Summarize and PowerPoint

Make a summary and PowerPoint and answer case study (A,B) (Please use easy words and easy sentences)

A. prepare a 3 or 4 pages summary of each paper (3 Article) (including the Libby box)

Focus on summarizing the most salient points of the article.

. After that make sure the summaries address the following questions (I will send them).

B.and make PowerPoint slides for each study.

Three Articles in the attachments

Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
Contents lists available at ScienceDirect
Advances in Accounting, incorporating Advances in
International Accounting
journal homepage: www.elsevier.com/locate/adiac
Managerial ability and goodwill impairment☆,☆☆
Li Sun
School of Accounting, Collins College of Business, University of Tulsa, Tulsa, OK, 74104, USA
a r t i c l e
i n f o
Available online 29 March 2016
Keywords:
Managerial ability
Goodwill
Goodwill impairment
a b s t r a c t
This study examines the relationship between managerial ability and goodwill impairment. I predict a negative
relationship because prior studies suggest that more-able managers better prevent or reduce goodwill impairment, relative to less-able managers. Regression analysis reveals a signiﬁcant and negative relationship between
managerial ability and goodwill impairment measured as the likelihood of goodwill impairment and the magnitude of goodwill impairment losses. Overall, evidence suggests that managers with greater ability play an important role in preventing or reducing goodwill impairment.
© 2016 Elsevier Ltd. All rights reserved.
1. Introduction
In June 2001, the Financial Accounting Standards Board (FASB) issued Statement of Financial Accounting Standards No. 142 (SFAS 142),
Goodwill and Other Intangible Assets.1 Prior to SFAS 142, any excess of
purchase price over the fair value of the acquired ﬁrm’s net assets was
recognized as goodwill. The value of goodwill in a purchase acquisition
was then amortized over a period of up to 40 years. SFAS 142 eliminates
the practice of systematic amortization of goodwill in business combinations; instead, the standard requires companies to assess goodwill
for impairment annually and to recognize a loss if goodwill is impaired.
Hayn and Hughes (2006) argue that the new goodwill impairment accounting practices under SFAS 142 put more responsibility on managers
to determine the fair value of goodwill, suggesting that management
plays an important role in the process of determining the fair value of
goodwill and the magnitude of goodwill impairment losses if goodwill
impairment exists.
The purpose of this study is to examine the relationship between
managerial ability and goodwill impairment. This study focuses on
goodwill impairment for the following reasons: First, goodwill accounts
for a signiﬁcant amount of a ﬁrm’s balance sheet and thus it is an
☆ I thank Roger Graham (editor), Kathleen Rupley (associate editor), and one
anonymous reviewer. I am especially indebted to the associate editor for her work and
care in the review process. I appreciate the comments and suggestions provided by
Henry Huang, Sandeep Nabar, Michelle Draeger, Mollie Mathis, and the doctoral
students at Oklahoma State University. All errors are my own.
☆☆ Data Availability: Data are available from sources identiﬁed in the paper.
E-mail address: li-sun@utulsa.edu.
1
Under FASB Accounting Standards Update 350–20 (ASU 350–20), the annual assessment of goodwill impairment is no longer required after 12/15/2011 but rather when conditions exist that it is more likely than not that the reporting unit fair value is less than its
carrying value.
http://dx.doi.org/10.1016/j.adiac.2016.02.002
0882-6110/© 2016 Elsevier Ltd. All rights reserved.
important corporate asset (Jennings, Robinson, Thompson, & Duvall,
1996). Second, goodwill valuation is a key input when assessing a ﬁrm’s
future cash ﬂows (Hayn & Hughes, 2006). Investors extract goodwill information to form appropriate perceptions concerning a ﬁrm’s intangible assets. Third, SFAS 142 requires the goodwill impairment test if
there is a decline in the fair value of a reporting unit. Thus, goodwill is
regarded as the most sensitive asset to a decline in ﬁrm value (Filip,
Jeanjean, & Paugam, 2015). Fourth, goodwill impairment reﬂects managerial inability to extract value from prior acquisitions. Fifth, goodwill
impairment is a leading indicator of future ﬁrm performance stemming
from the failure to realize the expected beneﬁts of prior acquisitions (Li,
Shroff, Venkataraman, & Zhang, 2011). Last, the frequency of goodwill
impairments has drastically increased and goodwill impairment losses
have become economically signiﬁcant events (Darrough, Guler, & Wang,
2014).
Demerjian, Lev, and McVay (2012) argue that more-able managers
better foresee business opportunities, make better decisions, and better
manage their ﬁrms to maximize shareholders’ beneﬁts, relative to lessable managers. Other studies on managerial ability document that
more-able managers better smooth earnings to maximize shareholders’
beneﬁts (Demerjian, Lewis-Western, & McVay, 2015), engage in fewer
tax-avoidance activities (Francis, Sun, & Wu, 2014) and fewer
earnings-management activities (Demerjian, Lewis, Lev, & McVay,
2013), and better reduce audit fees (Krishnan & Wang, 2015). Taken together, the above studies suggest that managers with greater ability better manage their companies. Goodwill impairment is viewed as
negative news that signals declining ﬁrm performance (Hirschey &
Richardson, 2002). Hence, companies have incentives to prevent or reduce goodwill impairment losses (Li et al., 2011). Whether more-able
managers can better prevent or reduce goodwill impairment losses is
an interesting question that has not been examined previously. Based
on prior studies, I posit that more-able managers better prevent or
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
reduce goodwill impairment losses through more efﬁcient management than less-able managers.2
I ﬁrst identify a full sample including goodwill impairment ﬁrms and
no goodwill impairment ﬁrms from 2002 to 2011. The full sample is
restricted to ﬁrms that have goodwill.3 After controlling for managers’
opportunistic behavior, the regression analysis reveals a negative relationship between managerial ability and the likelihood of goodwill
impairment, suggesting that more-able mangers can better prevent
goodwill impairment than can less-able managers. Next, using the
goodwill impairment sample ﬁrms, the regression analysis documents
a negative relationship between managerial ability and the magnitude
of goodwill impairment losses, suggesting that more-able managers
can better reduce the magnitude of goodwill impairment losses than
less-able mangers. Next, I perform various additional tests including alternative sample periods, ﬁxed-effects regression, and two-stage OLS
regression analysis (2SLS) to address potential endogeneity issues.
These additional tests provide consistent results. Overall, the ﬁndings
support a negative relationship between managerial ability and goodwill impairment, suggesting that managers with greater ability better
prevent or reduce impairment losses. Last, I incorporate CEO tenure
into the regression analysis and ﬁnd that CEO tenure is negatively related to goodwill impairment losses. This ﬁnding is consistent with Beatty
and Weber (2006); Ramanna and Watts (2012). Furthermore, I ﬁnd
that capable CEOs with longer tenure better reduce the magnitude of
goodwill impairment losses.
This study makes several contributions. First, this study extends and
links two distinct research streams: managerial ability studies in management and goodwill literature in accounting. Speciﬁcally, this study
extends literature on the impact of managerial ability on various ﬁrm
characteristics and contributes to research regarding the determinants
of goodwill impairment, a major research stream in goodwill accounting (Li & Sloan, 2015). To the best of my knowledge, this is the ﬁrst
study that performs a direct empirical test on the relationship between
managerial ability and goodwill impairment. Second, this study contributes to the literature on goodwill impairment prediction models
(e.g., Hayn & Hughes, 2006) by examining managerial ability and the
likelihood of goodwill impairment. Although this study does not attempt to construct a prediction model for goodwill impairment, the
ﬁndings from this study may provide an avenue for future research on
goodwill impairment. The inclusion of managerial ability may help
users of ﬁnancial statements better assess the likelihood of goodwill
impairment. Third, Ramanna and Watts (2012); Li and Sloan (2015)
suggest that SFAS 142 provides managers with discretion in respect to
the timing and the magnitude of goodwill losses recognized. This
study complements the ﬁndings and associated interpretations in
Ramanna and Watts (2012); Li and Sloan (2015) by providing another
explanation. That is, it is possible that managers with greater ability better prevent or reduce goodwill impairment. Last, from a practical perspective, the results should interest policy makers who design and
implement guidelines on goodwill impairment decisions. Results
should also interest shareholders by showing the importance of moreable managers in preventing goodwill impairment and reducing the
magnitude of goodwill impairment losses after goodwill impairment
occurs.
The rest of this paper is organized as follows. Section 2 describes the
institutional background, while Section 3 presents literature review and
hypothesis development. Section 4 describes the research design, and
Section 5 presents the results of the empirical analyses. Section 6 presents the results of additional analyses, and Section 7 concludes this
study.
2
It is possible that mangers with greater ability better exploit the discretion by SFAS
142 to avoid/delay goodwill impairment or understate goodwill impairment losses. To
purge this possible explanation, I follow Ramanna and Watts (2012) by including variables
to control for managers’ opportunistic behavior in the regression analysis.
3
Firms have goodwill reported in at least one year during the sample period.
43
2. Background
Prior to 2001, goodwill accounting in the U.S. was governed by
Accounting Principles Board (APB) Opinion No. 16. Under APB 16, any
excess of purchase price over the fair value of the acquired ﬁrm’s net assets was recognized as goodwill. Goodwill was viewed as a depreciating
asset. The value of goodwill in a purchase acquisition was then amortized over a period of up to 40 years. To avoid the impact of goodwill
amortization expenses on earnings, many ﬁrms chose the pooling of interest acquisition method in which purchased goodwill was not recognized and amortized.
In June 2001, FASB issued Statement of Financial Accounting
Standards 142 (SFAS 142), Goodwill and Other Intangible Assets. SFAS
142 eliminated the pooling of interest acquisition method and required
all business acquisitions be accounted for by the purchase acquisition
method. In addition, SFAS 142 required sufﬁcient disclosure of the
allocation of the purchase price among the assets acquired. SFAS
142 required annual tests for goodwill and other intangible assets.
Speciﬁcally, it stated that goodwill should be tested for impairment
using a two-step process. In the ﬁrst step, companies compare the
carrying value of the reporting unit (including goodwill) to the estimated fair value of the reporting unit. If the carrying value of the
reporting unit is less than the estimated fair value of the reporting
unit, no impairment in goodwill exists. If the carrying value of the
reporting unit exceeds the estimated fair value of the reporting
unit, companies perform the second step: to determine and recognize the amount of goodwill impairment loss, which is recorded
against earnings. The impairment loss is measured as the difference
between the implied value and the carrying value of goodwill. In addition, any reversals of goodwill impairment losses are prohibited.
SFAS 142 also required ﬁrms to disclose the carrying value and any
changes in carrying value of goodwill. In 2011, FASB issued Accounting Standard Update 350 (ASU 350), which permits companies to
ﬁrst assess qualitative factors to determine whether it is more likely
than not that the fair value of a reporting unit is less than its carrying
value. Based on the assessment of qualitative factors, companies
then determine whether it is necessary to perform the goodwill impairment test. ASU 350 became effective for ﬁscal year beginning
after 12/15/2011.
3. Literature review and hypothesis development
3.1. Managerial ability
Upper echelons theory (Hambrick, 2007; Hambrick & Mason, 1984)
states that organizational outcomes are partially inﬂuenced by managers’ differing background characteristics. Bertrand and Schoar
(2003) ﬁnd that chief executive ofﬁcers (CEOs) have different managerial styles, and these styles inﬂuence a wide range of corporate decisions. Other similar studies investigate the relationship between chief
ﬁnancial ofﬁcer (CFO) expertise and restatements (Aier, Comprix,
Gunlock, & Lee, 2005), CEO reputation and earnings quality (Francis,
Nanda, & Olsson, 2008), managerial style and ﬁrm voluntary disclosure
(Bamber & Wang, 2010), managerial style and corporate tax avoidance
(Dyreng, Hanlon, & Maydew, 2010), and CFO style and accounting policies (Ge, Matsumoto, & Zhang, 2011). Taken together, this research
supports the important role of individual managers in accounting
choices and ﬁrm performance.
Demerjian et al. (2012) introduce a new measure of managerial
ability based on managers’ efﬁciency in generating revenues. They
argue that more-able managers “better understand technology and
industry trends, reliably predict product demand, invest in higher
value projects, and manage their employees more efﬁciently than
less-able managers” (page 1229). Demerjian et al. (2012) argue
that their measure (a comprehensive summary measure on managerial ability) outperforms existing managerial ability measures. In
44
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
addition, their measure is robust to CEO switch and is valued by the
market.
Using their managerial ability measure, Demerjian et al. (2013)
examine the relationship between managerial ability and earnings
quality. They ﬁnd that more-able managers are associated with
fewer subsequent restatements, higher earnings and accruals persistence, lower errors in the bad debt provision, and higher-quality
accrual estimations. Baik, Farber, and Lee (2011) ﬁnd a positive relationship between CEO ability and management earnings forecast issuance. Wang (2013) examines the informativeness of insider trades
conditional on managerial ability and ﬁnds that more-able managers
have greater net insider sales before the earnings break than do lessable mangers. Demerjian et al. (2015) ﬁnd that more-able managers
better smooth earnings to beneﬁt shareholders than do less-able
managers. Francis et al. (2014) ﬁnd a signiﬁcant negative relationship between managerial ability and corporate tax avoidance, suggesting that more-able managers engage in fewer tax-avoidance
activities, relative to less-able managers. Krishnan and Wang (2015)
ﬁnd negative relationships between managerial ability and both audit
fees and going concern options, suggesting that managerial ability
plays an important role in auditors’ decisions.
3.2. Goodwill impairment
Prior studies on goodwill impairment can be classiﬁed into two categories: The ﬁrst category examines the impact of goodwill impairment
on the stock market and on various ﬁrm characteristics. Prior studies
(e.g.Francis, Hanna, & Vincent, 1996; Henning & Shaw, 2003; Hirschey
& Richardson, 2002; Li et al., 2011; Xu, Anandarajana, & Curatolab,
2011) ﬁnd that goodwill impairment is value relevant to the market,
and normally investors view goodwill impairment as negative news.
For instance, Li et al. (2011) ﬁnd that investors react negatively to
goodwill impairment and conclude that goodwill impairment is a
leading indicator of a decline in future ﬁrm performance. Regarding
the impact of impairment on ﬁrm characteristics, Darrough et al.
(2014) examine the relationship between goodwill impairment
losses and CEO compensation and document that goodwill impairment losses lead to reduced CEO compensation. Sun and Zhang
(2016) ﬁnd a negative impact of goodwill impairment on bond credit
ratings.
The second category investigates the determinants of goodwill
impairment. Prior studies examine and ﬁnd that the cause of many
goodwill impairment losses is that the target ﬁrm is overpaid at the
time of acquisition (e.g.Beatty & Weber, 2006; Gu & Lev, 2011;
Hayn & Hughes, 2006; Li et al., 2011; Olante, 2013). Speciﬁcally,
Beatty and Weber (2006) examine a sample of ﬁrms that are likely
to have recorded a goodwill impairment loss and show that a ﬁrm’s
decision to accelerate or delay recognition of the loss is related to
managerial incentives. They ﬁnd evidence suggesting that ﬁrms are
less likely to accelerate recognition of goodwill impairment if they
have debt covenants affected by impairment, are listed on an exchange with delisting requirements, or have earnings-based bonus
plans, and more likely to accelerate recognition when they have a
CEO with a short tenure or a high earnings multiple. Olante (2013)
estimates that approximately 40% of goodwill impairment losses
are caused by overpayment at acquisition. Some studies investigate
whether goodwill impairment is associated with economic factors
at the ﬁrm level. For example, Chen, Kohlbeck, and Warﬁeld (2008)
and Chalmers, Godfrey, and Webster (2011) ﬁnd that goodwill impairments better reﬂect the underlying economics of goodwill after
the adoption of SFAS 142, supporting the FASB’s claim that SFAS
142 “will improve ﬁnancial reporting because the ﬁnancial statements of entities that acquire goodwill and other intangible assets
will better reﬂect the underlying economics of those assets” (SFAS
142, page 7). Other studies examine the role of managers’ opportunistic behavior in determining goodwill impairment. Ramanna and
Watts (2012) suggest that managers may avoid goodwill impairment under SFAS 142 when they have agency-based private information, because the current fair value of goodwill is a function of
management’s future actions such as ﬁrm strategy implementation.
They also ﬁnd a negative relationship between CEO tenure and goodwill impairment. Similarly, Li and Sloan (2015) argue that managers
exploit the discretion granted by SFAS 142 to delay goodwill
impairment.
3.3. Hypothesis development
Taken together, the literature review on managerial ability
suggests that more-able managers better manage their ﬁrms to maximize shareholders’ beneﬁts, relative to less-able managers. Goodwill impairment is viewed as negative news that signals declining
ﬁrm performance (Hirschey & Richardson, 2002). Hence, companies
have incentives to prevent or reduce goodwill impairment losses (Li
et al., 2011). If more-able managers better manage their ﬁrms to
maximize shareholders’ beneﬁts, I predict that managers with greater ability are more likely to effectively ﬁnd ways to prevent or reduce
goodwill impairment. Therefore, I expect a negative relationship between managerial ability and goodwill impairment. I propose the
following hypothesis:
H1. Managerial ability is negatively related to goodwill impairment.
4. Research design
4.1. Measurement of the primary independent variable—managerial ability
I use both the original managerial ability scores (MA) and decile
rankings (MARANK) developed by Demerjian et al. (2012) as proxies
for managerial ability in this study. Their managerial ability measure is
a performance-based measure of managers’ efﬁciency in using
ﬁrms’ resources to generate revenue. Demerjian et al. (2012) use a
two-step approach to develop their managerial ability measure.
First, they rely on Data Envelopment Analysis (DEA) to estimate
total ﬁrm efﬁciency by industry and year. Given a collection of points
in a multidimensional space, DEA ﬁts a piecewise linear envelope or
frontier to the given data. The envelope indicates a normative ideal
given the existing data. Points located on the envelope are optimally
efﬁcient, while points below the envelope are inefﬁcient. DEA evaluates all points with respect to their deviation from the frontier. The
values of the points on the frontier equal 1, and the values of other
points which operate beneath the frontier are between 0 and 1.
DEA requires identifying input and output variables. Demerjian
et al. (2012) use seven input variables: cost of goods sold; selling,
general and administrative expenses; property, plant and equipment; operating lease; research and development cost; goodwill; and
other intangibles. The output variable in Demerjian et al. (2012) is net
sales.
Demerjian et al. (2012) acknowledge that total ﬁrm efﬁciency can
be attributed to both manager-speciﬁc characteristics and ﬁrmspeciﬁc characteristics. Therefore, their second step is to identify
the manager-speciﬁc characteristics of the total ﬁrm efﬁciency
from DEA results. Thus, Demerjian et al. (2012) regress the total
ﬁrm efﬁciency on six ﬁrm-speciﬁc variables that could aid or hinder
managers’ ability. These six variables include ﬁrm size, ﬁrm market
share, cash available, ﬁrm age, operational complexity, and foreign
operations. This regression is run by industry and with year ﬁxed effects to purge industry and year effects. Demerjian et al. (2012) use
the residuals from the regression as proxy for managerial ability.
Demerjian et al. (2012) also transform the raw residual scores from
the above regression into an industry-based decile ranking for a
given year.
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
4.2. Empirical speciﬁcation
I use the following equations to test the inﬂuence of managerial ability on goodwill impairment:
GWI ¼ β0 þ β1 MA=MARANK þ β2 UNVA þ β3 DCOVPRO þ β4
LIST þ β5 APC þ β6 FOG þ β7 SIZE þ β8 ROA þ β9
LEV þ β10 MTB þ β11 GDW þ β12 WD þ β13 RC
þ β14 OSI þ ε:
ð1Þ
GWILOSS ¼ β0 þ β1 MA=MARANK þ β2 UNVA þ β3 DCOVPRO
þ β4 LIST þ β5 APC þ β6 FOG þ β7 SIZE þ β8
ROA þ β9 LEV þ β10 MTB þ β11 GDW þ β12
WD þ β13 RC þ β14 OSI þ ε:
ð2Þ
In Eq. (1), the dependent variable (GWI) captures the likelihood of
goodwill impairment. It is an indicator variable which takes 1 if the
ﬁrm-year observation has goodwill impairment loss and otherwise 0.
Hence, I use logistic regression. In Eq. (2), the dependent variable
(GWILOSS) measures the magnitude of goodwill impairment losses4
scaled by total assets. Consistent with prior studies (e.g., Li et al., 2011),
I use Tobit regression5 in Eq. (2) because the values of GWILOSS are between 0 and 1. All variables are deﬁned in Appendix 1. To test the hypothesis, I analyze the coefﬁcient β1 on MA and MARANK. If the hypothesis is
valid, I expect a negative and signiﬁcant coefﬁcient on managerial ability
(MA and MARANK).
Prior research (e.g., Ramanna & Watts, 2012) suggests that managers use their discretion opportunistically to delay/avoid goodwill impairment or understate the magnitude of goodwill impairment losses.
Hence, it is possible for more-able managers to better exploit the discretion opportunistically by SFAS 142, relative to less-able managers. This
can be a possible explanation for the negative relationship between
managerial ability and goodwill impairment. To purge this possible explanation, I follow Ramanna and Watts (2012) by including variables to
control for managers’ opportunistic behavior in the regression analysis.
Speciﬁcally, I ﬁrst use unveriﬁable net assets (UNVA) to control for
managers’ ﬂexibility in goodwill reporting. Ramanna and Watts
(2012) argue that ﬁrms with more unveriﬁable net assets have smaller
goodwill impairment losses. I calculate UNVA using the model in
Richardson, Sloan, Soliman, and Tuna (2005). Second, I use debt covenant probability (DCOVPRO) and whether a ﬁrm is listed on NASDAQ
or AMEX (LIST) to control for managers’ contracting motive. Ramanna
and Watts (2012) argue that the probability of debt covenant violation
is high for ﬁrms with 2 years of market-to-book ratio (MTB) b 1. Beatty
and Weber (2006) argue that ﬁrms listed on the NASDAQ or AMEX are
subject to goodwill-inclusive delisting requirements. Third, I use asset
pricing concerns (APC) to control for managers’ valuation motives in
goodwill impairment. Consistent with Beatty and Weber (2006), I calculate APC as the coefﬁcient from a regression of the ﬁrm’s quarterly
share price on its operating income using at least 16 quarters of data
prior to the ﬁrm-year. Last, I use the readability6 (FOG) of 10Ks to control for managers’ private information motive in goodwill impairment.
Ramanna and Watts (2012) argue that managers with positive private
information are less likely to engage in opportunistic behavior. As a result, such managers generate more readable 10Ks.
In addition to the variable of interest, I also control for factors associated with goodwill impairment losses established in prior literature. Gu
and Lev (2011) control market-to-book ratio, return on assets and
goodwill. Ramanna and Watts (2012) control leverage ratio. Darrough
4
Goodwill impairment loss (GDWLIP) is reported as a negative number in Compustat. I
multiply GDWLIP by −1.
5
Tobit regression does not compute an adjusted R squared. I obtain the adjusted R
squared from OLS regression.
6
Readability data (FOG Index) is provided by Dr. Feng Li. Li (2008) examines the relationship between readability of 10 K and earnings.
45
et al. (2014) suggest goodwill impairment is related to contemporaneous ﬁrm events such as long-term asset write-downs, restructuring
charges, and other special items. Following Gu and Lev (2011),
Ramanna and Watts (2012), and Darrough et al. (2014), I control for
ﬁrm size (SIZE), return on assets (ROA), leverage ratio (LEV), marketto-book ratio (MTB), goodwill (GDW), long-term assets write-downs
(WD), restructuring charges (RC), and other special items (OSI). I
winsorize the variables at level 1% and 99% and control for year and industry ﬁxed effects (Fama and French 48 industries) in the regression
analysis. See Appendix 1 for variable deﬁnitions.
4.3. Sample selection and descriptive statistics
I use 2002 as the initial testing year because SFAS 142 became effective in 2002. Consistent with Li and Sloan (2015), I end my sample in
2011 as Accounting Standards Update (ASU) 350–20 became effective
after 12/15/2011. I begin the sample selection process by using the
managerial ability scores and ranks by Demerjian et al. (2012). There
are 53,766 ﬁrm-year observations from 2002 to 2011. Next, I use
Compustat to obtain ﬁnancial statement data, which includes total assets (AT, #6), book value of equity (CEQ, #60), cash (CHE, #1), common
stock shares (CSHO, #25), debt in current liabilities (DLC, #34), longterm debt (DLTT, #9), goodwill (GDWL, #204), goodwill impairment
loss (GDWLIP, #368), investments and advances (IVAO, #32), shortterm investments (IVST, #193), total liabilities (LT, #181), net income
(NI, #172), stock price at ﬁscal year end (PRCC_F, #24), preferred
stock (PSTK, #130), restructuring costs (RCP, #376), sales (SALE, #12),
special items (SPI, #17), and long-term assets write-downs (WDP,
#380). The initial sample from Compustat including the above 18 variables consists of 110,991 observations from 2002 to 2011. I merge the
above two samples. Some observations are lost due to missing observations in Compustat. Next, I remove observations that do not have goodwill. The ﬁnal sample with complete data consists of 30,426 ﬁrm-year
observations, of which 4576 observations are ﬁrm-years with goodwill
impairment losses (the goodwill impairment sample) and 25,850 observations are ﬁrm-years without goodwill impairment losses (the no
goodwill impairment sample).
Panel A of Table 1 reports the distribution of ﬁrm-year observations
by year for the goodwill impairment sample ﬁrms and no goodwill impairment sample ﬁrms. For goodwill impairment sample ﬁrms, there
are 599 ﬁrm-year observations in 2002 and 415 observations in 2011.
2008 has the largest number of observations (959). This is consistent
with Darrough et al. (2014), who also ﬁnd that 2008 has the largest
number of goodwill impairments. For no goodwill impairment sample
ﬁrms, there are 2678 ﬁrm-year observations in 2002 and 2234 observations in 2011. 2004 has the largest number of observations (2953).
Panel B of Table 1 reports the distribution of ﬁrm-year observations by
industry for the top 10 industries. For the goodwill impairment sample,
the most heavily represented industry is business services (16.30%, SIC
73), followed by electric equipment (11.01%, SIC 36) and communications (9.27%, SIC 48). For the no goodwill impairment sample, the
most heavily represented industry is business services (17.78%, SIC
73), followed by electric equipment (8.80%, SIC 36) and chemical
(8.43%, SIC 28).
Table 2 presents descriptive statistics for the full sample partitioned
based on goodwill impairment (obs. = 4576) and no goodwill impairment losses observations (obs. = 25,850). Speciﬁcally, Table 2 reports
the mean, standard deviation, median, 25th percentile and 75th percentile of the following variables: GWILOSS, MA, MARANK, UNVA,
DCOVPRO, LIST, APC, FOG, SIZE, ROA, LEV, MTB, GDW, WD, RC, and
OSI for both subsamples (the goodwill impairment sample vs. the no
goodwill impairment sample). For the goodwill impairment sample
(no goodwill impairment sample), the mean values of MA and
MARANK are − 0.035 (− 0.001) and 0.493 (0.557), respectively. The
no goodwill impairment sample ﬁrms have higher managerial ability
relative to the goodwill impairment sample ﬁrms. For the goodwill
46
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
Table 1
Distribution of ﬁrm-year observations.
Panel A: Distribution of ﬁrm-year observations by year
Goodwill impairment
No goodwill impairment
Year
Obs.
% of sample
Cumulative %
Obs.
% of sample
Cumulative %
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
599
349
326
356
324
371
959
575
302
415
4576
13.09%
7.63%
7.12%
7.78%
7.08%
8.11%
20.96%
12.57%
6.60%
9.07%
100.00%
13.09%
20.72%
27.84%
35.62%
42.70%
50.81%
71.77%
84.33%
90.93%
100.00%
2678
2856
2953
2897
2883
2757
2061
2161
2370
2234
25,850
10.36%
11.05%
11.42%
11.21%
11.15%
10.67%
7.97%
8.36%
9.17%
8.64%
100.00%
10.36%
21.41%
32.83%
44.04%
55.19%
65.86%
73.83%
82.19%
91.36%
100.00%
Panel B: Distribution of ﬁrm-year observations by industry: top 10 industries
Goodwill impairment
No goodwill impairment
2 SIC
Industry description
Obs.
% of sample
2 SIC
Industry description
Obs.
% of sample
73
36
48
28
35
38
37
20
13
50
Business services
Electronic equipment
Communications
Chemicals products
Industrial machinery
Measuring instruments
Transportation equipment
Food products
Oil and gas extraction
Wholesale durable goods
746
504
424
275
272
198
143
118
105
98
16.30%
11.01%
9.27%
6.01%
5.94%
4.33%
3.13%
2.58%
2.29%
2.14%
73
36
28
38
35
48
13
20
80
37
Business services
Electronic equipment
Chemicals products
Measuring instruments
Industrial machinery
Communications
Oil & gas extraction
Food products
Health services
Transportation equipment
4442
2276
2178
1892
1679
1340
746
737
668
655
17.18%
8.80%
8.43%
7.32%
6.50%
5.18%
2.89%
2.85%
2.58%
2.53%
are signiﬁcant (p-value b 0.0001). For example, average managerial
ability (MA/MARANK) of the no goodwill impairment sample ﬁrms is
signiﬁcantly higher than the average managerial ability of the goodwill
impairment sample ﬁrms, suggesting that more-able managers better
prevent goodwill impairment.
Panel A (B) of Table 3 provides the correlation matrices for selected
variables for the full sample (the goodwill impairment sample). For
each pair of variables, the Spearman correlation coefﬁcients and related
p-values are provided. I use Spearman correlation in this study because
of the discrete nature of the variables such as GWI and MARANK. Panel
A of Table 3 reports a signiﬁcant and negative (p-value b 0.0001) relationship between GWI and managerial ability (MA and MARANK).
The negative association suggests that more-able managers better
prevent goodwill impairment. Using the goodwill impairment
impairment sample, the mean and median values of goodwill impairment losses (GWILOSS) are 0.076 and 0.035, respectively. The mean
and median values of unveriﬁable net assets (UNVA) are 0.245 (0.168)
and 0.079 (0.042) in the goodwill impairment sample (no goodwill impairment sample), suggesting that goodwill impairment sample ﬁrms
have more unveriﬁable net assets. The mean value of 10 K readability
(FOG) is 0.860 (0.812) in the goodwill impairment sample (no goodwill
impairment sample), suggesting that 10Ks of goodwill impairment
sample ﬁrms are less readable. The median value of ROA in the goodwill
impairment sample (no goodwill impairment sample) is − 0.076
(0.045), suggesting that goodwill impairment sample ﬁrms demonstrate worse accounting performance. Using a t-test, I also test the signiﬁcance of the differences in means of the variables. For all variables
in Table 2, (two-tailed) p-values suggest that the differences in means
Table 2
Descriptive statistics goodwill impairment vs. no goodwill impairment.
Goodwill impairment
No goodwill impairment
Variable
Obs.
Mean
25P
Median
75P
Obs.
Mean
25P
Median
75P
Difference in means
GWILOSS
MA
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
4576
0.076
−0.035
0.493
0.245
0.429
0.384
15.505
0.860
6.133
−0.196
0.203
1.552
0.137
−0.012
−0.007
−0.253
0.006
−0.126
0.300
−0.503
0.000
0.000
−3.456
1.000
4.367
−0.266
0.004
0.593
0.000
−0.006
−0.006
−0.333
0.035
−0.044
0.500
0.079
0.000
0.000
9.726
1.000
6.166
−0.076
0.141
1.149
0.083
0.000
0.000
−0.175
0.124
0.047
0.700
0.723
1.000
1.000
24.883
1.000
7.931
0.020
0.314
2.064
0.225
0.000
0.000
−0.062
–
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
25,850
–
−0.001
0.557
0.168
0.174
0.452
12.428
0.812
6.281
0.006
0.177
2.702
0.164
−0.003
−0.003
−0.219
–
−0.093
0.300
−0.457
0.000
0.000
−7.223
1.000
4.795
−0.008
0.002
1.241
0.040
0.000
−0.001
−0.289
–
−0.013
0.600
0.042
0.000
0.000
3.583
1.000
6.291
0.045
0.128
2.015
0.117
0.000
0.000
−0.130
–
0.079
0.800
0.546
0.000
1.000
26.677
1.000
7.768
0.093
0.278
3.360
0.250
0.000
0.000
−0.044
–
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
b0.0001
This table reports the descriptive statistics for goodwill impairment and no goodwill impairment samples over the period of 2002–2011. The goodwill impairment sample consists of 4576
ﬁrm-year observations and the no goodwill impairment sample consists of 25,850 ﬁrm-year observations. Two-tailed p-values are provided in the last column for the difference in means
test. Refer to Appendix 1 for variable deﬁnitions.
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
47
Table 3
Correlations among selected variables.
Panel A: Full sample (Obs. = 30,426)
GWI
MA
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
MA
−0.086
p-value
b.0001
MARANK
−0.082 0.950
p-value
b.0001
b.0001
UNVA
0.017
0.004
−0.013
p-value
0.003
0.533
0.029
DCOVPRO 0.223
−0.108 −0.110
−0.047
p-value
b.0001
b.0001
b.0001
b.0001
LIST
−0.048 −0.045 −0.040
−0.270 −0.068
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
APC
0.273
0.260
0.236
0.099
−0.326
−0.204
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
FOG
−0.044 −0.636 −0.670
0.040
0.063
0.003
−0.121
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.575
b.0001
SIZE
−0.018 0.090
0.066
0.261
−0.186
−0.297 0.588
−0.022
p-value
0.002
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
ROA
−0.309 0.322
0.302
−0.022 −0.338
−0.012 0.772
−0.195 0.327
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.040
b.0001
b.0001
b.0001
LEV
0.031
−0.040 −0.050
0.593
0.095
−0.261 0.082
0.049
0.362
−0.063
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
MTB
−0.213 0.147
0.156
0.019
−0.708
0.060
0.309
−0.112 0.124
0.363
−0.079
p-value
b.0001
b.0001
b.0001
0.001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
GDW
−0.115 0.023
0.030
0.152
−0.088
−0.008 0.055
−0.008 0.034
0.006
0.110
0.058
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.140
b.0001
0.150
b.0001
0.277
b.0001
b.0001
WD
−0.229 0.089
0.079
−0.010 −0.094
0.026
0.151
−0.043 −0.030 0.198
−0.033 0.096
0.042
p-value
b.0001
b.0001
b.0001
0.079
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
RC
−0.117 0.086
0.074
−0.016 −0.021
0.041
0.066
−0.072 −0.204 0.145
−0.073 0.055
−0.060 0.143
p-value
b.0001
b.0001
b.0001
0.005
0.000
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
OSI
−0.061 −0.037 −0.042
−0.128 0.040
0.004
0.032
0.031
−0.003 0.058
−0.098 −0.032 −0.857 −0.004 0.060
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.541
b.0001
b.0001
0.559
b.0001
b.0001
b.0001
b.0001
0.443
b.0001
This table presents the Spearman correlations based on the full sample of 30,426 ﬁrm-year observations over the period of 2002–2011. Two-tailed p-values are provided. Refer
to Appendix 1 for variable deﬁnitions.
Panel B: Goodwill impairment sample (Obs. = 4576)
GWILOSS
MA
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
MA
−0.038
p-value
0.001
MARANK
−0.032
0.949
p-value
0.003
b.0001
UNVA
−0.134
0.054
0.028
p-value
b.0001
0.000
0.059
DCOVPRO 0.228
−0.066 −0.083
−0.135
p-value
b.0001
b.0001
b.0001
b.0001
LIST
−0.085
−0.029 −0.031
−0.168 −0.025
p-value
b.0001
0.047
0.036
b.0001
0.096
APC
0.524
0.199
0.190
0.030
−0.249
−0.102
p-value
b.0001
b.0001
b.0001
0.043
b.0001
b.0001
FOG
−0.003
−0.584 −0.605
0.004
0.035
0.002
−0.085
p-value
0.826
b.0001
b.0001
0.784
0.018
0.886
b.0001
SIZE
−0.421
0.072
0.057
0.303
−0.244
−0.164 0.075
−0.012
p-value
b.0001
b.0001
0.000
b.0001
b.0001
b.0001
b.0001
0.422
ROA
−0.698
0.216
0.194
0.186
−0.313
−0.045 0.672
−0.088 0.560
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.003
b.0001
b.0001
b.0001
LEV
−0.085
0.005
−0.002
0.481
−0.009
−0.187 −0.071 0.018
0.401
0.138
p-value
b.0001
0.733
0.885
b.0001
0.563
b.0001
b.0001
0.235
b.0001
b.0001
MTB
−0.260
0.081
0.094
0.195
−0.857
0.027
0.277
−0.035 0.252
0.349
0.002
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.067
b.0001
0.019
b.0001
b.0001
0.904
GDW
−0.048
0.104
0.114
0.210
−0.236
−0.036 0.122
−0.062 0.251
0.199
0.178
0.231
p-value
0.001
b.0001
b.0001
b.0001
b.0001
0.015
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
WD
−0.137
0.094
0.085
0.059
−0.056
−0.036 0.197
−0.034 0.075
0.210
0.020
0.057
0.061
p-value
b.0001
b.0001
b.0001
b.0001
0.000
0.016
b.0001
0.021
b.0001
b.0001
0.173
0.000
b.0001
RC
−0.045
0.075
0.063
0.037
0.040
−0.062 0.245
−0.045 −0.201 0.074
−0.020 −0.029 −0.028 0.085
p-value
0.002
b.0001
b.0001
0.013
0.007
b.0001
b.0001
0.002
b.0001
b.0001
0.178
0.048
0.059
b.0001
OSI
−0.502
−0.094 −0.106
−0.073 0.067
−0.003 0.130
0.090
0.040
0.269
−0.079 −0.044 −0.661 0.003
0.022
p-value
b.0001
b.0001
b.0001
b.0001
b.0001
0.831
b.0001
b.0001
0.007
b.0001
b.0001
0.003
b.0001
0.834
0.133
This table reports the Spearman correlations based on the goodwill impairment sample of 4576 ﬁrm-year observations over the period of 2002–2011. Two-tailed p-values are
provided. Refer to Appendix 1 for variable deﬁnitions.
sample (obs. = 4576), Panel B of Table 3 reports a signiﬁcant and
negative relationship between GWILOSS and managerial ability
(MA and MARANK). The negative association suggests that more-
able managers can better reduce the magnitude of goodwill impairment losses after goodwill impairment occurs. Overall, results in
Table 3 lend support to the hypothesis.
48
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
5. Results
Using the full sample (obs. = 30,426), Table 4 reports the Logistic regression results testing the hypothesis. The coefﬁcient on MA is −0.847
(p-value b 0.0001) and on MARANK is −0.409 (p-value b 0.0001). The
negative and signiﬁcant coefﬁcients support the hypothesis that managerial ability is negatively related to the likelihood of goodwill impairment. This evidence suggests that more-able managers can better
prevent goodwill impairment relative to less-able managers. For the
control variables, GWI is signiﬁcantly and positively associated with
UNVA, DCOVPRO, APC, and LEV, but negatively associated with LIST,
SIZE, ROA, MTB, GDW, WD, RC, and OSI.
Using the goodwill impairment sample (obs. = 4576), Table 5 reports the Tobit regression results7 testing the hypothesis. The coefﬁcient
on MA is − 0.020 (p-value = 0.009) and on MARANK is − 0.008
(p-value = 0.022). The negative and signiﬁcant coefﬁcients support
the hypothesis that managerial ability is negatively related to goodwill
impairment losses, suggesting that more-able managers can better
reduce the magnitude of goodwill impairment losses, relative to
less-able managers. For the control variables, GWILOSS is signiﬁcantly and positively associated with APC, but negatively associated with
LIST, FOG, SIZE, ROA, LEV, MTB, GDW, RC, and OSI. The signiﬁcantly
positive relationship between GWILOSS and APC and the signiﬁcantly negative relationships between GWILOSS and both LIST and FOG
are consistent with the ﬁndings in Ramanna and Watts (2012). For
example, Ramanna and Watts (2012) ﬁnd that ﬁrms listed on
NASDAQ or AMEX report smaller goodwill impairment losses. The
signiﬁcantly negative relationships between GWI and both RC and
OSI are consistent with the ﬁndings in Darrough et al. (2014). For example, Darrough et al. (2014) ﬁnd that contemporaneous ﬁrm
events such as long-term assets write-downs (WD), restructuring
charges (RC), and other special items (OSI) result in smaller goodwill
impairment losses.
6. Additional tests
Table 4
Managerial ability and likelihood of goodwill impairment.*, **
Model: GWI = f (MA/MARANK; control variables).
Variable
Estimate
Pr N ChiSq
Variable
Estimate
Pr N ChiSq
Intercept
MA
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Pseudo R2
−2.051
−0.847⁎⁎⁎
0.049⁎⁎⁎
0.812⁎⁎⁎
−0.117⁎⁎⁎
0.001⁎⁎⁎
b.0001
b.0001
b.0001
b.0001
0.003
b.0001
0.339
b.0001
b.0001
0.004
0.001
b.0001
b.0001
b.0001
0.001
Intercept
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Pseudo R2
−1.836
−0.409⁎⁎⁎
0.048⁎⁎⁎
0.808⁎⁎⁎
−0.117⁎⁎⁎
0.001⁎⁎⁎
b.0001
b.0001
b.0001
b.0001
0.003
b.0001
0.466
b.0001
b.0001
0.004
0.001
b.0001
b.0001
b.0001
0.001
−0.064
−0.114⁎⁎⁎
−1.406⁎⁎⁎
0.262⁎⁎⁎
−0.016⁎⁎⁎
−1.660⁎⁎⁎
−13.298⁎⁎⁎
−13.475⁎⁎⁎
−0.269⁎⁎⁎
Included
Included
30,426
0.2069
−0.048
−0.113⁎⁎⁎
−1.409⁎⁎⁎
0.260⁎⁎⁎
−0.016⁎⁎⁎
−1.650⁎⁎⁎
−13.329⁎⁎⁎
−13.493⁎⁎⁎
−0.262⁎⁎⁎
Included
Included
30,426
0.2068
This table presents the results of logistic regressions with industry and year effects based
on the full sample, including goodwill impairment sample and no goodwill impairment
sample, over the period of 2002–2011. The dependent variable (GWI), capturing the likelihood of goodwill impairment losses, takes a value of one if the ﬁrm-year observation has
a goodwill impairment loss and zero otherwise. The industry-speciﬁc and year-speciﬁc intercepts are omitted for brevity. Continuous control variables are winsorized at 1% and 99%
percentiles each year before entering the regression tests. Refer to Appendix 1 for variable
deﬁnitions.
⁎ Signiﬁcance at the 10% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 5% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence levels.
which removes the cross-sectional variation and analyzes only the variation over time within a ﬁrm. Because industry dummies are timeinvariant, I exclude them in the ﬁxed-effects regression (Jiraporn,
Jiraporn, Boeprasert, & Chang, 2014).
6.1. Alternative sample periods
Due to the ﬁnancial crisis, I use two alternative testing periods:
pre-2008 vs. post-2008. This test examines the extent to which
changes in ﬁrm level and macroeconomic risk factors affect the relationship between managerial ability and goodwill impairment.
Table 6 reports the regression results testing the hypothesis for
both periods. In the pre-2008 period, the coefﬁcient on MA is
− 0.023 (p-value = 0.048) and on MARANK is − 0.008 (p-value =
0.054). In the post-2008 period, the coefﬁcient on MA is − 0.022
(p-value = 0.030) and on MARANK is − 0.010 (p-value = 0.032).
The negative and signiﬁcant coefﬁcients support the hypothesis
that managerial ability is negatively related to goodwill impairment
losses, consistent with my earlier ﬁndings.
6.2. Fixed-effects regression analysis
Although I control for several variables that are possibly related to
managerial ability and/or goodwill impairment losses, this procedure
may not effectively address the omitted-variable bias induced by unknown ﬁrm characteristics. For example, some unknown variable may
affect managerial ability and goodwill impairment simultaneously. To
mitigate the omitted-variable concern, I use ﬁxed-effects regression,
7
I also use clustered standard errors regression and obtain similar results. Petersen
(2009) states that the residuals of a given ﬁrm may be correlated across years (ﬁrm effect)
and the residuals of a given year may be correlated across different ﬁrms (i.e., time effect)
in studies using panel data sets. To better control for the ﬁrm and time effects, the author
suggests the use of clustered standard errors regression.
Table 5
Managerial ability and magnitude of goodwill impairment.*
Model: GWILOSS = f (MA/MARANK; control variables).
Variable
Estimate
Pr N ChiSq
Variable
Estimate
Pr N ChiSq
Intercept
MA
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.019
−0.020⁎⁎⁎
−0.000
0.001
−0.004⁎⁎
0.000⁎⁎⁎
−0.007⁎⁎
−0.002⁎⁎⁎
−0.110⁎⁎⁎
−0.014⁎⁎⁎
−0.001⁎⁎⁎
−0.283⁎⁎⁎
0.001
0.009
0.368
0.551
0.015
b.0001
0.014
b.0001
b.0001
0.000
0.000
b.0001
0.414
0.042
b.0001
Intercept
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.016
−0.008⁎⁎
−0.000
0.001
−0.004⁎⁎
0.000⁎⁎⁎
−0.006⁎⁎
−0.002⁎⁎⁎
−0.110⁎⁎⁎
−0.014⁎⁎⁎
−0.001⁎⁎⁎
−0.283⁎⁎⁎
0.015
0.022
0.366
0.524
0.015
b.0001
0.031
b.0001
b.0001
0.000
0.000
b.0001
0.417
0.039
b.0001
−0.037
−0.177⁎⁎
−0.389⁎⁎⁎
Included
Included
4576
0.7029
−0.037
−0.179⁎⁎
−0.389⁎⁎⁎
Included
Included
4576
0.7028
This table presents the results of Tobit regressions with industry and year effects based on
the goodwill impairment sample over the period of 2002–2011. The dependent variable
(GWILOSS) measures the magnitude of goodwill impairment losses. The industry-speciﬁc
and year-speciﬁc intercepts are omitted for brevity. Continuous control variables are
winsorized at 1% and 99% percentiles each year before entering the regression tests.
Refer to Appendix 1 for variable deﬁnitions.
⁎ Signiﬁcance at the 10% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 5% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence levels.
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
49
Table 6
Managerial ability and magnitude of goodwill impairment alternative sample period.
Model: GWILOSS = f (MA/MARANK; control variables).
2002–2007
2008–2011
Variable
Estimate
Pr N ChiSq
Variable
Estimate
Pr N ChiSq
Variable
Estimate
Pr N ChiSq
Variable
Estimate
Pr N ChiSq
Intercept
MA
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.030
−0.023⁎⁎
−0.000
0.000
−0.002
0.000⁎⁎⁎
b.0001
0.048
0.678
0.930
0.416
0.002
0.184
b.0001
b.0001
0.097
b.0001
b.0001
0.255
0.277
b.0001
Intercept
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.027
−0.008⁎
−0.000
0.000
−0.002
0.000⁎⁎⁎
0.002
0.054
0.688
0.922
0.416
0.002
0.315
b.0001
b.0001
0.099
b.0001
b.0001
0.255
0.283
b.0001
Intercept
MA
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.008
−0.022⁎⁎
−0.001
0.007⁎⁎⁎
−0.006⁎⁎⁎
0.000⁎⁎⁎
−0.009⁎⁎
−0.002⁎⁎
−0.108⁎⁎⁎
−0.018⁎⁎⁎
0.302
0.030
0.251
0.002
0.004
0.002
0.015
0.010
b.0001
0.000
0.606
b.0001
0.883
0.011
b.0001
Intercept
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.003
−0.010⁎⁎
−0.001
0.007⁎⁎⁎
−0.006⁎⁎⁎
0.000⁎⁎⁎
−0.009⁎⁎
−0.002⁎⁎
−0.108⁎⁎⁎
−0.018⁎⁎⁎
0.769
0.032
0.231
0.002
0.004
0.002
0.019
0.012
b.0001
0.000
0.603
b.0001
0.892
0.012
b.0001
−0.006
−0.003⁎⁎⁎
−0.116⁎⁎⁎
−0.009⁎
−0.002⁎⁎⁎
−0.235⁎⁎⁎
−0.075
−0.129
−0.345⁎⁎⁎
Included
Included
2325
0.6799
−0.004
−0.003⁎⁎⁎
−0.115⁎⁎⁎
−0.009⁎
−0.002⁎⁎⁎
−0.236⁎⁎⁎
−0.075
−0.128
−0.346⁎⁎⁎
Included
Included
2325
0.6798
−0.000
−0.356⁎⁎⁎
−0.009
−0.323⁎⁎
−0.456⁎⁎⁎
Included
Included
2251
0.7288
−0.000
−0.356⁎⁎⁎
−0.009
−0.323⁎⁎
−0.456⁎⁎⁎
Included
Included
2251
0.7288
This table presents the results of Tobit regressions with industry and year effects based on the goodwill impairment sample for two alternative periods: pre-2008 and post-2008. The dependent variable (GWILOSS) measures the magnitude of goodwill impairment losses. The industry-speciﬁc and year-speciﬁc intercepts are omitted for brevity. Continuous control variables are winsorized at 1% and 99% percentiles each year before entering the regression tests. Refer to Appendix 1 for variable deﬁnitions.
⁎ Signiﬁcance at the 10% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 5% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence level.
Table 7 reports that the coefﬁcient on MA is − 0.029 (p-value =
0.063) and on MARANK is −0.017 (p-value = 0.014). The ﬁxed-effect
regression suggests that, within ﬁrms, managerial ability is negatively
related to goodwill impairment losses. Because the ﬁxed-effects result
is consistent with the primary result by Tobit regression, it does not appear that the conclusion is affected by endogeneity due to omittedvariable bias.
Table 7
Managerial ability and likelihood of goodwill impairment ﬁxed effects regression.
Model: GWILOSS = f (MA/MARANK; control variables).
Variable
Estimate
MA
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
−0.029⁎
−0.002⁎⁎
0.003
−0.005⁎⁎⁎
0.000⁎⁎⁎
−0.009⁎
−0.004
−0.144⁎⁎⁎
−0.002
−0.001⁎⁎⁎
−0.316⁎⁎⁎
−0.105
0.304⁎⁎
−0.352⁎⁎⁎
Not included
Included
4576
0.7032
Pr N |t|
0.063
0.015
0.270
0.002
0.002
0.054
0.189
b.0001
0.837
0.004
b.0001
0.168
0.047
b.0001
Variable
Estimate
Pr N |t|
MARANK
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
−0.017⁎⁎
−0.002⁎⁎
0.014
0.013
0.237
0.048
0.002
0.031
0.213
b.0001
0.829
0.004
b.0001
0.166
0.050
b.0001
0.004
−0.005⁎⁎
0.000⁎⁎⁎
−0.010⁎⁎
−0.004
−0.145⁎⁎⁎
−0.002
−0.001⁎⁎⁎
−0.316⁎⁎⁎
−0.106
0.301⁎
−0.351⁎⁎⁎
Not included
Included
4576
0.7031
This table presents the results of ﬁxed effects regressions with year effect based on the
goodwill impairment sample over the period of 2002–2011. Company identiﬁer used in
ﬁxed effects regression is GVKEY. The dependent variable (GWILOSS) measures the magnitude of goodwill impairment losses. The year-speciﬁc intercepts are omitted for brevity.
Continuous control variables are winsorized at 1% and 99% percentiles each year before
entering the regression tests. Refer to Appendix 1 for variable deﬁnitions.
⁎ Signiﬁcance at the 10% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 5% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence level.
6.3. Two-stage OLS regression analysis (2SLS)
I explore the possibility of a reverse causality (self-selection) issue.
For example, ﬁrms with large goodwill impairment losses are perhaps
more likely to seek more-able managers. Following Jiraporn et al.
(2014), I perform a two-stage OLS regression analysis, which controls
for possible reverse causality. Two-stage regression analysis requires
identifying an instrumental variable (IV) which is highly correlated to
a ﬁrm’s managerial ability but does not inﬂuence ﬁrm performance except through managerial ability. Consistent with Jiraporn et al. (2014), I
use the average managerial ability performance of the ﬁrms in the same
industry (ﬁrst 2 SIC code). This variable is clearly related to the managerial ability of a given ﬁrm, but it does not relate to the goodwill impairment losses of a given ﬁrm. In the ﬁrst stage of 2SLS, I estimate
managerial ability score (MA) and rank (MARANK) using the average
score (MA) and rank (MARANK) of the ﬁrms in the same industry. I
include all of the control variables, as well as the industry and year
dummy variables. In the second stage of 2SLS, I use the instrumented
values of MA and MARANK from the ﬁrst stage and include them as independent variables in the second-stage regression. I use the same control variables in the second-stage regression.
Table 8 reports the 2SLS results for testing the hypothesis. For the relationship between MA and GWILOSS, the ﬁrst-stage regression reports
the average MA is positively related (0.479) to individual MA at a significant level (p-value b 0.0001). The second stage reports that the coefﬁcient of the instrumented MA is negative (− 0.020) and highly
signiﬁcant (p-value = 0.009), suggesting that managers with greater
ability better reduce goodwill impairment losses. For the relationship
between MARANK and GWILOSS, the ﬁrst-stage regression reports the
average MARANK is positively related (0.429) to individual MARANK
at a signiﬁcant level (p-value b 0.0001). The second stage reports the coefﬁcient of the instrumented MARANK is negative (−0.008) and significant (p-value = 0.023), suggesting that more-able managers better
reduce the magnitude of goodwill impairment losses, relative to lessable managers. Overall, the two-stage OLS regression analysis (2SLS)
lends support to the main results.
50
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
Table 8
Managerial ability and likelihood of goodwill impairment two-stage regression analysis
(2SLS).
Stage 1: MA/MARANK = f (average MA/MARANK; control variables).
Stage 2: GWILOSS = f (instrumented MA/MARANK; control variables).
Intercept
p-value
Average MA
p-value
MA (instrumented)
p-value
Average MARANK
p-value
MARANK (instrumented)
p-value
UNVA
p-value
DCOVRPO
p-value
LIST
p-value
APC
p-value
FOG
p-value
SIZE
p-value
ROA
p-value
LEV
p-value
MTB
p-value
GDW
p-value
WD
p-value
RC
p-value
OSI
p-value
Industry
Year
Obs.
Adj. R2
MA
GWILOSS
MARANK
GWILOSS
Stage 1
Stage 2
Stage 1
Stage 2
0.226
b.0001
0.479⁎⁎⁎
0.019
0.001
0.723
b.0001
0.016
0.016
b.0001
−0.001
0.132
−0.002
0.642
−0.012⁎⁎⁎
0.000
0.000
0.641
−0.255⁎⁎⁎
b.0001
−0.003⁎⁎⁎
0.001
0.110⁎⁎⁎
b.0001
−0.001
0.901
0.001⁎
0.055
−0.053⁎⁎⁎
0.000
0.096
0.271
0.065
0.699
−0.119⁎⁎⁎
b.0001
Included
Included
4576
0.5460
−0.020⁎⁎⁎
0.009
−0.000
0.372
0.001
0.554
−0.004⁎⁎
0.016
0.000⁎⁎⁎
b.0001
−0.007⁎⁎
0.015
−0.002⁎⁎⁎
b.0001
−0.110⁎⁎⁎
b.0001
−0.014⁎⁎⁎
0.000
−0.001⁎⁎⁎
0.000
−0.282⁎⁎⁎
b.0001
−0.037
0.417
−0.177⁎⁎
0.043
−0.389⁎⁎⁎
b.0001
Included
Included
4576
0.7029
0.429⁎⁎⁎
b.0001
−0.003⁎
0.057
−0.014⁎
0.063
−0.028⁎⁎⁎
b.0001
0.000
0.230
−0.492⁎⁎⁎
b.0001
−0.007⁎⁎⁎
b.0001
0.228⁎⁎⁎
b.0001
−0.006
0.696
0.002⁎⁎
0.044
−0.105⁎⁎⁎
0.001
0.295
0.126
0.398
0.281
−0.247⁎⁎⁎
b.0001
Included
Included
4576
0.4657
−0.008⁎⁎
0.023
−0.000
0.370
0.001
0.527
−0.004⁎⁎
0.016
0.000⁎⁎⁎
b.0001
−0.006⁎⁎
0.033
−0.002⁎⁎⁎
b.0001
−0.110⁎⁎⁎
b.0001
−0.014⁎⁎⁎
0.000
−0.001⁎⁎⁎
0.000
−0.283⁎⁎⁎
b.0001
−0.037
0.420
−0.179⁎
0.041
−0.389⁎⁎⁎
b.0001
Included
Included
4576
0.7027
The table presents the results of two-stage OLS regression analysis (2SLS) with industry
and year effects based on the goodwill impairment sample. In the ﬁrst stage of 2SLS, I estimate managerial ability score (MA) and rank (MARANK) using the average score (MA)
and rank (MARANK) of the ﬁrms in the same industry. I include all of the control variables,
as well as the industry and year dummy variables. In the second stage of 2SLS, I use the instrumented values of MA and MARANK from the ﬁrst stage and include them as independent variables in the second-stage regression. I use the same control variables in
the second-stage regression. The above procedures are applied in previous studies
(e.g., Jiraporn et al., 2014). The industry-speciﬁc and year-speciﬁc intercepts are omitted
for brevity. Continuous control variables are winsorized at 1% and 99% percentiles each
year before entering the regression tests. Refer to Appendix 1 for variable deﬁnitions.
⁎ Signiﬁcance at the 10% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 5% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence levels.
6.4. Managerial ability, CEO tenure, and goodwill impairment
Motivated by Beatty and Weber (2006) and Ramanna and Watts
(2012), I incorporate CEO tenure into Eq. (2)8 and ﬁnd a signiﬁcant
(p-value = 0.010) and negative (− 0.001) relationship between CEO
tenure and the magnitude of goodwill impairment losses, suggesting
that CEOs with longer tenure better reduce the magnitude of goodwill
8
Following Ge et al. (2011), I hand collected CEO tenure data from various sources such
as SEC’s Edgar database, company websites, and internet search. I managed to collect CEO
tenure data for 1390 ﬁrm-year observations. Hence, the sample consists of 1390 ﬁrm-year
observations.
Table 9
Managerial ability, CEO tenure, and magnitude of goodwill impairment.
Model: GWILOSS = f (MARANK, TENURE, MARANK × TENURE; control variables).
Variable
Estimate
Pr N ChiSq
Intercept
MARANK
TENURE
MARANK × TENURE
UNVA
DCOVPRO
LIST
APC
FOG
SIZE
ROA
LEV
MTB
GDW
WD
RC
OSI
Industry
Year
Obs.
Adj. R2
0.073
−0.016⁎
−0.001⁎⁎
−0.002⁎⁎
−0.000
0.000
−0.002
0.000⁎⁎⁎
b.0001
0.068
0.010
0.024
0.664
0.901
0.556
b.0001
0.184
0.000
b.0001
0.582
0.835
b.0001
0.000
0.077
b.0001
−0.006
−0.004⁎⁎⁎
−0.151⁎⁎⁎
−0.005
−0.000
−0.169⁎⁎⁎
0.245⁎⁎⁎
−0.179⁎
−0.195⁎⁎⁎
Included
Included
1390
0.6323
This table presents the results of Tobit regression with industry and year effects based on
the goodwill impairment sample over the period of 2002–2011. The above regression incorporates CEO tenure and the interaction term of managerial ability rank and CEO tenure.
The dependent variable (GWILOSS) measures the magnitude of goodwill impairment
losses. TENURE is the number of years since the CEO assumed the ofﬁce. The industryspeciﬁc and year-speciﬁc intercepts are omitted for brevity. Continuous control variables are winsorized at 1% and 99% percentiles each year before entering the regression tests. Refer to Appendix 1 for variable deﬁnitions.
⁎ Signiﬁcance at the 101% (two-tailed) conﬁdence level.
⁎⁎ Signiﬁcance at the 51% (two-tailed) conﬁdence level.
⁎⁎⁎ Signiﬁcance at the 1% (two-tailed) conﬁdence levels.
impairment losses. This is consistent with Beatty and Weber (2006)
and Ramanna and Watts (2012). Furthermore, I ﬁnd a signiﬁcant
(p-value = 0.024) and negative (−0.002) relationship between the interaction term (MARANK × TENURE) and the magnitude of goodwill
impairment losses, suggesting that capable CEOs with longer tenure
better reduce the magnitude of goodwill impairment losses than capable CEOs with shorter tenure. This ﬁnding offers another explanation
for the negative relationship between managerial ability and goodwill
impairment. That is, it is possible that managers with greater ability initially make better acquisition decisions when determining whether to
make an acquisition that leads to the booking of goodwill. These better
decisions at acquisition lead to smaller goodwill impairment losses. This
explanation is in line with Gu and Lev (2011), who argue that many
goodwill impairment losses are caused by managers’ poor acquisition
decisions. (See Table 9).
7. Conclusion
In this study, I examine the relationship between managerial ability
and goodwill impairment. After controlling for managers’ opportunistic
behavior, the regression analysis reveals a negative relationship between managerial ability and goodwill impairment measured as the
likelihood of goodwill impairment and the magnitude of goodwill impairment losses after goodwill impairment occurs. Findings suggest
that more-able managers better prevent goodwill impairment and
better reduce the magnitude of goodwill impairment losses, relative to
less-able managers. I also perform various additional tests to address
potential endogeneity issues. Additional tests provide consistent results.
It is difﬁcult to measure managerial ability because it is multidimensional. The managerial ability index scores by Demerjian et al. (2012) are an
approximate measure of management performance. More precise measures of management performance may yield stronger results. Readers
need to exercise caution when generalizing the conclusions.
L. Sun / Advances in Accounting, incorporating Advances in International Accounting 32 (2016) 42–51
Appendix 1. Variable deﬁnition
A.1. Dependent variables
GWI
= An indicator variable that takes a value of one if the ﬁrm-year has a
goodwill impairment loss and zero otherwise
GWILOSS = Goodwill impairment losses [GDWLIP (#368) × (−1)] scaled by
total assets at t − 1
A.2. Primary variables of interest
MA
=
Managerial ability score by Demerjian et al. (2012).
MARANK
=
Decile ranking of managerial ability score by Demerjian et al. (2012).
A.3. Control variables
UNVA
=
(−1) × [Cash (CHE, #1) + Short-term investment (IVST, #193)
+Investments and advances (IVAO, #32) − Debt in current liabilities (DLC, #34) − Long-term liabilities (DLTT, #9) − Preferred
stock (PSTK, #130)] divided by [Total assets (AT, #6) − Total
liabilities (LT, #181)]
DCOVPRO
=
LIST
=
APC
=
FOG
=
SIZE
ROA
LEV
MTB
=
=
=
=
GDW
WD
=
=
RC
=
OSI
=
TENURE
=
An indicator variable set to one if market to book ratio is less than
one in year t − 1 and year t, and zero otherwise
An indicator variable set to one if the ﬁrm trades on the NASDAQ
or AMEX, and zero otherwise
The coefﬁcient from regressing a ﬁrm’s price on its operating
income using at least 16 and up to 20 quarters of data prior to
year t
An indicator variable set to one if the Fog index is greater than 18
(unreadable), and zero otherwise
Natural log of total assets (AT, #6)
Net income (NI, #172) scaled by total assets (AT, #6) at t − 1
Long-term liabilities (DLTT, #9) divided by total assets (AT, #6)
[Outstanding common shares (CHSO, #25)×Stock price at ﬁscal
year end (PRCC_F, #24)] divided by total book value (CEQ, #60)
Total goodwill (GDWL, #204) scaled by total assets (AT, #6)
Total long-term assets write-downs (WDP, #380) scaled by total
assets (AT, #6) at t − 1
Restructure charges (RCP, #376) scaled by the total assets (AT,
#6) at t − 1
[Special items (SPI, #17) − Goodwill impairment losses
(GDWLIP, #368) − Long-term assets write-downs (WDP, #380)
− Restructure charges (RCP, #376)] scaled by total assets (AT,
#6) at t − 1
The number of years since the CEO assumed the ofﬁce
References
Aier, J., Comprix, J., Gunlock, M., & Lee, D. (2005). The ﬁnancial expertise of CFOs and accounting restatements. Accounting Horizons, 19(3), 123–135.
Baik, B., Farber, D. B., & Lee, S. (2011). CEO ability and management earnings forecasts.
Contemporary Accounting Research, 28(5), 1645–1668.
Bamber, L., & Wang, J. I. (2010). What’s my style? The inﬂuence of top managers on voluntary corporate ﬁnancial disclosure. The Accounting Review, 85(4), 1131–1162.
Beatty, A., & Weber, J. (2006). Accounting discretion in fair value estimates: An examination of SFAS 142 goodwill impairments. Journal of Accounting Research, 44(2),
257–288.
Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on ﬁrm
policies. The Quarterly Journal of Economics, 118(4), 1169–1208.
51
Chalmers, K. G., Godfrey, J. M., & Webster, J. C. (2011). Does a goodwill impairment regime
better reﬂect the underlying economic attributes of goodwill? Accounting and
Finance, 51, 634–650.
Chen, C., Kohlbeck, M., & Warﬁeld, T. (2008). Goodwill valuation effects of the initial
adoption of SFAS 142. Advances in Accounting, 24(1), 72–81.
Darrough, M., Guler, L., & Wang, P. (2014). Goodwill impairment losses and CEO compensation. Journal of Accounting, Auditing and Finance, 29(4), 435–463.
Demerjian, P., Lev, B., & McVay, S. (2012). Quantifying managerial ability: A new measure
and validity tests. Management Science, 58(7), 1229–1248.
Demerjian, P., Lewis, M., Lev, B., & McVay, S. (2013). Managerial ability and earnings quality. The Accounting Review, 88(2), 463–498.
Demerjian, P., Lewis-Western, M., & McVay, S. (2015). Earnings smoothing: For good or
evil? Working paper. University of Washington.
Dyreng, S., Hanlon, M., & Maydew, E. (2010). The effects of executives on corporate tax
avoidance. The Accounting Review, 85(4), 1163–1189.
Fama, E. F., & French, K. R. (1997). Industry costs of equity. Journal of Financial Economics,
43, 153–193.
Filip, A., Jeanjean, T., & Paugam, L. (2015). Using real activities to avoid goodwill impairment losses: Evidence and effect on future performance. Journal of Business, Finance
and Accounting, 42(3–4), 515–554.
Financial Accounting Standards Board (FASB) (2001). Goodwill and other intangible assets. Statement of ﬁnancial accounting standards no. 142. CT: Norwalk.
Financial Accounting Standards Board (FASB) (2011). Goodwill and other. Accounting
standards update no. 350–20. CT: Norwalk.
Francis, B. B., Sun, X., & Wu, Q. (2014). Managerial ability and tax avoidance. Working
paper. Rensselaer Polytechnic Institute.
Francis, J., Hanna, J. D., & Vincent, L. (1996). Causes and effects of discretionary assets
write-offs. Journal of Accounting Research, 34, 117–134.
Francis, J., Nanda, D., & Olsson, P. (2008). Voluntary disclosure, earnings quality, and cost
of capital. Journal of Accounting Research, 46(1), 53–99.
Ge, W., Matsumoto, D., & Zhang, J. (2011). Do CFOs have style? An empirical investigation
of the effect of individual CFOs on accounting practices. Contemporary Accounting
Research, 28(4), 1141–1179.
Gu, F., & Lev, B. (2011). Overpriced shares, ill-advised acquisitions, and goodwill impairment. The Accounting Review, 86(6), 1995–2022.
Hambrick, D. (2007). Upper echelons theory: An update. Academy of Management Review,
32(2), 334–343.
Hambrick, D., & Mason, P. (1984). Upper echelons: The organization as a reﬂection of its
top managers. The Academy of Management Review, 9(2), 193–206.
Hayn, C., & Hughes, P. J. (2006). Leading indicators of goodwill impairment. Journal of
Accounting, Auditing and Finance, 21, 223–265.
Henning, S. L., & Shaw, W. H. (2003). Is the selection of the amortization period of goodwill a strategic choice? Review of Quantitative Finance and Accounting, 20, 315–333.
Hirschey, M., & Richardson, V. J. (2002). Information content of accounting goodwill numbers. Journal of Accounting and Public Policy, 21, 173–191.
Jennings, R., Robinson, J., Thompson, R., & Duvall, L. (1996). The relation between accounting goodwill numbers and equity values. Journal of Business, Finance and
Accounting(June), 513–533.
Jiraporn, P., Jiraporn, N., Boeprasert, A., & Chang, K. (2014). Does corporate social responsibility improve credit ratings? Evidence from geographic identiﬁcation. Financial
Management, 43(3), 505–531.
Krishnan, G. V., & Wang, C. (2015). The relation between managerial ability and audit fees
and going concern opinions. Auditing: A Journal of Practice and Theory (in press).
Li, F. (2008). Annual report readability, current earnings, and earnings persistence. Journal
of Accounting and Economics, 45, 221–247.
Li, K., & Sloan, R. G. (2015). Has goodwill accounting gone bad? Working paper. University
of Toronto.
Li, Z., Shroff, P. K., Venkataraman, R., & Zhang, I. X. (2011). Causes and consequences of
goodwill impairment losses. Review of Accounting Studies, 16, 745–778.
Olante, M. E. (2013). Overpaid acquisitions and goodwill impairment losses—Evidence
from the U.S. Advances in Accounting, 29, 243–254.
Petersen, M. (2009). Estimating standard errors in ﬁnance panel data sets: Comparing
approaches. Review of Financial Studies, 22(1), 435–480.
Ramanna, K., & Watts, R. (2012). Evidence on the use of unveriﬁable estimates in required
goodwill impairment. Review of Accounting Studies, 17(4), 749–780.
Richardson, S., Sloan, R., Soliman, M., & Tuna, I. (2005). Accrual reliability, earnings persistence and stock prices. Journal of Accounting and Economics, 39, 437–485.
Sun, L., & Zhang, J. H. (2016). The impact of goodwill impairment losses on bond credit
ratings. Working paper. University of Tulsa.
Wang, G. (2013). Smarter money: How high ability managers trade before ﬁrm-speciﬁc
events. Working paper. York at Buffalo: State University of New.
Xu, W., Anandarajana, A., & Curatolab, A. (2011). The value relevance of goodwill impairment. Research in Accounting Regulation, 23(2), 145–148.
Journal of Business Finance & Accounting
Journal of Business Finance & Accounting, 37(3) & (4), 456–485, April/May 2010, 0306-686X
doi: 10.1111/j.1468-5957.2010.02203.x
The Impact of Dual Class Structure on
Earnings Management Activities
VAN THUAN NGUYEN AND LI XU∗
Abstract: This paper hypothesizes and finds that firms with dual class structure are less likely
to engage in earnings management activities than firms with single class structure. The paper
also finds that within the sample of firms with dual class structure, earnings management
activities are positively associated with managerial cash flow rights, but marginally and negatively
associated with managerial voting rights. In addition, the divergence between voting rights and
cash flow rights has a marginally negative impact on earnings management. Finally, in a sample
of firms switching from dual class structure to single class structure, earnings management
activities increase following the switch.
Keywords: dual class, earnings management, agency cost, ownership structure, voting rights
1. INTRODUCTION
A nontrivial number of publicly-traded companies in the United States have more than
one class of common stock, and many executives at these companies claim that one of
the virtues of the dual class system is that it can allow management to concentrate on
their core, long-term interests, despite fluctuations in quarterly results.1 For example,
Google’s initial public offering prospectus states:
Dual class ownership has allowed these companies to concentrate on their core, longterm interest in serious news coverage, despite fluctuations in quarterly results . . . .
∗ The authors are respectively from the Department of Accounting and Finance, Morgan State University;
and the School of Accountancy, Southern Illinois University at Carbondale. Li Xu gratefully acknowledges
the financial support from Southern Illinois University at Carbondale. Van Thuan Nguyen gratefully
acknowledges that this research is supported in part by funds from the Morgan State University Office
of Faculty Professional Development under a Title III Grant from the US Department of Education. The
authors appreciate the comments from participants at the 2009 JBFA Conference, 2009 FMA Conference,
2009 AAA Midwest Conference, an anonymous referee and the editor, Steven Young. They also thank Paul A.
Gompers, Joy Ishii and Andrew Metrick for generously providing them with the dual-class sample data. They
also appreciate the research assistance of Alex Igolnikov and Kavya Yenigalla. All remaining errors are the
authors’ own. Li Xu and Van Thuan Nguyen contributed equally to this work. (Paper received November
2008, revised version accepted February 2010)
Address for correspondence: Li Xu, School of Accountancy, Rehn Hall, Room 220A – Mail Code
4631, Southern Illinois University at Carbondale, 1025 Lincoln Drive, Carbondale, IL 62901, USA.
e-mail: lixu@cba.siu.edu
1 According to Gompers et al. (2009), 6% of all Compustat firms are dual-class, comprising about 8% of
the market capitalization of all firms.

C 2010 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK
and 350 Main Street, Malden, MA 02148, USA.
456
THE IMPACT OF DUAL CLASS STRUCTURE
457
From the point of view of long-term success in advancing a company’s core values, the
structure has clearly been an advantage.
However, there is little empirical evidence regarding whether dual class structure
enables managers to focus on long-term development of firms and ignore short-term
earnings targets. The purpose of this paper is to examine whether managers from firms
with dual class structure are less likely to engage in earnings management activities
to achieve short-term earnings targets (compared to managers from firms with single
class structure); this paper provides evidence that can be used to support an executive’s
decision to choose a dual class structure.
In a typical dual class company, there are two classes of stock: one publicly-traded
‘inferior’ class with one vote per share, and another non-publicly-traded ‘superior’
class of stock with more than one vote per share. The most common structure is ten
votes per share for the superior class (Smart and Zutter, 2003). Since the superior
class is usually owned mostly by the insiders (managers and directors), the dual class
share structure creates a significant divergence between cash flow rights and voting
rights. This divergence provides insiders with a majority of voting rights, despite their
smaller residual claims. The dual class structure essentially creates a concentration of
control that differs from concentrated ownership in general (Ben-Amar and André,
2006). We conjecture that the concentrated voting rights created by the dual class
structure can effectively reduce the likelihood that management is displaced in a
hostile takeover. With no need to worry about dismissal, managers will have less
incentive to manage earnings at the expense of long-term value. The smaller residual
claims will also reduce the possible benefits associated with earnings management
activities. By examining the dual class firms, we are able to observe the effect of the
concentration of control created by dual class structure on earnings management
activities.
We begin by documenting whether managers of dual class firms are less likely to
engage in earnings management activities compared to a sample of single class firms
(using a cross- firm sample test). We measure earnings management using both the
magnitude of absolute abnormal accruals and the frequency of earnings meeting or
just beating analysts’ forecasts. To control for any simultaneity between dual class
structure and earnings management activities, we use a two-stage probit least squares
regression model (2SPLS). We find that managers of dual class firms engage in less
earnings management activities, as evidenced by lower frequency of earnings meeting
or just beating analysts’ forecasts, and a smaller magnitude of absolute abnormal
accruals.
We next examine whether managers of dual class firms with lower cash flow rights
are less likely to engage in earnings management activities compared to managers
of dual class firms with higher cash flow rights (using a within-firm sample test). We
find higher managerial cash flow rights are associated with a higher likelihood of
earnings meeting or just beating analysts’ forecasts, and a larger magnitude of absolute
abnormal accruals. Higher managerial voting rights are marginally associated with a
lower likelihood of earnings meeting or just beating analysts’ forecasts. In addition, the
difference between voting rights and cash flow rights has a marginally negative impact
on earnings management activities. These results are consistent with the argument
that higher managerial cash flow rights and lower managerial voting rights lead to
more earnings management activities.

C 2010 Blackwell Publishing Ltd.
458
NGUYEN AND XU
In addition, in a sample of firms switching from dual class structure to single
class structure, we find that the magnitude of absolute abnormal accruals increases
significantly in the three-year period following the switch. Our interpretation of these
results is that switching to single class structure increases managerial incentives to
manipulate earnings.
Francis et al. (2005) find that earnings are less informative for dual class firms
due to credibility concerns. In this paper, we find that the managers from dual class
firms are less likely to manipulate earnings compared to those from single class firms.
One possible explanation of these results is that due to the lack of credibility, the
investors attach lower weight to reported earnings of dual class firms. Managers of dual
class firms, therefore, have fewer incentives to manipulate earnings, since the possible
benefits associated with such behavior are smaller than they would be in single class
firms. This explanation is consistent with our claim that managers from dual class firms
place less importance on meeting certain earnings benchmarks compared to managers
from single class firms.
This paper contributes to the earnings management literature in the following ways.
Although earnings management has received considerable attention in accounting
and finance literature, little is known about the association between share structure
and earnings management. Prior studies on the motivations surrounding earnings
management tend to focus on how earnings management can be explained by the
contracting incentives (e.g., Sweeney, 1994; and Healy, 1985) and by the regulatory
incentives (e.g., Collins et al., 1995). This paper therefore enriches the literature
on earnings management by providing evidence that ownership structure can affect
earnings management. Secondly, as pointed out by Healy and Wahlen (1999),
standard setters are interested in evidence on the magnitude, frequency and motives
for earnings management. Our study provides direct evidence that managers of dual
class firms engage in less earnings management activities. These findings can help
regulators better allocate scarce resources for the enforcement of standards. Finally,
our findings are important to investors who are interested in the capital market
implications of dual class ownership structure. Studies in this area have examined the
liquidity effect (Kim et al., 2007), the wealth effect (e.g., Gompers et al., 2009), and the
disclosure effect (Tinaikar, 2008) of dual class ownership structure. Our study adds to
this literature by showing that dual class ownership structure can influence managers’
reporting judgment.
The remainder of the paper is organized as follows. In Section 2, we present
theoretical backgrounds and our predictions. We discuss our sample selection in
Section 3 and empirical results in Section 4 and Section 5. We summarize and offer
our conclusions in Section 6.
2. THEORETICAL BACKGROUNDS AND PREDICTIONS
By nature, public companies are characterized by a separation of ownership from
control. In return for their equity, the owners (shareholders) profit from a company’s
performance through price appreciation and dividend distribution. Shareholders
delegate decision-making rights to a board of directors; directors, in turn, delegate dayto-day responsibilities to the corporate managers. To effectively control management’s
behavior, shareholders are granted the right to vote on major issues. Normally, these
rights are proportional to the shareholder’s equity stake in the company. Some firms’

C 2010 Blackwell Publishing Ltd.
THE IMPACT OF DUAL CLASS STRUCTURE
459
founders may want to avoid the dilution of control that normally accompanies the
public issuance of shares by issuing different classes of shares that confer different
voting rights on the holders. Such a scenario is known as a dual class share structure.2
Dual class share structure usually creates concentrated control within firms. As pointed
out by Gompers et al. (2009), using a comprehensive sample of dual class firms from
1995–2002, insiders have approximately 60% of the voting rights and 40% of the cash
flow rights in dual class firms. For almost 40% of the dual class firms, insiders have
more than half of the voting rights, but less than half of the cash flow rights.
Dual class structure can be created either through initial public offerings (IPOs)
or recapitalizations. In either situation, dual class structure is viewed as explicitly
defensive, in that it discourages hostile takeovers and proxy contests. Research on
whether the defensive aspects of dual class structure harm shareholders yields mixed
results: One stream of research claims that dual class structure is value enhancing
because the defensive mechanism allows managerial continuity (due to the large
voting power of insiders), which gives managers more incentives to focus on the
firm’s long-run performance (DeAngelo and DeAngelo, 1985). Consistently, Lehn
et al. (1990) and Dimitrov and Jain (2006) demonstrate that firms with better growth
opportunities are more likely to adopt a dual class structure. Such firms subsequently
engage in seasoned equity offerings (SEOs) of inferior-voting shares to finance their
projects, and show positive long-term performance. Another stream of research
concludes that dual class structure has value decreasing consequences because the
defensive mechanism weakens the capital markets’ effective monitoring of managers,
who are entrenched and have incentive to divert corporate resources. Jarrell and
Poulsen (1988) and Mikkelson and Partch (1994) provide evidence that the markets
respond negatively to the adoption of dual class structure and document significant
negative abnormal returns around the announcement day. Moreover, Lins (2003) and
Gompers et al. (2009) find that firm values are negatively related to insiders’ voting
rights, but positively related to insiders’ cash flow rights.
We conjecture that dual class structure will reduce managerial incentive to manipulate earnings for several reasons. Managers manipulate earnings in order to meet some
benchmarks. Mohanram (2003) states that:
Such benchmarks could be the previous period’s performance (the desire to show an
improving trend), analysts’ expectations (the desire to meet or beat expectations), zero
(the desire to remain profitable), or whatever benchmark is specified in a manager’s
compensation contract (the desire to meet a bonus threshold).
It is extremely costly to miss these benchmarks because the relationship between
stock price or compensation) and earnings is very non-linear around the benchmarks
(Burgstahler and Dichev, 1997). Missing these benchmarks may lead to a sharp decline
in stock price, which in turn leads to higher likelihood of management turnover and
hostile takeovers. In addition, managers can suffer greatly from a reduction in stock
price when they own a large position of company stocks (Cheng and Warfield, 2005).
When firms are extremely close to a target, the incentives to take earnings
just over the target become exceedingly strong. In these cases, the firms will try
2 Note that this is substantially different from the situation where one group of shareholders or a single
shareholder holds a significantly large share position to control the company. In this latter case, control is
proportional to the financial risk conferred by share ownership. In contrast, dual-class share structure alters
the normal 1:1 relationship between cash flow rights and voting rights.

C 2010 Blackwell Publishing Ltd.
460
NGUYEN AND XU
and use some form of upward earnings management to ‘bump up’ earnings over
the target (Burgstahler and Dichev, 1997). For dual class firms, the incentives to
manipulate earnings upward are significantly reduced, since dual class structure is
characterized by higher voting rights and lower cash flow rights, compared to single
class structure. The higher voting rights can be considered long-term employment
contracts in a sense because the controlling voting power reduces the likelihood that
the management is displaced. Thus, managers may place less importance on meeting
the benchmark. Further, one feature of accounting accruals (which is a proxy for
earnings management) is that they reverse over time. Thus, when managers expect to
stay on the job for an extended period of time, they may be less motivated to manage
earnings upward because they realize that, eventually, the accruals will unwind and
they will probably still be holding their jobs and will have to deal with the unwinding
accruals. Moreover, the higher voting rights associated with dual class structure should
reduce the benefits associated with the manipulated stock price (which is linked to the
manipulated earnings). This is consistent with Cheng and Warfield (2005), who show
that earnings management activities are positively related to cash flow rights.
Managers may also intentionally manipulate earnings downward. This is especially
likely when firms are either far above or far below their targets. For instance, when
firms are far below their targets, managers have an incentive to report even worse
performance because the costs of doing so are typically minimal. Such earnings
management is referred to as ‘big-bath’ accounting. When firms are far above their
targets, managers may again have an incentive to reduce earnings because there is
little benefit in going far above a benchmark. Such earnings management is referred to
as ‘cookie-jar’ accounting. In both cases, given the self-adjusting nature of accounting,
the manipulated downward earnings will lead to boosts in future income, which makes
it easier for managers to meet the benchmark in the future. Dual class firms with
higher voting rights and lower cash flow rights (compared to single class structure)
face less pressure from capital markets to signal firm value to the market. Therefore,
management may place less importance on meeting the benchmark, thereby reducing
the incentives to manipulate earnings downward (Klassen, 1997).
Taken together, the unique characteristics of the dual class structure are predicted
to reduce managerial incentives to manipulate short-term earnings. This is consistent
with Stein (1989), who claims that managers who are more certain that their jobs are
secure are more likely to make long-term plans for the benefit of the firm, rather than
engaging in short-term window dressing by managing earnings.
Based on these arguments, we predict that, relative to managers from single class
firms, managers from dual class firms have fewer incentives to manipulate earnings.
Accordingly, our first hypothesis (stated in alternative form) is as follows:
H1 : The incidence of earnings management is negatively associated with dual class
structure.
If dual class structure reduces managers’ earnings management activities, we expect
that managers from dual class firms with higher voting rights or lower cash flow
rights are less likely to engage in earnings management (compared to those from
single class firms). The rationale for the above arguments follows: If the concentrated
control characteristic of dual class structure mitigates managers’ incentives to manage
earnings by increasing their job security, then the more concentrated control is,
the fewer incentives there are for managers to manage earnings. In contrast, firms

C 2010 Blackwell Publishing Ltd.
THE IMPACT OF DUAL CLASS STRUCTURE
461
with higher cash flow rights will have more incentives to manage earnings, since the
benefits are larger compared to the financial risks involved (Cheng and Warfield,
2005). Hence, we expect that, within dual class firms, managers from firms with
higher voting rights are less likely to engage in earnings management activities,
while managers from firms with higher cash flow rights are more likely to engage in
earnings management activities. In addition, since cash flow rights and voting rights
have different implications for earnings management, managers probably consider
both cash flow rights and voting rights jointly when they make earnings management
decisions. In other words, since these cash flow rights and voting rights have opposite
effects on earnings management, managers may take the divergence between cash
flow rights and voting rights into consideration. We predict that the divergence has
a negative impact on earnings management activities. This leads to the following
predictions (presented in alternative form):
H2a : For dual class firms, the incidence of earnings management is negatively
associated with voting rights and positively associated with cash flow rights.
H2b : For dual class firms, the incidence of earnings management is negatively
associated with the difference between voting rights and cash flow rights.
Finally, if a dual class structure reduces managers’ incentives to manipulate earnings, we should see a change in these incentives when a firm switches from dual class
to single class. Specifically, managers from dual class firms are more likely to engage in
earnings management activities when these firms switch from dual class to single class
structure. Accordingly, we test the following prediction:
H3 : When a firm switches from dual class to single class, the incidence of earnings
management increases.
3. SAMPLE SELECTION PROCEDURES
Our dual class sample firms are based on Gompers et al.’s (2009) study. Gompers et al.
build a comprehensive set of dual class firms from 1995–2002, using data from
the Securities Data Company (SDC) (as amended by Jay Ritter), S&P’s Compustat,
the Center for Research in Security Prices (CRSP), and the Investor Responsibility
Research Center (IRRC).3 Using the same methodology as in Gompers et al. (2009),
we extend the dual class sample to 2006.
We start out with 4,491 dual class firm-years from 1995–2006. We eliminate the IPO
year, together with all financial institutions (SIC 6000-6999) and utilities firms (SIC
4400-5000). We also exclude all firms within any industry with less than 10 firms. The
final sample has 2,797 firm-years and the year distributions are: 258 firms in 1995; 277
firms in 1996; 302 firms in 1997; 302 firms in 1998; 295 firms in 1999; 275 firms in
2000; 254 firms in 2001; 220 firms in 2002; 171 firms in 2003; 157 firms in 2004; 150
firms in 2005; and 136 firms in 2006. It appears that the number of dual class firms is
decreasing in recent years. Fourteen firms in our final sample switch from single class
to dual class structure (10 firms have sufficient data), and 72 firms switch from dual
class to single class structure (70 firms have sufficient data).
3 See Gompers et al. (2009) for more detailed discussion on dual-class sample construction.

C 2010 Blackwell Publishing Ltd.
462
NGUYEN AND XU
We use two proxies of earnings management activities. The first proxy is the
frequency of earnings meeting or just beating analysts’ forecasts. Recent studies of
earnings management suggest that the disproportionate likelihood of meeting or just
beating analysts’ forecasts is an important manifestation of earnings management
(Degeorge et al., 1999; and Burgstahler and Eames, 2003, among others). Relative
to other proxies, such as abnormal accruals, this proxy has more direct market consequences. Prior research documents rewards for meeting analysts’ forecasts (Bartov
et al., 2002; and Kasznik and McNichols, 2002) and negative market consequences
for missing analysts’ forecasts (Skinner and Sloan, 2002). Both actual earnings and
analysts’ forecasts are collected from the Institutional Brokers’ Estimate System (IBES)
database. We exclude all dual class firms without the requisite data, resulting in a
sample of 2,095 firm-years. We name this sample ‘the earnings forecast sample’.4
The second earnings management proxy is unsigned (absolute) abnormal accruals,
which are defined as the absolute difference between total accruals and normal
accruals. Normal accruals are estimated using the cross-sectional Jones (1991) model
as described in DeFond and Jiambalvo (1994). We choose to use unsigned abnormal
accruals instead of signed abnormal accruals because we predict that dual class
structure will reduce both income-increasing and income-decreasing earnings management. According to Warfield et al. (1995), the extent to which companies use
accruals to manage earnings is best measured by the unsigned value of accruals. The
magnitude of unsigned accruals measures a company’s success in managing earnings
either up or down as required, depending on year-specific situations (DeFond and
Park, 1997; and Healy, 1985). Data for this second measure are obtained from the
Compustat annual industry database and comprise 2,797 firm-years. We name this
sample ‘the abnormal accruals sample’.
Note that these two proxies do not measure the same type of earnings management.
The meeting or just beating analysts’ forecasts proxy captures earnings management
activities that manipulate earnings upward to beat analysts’ forecasts. The absolute
abnormal accruals proxy captures all earnings management activities (i.e., earnings
can be manipulated upward or downward). As discussed in Section 2, we expect that
dual class structure is negatively associated with all types of earnings management
activities.
We begin our analyses by comparing the dual class sample with the single class
firms from the entire Compustat population for the entire sample period. Panels A
and B of Table 1 compare selected characteristics of dual class and single class firms
for the accrual test and earnings forecasts test, respectively. Compared to single class
firms, dual class firms on average have higher past growth and lower return volatility,
earnings volatility and cash flow volatility.
Panel C of Table 1 compares the corporate governance statistics for dual class and
single class sample firms. Both board structure and executive compensation structure
are examined. Board members from dual class firms are on average less independent.
Compared to dual class firms, single class firms appear to have better corporate
governance, as evidenced by a significantly higher G-Index. In terms of compensation
structure, executives from dual class firms have, on average, a higher percentage of
bonuses and salaries and a lower percentage of options, compared to executives from
4 Note that by excluding firms without analyst coverage, our results may be subject to bias because firms
without analyst coverage tend to be smaller and less profitable. That is, only firms which are big or profitable
enough to attract analysts are included in our sample. It is not certain how this bias will affect our results.

C 2010 Blackwell Publishing Ltd.
THE IMPACT OF DUAL CLASS STRUCTURE
463
single class firms. Finally, 42.60% of our dual class sample firms have their founders
as their CEOs. This suggests that close to half of dual class firms are headed by their
founders.5
Table 2 reports the voting structure of the dual class firms. The most common
arrangement is a 10:1 voting structure, in which the superior class has ten votes per
share and the inferior class has one vote per share. Demonstrating how such a structure
can affect the eventual ownership and control of the firms, Table 2 displays the
fractions of cash flow and voting rights held by the insiders. On average, the insiders
of dual class firms own a majority of the voting rights. In the abnormal accruals sample
(reported in Panel A), 64.34% of the insiders own more than 50% of the voting rights,
while the corresponding number is 65.57% for the earnings forecast sample (reported
in Panel B). In addition, for the abnormal accrual (earnings forecast) sample, 38.67%
(38.54%) of the insiders own more than 50% of the voting rights and less than 50% of
the cash flow rights.
4. EMPIRICAL TESTS RESULTS
(i) Cross-Sample Empirical Results
We begin by comparing dual class firms to the single class firms in the entire Compustat population in subsection (i) (cross-sample tests).6 We then focus exclusively on dual
class firms in subsection (ii) (within-sample tests) and subsection (iii) (switch-sample
tests).
We start by examining the earnings management proxies without considering other
firm characteristics. Table 3 , Panel A reports the magnitude of absolute abnormal
accruals by comparing the dual ownership sample with the single ownership sample.
As evidenced by these results, firms with a dual class ownership structure tend to
have smaller absolute abnormal accruals: mean (median) absolute abnormal accruals
(ACCRUAL) of the dual class sample are 0.2486 (0.0668), compared to 0.7339
(0.1161) for the single class sample. Differences are statistically significant using a twosample t-test for the means at a p-value of 0.01, and a two-sample Wilcoxon test for the
medians at a p-value of 0.01.
Table 3, Panel B reports the frequency of firms meeting or beating analyst forecasts
by comparing the dual ownership sample with the single ownership sample. We do
not find significant differences in the frequency of firms meeting or beating analysts’
forecasts between two samples.
Since these univariate tests do not take other firm characteristics into consideration,
we may draw incorrect conclusions. We next examine the association between dual
class structure and earnings management by estimating multivariat…

Turn in your highest-quality paper
Get a qualified writer to help you with

“ Summarize and PowerPoint ”

Get high-quality paper

Guarantee! All work is written by expert writers!

Still stressed from student homework?

Get quality assistance from academic writers!

Order now