Description
This study examines whether firms with large conservative allowances for uncollectible accounts on their
balance sheets make significant adjustments to increase net assets in response to Staff Accounting
Bulletin No. 108 (SAB 108). To the extent adjustments are not made, we examine the qualitative reasons
SAB 108 may have had a limited effect on these large conservative accounting estimates. In the year of
enactment, SAB 108 required public companies to adjust material accumulations on the balance sheet
that may have occurred prior to enactment of SAB 108 after considering all relevant quantitative and
qualitative factors. Our archival results find that no firms made a SAB 108 adjustment and while sample
firms tend to, at best, make small reductions to the allowance subsequent to SAB 108’s effective date,
most sample firms continue to have potentially overestimated allowances several years later. To understand
the qualitative forces underlying this finding, we interview key individuals involved in the financial
reporting process and we survey experienced financial executives employed at our sample firms. The
findings reveal that organizational forces favor conservatism because being over-reserved helps firms
avoid income-decreasing surprises while being under-reserved appears careless. We conclude that conservatism
remains a deeply engrained feature of accounting thought and practice, in contrast to recent
emphasis by accounting standard setters on neutrality over conservatism.
Conservatism and Staff Accounting Bulletin No. 108
q
Thomas G. Canace
a
, Richard C. Hat?eld
b,?
, Scott B. Jackson
c
a
Wake Forest University, United States
b
University of Alabama, United States
c
University of South Carolina, United States
a r t i c l e i n f o
Article history:
Received 22 August 2012
Revised 14 August 2015
Accepted 19 August 2015
Available online xxxx
a b s t r a c t
This study examines whether ?rms with large conservative allowances for uncollectible accounts on their
balance sheets make signi?cant adjustments to increase net assets in response to Staff Accounting
Bulletin No. 108 (SAB 108). To the extent adjustments are not made, we examine the qualitative reasons
SAB 108 may have had a limited effect on these large conservative accounting estimates. In the year of
enactment, SAB 108 required public companies to adjust material accumulations on the balance sheet
that may have occurred prior to enactment of SAB 108 after considering all relevant quantitative and
qualitative factors. Our archival results ?nd that no ?rms made a SAB 108 adjustment and while sample
?rms tend to, at best, make small reductions to the allowance subsequent to SAB 108’s effective date,
most sample ?rms continue to have potentially overestimated allowances several years later. To under-
stand the qualitative forces underlying this ?nding, we interview key individuals involved in the ?nancial
reporting process and we survey experienced ?nancial executives employed at our sample ?rms. The
?ndings reveal that organizational forces favor conservatism because being over-reserved helps ?rms
avoid income-decreasing surprises while being under-reserved appears careless. We conclude that con-
servatism remains a deeply engrained feature of accounting thought and practice, in contrast to recent
emphasis by accounting standard setters on neutrality over conservatism.
Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Prior research suggests that many ?rms have accumulated large
conservative allowances for uncollectible accounts on their bal-
ance sheets and that those ?rms may use conservative balances
to manage earnings in future periods (Jackson & Liu, 2010). Staff
Accountant Bulletin No. 108 (SAB 108) requires public companies,
under the iron-curtain approach, to consider the materiality of
potential misstatements that have accumulated on the balance
sheet through multiple non-material income statement differ-
ences. In the year of enactment, SAB 108 required ?rms to make
adjustments for any material accumulations that may have
occurred prior to enactment after considering all relevant quantita-
tive and qualitative factors. The enactment of SAB 108 allows for a
‘‘natural experiment” to determine whether such speci?c guidance
would initiate ?nancial statement adjustments for large conserva-
tive allowances. Further, to the extent that adjustments are not
made, the enactment of SAB 108 allows us to use qualitative meth-
ods to investigate speci?c reasons why conservative allowances
persist on the balance sheet. Thus, we interview and survey the
key players (i.e., audit partners, ?rm managers and regulators) to
better understand why SAB 108 had a limited effect on large con-
servative allowance balances.
1
Evidence in various domains of accounting suggests that com-
panies may be reluctant to increase net assets because doing so
(1) runs counter to conservatism (Basu, 1997; Givoly & Hayn,
2000; Hirshleifer & Teoh, 2009; Watts, 2003a; Watts, 2003b) and
(2) may reduce their ability to manage earnings in the future
(Barton & Simko, 2002; Jackson & Liu, 2010; McNichols & Wilson,
1988; Penman, 2001). Thus, a better understanding of whether
such balances are adjusted under SAB 108 is important because
the SEC has indicated that uncorrected misstatements distort the
balance sheet and are contrary to the best interests of ?nancialhttp://dx.doi.org/10.1016/j.aos.2015.08.002
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
q
We thank the editor, two anonymous reviewers, Charles Boster, Steven Jackson,
Marsha Keune, Kelvin Liu, Joel Owens, Tammie Schaefer, Mike Shaub, and Mark
Taylor for providing insightful comments and suggestions. We also thank Guang
Rong for assistance with data collection. Professor Canace is especially grateful to
the Wake Forest University School of Business for providing research funding
through the summer research grant program.
?
Corresponding author at: Culverhouse School of Accounting, University of
Alabama, Tuscaloosa, AL 35487-0220, United States.
E-mail address: [email protected] (R.C. Hat?eld).
1
In our setting, the practice of conservatism refers to a preference for reporting
estimates in the direction of understatement rather than overstatement of net income
and net assets (Financial Accounting Standards Board (FASB), 1980).
Accounting, Organizations and Society xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er . com/ l ocat e/ aos
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
statement users, especially if ‘‘cookie jar” reserves facilitate earn-
ings management in the future (Levitt, 1998; Securities &
Exchange Commission (SEC), 2006). Indeed, the SEC identi?ed
these among the reasons for issuing SAB 108. Yet, SAB 108 also
requires managers and their auditors to consider all relevant quan-
titative and qualitative information, which raises the importance of
examining both factors before making conclusions about necessary
adjustments for material accumulations on the balance sheet.
Thus, we explore both aspects of this decision because there may
be qualitative considerations that help to explain managers’
estimates.
In the ?rst stage of this study, we focus on quantitative data
related to the allowance for uncollectible accounts to consider the
potential for conservative accumulations on the balance sheet. Prior
research shows that this account has grown over time because
?rms seemingly over-accrue bad debt expense, perhaps causing
the allowance to become overly conservative in the long-run
(Jackson & Liu, 2010). Using future realizations of the allowance
by examining ?rms’ publicly disclosed write-offs (McNichols &
Wilson, 1988), we calculate a quantitative difference (potential
overestimation) as a measure of allowance accuracy by subtracting
one-year leading write-offs from the allowance for uncollectible
accounts. We then consider whether these quantitative differences
remain on the balance sheet of our sample ?rms and if the large
conservative differences persist after the promulgation of SAB 108.
In the year immediately preceding the effective date of SAB 108,
our archival analyses suggest that 37.50% of the ?rms we analyze
have potential overestimations of their allowance in excess of ?ve
percent of net income. The mean allowance for these ?rms is suf-
?cient to cover ?ve years of leading write-offs. In fact, the 25 lar-
gest calculated quantitative differences average almost 60% of
net income. Despite the size of the accumulated allowance for
uncollectible accounts on the balance sheet, none of the ?rms we
analyze make a SAB 108 adjustment to correct a potential accumu-
lated error in the allowance account. Moreover, while we ?nd that
the allowance declined modestly, on average, following SAB 108’s
effective date, most sample ?rms continue to have a potentially
overestimated allowance four periods after SAB 108’s effective
date. The SEC expressly prohibits improper assets and liabilities
from remaining on the balance sheet in perpetuity (SEC, 2006).
Thus, the fact that our sample ?rms opted for no adjustment, or
at best, small delayed adjustment, of the allowance suggests that
the differences we calculate, though often quite large, are not being
perceived as material misstatements worthy of correction.
The second stage of the study employs qualitative methods to
better understand why ?rms, and their auditors, are not applying
SAB 108 to this apparent overestimation of the allowance for
uncollectible accounts. We conduct open-ended interviews with
ten key individuals involved in the ?nancial reporting process,
including ?nancial executives, audit partners, and regulators at
the SEC and PCAOB to shed light on the archival ?ndings. Experi-
enced ?nancial executives indicate that organizational forces favor
conservatism with respect to the allowance and auditors do little
to deter it. Indeed, these executives believe that being over-
reserved helps the ?rm to avoid income-decreasing surprises in
an uncertain environment, while being under-reserved appears
careless. In turn, auditors indicate that there is little risk associated
with having a conservative allowance and are unlikely to oppose
such a preference. These partners framed the issue not as one
involving an over-reserved allowance, but as one involving a con-
servative balance. The accounting regulators commented (1) that
some SEC reviewers of ?lings may evaluate the allowance in rela-
tion to the balance sheet rather than leading write-offs, which
might make the overestimation seem less consequential, and (2)
that a plausible explanation for overestimated allowances is the
desire to be conservative.
To get a better understanding of the speci?c actions of the ?rms
in our sample, we also survey experienced ?nancial executives
employed at our sample ?rms. Survey ?ndings reveal that many
individuals and groups internal to these ?rms strongly favor con-
servatism with respect to the allowance and no individuals or
groups oppose it. Moreover, the responses suggest that ?nancial
executives primarily resist drawing down the allowance because
doing so is at odds with maintaining a conservative balance. Thus,
a conservative allowance appears to have become a somewhat per-
manent feature of the balance sheet.
Taken as a whole, it appears that SAB 108 would certainly be
applicable to these ?rms if we simply considered the quantitative
materiality of these potential misstatements. The fact that these
?rms and their auditors do not perceive SAB 108 to apply suggests
that they do not perceive these overestimates as material misstate-
ments. That is, if a ?rm has a large allowance, but has some justi-
?cation for it, the auditor may just view this as a very conservative
balance, but not a misstatement. There appears to be some asym-
metry in the judgment of misstatement depending upon whether
the amount in question is income/asset increasing or decreasing.
It is also interesting that, despite the potential for overestimated
allowances to be used as ‘‘cookie jar” reserves, they still occur after
SAB 108 and ?rms’ management suggest they keep these reserves
for the very reason of having some cushion to avoid hits to
earnings.
Our study contributes to the accounting literature by providing
the ?rst evidence that conservative accumulations on the balance
sheet persist subsequent to speci?c guidance about eliminating
accumulations (i.e., SAB 108). Our archival evidence that ?rms
make small reductions to conservative allowance balances after
SAB 108 suggests that the SEC may have modestly attenuated
?rms’ conservatism in the reporting of these estimates on the bal-
ance sheet. However, interview and survey evidence also suggests
that managers believe the maintenance of large conservative
allowances is appropriate.
Further, discussions with audit partners and regulators reveal
that such accumulations are unlikely to cause them to object to
conservative balances. So, while regulators create speci?c guid-
ance about dealing with material accumulations (e.g., SAB 108),
our quantitative and qualitative results suggest that conservative
balances in the ?nancial statements are still viewed as appropri-
ate by ?rms and seemingly permissible by auditors/regulators.
These ?ndings may be important for future studies examining
accounting conservatism for two reasons. First, our evidence
highlights the need to consider both quantitative and qualitative
factors when making conclusions about ?rms’ accounting esti-
mates. Second, our evidence should be considered in light of
recent changes to the Conceptual Framework (Concepts State-
ment No. 8, ‘‘Conceptual Framework for Financial Reporting”
FASB, 2010) which places emphasis on neutrality over conser-
vatism. The revised framework does not include conservatism
as a qualitative characteristic because it has become associated
with the deliberate understatement of net assets and net income
and is in con?ict with the quality of neutrality which encom-
passes freedom from bias.
Moreover, the results have at least two practical implications
given recent regulatory concern and research that suggest such
conservative balances may increase the likelihood of earnings
management. First, the large accumulations that we identify seem
to have become part of the status quo and auditors may have
become uncritically accustomed to them as interview evidence
shows they view them as bearing little inherent risk. Our ?ndings
could sensitize auditors to the possibility that accumulations in the
allowance account are systematic and excessive when assessed
with commonly accepted quantitative benchmarks. Second, these
?ndings could prompt the SEC of?cials to scrutinize allowance
2 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
balances that appear to be ‘‘overly conservative” and issue
comment letters to ?rms, which, among other things, opens a
dialogue about conservatism.
The remainder of this paper proceeds as follows. The next sec-
tion provides background, discusses relevant prior research, and
develops our hypotheses. Section 3 describes the archival methods
and results. Section 4 describes our additional archival analyses.
Section 5 reports the results of open-ended interviews with ?nan-
cial executives, auditors, and regulators. Section 6 discusses our
survey of experienced ?nancial executives employed at our sample
?rms. The ?nal section summarizes the results and describes cer-
tain limitations.
2. Background, relevant prior research, and hypotheses
development
2.1. Summary of Staff Accounting Bulletin No. 108
The SEC issued SAB 108 in 2006 to address two concerns—(1)
the diversity in practice with respect to quantifying ?nancial state-
ment misstatements and (2) the potential for improper amounts to
accumulate on the balance sheet.
2
Prior to the issuance of SAB 108,
auditors elected to use the ‘‘rollover” approach or ‘‘iron curtain”
approach to quantify misstatements, but auditors generally did not
consider both approaches simultaneously. The rollover approach
quanti?es a misstatement by looking at the amount of the error orig-
inating on the income statement in the current year. This method
ignores any accumulated balance sheet misstatement originating
in prior years. Conversely, the iron curtain approach quanti?es a
misstatement by looking at the amount of the misstatement on
the balance sheet regardless of when the misstatement originated.
To illustrate the differences between the rollover approach and
iron curtain approach, consider Fig. 1, which contemplates the fol-
lowing circumstances. A ?rm records an excessive expense accrual
each year for ?ve years. The amount by which the expense accrual
exceeds the ‘‘correct” accrual is $10 per year. At the same time, the
?rm reports income of $1000 each year. If the rollover approach is
used to quantify ?nancial statement misstatements then the error
is $10 each year. Expressed as a percent of income, the excessive
expense accrual is only one percent of income each year. If the iron
curtain approach is used to quantify ?nancial statement misstate-
ments then the error is $10 in year 1, $20 in year 2, $30 in year
3, $40 in year 4, and $50 in year 5. Expressed as a percent of
income, the excess accrual is one, two, three, four, and ultimately
?ve percent of earnings in years 1 through 5, respectively.
The provisions of SAB 108 prohibit exclusive reliance on either
the rollover approach or iron curtain approach. Instead, SAB 108
requires use of the ‘‘dual” approach, which is the approach initially
recommended by Nelson, Smith, and Palmrose (2005). The dual
approach requires auditors and their clients to consider the magni-
tude of a misstatement under both the rollover approach and iron
curtain approach and then make an adjustment whenever the mis-
statement is material under either approach after considering the
relevant quantitative and qualitative factors. Thus, in the illustra-
tive example above, the dual approach would seemingly require
an adjustment to correct the accumulated misstatement in the
allowance account based upon the quantitative materiality of the
error under the iron curtain approach.
In the year SAB 108 became effective, ?rms are not required to
restate prior period ?nancial statements if management properly
applied either the rollover approach or iron curtain approach. This
transition feature allows ?rms that have accumulated misstatements
on the balance sheet to avoid restating ?nancial statements in
the years that led to the misstatement. However, ?rms must re?ect
the cumulative effect of the initial application of SAB 108 in the
carrying amounts of assets and liabilities as of the beginning
of the ?rst ?scal year after SAB 108 became effective, with an off-
setting adjustment to the opening balance of retained earnings.
SAB 108 expressly prohibits delayed or partial corrections of mis-
statements. Errors that existed at the time SAB 108 became effec-
tive but were not corrected on a timely basis must be corrected by
restating the prior period ?nancial statements in accordance with
Statement of Financial Accounting Standards (SFAS) No. 154,
Accounting Changes and Error Corrections (FASB, 2006; ASC §250).
2.2. Research related to Staff Accounting Bulletin No. 108
Nelson et al. (2005) examine the effect of the two methods of
quantifying ?nancial statement misstatements, ?nding that audi-
tors are more likely to require their clients to book misstatements
under the approach that makes the misstatement appear more
material. Also, while Nelson et al. (2005) provide participants
with the information necessary to evaluate misstatements using
either the rollover approach or iron curtain approach, partici-
pants’ audit adjustment decisions were strongly in?uenced by
how the information was presented to them (i.e., presented in
accordance with the rollover approach or the iron curtain
approach). This ?nding is consistent with the SEC’s concern that
the rollover approach may allow improper amounts to accumu-
late on the balance sheet. Nelson et al. (2005) also consider
whether income increasing adjustments would be affected by
employing the iron curtain approach. In their cases with income
increasing adjustments, auditors are least likely to require the
adjustment. They suggest that auditors may waive these
adjustments, thereby permitting conservatism, because auditor
training and litigation risk encourage detecting income overstate-
ments, not understatements.
Keune and Johnstone (2009) examine SAB 108 disclosures, ?nd-
ing that misstatements disclosed under SAB 108 (1) tend to be
income-decreasing, (2) tend to be made by larger ?rms, (3) are
concentrated in regulated industries, (4) are underrepresented in
the manufacturing industry, (5) are clustered in current liabilities,
leases, revenue recognition, and deferred taxes, and (6) are not
overrepresented in reserve accounts. They point out that these
?ndings are consistent with prior evidence that auditors tend to
allow conservative reporting (Kinney & Martin, 1994; Wright &
Wright, 1997) and are less likely to require adjustments for subjec-
tive accounts that require estimation (Braun, 2001; Ng, 2007).
Therefore, their results suggest that considering both the direction
and account type of the adjustment are critical to a consideration
of why ?rms make SAB 108 adjustments.
Moreover, several studies examine the association between var-
ious market-related variables and SAB 108 corrections. Omer,
Shelley, and Thompson (2011) ?nd that ?rms reporting SAB 108
corrections are no more likely to meet or beat analysts’ consensus
forecasts than ?rms not reporting SAB 108 corrections. Keune and
Johnstone (2012) ?nd that auditors are more likely to waive qual-
itatively material misstatements as analyst following increases,
but they also ?nd that auditors are less likely to waive misstate-
ments as audit fees increase. Omer, Shelley, and Thompson
(2012) ?nd a negative association between factors that are posited
to reduce audit quality and investors’ response to SAB 108 correc-
tions, which suggests that investors react more negatively to
corrections when auditor independence appears weaker.
To date, prior research has empirically examined corrected mis-
statements that are disclosed in audited ?nancial statements; prior
research has not empirically examined whether large and poten-
tially material misstatements go uncorrected in audited ?nancial
2
SAB 108 was issued on September 13, 2006 and its provisions became effective
for the ?rst ?scal year ending after November 15, 2006.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 3
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
statements, and, if so, the qualitative reasons adjustments are not
made. We speci?cally identify large accumulations (potential mis-
statements based on the iron curtain approach) on the balance
sheet prior to and after the enactment of SAB 108 to consider
whether or not management and auditors perceive that SAB 108
applies.
2.3. Are large conservative accumulations of the allowance adjusted
under SAB 108?
We focus on a particular accrual account, the allowance for
uncollectible accounts, to evaluate whether or not ?rms applied
SAB 108 to conservative accumulations on the balance sheet.
The allowance captures the accumulated effect of small increases
due to bad debt expense on the income statement being consis-
tently in excess of write-offs. Jackson and Liu (2010) suggest that
over-accruals of bad debt expense have accumulated in the allow-
ance account during the past 25 years. By the ?nal year of their
sample (2004), which predates SAB 108, Jackson and Liu (2010)
?nd that most ?rms could endure write-offs of uncollectible
accounts for several years without exhausting the allowance. Like
Jackson and Liu (2010), we de?ne a potential overestimate of the
allowance by considering the extent to which it exceeds leading
write-offs.
Managers of ?rms that have a large allowance may be reluctant
to fully adjust this account in the year that SAB 108 became effec-
tive for two main reasons. First, if managers immediately adjusted
the allowance then they would be unable to draw it down in sub-
sequent years when earnings thresholds may be in danger of not
being met. Regulators and academics contend that understated
net assets increase management’s ability to manage earnings
(American Institute of Certi?ed Public Accountants (AICPA),
1939; Devine, 1963; DeFond, 2002; Levitt, 1998; Penman, 2001),
and empirical research provides evidence that managers are more
likely to manage earnings when net assets are understated (Barton
& Simko, 2002; Jackson & Liu, 2010; McNichols & Wilson, 1988).
Second, managers face strong institutional forces that promote
conservative accounting and generally oppose adjusting asset
understatements (Givoly & Hayn, 2000; Watts, 2003a; Watts,
2003b). Conservatism has evolved over many decades and it has
become one of the most deeply engrained features of accounting
thought and practice (Basu, 1997; Watts, 2003a). Hirshleifer and
Teoh (2009) argue that managers favor conservative accounting
because it reduces the likelihood that they (and others) will be dis-
appointed when the realization of signi?cant estimates becomes
known in the future. Thus, preferences for conservative accounting
by managers may override preferences for regulatory compliance
(Holthausen, 2009)
3
or may alter the perception of whether or not
SAB 108 applies.
In addition, auditors may not require their clients to correct an
overestimated allowance. The risk-based audit approach directs
auditors’ attention to the overstatement of net assets, not the
understatement of net assets (Arens, Elder, & Beasley, 2003). Con-
sistent with this view, evidence suggests that auditors are more
concerned about reducing asset-increasing misstatements than
reducing asset-decreasing misstatements (Francis & Krishnan,
1999; Kinney & Martin, 1994; Nelson, Elliott, & Tarpley, 2002;
Wright & Wright, 1997). If auditors are comfortable with a conser-
vative balance in the allowance, then it is likely they will accept the
amount, given that management will have some plausible ratio-
nale for the balance.
This discussion not only suggests that managers may oppose
adjusting a potentially overestimated allowance, but it also sug-
gests that auditors may not require an adjustment. Hypothesis 1,
stated in alternative form, is as follows:
Hypothesis 1. Large (conservative) accumulations in the allow-
ance for uncollectible accounts will not be fully adjusted under the
application of SAB 108 in the year it became effective.
2.4. Does the allowance gradually decline after SAB 108’s effective
date?
While Hypothesis 1 predicts that these large accumulations in
the allowance are not fully adjusted in the year that SAB 108
became effective, we nonetheless expect that the allowance will
decline in the years subsequent to SAB 108’s effective date.
Jackson and Liu (2010) indicate that the allowance is built-up over
time by ?rms over-accruing bad debt expense. In turn, they ?nd
that the allowance is subsequently drawn down by ?rms recording
small amounts of bad debt expense when they are in danger of
missing analysts’ forecasts. However, SAB 108 may make it dif?cult
for managers to continue to build up the allowance because over-
Illustration of Different Methods of Quantifying Financial Statement Misstatements
Financial statement
Excessive Financial statement misstatement amount as a
expense misstatement amount percentage of earnings
Year accrual Rollover Iron curtain Income Rollover Iron curtain
1 10 10 10 1,000 1% 1%
2 10 10 20 1,000 1% 2%
3 10 10 30 1,000 1% 3%
4 10 10 40 1,000 1% 4%
5 10 10 50 1,000 1% 5%
This figure illustrates the effect on the income statement and balance sheet of recording an excessive expense accrual
in five consecutive years. A hypothetical company is assumed to record an expense of $10 in excess of the needed
amount for five consecutive years. Each year the excess expense accumulates on the balance sheet. The misstatement
on the balance sheet is $10 at the end of year 1, $20 at the end of year 2, and ultimately $50 at the end of year 5. While
the adjustment necessary to correctly state earnings is only $10 in any given year, the adjustment necessary to
correctly state the balance sheet accumulates over time, resulting in a quantitatively material misstatement by the end
of year 5.
Fig. 1. Illustration of different methods of quantifying ?nancial statement misstatements.
3
Empirical evidence suggests that compliance with regulations does not always
occur. For example, Schwartz and Soo (1996) show that noncompliance with
disclosures related to auditor terminations is related to various management
incentives. Likewise, Robinson, Xue, and Yu (2011) and Scannell and Lublin (2008)
?nd considerable non-compliance with executive compensation disclosures despite
the fact that veri?cation of compliance is almost costless. Moreover, articles from the
business press point out instances of possible noncompliance with accounting
standards affecting ?nancial statement accounts such as recent rules on asset
retirement obligations (see www.goodwinprocter.com/Publications/Newsletters/
Environmental-and-Energy-Advisory/2007/0424_FIN-47-Poses-Hard-Questions-for-
Estimating-Future-Environmental-Liabilities.aspx).
4 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
accruing bad debt expense would increase the likelihood that,
under the iron curtain approach, this account is seen to be materi-
ally misstated. Thus, we expect SAB 108 to create asymmetry in the
movement of large allowance accumulations—?rms can readily
draw down the allowance in the post-SAB 108 years but they
may encounter resistance if they wish to replenish the allowance
in those years. At the same time, if SAB 108 signi?cantly reduces
?rms’ ability to replenish an overestimated allowance account
then bad debt expense should be lower in the years after SAB
108 became effective than in the years before SAB 108 became
effective. Hypotheses 2a and 2b, stated in alternative form, are as
follows:
Hypothesis 2a. The allowance account declines in the years after
SAB 108 became effective.
Hypothesis 2b. Bad debt expense declines in the years after SAB
108 became effective.
2.5. Does the allowance continue to be highly conservative?
While we expect that the allowance account will be drawn
down in the post-SAB 108 years, it seems unlikely that ?rms
will quickly draw down accumulations in the allowance during
that period because that action could attract unwanted scrutiny
by auditors, regulators, and investors. We therefore expect the
allowance to remain quantitatively large (conservative) several
years after SAB 108 became effective even though Hypotheses
2a and 2b predict that it will decline during the post-SAB
108 period. Hypothesis 3, stated in alternative form, is as
follows:
Hypothesis 3. Firms continue to have large (conservative) accu-
mulations in the allowance account several years after SAB 108
became effective.
3. Archival analyses
In the ?rst stage of the study, we employ archival data to test
our hypotheses about the potential for conservative accumulations
in the allowance for uncollectible accounts to survive subsequent
to the effective date of SAB 108.
3.1. Sample selection
We identify the 750 domestic ?rms on Compustat in 2002 with
the largest unscaled balances in net current accounts receivable.
Our rationale for selecting such ?rms is that we wish to identify
?rms that are likely to have material allowance balances. We elim-
inate (1) ?rms in regulated industries (primarily ?nancial institu-
tions and utilities), (2) ?rms whose receivables are mainly due
from consumers (i.e., non-business receivables), and (3) leasing
companies.
4
This eliminates 129 ?rms. We then attempt to obtain
?nancial data for the remaining 621 ?rms from two sources—Sched-
ule II of Form 10-K and Compustat. Write-offs, bad debt expense, and
recoveries of previously written-off accounts must all be obtained
from Schedule II because Compustat does not collect these items.
5
In some cases we could not locate a Schedule II or an equivalent
schedule in the 10-K. In total, we eliminate 132 ?rms because Sched-
ule II could not be located or because of missing ?nancial data on
Compustat.
Our sample period starts in 2002, although we also obtain
needed data for 2001 from Schedule II along with Compustat data
for that year (there is no incremental sample attrition due to add-
ing 2001). We also require ?rms to have continuous Schedule II
data and Compustat data during the period 2001 through 2006.
We impose this requirement to avoid having ?rms appear in our
sample during pre-SAB 108 years but not actually appear in our
sample during the year SAB 108 became effective. One hundred
sixty-one ?rms do not have continuous Schedule II data and Com-
pustat data during this period, which results in 328 ?rms from
which we attempt to calculate the quantitative accuracy of the
allowance balance (discussed next). In turn, we collect needed data
through 2009, thereby leaving four potential reporting periods
subsequent to the effective date of SAB 108. Due to data availabil-
ity and sample attrition, the number of ?rms analyzed declines
from 328 ?rms in 2006 to 259 ?rms by 2009.
3.2. Benchmarks
We need a quantitative benchmark for estimating the accuracy
of the allowance based on observable write-offs. We use material-
ity which is typically de?ned by reference to net income. Evidence
indicates that a misstatement above ?ve percent of net income is
almost always considered quantitatively material (Chewning,
Pany, & Wheeler, 1989; Eilifsen & Messier, 2015; Holstrum &
Messier, 1982; Messier, Martinov-Bennie, & Eilifsen, 2005; Pany
& Wheeler, 1989). Therefore, our analysis employs ?ve percent of
net income for determining whether the allowance is potentially
overestimated.
We develop a benchmark for an ‘‘accurate” allowance by con-
sidering leading write-offs as a reference point for what the allow-
ance would have been to perfectly cover bad debts for the year
(Cecchini, Jackson, & Liu, 2012; Fedyk & Singer, 2011; Jackson &
Liu, 2010; McNichols & Wilson, 1988). This benchmark matches
the allowance with its future realization.
6
We note, however, that
leading write-offs is not a perfect benchmark because it combines
write-offs of prior year credit sales (this is the portion of write-offs
that we would ideally include) with write-offs of current year credit
sales (this is the portion of write-offs that we would ideally exclude).
Because our benchmark for the allowance includes some write-offs
that ought to be excluded, it is a benchmark that may bias our tests
against concluding that the allowance is potentially overestimated.
Thus, we de?ne a potentially overestimated allowance as one in
which the allowance exceeds leading write-offs by an amount in
excess of materiality. Using the materiality and allowance bench-
mark noted above (?ve percent), Appendix A lists the ?rms that
we de?ne to have a potentially overestimated allowance in the
year before SAB 108 became effective. Column (a) provides the dif-
4
With respect to item (1), we eliminate ?nancial institutions because their
receivables (primarily loans) are fundamentally different from the receivables of non-
?nancial institutions. Also, eliminating ?nancial institutions removes ?rms that
experienced extreme distress during the ?nancial crisis of 2008. With respect to item
(ii), consumer receivables are different from trade (business) receivables in terms of
transaction size, incentives for prompt payment, and bankruptcy frequency. Elimi-
nating consumer receivables therefore increases the homogeneity of our sample.
5
There are several non-routine items on Schedule II—recoveries, reclassi?cations,
reserve adjustments, currency translation adjustments, miscellaneous adjustments,
etc. We use recoveries as a catch-all label for all of the non-routine items on Schedule
II. In some instances, ?rms combine write-offs with recoveries (or other non-routine
items) and report a negative value for write-offs to the extent that recoveries (or other
non-routine items) are oppositely signed and exceed write-offs. Such negative values
are uncommon. Only two such instances arose in Appendix A (this appendix is
discussed in detail later). In these few instances, we set a lower bound of zero on
negative write-offs.
6
There is a one-year time period assumption inherent in the allowance. The
allowance is for current receivables, so it should have a write-off horizon of one year
or less. As a result, our use of one year of leading write-offs (rather than, for example,
two years of leading write-offs) is appropriate.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 5
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
ference between the allowance and leading write-offs expressed as
a percent of net income. Firms are ranked in descending order by
column (a). Formal tests are discussed below, but notice that 123
?rms (37.50% of the 328 ?rms analyzed) have a difference in Col-
umn (a) in excess of ?ve percent of net income in the year preced-
ing SAB 108’s effective date which suggests their allowance may be
overestimated.
We also consider various alternative quantitative measures in
Appendix A because it is possible that both managers and their
auditors use other potential metrics when determining the amount
of the allowance to report in the ?nancial statements. Column (b)
is a variant of Column (a), the only difference being that leading
write-offs are multiplied by a factor of two, consistent with an
alternative benchmark discussed in Riley and Pasewark (2009).
Column (c) provides the percentage of accounts receivable that
each ?rm has reserved, and Column (d) provides the industry
adjusted percentage of accounts receivable that each ?rm has
reserved. Column (e) provides the allowance as a percentage of
total assets. Following Jackson and Liu (2010), Column (f) provides
the number of years of future write-offs that each ?rm has
reserved.
7,8
A value of ?ve would, for example, indicate that the ?rm
could forego recording bad debt expense for ?ve years before write-
offs exhaust the allowance. Thus, higher ratios might suggest ?rms
are overestimating the allowance, whereas a value of 1–2 for a typ-
ical year may be more appropriate (Riley & Pasewark, 2009). Column
(g) provides the standard deviation of Column (f), which is calculated
on a ?rm-by-?rm basis using all available observations for each ?rm
during the sample period. The standard deviation, relative to the
mean value, helps to give readers a sense of the variable’s temporal
stability. Column (h) indicates the percentage of the allowance that
has not been consumed by leading write-offs. Lower ratios,
approaching zero, suggest ?rms’ current allowance provides a better
estimate of future write-offs. Column (i) is the ratio of bad debt
expense to write-offs. This ratio should be approximately one in
the long-run. Column (j) is the standard deviation of the ratio of
bad debt expense to write-offs.
Our discussion focuses on the mean and median values in the
last two rows of Appendix A because our interest lies in character-
izing the sample of ?rms that appear to have a potentially overes-
timated allowance, not to characterize particular ?rms in the
sample. The mean (median) value of Column (a) is 20.00 (11.10),
which suggests that the overstatement of the allowance is poten-
tially large in relation to reported earnings. The mean (median)
value of Column (b) is smaller than Column (a), but nonetheless
fairly large. Firms reserve a substantial fraction of their receivables
as indicated by the mean (median) value of Column (c) of 4.18
(3.45). These values are large relative to the industry medians, as
indicated by the positive mean (median) values in Column (d) of
1.10 (0.50). The mean (median) value of Column (e) is 0.96
(0.68), which reveals that the allowance is moderately signi?cant
in relation to total assets. However, we note that SAB 108 requires
registrants to consider both the income statement and balance
sheet when making materiality judgments.
The mean (median) value of Column (f) is 5.51 (3.93) which
indicates that the allowance will not be exhausted any time soon
even if no additional reserve is recorded for several years. The
mean value of Column (g) of 5.38 suggests that the ratio in Column
(f) may be somewhat unstable over time. This is consequential
because if the relationship between the allowance and leading
write-offs is erratic, it is possible that conservatism in the
allowance is merely intended to create a ‘‘cookie jar” reserve that
is increased in good years and drawn down in bad years to
manipulate earnings. The mean (median) value of Column (h) is
71.23 (74.55) which indicates that about three quarters of the
allowance is unused by leading write-offs. The mean (median)
value of Column (i) is 1.98 (1.00) which indicates that bad debt
expense and write-offs are somewhat similar to one another. The
mean (median) value of Column (j) of 2.08 (0.75) suggests that
the ratio in Column (i) is fairly stable over time. Overall, the data
in Appendix A suggests that these ?rms have potentially
overestimated allowances even after considering various alterna-
tive quantitative measures.
3.3. Descriptive information
Table 1 provides the industry pro?le for the 123 ?rms that com-
prise the ?ve percent of net income sample. While the sample con-
sists of a broad cross-section of different industries, there is
clustering in SIC codes 28, 35, and 73. Table 2 provides descriptive
statistics for the 1079 ?rm-years (123 individual ?rms). Panel A
provides descriptive statistics for ?rm size and other variables fol-
lowed by descriptive statistics for the dependent and independent
variables. Panel B provides descriptive statistics for the dependent
variables partitioned by SAB 108’s effective date.
9
3.4. Test of whether ?rms make SAB 108 adjustments
Hypothesis 1 predicts that ?rms will tend not to make SAB 108
adjustments in the year it became effective. For each of the 123
?rms identi?ed above as having a potentially overestimated allow-
ance (see Appendix A), we examined the Form 10-K to determine
whether a SAB 108 adjustment related to the allowance was made
in the year that SAB 108 became effective. None of the 123 ?rms
made any such adjustment.
10
The absence of any SAB 108 adjust-
ments to the allowance is of interest because the mean calculated
potential overestimation of the allowance was approximately 20%
of net income and the mean number of years of leading write-offs
reserved was approximately six.
Further, many of the ?rms that we analyze have a potentially
overestimated allowance that is far in excess of various commonly
used benchmarks. In fact, the 25 largest overestimates are, on aver-
age, 58% of net income and the allowance associated with these
overestimates covers approximately seven years of leading write-
offs. Given the evidence in Appendix A, it appears that, at ?rst
blush, a large fraction of the ?rms we analyze needed to make
SAB 108 adjustments related to the allowance. However, none of
these ?rms made such adjustments, but, instead, continue to main-
tain conservative allowances in the year SAB 108 became effec-
tive.
11
Hypothesis 1 is therefore supported.
3.5. Tests of whether the allowance declines after the effective date of
SAB 108
Hypothesis 2a predicts that the allowance declines in the years
after SAB 108 became effective. To test this hypothesis, we
7
When leading write-offs are zero, Column (f) is unde?ned. The issue that we face
is what value to assign to Column (f) when leading write-offs are zero. Given that the
mean and median values in Column (f) of Appendix A are approximately 6 and 4,
respectively, we selected a value of 5.
8
Write-offs exhibit no signi?cant time-trend during our sample period.
9
To mitigate the effect of outliers, continuous regression variables are winsorized
at the 99th percentile.
10
We do, however, ?nd that 22 ?rms made a SAB 108 adjustment unrelated to the
allowance. Also, while none of the ?rms in our ?nal sample made a SAB 108
adjustment related to the allowance, three ?rms in our initial sample made such
adjustments. As a percentage of net income, the SAB 108 adjustments related to the
allowance were 20.66%, 5.02%, and 1.02%.
11
SAB 99 (SEC, 1999) and Auditing Standard No. 14 (AICPA, 2010) prohibit auditors
from offsetting one material misstatement against another material misstatement. As
a result, the absence of a SAB 108 adjustment related to the allowance is not the result
of concurrent offsetting misstatements.
6 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
estimate the following pooled, cross-sectional regression using the
model from Jackson and Liu (2010):
ALLOW
it
=SALE
it
¼ b
0
þb
1
AR
it
=SALE
it
þb
2
WO
it
=SALE
it
þb
3
WO
itþ1
=SALE
it
þb
4
ARTO IND
it
þb
5
ALT IND
it
þb
6
SD SALE IND
it
þb
7
BANKRUPT
it
þb
8
POST SAB 108
it
þe
it
ð1Þ
where ALLOW
it
is the allowance for uncollectible accounts; SALE
it
is
net sales; AR
it
is gross accounts receivable; WO
it
is write-offs of
uncollectible accounts; ARTO_IND
it
is the median industry accounts
receivable turnover ratio (de?ned as SALE
it
divided by the average
accounts receivable) computed using all ?rms with available data
on Compustat;
12
ALT_IND
it
is the median industry Altman (1968)
Z-score computed using all ?rms with available data on Compustat;
SD_SALE_IND
it
is the median industry standard deviation of sales
computed on a ?rm-by-?rm basis using the four quarters of each
year for all ?rms with available data on Compustat; BANKRUPT
it
is
the total number of business bankruptcies in the U.S. economy
scaled by the total number of corporate tax ?lers;
13
POST_SAB_108
it
is a dichotomous variable coded as 1 for the ?scal years after SAB
108 became effective (November 15, 2006), and 0 otherwise; i and
t are ?rm and year subscripts, respectively. Firm ?xed effects are
included in Eq. (1) but are not shown, and t-statistics are calculated
using standard errors clustered by ?rm. Hypothesis 2a predicts that
the coef?cient on POST_SAB_108
it
will be negative, indicating that
the allowance is lower in the post-SAB 108 years.
14
To enhance comparability over time and to control for potential
survivorship bias, we analyze two samples. The ?rst sample is
referred to as the ‘‘full sample.” This sample contains all observa-
tions during the sample period 2001 through 2009 for ?rms that
have an allowance exceeding our quantitative benchmark in the
year that SAB 108 became effective. The second sample is referred
to as the ‘‘constant sample.” This sample is similar to the full sam-
ple except that it only contains observations for ?rms that have
complete data for the entire nine-year sample period.
Table 3 reports regression results for Eq. (1) using the full and
constant samples. The coef?cient on each of the economic determi-
nants is positive and signi?cant (p-values 6 0.020) using both the
full and constant samples. The coef?cients on three of the controls
for temporal changes in the riskiness of receivables (ARTO_IND
it
,
ALT_IND
it
, and SD_SALE_IND
it
) are insigni?cant (p-values > 0.10),
while the coef?cient on one of the controls (BANKRUPT
it
) is signif-
icant in the predicted direction (p-values 6 0.002).
15
Table 1
Industry pro?le of sample.
SIC code SIC description # %
13 Oil and Gas Extraction 4 3.25
20 Food and Kindred Products Manufacturers 4 3.25
25 Furniture and Fixture Manufacturers 6 4.88
27 Printing and Publishing 5 4.07
28 Chemicals and Allied Products Manufacturers 12 9.76
30 Rubber and Misc. Plastics Manufacturers 3 2.44
35 Industrial and Commercial Machinery Manufacturers 11 8.94
36 Electronic and Electrical Equipment Manufacturers 3 2.44
37 Transportation Equipment Manufacturers 6 4.88
38 Measuring and Analyzing Equipment Manufacturers 7 5.69
50 Wholesale, Durable Goods 7 5.69
51 Wholesale, Non-Durable Goods 7 5.69
73 Business Services 15 12.20
Other 33 26.83
Total 123 100.00
The ‘‘Other” category captures count data for SIC codes that have fewer than three
different ?rms.
Table 2
Descriptive statistics.
Panel A: Descriptive statistics
(n = 1079)
Mean Median Std. dev.
Firm size and other variables
TA
it
($mil.) 6225.23 1859.08 12620.11
SALE
it
($mil.) 7279.60 2347.00 14121.54
MVE
it
($mil.) 5422.32 1641.22 15434.56
AR
it
/TA
it
(%) 24.32 20.72 13.14
AR
it
/CA
it
(%) 50.71 48.24 16.29
ALLOW
it
/AR
it
(%) 4.46 3.52 3.41
BDE
it
/NI
it
(%) 4.70 3.24 32.84
WO
it
/NI
it
(%) 4.38 3.31 46.91
BDE
it
/SHO
it
($ per share) 0.15 0.08 0.19
Dependent variables from regressions
ALLOW
it
/SALE
it
(Â100) 0.81 0.59 0.76
BDE
it
/SALE
it
(Â100) 0.32 0.19 0.51
Independent variables from regressions
AR
it
/SALE
it
(Â100) 18.43 16.90 8.38
WO
it
/SALE
it
(Â100) 0.35 0.18 0.57
WO
it+1
/SALE
it
(Â100) 0.34 0.18 0.55
ARTO_IND
it
6.64 6.06 1.98
ALT_IND
it
4.27 4.27 1.45
SD_SALE_IND
it
7.54 5.22 8.09
BANKRUPT
t
(Â100) 0.66 0.69 0.18
ALLOW
itÀ1
/SALE
it
(Â100) 0.79 0.58 0.78
TOP_DEC
it
(%) 10.00 0.00 30.15
BOT_DEC
it
(%) 10.00 0.00 30.15
RECOV
it
/SALE
it
(Â100) 0.02 0.00 0.11
Panel B: Descriptive statistics for dependent variables partitioned by the effective
date of SAB 108
?
Pre-SAB 108 years (n = 643)
Mean Median Std. dev.
ALLOW
it
/SALE
it
(Â100) 0.88 0.66 0.75
BDE
it
/SALE
it
(Â100) 0.36 0.21 0.51
Post-SAB 108 years (n = 436)
Mean Median Std. dev.
ALLOW
it
/SALE
it
(Â100) 0.70 0.49 0.75
BDE
it
/SALE
it
(Â100) 0.27 0.14 0.50
Variables are de?ned as follows: TA
it
is total assets; SALE
it
is net sales; MVE
it
is
market value of equity; AR
it
is gross accounts receivable; CA
it
is current assets;
ALLOW
it
is the allowance for uncollectible accounts; BDE
it
is bad debt expense; NI
it
is net income; SHO
it
is common shares outstanding; WO
it
is write-offs of uncol-
lectible accounts; ARTO_IND
it
is the median industry accounts receivable turnover
ratio (de?ned as SALE
it
divided by the average accounts receivable) computed using
all ?rms with available data on Compustat; ALT_IND
it
is the median industry
Altman (1968) Z-score computed using all ?rms with available data on Compustat;
SD_SALE_IND
it
is the median industry standard deviation of sales computed on a
?rm-by-?rm basis using the four quarters of each year for all ?rms with available
data on Compustat; BANKRUPT
it
is the total number of business bankruptcies in the
U.S. economy scaled by the total number of corporate tax ?lers; TOP_DEC
it
is a
dichotomous variable coded as 1 for ?rms in the sample top decile of net income
divided by total assets, and 0 otherwise; BOT_DEC
it
is a dichotomous variable coded
as 1 for ?rms in the sample bottom decile of net income divided by total assets, and
0 otherwise; RECOV
it
is recoveries of previously written-off accounts receivable; i
and t are ?rm and year subscripts, respectively. Continuous regression variables are
winsorized at the 99th percentile.
*
Means and medians are signi?cantly different between the pre- and post-SAB
108 years (p-values < 0.01).
12
Industry is de?ned at the two-digit SIC code level. When there are fewer than ten
total ?rms in a two-digit SIC code, we compute ALT_IND
it
, ARTO_IND
it
, and
SD_SALE_IND
it
, at the one-digit SIC code level.
13
We obtained the number of bankruptcies from the American Bankruptcy Institute
at www.abiworld.org under the heading ‘‘U.S. Bankruptcy Filings 1980–2009
(Business, non-Business, Total).”
14
Studies often include both ?rm and year ?xed effects in pooled, cross-sectional
regressions. Because the test variable in the regressions reported in this study have a
time dimension, year ?xed effects are excluded.
15
When we specify the industry adjusted controls for temporal changes in the
riskiness of receivables in their non-industry adjusted form, the variables remain
insigni?cant in all regressions and none of our inferences or conclusions change.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 7
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
In Table 3, the coef?cient on the test variable, POST_SAB_108
it
,
is negative in both the full sample (coef?cient = À0.0006,
t-statistic = À1.68, p-value = 0.094) and constant sample
(coef?cient = À0.0005, t-statistic = À1.42, p-value = 0.156). The
coef?cients on POST_SAB_108
it
are negative but, at best, marginally
signi?cant, indicating limited support for Hypothesis 2a. Moreover,
the economic signi?cance of the decline is modest. Notice that the
coef?cient on POST_SAB_108
it
for the full sample is only À0.0006,
which implies that the mean allowance has declined by approxi-
mately 0.06% of sales. Thus, although ?rms may not have felt the
need to make a SAB 108 adjustment immediately after its enact-
ment, SAB 108 does appear to have caused the allowance to decline
modestly, perhaps because managers of ?rms now ?nd it more
dif?cult to replenish an already large allowance balance by over-
accruing bad debt expense. We consider this possibility next.
3.6. Tests of whether bad debt expense declines after the effective date
of SAB 108
Hypothesis 2b predicts that bad debt expense declines in the
years after SAB 108 became effective. To test this hypothesis, we
estimate the following pooled, cross-sectional regression using
the model from Jackson and Liu (2010) and McNichols and
Wilson (1988):
BDE
it
=SALE
it
¼ b
0
þb
1
ALLOW
itÀ1
=SALE
it
þb
2
WO
it
=SALE
it
þb
3
WO
itþ1
=SALE
it
þb
4
TOP DEC
it
þb
5
BOT DEC
it
þb
6
RECOV
it
=SALE
it
þb
7
ARTO IND
it
þb
8
ALT IND
it
þb
9
SD SALE IND
it
þb
10
BANKRUPT
it
þb
11
POST SAB 108
it
þe
it
ð2Þ
where SALE
it
, ALLOW
it
, WO
it
, ARTO_IND
it
, ALT_IND
it
,
SD_SALE_IND
it
, BANKRUPT
it
, and POST_SAB_108
it
are de?ned
above; BDE
it
is bad debt expense; TOP_DEC
it
is a dichotomous vari-
able coded as 1 for ?rms in the sample top decile of net income
divided by total assets, and 0 otherwise; BOT_DEC
it
is a dichoto-
mous variable coded as 1 for ?rms in the sample bottom decile of
net income divided by total assets, and 0 otherwise; RECOV
it
is
recoveries of previously written-off accounts receivable (as noted
in Section 3, this is a catch-all label for non-routine items on Sched-
ule II). Firm ?xed effects are included in Eq. (2) but are not shown,
and t-statistics are calculated using standard errors clustered by
?rm. The ?rst six variables in Eq. (2) are economic determinants
from McNichols and Wilson (1988), while the next four variables
are controls for temporal changes in the riskiness of receivables
from Jackson and Liu (2010). Hypothesis 2b predicts that the coef-
?cient on POST_SAB_108
it
will be negative.
Table 4 reports regression results for Eq. (2) using the full and
constant samples. The coef?cient on each of the economic determi-
nants is signi?cant (p-values < 0.10) in the predicted direction
using the full and constant samples, except for the coef?cient on
TOP_DEC
it
which is insigni?cant in both samples (p-values > 0.10)
and the coef?cients on ALLOW
it-1
/SALE
it
, WO
it
/SALE
it
, and
BOT_DEC
it
which are insigni?cant (p-values > 0.10) in the constant
sample. The coef?cients on controls for temporal changes in the
riskiness of receivables (ARTO_IND
it
, ALT_IND
it
, and
SD_SALE_IND
it
) are insigni?cant (p-values > 0.10), while the coef?-
cient on one of the controls (BANKRUPT
it
) is signi?cant in the pre-
dicted direction (p-values < 0.01).
In Table 4, the coef?cient on the test variable, POST_SAB_108
it
, is
negative but, at best, marginally signi?cant in both the full sample
(coef?cient = À0.0005, t-statistic = À1.43, p-value = 0.154) and con-
stant sample (coef?cient = À0.0005, t-statistic = À1.40, p-value =
0.162). These results provide very limited support for Hypothesis
2b. Thus, not only do ?rms not make SAB 108 adjustments, but they
do not reduce their annual bad debt expense to gradually reduce the
allowance balance.
16
Table 3
Regressions of the allowance for uncollectible accounts on economic determinants, controls for temporal changes in the riskiness of receivables, and the test variable. ALLOW
it
/
SALE
it
= b
0
+ b
1
AR
it
/SALE
it
+ b
2
WO
it
/SALE
it
+ b
3
WO
it+1
/SALE
it
+ b
4
ARTO_IND
it
+ b
5
ALT_IND
it
+ b
6
SD_SALE_IND
it
+ b
7
BANKRUPT
it
+ b
8
POST_SAB_108
it
+ e
it
.
Variable Pred. Full sample (n = 1079) Constant sample (n = 873)
Sign Coeff. t-stat. p-value Coeff. t-stat. p-value
Economic determinants
Intercept ? À0.0040 À1.87 0.062 À0.0048 À1.85 0.064
AR
it
/SALE
it
+ 0.0341 9.21
This study examines whether firms with large conservative allowances for uncollectible accounts on their
balance sheets make significant adjustments to increase net assets in response to Staff Accounting
Bulletin No. 108 (SAB 108). To the extent adjustments are not made, we examine the qualitative reasons
SAB 108 may have had a limited effect on these large conservative accounting estimates. In the year of
enactment, SAB 108 required public companies to adjust material accumulations on the balance sheet
that may have occurred prior to enactment of SAB 108 after considering all relevant quantitative and
qualitative factors. Our archival results find that no firms made a SAB 108 adjustment and while sample
firms tend to, at best, make small reductions to the allowance subsequent to SAB 108’s effective date,
most sample firms continue to have potentially overestimated allowances several years later. To understand
the qualitative forces underlying this finding, we interview key individuals involved in the financial
reporting process and we survey experienced financial executives employed at our sample firms. The
findings reveal that organizational forces favor conservatism because being over-reserved helps firms
avoid income-decreasing surprises while being under-reserved appears careless. We conclude that conservatism
remains a deeply engrained feature of accounting thought and practice, in contrast to recent
emphasis by accounting standard setters on neutrality over conservatism.
Conservatism and Staff Accounting Bulletin No. 108
q
Thomas G. Canace
a
, Richard C. Hat?eld
b,?
, Scott B. Jackson
c
a
Wake Forest University, United States
b
University of Alabama, United States
c
University of South Carolina, United States
a r t i c l e i n f o
Article history:
Received 22 August 2012
Revised 14 August 2015
Accepted 19 August 2015
Available online xxxx
a b s t r a c t
This study examines whether ?rms with large conservative allowances for uncollectible accounts on their
balance sheets make signi?cant adjustments to increase net assets in response to Staff Accounting
Bulletin No. 108 (SAB 108). To the extent adjustments are not made, we examine the qualitative reasons
SAB 108 may have had a limited effect on these large conservative accounting estimates. In the year of
enactment, SAB 108 required public companies to adjust material accumulations on the balance sheet
that may have occurred prior to enactment of SAB 108 after considering all relevant quantitative and
qualitative factors. Our archival results ?nd that no ?rms made a SAB 108 adjustment and while sample
?rms tend to, at best, make small reductions to the allowance subsequent to SAB 108’s effective date,
most sample ?rms continue to have potentially overestimated allowances several years later. To under-
stand the qualitative forces underlying this ?nding, we interview key individuals involved in the ?nancial
reporting process and we survey experienced ?nancial executives employed at our sample ?rms. The
?ndings reveal that organizational forces favor conservatism because being over-reserved helps ?rms
avoid income-decreasing surprises while being under-reserved appears careless. We conclude that con-
servatism remains a deeply engrained feature of accounting thought and practice, in contrast to recent
emphasis by accounting standard setters on neutrality over conservatism.
Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Prior research suggests that many ?rms have accumulated large
conservative allowances for uncollectible accounts on their bal-
ance sheets and that those ?rms may use conservative balances
to manage earnings in future periods (Jackson & Liu, 2010). Staff
Accountant Bulletin No. 108 (SAB 108) requires public companies,
under the iron-curtain approach, to consider the materiality of
potential misstatements that have accumulated on the balance
sheet through multiple non-material income statement differ-
ences. In the year of enactment, SAB 108 required ?rms to make
adjustments for any material accumulations that may have
occurred prior to enactment after considering all relevant quantita-
tive and qualitative factors. The enactment of SAB 108 allows for a
‘‘natural experiment” to determine whether such speci?c guidance
would initiate ?nancial statement adjustments for large conserva-
tive allowances. Further, to the extent that adjustments are not
made, the enactment of SAB 108 allows us to use qualitative meth-
ods to investigate speci?c reasons why conservative allowances
persist on the balance sheet. Thus, we interview and survey the
key players (i.e., audit partners, ?rm managers and regulators) to
better understand why SAB 108 had a limited effect on large con-
servative allowance balances.
1
Evidence in various domains of accounting suggests that com-
panies may be reluctant to increase net assets because doing so
(1) runs counter to conservatism (Basu, 1997; Givoly & Hayn,
2000; Hirshleifer & Teoh, 2009; Watts, 2003a; Watts, 2003b) and
(2) may reduce their ability to manage earnings in the future
(Barton & Simko, 2002; Jackson & Liu, 2010; McNichols & Wilson,
1988; Penman, 2001). Thus, a better understanding of whether
such balances are adjusted under SAB 108 is important because
the SEC has indicated that uncorrected misstatements distort the
balance sheet and are contrary to the best interests of ?nancialhttp://dx.doi.org/10.1016/j.aos.2015.08.002
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
q
We thank the editor, two anonymous reviewers, Charles Boster, Steven Jackson,
Marsha Keune, Kelvin Liu, Joel Owens, Tammie Schaefer, Mike Shaub, and Mark
Taylor for providing insightful comments and suggestions. We also thank Guang
Rong for assistance with data collection. Professor Canace is especially grateful to
the Wake Forest University School of Business for providing research funding
through the summer research grant program.
?
Corresponding author at: Culverhouse School of Accounting, University of
Alabama, Tuscaloosa, AL 35487-0220, United States.
E-mail address: [email protected] (R.C. Hat?eld).
1
In our setting, the practice of conservatism refers to a preference for reporting
estimates in the direction of understatement rather than overstatement of net income
and net assets (Financial Accounting Standards Board (FASB), 1980).
Accounting, Organizations and Society xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er . com/ l ocat e/ aos
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
statement users, especially if ‘‘cookie jar” reserves facilitate earn-
ings management in the future (Levitt, 1998; Securities &
Exchange Commission (SEC), 2006). Indeed, the SEC identi?ed
these among the reasons for issuing SAB 108. Yet, SAB 108 also
requires managers and their auditors to consider all relevant quan-
titative and qualitative information, which raises the importance of
examining both factors before making conclusions about necessary
adjustments for material accumulations on the balance sheet.
Thus, we explore both aspects of this decision because there may
be qualitative considerations that help to explain managers’
estimates.
In the ?rst stage of this study, we focus on quantitative data
related to the allowance for uncollectible accounts to consider the
potential for conservative accumulations on the balance sheet. Prior
research shows that this account has grown over time because
?rms seemingly over-accrue bad debt expense, perhaps causing
the allowance to become overly conservative in the long-run
(Jackson & Liu, 2010). Using future realizations of the allowance
by examining ?rms’ publicly disclosed write-offs (McNichols &
Wilson, 1988), we calculate a quantitative difference (potential
overestimation) as a measure of allowance accuracy by subtracting
one-year leading write-offs from the allowance for uncollectible
accounts. We then consider whether these quantitative differences
remain on the balance sheet of our sample ?rms and if the large
conservative differences persist after the promulgation of SAB 108.
In the year immediately preceding the effective date of SAB 108,
our archival analyses suggest that 37.50% of the ?rms we analyze
have potential overestimations of their allowance in excess of ?ve
percent of net income. The mean allowance for these ?rms is suf-
?cient to cover ?ve years of leading write-offs. In fact, the 25 lar-
gest calculated quantitative differences average almost 60% of
net income. Despite the size of the accumulated allowance for
uncollectible accounts on the balance sheet, none of the ?rms we
analyze make a SAB 108 adjustment to correct a potential accumu-
lated error in the allowance account. Moreover, while we ?nd that
the allowance declined modestly, on average, following SAB 108’s
effective date, most sample ?rms continue to have a potentially
overestimated allowance four periods after SAB 108’s effective
date. The SEC expressly prohibits improper assets and liabilities
from remaining on the balance sheet in perpetuity (SEC, 2006).
Thus, the fact that our sample ?rms opted for no adjustment, or
at best, small delayed adjustment, of the allowance suggests that
the differences we calculate, though often quite large, are not being
perceived as material misstatements worthy of correction.
The second stage of the study employs qualitative methods to
better understand why ?rms, and their auditors, are not applying
SAB 108 to this apparent overestimation of the allowance for
uncollectible accounts. We conduct open-ended interviews with
ten key individuals involved in the ?nancial reporting process,
including ?nancial executives, audit partners, and regulators at
the SEC and PCAOB to shed light on the archival ?ndings. Experi-
enced ?nancial executives indicate that organizational forces favor
conservatism with respect to the allowance and auditors do little
to deter it. Indeed, these executives believe that being over-
reserved helps the ?rm to avoid income-decreasing surprises in
an uncertain environment, while being under-reserved appears
careless. In turn, auditors indicate that there is little risk associated
with having a conservative allowance and are unlikely to oppose
such a preference. These partners framed the issue not as one
involving an over-reserved allowance, but as one involving a con-
servative balance. The accounting regulators commented (1) that
some SEC reviewers of ?lings may evaluate the allowance in rela-
tion to the balance sheet rather than leading write-offs, which
might make the overestimation seem less consequential, and (2)
that a plausible explanation for overestimated allowances is the
desire to be conservative.
To get a better understanding of the speci?c actions of the ?rms
in our sample, we also survey experienced ?nancial executives
employed at our sample ?rms. Survey ?ndings reveal that many
individuals and groups internal to these ?rms strongly favor con-
servatism with respect to the allowance and no individuals or
groups oppose it. Moreover, the responses suggest that ?nancial
executives primarily resist drawing down the allowance because
doing so is at odds with maintaining a conservative balance. Thus,
a conservative allowance appears to have become a somewhat per-
manent feature of the balance sheet.
Taken as a whole, it appears that SAB 108 would certainly be
applicable to these ?rms if we simply considered the quantitative
materiality of these potential misstatements. The fact that these
?rms and their auditors do not perceive SAB 108 to apply suggests
that they do not perceive these overestimates as material misstate-
ments. That is, if a ?rm has a large allowance, but has some justi-
?cation for it, the auditor may just view this as a very conservative
balance, but not a misstatement. There appears to be some asym-
metry in the judgment of misstatement depending upon whether
the amount in question is income/asset increasing or decreasing.
It is also interesting that, despite the potential for overestimated
allowances to be used as ‘‘cookie jar” reserves, they still occur after
SAB 108 and ?rms’ management suggest they keep these reserves
for the very reason of having some cushion to avoid hits to
earnings.
Our study contributes to the accounting literature by providing
the ?rst evidence that conservative accumulations on the balance
sheet persist subsequent to speci?c guidance about eliminating
accumulations (i.e., SAB 108). Our archival evidence that ?rms
make small reductions to conservative allowance balances after
SAB 108 suggests that the SEC may have modestly attenuated
?rms’ conservatism in the reporting of these estimates on the bal-
ance sheet. However, interview and survey evidence also suggests
that managers believe the maintenance of large conservative
allowances is appropriate.
Further, discussions with audit partners and regulators reveal
that such accumulations are unlikely to cause them to object to
conservative balances. So, while regulators create speci?c guid-
ance about dealing with material accumulations (e.g., SAB 108),
our quantitative and qualitative results suggest that conservative
balances in the ?nancial statements are still viewed as appropri-
ate by ?rms and seemingly permissible by auditors/regulators.
These ?ndings may be important for future studies examining
accounting conservatism for two reasons. First, our evidence
highlights the need to consider both quantitative and qualitative
factors when making conclusions about ?rms’ accounting esti-
mates. Second, our evidence should be considered in light of
recent changes to the Conceptual Framework (Concepts State-
ment No. 8, ‘‘Conceptual Framework for Financial Reporting”
FASB, 2010) which places emphasis on neutrality over conser-
vatism. The revised framework does not include conservatism
as a qualitative characteristic because it has become associated
with the deliberate understatement of net assets and net income
and is in con?ict with the quality of neutrality which encom-
passes freedom from bias.
Moreover, the results have at least two practical implications
given recent regulatory concern and research that suggest such
conservative balances may increase the likelihood of earnings
management. First, the large accumulations that we identify seem
to have become part of the status quo and auditors may have
become uncritically accustomed to them as interview evidence
shows they view them as bearing little inherent risk. Our ?ndings
could sensitize auditors to the possibility that accumulations in the
allowance account are systematic and excessive when assessed
with commonly accepted quantitative benchmarks. Second, these
?ndings could prompt the SEC of?cials to scrutinize allowance
2 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
balances that appear to be ‘‘overly conservative” and issue
comment letters to ?rms, which, among other things, opens a
dialogue about conservatism.
The remainder of this paper proceeds as follows. The next sec-
tion provides background, discusses relevant prior research, and
develops our hypotheses. Section 3 describes the archival methods
and results. Section 4 describes our additional archival analyses.
Section 5 reports the results of open-ended interviews with ?nan-
cial executives, auditors, and regulators. Section 6 discusses our
survey of experienced ?nancial executives employed at our sample
?rms. The ?nal section summarizes the results and describes cer-
tain limitations.
2. Background, relevant prior research, and hypotheses
development
2.1. Summary of Staff Accounting Bulletin No. 108
The SEC issued SAB 108 in 2006 to address two concerns—(1)
the diversity in practice with respect to quantifying ?nancial state-
ment misstatements and (2) the potential for improper amounts to
accumulate on the balance sheet.
2
Prior to the issuance of SAB 108,
auditors elected to use the ‘‘rollover” approach or ‘‘iron curtain”
approach to quantify misstatements, but auditors generally did not
consider both approaches simultaneously. The rollover approach
quanti?es a misstatement by looking at the amount of the error orig-
inating on the income statement in the current year. This method
ignores any accumulated balance sheet misstatement originating
in prior years. Conversely, the iron curtain approach quanti?es a
misstatement by looking at the amount of the misstatement on
the balance sheet regardless of when the misstatement originated.
To illustrate the differences between the rollover approach and
iron curtain approach, consider Fig. 1, which contemplates the fol-
lowing circumstances. A ?rm records an excessive expense accrual
each year for ?ve years. The amount by which the expense accrual
exceeds the ‘‘correct” accrual is $10 per year. At the same time, the
?rm reports income of $1000 each year. If the rollover approach is
used to quantify ?nancial statement misstatements then the error
is $10 each year. Expressed as a percent of income, the excessive
expense accrual is only one percent of income each year. If the iron
curtain approach is used to quantify ?nancial statement misstate-
ments then the error is $10 in year 1, $20 in year 2, $30 in year
3, $40 in year 4, and $50 in year 5. Expressed as a percent of
income, the excess accrual is one, two, three, four, and ultimately
?ve percent of earnings in years 1 through 5, respectively.
The provisions of SAB 108 prohibit exclusive reliance on either
the rollover approach or iron curtain approach. Instead, SAB 108
requires use of the ‘‘dual” approach, which is the approach initially
recommended by Nelson, Smith, and Palmrose (2005). The dual
approach requires auditors and their clients to consider the magni-
tude of a misstatement under both the rollover approach and iron
curtain approach and then make an adjustment whenever the mis-
statement is material under either approach after considering the
relevant quantitative and qualitative factors. Thus, in the illustra-
tive example above, the dual approach would seemingly require
an adjustment to correct the accumulated misstatement in the
allowance account based upon the quantitative materiality of the
error under the iron curtain approach.
In the year SAB 108 became effective, ?rms are not required to
restate prior period ?nancial statements if management properly
applied either the rollover approach or iron curtain approach. This
transition feature allows ?rms that have accumulated misstatements
on the balance sheet to avoid restating ?nancial statements in
the years that led to the misstatement. However, ?rms must re?ect
the cumulative effect of the initial application of SAB 108 in the
carrying amounts of assets and liabilities as of the beginning
of the ?rst ?scal year after SAB 108 became effective, with an off-
setting adjustment to the opening balance of retained earnings.
SAB 108 expressly prohibits delayed or partial corrections of mis-
statements. Errors that existed at the time SAB 108 became effec-
tive but were not corrected on a timely basis must be corrected by
restating the prior period ?nancial statements in accordance with
Statement of Financial Accounting Standards (SFAS) No. 154,
Accounting Changes and Error Corrections (FASB, 2006; ASC §250).
2.2. Research related to Staff Accounting Bulletin No. 108
Nelson et al. (2005) examine the effect of the two methods of
quantifying ?nancial statement misstatements, ?nding that audi-
tors are more likely to require their clients to book misstatements
under the approach that makes the misstatement appear more
material. Also, while Nelson et al. (2005) provide participants
with the information necessary to evaluate misstatements using
either the rollover approach or iron curtain approach, partici-
pants’ audit adjustment decisions were strongly in?uenced by
how the information was presented to them (i.e., presented in
accordance with the rollover approach or the iron curtain
approach). This ?nding is consistent with the SEC’s concern that
the rollover approach may allow improper amounts to accumu-
late on the balance sheet. Nelson et al. (2005) also consider
whether income increasing adjustments would be affected by
employing the iron curtain approach. In their cases with income
increasing adjustments, auditors are least likely to require the
adjustment. They suggest that auditors may waive these
adjustments, thereby permitting conservatism, because auditor
training and litigation risk encourage detecting income overstate-
ments, not understatements.
Keune and Johnstone (2009) examine SAB 108 disclosures, ?nd-
ing that misstatements disclosed under SAB 108 (1) tend to be
income-decreasing, (2) tend to be made by larger ?rms, (3) are
concentrated in regulated industries, (4) are underrepresented in
the manufacturing industry, (5) are clustered in current liabilities,
leases, revenue recognition, and deferred taxes, and (6) are not
overrepresented in reserve accounts. They point out that these
?ndings are consistent with prior evidence that auditors tend to
allow conservative reporting (Kinney & Martin, 1994; Wright &
Wright, 1997) and are less likely to require adjustments for subjec-
tive accounts that require estimation (Braun, 2001; Ng, 2007).
Therefore, their results suggest that considering both the direction
and account type of the adjustment are critical to a consideration
of why ?rms make SAB 108 adjustments.
Moreover, several studies examine the association between var-
ious market-related variables and SAB 108 corrections. Omer,
Shelley, and Thompson (2011) ?nd that ?rms reporting SAB 108
corrections are no more likely to meet or beat analysts’ consensus
forecasts than ?rms not reporting SAB 108 corrections. Keune and
Johnstone (2012) ?nd that auditors are more likely to waive qual-
itatively material misstatements as analyst following increases,
but they also ?nd that auditors are less likely to waive misstate-
ments as audit fees increase. Omer, Shelley, and Thompson
(2012) ?nd a negative association between factors that are posited
to reduce audit quality and investors’ response to SAB 108 correc-
tions, which suggests that investors react more negatively to
corrections when auditor independence appears weaker.
To date, prior research has empirically examined corrected mis-
statements that are disclosed in audited ?nancial statements; prior
research has not empirically examined whether large and poten-
tially material misstatements go uncorrected in audited ?nancial
2
SAB 108 was issued on September 13, 2006 and its provisions became effective
for the ?rst ?scal year ending after November 15, 2006.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 3
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
statements, and, if so, the qualitative reasons adjustments are not
made. We speci?cally identify large accumulations (potential mis-
statements based on the iron curtain approach) on the balance
sheet prior to and after the enactment of SAB 108 to consider
whether or not management and auditors perceive that SAB 108
applies.
2.3. Are large conservative accumulations of the allowance adjusted
under SAB 108?
We focus on a particular accrual account, the allowance for
uncollectible accounts, to evaluate whether or not ?rms applied
SAB 108 to conservative accumulations on the balance sheet.
The allowance captures the accumulated effect of small increases
due to bad debt expense on the income statement being consis-
tently in excess of write-offs. Jackson and Liu (2010) suggest that
over-accruals of bad debt expense have accumulated in the allow-
ance account during the past 25 years. By the ?nal year of their
sample (2004), which predates SAB 108, Jackson and Liu (2010)
?nd that most ?rms could endure write-offs of uncollectible
accounts for several years without exhausting the allowance. Like
Jackson and Liu (2010), we de?ne a potential overestimate of the
allowance by considering the extent to which it exceeds leading
write-offs.
Managers of ?rms that have a large allowance may be reluctant
to fully adjust this account in the year that SAB 108 became effec-
tive for two main reasons. First, if managers immediately adjusted
the allowance then they would be unable to draw it down in sub-
sequent years when earnings thresholds may be in danger of not
being met. Regulators and academics contend that understated
net assets increase management’s ability to manage earnings
(American Institute of Certi?ed Public Accountants (AICPA),
1939; Devine, 1963; DeFond, 2002; Levitt, 1998; Penman, 2001),
and empirical research provides evidence that managers are more
likely to manage earnings when net assets are understated (Barton
& Simko, 2002; Jackson & Liu, 2010; McNichols & Wilson, 1988).
Second, managers face strong institutional forces that promote
conservative accounting and generally oppose adjusting asset
understatements (Givoly & Hayn, 2000; Watts, 2003a; Watts,
2003b). Conservatism has evolved over many decades and it has
become one of the most deeply engrained features of accounting
thought and practice (Basu, 1997; Watts, 2003a). Hirshleifer and
Teoh (2009) argue that managers favor conservative accounting
because it reduces the likelihood that they (and others) will be dis-
appointed when the realization of signi?cant estimates becomes
known in the future. Thus, preferences for conservative accounting
by managers may override preferences for regulatory compliance
(Holthausen, 2009)
3
or may alter the perception of whether or not
SAB 108 applies.
In addition, auditors may not require their clients to correct an
overestimated allowance. The risk-based audit approach directs
auditors’ attention to the overstatement of net assets, not the
understatement of net assets (Arens, Elder, & Beasley, 2003). Con-
sistent with this view, evidence suggests that auditors are more
concerned about reducing asset-increasing misstatements than
reducing asset-decreasing misstatements (Francis & Krishnan,
1999; Kinney & Martin, 1994; Nelson, Elliott, & Tarpley, 2002;
Wright & Wright, 1997). If auditors are comfortable with a conser-
vative balance in the allowance, then it is likely they will accept the
amount, given that management will have some plausible ratio-
nale for the balance.
This discussion not only suggests that managers may oppose
adjusting a potentially overestimated allowance, but it also sug-
gests that auditors may not require an adjustment. Hypothesis 1,
stated in alternative form, is as follows:
Hypothesis 1. Large (conservative) accumulations in the allow-
ance for uncollectible accounts will not be fully adjusted under the
application of SAB 108 in the year it became effective.
2.4. Does the allowance gradually decline after SAB 108’s effective
date?
While Hypothesis 1 predicts that these large accumulations in
the allowance are not fully adjusted in the year that SAB 108
became effective, we nonetheless expect that the allowance will
decline in the years subsequent to SAB 108’s effective date.
Jackson and Liu (2010) indicate that the allowance is built-up over
time by ?rms over-accruing bad debt expense. In turn, they ?nd
that the allowance is subsequently drawn down by ?rms recording
small amounts of bad debt expense when they are in danger of
missing analysts’ forecasts. However, SAB 108 may make it dif?cult
for managers to continue to build up the allowance because over-
Illustration of Different Methods of Quantifying Financial Statement Misstatements
Financial statement
Excessive Financial statement misstatement amount as a
expense misstatement amount percentage of earnings
Year accrual Rollover Iron curtain Income Rollover Iron curtain
1 10 10 10 1,000 1% 1%
2 10 10 20 1,000 1% 2%
3 10 10 30 1,000 1% 3%
4 10 10 40 1,000 1% 4%
5 10 10 50 1,000 1% 5%
This figure illustrates the effect on the income statement and balance sheet of recording an excessive expense accrual
in five consecutive years. A hypothetical company is assumed to record an expense of $10 in excess of the needed
amount for five consecutive years. Each year the excess expense accumulates on the balance sheet. The misstatement
on the balance sheet is $10 at the end of year 1, $20 at the end of year 2, and ultimately $50 at the end of year 5. While
the adjustment necessary to correctly state earnings is only $10 in any given year, the adjustment necessary to
correctly state the balance sheet accumulates over time, resulting in a quantitatively material misstatement by the end
of year 5.
Fig. 1. Illustration of different methods of quantifying ?nancial statement misstatements.
3
Empirical evidence suggests that compliance with regulations does not always
occur. For example, Schwartz and Soo (1996) show that noncompliance with
disclosures related to auditor terminations is related to various management
incentives. Likewise, Robinson, Xue, and Yu (2011) and Scannell and Lublin (2008)
?nd considerable non-compliance with executive compensation disclosures despite
the fact that veri?cation of compliance is almost costless. Moreover, articles from the
business press point out instances of possible noncompliance with accounting
standards affecting ?nancial statement accounts such as recent rules on asset
retirement obligations (see www.goodwinprocter.com/Publications/Newsletters/
Environmental-and-Energy-Advisory/2007/0424_FIN-47-Poses-Hard-Questions-for-
Estimating-Future-Environmental-Liabilities.aspx).
4 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
accruing bad debt expense would increase the likelihood that,
under the iron curtain approach, this account is seen to be materi-
ally misstated. Thus, we expect SAB 108 to create asymmetry in the
movement of large allowance accumulations—?rms can readily
draw down the allowance in the post-SAB 108 years but they
may encounter resistance if they wish to replenish the allowance
in those years. At the same time, if SAB 108 signi?cantly reduces
?rms’ ability to replenish an overestimated allowance account
then bad debt expense should be lower in the years after SAB
108 became effective than in the years before SAB 108 became
effective. Hypotheses 2a and 2b, stated in alternative form, are as
follows:
Hypothesis 2a. The allowance account declines in the years after
SAB 108 became effective.
Hypothesis 2b. Bad debt expense declines in the years after SAB
108 became effective.
2.5. Does the allowance continue to be highly conservative?
While we expect that the allowance account will be drawn
down in the post-SAB 108 years, it seems unlikely that ?rms
will quickly draw down accumulations in the allowance during
that period because that action could attract unwanted scrutiny
by auditors, regulators, and investors. We therefore expect the
allowance to remain quantitatively large (conservative) several
years after SAB 108 became effective even though Hypotheses
2a and 2b predict that it will decline during the post-SAB
108 period. Hypothesis 3, stated in alternative form, is as
follows:
Hypothesis 3. Firms continue to have large (conservative) accu-
mulations in the allowance account several years after SAB 108
became effective.
3. Archival analyses
In the ?rst stage of the study, we employ archival data to test
our hypotheses about the potential for conservative accumulations
in the allowance for uncollectible accounts to survive subsequent
to the effective date of SAB 108.
3.1. Sample selection
We identify the 750 domestic ?rms on Compustat in 2002 with
the largest unscaled balances in net current accounts receivable.
Our rationale for selecting such ?rms is that we wish to identify
?rms that are likely to have material allowance balances. We elim-
inate (1) ?rms in regulated industries (primarily ?nancial institu-
tions and utilities), (2) ?rms whose receivables are mainly due
from consumers (i.e., non-business receivables), and (3) leasing
companies.
4
This eliminates 129 ?rms. We then attempt to obtain
?nancial data for the remaining 621 ?rms from two sources—Sched-
ule II of Form 10-K and Compustat. Write-offs, bad debt expense, and
recoveries of previously written-off accounts must all be obtained
from Schedule II because Compustat does not collect these items.
5
In some cases we could not locate a Schedule II or an equivalent
schedule in the 10-K. In total, we eliminate 132 ?rms because Sched-
ule II could not be located or because of missing ?nancial data on
Compustat.
Our sample period starts in 2002, although we also obtain
needed data for 2001 from Schedule II along with Compustat data
for that year (there is no incremental sample attrition due to add-
ing 2001). We also require ?rms to have continuous Schedule II
data and Compustat data during the period 2001 through 2006.
We impose this requirement to avoid having ?rms appear in our
sample during pre-SAB 108 years but not actually appear in our
sample during the year SAB 108 became effective. One hundred
sixty-one ?rms do not have continuous Schedule II data and Com-
pustat data during this period, which results in 328 ?rms from
which we attempt to calculate the quantitative accuracy of the
allowance balance (discussed next). In turn, we collect needed data
through 2009, thereby leaving four potential reporting periods
subsequent to the effective date of SAB 108. Due to data availabil-
ity and sample attrition, the number of ?rms analyzed declines
from 328 ?rms in 2006 to 259 ?rms by 2009.
3.2. Benchmarks
We need a quantitative benchmark for estimating the accuracy
of the allowance based on observable write-offs. We use material-
ity which is typically de?ned by reference to net income. Evidence
indicates that a misstatement above ?ve percent of net income is
almost always considered quantitatively material (Chewning,
Pany, & Wheeler, 1989; Eilifsen & Messier, 2015; Holstrum &
Messier, 1982; Messier, Martinov-Bennie, & Eilifsen, 2005; Pany
& Wheeler, 1989). Therefore, our analysis employs ?ve percent of
net income for determining whether the allowance is potentially
overestimated.
We develop a benchmark for an ‘‘accurate” allowance by con-
sidering leading write-offs as a reference point for what the allow-
ance would have been to perfectly cover bad debts for the year
(Cecchini, Jackson, & Liu, 2012; Fedyk & Singer, 2011; Jackson &
Liu, 2010; McNichols & Wilson, 1988). This benchmark matches
the allowance with its future realization.
6
We note, however, that
leading write-offs is not a perfect benchmark because it combines
write-offs of prior year credit sales (this is the portion of write-offs
that we would ideally include) with write-offs of current year credit
sales (this is the portion of write-offs that we would ideally exclude).
Because our benchmark for the allowance includes some write-offs
that ought to be excluded, it is a benchmark that may bias our tests
against concluding that the allowance is potentially overestimated.
Thus, we de?ne a potentially overestimated allowance as one in
which the allowance exceeds leading write-offs by an amount in
excess of materiality. Using the materiality and allowance bench-
mark noted above (?ve percent), Appendix A lists the ?rms that
we de?ne to have a potentially overestimated allowance in the
year before SAB 108 became effective. Column (a) provides the dif-
4
With respect to item (1), we eliminate ?nancial institutions because their
receivables (primarily loans) are fundamentally different from the receivables of non-
?nancial institutions. Also, eliminating ?nancial institutions removes ?rms that
experienced extreme distress during the ?nancial crisis of 2008. With respect to item
(ii), consumer receivables are different from trade (business) receivables in terms of
transaction size, incentives for prompt payment, and bankruptcy frequency. Elimi-
nating consumer receivables therefore increases the homogeneity of our sample.
5
There are several non-routine items on Schedule II—recoveries, reclassi?cations,
reserve adjustments, currency translation adjustments, miscellaneous adjustments,
etc. We use recoveries as a catch-all label for all of the non-routine items on Schedule
II. In some instances, ?rms combine write-offs with recoveries (or other non-routine
items) and report a negative value for write-offs to the extent that recoveries (or other
non-routine items) are oppositely signed and exceed write-offs. Such negative values
are uncommon. Only two such instances arose in Appendix A (this appendix is
discussed in detail later). In these few instances, we set a lower bound of zero on
negative write-offs.
6
There is a one-year time period assumption inherent in the allowance. The
allowance is for current receivables, so it should have a write-off horizon of one year
or less. As a result, our use of one year of leading write-offs (rather than, for example,
two years of leading write-offs) is appropriate.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 5
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
ference between the allowance and leading write-offs expressed as
a percent of net income. Firms are ranked in descending order by
column (a). Formal tests are discussed below, but notice that 123
?rms (37.50% of the 328 ?rms analyzed) have a difference in Col-
umn (a) in excess of ?ve percent of net income in the year preced-
ing SAB 108’s effective date which suggests their allowance may be
overestimated.
We also consider various alternative quantitative measures in
Appendix A because it is possible that both managers and their
auditors use other potential metrics when determining the amount
of the allowance to report in the ?nancial statements. Column (b)
is a variant of Column (a), the only difference being that leading
write-offs are multiplied by a factor of two, consistent with an
alternative benchmark discussed in Riley and Pasewark (2009).
Column (c) provides the percentage of accounts receivable that
each ?rm has reserved, and Column (d) provides the industry
adjusted percentage of accounts receivable that each ?rm has
reserved. Column (e) provides the allowance as a percentage of
total assets. Following Jackson and Liu (2010), Column (f) provides
the number of years of future write-offs that each ?rm has
reserved.
7,8
A value of ?ve would, for example, indicate that the ?rm
could forego recording bad debt expense for ?ve years before write-
offs exhaust the allowance. Thus, higher ratios might suggest ?rms
are overestimating the allowance, whereas a value of 1–2 for a typ-
ical year may be more appropriate (Riley & Pasewark, 2009). Column
(g) provides the standard deviation of Column (f), which is calculated
on a ?rm-by-?rm basis using all available observations for each ?rm
during the sample period. The standard deviation, relative to the
mean value, helps to give readers a sense of the variable’s temporal
stability. Column (h) indicates the percentage of the allowance that
has not been consumed by leading write-offs. Lower ratios,
approaching zero, suggest ?rms’ current allowance provides a better
estimate of future write-offs. Column (i) is the ratio of bad debt
expense to write-offs. This ratio should be approximately one in
the long-run. Column (j) is the standard deviation of the ratio of
bad debt expense to write-offs.
Our discussion focuses on the mean and median values in the
last two rows of Appendix A because our interest lies in character-
izing the sample of ?rms that appear to have a potentially overes-
timated allowance, not to characterize particular ?rms in the
sample. The mean (median) value of Column (a) is 20.00 (11.10),
which suggests that the overstatement of the allowance is poten-
tially large in relation to reported earnings. The mean (median)
value of Column (b) is smaller than Column (a), but nonetheless
fairly large. Firms reserve a substantial fraction of their receivables
as indicated by the mean (median) value of Column (c) of 4.18
(3.45). These values are large relative to the industry medians, as
indicated by the positive mean (median) values in Column (d) of
1.10 (0.50). The mean (median) value of Column (e) is 0.96
(0.68), which reveals that the allowance is moderately signi?cant
in relation to total assets. However, we note that SAB 108 requires
registrants to consider both the income statement and balance
sheet when making materiality judgments.
The mean (median) value of Column (f) is 5.51 (3.93) which
indicates that the allowance will not be exhausted any time soon
even if no additional reserve is recorded for several years. The
mean value of Column (g) of 5.38 suggests that the ratio in Column
(f) may be somewhat unstable over time. This is consequential
because if the relationship between the allowance and leading
write-offs is erratic, it is possible that conservatism in the
allowance is merely intended to create a ‘‘cookie jar” reserve that
is increased in good years and drawn down in bad years to
manipulate earnings. The mean (median) value of Column (h) is
71.23 (74.55) which indicates that about three quarters of the
allowance is unused by leading write-offs. The mean (median)
value of Column (i) is 1.98 (1.00) which indicates that bad debt
expense and write-offs are somewhat similar to one another. The
mean (median) value of Column (j) of 2.08 (0.75) suggests that
the ratio in Column (i) is fairly stable over time. Overall, the data
in Appendix A suggests that these ?rms have potentially
overestimated allowances even after considering various alterna-
tive quantitative measures.
3.3. Descriptive information
Table 1 provides the industry pro?le for the 123 ?rms that com-
prise the ?ve percent of net income sample. While the sample con-
sists of a broad cross-section of different industries, there is
clustering in SIC codes 28, 35, and 73. Table 2 provides descriptive
statistics for the 1079 ?rm-years (123 individual ?rms). Panel A
provides descriptive statistics for ?rm size and other variables fol-
lowed by descriptive statistics for the dependent and independent
variables. Panel B provides descriptive statistics for the dependent
variables partitioned by SAB 108’s effective date.
9
3.4. Test of whether ?rms make SAB 108 adjustments
Hypothesis 1 predicts that ?rms will tend not to make SAB 108
adjustments in the year it became effective. For each of the 123
?rms identi?ed above as having a potentially overestimated allow-
ance (see Appendix A), we examined the Form 10-K to determine
whether a SAB 108 adjustment related to the allowance was made
in the year that SAB 108 became effective. None of the 123 ?rms
made any such adjustment.
10
The absence of any SAB 108 adjust-
ments to the allowance is of interest because the mean calculated
potential overestimation of the allowance was approximately 20%
of net income and the mean number of years of leading write-offs
reserved was approximately six.
Further, many of the ?rms that we analyze have a potentially
overestimated allowance that is far in excess of various commonly
used benchmarks. In fact, the 25 largest overestimates are, on aver-
age, 58% of net income and the allowance associated with these
overestimates covers approximately seven years of leading write-
offs. Given the evidence in Appendix A, it appears that, at ?rst
blush, a large fraction of the ?rms we analyze needed to make
SAB 108 adjustments related to the allowance. However, none of
these ?rms made such adjustments, but, instead, continue to main-
tain conservative allowances in the year SAB 108 became effec-
tive.
11
Hypothesis 1 is therefore supported.
3.5. Tests of whether the allowance declines after the effective date of
SAB 108
Hypothesis 2a predicts that the allowance declines in the years
after SAB 108 became effective. To test this hypothesis, we
7
When leading write-offs are zero, Column (f) is unde?ned. The issue that we face
is what value to assign to Column (f) when leading write-offs are zero. Given that the
mean and median values in Column (f) of Appendix A are approximately 6 and 4,
respectively, we selected a value of 5.
8
Write-offs exhibit no signi?cant time-trend during our sample period.
9
To mitigate the effect of outliers, continuous regression variables are winsorized
at the 99th percentile.
10
We do, however, ?nd that 22 ?rms made a SAB 108 adjustment unrelated to the
allowance. Also, while none of the ?rms in our ?nal sample made a SAB 108
adjustment related to the allowance, three ?rms in our initial sample made such
adjustments. As a percentage of net income, the SAB 108 adjustments related to the
allowance were 20.66%, 5.02%, and 1.02%.
11
SAB 99 (SEC, 1999) and Auditing Standard No. 14 (AICPA, 2010) prohibit auditors
from offsetting one material misstatement against another material misstatement. As
a result, the absence of a SAB 108 adjustment related to the allowance is not the result
of concurrent offsetting misstatements.
6 T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
estimate the following pooled, cross-sectional regression using the
model from Jackson and Liu (2010):
ALLOW
it
=SALE
it
¼ b
0
þb
1
AR
it
=SALE
it
þb
2
WO
it
=SALE
it
þb
3
WO
itþ1
=SALE
it
þb
4
ARTO IND
it
þb
5
ALT IND
it
þb
6
SD SALE IND
it
þb
7
BANKRUPT
it
þb
8
POST SAB 108
it
þe
it
ð1Þ
where ALLOW
it
is the allowance for uncollectible accounts; SALE
it
is
net sales; AR
it
is gross accounts receivable; WO
it
is write-offs of
uncollectible accounts; ARTO_IND
it
is the median industry accounts
receivable turnover ratio (de?ned as SALE
it
divided by the average
accounts receivable) computed using all ?rms with available data
on Compustat;
12
ALT_IND
it
is the median industry Altman (1968)
Z-score computed using all ?rms with available data on Compustat;
SD_SALE_IND
it
is the median industry standard deviation of sales
computed on a ?rm-by-?rm basis using the four quarters of each
year for all ?rms with available data on Compustat; BANKRUPT
it
is
the total number of business bankruptcies in the U.S. economy
scaled by the total number of corporate tax ?lers;
13
POST_SAB_108
it
is a dichotomous variable coded as 1 for the ?scal years after SAB
108 became effective (November 15, 2006), and 0 otherwise; i and
t are ?rm and year subscripts, respectively. Firm ?xed effects are
included in Eq. (1) but are not shown, and t-statistics are calculated
using standard errors clustered by ?rm. Hypothesis 2a predicts that
the coef?cient on POST_SAB_108
it
will be negative, indicating that
the allowance is lower in the post-SAB 108 years.
14
To enhance comparability over time and to control for potential
survivorship bias, we analyze two samples. The ?rst sample is
referred to as the ‘‘full sample.” This sample contains all observa-
tions during the sample period 2001 through 2009 for ?rms that
have an allowance exceeding our quantitative benchmark in the
year that SAB 108 became effective. The second sample is referred
to as the ‘‘constant sample.” This sample is similar to the full sam-
ple except that it only contains observations for ?rms that have
complete data for the entire nine-year sample period.
Table 3 reports regression results for Eq. (1) using the full and
constant samples. The coef?cient on each of the economic determi-
nants is positive and signi?cant (p-values 6 0.020) using both the
full and constant samples. The coef?cients on three of the controls
for temporal changes in the riskiness of receivables (ARTO_IND
it
,
ALT_IND
it
, and SD_SALE_IND
it
) are insigni?cant (p-values > 0.10),
while the coef?cient on one of the controls (BANKRUPT
it
) is signif-
icant in the predicted direction (p-values 6 0.002).
15
Table 1
Industry pro?le of sample.
SIC code SIC description # %
13 Oil and Gas Extraction 4 3.25
20 Food and Kindred Products Manufacturers 4 3.25
25 Furniture and Fixture Manufacturers 6 4.88
27 Printing and Publishing 5 4.07
28 Chemicals and Allied Products Manufacturers 12 9.76
30 Rubber and Misc. Plastics Manufacturers 3 2.44
35 Industrial and Commercial Machinery Manufacturers 11 8.94
36 Electronic and Electrical Equipment Manufacturers 3 2.44
37 Transportation Equipment Manufacturers 6 4.88
38 Measuring and Analyzing Equipment Manufacturers 7 5.69
50 Wholesale, Durable Goods 7 5.69
51 Wholesale, Non-Durable Goods 7 5.69
73 Business Services 15 12.20
Other 33 26.83
Total 123 100.00
The ‘‘Other” category captures count data for SIC codes that have fewer than three
different ?rms.
Table 2
Descriptive statistics.
Panel A: Descriptive statistics
(n = 1079)
Mean Median Std. dev.
Firm size and other variables
TA
it
($mil.) 6225.23 1859.08 12620.11
SALE
it
($mil.) 7279.60 2347.00 14121.54
MVE
it
($mil.) 5422.32 1641.22 15434.56
AR
it
/TA
it
(%) 24.32 20.72 13.14
AR
it
/CA
it
(%) 50.71 48.24 16.29
ALLOW
it
/AR
it
(%) 4.46 3.52 3.41
BDE
it
/NI
it
(%) 4.70 3.24 32.84
WO
it
/NI
it
(%) 4.38 3.31 46.91
BDE
it
/SHO
it
($ per share) 0.15 0.08 0.19
Dependent variables from regressions
ALLOW
it
/SALE
it
(Â100) 0.81 0.59 0.76
BDE
it
/SALE
it
(Â100) 0.32 0.19 0.51
Independent variables from regressions
AR
it
/SALE
it
(Â100) 18.43 16.90 8.38
WO
it
/SALE
it
(Â100) 0.35 0.18 0.57
WO
it+1
/SALE
it
(Â100) 0.34 0.18 0.55
ARTO_IND
it
6.64 6.06 1.98
ALT_IND
it
4.27 4.27 1.45
SD_SALE_IND
it
7.54 5.22 8.09
BANKRUPT
t
(Â100) 0.66 0.69 0.18
ALLOW
itÀ1
/SALE
it
(Â100) 0.79 0.58 0.78
TOP_DEC
it
(%) 10.00 0.00 30.15
BOT_DEC
it
(%) 10.00 0.00 30.15
RECOV
it
/SALE
it
(Â100) 0.02 0.00 0.11
Panel B: Descriptive statistics for dependent variables partitioned by the effective
date of SAB 108
?
Pre-SAB 108 years (n = 643)
Mean Median Std. dev.
ALLOW
it
/SALE
it
(Â100) 0.88 0.66 0.75
BDE
it
/SALE
it
(Â100) 0.36 0.21 0.51
Post-SAB 108 years (n = 436)
Mean Median Std. dev.
ALLOW
it
/SALE
it
(Â100) 0.70 0.49 0.75
BDE
it
/SALE
it
(Â100) 0.27 0.14 0.50
Variables are de?ned as follows: TA
it
is total assets; SALE
it
is net sales; MVE
it
is
market value of equity; AR
it
is gross accounts receivable; CA
it
is current assets;
ALLOW
it
is the allowance for uncollectible accounts; BDE
it
is bad debt expense; NI
it
is net income; SHO
it
is common shares outstanding; WO
it
is write-offs of uncol-
lectible accounts; ARTO_IND
it
is the median industry accounts receivable turnover
ratio (de?ned as SALE
it
divided by the average accounts receivable) computed using
all ?rms with available data on Compustat; ALT_IND
it
is the median industry
Altman (1968) Z-score computed using all ?rms with available data on Compustat;
SD_SALE_IND
it
is the median industry standard deviation of sales computed on a
?rm-by-?rm basis using the four quarters of each year for all ?rms with available
data on Compustat; BANKRUPT
it
is the total number of business bankruptcies in the
U.S. economy scaled by the total number of corporate tax ?lers; TOP_DEC
it
is a
dichotomous variable coded as 1 for ?rms in the sample top decile of net income
divided by total assets, and 0 otherwise; BOT_DEC
it
is a dichotomous variable coded
as 1 for ?rms in the sample bottom decile of net income divided by total assets, and
0 otherwise; RECOV
it
is recoveries of previously written-off accounts receivable; i
and t are ?rm and year subscripts, respectively. Continuous regression variables are
winsorized at the 99th percentile.
*
Means and medians are signi?cantly different between the pre- and post-SAB
108 years (p-values < 0.01).
12
Industry is de?ned at the two-digit SIC code level. When there are fewer than ten
total ?rms in a two-digit SIC code, we compute ALT_IND
it
, ARTO_IND
it
, and
SD_SALE_IND
it
, at the one-digit SIC code level.
13
We obtained the number of bankruptcies from the American Bankruptcy Institute
at www.abiworld.org under the heading ‘‘U.S. Bankruptcy Filings 1980–2009
(Business, non-Business, Total).”
14
Studies often include both ?rm and year ?xed effects in pooled, cross-sectional
regressions. Because the test variable in the regressions reported in this study have a
time dimension, year ?xed effects are excluded.
15
When we specify the industry adjusted controls for temporal changes in the
riskiness of receivables in their non-industry adjusted form, the variables remain
insigni?cant in all regressions and none of our inferences or conclusions change.
T.G. Canace et al. / Accounting, Organizations and Society xxx (2015) xxx–xxx 7
Please cite this article in press as: Canace, T. G., et al. Conservatism and Staff Accounting Bulletin No. 108. Accounting, Organizations and Society (2015),http://dx.doi.org/10.1016/j.aos.2015.08.002
In Table 3, the coef?cient on the test variable, POST_SAB_108
it
,
is negative in both the full sample (coef?cient = À0.0006,
t-statistic = À1.68, p-value = 0.094) and constant sample
(coef?cient = À0.0005, t-statistic = À1.42, p-value = 0.156). The
coef?cients on POST_SAB_108
it
are negative but, at best, marginally
signi?cant, indicating limited support for Hypothesis 2a. Moreover,
the economic signi?cance of the decline is modest. Notice that the
coef?cient on POST_SAB_108
it
for the full sample is only À0.0006,
which implies that the mean allowance has declined by approxi-
mately 0.06% of sales. Thus, although ?rms may not have felt the
need to make a SAB 108 adjustment immediately after its enact-
ment, SAB 108 does appear to have caused the allowance to decline
modestly, perhaps because managers of ?rms now ?nd it more
dif?cult to replenish an already large allowance balance by over-
accruing bad debt expense. We consider this possibility next.
3.6. Tests of whether bad debt expense declines after the effective date
of SAB 108
Hypothesis 2b predicts that bad debt expense declines in the
years after SAB 108 became effective. To test this hypothesis, we
estimate the following pooled, cross-sectional regression using
the model from Jackson and Liu (2010) and McNichols and
Wilson (1988):
BDE
it
=SALE
it
¼ b
0
þb
1
ALLOW
itÀ1
=SALE
it
þb
2
WO
it
=SALE
it
þb
3
WO
itþ1
=SALE
it
þb
4
TOP DEC
it
þb
5
BOT DEC
it
þb
6
RECOV
it
=SALE
it
þb
7
ARTO IND
it
þb
8
ALT IND
it
þb
9
SD SALE IND
it
þb
10
BANKRUPT
it
þb
11
POST SAB 108
it
þe
it
ð2Þ
where SALE
it
, ALLOW
it
, WO
it
, ARTO_IND
it
, ALT_IND
it
,
SD_SALE_IND
it
, BANKRUPT
it
, and POST_SAB_108
it
are de?ned
above; BDE
it
is bad debt expense; TOP_DEC
it
is a dichotomous vari-
able coded as 1 for ?rms in the sample top decile of net income
divided by total assets, and 0 otherwise; BOT_DEC
it
is a dichoto-
mous variable coded as 1 for ?rms in the sample bottom decile of
net income divided by total assets, and 0 otherwise; RECOV
it
is
recoveries of previously written-off accounts receivable (as noted
in Section 3, this is a catch-all label for non-routine items on Sched-
ule II). Firm ?xed effects are included in Eq. (2) but are not shown,
and t-statistics are calculated using standard errors clustered by
?rm. The ?rst six variables in Eq. (2) are economic determinants
from McNichols and Wilson (1988), while the next four variables
are controls for temporal changes in the riskiness of receivables
from Jackson and Liu (2010). Hypothesis 2b predicts that the coef-
?cient on POST_SAB_108
it
will be negative.
Table 4 reports regression results for Eq. (2) using the full and
constant samples. The coef?cient on each of the economic determi-
nants is signi?cant (p-values < 0.10) in the predicted direction
using the full and constant samples, except for the coef?cient on
TOP_DEC
it
which is insigni?cant in both samples (p-values > 0.10)
and the coef?cients on ALLOW
it-1
/SALE
it
, WO
it
/SALE
it
, and
BOT_DEC
it
which are insigni?cant (p-values > 0.10) in the constant
sample. The coef?cients on controls for temporal changes in the
riskiness of receivables (ARTO_IND
it
, ALT_IND
it
, and
SD_SALE_IND
it
) are insigni?cant (p-values > 0.10), while the coef?-
cient on one of the controls (BANKRUPT
it
) is signi?cant in the pre-
dicted direction (p-values < 0.01).
In Table 4, the coef?cient on the test variable, POST_SAB_108
it
, is
negative but, at best, marginally signi?cant in both the full sample
(coef?cient = À0.0005, t-statistic = À1.43, p-value = 0.154) and con-
stant sample (coef?cient = À0.0005, t-statistic = À1.40, p-value =
0.162). These results provide very limited support for Hypothesis
2b. Thus, not only do ?rms not make SAB 108 adjustments, but they
do not reduce their annual bad debt expense to gradually reduce the
allowance balance.
16
Table 3
Regressions of the allowance for uncollectible accounts on economic determinants, controls for temporal changes in the riskiness of receivables, and the test variable. ALLOW
it
/
SALE
it
= b
0
+ b
1
AR
it
/SALE
it
+ b
2
WO
it
/SALE
it
+ b
3
WO
it+1
/SALE
it
+ b
4
ARTO_IND
it
+ b
5
ALT_IND
it
+ b
6
SD_SALE_IND
it
+ b
7
BANKRUPT
it
+ b
8
POST_SAB_108
it
+ e
it
.
Variable Pred. Full sample (n = 1079) Constant sample (n = 873)
Sign Coeff. t-stat. p-value Coeff. t-stat. p-value
Economic determinants
Intercept ? À0.0040 À1.87 0.062 À0.0048 À1.85 0.064
AR
it
/SALE
it
+ 0.0341 9.21