Research in Accounting Regulation 24 (2012) 96–104

Contents lists available at SciVerse ScienceDirect

Research in Accounting Regulation

journal homepage: www.elsevier.com/locate/racreg

Research Report

The relation between aggressive financial reporting and aggressive

tax reporting: Evidence from ex-Arthur Andersen clients
Wendy Heltzer, Mary P. Mindak ⇑, Sandra W. Shelton
DePaul University, Chicago, IL, USA

a r t i c l e i n f o a b s t r a c t

Article history: We investigate the economic trade-offs managers face due to conflicting incentives to
Available online 15 June 2012 report high financial statement book income and, at the same time, report low taxable
income. Our setting involves Houston clients of Arthur Andersen (AA), who have been
Keywords: shown to exhibit a culture of aggressive financial reporting. Using our sample of AA Hous-
Auditing ton clients, we test two competing theories: (1) firms which have a culture of aggressive
Tax financial reporting are also aggressive in their tax reporting, versus (2) firms which are
Aggressive reporting
willing to pay real dollars (taxes) to report higher financial statement earnings. We do
not find support for either theory. Instead, our findings suggest a middle-ground: firms
may exhibit a culture of aggressive financial reporting without impacting their relative
tax reporting. Our findings not only shed light on the intersection of financial and tax
reporting, but they also add to the extant literature involving the culture of AA. To the best
of our knowledge, this is the first paper to investigate the tax ramifications of AA’s culture
of aggressive financial reporting.
Ó 2012 Elsevier Ltd. All rights reserved.

Introduction financial reporting on taxes is examined within the context

Financial reporting and tax reporting, in general, are
subject to contradictory pressures. Managers face pres-
sures from shareholders and analysts to report high finan-
cial income. At the same time, managers have a fiduciary
duty to act efficiently in their spending; one way cash
may be conserved is by reporting low taxable income.
Thus, the intersection of financial and tax reporting allows
academics to ponder whether managers will pay (in tax
dollars) for higher financial earnings, or, if aggressive
reporting is a pervasive trait across both sets of books.
Many studies involving the intersection of financial and
tax reporting are conducted in settings of known aggres-
sive financial reporting. The impact of such aggressive
⇑ Corresponding author. Address: School of Accountancy and MIS,
DePaul University, 1 E. Jackson Boulevard, Suite 6000, Chicago, IL 60604,
USA. Tel.: +1 312 362 5601; fax: +1 312 362 6208.
E-mail addresses: wheltzer@depaul.edu (W. Heltzer), mmindak1@
depaul.edu (M.P. Mindak), sshelton@depaul.edu (S.W. Shelton).
of such settings. Findings from these studies yield mixed
results. Erickson, Hanlon, and Maydew (2004) study the
tax implications of firms which committed financial
accounting fraud. By comparing original and restated
financial statements, the authors find that the mean (med-
ian) income taxes paid on each dollar of overstated earn-
ings was approximately 11 (eight) cents. These findings
suggest that firms are willing to pay for overstated income.
On the contrary, Frank, Lynch, and Rego (2009), in a larger
sample of 49,886 firm-year observations, find that firms
which are aggressive in their reporting of financial income
are simultaneously aggressive in their reporting of taxable
income. Thus, Frank et al.’s findings suggest that aggressive
reporting behavior may be a pervasive trait across both
sets of books.
Similar to prior studies of the intersection of aggressive
financial and tax reporting, we employ a setting where the
auditors have been known to allow aggressive financial
reporting, and examine relative differences in tax reporting
within the context of such setting. Specifically, we examine

1052-0457/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.racreg.2012.05.001
 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104
the tax reporting of Houston clients of Arthur Andersen
(AA), vis-à-vis a control sample of Houston clients of other
Big 5/61 audit firms, during the five-year period 1996–2000.
As discussed in more detail herein, former AA employees
and recent literature on the firm have discussed how the
firm’s culture changed over the years allowing clients to
make aggressive financial reporting decisions. In addition,
some partners in the AA Houston office went against man-
dates of the national AA office (Schmidt, 2002), exhibiting
their willingness to tolerate aggressive reporting practices.
Furthermore, a recent study finds that AA’s Houston clients
exhibited relatively less timely loss recognition (Krishnan,
2005) relative to Houston clients of other Big 5/6 auditors
during a similar time period.
In our analyses, we find that Houston clients of AA were
neither more nor less aggressive in their tax reporting, rel-
ative to a control sample of Houston clients of other Big 5/6
firms. Specifically, we find that the level of tax aggressive-
ness of Houston AA clients was similar to that of geograph-
ically close firms. As such, our findings suggest that firms
are making financial reporting decisions that do not cost
the (relative) tax dollars.
Our study contributes to three branches of literature.
Primarily, as mentioned above, we contribute to the litera-
ture which examines the intersection of financial and tax
reporting, and the economic tradeoffs which are faced in
these crossroads. We additionally contribute to the litera-
ture which studies the unique culture of AA, as well as
the quality of AA’s clients’ earning reports (see, for exam-
ple, Chaney & Philipich, 2002; Krishnan, 2005; Cahan &
Zhang, 2006). Finally, we extend the literature which
investigates the impact of an audit firm on tax reporting
(see, for example, Maydew & Shackelford, 2005; Omer,
Bedard, & Falsetta, 2006; Lassila, Omer, Shelley, & Smith,
2010; Cook & Omer, 2010).
Background and literature review
The rise and demise of Arthur Andersen
Arthur Andersen was founded by Arthur E. Andersen
and Clarence DeLaney, and was originally known as Ander-
sen, Delaney & Company. In 1918, Arthur E. Andersen
gained sole ownership of the company due to DeLaney’s
resignation. The founding principles of AA rested upon
the integrity of the firm and the responsibility of the firm
to serve investors. The original values of the firm entailed
the following three main principles: (1) integrity and hon-
esty; (2) one firm, one voice partnership model; and (3)
training to a shared method (Squires, Smith, McDougall,
& Yeack, 2003).
Over the years, until his death in 1947, Arthur E. Ander-
sen expanded the company nationally, establishing new
offices in every major U.S. city. During the 1970’s, the firms
consulting services practice grew and created a new chal-
lenge for AA, as it struggled to balance the high revenue
1
Price Waterhouse merged with Coopers & Lybrand on July 1, 1998,
causing the ‘‘Big 6’’ audit firms to become the ‘‘Big 5’’. Our sample period is
1996 through 2000, and thus spans both the Big 6 and Big 5 periods. As
such, we refer to the Big 5/6 throughout.
97
generating consulting business with the ‘‘bread-and-but-
ter’’ accounting practices. The firm adopted a more sales-
focused culture leading to a change in the firm’s values,
increasing levels of acceptable risk with clients, and raised
the potential for conflicts of interest (Squires et al., 2003).
There also was an internal competitiveness in the firm that
focused on generating revenues. For example, if fees were
opposed by the client, ‘‘the most likely scenario was a pro-
ject completed in as short a period of time as possible, with
as any inexperienced lower-paid people doing the work as
possible’’ (Toffler & Reingold, 2003).
In sum, there were many factors at play which caused
AA to shift from its original values to a more aggressive
mentality. It is this later, more aggressive mentality, which
became more prominent in the late 1990’s, which we study
herein. Specifically, we study what impact, if any, this
aggressiveness had on tax reports.
Arthur Andersen, Houston and aggressive financial reporting
We conduct our research by examining the levels of
aggressive tax reporting of Houston clients of AA, vis-à-
vis Houston clients of other Big 5/6 auditors. We have cho-
sen the Houston AA office as our setting for this study for a
number of reasons. First of all, empirical research suggests
that AA’s Houston based clients were more aggressive in
financial reporting relative to clients of other auditors
and other geographic locations. Krishnan (2005) finds that
clients of AA’s Houston office delayed recognition of pub-
licly available bad news, relative to a control sample of
Houston-based clients audited by other Big 5/6 auditors.
In a related, follow-up study, Krishnan (2007) finds that
financial statement conservatism increases for former
Houston-based AA clients, following the demise of AA.
As noted by Krishnan (2005, 2007), there are additional
reasons to focus on Houston-area AA clients when trying to
isolate aggressive financial reporting, beyond the empirical
findings above. First, some Houston AA partners chose to
ignore advice and guidance from AA headquarters, and dis-
played, at times, aggressive interpretation of accounting
rules (Schmidt, 2002). Second, Enron and Waste Manage-
ment were clients of the Houston AA office. Both firms
were charged with massive accounting fraud by the Secu-
rities and Exchange Commission (SEC). Enron’s restated
financials revealed that it overstated income during the
years 1997–2000 by approximately $600 million, while
Waste Management’s restated financials reveal that it
overstated income during the years 1992–1996 by approx-
imately $1 billion. Third, Chaney and Philipich (2002)
examine stock price changes of AA’s clients surrounding
AA’s admission of document destruction. They find that
the largest negative reaction occurred with the Houston of-
fice clients. Finally, Francis, Stokes, and Anderson (1999)
explore the usefulness of conducting audit research using
city-level markets, as opposed to using aggregate national
data. The authors find that, within the Big 5/6 auditor
group, national data obscures important city-level
variations.
In sum, our test setting is similar to that of other studies
regarding the financial statement implications of AA’s
unique culture. Firm chronicles and empirical research
98 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104

findings lead us to believe that Houston AA clients were aggressive financial reporting does not necessarily cost a
unique in their aggressive financial reporting. As such, by firm relative tax dollars, nor does it signal a pervasive
examining the aggressive tax reporting levels of these aggressiveness across both sets of books. Thus, we add to
firms, relative to geographically similar firms, our setting the literature which examines the intersection of financial
provides a natural experiment to examine the intersection and tax reporting by suggesting a middle-ground between
of aggressive financial reporting and aggressive tax the findings in Erickson et al. and those in Frank et al.: in-
reporting.2 stead of paying for overstated earnings (Erickson et al.) or
being aggressive across both sets of books (Frank et al.)
The Intersection of aggressive financial and tax reporting firms are able to be aggressive in their financial reports
in a manner which does not impact their relative tax
Due to the economic trade-offs between reporting high- aggressiveness.
er book income and lower taxable income, there is a grow-
ing sector of empirical literature which explores the
Research question and methodology
intersection of book income and taxable income. Erickson
et al. (2004) investigate a sample of firms which meet
Our primary research question involves one economic
two criteria: (1) the firm and/or management were ac-
impact of aggressive financial reporting: taxes. Past re-
cused of fraud by the SEC, and (2) the firm restated its
search has found evidence of taxes being paid on aggres-
earnings with the SEC. The authors find that for one dollar
sively reported financial income (Erickson et al., 2004) as
of overstated pretax earnings, the mean income taxes paid
well as a pervasive aggressiveness throughout both book
were approximately 11 cents, while the median was
and tax reporting (Frank et al., 2009). We call the first sit-
approximately 8 cents. Therefore, management paid for
uation ‘‘Willingness to Pay’’ and the later ‘‘Pervasive Aggres-
illegally aggressive book reporting with real tax dollars.
siveness’’. We additionally put forward the possibility that
Badertscher, Phillips, Pincus, and Rego (2009) extend
aggressive financial reporting is conducted up to a rela-
Erickson et al. by evaluating firms that restate income
tively ‘‘neutral’’ point, allowing firms to be aggressive in
downwards, thereby implying that they originally man-
their financial reporting without engaging in relatively dif-
aged earnings upwards. By comparing original and re-
ferent levels of aggressive tax reporting. We call this situ-
stated data, the authors findings suggests that firms
ation a ‘‘Neutral’’ strategy. A ‘‘Neutral’’ strategy may occur
which are aggressive in their reporting of book income
because a firm does not have the ability to engage in non-
may also be aggressive in their reporting of taxable in-
conforming transactions, a firm does not want to signal
come. Frank et al. (2009) employ a larger sample to study
aggressive behavior with a large divergence between book
the intersection of aggressive financial reporting and
income and taxable income, and/or a firm is willing to be
aggressive tax reporting. Frank et al. provide empirical evi-
aggressive in its financial reporting provided it does not
dence of a strong, positive relation between aggressive
cost it relative tax dollars.
book reporting and aggressive tax reporting. Their results
We conduct our analysis as an exploratory study; we
therefore suggest that there are insufficient costs in place
are agnostic as to the predictions regarding the above three
to offset the conflicting incentives between financial and
possible outcomes. Because of the conflicting theories out-
tax reporting.3
lined above we present the following unsigned research
Our paper contributes to the literature outlined above
question to advance our understanding of the economic
by exploring the questions: do firms with aggressive finan-
trade-offs between financial and tax reporting:
cial reporting characteristics additionally exhibit traits of
RQ1: did the Houston clients of AA engage in different
aggressive tax reporting (per Frank et al.)? Or, on the con-
levels of aggressive tax reporting, vis-à-vis Houston clients
trary, are firms with aggressive reporting behavior willing
of other Big 5/6 auditors?
to pay (in tax dollars) for their inflated financial statement
Similar to contemporaneous studies (Frank et al., 2009
earnings (per Erickson et al.)?
and Omer et al., 2006), we use three different measures
In our analyses, we find that the Houston clients of AA
of aggressive tax reporting: discretionary tax (DTAX); the
engaged in similar aggressive tax reporting, relative to a
book effective tax rate (B_ETR); and the cash effective tax
control sample of Houston clients of other Big 5/6 auditors.
rate (C_ETR). Our measure of aggressive tax reporting,
In other words, our findings suggest that a culture of
DTAX, was introduced by Frank et al. (2009). DTAX is the
unexplained portion of permanent book-tax differences.
2
It should be noted that not all empirical findings suggest that Houston More specifically, DTAX is determined by estimating the
clients were more aggressive in their financial reports, relative to other
following equation; residuals are deemed to be
geographic areas. Specifically, Cahan and Zhang (2006) study abnormal
accruals of AA clients in 2001 and 2002. They find an increase in income-
decreasing accruals among ex-AA clients in 2002, relative to 2001. These DTAX : PERMDIFF it
findings suggest that ex-AA clients were viewed as a litigation risk, possibly
¼ a0 þ a1 INTANGit þ a2 UNCON it þ a3 MIit
due to aggressive past reporting. The authors’ findings are not more
extreme in the case of Houston area clients. However, as the authors note, þ a4 CSTEit þ a5 DNOLit þ a6 LAGPERM it þ eit ð1Þ
their study is limited by the accuracy of identifying characteristics of
litigation risk; the low R2 values in their study hint at the possibly of where:4
correlated omitted variables.
3
Such costs include the ability to engage in nonconforming transactions,
4
and/or the willingness to increase the gap between book income and See Table 2 for detailed variable definitions, including COMPUSTAT
taxable income, which may increase audit risk. data items.
 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104
PERMDIFF = Total book-tax differences, less temporary
book-tax differences.
INTANG = Goodwill and other intangibles.
UNCON = Income (loss) reported under the equity
method.
MI = Income (loss) attributable to minority interest.
DCSTE = Current state income tax expense.
DNOL = Change in net operating loss carryforward LAG-
PERM = One-year lagged PERMDIFF e = Discretionary
permanent difference (DTAX).
For the sake of completeness, we employ two additional
measures of aggressive tax reporting in our analyses. Our
second and third measures of aggressive tax reporting
are the book effective tax rate (B_ETR) and the cash effec-
tive tax rate (C_ETR). B_ETR is equal to total income taxes
paid divided by pre-tax income less special items. C_ETR
is similar to B_ETR except that the numerator is income
taxes paid rather than total income taxes.5 While the
B_ETR has long been used in empirical literature, the C_ETR
was introduced by Dyreng, Hanlon, and Maydew (2008).6
We outline our sample selection in detail in the next section.
Once our aggressive tax proxies are determined, we con-
duct both univariate and multivariate analyses in Results
section. First, we conduct a univariate analysis to compare
the aggressive tax reporting of Houston clients of AA, rela-
tive to our sample of Houston clients of other Big 5/6 audit
firms. In these univariate analyses, Willingness to Pay is sug-
gested by Houston clients of AA exhibiting smaller (greater)
DTAX (B_ETR/C_ETR), relative to Houston clients of other Big
5/6 firms, while Pervasive Aggressiveness is suggested by
Houston clients of AA exhibiting greater (smaller) DTAX
(B_ETR/C_ETR), relative to Houston clients of other Big 5/6
firms. A Neutral strategy is suggested by Houston clients
of AA exhibiting similar levels of aggressive tax reporting,
relative to Houston clients of other Big 5/6 firms.
We additionally conduct multivariate analyses, control-
ling for known factors which impact aggressive tax report-
ing. Specifically, we conduct the following regressions:
DTAX it ¼ a0 þ a1 AAit þ a2 SIZEit þ a3 FORit þ a4 LEV1it
þ a5 MTBit þ þa6 PTROAit þ a7 DPTCFOit
þ a8 NOLit þ a9 AF it þ a10 NUMEST it
þ a11 EM1it þ a12 EM2it þ a13 EM3it
þ a1418 YRit þ e; ð2Þ
ETRit ¼ a0 þ a1 AAit þ a2 SIZEit þ a3 FORit þ a4 LEV2it
þ a5 MVEit þ a6 BM it þ a7 EP it þ a8 ROAit
þ a9 INT it þ a10 R&Dit þ a11 ADV it þ a12 CAP it
þ a13 INV it þ a1418 YRit þ e ð3Þ
where: 7 Houston office. We manually checked each of the signing
5
See Table 3 for detailed variable definitions, including COMPUSTAT
data items.
6 8
Dyreng et al. (2008) introduce a one-year, five-year and ten-year cash
effective tax rate. Due to the limited time period in our sample, we employ
a one-year rate, similar to Cook and Omer (2010).
7
See Table 3 for detailed variable definitions, including COMPUSTAT
data items.
99
Eq. (2) Variables DTAX = Residual from Eq. (1) AA = Dum-
my variable set to 1 if the auditor is AA; 0 otherwise
SIZE = Natural logarithm of total assets FOR = Dummy vari-
able set to 1 if the absolute value of foreign pretax income
is positive; 0 otherwise LEV1 = Leverage measure #1
MTB = Market-to-book ratio PTROA = Pretax return on assets
DPTCFO = Change in pretax cash flows from operations
NOL = Dummy variable set to 1 if the firm has a net operating
loss; 0 otherwise AF = Dummy variable set to 1 if the number
of analysts following the firm is positive; 0 otherwise
NUMEST = Number of analysts following the firm
EM1 = Earnings management #1: loss avoidance EM2 = Earn-
ings management #2: earnings decline avoidance
EM3 = Earnings management #3: meet/beat analysts fore-
cast YR = Dummy variable to control for year fixed-effects.
Eq. (3) Variables ETR = B_ETR (book effective tax rate)
and, separately, C_ETR (cash effective tax rate) AA = Dum-
my variable set to 1 if the auditor is AA; 0 otherwise
SIZE = Natural logarithm of total assets FOR = Dummy var-
iable set to 1 if the absolute value of foreign pretax income
is positive; 0 otherwise LEV2 = Leverage measure #2
MVE = Market value of equity BM = Book-to-market ratio
EP = Earnings-to-price ratio ROA = Return on assets
INT = Intangible asset intensity R&D = Research and devel-
opment intensity ADV = Advertising intensity CAP = Capital
intensity INV = Inventory intensity YR = Dummy variable to
control for year fixed-effects.
The variable of interest in both the above regressions is
AA, a dummy variable equal to one if the firm retained AA as
its auditor for the given year; zero otherwise. Per the Will-
ingness to Pay theory, AA will be negative in Eq. (2), and po-
sitive in Eq. (3). The Pervasive Aggressiveness theory will be
supported by a positive AA variable in Eq. (2), and a negative
AA variable in Eq. (3). Finally, the Neutral theory predicts AA
will be insignificant across all regression models.
Sample
Sample selection
As discussed above, our sample consists of Houston cli-
ents of AA, and our control consists of Houston clients of
non-AA Big 5/6 firms, during the period 1996–2000. We
have chosen to restrict our control sample to non-AA Big
5/6 audit firms, as Big 5/6 audit firms have been shown
to perform higher quality audits, vis-à-vis non-Big 5/6
firms (DeAngelo, 1981). We began our sample selection
process by gathering all firms on the COMPUSTAT tapes
in Houston, Texas, audited by a Big 5/6 auditor and with
positive total assets8 in the relevant time period. This initial
search yielded 1132 observations.
Next, we eliminated firms which were not audited by a
auditor’s office location by reviewing the company’s 10K
auditor Opinion for each firm-year observation.9 There
All variables used in our analyses are scaled by lagged assets,
necessitating positive total asset values.
9
When reviewing the auditor opinions, we found that 11 of the auditor
names identified by COMPUSTAT data149 were incorrect, so we updated
our data set with the correct auditor and office location.
100 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104

Table 1
Sample selection. This table reflects our sample selection process.

DTAX B_ETR C_ETR

Houston Area Big-Six Clients, 1996–2000, with positive assets (data6) and auditor data (data149) 1132 1132 1132
Non-Houston Auditors (104) (104) (104)
Elimination of observations in the finance, insurance and real estate industries (SIC 6000–6999) (46) (46) (46)
Elimination of observations in the electric gas and sanitary services industries (SIC 4900–4999) (120) (120) (120)
Subtotal 862 862 862
Incomplete data needed for dependent variables (286) (5) (135)
Incomplete data needed for independent variables (116) (175) (136)
Final sample 460 682 591

were 104 observations where the auditor’s office was not

Table 2 Houston (despite the company being located in Houston)
DTAX determination. DTAX is determined by calculating the error term from and these observations were eliminated. Due to unique
Eq. (1), by year.
incentives in regulated industries, we eliminate all financial
Intercept 0.55*** institutions (SIC codes 6000–6900) and utilities (SIC codes
(2.85) 4900–4999), eliminating 46 and 120 observations,
INTANG 0.03
respectively.
(0.29)
UNCON 1.30** The remaining 862 firm-year observations were further
(2.38) reduced due to data restrictions for each aggressive tax
MI 2.12** reporting variable. In relation to the DTAX variable, 286
(2.24)
observations were missing data necessary to estimate
CSTE 1.47*
(1.77)
Eq. (1),10 while an additional 116 observations were missing
DNOL 0.55** data necessary to estimate Eq. (2). In relation to the ETR vari-
(2.32) ables, we eliminated 5 (135) observations due to missing
LAGPERM 0.93⁄⁄ data for B_ETR (C_ETR), and an additional 175 (208) observa-
(2.52)
tions were eliminated due to missing data needed for Eq. (3).
Adj-R2 0.32 See Table 1 for a detailed outline of our sample selection
N 460
process.11
Variable definitions:
PERMDIFF = total book-tax differences  temporary book tax differences,
DTAX determination
={(pretax incomet(data170)  {[federal current income taxest
(data63) + foreign current income taxest (data64))]/0.35})[deferred taxest
(data50)/0.35]}/total assetst1 (data6). As discussed in Research question and methodology sec-
Intercept = intercept scaled, =1/total assetst1 (data6). tion above, our first measure of aggressive tax reporting,
INTANG(a) = goodwill and other intangibles, =intangiblest (data33)/total DTAX, is the residual from Eq. (1). Mean coefficient determi-
assetst1 (data6).
nants from estimating Eq. (1) may be found in Table 2. In or-
UNCON(b) = income (loss) reported under the equity method, =equity in
earningst (data55)/total assetst1 (data6). der to control for macro-economic impacts on taxes, we
MI(c) = income (loss) attributable to minority interest, =minority interestt estimated Eq. (1) by year. Our results are similar to the ori-
(data49)/total assetst1 (data6). ginal DTAX determination in Frank et al. (2009).
CSTE(d) = current state income tax expense, =state income taxest
(data173)/total assetst1 (data6)
DNOL = change in net operating losscarryforwards, = (unused net oper- Summary statistics
ating loss carry forwardt (data52)  unused net operating loss carry for-
wardt1)/total assetst1 (data6). Summary statistics for our sample may be found in
LAGPERM = last year’s PERMDIFF calculation scaled by total assetst1
Table 3. Panel A provides information regarding our three
(data6).
(a) If data33 had missing data, then we set the value to 0. If data33 = .C
aggressive tax variables. The mean of DTAX is zero, as DTAX
then we set the value equal to goodwill (data204). If data204 also had is the error term from Eq. (1).12 Both C_ETR and B_ETR have
missing data then we set the value to 0. means which are positive and significantly different than
(b) If data55 had missing data or data55 = .C then we set the value equal zero. It should be noted that the average B_ETR in our
to 0.
(c) If data49 had missing data or data49 = .C then we set the value equal
10
to 0. The vast majority of these observations (177 observations) were
(d) If data64 is missing data then we set the value equal to 0. eliminated due to missing NOL data. In order to reduce the amount of
(e) If data173 had missing data then we set the value equal to 0. firms that were lost due to missing data, we hand-collected NOL data from
(f) If data63 is missing data then we recalculate the value to be equal to the firms’ Form 10K tax footnote disclosures. Of the 322 observations that
total income taxes (data16)  foreign income taxes(data64)  state were missing NOL data in Compustat, we were able to maintain 145 of the
income taxes (data173)  deferred taxes (data50). If these data items are observations in our sample.
11
missing as well, then we eliminate the observation from our sample. Our sample has 150, 202, and 176 unique firm observations for the
Values in the table below represent mean coefficient estimates; values in sample period 1996-2000 for the DTAX, B_ETR, and C_ETR dependent
parentheses represent t-statistics. variables, respectively.
⁄ ⁄⁄ 12
, , and ⁄⁄⁄ denote significance at the 90%, 95% and 99% levels, The median of DTAX is statistically significant, due to rounding: the
respectively. actual median of DTAX is 0.004.
 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104 101

Table 3
Descriptive statistics. This table provides descriptive statistics on whether the mean and median are significantly different than zero for all sample variables.

Panel A: Dependent variables

Variable N Mean Std. Dev. Min 25% Median 75% Max
DTAX 460 0.00 0.09 0.87 -0.01 0.00*** 0.02 0.62
B_ETRi,t 682 0.26*** 0.23 0.00 0.00 0.31*** 0.38 1.00
C_ETRi,t 591 0.18*** 0.22 0.00 0.00 0.10*** 0.28 1.00
Variable definitions:
DTAX Residual = the residual, ei,t, is calculated from Eq. (1). See Table 2 for detailed definitions of each variable.
B_ETR(a) = measure of the amount of tax expense per dollar of pre-tax earnings, =total income taxest (data16)/[pretax incomet (data170)-special itemst
(data17)]
C_ETR(a) = measures the amount of cash taxes paid per dollar of pre-tax earnings, =income taxes paidt (data317)/[pretax incomet (data170) – special
itemst (data17)]
(a) If the calculated C_ETR or B_ETR are less than 0 (greater than 1), the value is set at 0 (1).

Panel B: Independent variables for Eq. (2)

Variable AA subsample Non-AA subsample Difference
N = 178 N = 282
39% of Total sample 61% of Total sample
Mean Std. Dev. Median Mean Std. Dev. Median
SIZE 6.34 1.66 6.27 5.97 1.74 6.17 0.38**
FOR 0.44 0.50 0.00 0.39 0.49 0.00 0.05
LEV1 0.32 0.21 0.31 0.33 0.26 0.31 0.01
MTB 2.30 4.06 2.10 2.41 2.97 2.10 0.12
PTROA 0.06 0.14 0.07 0.03 0.19 0.06 0.03*
DPTCFO 0.03 0.11 0.03 0.02 0.12 0.02 0.01
NOL 0.53 0.50 1.00 0.65 0.48 1.00 0.12***
AF 0.63 0.48 1.00 0.55 0.50 1.00 0.09*
NUMEST 0.01 0.02 0.01 0.01 0.02 0.00 1.65*
EM1 0.01 0.11 0.00 0.04 0.19 0.00 0.02
EM2 0.10 0.29 0.00 0.04 0.20 0.00 0.05**
EM3 0.04 0.21 0.00 0.04 0.19 0.00 0.01
Variable definitions:
SIZE = natural logarithm of total assetst (data6).
FOR = dummy variable set to 1 if the absolute value of foreign pretax incomet (data273) is greater than 0, and 0 otherwise.
LEV1 = leverage, =[long-term debtt (data9) + debt in current liabilitiest (data34)]/total assetst (data6).
MTB = market-to-book ratio, =[fiscal year closing pricet1(data199)  common shares outstandingt1(data25)]/total common equityt1(data60).
PTROA = pretax return on assets, =pretax incomet (data170)/total assetst1 (data6).
PTCFO = the change in pretax cash flows from operations, =(PTCFOt  PTCFOt1)/total assetst1 (data6); where PTCFO = net cash flow from operating
activitiest (data308) – extraordinary items and discontinued operationst (data124) + income taxes paidt (data317).
NOL = dummy variable set to 1 if the firm has net operating loss carryforwards (data52) in year t, and 0 otherwise.
(a)
AF = dummy variable set to 1 if the number of analyst following the firm is greater than 0, and 0 otherwise.
NUMEST(a) = the number of analysts following the firm/total assetst1 (data6).
EM1 = dummy variable set to 1 if {net incomet (data172)/[fiscal year closing pricet1 (data199)  common shares outstandingt1 (data25)]} > 0 and
60.01, and 0 otherwise.
EM2 = dummy variable set to 1 if {[net incomet (data172) – net incomet1]/[fiscal year closing pricet2 (data199)  common shares outstandingt2
(data25)]} > 0 and 60.01, and 0 otherwise.
EM3(a) = dummy variable set to 1 if (reported EPS valuet – median EPS estimatet) > 0 and 60.01, and 0 otherwise.
(a) We restricted the IBES data so that each firm has only one observation per analyst per year. We retained the most recent analyst forecasts prior to
the earnings announcement. We also restricted that analyst forecast had to be made within one year (after last year’s earnings announcement
date). We set NUMEST and AF to 0 if there was missing IBES data.

Panel C: Independent variables for Eq. (3)

Variable AA subsample Non-AA Difference
N = 244 N = 438
36% of Total sample 64% of Total sample
Mean Std. Dev. Median Mean Std. Dev. Median
SIZE 5.88 1.82 5.78 5.53 1.85 5.65 0.35**
FOR 0.42 0.49 0.00 0.34 0.48 0.00 0.08**
LEV2 0.25 0.20 0.26 0.27 0.24 0.24 0.02
MVE 1436 3983 282 1332 4197 190 103.91
BM 0.55 1.30 0.46 0.45 1.60 0.44 0.11
EP 0.05 0.63 0.06 0.18 0.91 0.05 0.12*
ROA 0.03 0.41 0.05 0.00 0.22 0.04 0.03
INT 0.11 0.17 0.01 0.08 0.13 0.00 0.03**
R&D 0.02 0.10 0.00 0.02 0.07 0.00 0.00
ADV 0.00 0.01 0.00 0.00 0.02 0.00 0.00*

(continued on next page)

102 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104

Table 3 (continued)
Panel C: Independent variables for Eq. (3)
Variable AA subsample Non-AA Difference
N = 244 N = 438
36% of Total sample 64% of Total sample
Mean Std. Dev. Median Mean Std. Dev. Median
CAP 0.44 0.29 0.39 0.45 0.29 0.40 0.01
INV 0.09 0.12 0.03 0.10 0.15 0.02 0.01
Variable definitions:
SIZE = natural logarithm of total assetst (data6).
FOR = dummy variable set to 1 if the absolute value of foreign pretax incomet (data273) is greater than 0, and 0 otherwise.
LEV2 = leverage, =total long-term debtt (data9)/total assetst (data6).
MVE = market value of equity, =fiscal year closing pricet (data199)  common shares outstandingt (data25).
BM = book-to market ratio, =total common equityt (data60)/MVEt; EP = earnings-to-price ratio, = [pretax incomet (data170) – special itemst (data17)]/
MVEt.
ROA = return on assets, =[pretax incomet (data170) – special itemst (data17)]/total assetst (data6).
INT = intangibles, =intangiblest (data33)/total assetst (data6).
R&D = research and development, =research and development expenset (data46)/total assetst (data6).
ADV = advertising, =advertising expenset (data45)/total assetst (data6).
CAP = capital intensity, =net property, plan, and equipmentt (data8)/total assetst (data6).
INV = inventory intensity = total inventoriest (data3)/total assetst (data6).
(a) If the R&D and ADV variables have missing data, then we set the value equal to 0.
⁄ ⁄⁄ ⁄⁄⁄
, , and denote significance at the 90%, 95% and 99% levels, respectively.

Table 4
Univariate analysis. This table provides the univariate analysis of aggressive tax measures.

AA Non-AA Differences across subsamples (AA less non-AA)

N Mean Median N Mean Median Pervasive aggressiveness Neutral Willingness to pay Mean difference
DTAXi,t 178 0.000 0.003 282 0.005 0.004 + 0  0.005
B_ETRi,t 244 0.266 0.308 438 0.260 0.302  0 + 0.006
C_ETRi,t 214 0.175 0.120 377 0.180 0.099  0 + 0.004

All variables used herein have been previously defined.

sample, 26%, is below the statutory tax rate of 35%, suggest- (NOL, stated in absolute value terms), possibly suggesting
ing the presence of aggressive tax reporting among all sam- greater maturity of AA clients. Finally, as expected based
ple firms. However, the median B_ETR of 31% is significantly upon prior research (Krishnan, 2005), in Panel B we find
greater than the mean (and closer to the statutory rate), sug- greater earnings management among the AA subsample, rel-
gesting the presence of aggressive tax reporting outliers. It ative to the non-AA subsample, where earnings manage-
should also be noted that there is a large overlap between ment is measured as the change in net income, scaled by
observations across the three aggressive tax variables.13 equity value (EM2).
Descriptive statistics regarding our control variables in
Eqs. (2) and (3)14 may be found in Panels B and C, respec-
Results
tively. Of particular noteworthiness is that both Eqs. (2)
and (3) contain approximately one-third AA firms (39% in
Our univariate results may be found in Table 4. The
Eq. (2) and 36% in Eq. (3)), suggesting that AA is well repre-
absolute values of differences cross subsamples are small
sented in the Houston area during our sample period. In gen-
for three aggressive tax variables. Further, none of these
eral, the AA and non-AA subsamples appear rather similar,
mean differences are statistically significant. However,
as there are few statistical differences across independent
these results should be interpreted with caution, as there
variables. There do appear to be differences across both
are many firm-specific variables correlated with aggressive
subsamples in regards to size: in both Panels B and C we find
tax measures. We control for these correlated variable in
that AA firms are slightly larger in terms of total assets (SIZE)
our multivariate analyses below.
stated as the natural logarithm. In Panel C, we find greater
Our multivariate results may be found in Table 5. The
intangibles (INT) and foreign pre-tax income (FOR) among
variable of interest across all three multivariate regressions
AA firms, relative to non-AA firms. Further, in Panel B we
is AA. While our models provide relatively good fits, as evi-
additionally find that non-AA firms have greater NOLs
denced by robust adjusted-R2 values, none of the AA vari-
ables are statistically significant. As such, it appears as if
13
In untablulated statistics, we find that 93.7% of observations in the the sample of firms known to have a culture of aggressive
DTAX subsample are in both of the ETR subsamples. Additionally, all of the
591 observations in the C_ETR subsample are included in the B_ETR
financial reporting (Houston AA clients) were not addition-
subsample. ally aggressive in their tax reporting, as suggested by Frank
14
We have included summary statistics for our larger B_ETR sample. et al. (2009). Further, the financially aggressive firms did
 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104 103

Table 5
Multivariate analysis. This table provides results from estimating Eqs. (2) and (3).

Dependent variable Dependent variable

DTAX B_ETR C_ETR
Intercept 0.02 Intercept 0.12⁄⁄⁄ 0.10⁄⁄
(0.91) (3.04) (2.36)
AA 0.00 AA 0.01 0.01
(0.63) (0.37) (0.72)
SIZE 0.00 SIZE 0.02⁄⁄⁄ 0.00
(1.14) (2.91) (0.01)
FOR 0.02⁄ FOR 0.03 0.05⁄⁄
(1.90) (1.52) (2.56)
LEV1 0.01 LEV2 0.02 0.00
(0.85) (0.52) (0.07)
MTB 0.00⁄ MVE 0.00 0.00
(1.95) (1.10) (1.03)
PTROA 0.33⁄⁄⁄ BM 0.01 0.01
(11.97) (0.82) (1.26)
DPTCFO 0.08⁄⁄ EP 0.03⁄⁄ 0.01
(2.27) (2.07) (0.78)
NOL 0.02⁄⁄ ROA 0.09⁄⁄ 0.19⁄⁄⁄
(2.63) (2.47) (3.20)
AF 0.01 INT 0.12⁄ 0.20⁄⁄⁄
(0.59) (1.70) (2.79)
NUMEST 0.36 R&D 0.01 0.54⁄⁄
(1.36) (0.08) (2.54)
EM1 0.03 ADV 0.59 0.41
(1.12) (1.25) (0.58)
EM2 0.02 CAP 0.00 0.06
(1.13) (0.04) (1.36)
EM3 0.00 INV 0.08 0.24⁄⁄⁄
(0.09) (1.05) (3.05)
No. Obs. 460 682 591
Adj-R2 0.269 0.087 0.149

Eq. (3) is estimated with both the book effective tax rate (B_ETR) and the cash effective tax rate (C_ETR) as the dependent variable.
Values in the table below represent coefficient estimates; values in parentheses represent t-statistics. All variables used herein have been previously
defined.
⁄ ⁄⁄
, , and ⁄⁄⁄ denote significance at the 90%, 95% and 99% levels, respectively.

not pay, in relative tax dollars, for their aggressive financial
reporting, as suggested by Erickson et al. (2004). Thus, it
appears as if our results do not support either the willing-
ness to pay or pervasive aggressiveness theories outlined
earlier. Instead, it appears as if the firms with a culture of
aggressive financial reporting are able to conduct their
aggressive financial reporting up to a relative tax-thresh-
old, which enabled them to break-even with their neigh-
borhood competition in terms of tax aggressiveness.
Conclusions
As noted by Frank et al. (2009), the intersection of
financial and tax reporting is ‘‘a relatively unexplored area
of accounting research.’’ We shed new light on this ‘‘rela-
tively unexplored area’’ herein by exposing the tax report-
ing behaviors of firms audited by AA. We find evidence that
Houston-based AA clients, who have been shown to allow
clients to be aggressive in their financial reporting, exhibit
similar aggressiveness in their tax reporting, relative to a
control sample of non-AA Big 5/6 Houston clients. Instead
of paying for additional earnings in tax dollars, as sug-
gested by Erickson et al. (2004), or being aggressive across
both sets of books, as suggested by Frank et al. (2009), our
findings suggest that firms may be aggressive in their
financial reporting up to a relatively neutral tax threshold.
The findings herein have implications for multiple par-
ties. Our findings show that AA Houston clients were not
relatively aggressive in their tax reports, suggesting that
costs exist to restrict aggressive reporting across both sets
of books. Thus, while critics call for increased alignment of
book and tax income (see Carnahan & Novack, 2002), it
may be the case that a balance already exists. Our findings
may additionally be of interest to current and potential
investors, as they suggest that, contrary to existing theo-
ries, aggressive financial reporting does not necessarily
cost a firm relative tax dollars.
Our study is subject to certain limitations. First, our
findings are only as strong as our measures of tax
aggressiveness. While our tax aggressive measures have
been shown to be correlated with tax sheltering (see
Frank et al., 2009) they are estimates of aggressive tax
reporting and subject to measurement error. Second,
the set of control variables used in Eqs. (2) and (3)
may not be complete, as tax reporting is both complex
and multidimensional. Finally, our interpretations in
relation to the overlap of financial reporting and tax
reporting aggressiveness rest upon the assumption that
AA clients served by the Houston office from 1996 –
2000 were indeed aggressive in their financial reporting.
We leave it to future studies to further scrutinize and ex-
pand upon the findings herein.
104 W. Heltzer et al. / Research in Accounting Regulation 24 (2012) 96–104

References Frank, M., Lynch, L., & Rego, S. (2009). Tax reporting aggressiveness and Its
relation to aggressive financial reporting. The Accounting Review,
84(2), 467–496.
Badertscher, B., Phillips, J., Pincus, M., & Rego, S. (2009). Earnings
Krishnan, G. (2005). Did Houston clients of Arthur Andersen recognize
management strategies: to conform or not to conform? The
publicly available bad news in a timely fashion? Contemporary
Accounting Review, 84(2), 63–98.
Accounting Research, 22(1), 165–193.
Cahan, S., & Zhang, W. (2006). After Enron: auditor conservatism and ex-
Krishnan, G. (2007). Did earnings conservatism increase for former
Andersen clients. The Accounting Review, 81, 49–82.
Andersen clients? Journal of Accounting, Auditing & Finance, 22(2),
Carnahan, I., & Novack, J. (2002). Two birds, one stone: would firms inflate
141–163.
their earnings if it meant more taxes? Forbes Magazine, 40.
Lassila, D., Omer, T., Shelley, M., & Smith, L. (2010). Do complexity,
Chaney, R., & Philipich, K. (2002). Shredded reputation: the cost of audit
governance, and auditor independence influence whether firms retain
failure. Journal of Accounting Research, 40(4), 1221–1245.
their auditors for tax services? Journal of the American Taxation
Cook, K., & Omer, T. (2010). The cost of independence. Evidence from
Association, forthcoming.
firms’ decisions to dismiss auditors as tax-service providers. Working
Maydew, E., & Shackelford, D. (2005). The Changing Role of Auditors in
paper, University of Arizona and Texas A&M University.
Corporate Tax Planning. Working paper, University of North Carolina
DeAngelo, L. (1981). Auditor size and audit quality. Journal of Accounting
at Chapel Hill.
and Economics, 3(3), 183–199.
Omer, T., Bedard, J., & Falsetta, D. (2006). Auditor-provided tax services:
Dyreng, S., Hanlon, M., & Maydew, E. (2008). Long-run corporate tax
the effects of a changing regulatory environment. The Accounting
avoidance. The Accounting Review, 83(1), 61–82.
Review, 81(5), 1095–1117.
Erickson, M., Hanlon, M., & Maydew, E. (2004). How much will firms
Schmidt, S. (2002). Tensions flare at trial of Andersen. Washington Post,
pay for earnings that do not exist? Evidence of taxes paid on
May 10, El.
allegedly fraudulent earnings. The Accounting Review, 79(2),
Squires, S., Smith, C., McDougall, L., & Yeack, W. (2003). Inside Arthur
387–408.
Andersen: shifting values, unexpected consequences. New Jersey:
Francis, J., Stokes, D., & Anderson, D. (1999). City markets as a unit of
Financial Times Prentice Hall.
analysis in audit research and the reexamination of Big Six market
Toffler, B., & Reingold, J. (2003). Final Accounting: ambition, greed and the
shares. Abacus, 35(2), 185–206.
fall of Arthur Andersen. New York: Broadway Books.