Description
Weexaminetherelationbetweenthetransparencyofdisclosuresaboutactivityinvaluationallowanceandreserveaccountsandaccruals-basedearningsmanagement.Weclassifydisclosuresasbeingtranspar-entiftheyprovidedetailedinformationaboutactivityintheallowanceandreserveaccountsduringthefiscalperiod.Wefindstrongevidencethattheextentofaccruals-basedearningsmanagementisloweramongcompanieswithtransparentdisclosuresthanamongcompanieswithouttransparentdisclosures.
Accounting, Organizations and Society 46 (2015) 23–38
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Disclosure transparency about activity in valuation allowance and
reserve accounts and accruals-based earnings management
Cory A. Cassell
a
, Linda A. Myers
a,?
, Timothy A. Seidel
b
a
University of Arkansas, Fayetteville, AR, United States
b
Utah State University, Logan, UT, United States
a r t i c l e i n f o
Article history:
Received 2 June 2014
Revised 18 March 2015
Accepted 19 March 2015
Available online 16 April 2015
Keywords:
Disclosure transparency
Valuation allowances and reserves
Schedule II
Earnings management
Discretionary accruals
Management discretion
a b s t r a c t
We examine the relation between the transparency of disclosures about activity in valuation allowance
and reserve accounts and accruals-based earnings management. We classify disclosures as being transpar-
ent if they provide detailed information about activity in the allowance and reserve accounts during the
?scal period. We ?nd strong evidence that the extent of accruals-based earnings management is lower
among companies with transparent disclosures than among companies without transparent disclosures.
We also investigate whether the extent of accruals-based earnings management is lower for companies
that provide transparent disclosures in one comprehensive schedule (i.e., the Schedule II) relative to those
that provide transparent disclosures spread throughout the notes to the ?nancial statements. Although
regulators have expressed concern that the omission of a Schedule II could indicate a greater likelihood
of earnings management, our results indicate that it is the omission of transparent disclosures rather
than the omission of a comprehensive schedule outlining activity in the allowance and reserve accounts
that affects earnings management. Our ?ndings suggest that regulators, auditors, and investors should
consider subjecting companies that fail to provide transparent disclosures to additional scrutiny.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Accruals-based earnings management can be costly because
questionable accounting practices are likely to be scrutinized by
external auditors, investors, the board of directors, regulators,
and other stakeholders. Valuation allowances and reserves pro-
vide managers with substantial ?exibility to manage earnings be-
cause they are based on subjective estimates and are evaluated
at higher levels of materiality, making them inherently di?cult to
audit (Gri?th, Hammersley, & Kadous, 2013; Peecher, Schwartz,
& Solomon, 2007; Peecher, Solomon, & Trotman, 2013). In addi-
tion, any differences identi?ed by the auditor are more likely to
be waived when the underlying accruals are more subjective (Joe,
Wright, & Wright, 2011; Knapp, 1987; Wright & Wright, 1997).
Thus, prior research suggests that companies are more likely to at-
tempt to manage reported earnings using these types of accounts.
1
We examine the relation between the transparency of
?
Corresponding author.
1
For example, Nelson, Elliot, and Tarpley (2002) use a ?eld-based questionnaire
in which 253 auditors from one Big 5 audit ?rm described 515 speci?c instances in
which they believed that their clients were attempting to manage earnings. The au-
thors indicate that “by far the most frequently identi?ed attempts involve reserves”
(Nelson et al., 2002, 176).
disclosure related to activity in valuation allowance and reserve
accounts and accruals-based earnings management, where disclo-
sures are classi?ed as being transparent if they provide detailed
information about activity in the accounts during the ?scal period
(e.g., information about the current period expense accrual, write-
offs, etc. for the allowance for doubtful accounts).
2
The U.S. Securi-
ties and Exchange Commission (SEC) has long required that activ-
ity in subjective accrual accounts be disclosed. Speci?cally, within
Regulation S-X, which prescribes the format and content of ?nan-
cial reports ?led under the Securities Act of 1933 and the Securi-
ties Exchange Act of 1934, Rule 5-04 and Rule 12-09 require that
companies ?le a Schedule II.
3
The Schedule II should provide the
beginning and ending balances and the current period activity in
2
Barth and Schipper (2008) de?ne ?nancial reporting transparency as the extent
to which ?nancial reports reveal an entity’s underlying economics in a way that is
understandable by those using the ?nancial reports. Bushman, Piotroski, and Smith
(2004) de?ne corporate transparency as the widespread availability of ?rm-speci?c
information about publicly listed ?rms to parties outside of the ?rm. Our de?nition
of transparency is speci?c to our setting and relates to just one aspect of overall re-
porting transparency, but it is consistent with de?nitions set forth in prior research.
3
Schedule II is a supplemental schedule ?led with the company’s form 10-K and
is typically located after the footnotes to the ?nancial statements. Laws requiring
the use of Schedule II can be traced to 37 FR 14602 (passed on July 21, 1972) and
re-designated and amended in 45 FR 63679 (on September 25, 1980).http://dx.doi.org/10.1016/j.aos.2015.03.004
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
24 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
all material valuation allowances and reserves. The SEC exempts
companies in certain industries,
4
companies with immaterial bal-
ances in the valuation allowances and reserves, and companies
that provide comparable disclosures in the notes to the ?nancial
statements from the Schedule II requirement. Thus, under current
SEC regulations, companies are required to provide transparent dis-
closures for material allowance and reserve balances in either a
Schedule II or in the notes to the ?nancial statements.
5
Although the SEC requires transparent disclosure for material
allowance and reserve balances, we ?nd that compliance is lack-
ing. Speci?cally, when we examine companies where the aggre-
gate prior year balance of the allowance for doubtful accounts, in-
ventory valuation allowance, and deferred tax asset valuation al-
lowance is greater than one cent per share, thus removing compa-
nies that would not be able to increase current year earnings per
share (EPS) by at least one cent even if they were to fully elimi-
nate these accruals, we ?nd that approximately 30% fail to provide
transparent disclosures (either in a Schedule II or in the notes to
the ?nancial statements).
6
Although the following statement made
by Ronald A. Kima, a former Assistant Chief Accountant for the
SEC, focuses on problems associated with Schedule II compliance,
it captures the potential implications of non-transparent disclosure
in general,
7
Unfortunately, despite such schedule being required of most
public companies, few companies seemingly fully comply. The
absence of otherwise required data, or worse the outright omis-
sion of the schedule in its entirety, should raise investor con-
cerns that a company may be engaging in some degree of inap-
propriate earnings management… Any immateriality assertion
by a company’s management that has been predicated exclu-
sively on balance sheet measures is, at best, inappropriate and,
at worst, an attempt to conceal inappropriate earnings manage-
ment practices.
Consistent with claims in Kima (2007), we expect that com-
panies will use allowances and reserves to manage earnings if
the probability of detection is su?ciently low and if the com-
bined magnitude of these accrual accounts is su?ciently large
(so that their manipulation can have a meaningful effect on re-
ported earnings). With respect to the ?rst condition, we posit that
non-transparent disclosures about highly subjective accrual ac-
counts provide managers with ?exibility to in?uence the market’s
perceptions of earnings. Because market participants face limita-
tions when processing accounting information (Hirshleifer & Teoh,
2003), they may be less likely to see through the earnings man-
agement when companies disclose allowance and reserve accounts
non-transparently. Although prior research does not investigate the
4
Excluded companies include registered investment companies, insurance com-
panies, bank holding companies and banks, and brokers and dealers.
5
Note that because Rule 5-04 does not provide a de?nition of materiality, mate-
riality is left to management’s discretion. In 2000, the SEC proposed a rule which
would have required a disclosure similar to Schedule II in the notes to the ?nancial
statements, rather than as a supplementary schedule (SEC, 2000). The relocation of
this disclosure would have provided increased prominence and auditor responsibil-
ity for the disclosure. The majority of comments received on the SEC’s proposed
rule opposed the proposal. The SEC did not ?nalize the proposed rule, so Rule 5-
04 of Regulation S-X remains applicable and materiality thresholds continue to be
subjective.
6
To further support this ?nding, we use the Audit Analytics SEC comment let-
ter database to identify SEC comment letters with speci?c reference to Schedule II
or to the allowances most frequently included in a Schedule II. Our search resulted
in many examples of the SEC requesting increased disclosure of the activity in al-
lowance and reserve accounts (in accordance with Rule 5-04 of Regulation S-X).
7
See SEC Schedule II – Visibility into the integrity of reported results. AICPA
CPA Insider Newsletter (October 1, 2007). Retrieved fromhttp://www.cpa2biz.com/
Content/media/newsletters/cpainsider/cpainsider071001.jsp.
association between accruals-based earnings management and the
transparency of accruals disclosures, evidence suggests that man-
agers use the ?exibility in disclosure rules to engage in real earn-
ings management (Lee, Petroni, & Shen, 2006) and that transparent
disclosures facilitate investors’ ability to detect earnings manage-
ment (Hirst & Hopkins, 1998).
Because disclosures of allowances and reserves must be hand
collected, we focus our analyses on the three accounts most com-
monly included in a Schedule II – the allowance for doubtful ac-
counts, the valuation allowance for deferred tax assets, and the val-
uation allowance for inventories.
8
We hand collect disclosures for
all companies in non-exempt industries that include at least one
of the three allowance and reserve balances in a Schedule II, in the
notes to the ?nancial statements, or parenthetically in the balance
sheet in their 2008, 2009, or 2010 annual reports.
We ?rst investigate whether disclosure transparency for a given
allowance or reserve account affects discretion in that account.
Following Marquardt and Wiedman (2004), we estimate the ex-
pected balance in each account by multiplying the prior year bal-
ance by the current year growth rate. To further re?ne this mea-
sure for the allowance for doubtful accounts, we also multiply
by the growth in days sales outstanding (to capture the change
in collectability). Similarly, for the inventory valuation allowance,
we multiply by the growth in days inventory outstanding (to cap-
ture the change in inventory turnover). Our measure of the discre-
tion in each allowance and reserve account is the difference be-
tween the actual account balance and the expected balance. For
each account examined, we limit the sample to companies where
the disclosed beginning balance is at least one cent per share.
We contrast discretion in the individual allowance or reserve ac-
count for companies that provide transparent disclosure (either in
a Schedule II or in the notes to the ?nancial statements) with that
for companies that provide non-transparent disclosure. We ?nd
that discretion in the each of the three accounts, the allowance
for doubtful accounts, the valuation allowance for inventories,
and the valuation allowance for deferred tax assets is greater
(i.e., more income-increasing) when these accounts are disclosed
non-transparently.
Although we focus on the three allowance accounts most com-
monly included in a Schedule II (and not on all valuation and
qualifying accounts that could be included in the schedule), we
supplement our primary tests by examining broad measures of
earnings management. In these tests, we contrast overall discre-
tion in accruals for companies that provide transparent disclo-
sures for all material (disclosed) allowance and reserve balances
(either in a Schedule II or in the notes to the ?nancial state-
ments) with that for companies that provide non-transparent dis-
closures for all disclosed allowance and reserve balances. We ?nd
that companies that provide transparent disclosures report smaller
signed, absolute value, and positive (income-increasing) discre-
tionary accruals than do companies that provide non-transparent
disclosures.
Having established a negative association between disclosure
transparency and the extent of earnings management in subjec-
tive accrual accounts, we next investigate whether the placement
of transparent disclosures matters. As evidenced by the rules sur-
rounding Schedule II and by subsequent (defeated) regulatory pro-
posals which would have required that a Schedule II be included in
8
In addition to being the accounts most commonly included in Schedule II dis-
closures, these accounts have also been the focus of prior research examining earn-
ings management in speci?c accounts. See, for example, McNichols and Wilson
(1988), Schrand and Wong (2003), Frank and Rego (2006), and Cecchini, Jackson,
and Liu (2012).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 25
the notes to the ?nancial statements, regulators appear to believe
that the transparent disclosure of all accrual accounts in a single
schedule will facilitate and enhance investors’ ability to process the
information relative to comparable, but separate, disclosure in the
notes. These beliefs are consistent with context theory (Medin &
Schaffer, 1978), which suggests that the presentation of an isolated
piece of information can lead to different interpretations relative to
when this information is combined with other information. These
beliefs are also consistent with the proximity compatibility princi-
ple (Carswell & Wickens, 1996; Wickens & Carswell, 1995), which
suggests that information processing can be enhanced when the
information is displayed in close proximity, and with cognitive load
theory (Sweller, 1988, 1989), which suggests that individuals will
be unable or unwilling to exert the cognitive effort necessary to
integrate information presented separately (Dietrich, Kachelmeier,
Kleinmuntz, & Linsmeier, 2001; Krische, 2005).
9
Thus, we expect
that the transparent disclosure of allowance and reserve accounts
in a single schedule (i.e., in a Schedule II) will aid users’ under-
standing of the (collective) information about activity in allowance
and reserve accounts, which should facilitate the detection of earn-
ings management in these accounts. Because managers have little
incentive to manage earnings if their actions will be detected, we
expect less earnings management among companies that provide a
Schedule II.
For tests using the estimated discretion in individual allowance
and reserve accounts, we limit our samples to companies that
provide transparent disclosures for the given account (either in a
Schedule II or in the notes to the ?nancial statements) and have
a prior year account balance greater than one cent per share. For
tests using the broad accruals measures, we limit our sample to
companies that provide transparent disclosures for all disclosed al-
lowance and reserve accounts where the aggregate prior year bal-
ance of these accounts exceeds one cent per share. We ?nd no sig-
ni?cant difference in the estimated discretion (either in individ-
ual accounts or in the broad accruals measure) between compa-
nies that provide a Schedule II and companies that provide trans-
parent disclosures in the notes to the ?nancial statements. Our in-
ability to ?nd support for this prediction could be due to a lack
of task complexity, the effect of a different (and competing) as-
pect of context being important in our setting, or a lack of power
in our tests. We discuss these possibilities in Section ‘Results
summary’.
We expect our study to be of interest to standard setters and
?nancial statement users as they consider proposals put forth in
the International Accounting Standard Board (IASB) and the Finan-
cial Accounting Standards Board (FASB) joint project related to ?-
nancial statement presentation. One such proposal would require
note disclosure of detailed activity in balance sheet accounts to fa-
cilitate user understanding of the current period change in posi-
tion (FASB, 2010). Our results suggest that the proposed require-
ment could be bene?cial because we document a negative asso-
ciation between the transparency of disclosure and earnings man-
agement. Our study should also be of interest to regulators, au-
ditors, and investors because our ?ndings suggest that the dis-
closure transparency of subjective accounts could be used as a
mechanism to identify companies that may warrant additional
scrutiny.
9
Several experimental studies provide evidence that the placement of equiva-
lent information affects investor and analyst judgments. See, for example, Hopkins
(1996), Hirst and Hopkins (1998), Maines and McDaniel (2000), and Hodge, Hop-
kins, and Wood (2010). Also see Libby and Emett (2014) for a broad overview of
the literature related to placement effects.
We also provide empirical evidence that addresses concerns
expressed by former SEC Assistant Chief Accountant Ronald
A. Kima, who suggests that the omission of a Schedule II
could indicate a greater likelihood of earnings management. We
?nd evidence that the omission of transparent disclosure, and
not necessarily the omission of a Schedule II, affects earnings
management.
The remainder of this paper is organized as follows: the next
section provides a review of related prior literature and develops
our hypotheses. The third section describes our empirical design.
We discuss the sample and present results from our analyses in
the fourth section. The ?nal section provides our conclusions.
2. Prior literature and development of hypotheses
Greater information asymmetry can provide companies greater
?exibility to manage earnings (Schipper, 1989). For example,
Richardson (2000) ?nds that increased information asymmetry,
measured by the bid-ask spread and analyst forecast dispersion, is
associated with higher levels of discretionary accruals. In addition,
Hodder and Hopkins (2014) ?nd that bank managers’ demand for
reporting opacity lead them to oppose accounting rules that would
increase disclosures about ?nancial instruments.
Some studies examine the relation between the transparency
of speci?c disclosures and earnings management (Hirst & Hopkins,
1998; Lee et al., 2006). These studies, however, focus exclusively
on the relation between the transparency of comprehensive in-
come disclosures and “real” earnings management decisions (i.e.,
the opportunistic selling of available-for-sale securities) as well as
the market’s ability to detect this real earnings management. Us-
ing archival data, Lee et al. (2006) ?nd that property-liability in-
surers with a history of selectively selling available-for-sale securi-
ties chose the less transparent format for disclosing comprehen-
sive income in the ?rst year of Statement of Financial Account-
ing Standards No. 130’s adoption. With respect to the detection of
earnings management, Hirst and Hopkins (1998) ?nd experimental
evidence that transparent comprehensive income disclosures en-
able buy-side ?nancial analysts to detect opportunistic selling of
available-for-sale securities and that they adjust their target prices
accordingly. Collectively, this stream of research suggests that in-
creased disclosure transparency about certain accounts or trans-
actions will reduce the likelihood that earnings are managed us-
ing real earnings management because increased transparency fa-
cilitates the detection of earnings management by market partici-
pants.
Lobo and Zhou (2001) and Shaw (2003) examine the relation
between disclosure quality and accruals-based earnings manage-
ment but they use a measure of overall disclosure quality rather
than a measure of the transparency of disclosures for speci?c ac-
cruals that may be used to manage earnings. Speci?cally, they
measure disclosure quality using the Association for Investment
Management Research (AIMR) scores. These scores are based on
industry-speci?c analyst evaluations of overall disclosure quality
and analysts consider annual and quarterly SEC ?lings, along with
other published information. Both Lobo and Zhou (2001) and Shaw
(2003) ?nd a negative association between AIMR scores and dis-
cretionary accruals, suggesting that companies with higher over-
all disclosure quality report lower discretionary accruals. However,
Shaw (2003) ?nds that this conservatism is limited to years with
good news (i.e., positive stock returns or cash ?ows from opera-
tions). In years with bad news, companies with higher AIMR scores
report higher levels of income-increasing discretionary accruals.
Thus, Shaw’s ?ndings suggest that companies with higher overall
disclosure scores smooth earnings more aggressively than do com-
panies with lower overall disclosure scores.
26 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
We extend prior research by investigating whether the dis-
closure transparency about speci?c accruals is associated with
accruals-based earnings management. While prior research pro-
vides evidence that real earnings management is lower in the pres-
ence of more transparent comprehensive income disclosures, prior
research does not provide evidence about accruals-based earnings
management in the presence of more transparent disclosure of
subjective accrual accounts. Additionally, regulators have expressed
concern that a lack of transparency in these accounts could be re-
lated to earnings management.
10
Because subjective accruals are evaluated at higher levels of
materiality (FASB, 1980) and any audit adjustments proposed to
these accruals are more likely to be waived (Joe et al., 2011; Knapp,
1987; Wright & Wright, 1997), auditors may not be effective at
identifying and constraining earnings management in these ac-
counts if disclosures about the activity in these accounts are not
transparent.
11
Because greater transparency about these accounts
could aid in the detection of accruals-based earnings management,
we expect companies that provide transparent disclosures about
activity in accrual accounts to engage in less earnings management
through these accounts. Our ?rst hypothesis, stated in the alterna-
tive form, is as follows:
Hypothesis 1. The extent of accruals-based earnings management
will be lower for companies that provide transparent disclosures
about activity in accrual accounts.
We also expect that earnings management may differ when
companies provide a Schedule II versus when companies provide
similar information about the activity in subjective accruals else-
where in the notes to the ?nancial statements. That is, there may
be differences in earnings management behavior within the set of
companies that provide transparent disclosures about activity in al-
lowance and reserve accounts.
Medin and Schaffer (1978) introduce context theory, which sug-
gests that the presentation of an isolated piece of information
can lead to different interpretations relative to when this infor-
mation is combined with other information. In addition, the prox-
imity compatibility principle (Carswell & Wickens, 1996; Wick-
ens & Carswell, 1995) coupled with cognitive load theory (Sweller,
1988, 1989) suggest that information processing can be enhanced
when the information is displayed in close proximity. Consistent
with these theories, experimental research provides evidence that
disclosure format and placement affects investors’ ability to pro-
cess information and/or detect earnings management (Hirst & Hop-
kins, 1998; Hodge et al., 2010; Krische, 2005; Maines & McDaniel,
2000). For example, Hirst and Hopkins (1998) show that profes-
sional investors and analysts often overlook information about un-
realized gains and losses on marketable securities when this infor-
mation is presented in a statement of stockholders’ equity rather
than in a statement of comprehensive income, and Maines and
McDaniel (2000) ?nd a similar effect among nonprofessional in-
vestors. Krische (2005) ?nds that investors adjust their evaluations
10
For example, Kima (2007) states, [g]iven their inherent uncertainty, contra-asset
accounts and liability reserves typically constitute the soft underbelly of any set of
consolidated ?nancial statements. Thus, contra asset accounts and liability reserves
pose the greatest accounting challenges to ethical managements as they attempt to
derive the best possible estimates. Unfortunately, ethically-challenged managements
often exploit the softness of contra-asset accounts and liability reserves to engage
in inappropriate earnings management, i.e., tucking away earnings for a rainy day.
11
Alternatively, increased disclosure transparency may not affect accruals-based
earnings management behavior because auditors have visibility into these accounts
regardless of the level of disclosure transparency and auditors have incentives to
constrain accruals-based earnings management in order to adhere to SEC guidelines
and auditing standards (PCAOB, 2003, 2010; SEC, 1999).
of current period earnings to a greater extent when transparent
disclosures about prior period transitory gains and losses are re-
peated in the current year versus when they are made only in the
prior year. Finally, Hodge et al. (2010) ?nd that the proximity of
?nancial information improves nonprofessional investors’ ability to
process such information, where improved information processing
is re?ected in more accurate cash ?ow forecasts.
This prior evidence suggests that the transparent disclosure
of allowance and reserve accounts in a single schedule (i.e., in
a Schedule II) will aid users’ understanding of the (collective)
information about activity in allowance and reserve accounts,
which should facilitate the detection of earnings management in
these accounts. Because managers have little incentive to man-
age earnings if their actions will be detected, we would expect
less earnings management among companies that provide a Sched-
ule II. Our second hypothesis, stated in the alternative form, is as
follows:
Hypothesis 2. The extent of accruals-based earnings management
will be lower for companies that provide a Schedule II than for
companies that provide comparable information in the notes to the
?nancial statements.
3. Empirical design
3.1. Self-selection bias
We model the decision to disclose allowances and reserves
transparently using variables that should in?uence the quality
of voluntary disclosure. Because Lang and Lundholm (1993) ?nd
that analyst ratings of companies’ disclosure practices (using AIMR
scores) are higher for larger companies, companies with better per-
formance, and companies that issue securities in the current or
future periods, we include measures of company size (Size), per-
formance (CFO, ROA, Loss), and the issuance of new securities (Is-
sue). We include measures of company sales growth (Growth) and
industry sales growth (Ind_Growth) because managers should in-
crease disclosure when it is more di?cult for analysts to assess
value based on historical numbers (Frankel, Johnson, & Skinner,
1999). Because prior research ?nds a negative association between
leverage and voluntary disclosure (Chow & Wong-Boren, 1987; Eng
& Mak, 2003), we include leverage (Leverage). Prior work also ?nds
that companies with greater information asymmetry are likely to
bene?t from additional disclosure (Chen, DeFond, & Park, 2002;
Gu & Li, 2007; Lang, 1991) so we follow Gu and Li (2007) and in-
clude company age (Age) and analyst coverage (Analyst_Coverage)
to proxy for the extent of information asymmetry. We include in-
stitutional ownership (Inst_Holdings) because prior research ?nds
a positive association between AIMR scores and institutional own-
ership (Bushee & Noe, 2000). We include a number of variables
to proxy for auditor quality (Big 4, Specialist, and Tenure) be-
cause prior work ?nds a positive association between auditor qual-
ity and disclosure quality (Almutairi, Dunn, & Skantz, 2009; Dunn
& Mayhew, 2004). Because companies committed to high-quality
internal controls might also be committed to high-quality disclo-
sure, we control for internal control material weaknesses (ICMW).
We include the Her?ndahl index to proxy for proprietary costs
(HERF) because companies should be less likely to commit to high-
quality disclosures when proprietary costs are high, and we in-
clude an indicator for litigious industry membership (Lit) because
managers might provide high-quality disclosures in an effort to
avoid litigation (Field, Lowry, & Shu, 2005; Skinner, 1994, 1997).
Finally, because Lennox, Francis, and Wang (2012) recommend the
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 27
use of an exclusion restriction when estimating selection models,
we also include the level of investment in research and devel-
opment (R&D). Research and development activities should affect
the proprietary costs of disclosure but should not affect the dis-
cretionary portion of current period working capital accruals be-
cause research and development activities relate to future product
or service offerings for which no sales have yet been made and no
costs have yet been capitalized. Our ?rst-stage probit model is as
follows:
Pr(Transparent_All
it
= 1) = ?
0
+?
1
(Size
it
) +?
2
(CFO
it
)
+?
3
(ROA
it
) +?
4
(Loss
it
) +?
5
(Issue
it
)
+?
6
(Growth
it
) +?
7
(Ind_Growth
it
)
+?
8
(Leverage
it
) +?
9
(Age
it
)
+?
10
(Analyst_Coverage
it
)
+?
11
(Inst_Holdings
it
) +?
12
(Big4
it
)
+?
13
(Specialistit) +?
14
(Tenure
it
)
+?
15
(ICMW
it
) +?
16
(HERF
it
)
+?
17
(Lit
it
) +?
18
(R&D
it
) (1)
where
Transparent_All an indicator variable set equal to one if the company provides transparent disclosures for all material (disclosed) valuation allowances and
reserves, either in a Schedule II or in the notes to the ?nancial statements, and zero otherwise
Size the natural log of total assets
CFO operating cash ?ow in the current year divided by total assets at the beginning of year
ROA return on assets, measured as net income divided by total assets
Loss an indicator variable set equal to one if net income is less than zero, and zero otherwise
Issue an indicator variable set equal to one if the company issued equity or debt in the year, and zero otherwise
Growth sales growth measured as sales less prior year sales, scaled by prior year sales
Ind_Growth industry sales (for the 2-digit SIC code industry) divided by prior year industry sales
Leverage long-term debt plus the current portion of long-term debt, divided by total assets
Age the number of years to date during which total assets is reported in Compustat
Analyst_Coverage an indicator variable equal to one if at least one analyst covers the company (from I/B/E/S), and zero otherwise
Inst_Holdings the percentage of company stock held by institutions
Big4 an indicator variable set equal to one if the auditor is from the Big 4, and zero otherwise
Specialist an indicator variable set equal to one if the auditor is an industry specialist, de?ned following Reichelt and Wang (2010) as an auditor
whose audit fee market share in the 2-digit SIC code exceeds 30% at the national level, and zero otherwise
Tenure the number of consecutive years to date of the auditor–client relationship
ICMW an indicator variable set equal to one if at least one material weakness in internal controls is disclosed in the year, and zero otherwise
HERF the Her?ndahl index estimated by summing the squared market shares of all companies in the 2-digit SIC code industry, where market
share is measured using sales
Lit an indicator variable set equal to one if the company operates in a high litigation risk industry (SIC codes 2833-2836, 3570-3577,
3600-3674, 5200-5961, and 7370), and zero otherwise
R&D research and development expense scaled by total assets
i and t company and year indicators
We include the inverse Mills ratio (IMR) derived from estimat-
ing this model in our second-stage models to mitigate concerns
relating to selection bias.
12
3.2. Tests of earnings management
We perform tests of our hypotheses using two approaches
to identify earnings management behavior. First, we investigate
whether companies that provide transparent disclosures for indi-
vidual allowance and reserve accounts use less discretion in those
accounts relative to companies that provide non-transparent dis-
12
Because many of the factors that in?uence managers’ disclosure choices are un-
observable and some of these unobservable factors are likely to be correlated with
the extent of earnings management, we control for these unobservable factors by
including the inverse Mills ratio in the second stage of a two-stage Heckman proce-
dure. Tucker (2010) argues that the two-stage Heckman procedure is useful in set-
tings where managers have considerable discretion and where the decision-making
process is more opaque to researchers. Relative to propensity score matching (PSM),
the Heckman approach is preferred in our setting because PSM relies exclusively on
observables to address the self-selection problem.
closures. Second, because we do not examine all valuation and
qualifying accounts, we supplement our primary tests by examin-
ing broad measures of earnings management. Speci?cally, we in-
vestigate whether companies that provide transparent disclosure
of all material (disclosed) allowances and reserves report discre-
tionary accruals that are less positive (i.e., less income-increasing)
or smaller in magnitude (absolute value) than companies that pro-
vide non-transparent disclosures. We discuss the construction of
our earnings management measures and the associated empirical
models in the remainder of this section.
3.2.1. Measures of discretion in individual accrual accounts
Prior research examines the use of discretion in speci?c ac-
counts to manage earnings. Although several studies model discre-
tion in the allowance for doubtful accounts (Cecchini et al., 2012;
Jackson & Liu, 2010; McNichols & Wilson, 1988), their models re-
quire information such as bad debt expense and write-offs that is
only available for companies that provide transparent disclosure.
Because our goal is to examine differences in discretion between
companies that provide transparent disclosure and those that do
not, we build on a simple model from Marquardt and Wiedman
(2004). Marquardt and Wiedman (2004) capture discretion in the
allowance for doubtful accounts by taking the difference between
the current year allowance and the prior year allowance, and ad-
justing the current year allowance for changes in gross accounts
receivable relative to the prior year. While changes in gross ac-
counts receivable can affect the allowance for doubtful accounts,
this model assumes that collectability of the underlying receivables
remains constant. We extend their model to incorporate changes
in the collectability of the underlying receivables. Speci?cally, be-
cause the ratio of days sales outstanding is a measure of the aver-
age number of days that a company takes to collect revenue after
a sale has been made, changes in this ratio should capture changes
in the collectability of the underlying receivables. We incorporate
this into the model as follows:
DISC ARALL
it
= [ARALL
it
? (ARALL
it?1
? (GROSSAR
it
/GROSSAR
it?1
) ? (DSO
it
/DSO
it?1
)] (2)
/Total Assets
it?1
28 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
where
DISC ARALL discretion in the allowance for doubtful accounts
ARALL the allowance for doubtful accounts
GROSSAR gross accounts receivable
DSO days sales outstanding, calculated as gross accounts
receivable divided by daily sales (sales/365)
Because prior research does not model the valuation allowance
for inventories, we model discretion in this account using an ap-
proach that is consistent with the approach we use to model dis-
cretion in the allowance for doubtful accounts. Speci?cally, we es-
timate discretion in the valuation allowance for inventories as the
difference between the current and prior year allowance, adjust-
ing the current year allowance for changes in gross inventories
and changes in the age or salability of the inventory. We proxy
for changes in the age or salability of the inventory using the days
inventory outstanding ratio. This ratio measures how long it takes
the company, on average, to sell its inventory.
13
Thus, we use the
following model to estimate discretion in the valuation allowance
for inventories:
DISC INVALL
it
= [INVALL
it
? (INVALL
it?1
? (GROSSINV
it
/GROSSINV
it?1
) ? (DIO
it
(3)
/DIO
it?1
)]/Total Assets
it?1
where
DISC INVALL discretion in the valuation allowance for inventory
INVALL the valuation allowance for inventory
GROSSINV gross inventory
DIO days inventory outstanding, calculated as gross inventory
divided by daily cost of sales (cost of sales/365)
Finally, for the valuation allowance for deferred tax assets, we
simply estimate discretion as the difference between the current
and prior year allowance, adjusting the current year allowance for
changes in gross deferred tax assets.
14
Thus, we use the follow-
13
Accounting Standards Codi?cation (ASC) Topic 330 states, “in accounting for in-
ventories, a loss shall be recognized whenever the utility of goods is impaired by
damage, deterioration, obsolescence, changes in price levels, or other causes. The
measurement of such losses shall be accomplished by applying the rule of pric-
ing inventories at the lower of cost or market.” Two methods for dealing with the
requirements in ASC Topic 330 have emerged in practice: (1) a direct (write-off)
method where ending inventory is recorded at net realizable value if it is less than
cost and the associated loss is recorded against cost of goods sold (COGS), and (2)
an indirect (allowance) method where a valuation allowance is established to record
any decline in value below cost, with the associated loss recorded against COGS. We
?nd that approximately 19% of the 7669 sample company-year observations with a
current or prior year inventory balance disclose an associated current or prior year
inventory valuation allowance. Thus, the use of the indirect (allowance) method ap-
pears to be common in practice.
14
In untabulated analysis, we re-estimate discretion in the deferred tax asset val-
uation allowance by not only adjusting for the change in the underlying deferred
tax asset, but also adjusting for the change in net operating losses (NOLs) because
Miller and Skinner (1998) suggest that NOLs are the most important component
of deferred tax assets for determining the size of the valuation allowance. We ac-
complish this by multiplying the prior year allowance by the ratio of the current to
prior year gross deferred tax assets and by the ratio of current to prior year NOLs.
When current or prior year NOLs are zero, we set this ratio equal to one. All results
using this measure of discretion in the deferred tax asset valuation allowance are
consistent with those tabulated.
ing model to estimate discretion in the valuation allowance for de-
ferred tax assets:
DISC DTAALL
it
= [DTAALL
it
? (DTAALL
it?1
? (GROSSDTA
it
(4)
/GROSSDTA
it?1
))]/Total Assets
it?1
where
DISC DTAALL discretion in the valuation allowance for deferred tax
assets
DTAALL the valuation allowance for deferred tax assets
GROSSDTA gross deferred tax assets
DISC ARALL, DISC INVALL, and DISC DTAALL are multiplied by
negative one to capture income-increasing discretion.
15
3.2.2. Measures of overall discretion in accruals
We employ two different models to estimate discretionary ac-
cruals. First, following Kothari, Leone, and Wasley (2005), we esti-
mate performance-matched discretionary accruals using the mod-
i?ed Jones (1991) model (Dechow, Sloan, & Sweeney, 1995) and
we incorporate controls for the asymmetric recognition of gains
and losses (Ball & Shivakumar, 2006). Speci?cally, we estimate dis-
cretionary accruals as the residual from the following model, esti-
mated by year and 2-digit SIC code industry:
TA
it
= ?
0
+?
1
(REV
it
?REC
it
) +?
2
(PPE
it
) +?
3
(CFO
it
)
+?
4
(NEG_CFO
it
) +?
5
(NEG_CFO
it
? CFO
it
) +?
it
(5)
where
TA total accruals (calculated as income before extraordinary
items less cash ?ow from operations)
REV the change in net revenues from year t ? 1 to year t
REC the change in net receivables from year t ? 1 to year t
PPE gross property, plant, and equipment
NEG_CFO an indicator variable set equal to one if CFO is less than
zero, and zero otherwise
15
To assess the validity of these measures, we performed internet searches to
identify instances where these speci?c allowance accounts were restated. Our
searches included key words “restate,” “restatement,” “allowance for doubtful,” “bad
debts,” “inventory reserve,” “valuation allowance,” and “deferred tax valuation al-
lowance.” We identi?ed nine such restatements in the ?rst few pages of the search
results (four, three, and two restatements relate to the allowance for doubtful
accounts, the valuation allowance for inventories, and the deferred tax asset al-
lowance, respectively). For each restatement identi?ed, we estimated discretion in
the affected allowance account separately using the original (unrestated) data and
the restated data. We then contrasted the estimated discretion in the allowance ac-
count using restated data relative to unrestated data. For all nine restatements ex-
amined, we ?nd a decrease in the estimated discretion in the allowance account as
a result of the restatement. These results suggest that the likelihood of a misstate-
ment in a given allowance account is increasing in the discretion in that account,
where the discretion in the account is estimated using Models (2) through (4).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 29
all other variables are as de?ned previously. With the exception of
NEG_CFO, all variables in Model (5) are scaled by total assets at
the beginning of the year. We identify company-year matches by
identifying the company-year observation within the same 2-digit
SIC code industry with the closest prior year return on assets. The
performance-matched discretionary accrual estimate is the differ-
ence between the company-year and the matched company-year
estimates.
Second, following recent studies (Francis, Pinnuck, & Watan-
abe, 2014; Hope, Thomas, & Vyas, 2013), we estimate performance-
adjusted discretionary accruals using the Jones (1991) model.
Speci?cally, we estimate discretionary accruals as the residual from
the following model, estimated by year and 2-digit SIC code indus-
try
16
:
TA
it
= ?
0
+?
1
(REV
it
) +?
2
(PPE
i
t ) +?
3
(ROA
it
) +e
it
(6)
where all variables are as previously de?ned. All variables in Model
(6) are scaled by total assets at the beginning of the year.
For both discretionary accruals measures, we separately exam-
ine signed and positive (income-increasing) discretionary accruals
to capture behavior that in?ates current period earnings. We also
examine the absolute value of discretionary accruals to capture
the magnitude of earnings management regardless of its direction
(Klein, 2002; Reynolds & Francis, 2000).
17
3.2.3. The association between disclosure transparency and earnings
management
We use the following model to test our ?rst hypothesis, where
the measure of earnings management is based on discretion in the
individual allowance accounts:
DISC ALL
it
= ?
0
+?
1
(Transparent_Ind
it
) +?
2
(Size
it
)
+?
3
(CFO
it
) +?
4
(ROA
it
) +?
5
(Loss
it
) +?
6
(Issue
it
)
+?
7
(Growth
it
) +?
8
(Ind_Growth
it
)
+?
9
(Leverage
it
) +?
10
(Age
it
)
+?
11
(Analyst_Coverage
it
) +?
12
(Inst_Holdings
it
)
+?
13
(Big4
it
) +?
14
(Specialist
it
) +?
15
(Tenure
it
)
+?
16
(ICMW
it
) +?
17
(HERF
it
) +?
18
(Lit
it
)
+?
19
(IMR
it
) +?
20
(ALL_Scaled
it
)
+?
j
(Industry FE) +?
k
(Year FE) +?
it
(7)
16
For both Models (5) and (6), we estimate discretionary accruals using all avail-
able companies with su?cient data in Compustat during the sample period. We ex-
clude industry-years with fewer than 20 observations and winsorize all continuous
variables at the 1% and 99% levels to mitigate the in?uence of outliers.
17
We also perform (untabulated) tests using a third broad accruals measure
based on the model developed by Dechow and Dichev (2002) and modi?ed by
McNichols (2002), Francis, LaFond, Olsson, and Schipper (2005), and Hope et al.
(2013). Speci?cally, following Hope et al. (2013), we use the absolute value of the
residual from the modi?ed Dechow and Dichev (2002) model as a proxy for accru-
als quality and we ?nd similar results using this measure.
where
DISC ALL one of three measures: (i) DISC ARALL – discretion in the
allowance for doubtful accounts estimated using Model (2),
(ii) DISC INVALL – discretion in the valuation allowance for
inventory estimated using Model (3), or (iii) DISC DTAALL
– discretion in the valuation allowance for deferred tax
assets estimated using Model (4);
Transparent_Ind an indicator variable set equal to one if the company
provides transparent disclosure of the individual allowance
or reserve account, either in a Schedule II or in the notes
to the ?nancial statements, and zero otherwise;
IMR the inverse Mills ratio estimated using Model (1);
ALL_Scaled one of three measures: (i) ARALL_Scaled, (ii)
INVALL_Scaled, or (iii) DTAALL_Scaled;
ARALL_Scaled ARALL, scaled by total assets;
INVALL_Scaled INVALL, scaled by total assets;
DTAALL_Scaled DTAALL, scaled by total assets;
Industry FE industry ?xed effects, where industries are de?ned
following Ashbaugh, LaFond, and Mayhew (2003)
a
Year FE year ?xed effects
a
We use SIC codes to de?ne industries as follows: agriculture (0100-0999),
mining and construction (1000-1999, excluding 1300-1399), food (2000-2111), tex-
tiles and printing/publishing (2200-2799), chemicals (2800-2824; 2840-2899), phar-
maceuticals (2830-2836), extractive (1300-1399; 2900-2999), durable manufactur-
ers (3000-3999, excluding 3570-3579 and 3670-3679), transportation (4000-4899),
retail (5000-5999), services (7000-8999, excluding 7370-7379), computers (3570-
3579; 3670-3679; 7370-7379), and utilities (4900-4999).
all other variables are as previously de?ned. We estimate Model
(7) separately for each of the three allowance accounts.We use the
following model to test our ?rst hypothesis, where the measure of
earnings management is based on one of the two broad accruals
measures:
DiscAcc
it
= ?
0
+?
1
(Transparent_All
t
) +?
2
(Size
it
) +?
3
(CFO
it
)
+?
4
(ROA
it
) +?
5
(Loss
it
) +?
6
(Issue
it
)
+?
7
(Growth
it
) +?
8
(Ind_Growth
it
)
+?
9
(Leverage
it
) +?
10
(Age
it
)
+?
11
(Analyst_Coverage
it
) +?
12
(Inst_Holdings
it
)
+?
13
(Big4
it
) +?
14
(Specialist
it
) +?
15
(Tenure
it
)
+?
16
(ICMW
it
) +?
17
(HERF
it
) +?
18
(Lit
it
)
+?
19
(IMR
it
) +?
j
(Industry FE) +?
k
(Year FE) +?
it
(8)
where
DiscAcc one of two measures of discretionary accruals as estimated using
Model (5) or (6); for both measures, we estimate Model (8)
separately for signed, positive (income-increasing) only, and
absolute value discretionary accruals
a
a
To examine positive (income-increasing) discretionary accruals, we use tobit
when estimating Model (8).
all other variables as previously de?ned. The coe?cient of interest
in Model (7) is ?
1
, the coe?cient on Transparent_Ind, and the co-
e?cient of interest in Model (8) is ?
1
, the coe?cient on Transpar-
ent_All. Consistent with Hypothesis 1, we expect the coe?cients on
Transparent_Ind and Transparent_All to be negative and signi?cant
in all speci?cations of Models (7) and (8).
30 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
Table 1
Sample selection.
ARALL INVALL DTAALL
Panel A: Samples for tests of discretion in individual allowance and reserve accounts
Company-year observations that provide a Schedule II 4405 1159 1754
Company-year observations that disclose the allowance or reserve balance transparently in the notes to the ?nancial statements 1790 63 170
Company-year observations that provide non-transparent disclosure for the allowance or reserve account 2382 216 630
Less: Company-year observations where the allowance is immaterial (i.e., where the prior year balance is less than once cent per share) (1987) (206) (94)
Sample for Tests of H1 6590 1232 2460
Less: Company-year observations where non-transparent disclosure is provided for the allowance or reserve account (1351) (141) (618)
Sample for Tests of H2 5239 1091 1842
Panel B: Samples for tests of overall discretion in accruals
Sample for
Tests of H1
Sample for
Tests of H2
Companies that provide transparent disclosures (Transparent_All = 1)
Company-year observations that provide a Schedule II 4664 4664
Company-year observations that disclose all material (disclosed) allowance and reserve balances transparently in the notes to
the ?nancial statements
1941 1941
Companies that provide non-transparent disclosures (Transparent_All = 0)
Company-year observations that provide non-transparent disclosure for all material (disclosed) allowance and reserve balances 2854 NA
Less: Company-year observations where prior year aggregate allowance and reserve balances are less than one cent per share (2032) (1009)
Samples for tests of overall discretion in accruals 7427 5596
The samples consist of company-year observations with necessary data from Compustat and Audit Analytics to construct the variables in our models. For all samples, we
exclude company-year observations where transparent disclosures are provided for some, but not all, material (disclosed) valuation allowance and reserve accounts.
Models (7) and (8) include controls for company size (Size),
age (Age), sales growth (Growth), and industry sales growth
(Ind_Growth) because prior research indicates that these company
characteristics affect the level of discretionary accruals (Anthony &
Ramesh, 1992; Becker, DeFond, Jiambalvo, & Subramanyam, 1998;
Myers, Myers, & Omer, 2003). To control for the effects of com-
pany performance, we include cash ?ow from operations (CFO)
and return on assets (ROA) (Dechow et al., 1995). Following prior
research, we also control for the issuance of debt and equity ?-
nancing (Issue) (Rangan, 1998), whether analysts cover the com-
Table 2
Descriptive statistics.
Variable
N Mean P25 Median P75 N Mean P25 Median P75
Diff
in
mean Transparent_Ind = 1 Transparent_Ind = 0
DISC ARALL 5239 0.000 ?0.001 0.000 0.002 1351 0.001 ?0.001 0.000 0.002 ?0.001
DISC INVALL 1091 0.001 ?0.003 0.000 0.003 141 0.016 ?0.003 0.000 0.002 ?0.014
?
DISC DTAALL 1842 0.001 ?0.002 0.000 0.001 618 0.005 0.000 0.000 0.000 ?0.004
?
Transparent_All = 1 Transparent_All = 0
DiscAcc
a
5596 0.043 ?0.018 0.021 0.073 1831 0.097 ?0.015 0.033 0.125 ?0.054
???
AbsDiscAcc
a
5596 0.125 0.019 0.048 0.107 1831 0.223 0.024 0.067 0.181 ?0.098
???
DiscAcc
b
5596 ?0.013 ?0.063 ?0.010 0.046 1831 ?0.003 ?0.060 0.006 0.068 ?0.010
??
AbsDiscAcc
b
5596 0.088 0.024 0.055 0.108 1831 0.129 0.027 0.064 0.135 ?0.042
???
Size 5596 6.542 5.222 6.500 7.808 1831 5.613 3.614 5.407 7.587 0.929
???
CFO 5596 0.078 0.043 0.091 0.137 1831 ?0.107 ?0.008 0.074 0.129 0.184
???
ROA 5596 ?0.004 ?0.028 0.034 0.075 1831 ?0.153 ?0.092 0.017 0.068 0.148
???
Loss 5596 0.327 0.000 0.000 1.000 1831 0.430 0.000 0.000 1.000 ?0.104
???
Issue 5596 0.125 0.000 0.000 0.000 1831 0.140 0.000 0.000 0.000 ?0.015
?
Growth 5596 0.127 ?0.093 0.038 0.154 1831 0.457 ?0.111 0.035 0.210 ?0.331
Ind_Growth 5596 0.029 ?0.038 0.029 0.096 1831 0.036 ?0.038 0.047 0.128 ?0.007
??
Leverage 5596 0.234 0.020 0.177 0.338 1831 0.367 0.015 0.184 0.383 ?0.133
???
Age 5596 23.200 12.000 18.000 30.000 1831 20.502 11.000 17.000 27.000 2.698
???
Analyst_Coverage 5596 0.786 1.000 1.000 1.000 1831 0.559 0.000 1.000 1.000 0.227
???
Inst_Holdings 5596 0.549 0.203 0.636 0.871 1831 0.335 0.000 0.172 0.699 0.214
???
Big 4 5596 0.773 1.000 1.000 1.000 1831 0.590 0.000 1.000 1.000 0.183
???
Specialist 5596 0.234 0.000 0.000 0.000 1831 0.190 0.000 0.000 0.000 0.045
???
Tenure 5596 10.460 5.000 8.000 13.000 1831 9.050 4.000 7.000 12.000 1.409
???
ICMW 5596 0.053 0.000 0.000 0.000 1831 0.102 0.000 0.000 0.000 ?0.048
???
HERF 5596 0.070 0.034 0.047 0.078 1831 0.066 0.031 0.042 0.084 0.004
??
Lit 5596 0.146 0.000 0.000 0.000 1831 0.193 0.000 0.000 0.000 ?0.047
???
a
Estimated using Model (5).
b
Estimated using Model (6).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 31
Table 3
The determinants of transparent disclosure.
Variable Pred.
DV = Transparent_All
Coe?cient estimate p-Value
Intercept ? 0.098 .133
Size + 0.016
?
.076
CFO + 0.326
???
.001
ROA + 0.076 .133
Loss + 0.033 .206
Issue + ?0.052 .849
Growth ? ?0.003 .310
Ind_Growth ? ?0.303
??
.030
Leverage ? ?0.109
???
.007
Age + 0.003
??
.023
Analyst_Coverage + 0.175
???
Weexaminetherelationbetweenthetransparencyofdisclosuresaboutactivityinvaluationallowanceandreserveaccountsandaccruals-basedearningsmanagement.Weclassifydisclosuresasbeingtranspar-entiftheyprovidedetailedinformationaboutactivityintheallowanceandreserveaccountsduringthefiscalperiod.Wefindstrongevidencethattheextentofaccruals-basedearningsmanagementisloweramongcompanieswithtransparentdisclosuresthanamongcompanieswithouttransparentdisclosures.
Accounting, Organizations and Society 46 (2015) 23–38
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Disclosure transparency about activity in valuation allowance and
reserve accounts and accruals-based earnings management
Cory A. Cassell
a
, Linda A. Myers
a,?
, Timothy A. Seidel
b
a
University of Arkansas, Fayetteville, AR, United States
b
Utah State University, Logan, UT, United States
a r t i c l e i n f o
Article history:
Received 2 June 2014
Revised 18 March 2015
Accepted 19 March 2015
Available online 16 April 2015
Keywords:
Disclosure transparency
Valuation allowances and reserves
Schedule II
Earnings management
Discretionary accruals
Management discretion
a b s t r a c t
We examine the relation between the transparency of disclosures about activity in valuation allowance
and reserve accounts and accruals-based earnings management. We classify disclosures as being transpar-
ent if they provide detailed information about activity in the allowance and reserve accounts during the
?scal period. We ?nd strong evidence that the extent of accruals-based earnings management is lower
among companies with transparent disclosures than among companies without transparent disclosures.
We also investigate whether the extent of accruals-based earnings management is lower for companies
that provide transparent disclosures in one comprehensive schedule (i.e., the Schedule II) relative to those
that provide transparent disclosures spread throughout the notes to the ?nancial statements. Although
regulators have expressed concern that the omission of a Schedule II could indicate a greater likelihood
of earnings management, our results indicate that it is the omission of transparent disclosures rather
than the omission of a comprehensive schedule outlining activity in the allowance and reserve accounts
that affects earnings management. Our ?ndings suggest that regulators, auditors, and investors should
consider subjecting companies that fail to provide transparent disclosures to additional scrutiny.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Accruals-based earnings management can be costly because
questionable accounting practices are likely to be scrutinized by
external auditors, investors, the board of directors, regulators,
and other stakeholders. Valuation allowances and reserves pro-
vide managers with substantial ?exibility to manage earnings be-
cause they are based on subjective estimates and are evaluated
at higher levels of materiality, making them inherently di?cult to
audit (Gri?th, Hammersley, & Kadous, 2013; Peecher, Schwartz,
& Solomon, 2007; Peecher, Solomon, & Trotman, 2013). In addi-
tion, any differences identi?ed by the auditor are more likely to
be waived when the underlying accruals are more subjective (Joe,
Wright, & Wright, 2011; Knapp, 1987; Wright & Wright, 1997).
Thus, prior research suggests that companies are more likely to at-
tempt to manage reported earnings using these types of accounts.
1
We examine the relation between the transparency of
?
Corresponding author.
1
For example, Nelson, Elliot, and Tarpley (2002) use a ?eld-based questionnaire
in which 253 auditors from one Big 5 audit ?rm described 515 speci?c instances in
which they believed that their clients were attempting to manage earnings. The au-
thors indicate that “by far the most frequently identi?ed attempts involve reserves”
(Nelson et al., 2002, 176).
disclosure related to activity in valuation allowance and reserve
accounts and accruals-based earnings management, where disclo-
sures are classi?ed as being transparent if they provide detailed
information about activity in the accounts during the ?scal period
(e.g., information about the current period expense accrual, write-
offs, etc. for the allowance for doubtful accounts).
2
The U.S. Securi-
ties and Exchange Commission (SEC) has long required that activ-
ity in subjective accrual accounts be disclosed. Speci?cally, within
Regulation S-X, which prescribes the format and content of ?nan-
cial reports ?led under the Securities Act of 1933 and the Securi-
ties Exchange Act of 1934, Rule 5-04 and Rule 12-09 require that
companies ?le a Schedule II.
3
The Schedule II should provide the
beginning and ending balances and the current period activity in
2
Barth and Schipper (2008) de?ne ?nancial reporting transparency as the extent
to which ?nancial reports reveal an entity’s underlying economics in a way that is
understandable by those using the ?nancial reports. Bushman, Piotroski, and Smith
(2004) de?ne corporate transparency as the widespread availability of ?rm-speci?c
information about publicly listed ?rms to parties outside of the ?rm. Our de?nition
of transparency is speci?c to our setting and relates to just one aspect of overall re-
porting transparency, but it is consistent with de?nitions set forth in prior research.
3
Schedule II is a supplemental schedule ?led with the company’s form 10-K and
is typically located after the footnotes to the ?nancial statements. Laws requiring
the use of Schedule II can be traced to 37 FR 14602 (passed on July 21, 1972) and
re-designated and amended in 45 FR 63679 (on September 25, 1980).http://dx.doi.org/10.1016/j.aos.2015.03.004
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
24 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
all material valuation allowances and reserves. The SEC exempts
companies in certain industries,
4
companies with immaterial bal-
ances in the valuation allowances and reserves, and companies
that provide comparable disclosures in the notes to the ?nancial
statements from the Schedule II requirement. Thus, under current
SEC regulations, companies are required to provide transparent dis-
closures for material allowance and reserve balances in either a
Schedule II or in the notes to the ?nancial statements.
5
Although the SEC requires transparent disclosure for material
allowance and reserve balances, we ?nd that compliance is lack-
ing. Speci?cally, when we examine companies where the aggre-
gate prior year balance of the allowance for doubtful accounts, in-
ventory valuation allowance, and deferred tax asset valuation al-
lowance is greater than one cent per share, thus removing compa-
nies that would not be able to increase current year earnings per
share (EPS) by at least one cent even if they were to fully elimi-
nate these accruals, we ?nd that approximately 30% fail to provide
transparent disclosures (either in a Schedule II or in the notes to
the ?nancial statements).
6
Although the following statement made
by Ronald A. Kima, a former Assistant Chief Accountant for the
SEC, focuses on problems associated with Schedule II compliance,
it captures the potential implications of non-transparent disclosure
in general,
7
Unfortunately, despite such schedule being required of most
public companies, few companies seemingly fully comply. The
absence of otherwise required data, or worse the outright omis-
sion of the schedule in its entirety, should raise investor con-
cerns that a company may be engaging in some degree of inap-
propriate earnings management… Any immateriality assertion
by a company’s management that has been predicated exclu-
sively on balance sheet measures is, at best, inappropriate and,
at worst, an attempt to conceal inappropriate earnings manage-
ment practices.
Consistent with claims in Kima (2007), we expect that com-
panies will use allowances and reserves to manage earnings if
the probability of detection is su?ciently low and if the com-
bined magnitude of these accrual accounts is su?ciently large
(so that their manipulation can have a meaningful effect on re-
ported earnings). With respect to the ?rst condition, we posit that
non-transparent disclosures about highly subjective accrual ac-
counts provide managers with ?exibility to in?uence the market’s
perceptions of earnings. Because market participants face limita-
tions when processing accounting information (Hirshleifer & Teoh,
2003), they may be less likely to see through the earnings man-
agement when companies disclose allowance and reserve accounts
non-transparently. Although prior research does not investigate the
4
Excluded companies include registered investment companies, insurance com-
panies, bank holding companies and banks, and brokers and dealers.
5
Note that because Rule 5-04 does not provide a de?nition of materiality, mate-
riality is left to management’s discretion. In 2000, the SEC proposed a rule which
would have required a disclosure similar to Schedule II in the notes to the ?nancial
statements, rather than as a supplementary schedule (SEC, 2000). The relocation of
this disclosure would have provided increased prominence and auditor responsibil-
ity for the disclosure. The majority of comments received on the SEC’s proposed
rule opposed the proposal. The SEC did not ?nalize the proposed rule, so Rule 5-
04 of Regulation S-X remains applicable and materiality thresholds continue to be
subjective.
6
To further support this ?nding, we use the Audit Analytics SEC comment let-
ter database to identify SEC comment letters with speci?c reference to Schedule II
or to the allowances most frequently included in a Schedule II. Our search resulted
in many examples of the SEC requesting increased disclosure of the activity in al-
lowance and reserve accounts (in accordance with Rule 5-04 of Regulation S-X).
7
See SEC Schedule II – Visibility into the integrity of reported results. AICPA
CPA Insider Newsletter (October 1, 2007). Retrieved fromhttp://www.cpa2biz.com/
Content/media/newsletters/cpainsider/cpainsider071001.jsp.
association between accruals-based earnings management and the
transparency of accruals disclosures, evidence suggests that man-
agers use the ?exibility in disclosure rules to engage in real earn-
ings management (Lee, Petroni, & Shen, 2006) and that transparent
disclosures facilitate investors’ ability to detect earnings manage-
ment (Hirst & Hopkins, 1998).
Because disclosures of allowances and reserves must be hand
collected, we focus our analyses on the three accounts most com-
monly included in a Schedule II – the allowance for doubtful ac-
counts, the valuation allowance for deferred tax assets, and the val-
uation allowance for inventories.
8
We hand collect disclosures for
all companies in non-exempt industries that include at least one
of the three allowance and reserve balances in a Schedule II, in the
notes to the ?nancial statements, or parenthetically in the balance
sheet in their 2008, 2009, or 2010 annual reports.
We ?rst investigate whether disclosure transparency for a given
allowance or reserve account affects discretion in that account.
Following Marquardt and Wiedman (2004), we estimate the ex-
pected balance in each account by multiplying the prior year bal-
ance by the current year growth rate. To further re?ne this mea-
sure for the allowance for doubtful accounts, we also multiply
by the growth in days sales outstanding (to capture the change
in collectability). Similarly, for the inventory valuation allowance,
we multiply by the growth in days inventory outstanding (to cap-
ture the change in inventory turnover). Our measure of the discre-
tion in each allowance and reserve account is the difference be-
tween the actual account balance and the expected balance. For
each account examined, we limit the sample to companies where
the disclosed beginning balance is at least one cent per share.
We contrast discretion in the individual allowance or reserve ac-
count for companies that provide transparent disclosure (either in
a Schedule II or in the notes to the ?nancial statements) with that
for companies that provide non-transparent disclosure. We ?nd
that discretion in the each of the three accounts, the allowance
for doubtful accounts, the valuation allowance for inventories,
and the valuation allowance for deferred tax assets is greater
(i.e., more income-increasing) when these accounts are disclosed
non-transparently.
Although we focus on the three allowance accounts most com-
monly included in a Schedule II (and not on all valuation and
qualifying accounts that could be included in the schedule), we
supplement our primary tests by examining broad measures of
earnings management. In these tests, we contrast overall discre-
tion in accruals for companies that provide transparent disclo-
sures for all material (disclosed) allowance and reserve balances
(either in a Schedule II or in the notes to the ?nancial state-
ments) with that for companies that provide non-transparent dis-
closures for all disclosed allowance and reserve balances. We ?nd
that companies that provide transparent disclosures report smaller
signed, absolute value, and positive (income-increasing) discre-
tionary accruals than do companies that provide non-transparent
disclosures.
Having established a negative association between disclosure
transparency and the extent of earnings management in subjec-
tive accrual accounts, we next investigate whether the placement
of transparent disclosures matters. As evidenced by the rules sur-
rounding Schedule II and by subsequent (defeated) regulatory pro-
posals which would have required that a Schedule II be included in
8
In addition to being the accounts most commonly included in Schedule II dis-
closures, these accounts have also been the focus of prior research examining earn-
ings management in speci?c accounts. See, for example, McNichols and Wilson
(1988), Schrand and Wong (2003), Frank and Rego (2006), and Cecchini, Jackson,
and Liu (2012).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 25
the notes to the ?nancial statements, regulators appear to believe
that the transparent disclosure of all accrual accounts in a single
schedule will facilitate and enhance investors’ ability to process the
information relative to comparable, but separate, disclosure in the
notes. These beliefs are consistent with context theory (Medin &
Schaffer, 1978), which suggests that the presentation of an isolated
piece of information can lead to different interpretations relative to
when this information is combined with other information. These
beliefs are also consistent with the proximity compatibility princi-
ple (Carswell & Wickens, 1996; Wickens & Carswell, 1995), which
suggests that information processing can be enhanced when the
information is displayed in close proximity, and with cognitive load
theory (Sweller, 1988, 1989), which suggests that individuals will
be unable or unwilling to exert the cognitive effort necessary to
integrate information presented separately (Dietrich, Kachelmeier,
Kleinmuntz, & Linsmeier, 2001; Krische, 2005).
9
Thus, we expect
that the transparent disclosure of allowance and reserve accounts
in a single schedule (i.e., in a Schedule II) will aid users’ under-
standing of the (collective) information about activity in allowance
and reserve accounts, which should facilitate the detection of earn-
ings management in these accounts. Because managers have little
incentive to manage earnings if their actions will be detected, we
expect less earnings management among companies that provide a
Schedule II.
For tests using the estimated discretion in individual allowance
and reserve accounts, we limit our samples to companies that
provide transparent disclosures for the given account (either in a
Schedule II or in the notes to the ?nancial statements) and have
a prior year account balance greater than one cent per share. For
tests using the broad accruals measures, we limit our sample to
companies that provide transparent disclosures for all disclosed al-
lowance and reserve accounts where the aggregate prior year bal-
ance of these accounts exceeds one cent per share. We ?nd no sig-
ni?cant difference in the estimated discretion (either in individ-
ual accounts or in the broad accruals measure) between compa-
nies that provide a Schedule II and companies that provide trans-
parent disclosures in the notes to the ?nancial statements. Our in-
ability to ?nd support for this prediction could be due to a lack
of task complexity, the effect of a different (and competing) as-
pect of context being important in our setting, or a lack of power
in our tests. We discuss these possibilities in Section ‘Results
summary’.
We expect our study to be of interest to standard setters and
?nancial statement users as they consider proposals put forth in
the International Accounting Standard Board (IASB) and the Finan-
cial Accounting Standards Board (FASB) joint project related to ?-
nancial statement presentation. One such proposal would require
note disclosure of detailed activity in balance sheet accounts to fa-
cilitate user understanding of the current period change in posi-
tion (FASB, 2010). Our results suggest that the proposed require-
ment could be bene?cial because we document a negative asso-
ciation between the transparency of disclosure and earnings man-
agement. Our study should also be of interest to regulators, au-
ditors, and investors because our ?ndings suggest that the dis-
closure transparency of subjective accounts could be used as a
mechanism to identify companies that may warrant additional
scrutiny.
9
Several experimental studies provide evidence that the placement of equiva-
lent information affects investor and analyst judgments. See, for example, Hopkins
(1996), Hirst and Hopkins (1998), Maines and McDaniel (2000), and Hodge, Hop-
kins, and Wood (2010). Also see Libby and Emett (2014) for a broad overview of
the literature related to placement effects.
We also provide empirical evidence that addresses concerns
expressed by former SEC Assistant Chief Accountant Ronald
A. Kima, who suggests that the omission of a Schedule II
could indicate a greater likelihood of earnings management. We
?nd evidence that the omission of transparent disclosure, and
not necessarily the omission of a Schedule II, affects earnings
management.
The remainder of this paper is organized as follows: the next
section provides a review of related prior literature and develops
our hypotheses. The third section describes our empirical design.
We discuss the sample and present results from our analyses in
the fourth section. The ?nal section provides our conclusions.
2. Prior literature and development of hypotheses
Greater information asymmetry can provide companies greater
?exibility to manage earnings (Schipper, 1989). For example,
Richardson (2000) ?nds that increased information asymmetry,
measured by the bid-ask spread and analyst forecast dispersion, is
associated with higher levels of discretionary accruals. In addition,
Hodder and Hopkins (2014) ?nd that bank managers’ demand for
reporting opacity lead them to oppose accounting rules that would
increase disclosures about ?nancial instruments.
Some studies examine the relation between the transparency
of speci?c disclosures and earnings management (Hirst & Hopkins,
1998; Lee et al., 2006). These studies, however, focus exclusively
on the relation between the transparency of comprehensive in-
come disclosures and “real” earnings management decisions (i.e.,
the opportunistic selling of available-for-sale securities) as well as
the market’s ability to detect this real earnings management. Us-
ing archival data, Lee et al. (2006) ?nd that property-liability in-
surers with a history of selectively selling available-for-sale securi-
ties chose the less transparent format for disclosing comprehen-
sive income in the ?rst year of Statement of Financial Account-
ing Standards No. 130’s adoption. With respect to the detection of
earnings management, Hirst and Hopkins (1998) ?nd experimental
evidence that transparent comprehensive income disclosures en-
able buy-side ?nancial analysts to detect opportunistic selling of
available-for-sale securities and that they adjust their target prices
accordingly. Collectively, this stream of research suggests that in-
creased disclosure transparency about certain accounts or trans-
actions will reduce the likelihood that earnings are managed us-
ing real earnings management because increased transparency fa-
cilitates the detection of earnings management by market partici-
pants.
Lobo and Zhou (2001) and Shaw (2003) examine the relation
between disclosure quality and accruals-based earnings manage-
ment but they use a measure of overall disclosure quality rather
than a measure of the transparency of disclosures for speci?c ac-
cruals that may be used to manage earnings. Speci?cally, they
measure disclosure quality using the Association for Investment
Management Research (AIMR) scores. These scores are based on
industry-speci?c analyst evaluations of overall disclosure quality
and analysts consider annual and quarterly SEC ?lings, along with
other published information. Both Lobo and Zhou (2001) and Shaw
(2003) ?nd a negative association between AIMR scores and dis-
cretionary accruals, suggesting that companies with higher over-
all disclosure quality report lower discretionary accruals. However,
Shaw (2003) ?nds that this conservatism is limited to years with
good news (i.e., positive stock returns or cash ?ows from opera-
tions). In years with bad news, companies with higher AIMR scores
report higher levels of income-increasing discretionary accruals.
Thus, Shaw’s ?ndings suggest that companies with higher overall
disclosure scores smooth earnings more aggressively than do com-
panies with lower overall disclosure scores.
26 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
We extend prior research by investigating whether the dis-
closure transparency about speci?c accruals is associated with
accruals-based earnings management. While prior research pro-
vides evidence that real earnings management is lower in the pres-
ence of more transparent comprehensive income disclosures, prior
research does not provide evidence about accruals-based earnings
management in the presence of more transparent disclosure of
subjective accrual accounts. Additionally, regulators have expressed
concern that a lack of transparency in these accounts could be re-
lated to earnings management.
10
Because subjective accruals are evaluated at higher levels of
materiality (FASB, 1980) and any audit adjustments proposed to
these accruals are more likely to be waived (Joe et al., 2011; Knapp,
1987; Wright & Wright, 1997), auditors may not be effective at
identifying and constraining earnings management in these ac-
counts if disclosures about the activity in these accounts are not
transparent.
11
Because greater transparency about these accounts
could aid in the detection of accruals-based earnings management,
we expect companies that provide transparent disclosures about
activity in accrual accounts to engage in less earnings management
through these accounts. Our ?rst hypothesis, stated in the alterna-
tive form, is as follows:
Hypothesis 1. The extent of accruals-based earnings management
will be lower for companies that provide transparent disclosures
about activity in accrual accounts.
We also expect that earnings management may differ when
companies provide a Schedule II versus when companies provide
similar information about the activity in subjective accruals else-
where in the notes to the ?nancial statements. That is, there may
be differences in earnings management behavior within the set of
companies that provide transparent disclosures about activity in al-
lowance and reserve accounts.
Medin and Schaffer (1978) introduce context theory, which sug-
gests that the presentation of an isolated piece of information
can lead to different interpretations relative to when this infor-
mation is combined with other information. In addition, the prox-
imity compatibility principle (Carswell & Wickens, 1996; Wick-
ens & Carswell, 1995) coupled with cognitive load theory (Sweller,
1988, 1989) suggest that information processing can be enhanced
when the information is displayed in close proximity. Consistent
with these theories, experimental research provides evidence that
disclosure format and placement affects investors’ ability to pro-
cess information and/or detect earnings management (Hirst & Hop-
kins, 1998; Hodge et al., 2010; Krische, 2005; Maines & McDaniel,
2000). For example, Hirst and Hopkins (1998) show that profes-
sional investors and analysts often overlook information about un-
realized gains and losses on marketable securities when this infor-
mation is presented in a statement of stockholders’ equity rather
than in a statement of comprehensive income, and Maines and
McDaniel (2000) ?nd a similar effect among nonprofessional in-
vestors. Krische (2005) ?nds that investors adjust their evaluations
10
For example, Kima (2007) states, [g]iven their inherent uncertainty, contra-asset
accounts and liability reserves typically constitute the soft underbelly of any set of
consolidated ?nancial statements. Thus, contra asset accounts and liability reserves
pose the greatest accounting challenges to ethical managements as they attempt to
derive the best possible estimates. Unfortunately, ethically-challenged managements
often exploit the softness of contra-asset accounts and liability reserves to engage
in inappropriate earnings management, i.e., tucking away earnings for a rainy day.
11
Alternatively, increased disclosure transparency may not affect accruals-based
earnings management behavior because auditors have visibility into these accounts
regardless of the level of disclosure transparency and auditors have incentives to
constrain accruals-based earnings management in order to adhere to SEC guidelines
and auditing standards (PCAOB, 2003, 2010; SEC, 1999).
of current period earnings to a greater extent when transparent
disclosures about prior period transitory gains and losses are re-
peated in the current year versus when they are made only in the
prior year. Finally, Hodge et al. (2010) ?nd that the proximity of
?nancial information improves nonprofessional investors’ ability to
process such information, where improved information processing
is re?ected in more accurate cash ?ow forecasts.
This prior evidence suggests that the transparent disclosure
of allowance and reserve accounts in a single schedule (i.e., in
a Schedule II) will aid users’ understanding of the (collective)
information about activity in allowance and reserve accounts,
which should facilitate the detection of earnings management in
these accounts. Because managers have little incentive to man-
age earnings if their actions will be detected, we would expect
less earnings management among companies that provide a Sched-
ule II. Our second hypothesis, stated in the alternative form, is as
follows:
Hypothesis 2. The extent of accruals-based earnings management
will be lower for companies that provide a Schedule II than for
companies that provide comparable information in the notes to the
?nancial statements.
3. Empirical design
3.1. Self-selection bias
We model the decision to disclose allowances and reserves
transparently using variables that should in?uence the quality
of voluntary disclosure. Because Lang and Lundholm (1993) ?nd
that analyst ratings of companies’ disclosure practices (using AIMR
scores) are higher for larger companies, companies with better per-
formance, and companies that issue securities in the current or
future periods, we include measures of company size (Size), per-
formance (CFO, ROA, Loss), and the issuance of new securities (Is-
sue). We include measures of company sales growth (Growth) and
industry sales growth (Ind_Growth) because managers should in-
crease disclosure when it is more di?cult for analysts to assess
value based on historical numbers (Frankel, Johnson, & Skinner,
1999). Because prior research ?nds a negative association between
leverage and voluntary disclosure (Chow & Wong-Boren, 1987; Eng
& Mak, 2003), we include leverage (Leverage). Prior work also ?nds
that companies with greater information asymmetry are likely to
bene?t from additional disclosure (Chen, DeFond, & Park, 2002;
Gu & Li, 2007; Lang, 1991) so we follow Gu and Li (2007) and in-
clude company age (Age) and analyst coverage (Analyst_Coverage)
to proxy for the extent of information asymmetry. We include in-
stitutional ownership (Inst_Holdings) because prior research ?nds
a positive association between AIMR scores and institutional own-
ership (Bushee & Noe, 2000). We include a number of variables
to proxy for auditor quality (Big 4, Specialist, and Tenure) be-
cause prior work ?nds a positive association between auditor qual-
ity and disclosure quality (Almutairi, Dunn, & Skantz, 2009; Dunn
& Mayhew, 2004). Because companies committed to high-quality
internal controls might also be committed to high-quality disclo-
sure, we control for internal control material weaknesses (ICMW).
We include the Her?ndahl index to proxy for proprietary costs
(HERF) because companies should be less likely to commit to high-
quality disclosures when proprietary costs are high, and we in-
clude an indicator for litigious industry membership (Lit) because
managers might provide high-quality disclosures in an effort to
avoid litigation (Field, Lowry, & Shu, 2005; Skinner, 1994, 1997).
Finally, because Lennox, Francis, and Wang (2012) recommend the
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 27
use of an exclusion restriction when estimating selection models,
we also include the level of investment in research and devel-
opment (R&D). Research and development activities should affect
the proprietary costs of disclosure but should not affect the dis-
cretionary portion of current period working capital accruals be-
cause research and development activities relate to future product
or service offerings for which no sales have yet been made and no
costs have yet been capitalized. Our ?rst-stage probit model is as
follows:
Pr(Transparent_All
it
= 1) = ?
0
+?
1
(Size
it
) +?
2
(CFO
it
)
+?
3
(ROA
it
) +?
4
(Loss
it
) +?
5
(Issue
it
)
+?
6
(Growth
it
) +?
7
(Ind_Growth
it
)
+?
8
(Leverage
it
) +?
9
(Age
it
)
+?
10
(Analyst_Coverage
it
)
+?
11
(Inst_Holdings
it
) +?
12
(Big4
it
)
+?
13
(Specialistit) +?
14
(Tenure
it
)
+?
15
(ICMW
it
) +?
16
(HERF
it
)
+?
17
(Lit
it
) +?
18
(R&D
it
) (1)
where
Transparent_All an indicator variable set equal to one if the company provides transparent disclosures for all material (disclosed) valuation allowances and
reserves, either in a Schedule II or in the notes to the ?nancial statements, and zero otherwise
Size the natural log of total assets
CFO operating cash ?ow in the current year divided by total assets at the beginning of year
ROA return on assets, measured as net income divided by total assets
Loss an indicator variable set equal to one if net income is less than zero, and zero otherwise
Issue an indicator variable set equal to one if the company issued equity or debt in the year, and zero otherwise
Growth sales growth measured as sales less prior year sales, scaled by prior year sales
Ind_Growth industry sales (for the 2-digit SIC code industry) divided by prior year industry sales
Leverage long-term debt plus the current portion of long-term debt, divided by total assets
Age the number of years to date during which total assets is reported in Compustat
Analyst_Coverage an indicator variable equal to one if at least one analyst covers the company (from I/B/E/S), and zero otherwise
Inst_Holdings the percentage of company stock held by institutions
Big4 an indicator variable set equal to one if the auditor is from the Big 4, and zero otherwise
Specialist an indicator variable set equal to one if the auditor is an industry specialist, de?ned following Reichelt and Wang (2010) as an auditor
whose audit fee market share in the 2-digit SIC code exceeds 30% at the national level, and zero otherwise
Tenure the number of consecutive years to date of the auditor–client relationship
ICMW an indicator variable set equal to one if at least one material weakness in internal controls is disclosed in the year, and zero otherwise
HERF the Her?ndahl index estimated by summing the squared market shares of all companies in the 2-digit SIC code industry, where market
share is measured using sales
Lit an indicator variable set equal to one if the company operates in a high litigation risk industry (SIC codes 2833-2836, 3570-3577,
3600-3674, 5200-5961, and 7370), and zero otherwise
R&D research and development expense scaled by total assets
i and t company and year indicators
We include the inverse Mills ratio (IMR) derived from estimat-
ing this model in our second-stage models to mitigate concerns
relating to selection bias.
12
3.2. Tests of earnings management
We perform tests of our hypotheses using two approaches
to identify earnings management behavior. First, we investigate
whether companies that provide transparent disclosures for indi-
vidual allowance and reserve accounts use less discretion in those
accounts relative to companies that provide non-transparent dis-
12
Because many of the factors that in?uence managers’ disclosure choices are un-
observable and some of these unobservable factors are likely to be correlated with
the extent of earnings management, we control for these unobservable factors by
including the inverse Mills ratio in the second stage of a two-stage Heckman proce-
dure. Tucker (2010) argues that the two-stage Heckman procedure is useful in set-
tings where managers have considerable discretion and where the decision-making
process is more opaque to researchers. Relative to propensity score matching (PSM),
the Heckman approach is preferred in our setting because PSM relies exclusively on
observables to address the self-selection problem.
closures. Second, because we do not examine all valuation and
qualifying accounts, we supplement our primary tests by examin-
ing broad measures of earnings management. Speci?cally, we in-
vestigate whether companies that provide transparent disclosure
of all material (disclosed) allowances and reserves report discre-
tionary accruals that are less positive (i.e., less income-increasing)
or smaller in magnitude (absolute value) than companies that pro-
vide non-transparent disclosures. We discuss the construction of
our earnings management measures and the associated empirical
models in the remainder of this section.
3.2.1. Measures of discretion in individual accrual accounts
Prior research examines the use of discretion in speci?c ac-
counts to manage earnings. Although several studies model discre-
tion in the allowance for doubtful accounts (Cecchini et al., 2012;
Jackson & Liu, 2010; McNichols & Wilson, 1988), their models re-
quire information such as bad debt expense and write-offs that is
only available for companies that provide transparent disclosure.
Because our goal is to examine differences in discretion between
companies that provide transparent disclosure and those that do
not, we build on a simple model from Marquardt and Wiedman
(2004). Marquardt and Wiedman (2004) capture discretion in the
allowance for doubtful accounts by taking the difference between
the current year allowance and the prior year allowance, and ad-
justing the current year allowance for changes in gross accounts
receivable relative to the prior year. While changes in gross ac-
counts receivable can affect the allowance for doubtful accounts,
this model assumes that collectability of the underlying receivables
remains constant. We extend their model to incorporate changes
in the collectability of the underlying receivables. Speci?cally, be-
cause the ratio of days sales outstanding is a measure of the aver-
age number of days that a company takes to collect revenue after
a sale has been made, changes in this ratio should capture changes
in the collectability of the underlying receivables. We incorporate
this into the model as follows:
DISC ARALL
it
= [ARALL
it
? (ARALL
it?1
? (GROSSAR
it
/GROSSAR
it?1
) ? (DSO
it
/DSO
it?1
)] (2)
/Total Assets
it?1
28 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
where
DISC ARALL discretion in the allowance for doubtful accounts
ARALL the allowance for doubtful accounts
GROSSAR gross accounts receivable
DSO days sales outstanding, calculated as gross accounts
receivable divided by daily sales (sales/365)
Because prior research does not model the valuation allowance
for inventories, we model discretion in this account using an ap-
proach that is consistent with the approach we use to model dis-
cretion in the allowance for doubtful accounts. Speci?cally, we es-
timate discretion in the valuation allowance for inventories as the
difference between the current and prior year allowance, adjust-
ing the current year allowance for changes in gross inventories
and changes in the age or salability of the inventory. We proxy
for changes in the age or salability of the inventory using the days
inventory outstanding ratio. This ratio measures how long it takes
the company, on average, to sell its inventory.
13
Thus, we use the
following model to estimate discretion in the valuation allowance
for inventories:
DISC INVALL
it
= [INVALL
it
? (INVALL
it?1
? (GROSSINV
it
/GROSSINV
it?1
) ? (DIO
it
(3)
/DIO
it?1
)]/Total Assets
it?1
where
DISC INVALL discretion in the valuation allowance for inventory
INVALL the valuation allowance for inventory
GROSSINV gross inventory
DIO days inventory outstanding, calculated as gross inventory
divided by daily cost of sales (cost of sales/365)
Finally, for the valuation allowance for deferred tax assets, we
simply estimate discretion as the difference between the current
and prior year allowance, adjusting the current year allowance for
changes in gross deferred tax assets.
14
Thus, we use the follow-
13
Accounting Standards Codi?cation (ASC) Topic 330 states, “in accounting for in-
ventories, a loss shall be recognized whenever the utility of goods is impaired by
damage, deterioration, obsolescence, changes in price levels, or other causes. The
measurement of such losses shall be accomplished by applying the rule of pric-
ing inventories at the lower of cost or market.” Two methods for dealing with the
requirements in ASC Topic 330 have emerged in practice: (1) a direct (write-off)
method where ending inventory is recorded at net realizable value if it is less than
cost and the associated loss is recorded against cost of goods sold (COGS), and (2)
an indirect (allowance) method where a valuation allowance is established to record
any decline in value below cost, with the associated loss recorded against COGS. We
?nd that approximately 19% of the 7669 sample company-year observations with a
current or prior year inventory balance disclose an associated current or prior year
inventory valuation allowance. Thus, the use of the indirect (allowance) method ap-
pears to be common in practice.
14
In untabulated analysis, we re-estimate discretion in the deferred tax asset val-
uation allowance by not only adjusting for the change in the underlying deferred
tax asset, but also adjusting for the change in net operating losses (NOLs) because
Miller and Skinner (1998) suggest that NOLs are the most important component
of deferred tax assets for determining the size of the valuation allowance. We ac-
complish this by multiplying the prior year allowance by the ratio of the current to
prior year gross deferred tax assets and by the ratio of current to prior year NOLs.
When current or prior year NOLs are zero, we set this ratio equal to one. All results
using this measure of discretion in the deferred tax asset valuation allowance are
consistent with those tabulated.
ing model to estimate discretion in the valuation allowance for de-
ferred tax assets:
DISC DTAALL
it
= [DTAALL
it
? (DTAALL
it?1
? (GROSSDTA
it
(4)
/GROSSDTA
it?1
))]/Total Assets
it?1
where
DISC DTAALL discretion in the valuation allowance for deferred tax
assets
DTAALL the valuation allowance for deferred tax assets
GROSSDTA gross deferred tax assets
DISC ARALL, DISC INVALL, and DISC DTAALL are multiplied by
negative one to capture income-increasing discretion.
15
3.2.2. Measures of overall discretion in accruals
We employ two different models to estimate discretionary ac-
cruals. First, following Kothari, Leone, and Wasley (2005), we esti-
mate performance-matched discretionary accruals using the mod-
i?ed Jones (1991) model (Dechow, Sloan, & Sweeney, 1995) and
we incorporate controls for the asymmetric recognition of gains
and losses (Ball & Shivakumar, 2006). Speci?cally, we estimate dis-
cretionary accruals as the residual from the following model, esti-
mated by year and 2-digit SIC code industry:
TA
it
= ?
0
+?
1
(REV
it
?REC
it
) +?
2
(PPE
it
) +?
3
(CFO
it
)
+?
4
(NEG_CFO
it
) +?
5
(NEG_CFO
it
? CFO
it
) +?
it
(5)
where
TA total accruals (calculated as income before extraordinary
items less cash ?ow from operations)
REV the change in net revenues from year t ? 1 to year t
REC the change in net receivables from year t ? 1 to year t
PPE gross property, plant, and equipment
NEG_CFO an indicator variable set equal to one if CFO is less than
zero, and zero otherwise
15
To assess the validity of these measures, we performed internet searches to
identify instances where these speci?c allowance accounts were restated. Our
searches included key words “restate,” “restatement,” “allowance for doubtful,” “bad
debts,” “inventory reserve,” “valuation allowance,” and “deferred tax valuation al-
lowance.” We identi?ed nine such restatements in the ?rst few pages of the search
results (four, three, and two restatements relate to the allowance for doubtful
accounts, the valuation allowance for inventories, and the deferred tax asset al-
lowance, respectively). For each restatement identi?ed, we estimated discretion in
the affected allowance account separately using the original (unrestated) data and
the restated data. We then contrasted the estimated discretion in the allowance ac-
count using restated data relative to unrestated data. For all nine restatements ex-
amined, we ?nd a decrease in the estimated discretion in the allowance account as
a result of the restatement. These results suggest that the likelihood of a misstate-
ment in a given allowance account is increasing in the discretion in that account,
where the discretion in the account is estimated using Models (2) through (4).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 29
all other variables are as de?ned previously. With the exception of
NEG_CFO, all variables in Model (5) are scaled by total assets at
the beginning of the year. We identify company-year matches by
identifying the company-year observation within the same 2-digit
SIC code industry with the closest prior year return on assets. The
performance-matched discretionary accrual estimate is the differ-
ence between the company-year and the matched company-year
estimates.
Second, following recent studies (Francis, Pinnuck, & Watan-
abe, 2014; Hope, Thomas, & Vyas, 2013), we estimate performance-
adjusted discretionary accruals using the Jones (1991) model.
Speci?cally, we estimate discretionary accruals as the residual from
the following model, estimated by year and 2-digit SIC code indus-
try
16
:
TA
it
= ?
0
+?
1
(REV
it
) +?
2
(PPE
i
t ) +?
3
(ROA
it
) +e
it
(6)
where all variables are as previously de?ned. All variables in Model
(6) are scaled by total assets at the beginning of the year.
For both discretionary accruals measures, we separately exam-
ine signed and positive (income-increasing) discretionary accruals
to capture behavior that in?ates current period earnings. We also
examine the absolute value of discretionary accruals to capture
the magnitude of earnings management regardless of its direction
(Klein, 2002; Reynolds & Francis, 2000).
17
3.2.3. The association between disclosure transparency and earnings
management
We use the following model to test our ?rst hypothesis, where
the measure of earnings management is based on discretion in the
individual allowance accounts:
DISC ALL
it
= ?
0
+?
1
(Transparent_Ind
it
) +?
2
(Size
it
)
+?
3
(CFO
it
) +?
4
(ROA
it
) +?
5
(Loss
it
) +?
6
(Issue
it
)
+?
7
(Growth
it
) +?
8
(Ind_Growth
it
)
+?
9
(Leverage
it
) +?
10
(Age
it
)
+?
11
(Analyst_Coverage
it
) +?
12
(Inst_Holdings
it
)
+?
13
(Big4
it
) +?
14
(Specialist
it
) +?
15
(Tenure
it
)
+?
16
(ICMW
it
) +?
17
(HERF
it
) +?
18
(Lit
it
)
+?
19
(IMR
it
) +?
20
(ALL_Scaled
it
)
+?
j
(Industry FE) +?
k
(Year FE) +?
it
(7)
16
For both Models (5) and (6), we estimate discretionary accruals using all avail-
able companies with su?cient data in Compustat during the sample period. We ex-
clude industry-years with fewer than 20 observations and winsorize all continuous
variables at the 1% and 99% levels to mitigate the in?uence of outliers.
17
We also perform (untabulated) tests using a third broad accruals measure
based on the model developed by Dechow and Dichev (2002) and modi?ed by
McNichols (2002), Francis, LaFond, Olsson, and Schipper (2005), and Hope et al.
(2013). Speci?cally, following Hope et al. (2013), we use the absolute value of the
residual from the modi?ed Dechow and Dichev (2002) model as a proxy for accru-
als quality and we ?nd similar results using this measure.
where
DISC ALL one of three measures: (i) DISC ARALL – discretion in the
allowance for doubtful accounts estimated using Model (2),
(ii) DISC INVALL – discretion in the valuation allowance for
inventory estimated using Model (3), or (iii) DISC DTAALL
– discretion in the valuation allowance for deferred tax
assets estimated using Model (4);
Transparent_Ind an indicator variable set equal to one if the company
provides transparent disclosure of the individual allowance
or reserve account, either in a Schedule II or in the notes
to the ?nancial statements, and zero otherwise;
IMR the inverse Mills ratio estimated using Model (1);
ALL_Scaled one of three measures: (i) ARALL_Scaled, (ii)
INVALL_Scaled, or (iii) DTAALL_Scaled;
ARALL_Scaled ARALL, scaled by total assets;
INVALL_Scaled INVALL, scaled by total assets;
DTAALL_Scaled DTAALL, scaled by total assets;
Industry FE industry ?xed effects, where industries are de?ned
following Ashbaugh, LaFond, and Mayhew (2003)
a
Year FE year ?xed effects
a
We use SIC codes to de?ne industries as follows: agriculture (0100-0999),
mining and construction (1000-1999, excluding 1300-1399), food (2000-2111), tex-
tiles and printing/publishing (2200-2799), chemicals (2800-2824; 2840-2899), phar-
maceuticals (2830-2836), extractive (1300-1399; 2900-2999), durable manufactur-
ers (3000-3999, excluding 3570-3579 and 3670-3679), transportation (4000-4899),
retail (5000-5999), services (7000-8999, excluding 7370-7379), computers (3570-
3579; 3670-3679; 7370-7379), and utilities (4900-4999).
all other variables are as previously de?ned. We estimate Model
(7) separately for each of the three allowance accounts.We use the
following model to test our ?rst hypothesis, where the measure of
earnings management is based on one of the two broad accruals
measures:
DiscAcc
it
= ?
0
+?
1
(Transparent_All
t
) +?
2
(Size
it
) +?
3
(CFO
it
)
+?
4
(ROA
it
) +?
5
(Loss
it
) +?
6
(Issue
it
)
+?
7
(Growth
it
) +?
8
(Ind_Growth
it
)
+?
9
(Leverage
it
) +?
10
(Age
it
)
+?
11
(Analyst_Coverage
it
) +?
12
(Inst_Holdings
it
)
+?
13
(Big4
it
) +?
14
(Specialist
it
) +?
15
(Tenure
it
)
+?
16
(ICMW
it
) +?
17
(HERF
it
) +?
18
(Lit
it
)
+?
19
(IMR
it
) +?
j
(Industry FE) +?
k
(Year FE) +?
it
(8)
where
DiscAcc one of two measures of discretionary accruals as estimated using
Model (5) or (6); for both measures, we estimate Model (8)
separately for signed, positive (income-increasing) only, and
absolute value discretionary accruals
a
a
To examine positive (income-increasing) discretionary accruals, we use tobit
when estimating Model (8).
all other variables as previously de?ned. The coe?cient of interest
in Model (7) is ?
1
, the coe?cient on Transparent_Ind, and the co-
e?cient of interest in Model (8) is ?
1
, the coe?cient on Transpar-
ent_All. Consistent with Hypothesis 1, we expect the coe?cients on
Transparent_Ind and Transparent_All to be negative and signi?cant
in all speci?cations of Models (7) and (8).
30 C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38
Table 1
Sample selection.
ARALL INVALL DTAALL
Panel A: Samples for tests of discretion in individual allowance and reserve accounts
Company-year observations that provide a Schedule II 4405 1159 1754
Company-year observations that disclose the allowance or reserve balance transparently in the notes to the ?nancial statements 1790 63 170
Company-year observations that provide non-transparent disclosure for the allowance or reserve account 2382 216 630
Less: Company-year observations where the allowance is immaterial (i.e., where the prior year balance is less than once cent per share) (1987) (206) (94)
Sample for Tests of H1 6590 1232 2460
Less: Company-year observations where non-transparent disclosure is provided for the allowance or reserve account (1351) (141) (618)
Sample for Tests of H2 5239 1091 1842
Panel B: Samples for tests of overall discretion in accruals
Sample for
Tests of H1
Sample for
Tests of H2
Companies that provide transparent disclosures (Transparent_All = 1)
Company-year observations that provide a Schedule II 4664 4664
Company-year observations that disclose all material (disclosed) allowance and reserve balances transparently in the notes to
the ?nancial statements
1941 1941
Companies that provide non-transparent disclosures (Transparent_All = 0)
Company-year observations that provide non-transparent disclosure for all material (disclosed) allowance and reserve balances 2854 NA
Less: Company-year observations where prior year aggregate allowance and reserve balances are less than one cent per share (2032) (1009)
Samples for tests of overall discretion in accruals 7427 5596
The samples consist of company-year observations with necessary data from Compustat and Audit Analytics to construct the variables in our models. For all samples, we
exclude company-year observations where transparent disclosures are provided for some, but not all, material (disclosed) valuation allowance and reserve accounts.
Models (7) and (8) include controls for company size (Size),
age (Age), sales growth (Growth), and industry sales growth
(Ind_Growth) because prior research indicates that these company
characteristics affect the level of discretionary accruals (Anthony &
Ramesh, 1992; Becker, DeFond, Jiambalvo, & Subramanyam, 1998;
Myers, Myers, & Omer, 2003). To control for the effects of com-
pany performance, we include cash ?ow from operations (CFO)
and return on assets (ROA) (Dechow et al., 1995). Following prior
research, we also control for the issuance of debt and equity ?-
nancing (Issue) (Rangan, 1998), whether analysts cover the com-
Table 2
Descriptive statistics.
Variable
N Mean P25 Median P75 N Mean P25 Median P75
Diff
in
mean Transparent_Ind = 1 Transparent_Ind = 0
DISC ARALL 5239 0.000 ?0.001 0.000 0.002 1351 0.001 ?0.001 0.000 0.002 ?0.001
DISC INVALL 1091 0.001 ?0.003 0.000 0.003 141 0.016 ?0.003 0.000 0.002 ?0.014
?
DISC DTAALL 1842 0.001 ?0.002 0.000 0.001 618 0.005 0.000 0.000 0.000 ?0.004
?
Transparent_All = 1 Transparent_All = 0
DiscAcc
a
5596 0.043 ?0.018 0.021 0.073 1831 0.097 ?0.015 0.033 0.125 ?0.054
???
AbsDiscAcc
a
5596 0.125 0.019 0.048 0.107 1831 0.223 0.024 0.067 0.181 ?0.098
???
DiscAcc
b
5596 ?0.013 ?0.063 ?0.010 0.046 1831 ?0.003 ?0.060 0.006 0.068 ?0.010
??
AbsDiscAcc
b
5596 0.088 0.024 0.055 0.108 1831 0.129 0.027 0.064 0.135 ?0.042
???
Size 5596 6.542 5.222 6.500 7.808 1831 5.613 3.614 5.407 7.587 0.929
???
CFO 5596 0.078 0.043 0.091 0.137 1831 ?0.107 ?0.008 0.074 0.129 0.184
???
ROA 5596 ?0.004 ?0.028 0.034 0.075 1831 ?0.153 ?0.092 0.017 0.068 0.148
???
Loss 5596 0.327 0.000 0.000 1.000 1831 0.430 0.000 0.000 1.000 ?0.104
???
Issue 5596 0.125 0.000 0.000 0.000 1831 0.140 0.000 0.000 0.000 ?0.015
?
Growth 5596 0.127 ?0.093 0.038 0.154 1831 0.457 ?0.111 0.035 0.210 ?0.331
Ind_Growth 5596 0.029 ?0.038 0.029 0.096 1831 0.036 ?0.038 0.047 0.128 ?0.007
??
Leverage 5596 0.234 0.020 0.177 0.338 1831 0.367 0.015 0.184 0.383 ?0.133
???
Age 5596 23.200 12.000 18.000 30.000 1831 20.502 11.000 17.000 27.000 2.698
???
Analyst_Coverage 5596 0.786 1.000 1.000 1.000 1831 0.559 0.000 1.000 1.000 0.227
???
Inst_Holdings 5596 0.549 0.203 0.636 0.871 1831 0.335 0.000 0.172 0.699 0.214
???
Big 4 5596 0.773 1.000 1.000 1.000 1831 0.590 0.000 1.000 1.000 0.183
???
Specialist 5596 0.234 0.000 0.000 0.000 1831 0.190 0.000 0.000 0.000 0.045
???
Tenure 5596 10.460 5.000 8.000 13.000 1831 9.050 4.000 7.000 12.000 1.409
???
ICMW 5596 0.053 0.000 0.000 0.000 1831 0.102 0.000 0.000 0.000 ?0.048
???
HERF 5596 0.070 0.034 0.047 0.078 1831 0.066 0.031 0.042 0.084 0.004
??
Lit 5596 0.146 0.000 0.000 0.000 1831 0.193 0.000 0.000 0.000 ?0.047
???
a
Estimated using Model (5).
b
Estimated using Model (6).
C.A. Cassell et al. / Accounting, Organizations and Society 46 (2015) 23–38 31
Table 3
The determinants of transparent disclosure.
Variable Pred.
DV = Transparent_All
Coe?cient estimate p-Value
Intercept ? 0.098 .133
Size + 0.016
?
.076
CFO + 0.326
???
.001
ROA + 0.076 .133
Loss + 0.033 .206
Issue + ?0.052 .849
Growth ? ?0.003 .310
Ind_Growth ? ?0.303
??
.030
Leverage ? ?0.109
???
.007
Age + 0.003
??
.023
Analyst_Coverage + 0.175
???