝑠𝑢𝑒) and equity issuances (𝐸𝑞𝑢𝑖𝑡𝑦 𝐼𝑠𝑠𝑢𝑒) are included to control for changes in
RI
debt structure. Asset turnover (𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟) is included to assess the health of the firm. To measure
insider ownership, I include the percentage of closely held shares (𝐶𝑙𝑜𝑠𝑒𝑙𝑦 𝐻𝑒𝑙𝑑). I control for quality of
SC
auditor (𝐵𝑖𝑔4), the number of exchanges on which the firm is listed (𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝐶𝑜𝑢𝑛𝑡), and whether it is
listed on a U.S. stock exchange (𝑈. 𝑆. 𝐶𝑟𝑜𝑠𝑠 − 𝑙𝑖𝑠𝑡𝑒𝑑). Finally, I attempt to control for systematic
U
differences across countries and industries by including country and industry fixed effects, respectively.
accruals using a combination of two residual regression models and one linear expectation model that
range in terms of sophistication and data requirements. The first discretionary accrual model is the linear
D
where
𝑇𝑂𝑇𝐴𝐶𝐶𝑖𝑡 =
𝑇𝐴𝑖,𝑡−1
𝑇𝑂𝑇𝐴𝐶𝐶 measures total accruals. The accounts use to measure total accruals include: 𝐶𝐴 as current assets
A
(WC02201), 𝐶𝐿 as current liabilities (WC03101), 𝐶𝐴𝑆𝐻 as total cash reserve (WC02003), 𝑆𝑇𝐷𝐸𝐵𝑇 as
short-term debt (WC03051), and 𝐷𝐸𝑃 as depreciation expense (WC01151). 𝑅𝐸𝑉 and 𝑅𝐸𝐶 represent
revenues (WC01001) and accounts receivable (WC02051), respectively. The level of property, plant, and
20
All control variable definitions are found in Appendix A.
13
equipment (WC02301) is captured in 𝑃𝑃𝐸. 𝑅𝑂𝐴, measured as net income divided by total assets, is the
return on assets. All variables except 𝑅𝑂𝐴 are scaled by lagged total assets (𝑇𝐴) to control for
heteroskedasticity. The residuals capture discretionary accruals. Regressions 2 and 3 are run over each
industry-year group with at least 15 observations and control for country-level variation by including
PT
The second accounting quality measure is another classic discretionary accrual residual regression
RI
model, the modified Dechow & Dichev (2002) model (Mod. D&D):
SC
𝑖,𝑡−1 𝑖,𝑡−1
The measure of total accruals (𝑇𝑂𝑇𝐴𝐶𝐶), revenues (𝑅𝐸𝑉), accounts receivable (𝑅𝐸𝐶), property,
U
plant, and equipment (𝑃𝑃𝐸), and total assets (𝑇𝐴) are as defined in the prior model. 𝑂𝐶𝐹 represents
N
operating cash flows calculated indirectly as 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒 − 𝑇𝑂𝑇𝐴𝐶𝐶. All variables are scaled by lagged
A
total assets. The residuals proxy for discretionary accruals.
M
In order to incorporate facets of each model, as well as attempt to mitigate measurement errors, I
constructed an aggregate accrual earnings management measure (Agg. AEM), calculated as the firm-year
D
mean of the standardized values of the prior two models (Biddle et al. 2009).
TE
The third measure is an expectation model created by DeFond & Park (2001). The linear benchmark
discretionary accrual model (Benchmark DA) uses the firm's accrual reporting behavior in the prior
EP
period as the benchmark to determine the expected current level of non-discretionary accruals. This
model eliminates country and industry difference concerns and sample size requirements by construction.
CC
The model defines discretionary accruals as the difference between total accruals (𝑇𝑂𝑇𝐴𝐶𝐶, as defined
𝐶𝐴𝐶𝐶𝑖,𝑡−1 𝐷𝐸𝑃𝑖,𝑡−1
[𝑅𝐸𝑉𝑖𝑡 𝑥 ]−[𝑃𝑃𝐸𝑖𝑡 𝑥 ]
𝑅𝐸𝑉𝑖,𝑡−1 𝑃𝑃𝐸𝑖,𝑡−1
𝑁𝐷𝐴𝐶𝐶𝑖𝑡 = 𝑇𝐴𝑖,𝑡−1
(4)
21
Annual levels of inflation (annual %) and GDP growth (annual %) are collected from the World Bank's World
Development Indicators databank at https://data.worldbank.org/indicator.
14
where the current portion of total accruals, 𝐶𝐴𝐶𝐶, is calculated as:
𝑅𝐸𝑉, 𝑃𝑃𝐸, 𝐷𝐸𝑃, 𝐶𝐴, 𝐶𝐴𝑆𝐻, 𝐶𝐿, 𝑆𝑇𝐷𝐸𝐵𝑇, and 𝑇𝐴 are as defined above. 𝑇𝑆 is the treasury stock portion
PT
The first measure of real earnings management, developed by Zang (2012), captures the change in
RI
bottom-line earnings through non-accrual-based decisions such as selling PP&E, altering production
levels, and cutting discretionary spending. REM1 is a residual model that captures the sum of abnormal
SC
discretionary expenses and abnormal production costs. The abnormal discretionary expenses and
abnormal production costs are the residuals of the following two equations:
U
𝐷𝐼𝑆𝐶 𝐸𝑋𝑃𝑖𝑡 𝑆𝑎𝑙𝑒𝑠𝑖,𝑡−1
= 𝛼0 + 𝛼1 + 𝜖𝑖𝑡 (5)
𝑇𝐴𝑖,𝑡−1 N𝑇𝐴𝑖,𝑡−1
A
𝑃𝑅𝑂𝐷𝑖𝑡 𝑆𝑎𝑙𝑒𝑠𝑖𝑡 ∆𝑆𝑎𝑙𝑒𝑠𝑖𝑡 𝑆𝑎𝑙𝑒𝑠𝑖,𝑡−1
= 𝛼0 + 𝛼1 + 𝛼2 + 𝛼3 + 𝜖𝑖𝑡 (6)
M
𝐷𝐼𝑆𝐶 𝐸𝑋𝑃 is measured as the sum of R&D expenses (WC01201) and SG&A expenses (WC01101).
D
𝑃𝑅𝑂𝐷 is measured as the sum of cost of goods sold (WC01051) and changes in inventory (WC02101).
TE
All variables are scaled by lagged total assets. The residuals of the regressions capture abnormal
EP
discretionary expenses and abnormal production costs, respectively. Regressions 5-7 are run over each
industry-year group with at least 15 observations controlling for country differences with lagged inflation
CC
and GDP growth. Abnormal discretionary expenses are multiplied by negative one so that higher values
The second real earnings management residual model (REM2) is the sum of abnormal discretionary
expenses and abnormal operating cash flows (multiplied by -1). Abnormal discretionary expenses are the
22
Due to low database coverage of 𝑇𝑆 and 𝐷𝐼𝑉, all missing variables are replaced with zero following prior
literature (Houqe et al., 2012).
15
residuals of Equation 5. Operating cash flows (𝑂𝐶𝐹) are calculated as in DeFond & Park (2001).
Abnormal operating cash flows, which are the residuals of the following regression, are indicators of
additional real earnings management activities that will affect final earnings23.
PT
To control for potential measurement error and capture different aspects of real earnings
RI
management, I aggregate the two REM variables (Agg. REM) by standardizing and taking the mean of
SC
3.3 Descriptive Statistics and Correlations
I use a sample of 46,939 firm-year observations from 46 countries from 1997 to 2013. All monetary
U
variables are standardized to the United States Dollar (USD) and continuous variables are winsorized at
N
the annual 2% and 98%24. Table 4 presents descriptive statistics of the sample split into the pre- and post-
A
IFRS adoption periods. Mean and median statistical comparisons are provided in the last two columns.
M
In terms of accounting quality proxies, signed accrual levels have somewhat mixed results with the
TE
median Mod. Jones, Mod. D&D, and Agg. AEM values all being significantly higher in the post-IFRS
adoption period. The Benchmark DA model, however, shows a median decrease in signed accrual levels
EP
from 0.004 to -0.001, significant at the 1% level. The real earnings management proxies are more
CC
unanimous. REM1, REM2, and Agg. REM all have significantly lower mean and median levels in the
post-IFRS adoption period. The absolute AEM and REM measures are more consistent than their signed
A
23
One caveat of using OCF in REM2 is that the sign of the abnormal OCF is hard to predict. As Zang (2012) notes
in her study, “price discount, channel stuffing, and overproduction all decrease cash flows from operations, while
cutting discretionary expenditures increases them” (pg. 682).
24
Winsorizing all continuous variables is especially important in this setting due to the calculations of accruals
being especially sensitive to outliers (Barth et al., 2008); however, the results remain when using a more
conservative winsorizing of 1 and 99%.
16
counterparts. For the four absolute discretionary accrual proxies, all means and medians decrease in the
Overall, these results provide preliminary evidence that IFRS adoption limits discretionary activities
by managers. This may be because of the high quality IFRS regulation or, alternatively, because of the
increased scrutiny of firms during the global financial crisis, which is present in the post-IFRS period for
PT
the majority of countries. I mitigate this latter concern by using rolling IFRS adoption dates and
RI
controlling for year fixed effects in a later empirical analysis.
SC
The Spearman rank (below the diagonal) and Pearson (above the diagonal) correlation matrix is
displayed in Table 5. The measures of AEM are all positively correlated and significant at the 1% level as
U
indicated by a star (*). The Benchmark DA measure is significantly correlated with the aggregated
N
residual models (Agg. AEM) at 0.3064. Real earnings management proxies are highly correlated at
A
0.7055. These findings add validity to the measurement of the selected accounting quality proxies.
M
In terms of the correlation of IFRS and accounting quality measures, signed accruals are positively
correlated, but in most cases not significant. However, all signed real earnings management measures are
D
negatively and mostly significantly correlated with IFRS, although weakly (e.g., a range of -0.0144 to
TE
-0.0782). These correlations provide very early evidence that mandatory IFRS adoption is associated
with higher levels of discretionary accruals, but lower levels of real earnings management.
EP
4. Results
CC
The total distribution of scaled earnings is plotted in Figure 1. The earnings distribution shows a large
discontinuity at the zero earnings threshold with a standardized difference below zero of -10.14 and above
zero of 6.59, both significant at the 1% level. When the total sample distribution is split into the pre- and
post-IFRS adoption periods per country, the discontinuity is significant in both samples. However, there
17
is a noticeable decrease in the size of the standardized difference test statistics as well as a decrease in the
To control for issues related to differences in sample composition between the pre- and post-IFRS
adoption periods, I replicate the previous test using a constant sample with equal observations from each
mandatory adopting firm in both samples. This methodology ensures that countries, industries, and static
PT
firm characteristics remain stationary, better isolating the role of IFRS adoption. Figure 2 provides a
RI
clearer comparison of firm benchmark beating behavior before and after mandated IFRS adoption.
SC
Consistent with the total sample distributions, the discontinuity, though still present, appears to have
decreased following IFRS adoption (standardized difference test statistics pre-IFRS: -6.81/4.18 and post-
U
IFRS: -6.05/3.32). Furthermore, the ratio of small profit to small loss firms substantially decreased from
material increase in accounting quality that was theorized from its inception, (2) earnings distributions are
overly simple and unable to control for potential omitted variable biases and time trends necessary to
D
identify the complex IFRS effect, or (3) earnings distributions potentially capture a substitution of real for
TE
accrual earning management (or vice versa) resulting in no noticeable change across periods. While
concern (1) is the main focus of the paper, concerns (2) and (3) are addressed in Sections 4.3 and 4.4,
EP
respectively.
CC
Given that a major contribution of this study is investigating a worldwide view of IFRS adoption, it is
A
imperative to explore whether the findings are consistent within the sample. In the first cross-sectional
25
To ensure that no specific country is driving the results, I replotted the baseline distribution analysis removing one
country at a time and ran the logit regression analysis for each country with at least 100 firm-year observations. In
this untabulated analysis, the marginal decrease in the earnings discontinuity remains consistent across the different
sample compositions.
18
analysis, I separate developed EU countries from non-EU countries and replot the scaled earnings
distributions for both the pre- and post-IFRS adoption periods using a constant sample of firms26.
Given that the earliest adoption date in the sample is 2005, which includes mainly EU countries, this
analysis can also be viewed partly as an analysis of 2005 adoption, which coincides with a number of
changes to the accounting environment (Christensen et al., 2013), versus later adoption dates.
PT
Additionally, adoption procedures and requirements are heterogeneous outside of the EU sample, thus
RI
this analysis indirectly investigates changes in accounting quality following IFRS adoption across
SC
[INSERT FIGURE 3 ABOUT HERE]
Overall, the results of Figure 3 are consistent with the original findings of Figure 2. The discontinuity
U
in the earnings distributions decreases following mandatory IFRS adoption for both EU and non-EU
N
countries, however, the decrease is larger for the non-EU countries as the ratio of small profit-to-loss
A
firms decreases from 2.87 to 1.98 for non-EU countries and from 2.10 to 2.01 for EU countries. The
M
significance of the standardized difference test statistics decrease significantly and are only partly
significant in the post adoption period for non-EU countries. These results provide evidence that IFRS
D
adoption had a stronger positive effect on accounting quality in countries outside of the EU. This suggests
TE
that reporting quality is initially lower for non-EU countries that generally have lower reporting
requirements, user demand, and regulator attention and, thus, IFRS adoption has a larger effect for non-
EP
EU countries relative to EU countries (Bae et al., 2008). Alternatively, firms in non-EU countries may
rely on IFRS adoption as a quality signal to be competitive in international markets (Judge et al., 2010).
CC
Next, I disaggregate the constant sample among characteristics classically assumed to affect
accounting quality. This analysis helps to both validate the main findings of the paper was well as test
A
whether these documented determinants hold in more global and diverse samples.
26
All analyses in Section 4.2 are run on both the full sample and constant sample of firm-years with similar results.
19
Panel A separates the sample of countries by enforcement strength27. High quality accounting
standards are ineffective without sufficient enforcement (Holthausen, 2009). Unsurprisingly, prior to the
introduction of IFRS, the discontinuity in the earnings distribution is substantially larger for firms in low
enforcement countries. Christensen et al. (2013) find that capital market benefits following IFRS
adoption in the EU are limited to countries with simultaneous, positive changes to accounting
PT
enforcement, introducing ambiguity as to which event drives the findings. I find a more notable decrease
RI
in the size of the discontinuity following IFRS adoption for the high enforcement countries relative to the
low enforcement countries as evident by a decrease in small profit-to-loss ratio from 1.92 to 1.57 versus a
SC
decrease from 2.56 to 2.27, respectively. This provides evidence that IFRS adoption increases accounting
U
Panel B disaggregates the constant sample across the legal origin of each country, i.e., common law
N
versus code law countries28, to investigate users' demand for high quality accounting. Common law
A
countries generally acquire capital from the equity markets, which demand high quality financial
M
reporting. In contrast, code law countries generally rely on private funding options that communicate
through internal channels (Ball et al., 2000). I find that common law countries have higher accounting
D
quality relative to code law countries prior to IFRS adoption as evident by the lower profit-to-loss ratio
TE
(2.53 versus 1.76). Interestingly, mandatory IFRS adoption causes the discontinuity in earnings
distributions to completely disappear for common law countries29. In contrast, code law countries have
EP
27
Enforcement is measured as a median split of an aggregate of enforcement measures for corruption, governance,
political, regulatory quality, rule of law, and voice and accountability from the World Book databank:
http://databank.worldbank.org/data/reports.aspx?source=worldwide-governance-indicators.
28
Legal origin is collected from the CIA's World Factbook: https://www.cia.gov/library/publications/the-world-
factbook/fields/2100.html.
29
Common law countries include Australia, Canada, Ghana, Hong Kong, Ireland, Israel, Malaysia, New Zealand,
Nigeria, Singapore, South Africa, Sri Lanka, and United Kingdom.
20
In an attempt to validate my chosen methodology and explore the relation between different styles of
earnings management in this setting, I correlate firms in the intervals around the profit threshold with
classic accounting quality proxies from prior literature. This is a contribution because prior research
investigating changes in accounting quality following IFRS adoption have mainly utilized discretionary
accruals models, whereas I investigate both signed discretionary accrual and real earnings management
PT
measures.
RI
[INSERT TABLE 6 ABOUT HERE]
Similar to the results of Dechow et al. (2003)30, Table 6 shows that there are limited differences
SC
between both accrual and real earnings management efforts when comparing firms in the small profit and
loss intervals. In Panel A, there is no significant evidence that small profit firms report differently than
U
small loss firms in the pre-IFRS period. As a result, there is no significant difference post-IFRS adoption.
N
Only REM1 has the expected result of significantly higher levels for small profit firms relative to small
A
loss firms in the pre adoption period (0.054 vs. -0.027, respectively, t-stat=1.98), however, this limited
M
result dissipates following IFRS adoption. Panel B and C, which compare small profit firms to all other
firms and small loss firms to all other firms, respectively, yield similarly inconclusive results31.
D
Overall, there is insufficient evidence that discretionary accrual or real earnings management
TE
activities are definitively causing the discontinuities observed in Figures 1 and 2. These inconsistencies
may be due to the discontinuities capturing activities other than earnings manipulation or because the
EP
proxies for measuring accounting quality are too weak to capture small levels of earnings manipulation,
which is a common critique of their design, especially for an international sample (Bernard & Skinner,
CC
1996; Dechow et al. 2003; Wysocki, 2004; Burgstahler & Chuk, 2015).
A
30
Dechow et al. (2003) finds that there are high levels of discretionary accruals for both small profit and small loss
firms, with no significant difference.
31
In an untabulated analysis, I investigate changes in discontinuities pre- and post-IFRS adoption using absolute
levels of AEM and REM. In terms of small profit firms and small loss firms during the pre-adoption period, there
are no significant differences across the vast majority of both AEM and REM measures. However, in the post-IFRS
adoption period, the levels of accruals for small profit firms became significantly lower compared to small loss firms
all four AEM measures. Additionally, the REM differences, although not significant, are also lower for small profit
firms in the post adoption phase.
21
4.4 Empirical Benchmark Beating Evidence
Table 7 employs a logit analysis to better isolate whether IFRS adoption affects benchmark beating
tendencies while attempting to control for firm, industry, and country differences.
For the total and constant samples (Columns 1 and 4), the IFRS variable is -0.143 and -0.094,
PT
respectively, and significant at the 5% level for the total sample, providing mild evidence that global
RI
mandatory IFRS adoption lowers the probability of firms benchmark beating. The introduction of the
level of signed discretionary accrual in Columns 2 and 5 shows that there is an inverse relationship
SC
between positive discretionary accruals and probability of benchmark beating, similar to the univariate
results in Table 6, although not statistically significant. The interaction of IFRS and AEM in the constant
U
sample yields a coefficient of -1.036 (t-stat: -2.55) indicating that firms with higher levels of AEM have a
N
smaller decrease in the likelihood of a firm engaging in threshold-beating post-IFRS adoption relative to
A
firms with lower levels of discretionary accruals32. In contrast, the Agg. REM variable in the pre-IFRS
M
period, as reported in Columns 3 and 6, shows a mild, positive relation between REM and the propensity
of firms to engage in benchmark beating accounting practices. This is in line with research that finds top
D
executives actually prefer to use REM to achieve reporting targets as opposed to AEM (Graham et al.,
TE
2005; Zang, 2012). The control variables, when significant, are mainly in line with expectations. In short,
although IFRS does seem to somewhat deter benchmark beating behavior, there is not a clear relationship
EP
5. Robustness Analyses
Concerns have been raised regarding the use of earnings distributions to adequately measure
A
accounting quality by capturing benchmark beating behavior. Critics Durtschi & Easton (2005, 2009) and
Brown & Caylor (2005), among others, caution that the variables used as the numerator and denominator
32
While the total sample has the same sign for 𝛽3 with respect to the AEM measure, it is not significant.
22
in the scaled earnings measure must be deliberately chosen to assure that the underlying distribution is not
Beaver et al. (2007) argue that there are asymmetric differences in taxes and special item reporting for
profit and loss firms, which artificially inflates the discontinuity in distributions. Their argument is that
profit firms have higher tax rates thus attempt to drive their earnings values closer to the zero threshold,
PT
whereas there is a propensity for loss firms to report large, negative special items (e.g., big bath
RI
accounting) that drive their observations further away from zero and positive earnings intervals. I
SC
[INSERT FIGURE 5 ABOUT HERE]
Interestingly, when investigating operating earnings (EBIT) as opposed to bottom-line earnings, there
U
still exists a statistically significant discontinuity around the zero earnings threshold in the pre-IFRS
N
period. However, in the post-IFRS adoption period, the discontinuity disappears and the standardized
A
differences are mostly insignificant. The small profit-to-loss ratio decreases substantially from 1.60 to
M
0.92. Therefore, it appears that the discontinuity is not artificially created and that IFRS adoption
improves operating income to a greater extent than bottom-line earnings. Furthermore, perhaps the
D
managing of tax bases and the classification shifting of non-operating activities (e.g., special items) are
TE
better explanations for the continued existence of the discontinuity in earnings in the post-IFRS adoption
period.
EP
Durtschi & Easton (2005, 2009) claim that the use of market value of equity as a deflator is
problematic due to systematic differences in market valuation of profit and loss firms and less data
CC
coverage for small loss firms, which exacerbates the discontinuity. To address the scaling concern, I
replot the distribution of earnings in the pre- and post-IFRS adoption samples using a constant sample of
A
firms that are deflated by lagged sales (Panel A) and lagged total assets (Panel B).
Similar to the findings from Figure 2 the scaling of earnings by lagged sales and lagged total assets
yield consistent results. The discontinuities decrease in size, however, the standardized differences, while
23
smaller, remain significant in the post-IFRS adoption period. Furthermore, the ratios of small profit to
loss firms decrease for both robustness samples. Hence, it does not appear that the choice of scaler affects
6. Conclusion
To help settle an existing debate in the IFRS literature, this paper reinvestigates whether the global
PT
introduction of IFRS has achieved its goal of improving financial reporting by using an alternative, more
RI
internationally comparable measure of accounting quality: earnings distributions around the zero earnings
threshold. Additionally, this study investigates the interplay between accrual and real earnings
SC
management levels around the zero earnings benchmark and whether this relationship changes post-IFRS
adoption.
U
Using a global sample of 46 mandatory IFRS adopting countries over 17 years, I find that, while the
N
discontinuity around the zero earnings benchmark remains statistically significant following IFRS
A
adoption, the size of the discontinuity decreases suggesting a moderate increase in accounting quality.
M
Including IFRS adoptions from both developed and developing countries allows for rolling adoption
dates, avoids transitional effects, and is a better representation of the current vision of the IASB. I find
D
discontinuities decrease for both EU and non-EU countries suggesting that the increase in accounting
TE
quality for EU countries is driven by mandatory IFRS adoption and not concurrent changes to the
reporting environment (Christensen et al., 2013). Furthermore, the decrease in the distribution
EP
discontinuities following IFRS adoption is more evident for high enforcement countries and completely
CC
disappears for common law countries, which helps to rectify the findings that, in a meta setting, reporting
environment does not have an impact on the effectiveness of IFRS adoption (Ahmed et al., 2013b).
A
In a second step, I attempt to isolate the role that signed accrual and real earnings management play in
the discontinuity of earnings distributions pre- and post-IFRS adoption. Using both univariate and logit
regression analyses, I find no discernible trend of accrual or real discretionary reporting for firms in the
benchmark beating intervals around the zero earnings threshold, which aligns with Dechow et al. (2003)
24
and Doukakis (2014). This is not surprising given the focus on subtle benchmark beating, which is
difficult to capture in a heterogenous, global setting (DeFond & Park, 2001; Wysocki, 2004).
This study contributes to the existing IFRS and earnings distribution literatures in a number of ways.
For one, measuring accounting quality with earnings distributions is not an arbitrary choice. This
methodology avoids the applicability and validity critiques of discretionary accruals in an international
PT
setting, as well as serves as a reference point for conclusions drawn in prior works. Additionally, this
RI
study adds to the recent trend of using earnings distributions to measure the effectiveness of accounting
policy changes, whereby keeping the sample constant reduces endogeneity concerns and better isolates
SC
the effect. While I use earnings distributions as a tool to evaluate the introduction of IFRS, this measure
could be used in future research to investigate specific changes to IFRS regulation as it evolves over time.
U
The results and limitations of this paper expose many potential avenues for future research. For one,
N
the use of earnings distributions provides an isolated setting to identify what exact reporting practices are
A
being curbed via IFRS adoption since this paper was not able to identify those practices using classic
M
discretionary accrual and real earnings management measures. Further, that finding that the discontinuity
disappears post-IFRS adoption when plotting EBIT, but still remains when plotting bottom-line earnings
D
provides a narrowed focus for future investigation into what exact areas of reporting are vulnerable to
TE
growth or analyst and manager forecast estimates post-IFRS adoption, given that different thresholds have
alternative reporting incentives that could be affected differently by the mandatory introduction of IFRS.
CC
Finally, because this study includes over 40 countries with varying IFRS adoption dates and styles, this
paper should be replicated when those countries with more recent adoption dates have longer post
A
adoption periods and additional countries adopt and further converge with IFRS. In a similar vein,
investigating adoption styles and processes across the world in greater detail, following the call of Zeff &
Nobes (2010) and Zeff (2015), would be an interesting and welcomed extension of this work.
25
Acknowledgments
I would like to thank seminars participants at the University of Mannheim and at the 2017 American
Accounting Association annual meeting for their helpful comments. Excellent research assistance was
provided by Katharina Jung and Jiwon Park. This work was supported by the German Research
Foundation sponsored Graduate School of Economic and Social Sciences of the University of Mannheim
PT
References
RI
Ahmed, A.S., Neel, M., & Wang, D. (2013a). Does mandatory adoption of IFRS improve accounting
SC
quality? Preliminary evidence. Contemporary Accounting Research, 30 (4), 1344-1372.
Ahmed, K., Chalmers, K., & Khlif, H. (2013b). A meta-analysis of IFRS adoption effects. The
International Journal of Accounting, 48 (2), 173-217.
U
Altamuro, J., & Beatty, A. (2010). How does internal control regulation affect financial reporting?
N
Journal of Accounting and Economics, 49 (1-2), 58-74.
Bae, K.-H., Tan, H., & Welker, M. (2008). International GAAP differences: The impact on foreign
A
analysts. The Accounting Review, 83 (3), 593-628.
M
Ball, R., Kothari, S., & Robin, A. (2000). The effect of international institutional factors on properties of
accounting earnings. Journal of Accounting and Economics, 29 (1), 1-51.
D
Barth, M. (2015). Commentary on prospects for global financial reporting. Accounting Perspectives, 14
(3), 154-167.
TE
Barth, M. E., Landsman, W. R., & Lang, M. H. (2008). International Accounting Standards and
accounting quality. Journal of Accounting Research, 46 (3), 467-498.
EP
Bartov, E., & Cohen, D. A. (2009). The numbers game in the pre- and post-Sarbanes-Oxley eras. Journal
of Accounting, Auditing, and Finance, 24 (4), 505-534.
Beaver, W. H., McNichols, M. F., & Nelson, K. K. (2007). An alternative interpretation of the
CC
Bernard, V. L., & Skinner, D. J. (1996). What motivates managers' choice of discretionary accruals?
Journal of Accounting and Economics, 22 (1-3), 313-325.
A
Biddle, G. C., Hilary, G., & Verdi, R. S. (2009). How does financial reporting quality relate to investment
efficiency? Journal of Accounting and Economics, 48 (2-3), 112-131.
Brown, L. D., & Caylor, M. L. (2005). A temporal analysis of quarterly earnings thresh-olds: Propensities
and valuation consequences. The Accounting Review, 80 (2), 423-440.
26
Burgstahler, D., & Chuk, E. (2015). Do scaling and selection explain earnings discontinuities? Journal of
Accounting and Economics, 60 (1), 168-186.
Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and losses.
Journal of Accounting and Economics, 24 (1), 99-126.
Cai, L., Rahman, A., & Courtenay, S. (2014). The effect of IFRS adoption conditional upon the level of
pre-adoption divergence. The International Journal of Accounting, 49 (2), 147-178.
PT
Callao, S., & Jarne, J. I. (2010). Have IFRS affected earnings management in the European Union?
Accounting in Europe, 7 (2), 159-189.
RI
Capkun, V., Collins, D.W., & Jeanjean, T. (2012). Does adoption of IAS/IFRS deter earnings
management? Working Paper, [retrieved from:] http://ssrn.com/sol3/papers.cfm?abstract
id=1850228.
SC
Chaney, P. K., Faccio, M., & Parsley, D. (2011). The quality of accounting information in politically
connected firms. Journal of Accounting and Economics, 51 (1-2), 58-76.
U
Chen, H., Tang, Q., Jiang, Y., & Lin, Z. (2010). The role of International Financial Reporting Standards
in accounting quality: Evidence from the European Union. Journal of International Financial
Management and Accounting, 21 (3), 220-278.
N
A
Christensen, H. B., Hail, L., & Leuz, C. (2013). Mandatory IFRS reporting and changes in enforcement.
Journal of Accounting and Economics, 56 (2-3), 147-177.
M
Christensen, H. B., Lee, E., Walker, M., & Zeng, C. (2015). Incentives or standards: What determines
accounting quality changes around IFRS adoption? European Accounting Review, 24 (1), 31-61.
D
Cohen, D. A., Dey, A., & Lys, T. Z. (2008). Real and accrual-based earnings management in the pre- and
post-Sarbanes-Oxley periods. The Accounting Review, 83 (3), 757-787.
TE
Daske, H., Hail, L., Leuz, C., & Verdi, R. (2013). Adopting a label: Heterogeneity in the economic
consequences around IAS/IFRS adoptions. Journal of Accounting Research, 51 (3), 495-547.
EP
Dechow, P. M., & Dichev, I. (2002). The quality of accruals and earnings: The role of accrual estimation
errors. The Accounting Review, 77 (supplement), 35-59.
Dechow, P. M., Richardson, S. A., & Tuna, I. (2003). Why are earnings kinky? An examination of the
CC
Dechow, P. M., & Skinner, D. J. (2000). Earnings management: Reconciling the views of accounting
academics, practitioners, and regulators. Accounting Horizons, 14 (2), 235-250.
A
Dechow, P. M., Sloan, R., & Sweeney, A. (1995). Detecting earnings management. The Accounting
Review, 70 (2), 193-225.
DeFond, M. L., & Park, C. W. (2001). The reversal of abnormal accruals and the market valuation of
earnings surprises. The Accounting Review, 76 (6), 375-404.
27
Degeorge, F., Patel, J., & Zeckhauser, R. (1999). Earnings management to exceed thresholds. The Journal
of Business, 72 (1), 1-33.
Donelson, D. C., Mcinnis, J. M., & Mergenthaler, R. D. (2013). Discontinuities and earnings
management: Evidence from restatements related to securities litigation. Contemporary
Accounting Research, 30 (1), 242-268.
Doukakis, L. C. (2014). The effect of mandatory IFRS adoption on real and accrual-based earnings
management activities. Journal of Accounting and Public Policy, 33 (6), 551-572.
PT
Durtschi, C., & Easton, P. (2005). Earnings management? The shapes of the frequency distributions of
earnings metrics are not evidence ipso facto. Journal of Accounting Research, 43 (4), 557-592.
RI
Durtschi, C., & Easton, P. (2009). Earnings management? Erroneous inferences based on earnings
frequency distributions. Journal of Accounting Research, 47 (5), 1249-1281.
SC
Enomoto, M., & Yamaguchi, T. (2016). Discontinuity of earnings and earnings changes distribution after
J-SOX implementation: Empirical evidence from Japan. Journal of Accounting and Public
Policy, 36 (1), 82-98.
U
Ezzamel, M., & Xiao, J. Z. (2011). Accounting in transitional and emerging market economies. European
Accounting Review, 20 (4), 625-637.
N
A
Gilliam, T. A., He in, F., & Paterson, J. S. (2015). Evidence that the zero-earnings discontinuity has
disappeared. Journal of Accounting and Economics, 60 (1), 117-132.
M
Graham, J. R., Harvey, C. R., & Rajgopal, S. (2005). The economic implications of corporate financial
reporting. Journal of Accounting and Economics, 40 (1-3), 3-73.
D
Haller, A., & Wehrfritz, M. (2013). The impact of national GAAP and accounting traditions on IFRS
policy selection: Evidence from Germany and the UK. Journal of International Accounting,
TE
Holthausen, R. W. (2009). Accounting standards, financial reporting outcomes, and enforcement. Journal
of Accounting Research, 47 (2), 447-458.
EP
Houqe, M. N., van Zijl, T., Dunstan, K., & Karim, A. W. (2012). The effect of IFRS adoption and
investor protection on earnings quality around the world. The International Journal of
CC
Jacob, J., & Jorgensen, B. N. (2007). Earnings management and accounting income aggregation. Journal
of Accounting and Economics, 43 (2-3), 369-390.
A
Jeanjean, T., & Stolowy, H. (2008). Do accounting standards matter? An exploratory analysis of earnings
management before and after IFRS adoption. Journal of Accounting and Public Policy, 27 (6),
480-494.
Jermakowicz, E. K., & Gornik-Tomaszewski, S. (2006). Implementing IFRS from the perspective of EU
publicly traded companies. Journal of International Accounting, Auditing and Taxation, 15 (2),
170-196.
28
Judge, W., Li, S., & Pinsker, R. (2010). National adoption of international accounting standards: An
institutional perspective. Corporate Governance: An International Review, 18 (3), 161-174.
Kothari, S. P., Leone, A. J., & Wasley, C. E. (2005). Performance matched discretionary accrual
measures. Journal of Accounting and Economics, 39 (1), 163-197.
Lamoreaux, P. T., Michas, P. N., & Schultz, W. L. (2015). Do accounting and audit quality affect World
Bank lending? The Accounting Review, 90 (2), 703-738.
PT
Larson, R. K., & Street, D. L. (2004). Convergence with IFRS in an expanding Europe: progress and
obstacles identified by large accounting firms' survey. Journal of International Accounting,
Auditing and Taxation, 13 (2), 89-119.
RI
Leuz, C., Nanda, D., & Wysocki, P. D. (2003). Earnings management and investor protection: an
international comparison. Journal of Financial Economics, 69 (3), 505-527.
SC
McNichols, M. F. (2000). Research design issues in earnings management studies. Journal of Accounting
and Public Policy, 19 (4-5), 313-345.
U
Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal of
Accounting and Economics, 42 (3), 335-370.
N
Soderstrom, N. S., & Sun, K. J. (2007). IFRS adoption and accounting quality: A review. European
A
Accounting Review, 16 (4), 675-702.
Wysocki, P. D. (2004). Discussion of ultimate ownership, income management, and legal and extra-legal
M
Zang, A. Y. (2012). Evidence on the trade-o between real activities manipulation and accrual-based
D
Zeff, S. A. (2007). Some obstacles to global financial reporting comparability and convergence at a high
level of quality. The British Accounting Review, 39 (4), 290-302.
Zeff, S. A. (2015). “In the Literature” but wrong: Switzerland and the adoption of IFRS. Journal of
EP
Zeff, S. A., & Nobes, C. W. (2010). Commentary: Has Australia (or any other jurisdiction) adopted
CC
Zeghal, D., Chtourou, S. M., & Fourati, Y. M. (2012). The effect of mandatory adoption of IFRS on
earnings quality: Evidence from the European Union. Journal of International Accounting
Research, 11 (2), 1-25.
A
Zeghal, D., Chtourou, S. M., & Sellami, Y. M. (2011). An analysis of the effect of mandatory adoption of
IAS/IFRS on earnings management. Journal of International Accounting, Auditing and Taxation,
20 (2), 61-72.
29
Figure 1
PT
RI
SC
Panel B: Pre-IFRS Adoption Period
U
N
A
M
D
30
PT
RI
SC
U
Distributions of Scaled Earnings using the Constant Sample
N
A
Figure 2 displays the distributions of earnings scaled by lagged market value of equity. The bin size is 0.01 calculated from the Freedman
Diaconis measure and the distributions are truncated at -0.25 and 0.35 to give the clearest picture of the zero earnings threshold, as indicated by
the red dashed line. The earnings and the beginning market value of equity are collected from Thomson Reuters Datastream. Both variables are
M
winsorized at the yearly 2nd and 98th percentiles. The sample is limited to a subset of constant firms with equal numbers of observations in both
the pre- and post-IFRS adoption periods. ***, **, and * represent significance levels of 10, 5, and 1%, respectively.
D
TE
EP
CC
A
31
Figure 3
PT
RI
SC
U
N
A
Distributions of Scaled Earnings for EU vs. non-EU Countries
Figure 3 displays the distributions of earnings scaled by lagged market value of equity in the pre- and post-IFRS adoption periods for EU
(n=4,530 per period) vs. non-EU countries (n=6,155 per period). The bin size is 0.01 calculated from the Freedman Diaconis measure and the
M
distributions are truncated at -0.25 and 0.35 to give the clearest picture of the zero earnings threshold, as indicated by the red dashed line. The
earnings and the beginning market value of equity are collected from Thomson Reuters Datastream. Both variables are winsorized at the yearly
2nd and 98th percentiles. The sample is limited to a subset of constant firms with equal numbers of observations in both the pre-and post-IFRS
adoption periods. ***, **, and * represent significance levels of 10, 5, and 1%, respectively.
D
TE
EP
CC
A
32
Figure 4
PT
RI
SC
U
N
A
Panel B: Legal Origin
M
D
TE
EP
CC
A
33
Figure 5
PT
RI
SC
U
N
A
Alternative Numerator Measure for Scaled Earnings Distributions
Figure 5 displays the earnings distributions of the constant sample of firms using an alternative numerator measure of earnings before interest and
taxes (EBIT) scaled by the lagged market value of equity. The bin size is 0.01 calculated from the Freedman Diaconis measure and the
M
distributions are truncated at -0.25 and 0.35 to give the clearest picture of the zero earnings threshold, as indicated by the red dashed line. All
variables are collected from Thomson Reuters Datastream and are winsorized at the yearly 2nd and 98th percentiles. ***, **, and * represent
significance levels of 10, 5, and 1%, respectively.
D
TE
EP
CC
A
34
Figure 6
PT
RI
SC
U
N
Panel B: Alternative Denominator – Lagged Total Assets
A
M
D
TE
EP
CC
A
35
Table 1: Mandatory IFRS Adoption Identification Strategy
PT
Local GAAP IFRS Local GAAP Exclude
IFRS Local GAAP IFRS Exclude
IFRS Local GAAP Local GAAP Exclude
Local GAAP/IFRS - - Exclude
RI
- Local GAAP/IFRS Local GAAP/IFRS Exclude
- - Local GAAP/IFRS Exclude
SC
*Included for convergence countries only
Note: Table 1 displays the identification strategy for the mandatory IFRS adoption variable (binary). Since this paper focuses on a global IFRS
adoption perspective, the identification of IFRS adoption at the firm level includes distinguishing between different country-specific adoption
dates and multiple adoption types (e.g., convergence, required for some firms, required for all firms, etc.). For the cleanest measure, I only
U
include firms with a definite switch to IFRS at the country-level mandated adoption date. For convergence countries, I include firms that adopt
IFRS after the convergence date due to potential misclassification in databases (Zeff, 2007). All other adoption variations are excluded. Adoption
dates and types are collected via cross-referencing the IASB jurisdiction profiles, Deloitte's “Use of IFRS by Jurisdiction" webpage, and
PricewaterhouseCooper's “IFRS Adoption by Country" survey.
N
A
M
Omitted Total
Observations Sample Size
D
Note: Table 2 outlines the data cleaning procedure that results in the final sample used for the distribution analyses. The sample begins with all
available firms in IFRS adopting countries from Thomson Reuters Datastream that have net income, market value of equity, sales, and total assets
available each year. Firm-level adoption details are identified via the “Accounting Standards Followed" variable. Firms following “U.S. standards
(GAAP)" and “U.S. GAAP reclassified from local standards" are excluded. Financial institutions with Datastream industry codes between 4300-
4600 are excluded. Country-level adoption dates and types are collected via cross-referencing the IASB jurisdiction profiles, Deloitte's “Use of
IFRS by Jurisdiction" webpage, and PricewaterhouseCooper's “IFRS Adoption by Country" survey. The truncated endpoints for the earnings
distributions are -0.25 and 0.35, following prior literature.
36
Table 3: Firm-year Observations by Country
Country No. Obs. % Obs. Development Level Adoption Year Adoption Type
Argentina 381 0.81 Developing 2012 Required (some)
Australia 1,010 2.15 Developed 2005 Required
Austria 102 0.22 Developed 2005 Required (EU)
Belgium 584 1.24 Developed 2005 Required (EU)
Brazil 3,558 7.58 Developing 2010 Required
Canada 6,065 12.92 Developed 2011 Required (some)
Chile 890 1.90 Developing 2009 Required
China 95 0.20 Developing 2007 Convergence
PT
Czech Republic 72 0.15 Developed 2005 Required (EU)
Denmark 721 1.54 Developed 2005 Required (EU)
Egypt 19 0.04 Developing 2007 Required
Finland 1,190 2.54 Developed 2005 Required (EU)
France 3,920 8.35 Developed 2005 Required (EU)
RI
Germany 1,384 2.95 Developed 2005 Required (EU)
Ghana 12 0.03 Developing 2007 Required
Greece 2,307 4.91 Developed 2005 Required (EU)
Hong Kong 25 0.05 Developed 2005 Required
SC
Hungary 46 0.10 Developed 2005 Required (EU)
Ireland 238 0.51 Developed 2005 Required (EU)
Israel 1,537 3.27 Developed 2008 Required (some)
Italy 2,316 4.93 Developed 2005 Required (EU)
Luxembourg 74 0.16 Developed 2005 Required (EU)
U
Macedonia 8 0.02 Developing 2009 Required
Malaysia 47 0.10 Developing 2012 Convergence
Mexico
Morocco
Netherlands
1,137
25
994
2.42
0.05
2.12
N
Developing
Developing
Developed
2012
2008
2005
Required (some)
Required (some)
Required (EU)
A
New Zealand 58 0.12 Developed 2005 Required
Nigeria 87 0.19 Developing 2012 Required
Norway 1,028 2.19 Developed 2005 Required (EU)
M
Peru 14 0.03 Developing 2012 Required
Philippines 180 0.38 Developing 2005 Convergence
Poland 809 1.72 Developing 2005 Required (EU)
Portugal 406 0.86 Developed 2005 Required (EU)
Russian Federation 66 0.14 Developing 2012 Required (some)
D
Note: Table 3 lists the 46 countries with truncated scaled earnings data available that are used in the main analyses. The table gives the number of
firm-year observations with net income and market value data available. The country development level is established from cross-referencing the
United Nations' UNCTAD survey and the World Trade Organization classification. Adoption dates and types are collected via cross-referencing
the IASB jurisdiction profiles, Deloitte's “Use of IFRS by Jurisdiction" webpage, and PricewaterhouseCooper's “IFRS Adoption by Country"
A
survey.
37
PT
RI
Table 4: Descriptive Statistics across IFRS Adoption Periods
SC
Mean Median
Pre-IFRS Adoption Period Post-IFRS Adoption Period Differences Differences
No. Obs. Mean Median SD P25 P75 No. Obs. Mean Median SD P25 P75 t-stat χ2
U
Main Test Variable
NI/MV 28,986 0.054 0.058 0.108 0.003 0.113 17,953 0.044 0.056 0.101 0.000 0.097 10.04*** 2.85*
N
Earnings Management Variables
Mod. Jones 21,186 0.007 -0.002 0.976 -0.068 0.061 13,873 0.008 0.002 0.224 -0.055 0.056 -0.05 10.41***
A
Mod. D&D 17,731 0.006 0.003 0.121 -0.031 0.041 9,891 0.009 0.006 0.098 -0.025 0.039 -2.10** 15.73***
Agg. AEM 17,731 0.009 0.000 0.349 -0.105 0.113 9,891 0.019 0.010 0.308 -0.088 0.111 -2.55*** 16.13***
Benchmark DA 17,825 0.020 0.004 0.717 -0.071 0.088 12783 0.071 -0.001 6.223 -0.071 0.068 -1.08 21.53***
M
REM1 14,921 -0.004 0.014 0.431 -0.143 0.148 9,221 -0.060 -0.014 0.649 -0.193 0.123 8.01*** 70.98***
REM2 15,326 -0.012 -0.007 0.335 -0.125 0.109 10,749 -0.063 -0.036 0.416 -0.162 0.076 10.84*** 104.77***
Agg. REM 11,724 0.018 0.022 0.238 -0.067 0.109 7,107 -0.019 0.002 0.317 -0.098 0.086 9.11*** 72.30***
Equity Issuance 19,893 0.255 0.037 6.721 -0.055 0.164 13,360 0.340 0.020 22.713 -0.049 0.109 -0.49 32.11***
Turnover 19,893 1.110 0.949 0.888 0.591 1.400 13,360 1.031 0.886 0.721 0.562 1.325 8.59*** 52.64***
Closely held 19,893 0.328 0.274 0.642 0.000 0.542 13,360 0.388 0.361 0.699 0.043 0.600 -8.11*** 239.20***
Big 4 19,893 0.743 1.000 0.437 0.000 1.000 13,360 0.756 1.000 0.430 1.000 1.000 -2.62***
Exchange Count 19,893 1.237 1.000 0.711 1.000 1.000 13,360 1.294 1.000 0.833 1.000 1.000 -6.73*** 16.64***
E
U.S. Cross-listed 19,893 0.032 0.000 0.176 0.000 0.000 13,360 0.018 0.000 0.132 0.000 0.000 7.99***
Robustness Variables
CC
EBIT/MV 24,769 0.097 0.103 0.121 0.026 0.177 16,492 0.090 0.099 0.111 0.030 0.158 5.94*** 13.28***
NI/SALE 24,991 0.054 0.042 0.084 0.011 0.089 15,881 0.048 0.040 0.081 0.007 0.084 7.06*** 15.99***
NI/TA 27,924 0.038 0.035 0.077 0.005 0.075 17,395 0.035 0.034 0.073 0.002 0.069 3.44*** 11.93***
Mod. Jones (abs) 21,186 0.117 0.065 0.969 0.029 0.129 13,873 0.098 0.056 0.202 0.024 0.111 2.27** 83.69***
Mod. D&D (abs) 17,731 0.060 0.036 0.105 0.016 0.072 9,891 0.052 0.032 0.084 0.015 0.063 6.16*** 40.01***
A
Agg. AEM (abs) 17,731 -0.040 -0.115 0.318 -0.177 -0.002 9,891 -0.061 -0.126 0.278 -0.185 -0.027 5.60*** 35.39***
Benchmark DA (abs) 17,825 0.155 0.080 0.700 0.034 0.163 12,783 0.199 0.069 6.220 0.029 0.143 -0.95 80.89***
REM1 (abs) 14,921 0.238 0.146 0.359 0.065 0.284 9,221 0.271 0.153 0.593 0.067 0.311 -5.46*** 7.81***
REM2 (abs) 15,326 0.185 0.117 0.280 0.053 0.219 10,749 0.200 0.119 0.371 0.054 0.231 -3.62*** 0.34
Agg. REM (abs) 11,724 -0.044 -0.096 0.201 -0.143 -0.015 7,107 -0.033 -0.095 0.293 -0.144 -0.006 -3.13*** 0.33
Note: Table 4 presents the descriptive statistics in both the pre- and post-IFRS adoption periods. All variable definitions are described in Appendix A. All non-ratio variables are winsorized annually at
the 2nd and 98th percentiles. The mean comparison (t-stat) and median comparison (χ2) are included in the last two columns. The difference in means is calculated using robust standard errors. ***,
**, and * represent significance levels of 10, 5, and 1%, respectively.
PT
RI
Table 5: Spearman and Pearson Correlation Table
SC
I II III IV V VI VII VIII IX X
I IFRS 1 -0.0603* 0.001 0.0148* 0.0169* 0.0068 -0.0148 -0.0623* -0.0609* 0.0521*
II NI/MV -0.0391* 1 0.0128* 0.1891* 0.1707* -0.0017 -0.0082 -0.0418* -0.0382* 0.1976*
U
III Mod. Jones 0.0029 0.0710* 1 0.4128* 0.6865* 0.0308* 0.0751* 0.4127* 0.3540* 0.0013
IV Mod. D&D 0.0193 0.2976* 0.5369* 1 0.9457* 0.0903* 0.0111 0.1415* 0.1338* -0.0226*
V Agg. AEM 0.0136 0.2443* 0.7652* 0.9318* 1 0.1384* 0.0374* 0.2695* 0.2451* -0.0173*
N
VI Benchmark DA -0.0501* 0.0714* 0.4757* 0.1880* 0.3064* 1 -0.0087 0.0179* 0.1044* -0.0048
VII REM1 -0.0730* -0.0453* 0.0565* -0.0099 0.0135 -0.0169 1 0.5760* 0.8035* 0.009
VIII REM2 -0.0782* -0.0970* 0.4234* 0.1575* 0.2671* 0.2352* 0.7055* 1 0.9495* 0.0310*
A
IX Agg. REM -0.0780* -0.0831* 0.3251* 0.1096* 0.1977* 0.1692* 0.8516* 0.9590* 1 0.0326*
X Size 0.1013* 0.0317* -0.0117 -0.0670* -0.0519* -0.0173 -0.0607* -0.0358* -0.0332* 1
XI Leverage 0.1033* -0.1117* -0.0167 -0.0959* -0.0765* -0.0126 -0.0392* 0.0166 0.0059 0.4813*
M
XII Sales Growth -0.0836* 0.1782* 0.0511* 0.0290* 0.0368* 0.1802* -0.0226* -0.0283* -0.0240* 0.0625*
XIII Op. Cash Flows -0.0247* 0.2737* -0.6233* -0.2367* -0.3905* -0.5188* -0.1857* -0.5206* -0.4365* 0.0614*
XIV Debt Issuance -0.0728* 0.0520* 0.0428* 0.0241* 0.0296* 0.1371* 0.0145 -0.0053 0.0015 0.0650*
XV Equity Issuance -0.0975* 0.0364* 0.0683* 0.0467* 0.0580* 0.2051* -0.0132 0.0277* 0.02 0.0108
XVI Turnover -0.0287* 0.2402* 0.0434* 0.0588* 0.0592* 0.0941* -0.0272* 0.0023 -0.0106 -0.1349*
ED
XVII Closely Held 0.0965* 0.0500* 0.0020 0.0117 0.0087 0.0112 -0.0731* -0.0240* -0.0435* 0.1442*
XVIII Big 4 0.0234* 0.0747* -0.0422* -0.0323* -0.0384* -0.016 -0.1277* -0.1045* -0.1129* 0.4207*
XIX Exchange Count 0.0537* -0.0079 -0.0156 -0.0121 -0.0129 -0.0109 -0.0929* -0.0806* -0.0817* 0.3200*
XX U.S. Cross-listed -0.0236* -0.0085 -0.016 -0.0229* -0.0203 -0.0021 0.0120 -0.0383* -0.0123 0.1667*
XI XII XIII XIV XV XVI XVII XVIII XIX XX
PT
I IFRS 0.0343* -0.0052 0.0014 -0.0088 0.0064 -0.0409* 0.0074 -0.0411* -0.0019 -0.0604*
II NI/MV -0.0159* 0.0003 0.0866* -0.0016 -0.0028 0.1931* 0.0181* 0.0969* -0.0123* -0.0051
III Mod. Jones -0.009 -0.0507* -0.9366* 0.1088* 0.0002 0.4437* -0.0003 -0.0016 0.0042 0.0253*
IV Mod. D&D -0.0392* 0.0042 -0.1773* 0.0192* 0.0176* -0.002 0.0002 -0.0137 -0.0091 -0.0125
V Agg. AEM -0.0337* 0.0160* -0.3176* 0.0268* 0.0141 0.0215* 0.0003 -0.0156* -0.0092 -0.0143
E
VI Benchmark DA 0.0074 0.6086* -0.0498* 0.0124 0.0011 0.0017 0.0002 -0.0113 -0.0032 -0.0012
VII REM1 -0.014 0.0011 -0.0644* 0.0129 0.0004 0.0235* -0.0055 -0.0075 -0.0037 0.0023
CC
VIII REM2 0.0328* -0.0355* -0.4967* -0.0246* 0.0008 0.0372* 0.0011 -0.0427* -0.0209* 0.0042
IX Agg. REM 0.0297* -0.0336* -0.4190* 0.0101 0.0077 0.1002* -0.0006 -0.0576* -0.0255* 0.0024
X Size 0.3024* 0.0035 0.0534* 0.0016 0.0043 -0.0051 0.011 0.3976* 0.2684* 0.1711*
XI Leverage 1 0.009 0.0032 -0.0056 0.0118* -0.0705* 0.0119* 0.1514* 0.0862* 0.0704*
XII Sales Growth 0.0484* 1 -0.0158* 0.3340* 0.0024 0.1934* -0.0003 -0.0073 -0.0022 -0.0008
XIII Op. Cash Flows -0.0373* -0.0748* 1 -0.1213* -0.0035 -0.3920* 0.0024 0.0306* 0.004 -0.0226*
A
XIV Debt Issuance 0.0053 0.5332* -0.0863* 1 0.0024 0.2073* -0.0018 -0.002 -0.0069 0.0022
XV Equity Issuance 0.0434* 0.5227* -0.1702* 0.3553* 1 -0.0013 -0.0022 0.0027 -0.0013 0.0001
XVI Turnover -0.1638* 0.3528* 0.0187 0.2453* 0.2061* 1 0.0021 0.0311* -0.0473* -0.0561*
XVII Closely Held 0.0731* -0.0143 0.0501* -0.0241* -0.0068 -0.0260* 1 0.0097 -0.0008 -0.0031
XVIII Big 4 0.2179* 0.0212 0.1338* 0.0091 -0.0198 0.0271* 0.1499* 1 0.1226* 0.0843*
XIX Exchange Count 0.1875* -0.0256* 0.0580* -0.0178 0.0102 -0.0739* 0.0340* 0.1397* 1 0.2797*
XX U.S. Cross-listed 0.1033* -0.0014 0.0314* 0.0014 0.0152 -0.0922* -0.0389* 0.0690* 0.2472* 1
Note: Table 5 displays the correlation matrix between the dependent variables, independent variables, and firm control variables. All variables definitions are found in Appendix A. All non-ratio
variables are winsorized annually at the 2nd and 98th percentiles. Below the diagonal are the Spearman ranked correlations and above the diagonal are the Pearson correlations. A star (*) indicates
significant correlations at the 1% level.
39
PT
RI
Table 6: Signed Earnings Management Measures around Distribution Thresholds
SC
Panel A: Small Profit Firms Compared to Small Loss Firms
Pre-IFRS Adoption Post-IFRS Adoption
EM Measures N Small Loss N Small Profit t-stat N Small Loss N Small Profit t-stat
U
Mod. Jones 208 0.038 482 -0.010 -1.83* 115 0.014 246 -0.010 -1.26
Mod. D&D 172 -0.005 378 -0.007 -0.26 79 -0.001 171 -0.001 -0.02
N
Agg. AEM 172 -0.003 378 -0.025 -0.87 79 -0.010 171 -0.017 -0.19
Benchmark DA 168 0.035 381 0.000 -1.16 101 -0.038 220 0.005 1.00
REM1 123 -0.027 329 0.054 1.98** 95 -0.061 163 -0.011 0.69
A
REM2 132 0.007 329 0.017 0.29 100 -0.067 184 -0.048 0.40
Agg. REM 93 0.016 250 0.047 1.02 73 -0.037 126 0.003 0.85
M
Panel B: Small Profit Firms Compared to All Other Firms
Pre-IFRS Adoption Post-IFRS Adoption
EM Measures N Small Profit N All Others t-stat N Small Profit N All Others t-stat
ED
Mod. Jones 482 -0.010 30,812 0.007 1.73* 246 -0.010 15,606 0.005 1.50
Mod. D&D 378 -0.007 25,491 -0.001 1.32 171 -0.001 10,946 0.004 0.67
Agg. AEM 378 -0.025 25,491 -0.007 1.29 171 -0.017 10,946 0.008 1.32
Benchmark DA 381 0.000 25,714 0.111 2.19** 220 0.005 14,286 0.069 1.27
REM1 329 0.054 20,066 0.020 -0.95 163 -0.011 10,902 -0.055 -1.42
PT
REM2 329 0.017 20,731 -0.005 -0.99 184 -0.048 12,318 -0.050 -0.09
Agg. REM 250 0.047 15,623 0.021 -1.25 126 0.003 8,194 -0.014 -0.98
Mod. Jones 208 0.038 31,086 0.007 -1.22 115 0.014 15,737 0.005 -0.56
Mod. D&D 172 -0.005 25,697 -0.001 0.54 79 -0.001 11,038 0.004 0.39
Agg. AEM 172 -0.003 25,697 -0.007 -0.19 79 -0.010 11,038 0.008 0.54
Benchmark DA 168 0.035 25,927 0.110 1.32 101 -0.038 14,405 0.069 1.66*
REM1 123 -0.027 20,272 0.021 1.37 95 -0.061 10,970 -0.054 0.10
A
REM2 132 0.007 20,928 -0.005 -0.48 100 -0.067 12,402 -0.050 0.40
Agg. REM 93 0.016 15,780 0.021 0.22 73 -0.037 8,247 -0.013 0.54
Note: Table 6 displays the comparison of signed levels of accrual and real earnings management measures across different subsections of small profit firms, small loss firms, and all other firms using the
full sample of observations. All variables are defined in Appendix A and non-ratio variables are winsorized annually at the 2nd and 98 th percentiles. The small profit and small loss categories are
defined as scaled earnings in the intervals [0, 0.005) and (-0.005, 0], respectively. The difference in means is calculated using robust standard errors. ***, **, and * represent significance levels of 10, 5,
and 1%, respectively.
40
Table 7: Probability of Benchmark Beating Regression Results
𝑃𝑟𝑜𝑓𝑖𝑡𝑖𝑡 = 𝛽0 + 𝛽1 𝐼𝐹𝑅𝑆𝑖𝑡 + 𝛽2 𝐸𝑀𝑖𝑡 + 𝛽3 𝐼𝐹𝑅𝑆𝑖𝑡 ∗ 𝐸𝑀𝑖𝑡 + 𝛽4 𝑆𝑖𝑧𝑒𝑖𝑡 + 𝛽5 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 + 𝛽6 𝑆𝑎𝑙𝑒𝑠 𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡 + 𝛽7 𝑂𝐶𝐹𝑖𝑡
+ 𝛽8 𝐷𝑒𝑏𝑡 𝐼𝑠𝑠𝑢𝑒𝑖𝑡 + 𝛽9 𝐸𝑞𝑢𝑖𝑡𝑦 𝐼𝑠𝑠𝑢𝑒𝑖𝑡 + 𝛽10 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟𝑖𝑡 + 𝛽11 𝐶𝑙𝑜𝑠𝑒𝑙𝑦 𝐻𝑒𝑙𝑑𝑖𝑡 + 𝛽12 𝐵𝑖𝑔4𝑖
+ 𝛽13 𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝐶𝑜𝑢𝑛𝑡𝑖 + 𝛽14 𝑈. 𝑆. 𝐶𝑟𝑜𝑠𝑠 − 𝑙𝑖𝑠𝑡𝑒𝑑𝑖 + 𝑓𝑖𝑥𝑒𝑑 𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜖𝑖𝑡
Profit = [0 - 0.005)
Total Sample Constant Sample
+/- (1) (2) (3) (4) (5) (6)
PT
(-8.97) (-7.00) (-7.08) (-4.11) (-3.24) (-2.63)
IFRS - -0.143** -0.095 -0.129 -0.094 -0.110 -0.021
(-2.37) (-1.30) (1.53) (-1.09) (-1.08) (-0.17)
Agg. AEM + -0.088 -0.256
RI
(-1.13) (-1.08)
IFRS x Agg. AEM - -0.028 -1.036***
(-0.16) (-2.55)
Agg. REM + 0.371* 0.577
SC
(1.76) (1.31)
IFRS x Agg. REM - -0.195 -0.401
(-0.79) (-0.70)
Size - -0.068*** -0.088*** -0.048* -0.097*** -0.128*** -0.111***
U
(-3.87) (-4.21) (-1.84) (-3.40) (-3.87) (-2.57)
Leverage + -0.013 0.041 -0.002 0.037 0.014 0.065
Sales Growth +
(-0.13)
-0.004
(-0.05)
(0.24)
-0.014
(-0.58)
N (-0.02)
-0.008
(-0.38)
(0.14)
-0.005
(-0.42)
(0.05)
-0.001
(-0.10)
(0.19)
-0.002
(-0.15)
A
Operating Cash Flow - -0.027 -0.148 0.079 -0.211* -0.494*** -0.148
(0.30) (-1.35) (0.49) (-1.84) (-3.38) (-0.55)
M
Note: Table 7 displays the logit regression results regarding the probability of a firm observation appearing in the interval direction above the
zero earnings threshold. All model specifications use the dependent binary variable Pro t that equals one if the scaled earnings value falls in the
interval directly above zero, or [0-0.005). Columns 1-3 use the total sample of firm-year observations from 1997-2013, whereas Columns 4-6 use
the constant sample as described in Figure 2. The variable definitions are found in Appendix A and non-ratio variables are winsorized annually at
the 2nd and 98th percentiles. EM is a placeholder variable for either the Agg. AEM or Agg. REM measures of accrual or real earnings
management, respectively. Industry classifications are derived from the Datastream industry codes. The t-statistics are in parentheses and are
calculated using robust standard errors. ***, **, and * represent significance levels of 10, 5, and 1%, respectively.
Appendix A: Definitions of Key Variables
PT
Mod. Jones Modified Jones Model (Kothari et al., 2005)
RI
+ ϵit
where
SC
∆𝐶𝐴𝑖𝑡 − ∆𝐶𝐿𝑖𝑡 − ∆𝐶𝐴𝑆𝐻𝑖𝑡 + ∆𝑆𝑇𝐷𝐸𝐵𝑇𝑖𝑡 − 𝐷𝐸𝑃𝑖𝑡
𝑇𝑂𝑇𝐴𝐶𝐶𝑖𝑡 =
𝑇𝐴𝑖,𝑡−1
U
The regression is run across each industry-year group with at least 15 observations
and includes country control variables (lagged GDP growth and inflation)
Mod. D&D N
Modified Dechow & Dichev (2002) Model
A
OCFi,t−1 OCFi,t OCFi,t+1 ∆(Revit − RECit )
TOTACCit = α0 + α1 + α2 + α3 + α4
TAi,t−1 TAi,t−1 TAi,t−1 TAi,t−1
M
PPEi,t
+ α5 + ϵit
TAi,t−1
The regression is run across each industry-year group with at least 15 observations
D
and includes country control variables (lagged GDP growth and inflation)
TE
Agg. AEM Standardized and aggregated summary variable of Mod. Jones and Mod. D&D
calculated using the method of Biddle et al. (2009)
Benchmark DA Benchmark Discretionary Accrual Model (DeFond & Park, 2001)
EP
where
CC
𝐶𝐴𝐶𝐶 𝐷𝐸𝑃
[𝑅𝐸𝑉𝑖𝑡 𝑥 𝑅𝐸𝑉 𝑖,𝑡−1 ] − [𝑃𝑃𝐸𝑖𝑡 𝑥 𝑃𝑃𝐸𝑖,𝑡−1 ]
𝑖,𝑡−1 𝑖,𝑡−1
𝑁𝐷𝐴𝐶𝐶𝑖𝑡 =
𝑇𝐴𝑖,𝑡−1
A
42
REM1 Sum of abnormal discretionary expenses (multiplied by -1) and abnormal production
costs
PT
where 𝐷𝐼𝑆𝐶 𝐸𝑋𝑃 is measured as the sum of research and development expense
(WC01201) and selling, general, and administrative expenses (WC01101) and 𝑃𝑅𝑂𝐷
is measured as the sum of cost of goods sold (WC01051) and inventory (WC02101).
RI
All variables are scaled by lagged total assets (WC02999)
The regression is run across each industry-year group with at least 15 observations
SC
and includes country control variables (lagged GDP growth and inflation)
REM2 Sum of abnormal discretionary expenses and abnormal operating cash flows
(multiplied by -1)
U
Note: abnormal discretionary expenses are calculated the same as in REM1 and OCF
OCFit Salesit
N
is calculated the same as the firm control variable (see below)
∆Salesit
A
= α0 + α1 + α2 + ϵit
TAi,t−1 TAi,t−1 TAi,t−1
M
The regression is run across each industry-year group with at least 15 observations
D
and includes country control variables (lagged GDP growth and inflation)
Agg. REM Standardized and aggregated summary variable of REM1 and REM2 calculated
TE
Sales Growth The difference between current and prior period sales (WC01001) divided by sales of
the prior period
Op. Cash Flows Operating cash flows are calculated indirectly as: 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒 − 𝑇𝑂𝑇𝐴𝐶𝐶, divided
A
43
Closely Held Shares The percentage of total common shares outstanding (WC05301) that is considered
“closely held” by insiders and major shareholders (WC05475)
Big 4 A binary variable equal to one if the firm employs PricewaterhouseCoopers
(PwC), Deloitte, Ernst & Young (E&Y), or KPMG (WC07800)
Exchange Count A static count variable of the number of stock exchanges on which the firm is listed
(WC05427)
U.S. Cross-listed A binary variable equal to one if the firm is listed on a U.S. stock exchange as
PT
coded by NYSE, BSE, CIN, NSE, PWB, NAS, ASE, or PCS (WC05427)
Robustness Variables
RI
EBIT/MV Earnings before interest and taxes (WC18191) scaled by lagged market value of
equity (WC08002)
NI/SALE Net income (WC01551) scaled by lagged sales (WC01001)
SC
NI/TA Net income (WC01551) scaled by lagged total assets (WC02999)
Absolutes Several earnings management variables are also calculated in absolute terms, which
U
are signified with an “(abs)” distinction
N
A
M
D
TE
EP
CC
A
44