Corporate Tax Compliance: The Role of Internal and External Preparers
Kenneth Klassen
University of Waterloo
kklassen@uwaterloo.ca
Petro Lisowsky
University of Illinois
at Urbana-Champaign
lisowsky@illinois.edu
Devan Mescall
University of Saskatchewan
mescall@edwards.usask.ca
Abstract:
Using a survey of tax executives and proprietary data on who signs the corporate tax return, we
investigate tax reporting strategies to understand how firms are managing this challenging task,
and what effect tax preparers have on corporate tax positions. Specifically, we investigate
whether tax preparer type—whether internal, external auditor, or external non-auditor—is related
to tax aggressiveness. Using IRS data on who signs a firm’s tax return, we find that (1) internal
and external non-auditor preparers exhibit greater tax aggressiveness than external auditor
preparers; and (2) publicly disclosed tax fees paid to a firm’s auditor do not provide infomration
sufficient to replicate this result. In our survey, tax executives report that their firms outsource
only 30% of their compliance and planning work, and seldom utilize their auditor exclusively for
such work. Applying conventions in tax fee research to infer preparer type, we estimate that tax
fees incorrectly classify between 20 and 62 percent of firms into tax preparer types that do not
match those reported on a firm’s tax return. Our findings are important given that the paucity of
archival research on tax preparers and the importance of tax advisors to the tax-related decisions
made by companies in the U.S. economy.
Keywords: paid preparer, tax expert, audit expert, FIN 48, tax reserve, tax aggressiveness
______________________________________________________________________________
We obtain confidential tax return data from the Internal Revenue Service (IRS) Large Business & International
Division’s (LB&I); these data are not publicly available. Because tax return data are confidential and protected by
data non-disclosure agreements under the Internal Revenue Code, all statistics are presented in the aggregate; no
statistics with fewer than three observations are disclosed. Any opinions are those of the authors and do not
necessarily reflect the views of the IRS. The authors thank Stephen Powers for excellent research assistance.
1. Introduction
The current corporate tax environment is characterized by a high degree of uncertainty.
Uncertainty in the tax law can create opportunities for planning, but can also create challenges in
compliance. To evaluate the compliance responsibilities and implications surrounding business
transactions, companies employ internal tax specialists and/or hire external advisory firms. The
effect of the tax advisor in compliance decisions of corporations is very important, but has not
been subject to much archival empirical research due to the lack of available data. This paper
begins to overcome that shortcoming by using proprietary data on who signs the corporate tax
return and survey responses of 170 tax executives to examine the effect of the tax service
provider on the compliance of tax positions claimed by corporations.
An extensive literature has examined the relation between auditors providing tax services
and the effect on the audit process.1 There continues to be a debate over whether auditors
providing tax services impairs independence, creates knowledge spillover, or both. For example,
Davis, et al. (1993) find no evidence of a knowledge spillover, while 43% of CFOs surveyed by
Cripe and McAllister (2009) identify knowledge spillover as their primary reason for integrating
their choice of auditor and tax preparer. Frankel, et al. (2002) show the provision of non-audit
services generally is positively associated with earnings management, but Ashbaugh, et al.
(2003), Chung and Kallapur (2003), and Larcker and Richardson (2004) refute this relation.
Kinney, et al. (2004) show that the provision of tax non-audit services in particular reduces
restatements. Francis and Ke (2006) show that the equity market uses non-audit service fees to
assess audit quality (negatively), but Pittman and Fortin (2008) show that increasing tax nonaudit services lowers cost of debt financing. Despite this lack of resolution, a majority of studies
1
For example, see Davis et al. (1993); DeFond et al. (2002); Abbott et al. (2003); Kinney et al. (2004); Francis and
Ke (2006); Pittman and Fortin (2008); Lim and Tan (2008); and Zamar et al. (2011).
1
in the accounting literature continue to look at the decision to choose the auditor as the firm’s tax
service provider through the lens of the audit and auditor-related consequences. The literature
remains mostly silent on the role of the tax preparer and its effect on tax-related decisions.
Two notable exception are Neuman, et al. (2011) and McGuire et al. (2011). Neuman et
al. explore the not-for-profit sector where the identity of the tax preparer is disclosed in publicly
available tax returns.2 In testing the effect of paid preparer type on donations (their measure of
credibility), they find that external preparers are positively associated with donations. They
interpret this evidence to suggest that the public views self-prepared returns as less credible.
Although they infer that their results generalize to the for-profit sector, direct evidence on the
role of tax preparer type in a corporate setting remains elusive. Our objective is to fill this void.
In particular, our research questions ask whether tax preparer type is associated with tax
aggressiveness; and how accurately public data on tax fees paid to a firm’s auditor capture the
actual tax return preparer type.
McGuire et al. (2011) use auditor fee data, as descibed more fully below, to show that
firms who obtain any tax services from their auditors, that are also local industry tax experts, are
more aggressive in their tax reporting. Our research is a complement to McGuire et al. because
we ask whether using the company’s auditor to be primarily responsible for tax compliance in
particular affects tax-related decisions. Secondly, we focus on a measure of tax decision making
(uncertain tax benefits) that is generally regarded as more extreme than those analyzed by
McGuire et al. (measures based on effective tax rates and permanent tax-book differences)
because, guided by the theoretical model of Phillips and Sansing (1998) the advisor should
matter most where the tax positions are particularly uncertain.
2
In particular, IRS Form 990 is required for tax-exempt entities and is available for public examination.
2
To answer these questions, we use confidential data from the Internal Revenue Service
(IRS) on who signs the tax return for 1,533 firm years during 2008 and 2009. The tax return data
allow us to focus on whether the compliance activities of the firm are primarily administered by
the internal tax department, the firm’s financial statement auditor, or an external non-auditor
preparer. Until now, researchers have only been able to observe the dichotomous choice of a
corporation using or not using its auditor for tax services (e.g., Lassila, et al. 2011); and even so,
it remains unclear whether the services pertain to tax compliance or planning.
To set the context for this analysis, we survey 170 tax executives from the Tax
Executives Institute (TEI) who disclose their use of external firms for compliance work. Our
survey reveals that 20% of the respondents do not use any external firm for compliance work,
16% use their auditors only, 52% use another type of consultant (either an accounting firm that is
not their auditor, or another type of provider), and 12% use both their auditor and other
providers. The survey also reveals that external providers perform only 30% of compliance
work, and external providers are used for a similar percentage of tax planning work. Finally,
companies who do not outsource their tax compliance work, and to a somewhat lesser extent
those who use accounting firms that are not their auditor, are generally larger firms.
With this background, we employ the theoretical model of Phillips and Sansing (1998) to
form predictions around the relation between the type of tax preparer and the observed
aggressiveness of the company’s tax-related decisions. As described in Section 3, we predict that
internally-prepared tax returns and returns prepared by external firms that are not the companies’
auditors, will be more aggressive, on average, because theory suggests that companies who do
not hire an external preparer tend to have a high tolerance for reassessment and are willing to
3
claim aggressive tax positions.3 Phillips and Sansing (1998) also predict that, in equilibrium,
companies’ aggressiveness will be increasing in their preparer’s expertise and we assume that
hiring another firm raises the average quality of the tax preparer.
Using the FIN 48 tax reserves as our proxy for tax aggressiveness, and linking them to
the tax preparer identity from the tax returns, our empirical tests support these hypotheses. In
particular, we find that external auditor-prepared tax returns are the least aggressive compared to
returns prepared by internal tax departments or other external preparers. Including proxies for
both internally prepared returns and non-auditor prepared returns in our models, their
coefficients are similar, and both types of companies have tax returns that are more aggressive
than auditor-prepared tax returns.
In a final analysis, we examine how accurate using publicly available data on tax fees
paid to a firm’s auditors are with respect to inferring tax preparer type from the tax return. We do
so to assess whether the window available to investors and researchers in financial reports on tax
compliance is useful. We find that more than 80% of companies in our sample publicly disclose
tax fees paid to their auditor, but only 20% of tax returns are in fact prepared by the auditor.
Using a variety of proxies generated from the tax fee data to infer preparer type, we attempt to
replicate our main results using tax returns, but are unable to do so. We estimate that tax fees
incorrectly classify between 20 and 62 percent of firms into tax preparer types that do not match
those reported on a firm’s tax return. This finding suggests that tax fee data has limited value in
correctly identifying the tax return preparer, and thus the tax compliance advisor of the company.
Given the paucity of archival research on tax preparers generally, our research makes a
significant contribution to understanding their important role in the U.S. tax system. Specifically,
3
We acknowledge that the choice of preparer is likely endogenous and account for this possibility in our empirical
tests.
4
our research is the first to document the identity of tax return preparers for a large sample of U.S.
companies, reports its links to tax aggressiveness, and evaluate how tax fees mostly do not
accurately capture tax compliance activity that is observed directly from the tax return.
The paper proceeds as follows. Section 2 describes our setting of tax preparers by
reporting results of our survey of tax executives at TEI. Sections 3 and 4 develop hypotheses
and describe the research design. Section 5 reports the results of tests using large-sample tax
return data. Section 6 explores the accuracy of public fee data in the context of testing our
research questions. Section 7 concludes.
2. Institutional Setting
2.1
SURVEY OF TAX EXECUTIVES ON TAX COMPLIANCE
To gain an understanding of the general use of external tax advisors in a corporate
compliance environment, we surveyed a sample of the Tax Executives Institute (TEI)
membership on this topic. The survey was conducted online between October and December,
2010, and members were contacted directly by TEI in two emails. Both the original email and
the second follow-up email were sent to tax executives at 2,700 multinational firms.
We
received responses from 218 tax executives resulting in a response rate of 8.1%.4 The response
rate is comparable to previous surveys of senior executives, including the 8.8% response rate in
Graham and Harvey (2001) and 9% in Slemrod and Venkatesch (2002), although it is lower than
the exceptional 26.6% rate in Hanlon, et al. (2010).5
4
Of the respondents, 40% identified themselves as their firm’s Tax Director, 39% identified themselves as VP Tax
or CTO, 19% as Tax Manager, and 2% as CFO.
5
To allow the researchers to ask the questions of interest and to insure the responding members of TEI felt
comfortable responding, it was agreed that all questions would be optional and members should feel free to skip any
questions. As a result, the response rate varies across questions.
5
Pursuant to our agreement with TEI, participation in the survey was optional. Therefore,
we evaluate whether our inferences suffer from non-response bias. The survey was conducted
by sending out the request for participation by TEI leadership to its members, and followed up
by a second request 42 days later. This approach allows us to test for evidence of a non-response
bias on the overall survey using Wallace and Mellor’s (1988) method (see also, for example,
Graham and Harvey, 2001). Our results show no statistical difference in the characteristics of the
early versus late respondent firms in terms of size, industry, or geography, suggesting the effect
of any non-response bias is likely minimal.
We also compare our respondents to the overall population of TEI member firms. In
Table 1, we compare the size and industry of our sample to the 2005 survey that TEI required its
total membership to answer. Our sample of surveyed firms is generally comparable to slightly
larger in revenue, total assets, and tax budget. For example, we find 17% and 4% of our sample
are in the largest two revenue categories of $10-$50 billion and greater than $50 billion,
respectively, while the 2005 population had 11% and 2%, respectively. This may suggest our
responses come from larger firms than the overall TEI membership.
[Insert Table 1 about here]
The respondents to our survey and the TEI membership have a similar industry
composition with the exception of a higher concentration of manufacturing firms; i.e., 49%
compared to 36% in the 2005 survey, and 11% in industries not identified in the 2005 survey.
Although we could not identify any obvious bias stemming from this difference in industry
composition or the somewhat larger respondents, we cannot rule out a bias of unknown severity
that may affect our inferences. Nevertheless, the survey data provide additional context against
which we can evaluate the different types of tax preparers used by U.S. corporations.
6
Summary statistics of survey questions are presented in Table 2. One question asked
respondents to identify who they use to provide assistance with tax compliance.
Possible
responses include the company’s auditor, other accounting firms, and/or non-accounting firms
such as lawyers, consultants, etc. Respondents indicated that they use one, two, or all three of
these alternative providers. Of the seven possible combinations listed, the exclusive use of one
accounting firm other than the company’s auditor was the most common response at 41%.
Twenty percent do not outsource any compliance work. Interestingly, only 16% of the
respondents outsourced their compliance work exclusively to their auditor. Seventeen percent
indicate that they use more than one type of outside firm to perform compliance work.
[Insert Table 2 about here]
The next column of statistics summarizes the amount of compliance work that is
outsourced. Overall, only approximately 30% of compliance work is outsourced. However,
when only the company’s auditor or another accounting firm is used, approximately 40% of
compliance work is outsourced. On the other hand, only 22% of compliance work is outsourced
when only a non-accounting firm is used.
Using more than one type of provider does not dramatically increase the proportion of
work outsourced. Column 5 lists the percentage of planning work that is outsourced.
Comparison of columns 4 and 5 reveals that on average, the same proportion of work is
outsourced, 30%, but in almost all categories, the average amount of planning work outsourced
is smaller (for example, for auditor only, 40% of the compliance work is outsourced, but only
31% of planning work is outsourced). The only group for which planning work is a higher
percentage is when only a non-accounting firm is used for compliance work. Overall, these
statistics suggest that much of the typical companies’ tax work is done in-house, and that if
7
anything, planning work is even more commonly done in house too. This provides some
validation that firms with internal tax departments are well-resourced, indicating an ability to
pursue advantageous tax positions.
The last two columns of Table 1 provide some summary statistics on respondent size in
terms of the number of internal tax personnel employed and total assets. On average, the firms
who do not outsource any compliance, i.e., “in-house,” are on average the largest in asset size,
while the group that does not use its auditors (both non-auditor accounting firms and nonaccounting firms) for compliance work has the largest internal tax personnel. Firms using only a
non-auditor accounting firm are, on average, large, and also have sizable tax departments. Those
firms that use only a non-accounting firm are the smallest, on average, but have a large number
of internal tax personnel, suggesting that firms generally use internal personnel, then supplement
with specialists (such as lawyers) as necessary. Those who use only their auditor are both small
in size and have the fewest internal tax personnel (though on a size-adjusted basis these firms
may also have sizable tax departments).
Overall, our analysis reveals that survey respondent-firms generally do not outsource the
majority of either their compliance or planning work. It also shows that while many firms
choose a single type of firm for compliance, there is a noteworthy percentage that uses more than
one type. Firms using only their auditor tend to be small while firms using only a non-audit
accounting firm or not outsourcing at all tend to be large in terms of asset size and tax personnel.
Other than those who do not outsource any compliance, other categories generally outsource a
smaller proportion of their planning than the proportion of their compliance work. In the next
section, we describe tax return information on the party primarily responsible for the tax
compliance work of the corporation, which we later employ in our empirical tests.
8
2.2 SIGNATURE INFORMATION ON THE TAX RETURN
To our knowledge, there is no archival research on the importance of a specific type of
tax preparer in a corporate setting.6 Studies such as Christian, et al. (1993), Hite and McGill
(1992), and Long and Caudhill (1987) investigate various aspects of the decision by individual
taxpayers to use paid tax preparers, and Neuman, et al. (2011) examine tax preparer use in the
non-profit sector. Although some factors identified in the individual tax preparer literature, such
as time, cost, and complexity (Christian et al. 1993) may impact corporate decisions to seek
external tax advisors as well, the tax compliance environment for corporations is certainly
different from individuals and non-profits along other dimensions. For example, differences in
audit probabilities, agency and reputational costs, resource availability, and profit-motive, all
create a unique environment in which corporations evaluate and implement tax compliance
activities.
The main challenge facing external observers who assess corporate tax compliance is
observing its qualitative input. For example, one cannot observe the set of transactions available
to the corporation; design details of the transactions; or the labor, time, and management talent
required to implement the transactions. To overcome this shortcoming, interested parties are
forced to utilize noisy observable outputs to infer unobservable inputs. For example, disclosures
in securities documents of fees paid to corporate auditors for tax services are sometimes used as
proxies for tax planning activities (e.g., Cook, et al. 2008). However, it is unclear whether tax
fees (a) in fact represent compliance-related activities at all; (b) accurately reflect the auditor’s
relative role and importance in the context of all the corporation’s tax compliance activities, and
(c) appropriately describe how much the corporation also uses its internal tax department or other
6
We discuss related theoretical research in Section 3.
9
external non-auditor parties to prepare and report the tax return to the tax authorities.
Unfortunately, these issues only become more opaque when tax fees paid to auditors are not
disclosed at all.
One avenue toward obtaining a clearer picture of a corporation’s tax compliance
activities is by observing the identity of the party that signs the tax return. Although this
information is not publicly available for corporations, it is for the first time made available to one
of the authors on a limited, confidential basis for purposes of this study. Importantly, by using
tax returns, the signature’s observability is not conditional on the corporation using its auditor for
tax services. Also, the tax return preparer’s identity provides a more direct signal of compliance
work than do tax fees.7
Most notably, Internal Revenue Code (IRC) Section 6694 outlines penalties related to
external preparers signing tax returns containing unreasonable positions that result in an
understatement of tax liability.8 The related penalty regime for non-compliance requires the
preparer to pay the greater of $1,000 or half of the income derived (or expected to be derived)
with respect to preparing the tax return (§6694(a)(1)). Clearly this latter penalty, coupled with
related litigation and reputational costs, can be significant for large external paid preparers, such
as those in our sample. The effect of the penalty regime is that an external preparer will not sign
the tax return unless it is confident it has obtained enough information to properly defend the
underlying tax positions and advocate for the client in the event of a dispute with the tax
authority. Therefore, if the external advisor does not sign the tax return, it is either because it did
not provide any tax work at all, or it only provided limited, focused compliance work on a
7
We thank a tax partner at a large accounting firm for extensive input on this subject. We also conduct various tests
on whether tax fees are informative for tax compliance. We report those results in Section 6.
8
An unreasonable position is one that (a) the tax preparer knew or reasonably should have known of the position;
(b) there was not a reasonable belief that the position would more likely than not be sustained on its merits; and (c)
the position was not disclosed or there was no reasonable basis for the position (§6694 (a)(2)).
10
particular transaction (e.g., calculated the foreign tax credit only). A narrow scope of work does
not require the external advisor to sign the tax return. In fact, the advisor will prefer not to sign
the return in this situation because it has not adequately evaluated the merits underlying the rest
of the tax return and thus does not want to incur additional risks by claiming material
responsibility over that tax return. If there is no other external party claiming responsibility over
the tax return compliance work, then the tax return is only signed by the corporation, typically
the senior tax officer (e.g., the Tax Director or Vice President of Tax).
However, if the corporation would like the external party to sign the tax return, the §6694
penalty regime provides strong incentives for the external party to require a much larger scope of
work (and thus more fees) to gather adequate documentation and support underlying the entire
tax return. If an external party reasons that through the additional work, its compliance-related
requirements are met, it signs the tax return (in the “Paid Preparer” section) alongside the
corporation’s tax officer.9 In fact, if the work by the external advisor is substantial, the external
party is required to sign the tax return; conditional on the work performed, the signature itself is
not an election.10 In all, the strict regulatory regime underlying §6694, which exposes an external
preparer to substantial risk if sufficient support is not gathered in preparing the return, provides
strong institutional incentives that ensure the signature information on the tax return represents
the party most substantially associated with the corporation’s tax compliance work. In the next
section, we develop hypotheses regarding the links between tax preparer type and aggressive tax
transactions.
9
Although technically the corporation must always sign the tax return, the absence (presence) of a paid preparer
signature implies that the compliance work is predominantly executed internally (externally). We use this fact in
our research design.
10
The tax partner at a large accounting firm stated to one of the authors that if a client wants his firm to sign the tax
return, his firm has to be “comfortable that it has substantially impacted the development of the return. If we are
signing the return, we have to feel in good conscience that the return is properly prepared. We will not sign a return
that we have not substantially worked on, and similarly, if we have substantially worked on the return, we cannot
duck the responsibility by not signing the tax return.”
11
3. Hypothesis Development
3.1 USING AN EXTERNAL TAX RETURN PREPARER
The effect of a company’s tax preparer on its tax positions is an important, but understudied area of corporate decision making. Recently, Neuman et al. (2011) explore the not-forprofit sector where the identity of the tax preparer is publicly disclosed. In tests relating paid
preparer type and donations (their measure of credibility), they find that external preparers are
positively associated with donations. They interpret this evidence to suggest that the public
views self-prepared returns as less credible.
McGuire et al. (2011) explore the role of expertise in tax aggressiveness. They find a
positive relation between tax aggressiveness and the use of expert tax providers, where expertise
is based on the proportion of all tax fees paid to auditors in the industry-city of the client.11 In
particular, firms paying tax-related fees to an auditor who is also a tax expert have lower
effective tax rates (ETR) and higher book-tax differences, and marginally lower cash ETR and
marginally higher discretionary permanent book-tax differences, consistent with such firms
being more tax aggressive.
To guide our predictions as to when firms choose among preparing their tax return
internally or engaging their auditor or another external provider, we explore the theoretical
model of Phillips and Sansing (1998) (hereafter PS). PS analyze the effects of external tax
preparers in the tax compliance of a firm. The focus of their study is on the effects of requiring a
fixed fee for compliance work, rather than a contingent fee based on the filing position. PS
conclude that a fixed fee contract, relative to a contingent fee contract, raises the expected cost of
tax return preparation, and leads to more taxpayers filing a favorable (aggressive) tax position,
11
Reichelt and Wang (2010) suggest expertise can also be measured at a national level.
12
increasing under-compliance overall. They also note that the use of a higher quality preparer
will lead to greater observed tax aggressiveness.
While not discussed in their paper, the PS model can also provide predictions for the
relation between using an external preparer and the aggressiveness of firms’ tax positions. In
their model, taxpayers are endowed with an exogenous, unobservable aversion to enduring a tax
audit adjustment (that is, an aversion to having a tax position reversed on audit). They use the
parameter λ to denote this characteristic. Thus, firms can be sorted along this dimension, with
lower values of λ denoting firms that are inherently less sensitive to audit adjustments and more
willing to be tax aggressive (i.e., tax aggressive taxpayers use favorable tax positions when the
outcome of a tax audit is uncertain, ex ante).
Use of an external tax preparer reduces uncertainty over the legitimacy of uncertain tax
positions. In a first-best solution (subscripted ‘FB’), where the taxpayer does not need to
motivate the tax preparer to work, taxpayers with λ below a critical value, λ FB, report the
favorable position (more aggressive) when the preparer remains uncertain, whereas those above
the critical value report the unfavorable position (more conservative) when uncertainty
remains.12 With the need to motivate the preparer to investigate the appropriate treatment and to
report as desired by the taxpayer, the critical value, λ FF, is higher than λ FB (with ‘FF’ denoting
the use of a fixed fee contract for the external advisor). Thus, more taxpayers report aggressively
because more taxpayers have λ values less than the critical value.
One can also allow the possibility for internally prepared tax returns. In this case, it can
be shown that the point of divide between reporting aggressively versus conservatively is equal
12
In Phillips and Sansing (1998), taxpayers above the critical value (who are labeled conservative) only report the
favorable tax position if the tax preparer’s research reveals that the favorable position will withstand audit. If the
audit outcome is uncertain, they will report the unfavorable position. Taxpayers below the threshold (who are
labeled aggressive) will report the favorable position if the outcome is uncertain and will only report the unfavorable
position if the preparer’s research shows the position will not withstand an audit.
13
to the first-best critical value. Thus, with no external preparers, taxpayers with λ < λ FB will not
choose to use an external preparer, and always report the favorable (aggressive) position.13 In
the intermediary case, where λ FB < λ < λ FF, the taxpayer will never report the favorable position
without the use of an external preparer. However, with an external preparer, the taxpayer will
report the favorable position unless the knowledgeable external professional discovers that the
favorable position will be overturned on audit. With the availability of an external preparer,
taxpayers below λ FB will trade off the greater certainty that the knowledgeable external preparer
provides with the fees charged by such a preparer. Depending on the parameters of the setting,
there will be a critical value, λ* , below which an external preparer is not hired, and above which
an external preparer is used. Thus, in cases with (1) low values of λ , (2) lower audit probabilities
or (3) lower tax benefit to the favorable position, it is value maximizing to self-prepare because
the expected benefit arising from the higher probability of certainty does not exceed the fees paid
to the preparer.14
In the conservative case, λ FF < λ , the taxpayer will also never report the favorable
position without hiring a professional, but will report the favorable position when the external
preparer is certain that the favorable position is correct. Thus, when λ* < λ , the taxpayer’s
wealth is improved by hiring an external preparer because the expected benefit of reporting the
favorable position some of the time exceeds the cost of the preparer.
From this extension of the Phillips and Sansing (1998) model, the only situation in which
the taxpayer will internally prepare is when the taxpayer’s sensitivity to reassessment is low (i.e.,
13
We assume here that the taxpayer, in the absence of a paid preparer, has no tax knowledge on the issue at hand;
however, the analysis holds for internal tax staff that is less knowledgeable than the external preparer. If the internal
staff has some ability to discern the legitimacy of the position, the company will choose the favorable position
except in the rare cases the internal analysis shows that the position will not withstand audit.
14
Anecdotal discussions with tax directors at two large publicly traded companies confirm this analysis. One reason
their companies do not employ external preparers is because their transactions are complex enough that the cost of
educating and managing a team of external preparers is not worth the potential benefit that the team might provide a
better understanding of the merits of the company’s positions, i.e., the effect on the company’s tax uncertainties.
14
when λ < λ* ). In that case, the taxpayer always reports the favorable (aggressive) position.
However, when hiring the external preparer is optimal for a particular taxpayer, the taxpayer
either reports the favorable position or the unfavorable position depending on the findings of the
professional’s research and the taxpayer’s underlying tolerance parameter (i.e., depending if
λ*
< λ < λ FF or if λ FF < λ ). Thus there is less aggressive reporting with an external preparer, on
average. This leads us to our first hypothesis:
H1:
Tax returns filed without the assistance of an external preparer will be more
aggressive than tax returns filed with the assistance of an external preparer, on
average.
Hypothesis H1 arises from the model of Phillips and Sansing (1998) using the
assumption that the choice to hire an external preparer is endogenous to the model. However, it
remains possible that the firm decides to hire a preparer for other reasons, and this decision is
independent of the firm’s parameter λ .15 Whatever the other dimensions involved in the choice to
self-prepare or to use an external preparer, it continues to be reasonable to assume that the
external preparer is of higher quality than the internal staff (to justify the fees). If the decision to
hire an external preparer is independent of λ and the external preparer is of higher quality than
the internal staff, then consistent with the analysis in Phillips and Sansing (1998), the use of a
higher-quality external provider would raise the threshold value of λ needed to claim aggressive
tax positions. In this setting, conditional on hiring an external preparer, the taxpayer will, on
average, be aggressive more often with an external preparer, contrary to hypothesis H1.
15
Firms that internally prepare tax returns might include those that are less sophisticated and resource constrained,
such that they will not have the capability or talent to develop and execute aggressive tax planning strategies,
resulting in less aggressive reporting overall, than for firms that engage an external preparer.
15
3.2 ROLE OF NON-AUDITOR PROVIDERS
As described in the introduction, the use of auditors as tax advisors has been the subject
of much research. This research suggests that firms choose to use, or not use, their auditor as a
tax preparer for a variety of reasons (see, for example, Omer et al., 2006, and Lassila et al.,
2010). Cripe and McAllister (2009) survey 42 CFOs about their decisions to use their auditor as
a tax preparer. When asked for their primary reason why a firm would choose their auditor as
their tax preparer, the most common response, at 48%, was lower cost. Alternatively, when firms
who used a non-auditor external preparer were asked about the effect of this choice on fees, 31%
identified that they pay higher fees by foregoing their auditor as tax preparer. So if, on average,
and using a non-auditor for tax compliance is more costly, then to make this choice rational, the
benefits of the external non-auditor preparer must exceed the incremental cost.
One of the benefits will be that the external non-auditor preparer will be of higher quality,
on average. Respondents to the Cripe and McAllister (2009) study identified non-auditor tax
expertise as the second most popular reason for choosing to acquire their audit and tax preparer
services from separate providers.16 Even if this is not the main reason to choose another preparer,
when choosing among the alternative firms, we assume that the highest quality preparer is most
likely to be chosen. Thus, given that an external, high quality preparer is chosen, on average, the
non-auditor external preparer will be of higher quality than the auditor.17
16
Auditor independence was the most popular response which is not surprising as the survey was conducted in early
2007 shortly after new PCAOB rules on auditor provided tax services were introduced.
17
For example, assume that there are four preparers possible, denoted A, B, C, and D in order of quality for a given
industry, location, etc. For each taxpayer, one is the auditor. If the taxpayer does not use the auditor, we assume it
always chooses the highest quality preparer among the remaining three choices. In this case, only audit clients of A
will choose a non-auditor preparer that is of lower quality (that is, B). However, clients of all other firms will
choose A, a higher quality tax preparer than their auditor. Given the relatively even distribution of companies
among the major audit firms, on average, the non-auditor preparer will be of higher quality than the auditor.
16
If we consider non-auditor preparers to generally have greater ability, the threshold
between what Phillips and Sansing (1998) refer to as “aggressive taxpayers” versus those labeled
“conservative taxpayers” is increasing in the ability of the tax preparer. In their model, the higher
quality preparers allow marginal taxpayers to report aggressively because reducing the likelihood
that the filing position is uncertain allows more taxpayers to overcome the trade-off between
receiving the expected benefits of the favorable (aggressive) reporting position and suffering the
expected cost of being overturned on audit.
As discussed above, McGuire et al. (2011) study the importance of expertise on tax
aggressiveness. Their empirical work is consistent with a positive relation between using an
expert for tax work and the aggressiveness of the tax position. McGuire et al. (2011) follow
extant research by classifying accounting firms as tax experts if the reported tax fees paid to that
firm (for audit clients) are “high” in the metropolitan area, where “high” is defined as greater
than 30% of reported audit and non-audit fees. One shortcoming of this measure is that only fees
paid to auditors are considered because these are the only fees that are publicly disclosed.
Because we can directly observe the compliance work through the tax return signatures, we can
further refine this literature by classifying a tax preparer as an expert if the external preparer is
not the auditor, as suggested by Cripe and McAllister (2009). Thus, we extend our analysis by
further delineating the external tax preparers into auditors and non-auditors, while still estimating
the effects of internally preparing the tax returns. Thus, assuming non-auditors are experts on
average, the predictions of Phillips and Sansing (1998) lead to the hypothesis below:
H2:
Tax returns filed with the assistance of an external preparer who is not the firm’s
auditor will, on average, be more aggressive than tax returns filed with the
assistance of the firm’s auditor.
17
4. Method
4.1 RESEARCH DESIGN
We design our empirical tests to examine whether tax returns filed without an external
preparer or with a non-auditor external preparer will be more aggressive than tax returns filed
with the assistance of an external preparer. We alternately estimate regressions of the logged FIN
48 tax reserve ending balance (Log_UTB_EB) on INTERNAL_PREP, OTHER_PREP, and
control variables for firm i at time t as follows:18
β0 + β1 INTERNAL_PREPit + β2 OTHER_PREPit
+ β3 LOG_ASSETSit + β4 PRETAX_ROAit
Log_UTB_EBit = + β5 FOR_INCOMEit + β6 NOLit + β7 R&Dit
+ β8 LEVERAGEit + β9 YR2008it +
(1)
16
∑ βkIndit + eit
k =9
We define the model variables in the Appendix.
We use Log_UTB_EB as our proxy for tax aggressiveness because Lisowsky, Robinson,
and Schmidt (2011) find that it is a reliable proxy for corporate tax aggressiveness, namely tax
shelters disclosed as reportable transactions (obtained from the tax return). A key reason
underlying their finding is that both tax shelters and the FIN 48 tax reserve (UTB) contain
information on tax benefits at the extreme end of the tax avoidance continuum, regardless if the
transactions conform or do not conform between financial and tax reporting. Supporting this
notion, they find that the UTB is the only key variable from the tax avoidance literature that
contains information on tax shelter use because the other measures focus on a set (or even subset) of non-conforming transactions only. In particular, Lisowsky, et al. (2011) compare the
18
Since our sample period is short (2008-2009), we follow Petersen (2009) and cluster the standard errors by firm
and include a 2008 time indicator variable in all of our regressions. All regressions also include industry fixed
effects at the one-digit SIC code level. We winsorize continuous variables at the 1st and 99th percentiles to mitigate
the effect of outliers. We employ both OLS and Tobit to ensure that the 154 firms with zero values in the dependent
variable do not alter our inferences.
18
explanatory power of the logged UTB to the GAAP effective tax rate (Rego 2003), cash effective
tax rate (Dyreng, et al. 2008), total book-tax differences (Mills 1998), permanent book-tax
differences, and discretionary permanent book-tax differences (Frank, et al. 2009), and find that
the non-UTB measures—either alone or together, and with or without the UTB—are not
associated with tax shelter use.
Lisowsky, et al. (2011) also demonstrate that the information on tax aggressiveness
contained in the UTB is not eliminated due to reporting discretion arising from non-tax factors,
such as financial reporting preferences (Hanlon and Heitzman 2010), corporate governance
mechanisms (Desai and Dharmapala 2006), independent auditor certification process (Gleason
and Mills 2011), capital market incentives to manage earnings (Gupta, et al. 2011), and tax
director compensation (Armstrong, et al. 2011). They conclude that the tax reserve itself, even if
subject to managerial discretion about reporting aggressiveness or conservatism, is a reliable
proxy for the aggressiveness of the underlying tax position claimed on the firm’s tax return.
Therefore, following advice in Hanlon and Heitzman (2010) to carefully select the tax
aggressiveness proxy to suit the particular research question and empirical evidence in Lisowsky,
et al. (2011), we employ the ending balance UTB to examine the relation between preparer type
and tax aggressiveness. The favorable positions modeled in Phillips and Sansing (1998) are
those with material benefit and uncertainty, and we therefore view UTB as the closest
approximation of this construct.
Our first independent variable of interest is INTERNAL_PREP, equal to one if the firm
does not use an external tax preparer. Because hypothesis H1 seeks to examine the relative
aggressiveness of internal preparers over external ones, we specify INTERNAL_PREP as our test
19
variable. If tax returns filed without an external preparer are more aggressive than those filed
with the assistance of the auditor, β1 will have a positive sign.19
We are also interested in the relation between tax aggressiveness and other external
preparers, as specified in hypothesis H2. We include OTHER_PREP to examine the relation
between using an external preparer that is not the firm’s auditor, rather than the firm’s auditor, to
explain tax aggressiveness. To do so, we specify OTHER_PREP as equal to one if the external
tax preparer is not the firm’s auditor, zero otherwise. We expect the coefficient on
OTHER_PREP, β2, will have a positive sign.
We draw on prior literature regarding investments in tax planning (Mills, et al. 1998),
auditor-provided tax services (McGuire, et al. 2011), and tax reserves (Lisowsky, et al. 2011) to
select our control variables. They include firm size (LOG_ASSETS), profitability
(PRETAX_ROA), foreign income (FOR_INCOME), existence of net operating losses (NOL),
research and development activities (R&D), and debt burden (LEVERAGE). We expect positive
coefficients on all the control variables.
In the hypothesis development, the choice to use an external preparer and the tax
positions taken by the taxpayer were both related to characteristics of the firm. Further, the
choice to use the auditor, versus other options (internally prepare or use another tax provider)
may relate to firm characteristics that also affect the aggressiveness of the firm’s tax positions.
Thus, to estimate equation (1), we explicitly model the endogeneity of this multiple-alternative
preparer choice. To do so, we employ a treatment-effects model (see, for example, Greene,
19
Hypothesis H1 is motivated by internal versus external preparers, but to jointly test H1 and H2, we use the auditprepared firm-years as the base group in both tests. In untabulated regressions without OTHER_PREP, the
coefficient on INTERNAL_PREP is approximately 60% as large but continues to be statistically significant at the
5% level.
20
2008), as extended by Deb and Trivedi (2006) to a multinomial treatment. To summarize their
work, the firm has an indirect utility, EV*, associated with the jth choice of preparer:
EVij* = zi′ α j + δ j1 li1 + δ j2 li2 + ηij
EVi0* is assumed to be zero for the base choice, j = 0, and zi are exogenous covariates with
parameters α. EV* includes latent factors lik which incorporate unobserved characteristics
common to firm i’s preparer choice and the outcome, yi (observed tax aggressiveness). η is an
i.i.d. error term. If dij are binary variables representing the observed choice of preparer of firm i,
then di = [di1, di2], with the probability of the choice represented as
(
) (
Pr di zi ,l i = g zi′ α1 + δ11 l11 + δ12 l12 , zi′ α 2 + δ 21 l21 + δ 22 l22
)
g is a multinomial probability function. Finally, the outcome equation for firm i, formulated in
linear form, is
yi = x i′ β + γ 1 di1 + γ 2 di2 + λ1 li1 + λ2 li2 + ε i
This model can be estimated, as described in Deb and Trivedi (2006), using maximum simulated
likelihood estimation.20
To implement this model in our setting, we identify the choices as 0 if the preparer is the
auditor (the base alternative), di = [1, 0] if the firm does not use an external preparer (i.e.,
INTERNAL_PREP = 1), and di = [0, 1] if the preparer is an external preparer other than the
auditor (i.e., OTHER_PREP = 1). In addition to the exogenous variables from equation (1),
namely firm characteristics (size, profitability, foreign income, NOL, R&D, and leverage), year,
and one-digit SIC industry code, we include NON_TAX_FEE as an additional variable in the
choice equation. NON_TAX_FEE is the fees paid to the auditor for services other than audit and
20
Stata uses this estimation approach in its mtreatreg procedure.
21
tax, deflated by the total non-tax fees paid to the auditor. NON_TAX_FEE is a measure of the
willingness of the firm to use the auditor for non-audit services and has been used in, for
example, Omer et al. (2011) in a similar choice equation. We anticipate that the coefficient on
NON_TAX_FEE will be negative for both selection equations because greater reluctance to use
the auditor for other services will lead to higher likelihood of internally preparing or using
another external preparer (i.e., not using the auditor for tax services).21
4.2 SAMPLE SELECTION
We obtain data from four sources: (1) FIN 48 tax reserve (UTB) data from the IRS Large
Business & International Division (LB&I);22 (2) tax return preparer identity from IRS-LB&I; (3)
financial statement data from Compustat; and (4) auditor identity, audit fee, and tax fee data
from Audit Analytics. We determine the sample used for our analysis in several steps. First, we
obtain the intersection of LB&I and Compustat data during 2008-2009, consisting of 10,881
firm-years.23 Second, we obtain IRS data on tax preparer identity from the face of the U.S.
corporate income tax return, Form 1120, for a restricted set of 805 calendar year-end firms in the
S&P 1500 as of the end of 2008.24 We are able to identify the tax preparer by name and tax
identification number (PTIN) in the “Paid Preparer” signature area; if preparer information is
21
We confirm that this variable does not have a statistically significant coefficient when also included in the main
regression (equation (1)). We also included a variety of other variables in the first stage, including other
aggressiveness measures such as Cash ETR; other firm attributes such as the corporate governance G-Score, stock
exchange listing, and IRS audit probability; and auditor characteristics such as Big 4, each Big 4 firm individually,
and audit fees. The coefficient on none of these variables were statistically significant, either when included alone
or together. For parsimony, we only include NON_TAX_FEE in tabulated tests.
22
The UTB ending balance is also available to empirical researchers from a public data source (i.e., Compustat
mnemonic TXTUBEND) or can be retrieved from financial statement footnotes. As in Lisowsky, et al. (2011), we
use LB&I data because they undergo several layers of review, and are believed to be highly accurate. Confirming
this point, Lisowsky, et al. (2011) report that the correlation between the authors’ hand-collected UTB ending
balances and the amounts collected by LB&I (Compustat) is 0.96 (0.86).
23
Specifically, we obtain 5,539 (5,342) firm-years when intersecting LB&I and Compustat data for 2008 (2009).
24
Access to Form 1120 data is more restricted than access to the UTB data because the former is confidential while
the latter can be obtained from other sources. Therefore, IRS non-disclosure agreements allowed us to only merge
805 (785) calendar year end S&P 1500 firms (excluding REITS) in 2008 (2009).
22
missing, we assign the firm’s tax return as being internally prepared.25 We then merge the LB&I
FIN 48/Form 1120 data with Compustat and Audit Analytics over the same years, yielding a
final test sample of 1,533 firm-years with non-missing observations for all of our multivariate
analyses.
5. Main Results
5.1 DESCRIPTIVE STATISTICS
Table 3 Panel A presents descriptive statistics for the tax preparers in our sample. In our
sample of 1,533 firm-years, 690 (45 percent) of the tax returns are externally prepared, of which
545 (145) are prepared by Big 4 (non-Big 4) accounting firms. The remaining 843 (55 percent)
are internally prepared. The proportions of externally and internally prepared returns are
consistent from 2008 to 2009, and reflect our survey results that most tax compliance work is
done in-house. Also consistent with the survey results is how large our sample firms are that
internally prepare their tax returns, with mean (median) assets of $24.8 ($5.9) billion, compared
to firms with externally prepared tax returns, either by Big 4 firms (mean [median] assets are
$8.9 [$1.8] billion) or non-Big 4 firms (mean [median] assets are $5.3 [$1.1] billion). The
descriptive statistics suggest that firms internally preparing their tax returns are well-resourced,
and likely sophisticated enough to not exhibit demand for outside tax preparer expertise, even if
it is a Big 4 preparer.
[Insert Table 3 about here]
25
The IRS employee within LB&I supplying the tax preparer data confirmed that missing preparer information on
the Form 1120 is consistent with internally prepared tax returns rather than the external tax preparer omitting its
information for whatever reason. This fact reflects that a corporate officer must always sign the tax return, whether
an external preparer is used or not. Therefore, the internal preparation of the tax return is determined by the absence
of a paid preparer signature. See Section 2.2 for further details.
23
In Table 3 Panel B, we report the count of internally and externally prepared tax returns
by outside auditor (note that each firm in our sample is publicly traded and requires a financial
statement audit). The diagonal values (bolded) in this Panel highlight where the auditor is also
the tax preparer. For example, of the 415 company financial statements that Firm A audits during
our sample period, 275 self-prepare their tax returns, 71 hire Firm A to also prepare the tax
return (i.e., auditor-provided tax service), and the remaining 69 firms hire an external tax
preparer that is not Firm A.26
Examining the data from a different perspective, Firm A is the tax preparer for 128 firmyears in our sample, meaning that they prepare tax returns for 57 non-audit firm-years (128
minus 71), of which 17 are audited by Firm B, 25 by Firm C, 10 by Firm D, and 4 by Firm F.
The Panel suggests that auditor-provided tax services highlighted as the bolded diagonal values
are the most common combination for externally prepared tax returns, at least for Big 4 users.
However, the greatest incidence of firms prepares their returns internally without the help of an
external auditor or non-auditor.
In summary, Panel B reports that 312 firm-year tax returns are prepared externally by the
firm’s auditors (i.e., the sum of the bolded diagonal values); 378 firm-year tax returns are
prepared by an external preparer that is not the auditor (i.e., the sum of the non-bolded, offdiagonal values, not including internally prepared); and 843 firm-year tax returns are prepared
internally, of which 829 use Big 4 firms as their auditor and 14 use non-Big 4 firms as their
auditors. These results are interesting in that we observe that auditor-provided tax services are by
far the minority of prepared tax returns.
26
Due to confidentiality agreements, we are unable to name the paid preparer, so we use the more generic terms
“Firm A,” “Firm B,” and so forth to denote the external preparers (who may or may not also be the corporation’s
auditor). We are also unable to disclose counts that are less than three; we denote these values as “ND” in the tables.
Although it would be interesting to examine the time-series trends in tax preparer use, especially before and after the
passage of Sarbanes-Oxley (SOX), unfortunately our data are limited to analyzing cross-sectional relations only.
24
In Table 3 Panel C, we also report the industry composition for the three types of
preparers. In each industry, the internally prepared return is the most common. In five (three) of
the disclosed eight industries, the next most common combination is for the external preparer to
not be the auditor (to be the auditor). Again, the majority of firms do not employ their auditors to
provide tax services.
Table 4 Panel A reports descriptive statistics for our dependent variable, control
variables, and other audit and tax fee-related data. The mean (median) raw UTB ending balance
is $97.98 ($16.35) million, or 1.1% (0.60%) of total assets. The mean (median) total assets for
our sample is $17.4 ($3.4) billion, which is consistent with our sample composition of large,
publicly traded firms. On average, the firms are profitable (mean [median] Pretax ROA of 4.7%
[6.2%]), but with foreign operations in only a limited number of firms (mean [median] scaled
Foreign Income of 1.7% [0%]). A large share of firms—45%—report net operation loss
carryforwards, indicative of the economic crisis during the sample period. R&D activities are, as
typically the case, also concentrated in a limited number of firms (mean [median] R&D of 2.4%
[0%]), and debt levels show mean (median) values of 19.6% (17.3%) of total assets.
[Insert Table 4 about here]
With respect to tax fee data disclosed by firms who engage their auditor for tax work,
80.6% of firms in our sample report positive tax fees paid to their auditors. Combined with our
tax return preparer information, it appears that many firms do not have their tax return prepared
by their auditor, although a vast majority of firms hire their auditor for some type of tax work,
perhaps limited consulting as consistent by our survey results. Tax fees account for a mean
(median) of $532,000 ($117,000), or 8.4% (4.8%) of total fees paid to auditors.
25
Table 4 Panel B reports related values, where pertinent, for the preparer groupings we
examine in our study. Most notably, the raw and logged UTB ending balances are significantly
higher for the internal tax department than for either of the external (is or is not auditor) groups,
while the scaled values for the internal tax department and external non-auditor preparer are
higher than the external auditor preparer.
Finally, the correlation table in Panel C reports a significantly positive (negative)
association between the UTB ending balance and the internally (externally) prepared tax returns,
providing initial support for hypothesis H1. We also find significantly positive correlations
between the UTB and OTHER_PREP suggesting support for hypothesis H2.
5.2 MULTIVARIATE RESULTS: HYPOTHESIS H1 AND H2
Table 5 presents the results of estimating Equation (1). We find significant results
supporting hypothesis H1 that tax returns filed without the assistance of an external preparer are
more aggressive than tax returns filed with the help of an external preparer, on average. In
particular, the coefficient on INTERNAL_PREP is positive and significant in all specifications.
In this setting, the OLS provides the most conservative estimates, in terms of the smallest
coefficients and the lowest t-values. The estimate for the coefficient on INTERNAL_PREP using
OLS is 0.29. This suggests that, relative to tax returns prepared by auditors, tax returns that are
internally prepared have uncertain tax benefits that are 29% higher. Second, the coefficient
estimate on INTERNAL_PREP using a Tobit estimation approach, which explicitly models the
truncation of the UTB balance at zero, is 0.35. Third, when the estimates explicitly control for
the endogeneity of the firm’s selection process using a multinomial treatment effect
specification, the coefficient on INTERNAL_PREP is 0.67. Overall, these tests provide evidence
26
that not using an external preparer, relative to use the auditor as the tax preparer, is associated
with an increase in UTB balance from between 29% and 67%.
[Insert Table 5 about here]
Table 5 concurrently presents the results of estimating whether an external non-auditor
preparer, rather than the auditor, is related to greater tax aggressiveness. We find evidence
supportive of H2 that tax returns filed with the assistance of a non-auditor external preparer will,
on average, be more aggressive than tax returns filed with the assistance of other external
preparers. Using both OLS and Tobit specifications, our tests yield significant and positive
coefficients on OTHER_PREP, suggesting that outside, non-auditor preparer-experts are related
to tax aggressiveness. The coefficient estimates on OTHER_PREP are smaller than those on
INTERNAL_PREP, and range from 0.22 using OLS to 0.55 using the multinomial treatment
effects model.27
The control variables, where significant, are in the expected directions; firm size
(LOG_ASSETS), the share of foreign income (FOR_INCOME), existence of net operating losses
(NOL), and research and development activities (R&D) are all increasing in the tax reserve,
while profitability and leverage are not significantly related to the FIN 48 tax reserve. In the
selection models, the non-tax fee ratio has a negative coefficient, statistically significant at the
10% level using a two-tailed test. This result confirms that firms less willing to use their auditor
for services other than tax and audit are more likely to choose a non-auditor alternative preparer,
either internally prepare or use another firm. Size and foreign income are also positive related to
internal preparation, while R&D is negatively related. None of the equation (1) control variables
are related to the choice to use a a tax preparer other than the audit firm for tax preparation.
27
A test of differences between the coefficients on INTERNAL_PREP and OTHER_PREP reveals they are not
significantly different (significance levels range from 21% to 51% using two-tailed tests).
27
The results on INTERNAL_PREP and OTHER_PREP are the first that provide empirical
evidence supportive of differences in tax aggressiveness across preparer type. Until now,
researchers have been unable to definitively observe the preparer type to examine whether tax
returns filed with the help of external preparers are more or less aggressive that those that are
internally prepared. By using tax return data on preparer type, we are, for example, able to
provide clear support that, as our survey results also suggest, internally prepared tax returns are
indicative of tax departments that are linked to more aggressive tax positions.
6. Comparing Tax Return Data to 10-K Inferences on Tax Preparer Type
Because tax preparer type is only observable from confidential tax returns, we examine
the extent to which publicly available information can aid interested parties in inferring preparer
type reported on the tax return. Specifically, we examine whether disclosures of auditor-provided
tax service fee data in financial statements can provide a window onto tax preparers, and thus,
tax compliance activities of the corporation.
Since the Sarbanes-Oxley Act of 2002, firms are required to disclose in their financial
statements audit and non-audit fees paid to their auditor. Although firms must disclose tax fees
paid to their auditors, they are not required to disclose fees paid to non-auditors, or for that
matter, whether the tax return is prepared in-house and what resources are used to do so. Because
of this limitation, outside researchers can only observe the dichotomous outcome of whether the
firm’s auditor does or does not receive a fee for tax services. This information is then used to
analyze the propensity to engage, retain, or dismiss the auditor as the tax service provider (e.g.,
Lassila, et al. 2010; Omer, et al. 2006; and McGuire, et al. 2011). Of course, information on
whether the provider for tax services other than the auditor is either a non-auditor or the internal
28
tax department—and whether “services” have to do with compliance or planning—have until
now been unobservable in a corporate setting.28 Because we can observe the identity of the tax
preparer, we design an archival experiment that compares how the accuracy of preparer type
classification using tax fees versus tax returns might affect our established inferences about tax
aggressiveness.
To begin, we re-specify Equation (1) by creating one variable, NONAUD_PREP_TR, that
combines the INTERNAL_PREP and OTHER_PREP firms into one proxy for non-audit tax
service provider that we can observe directly from tax returns. This re-specification of the
preparer variable of interest exactly mirrors the dichotomous specification using the tax fee data
as to whether the firm uses (or does not use) the auditor as the tax service provider. As in all of
our previous specifications, we leave the auditor tax service provider as the reference category
and continue to employ the same firm characteristics, year, and industry control variables.
Table 8 Panel A reports our replication of previous results using the modified Equation
(1). As expected, we continue to obtain a significantly positive sign on NON_AUD_PREP_TR,
indicating that tax service providers that are not the auditor (i.e., either external non-auditors or
internal tax departments) are linked to greater tax aggressiveness. These results in Panel A based
on tax return classification of preparer type serve as our benchmark case.
Using a similar approach when specifying our variable of interest, we then designate the
variable NON_AUD_PREP_10K as capturing whether the auditor does not provide tax services
to the client; this variable equals one if tax fees disclosed in the financial statements are equal to
zero; 0 otherwise. As reported in Panel B as “Case 1,” we re-estimate our modified Equation (1),
28
As noted above, Neuman, et al. (2011) are able to observe the choice of non-auditor tax service provider for notfor-profit organizations because the IRS Form 990 is publicly available. However, data access limitations to
corporate tax returns preclude them from evaluating the for-profit sector as we do here.
29
but find no significant relation between this financial statement-based variable and tax
aggressiveness.
[Insert Table 8 about here]
Because using a non-zero tax fee might unnecessarily designate firms that use their
auditor very little for tax services as similar to those that use their auditors for large amounts of
tax services, we designate several cut-off values on the size of the ratio of tax fees to total fees
paid to the auditor (or “tax fee ratio”). Specifically, we designate a new variable in “Case 2” of
Panel B as NON_AUDIT_PREP_10K_1% equal to one if the firm discloses a tax fee ratio of less
than or equal to 1%, zero otherwise. We re-estimate the modified Equation (1) and again obtain
no significant results.
In successive specifications, we continue to increase the tax fee ratio cut-off at less than
or equal to 5%, 10%, 25%, and 50% when defining our non-auditor preparer variables of
interest. In every case, the results are either not significant (10% and 50% designations) or
significantly negative (5% and 25% designations). These results are in stark contrast to those
obtained from using tax preparer data directly observable from tax returns. We repeat these
specifications at levels less than or equal to 2%, 3%, 4%, 15%, 20%, 30%, 35%, 40%, 45%, and
55%, but to no avail (untabulated). Not once do the results using the presence or magnitude of
the tax fee yield inferences that are consistent with our original results using tax preparer
designations directly from the tax return.
In Table 8 Panel C, we report error rates inherent in using public tax fee data in our
sample to classify the preparer type as either non-auditor (Non_Aud_Prep) or auditor
(Aud_Prep). Recall that our data report that 1,221 firm-years do not use their auditor to prepare
the tax return, while 312 do. Using these figures as benchmarks, we report for each tax fee ratio
30
cut-off the number and percent of firms that are correctly and incorrectly classified by preparer
type. For example, using NON_AUD_PREP_10K_1% derived from the tax fee data classifies
428 firm-years as having a non-auditor preparer, while 1,105 firm-years are classified as having
an auditor preparer. Of the 428, 412 indeed have a non-auditor tax preparer as reported on the tax
return. Of the 1,105, only 296 indeed have an auditor-preparer as reported on the tax return.
These figures illustrate that of the 1,533 total firm-year observations in our sample, only 708
(46.2%) are correctly classified into the two auditor and non-auditor preparer groups.
In particular, using the tax fees, sixteen firm-years are incorrectly classified as using nonauditor preparers, since per the tax return, they are in fact using auditor preparers; and an
astonishing 809 firm-years are classified as having an auditor preparer, when in fact they use
non-auditor preparers as confirmed by the tax return. These figures illustrate that of the 1,533
firm-year observations, 825 (53.8%) are incorrectly classified. To be sure, the classification
accuracy improves as the cut-off values increase (because increasing sample sizes ensnare
enough correct tax preparer types). However, the fee data continue to incorrectly classify
between 20 and 62 percent of firm-year observations.
Overall, our archival experiment implies that using tax fee data to classify the tax
preparer as the auditor (or not the auditor) is not only materially inaccurate, it is inaccurate
enough to yield starkly different inferences from tests that are otherwise designed exactly the
same way. As a corollary, we demonstrate that using tax return data is materially beneficial in
our setting when drawing stable conclusions about links between preparer type, expertise, and
tax aggressiveness. Admittedly, the original aim of this exercise was to provide guidance to
researchers when using tax fees to help them infer tax preparer type, and thus compliance
activities of the firm. However, the implication that tax fees are not informative of tax
31
compliance activities, namely the preparation of the tax return, suggests that researchers must
remain cautious when drawing inferences on the type of services supplied by tax providers when
tax fee data are used as the basis for such an analysis.
7. Conclusion
Our study examines companuies’ choices regarding advisors for their tax compliance
work and how these decisions relate to companies’ tax positions. We survey 170 tax executives
from the Tax Executives Institute (TEI) to determine firms’ strategies for tax compliance. The
survey reveals that 20% of companies do not use any external firm for compliance work, 16%
use their auditors only, 52% use another type of consultant (either an accounting firm that is not
their auditor, or another type of provider), and 12% use both their auditor and other providers.
The survey also reveals that external providers perform only 30% of compliance work, and
external providers are used for a similar percentage of planning work.
We then test whether the choice of tax preparer type—whether internal, external auditor,
or external non-auditor—is related to the aggressiveness of positions claimed on a firm’s tax
return, as suggested by theory in Phillips and Sansing (1998). Using confidential data obtained
from the IRS on who signs a firm’s tax return, we find a positive and significant relation between
tax aggressiveness and not using an external tax preparer or using an external non-auditor
preparer, compared to using an auditor preparer. Finally, we find that tax fees paid to a firm’s
auditors as publicly disclosed in a firm’s securities filings are unable to reliably measure the tax
compliance and preparation activities of the firm, namely the identity of the tax preparer. In
particular, we estimate that the tax fee data incorrectly classify between 20 and 62 percent of
firms into tax preparer types that do not match those reported on a firm’s tax return. This finding
32
should caution other researchers when attempting to use the tax fee data to infer the tax
compliance attributes of firms.
Our findings are important given the paucity of archival research on tax preparers
generally. We make a significant contribution to the literature by better understanding this
important part of the U.S. tax system and companies’ decision making. Specifically, we are the
first to document the identity, distribution, and attributes of tax return preparers for a large
sample of U.S. companies. This analysis helps us evaluate the compliance responsibilities and
implications surrounding business transactions, and examine the extent to which companies
employ internal tax specialists and/or hire external advisory firms. We are also the first study to
provide empirical archival evidence in support of theoretical predictions in Phillips and Sansing
(1998) concerning the relation between tax preparers and tax aggressiveness. Documenting the
effect of the tax preparer in corporate tax compliance decisions with respect to tax
aggressiveness is of interest to tax authorities, tax service providers, corporate tax executives,
and tax researchers, as the archival literature remains largely silent on the role of the corporate
tax preparer and its effect on compliance.
33
APPENDIX
Variable Definitions
Variable
Dependent Variable
Log_UTB_EB
Definition
= The natural log of (1 + UTB), where UTB = Ending balance (in
$millions) of the FIN 48 unrecognized tax benefit (UTB). Source:
IRS-LB&I.
Tax Preparer and Expert Measures
INTERNAL_PREP
= 1 if the firm’s Form 1120 tax return is not signed by an external
preparer; 0 otherwise. Source: IRS-LB&I.
OTHER_PREP
= 1 if the firm’s Form 1120 tax return is signed by an external preparer
that is not the firm’s external auditor; 0 otherwise. Source: IRSLB&I, Audit Analytics.
Control Variables
LOG_ASSETS
PRETAX_ROA
FOR_INCOME
NOL
R&D
LEVERAGE
NON-TAX FEE RATIO
= The natural log of Total Assets (AT). Source: Compustat.
= The ratio of Pretax Income (PI) / Total Assets (AT). Source:
Compustat.
= The ratio of Foreign Pretax Income (PIFO) / Total Assets (AT).
Source: Compustat.
= Indicator variable equal to one if Tax Loss Carry Forward (TLCF) is
non-zero; 0 otherwise. Source: Compustat.
= The ratio of R&D Expense (XRD) / Total Assets (AT). Source:
Compustat.
= The ratio of Long-Term Debt (DLTT) / Total Assets (AT). Source:
Compustat.
= The ratio of fees paid to the auditor for services other than tax or
audit, divided by fees paid to the auditor for services other than tax.
Source: Audit Analytics.
34
REFERENCES
ABBOTT, L., S. PARKER, G. PETERS, AND K. RAGHUNANDAN. “An Empirical Investigation of
Audit Fees, Nonaudit Fees, and Audit Committees.” Contemporary Accounting Research 20
(2) (2003): 215-34.
ARMSTRONG, C., J. BLOUIN, AND D. LARCKER. “The Incentives for Tax Planning.” Journal of
Accounting and Economics (2011): forthcoming.
ASHBAUGH, H., R. LAFOND, AND B. MAYHEW. “Do Non-Audit Services Compromise Auditor
Independence? Further Evidence.” The Accounting Review 78 (3) (2003): 611-39.
CHRISTIAN, C., S. GUPTA, AND S. LIN. “Determinants of Tax Preparer Usage: Evidence from
Panel Data.” National Tax Journal 46 (1993): 487-503.
CHUNG, H., AND KALLAPUR, S. “Client Importance, Nonaudit Services, and Abnormal Accruals.”
The Accounting Review 78 (2003): 931-55.
COOK, K., G. RYAN HUSTON, AND T. OMER.”Earnings Management through Effective Tax Rates:
The Effects of Tax-Planning Investment and the Sarbanes-Oxley Act of 2002.”
Contemporary Accounting Research 25 (2): 447-71.
CRIPE, B., AND B. MCALLISTER. “Determinants of Audit/Tax Separation Decisions.” American
Journal of Business 24 (2009): 47-56.
DAVIS, L., D. RICCHIUTE, AND G. TROMPETER. “Audit Effort, Audit Fees, and the Provision of
Nonaudit Services to Audit Clients.” The Accounting Review 68 (1) (1993): 135-50.
DEB, P., AND P.K. TRIVEDI, “Specification and Simulated Likelihood Estimation of a Nonnormal Treatment Outcome Model with Selection: Application to Health Care
Utilization,” Econometrics Journal 9 (2) (2006): 307-331.
DEFOND, M., K. RAGHUNANDAN, AND K. SUBRAMANYAM. “Do Non-Audit Service Fees Impair
Auditor Independence? Evidence from Going Concern Audit Opinions.” Journal of
Accounting Research 40 (4) (2002): 1247-74.
DESAI, M., AND D. DHARMAPALA. “Corporate Tax Avoidance and High-Powered Incentives.”
Journal of Financial Economics 79 (2006): 145-79.
DYRENG, S., M. HANLON, AND E. MAYDEW. “Long-Run Corporate Tax Avoidance.” The
Accounting Review 83 (2008): 61-82. FINANCIAL ACCOUNTING STANDARDS BOARD (FASB). Statement of Financial Accounting
Standards (SFAS) No. 109, Accounting for Income Taxes. Norwalk, CT (1992).
35
FINANCIAL ACCOUNTING STANDARDS BOARD (FASB). Financial Interpretation No. 48 (FIN
48), Accounting for Uncertainty in Income Taxes—an Interpretation of FASB Statement
No. 109. Norwalk, CT (2006).
FRANCIS, J., AND B. KE. “Disclosure of Fees Paid to Auditors and the Market Valuation of
Earnings Surprises.” Review of Accounting Studies 11 (2006): 495-523.
FRANK, M. M., L. LYNCH, AND S. REGO. “Tax Reporting Aggressiveness and Its Relation to
Aggressive Financial Reporting.” The Accounting Review 84 (2009): 467-96.
FRANKEL, R., M. JOHNSON, AND K. NELSON. “The Relation between Auditors’ Fees for Nonaudit
Services and Earnings Management.” The Accounting Review 77 (Supplement) (2002): 71105.
GLEASON, C., AND L. MILLS. “Do Auditor-Provided Tax Services Improve the Estimate of
Tax Reserves?” Contemporary Accounting Research, forthcoming 2011.
GRAHAM, J., M. HANLON, AND T. SHEVLIN. “Inside the Corporate Tax Department: Insights on
Corporate Decision Making and Tax Aggressiveness.” Working paper, 2011.
GRAHAM, J., AND C. HARVEY. “The Theory and Practice of Corporate Finance: Evidence from
the Field.” Journal of Financial Economics 60 (2001): 187-243.
GUPTA, S., R. LAUX, AND D. LYNCH. “Do Firms Use Tax Cushion Reversals to Meet Earnings
Targets?” Working paper, 2011.
HANLON, M., AND S. HEITZMAN. “A Review of Tax Research.” Journal of Accounting and
Economics 50 (2010): 127-178.
HITE, P. AND G. MCGILL. “An Examination of Taxpayer Preference for Aggressive Tax Advice.”
National Tax Journal 45 (4) (1992): 389-403.
KINNEY, W., JR., Z.-V. PALMROSE, AND S. SCHOLZ. “Auditor Independence, Non-Audit Services,
and Restatements: Was the U.S. Government Right?” Journal of Accounting Research 42
(June) (2004): 561–88.
LARCKER, D., AND S. RICHARDSON. “Fees Paid to Audit Firms, Accrual Choices, and Corporate
Governance.” Journal of Accounting Research 42 (3) (2004): 625-58.
LASSILA, D., T. OMER, M. SHELLEY, AND L. SMITH. “Do Complexity, Governance, and Auditor
Independence Influence Whether Firms Retain Their Auditors for Tax Services?” Journal of
the American Taxation Association 32 (Spring) (2010): 1–23.
LIM, C-Y., AND H-T. TAN. “Non-audit Service Fees and Audit Quality: The Impact of Auditor
Specialization.” Journal of Accounting Research 46 (1) (2008): 199-246.
36
LISOWSKY, P., L. ROBINSON, AND A. SCHMIDT. “Do Publicly Disclosed Tax Reserves Tell Us
About Privately Disclosed Tax Shelter Activity?” Working paper, 2011.
LONG, J., AND S. CAUDHILL. “The Usage and Benefit of Paid Tax Return Preparation.” National
Tax Journal 40 (1987): 35-46.
MAYDEW, E. AND D.SHACKELFORD. “The Changing Role of Auditors in Corporate Tax
Planning.” In Taxing Corporate Income in the 21st Century, edited by Auerbach, A., J.
Hines, and J. Slemrod. Cambridge, U.K.: Cambridge University Press. (2007): 307-34.
MCGUIRE, S., T. OMER, AND D. WANG. “Tax Avoidance: Does Tax-Specific Industry
Expertise Make a Difference?” The Accounting Review (2011): forthcoming.
MILLS, L. “Book-Tax Differences and Internal Revenue Service Adjustments.” Journal of
Accounting Research 36 (1998): 343-56.
MILLS, L., M. ERICKSON, AND E. MAYDEW. “Investments in Tax Planning.” Journal of the
American Taxation Association 20 (1) (1998): 1-20.
MILLS, L., L. ROBINSON, AND R. SANSING. “FIN 48 and Tax Compliance.” The Accounting
Review 85 (5) (2010): 1721-42.
NEUMAN, S., T. OMER, AND A. THOMPSON. “Competition for Non-Audit Services among
Professional Services Firms: Determinants and Consequences of Tax Service Provider
Choice in the Not-for-Profit Sector.” Working Paper, 2011.
OMER, T., J. BEDARD, AND D. FALSETTA. “Auditor-Provided Tax Services: The Effects of a
Changing Regulatory Environment.” The Accounting Review 81 (October) (2006): 10951117.
PETERSEN, M. “Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches.”
Review of Financial Studies 22 (2009): 435-80.
PHILLIPS, J., AND R. SANSING. “Contingent Fees and Tax Compliance.” The Accounting Review
73 (1) (1998): 1-18.
PITTMAN, J., AND S. FORTIN. “Auditor Choice and the Cost of Debt Capital for Newly Public
Firms.” Journal of Accounting and Economics 37 (2004): 113-36.
REICHELT, K., AND D. WANG. “National and Office-Specific Measures of Auditor Industry
Expertise and Effects on Audit Quality.” Journal of Accounting Research 48 (3) (2010): 64786.
SLEMROD, J., AND V. VENKATESCH. “The Income Tax Compliance Cost of Large and Midsize
Businesses.” A Report to the IRS LMSB Division (2002).
37
WALLACE, R. AND C. MELLOR. “Nonresponse Bias in Mail Accounting Surveys: A Pedagogical
Note.” British Accounting Review 20 (1988): 131-39.
ZAMAN, M., M. HUDAI, AND R. HANIFFA. “Corporate Governance Quality, Audit Fees and NonAudit Services Fees.” Journal of Business Finance and Accounting 58 (2011): 165-197.
38
TABLE 1
Descriptive Statistics for Survey Respondents and Comparison to TEI Membership Population
TEI
Survey
Total Assets
membership respondents
<$10M
1%
1%
$10M-$500M
24%
22%
$500M-$1 Billion
17%
15%
$1 Billion- $5 Billion
32%
29%
$5 Billion-$50 Billion
21%
26%
$50 Billion-$500 Billion
4%
5%
> $500 Billion
1%
2%
100%
100%
TEI
Survey
Total Revenues
membership respondents
< $500 Million
24%
25%
$500 Million - $1 Billion
17%
10%
$1 Billion - $5 Billion
37%
34%
$5 Billion - $10 Billion
10%
10%
$10 Billion - $50 Billion
11%
17%
Greater than $50 Billion
2%
4%
100%
100%
Tax Budgets
< $250,000
$250,000-$500,000
$500,000- $1 Million
$1 Million - $5 Million
>$5 Million
TEI
Survey
Industry
membership respondents
Mining and resources
6%
2%
Utilities
4%
0%
Construction
1%
1%
Manufacturing
36%
49%
Wholesale and retail
11%
8%
Transportation
3%
2%
Information and telecomm
5%
8%
Financial and Insurance
10%
6%
Real estate
2%
1%
Professional services & Education
6%
6%
Health
6%
2%
Media and Entertainment
3%
1%
Food and Accommodation
7%
3%
Other
0%
7%
None of the above
0%
4%
100%
100%
TEI
Survey
membership respondents
15%
10%
21%
18%
26%
16%
32%
40%
5%
16%
100%
100%
These data are drawn from the Tax Executives Institute’s (TEI) mandatory survey of its membership in 2005 and from
our Tax Survey conducted in 2010.
TABLE 2
Survey Statistics of Compliance Outsourcing
Compliance Work Outsourced To:
N
Not outsourced
34
Average
Average
Average
Percentage
Percentage
Number of
Of Compliance Of Planning Work Internal Tax Total Assets
Percentage Work Outsourced
Outsourced
Personnel
($ Billion)
20%
0%
16%
13
37.9
Auditor only
27
16%
40%
31%
5
2.2
Non-auditor accounting firm
only
70
41%
39%
35%
14
29.5
Non-accounting firm only
9
5%
22%
37%
18
1.8
Auditor and non-auditor
accounting firm
11
6%
27%
24%
16
7.4
Auditor and non-accounting
firm
3
2%
52%
17%
9
2.8
Non-auditor accounting firm
and non-accounting firm
11
6%
40%
38%
21
15.2
All three types
5
3%
37%
28%
12
7.5
100%
30%
30%
13
22.2
Totals
170
These data are drawn from our Tax Survey. Respondents are classified according to their responses to the question "If your tax
compliance is outsourced, to whom is it outsourced?" Tabulations in columns 4, 5, and 7 are based on the mid-point of categorical
responses given in 11 categories; and column 6 are based on the mid-point of categorical responses given in 5 categories. The highest,
open-ended category is not considered. Column 4 is based on responses to the question, "What percentage of overall tax compliance
work (both transfer pricing and non-transfer pricing) is outsourced?" Column 5 is based on responses to the question, "What percentage
of overall tax planning work (both transfer pricing and non-transfer pricing work) is outsourced?" Column 6 is based on responses to
the question, "How many full time tax personnel are employed company wide?" Column 7 is based on responses to the question, "My
company's asset size as of the most recent annual fiscal period is." While only 120 responses to this last question were received, the
distribution across classifications is almost identical to that shown in column 2 (for example, 25 of 120 response, or 21%, are in category
1-not outsourced; 18 of 120 responses, or 15%, are in category 2-auditor only; etc.).
TABLE 3
Descriptive Statistics on Firms by Tax Return Preparer Type and Year
Panel A. Distribution of Sample and Assets by Tax Return Preparer and Year
Big 4
Firm A
Firm B
Firm C
Firm D
Total Big 4
Non-Big 4
Firm E
Firm F
Other
Total Non-Big 4
2008
mean median
Assets Assets
n
% of
Total
69
73
89
51
282
8.6% 8,366
9.1% 8,462
11.1% 10,663
6.3% 5,811
35.1% 8,654
0.9%
2.0%
7.1%
10.0%
7
16
57
80
2009
mean median
Assets Assets
n
% of
Total
1,754
2,018
1,417
1,439
1,670
59
63
90
51
263
8.1% 15,106
8.6% 5,712
12.3% 9,581
7.0% 6,703
36.1% 9,336
2,958
3,527
3,832
3,695
1,420
629
1,679
1,138
7
11
47
65
n
2,662
1,920
1,661
1,636
1,955
128
136
179
102
545
2,544
4,031
8,808
7,325
631
833
1,857
1,352
14
27
104
145
55.0% 25,680
6,143
843
1.0%
1.5%
6.4%
8.9%
Internal
Tax Dept.
442
55.0% 24,158
5,544
401
TOTAL
804 100.0% 16,684
3,215
729 100.0% 18,147
Total By Preparer
% of
mean median
Total Assets Assets
8.3% 11,473
8.9% 7,188
11.7% 10,119
6.7% 6,257
35.6% 8,983
0.9%
1.8%
6.8%
9.5%
2,751
3,732
6,081
5,322
1,067
702
1,761
1,146
55.0% 24,882
5,884
3,601 1,533 100.0% 17,379
3,380
Asset values are in $Millions.
Panel B. Count of Sample Firms by Tax Return Preparer Type and Auditor
Auditor Firm
Tax Preparer
A
B
C
D
E
F Other TOTAL
Firm A
71
17
25
10
ND
4
ND
128
Firm B
21
54
27
19
ND
ND
10
136
Firm C
19
17 103
32
4
ND
ND
179
Firm D
10
11
16
56
ND
6
ND
102
Firm E
0
0
5
0
7
ND
ND
14
Firm F
5
3
7
7
ND
5
ND
27
Other
14
23
24
18
6
3
16
104
Internal
275
190 211
153
4
7
3
843
TOTAL
415
315 418
295
25
30
35
1,533
Bolded numbers on the Diagonal represent that the outside Tax Preparer is also the Auditor
ND = Not disclosed to do small sample sizes.
Summary
External Tax Preparer:
Is the Auditor
Is Not the Auditor
Total External
Internal Preparer Uses:
Big 4 Auditor
Non-Big 4 Auditor
Total Internal
Total Sample
Totals
312
378
690
829
14
843
1,533
2,346
1,988
1,585
1,506
1,834
TABLE 3 (continued)
Panel C. Distribution of Sample Firms by Tax Return Preparer Type and Industry
Preparer Type
Internal Preparer
External Preparer is Auditor
External Preparer is not Auditor
Total
SIC0
ND
ND
ND
SIC1
58
11
25
SIC2
166
31
48
SIC3
217
74
104
SIC4
136
40
38
SIC5
37
29
30
SIC6
122
76
68
SIC7
61
34
50
SIC8
39
16
14
SIC9
ND
ND
ND
ND
94
245
395
214
96
266
145
69
ND
ND Total
9
Total
836
311
377
9
1,533
ND Total
9
Total
836
311
69
35
101
52
26
377
40
73
73
35
117
9
1,533
Panel D. Distribution of Sample Firms by Tax Return Preparer Type, Firm, and Industry
Preparer Type
Internal Preparer
External Preparer is Auditor
Firm A
Firm B
Firm C
Firm D
Non-Big4
External Preparer is not Auditor
Firm A
Firm B
Firm C
Firm D
Non-Big4
Total
SIC0
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
SIC1
58
11
5
ND
ND
ND
ND
25
ND
ND
7
7
5
SIC2
166
31
6
10
9
3
3
48
6
7
15
12
8
SIC3
217
74
20
ND
33
11
ND
104
ND
31
21
ND
31
SIC4
136
40
11
ND
13
10
ND
38
7
12
ND
ND
15
SIC5
37
29
ND
ND
15
6
4
30
6
ND
10
ND
8
SIC6
122
76
18
6
18
18
16
68
6
14
11
9
28
SIC7
61
34
4
15
8
4
3
50
10
9
9
7
15
SIC8
39
16
5
4
5
ND
ND
14
5
ND
ND
ND
7
SIC9
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
94
245
395
214
96
266
145
69
ND
ND = Not disclosed to do small sample sizes.
SIC 0 = Agriculture
SIC 1 = Construction
SIC 2 = Chemicals
SIC 3 = Manufacturing
SIC 4 = Transportation
SIC 5 = Retail
SIC 6 = Financial
SIC 7 = Business Services
SIC 8 = Health
SIC 9 = Diversified/Other
TABLE 4
Descriptive Statistics for Regression Variables
Panel A. Summary Statistics
Variable
Dependent Variable
UTB Ending Balance ($M)
UTB EB / Total Assets
Log_UTB_EB
n
Median
Full Sample
Mean
1,533
1,533
1,533
16.347
0.006
2.853
97.982
0.011
2.912
199.177
0.017
1.953
Control Variables
Total Assets ($M)
LOG_ASSETS
PRETAX_ROA
FOR_INCOME
NOL
R&D
LEVERAGE
1,533
1,533
1,533
1,533
1,533
1,533
1,533
3,379.889
8.126
0.062
0.000
0.000
0.000
0.173
17,379.490
8.217
0.047
0.017
0.450
0.024
0.196
44,622.450
1.725
0.143
0.042
0.498
0.055
0.170
Tax Fee Data
Tax Non-Audit Fees > 0 (0/1)
Tax Non-Audit Fees / Total Assets
Tax Fee Ratio (Tax Fees / Total Fees)
Tax Non Audit Fees ($M)
Log(1+Tax Fees)
1,533
1,533
1,533
1,533
1,533
1.000
0.0000
0.048
0.117
0.111
0.806
0.0001
0.084
0.532
0.298
0.396
0.0002
0.096
1.129
0.431
1,533
0.050
0.082
0.098
1,533
1,533
1,533
1,533
0.000
0.000
0.000
1.000
0.450
0.204
0.247
0.550
0.498
0.403
0.431
0.498
Non-Tax Fee Ratio (Non-Audit Fees - Tax
Fees)/(Total Fees - Tax Fees)
Categories
Preparer Type
External Preparer
is Auditor (AUD_PREP)
is not Auditor (OTHER_PREP)
Internal Preparer (INTERNAL_PREP)
STD
Notes: UTB EB is the total value (ending balance) of uncertain tax benefits recorded by the IRS Large
Business & International Division. LOG_ASSETS is the natural logarithm of total assets. PRETAX_ROA is
pretax income deflated by total assets. FOR_INCOME is the ratio of the company's disclosure of foreign
pretax income, deflated by total assets. NOL is an indicator variable if the firm has a loss carryforward. R&D
is research and development expenses, deflated by total assets. LEVERAGE is the ratio of long-term debt to
total assets. Tax Non-Audit Fees and Total Fees are the fees disclosed by the company that are paid to the
auditor for tax services and all services, respectively. AUD_PREP is an indicator variable that takes on the
value 1 if the tax return is signed by the company's audit firm; NONAUD_PREP is an indicator variable that
takes on the value 1 if the tax return is signed by a preparer who is not the firm's auditor; and
INTERNAL_PREP is an indicator variable that takes on the value 1 if the firm's tax return is not signed by an
external preparer. AUD_PREP, OTHER_PREP and INTERNAL_PREP are mutually exclusive and jointly
exhaustive.
TABLE 4 (continued)
Panel B. Summary Statistics by Tax Preparer and Expert Type
Variable
Dependent Variable
UTB Ending Balance
($M)
UTB EB /
Total Assets
Log_UTB_EB
Control Variables
Total Assets ($M)
LOG_ASSETS
PRETAX_ROA
FOR_INCOME
NOL
R&D
LEVERAGE
External Tax Preparer:
Not the Auditor
Auditor (AUD_PREP = 1)
(OTHER_PREP = 1)
n
Median Mean STD
n Median Mean STD
Internal Tax Department
(INTERNAL_PREP = 1)
n Median Mean STD
312
5.79
46.86 127.66
378
6.65
46.77 134.40
843
37.30
312
0.00
0.01
0.02
378
0.01
0.01
0.02
843
0.01
0.01
0.01
312
1.19
2.10
1.79
378
2.03
2.23
1.70
843
3.65
3.52
1.91
312
312
312
312
312
312
312
1,956
7.58
0.06
0.00
0.00
0.00
0.13
9,472 28,610
7.71
1.58
0.04
0.15
0.01
0.03
0.45
0.50
0.02
0.06
0.17
0.17
378
378
378
378
378
378
378
1,339
7.20
0.04
0.00
0.00
0.00
0.11
7,175 18,697
7.42
1.61
0.02
0.19
0.01
0.04
0.49
0.50
0.04
0.07
0.16
0.18
843
843
843
843
843
843
843
5,884
8.68
0.07
0.00
0.00
0.00
0.21
5.
6.
Panel C. Correlation Table
1.
1. Log_UTB_EB
-0.21
2. AUD_PREP
-0.20
3. OTHER_PREP
4. INTERNAL_PREP 0.34
5. LOG_ASSETS
0.63
6. PRETAX_ROA
0.06
7. FOR_INCOME
0.28
8. NOL
0.07
9. R&D
0.04
10. LEVERAGE
0.18
2.
-0.29
-0.56
-0.15
-0.01
-0.09
0.00
0.00
-0.08
3.
-0.63
-0.26
-0.13
-0.11
0.05
0.13
-0.12
4.
0.35
0.12 0.11
0.16 0.14
-0.04 -0.19
-0.12 -0.21
0.16 0.18
7.
0.38
-0.07 0.09
-0.13 0.11
-0.02 -0.09
8.
9.
0.13
0.09 -0.16
This Panel reports Pearson correlation coefficients. Correlations in excess of 0.04, in absolute value, are generally
statistically significant at a 5% level.
139.87 232.75
24,882 55,116
8.76
1.63
0.06
0.11
0.02
0.04
0.43
0.50
0.02
0.04
0.22
0.16
TABLE 5
Multivariate Regression Results on the Link between Tax Aggressiveness and Tax Preparer Type
Multinomial Self-Selection Model
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Coefficient
(t-statistic)
INERNAL_PREP
+
0.287 **
(2.49)
OTHER_PREP
+
0.216 *
(1.70)
LOG_ASSETS
+
0.917 ***
(27.82)
PRETAX_ROA
+
-0.463
(-1.39)
FOR_INCOME
+
3.476 ***
(2.72)
NOL
+
0.276 ***
(3.23)
R&D
+
2.501 **
(2.13)
LEVERAGE
+
0.084
(0.25)
NON-TAX FEE RATIO
Constant
Year Control
Industry Controls
λ_INTERNAL_PREP
+
-4.986 ***
(-10.55)
YES
YES
Tobit
Coefficient
(t-statistic)
0.353 ***
(2.73)
0.274 *
(1.89)
0.963 ***
(27.13)
-0.536
(-1.49)
3.481 **
(2.55)
0.301 ***
(3.25)
2.542 *
(1.93)
0.113
(0.32)
Stage 1
INTERNAL_
OTHER_
PREP
PREP
Coefficient Coefficient
(z-statistic) (z-statistic)
-5.448 ***
(-10.72)
YES
YES
0.663 ***
(7.99)
-0.335
(-0.43)
5.147 *
(1.77)
-0.257
(-1.11)
-5.055 *
(-1.69)
0.800
(1.14)
-1.635 *
(-1.69)
-2.581 *
(-1.68)
YES
YES
1,533
843
-0.129
(-1.32)
-0.881
(-1.13)
-0.182
(-0.06)
0.083
(0.34)
2.512
(1.02)
-0.078
(-0.10)
-2.067 *
(-1.88)
0.734
(0.45)
YES
YES
λ_OTHER_PREP
Observations
Adjusted R2
Model F-Stat
Model Chi-Square
Pseudo-R2
Log Pseudo-Likelihood
Tests between Coefficients:
INTERNAL_PREP vs.
EXT_ISNT_AUD
1,533
60.1%
84.92 ***
378
Stage 2
Coefficient
(z-statistic)
0.667 ***
(4.31)
0.550 **
(2.21)
0.897 ***
(22.11)
-0.414
(-1.25)
3.265 **
(2.55)
0.279 ***
(3.26)
2.598 **
(2.26)
0.054
(0.16)
-5.150 ***
(-10.20)
YES
YES
-0.432 ***
(-3.34)
-0.359
(-1.47)
1,533
78.81 ***
1,845.67 ***
β1 = β2
F = 0.43
p=0.514
20.5%
-2,534.04
β1 = β2
F = 0.43
p=0.513
-3,838.08
β1 = β2
2
χ = 1.54
p=0.21
All models use robust standard errors clustered by firm. Continuous variables are winsorized at the 1 and 99 percentile
levels. *, **, and *** denote significance at the p < 0.10, 0.05, and 0.01 levels (all two-tailed), respectively. See Appendix
for full variable definitions.
TABLE 6
Sensitivity Test Summary Results: Classifying Type of Preparer from Tax Returns vs. from Financial
Statement Disclosures of Tax Fee Data
Panel A. Benchmark Case - Using Tax Returns to Classify Preparer as (Internal or External) Non-Auditor
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Coefficient
(t-statistic)
NONAUD_PREP_
TAXRET
+
0.259 **
(2.41)
All Controls
YES
Observations
1,533
Adjusted R2
60.1%
Model F-Stat
90.14 ***
Pseudo-R2
Log Pseudo-Likelihood
Tobit
Coefficient
(t-statistic)
0.323 ***
(2.64)
YES
1,533
83.57 ***
20.5%
-2,534.41
Panel B. Using Tax Fee Data to Classify Preparer as (Internal or External) Non-Auditor
Case 1: Non-Auditor Preparer as Zero Tax Fees
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Tobit
Coefficient Coefficient
(t-statistic) (t-statistic)
NONAUD_PREP_
10K
+
-0.012
-0.025
(-0.11)
(-0.21)
All Controls
YES
YES
Observations
1,533
1,533
Adjusted R2
59.8%
Model F-Stat
89.74 ***
83.03 ***
Pseudo-R2
20.3%
Log Pseudo-Likelihood
-2,540.91
Case 2: Non-Auditor Preparer as <1% Tax/Total Fee Ratio
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Tobit
Coefficient Coefficient
(t-statistic) (t-statistic)
NONAUD_PREP_
10K_1%
+
-0.059
-0.060
(-0.61)
(-0.56)
All Controls
YES
YES
Observations
1,533
1,533
Adjusted R2
59.8%
Model F-Stat
90.11 ***
83.25 ***
Pseudo-R2
20.3%
Log Pseudo-Likelihood
-2,540.67
Case 3: Non-Auditor Preparer as <5% Tax/Total Fee Ratio
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Tobit
Coefficient Coefficient
(t-statistic) (t-statistic)
NONAUD_PREP_
10K_5%
+
-0.141 *
-0.144
(-1.69)
(-1.53)
All Controls
YES
YES
Observations
1,533
1,533
Adjusted R2
59.9%
Model F-Stat
91.12 ***
84.12 ***
Pseudo-R2
20.3%
Log Pseudo-Likelihood
-2,538.93
Case 4: Non-Auditor Preparer as <10% Tax/Total Fee Ratio
Dependent Variable: Log_UTB_EB
Pred.
Variable
Sign
OLS
Tobit
Coefficient Coefficient
(t-statistic) (t-statistic)
NONAUD_PREP_
10K_10%
+
-0.127
-0.133
(-1.52)
(-1.46)
All Controls
YES
YES
Observations
1,533
1,533
Adjusted R2
59.9%
Model F-Stat
92.80 ***
85.25 ***
Pseudo-R2
20.3%
Log Pseudo-Likelihood
-2,539.49
TABLE 6 Panel B (continued)
Case 5: Non-Auditor Preparer as <25% Tax/Total Fee Ratio Case 6: Non-Auditor Preparer as <50% Tax/Total Fee Ratio
Dependent Variable: Log_UTB_EB
Dependent Variable: Log_UTB_EB
Pred.
Pred.
Variable
Sign
Variable
Sign
OLS
Tobit
OLS
Tobit
Coefficient Coefficient
Coefficient Coefficient
(t-statistic) (t-statistic)
(t-statistic) (t-statistic)
NONAUD_PREP_
NONAUD_PREP_
10K_25%
+
-0.301 **
-0.351 ***
10K_50%
+
0.858
1.171
(-2.42)
(-2.63)
(1.39)
(1.39)
All Controls
YES
YES
All Controls
YES
YES
Observations
1,533
1,533
Observations
1,533
1,533
Adjusted R2
59.9%
Adjusted R2
59.8%
Model F-Stat
91.67 ***
84.78 ***
Model F-Stat
90.06 ***
83.13 ***
Pseudo-R2
20.4%
Pseudo-R2
20.3%
Log Pseudo-Likelihood
-2,537.37
Log Pseudo-Likelihood
-2,539.96
All models use robust standard errors clustered by firm. Continuous variables are winsorized at the 1 and 99 percentile levels. *, **,
and *** denote significance at the p < 0.10, 0.05, and 0.01 levels (all two-tailed), respectively.
TABLE 6 (continued)
Panel C. Analysis of Accuracy in Classifying Preparer Type from Financial Statement Disclosures of Tax
# incorrectly classified
Is not
NONAUD_
Is NONAUD_
# correctly classified
# as
PREP but
PREP but
NONAUD # as AUD_ NONAUD
AUD_
10-K implies 10-K implies
_PREP
PREP
_PREP
PREP
Total
it is
it is not
Tax Return
NON_AUD_PREP_TAXRET
1,221
312
1,221
312
1,533
0
0
10-K
NON_AUD_PREP_10K
NON_AUD_PREP_10K_1%
NON_AUD_PREP_10K_2%
NON_AUD_PREP_10K_3%
NON_AUD_PREP_10K_4%
NON_AUD_PREP_10K_5%
NON_AUD_PREP_10K_10%
NON_AUD_PREP_10K_15%
NON_AUD_PREP_10K_20%
NON_AUD_PREP_10K_25%
NON_AUD_PREP_10K_30%
NON_AUD_PREP_10K_35%
NON_AUD_PREP_10K_40%
298
428
538
640
718
777
1,055
1,203
1,320
1,417
1,471
1,496
1,514
1,235
1,105
995
893
815
756
478
330
213
116
62
37
19
288
412
510
601
668
717
933
1,025
1,104
1,159
1,190
1,206
1,216
302
296
284
273
262
252
190
134
96
54
31
22
14
590
708
794
874
930
969
1,123
1,159
1,200
1,213
1,221
1,228
1,230
10
16
28
39
50
60
122
178
216
258
281
290
298
933
809
711
620
553
504
288
196
117
62
31
15
5
Total
% Sample
Incorrectly
Classified
0
0.0%
943
825
739
659
603
564
410
374
333
320
312
305
303
61.5%
53.8%
48.2%
43.0%
39.3%
36.8%
26.7%
24.4%
21.7%
20.9%
20.4%
19.9%
19.8%
NONAUD_PREP_TAXRET = 1 if tax return does not report the firm's auditor as its tax preparer; 0 otherwise.
NONAUD_PREP_10K = 1 if firm does not report in its 10-K any tax fees paid to its auditor; 0 otherwise.
NONAUD_PREP_10K_1% (_2%, _3% …) = 1 if firm reports in its 10-K any tax fees paid to its auditor less than or equal to 1% (2%, 3%..., respectively) of
total fees paid to the auditor; 0 otherwise.