did pcaob rules concerning independence and tax services ...€¦ · realigned clients and tax...
TRANSCRIPT
Did PCAOB Rules Concerning Independence and Tax Services Improve Audit Quality? Further Evidence
Kellie M. Carr Western Michigan University
Jian Cao * Florida Atlantic University
Joseph R. Rakestraw Florida Atlantic University
June 2019
* Corresponding Author
We thank Mark Kohlbeck, Ken Orbach, Maya Thevenot, George Young, as well as workshop participants at the Western Michigan University and the 2017 PhD Project Conference for helpful comments. Carr also thanks the KPMG Foundation Doctoral Scholarship and the AICPA Doctoral Fellowship for their financial support of her dissertation based upon which this paper is developed.
Did PCAOB Rules Concerning Independence and Tax Services Improve Audit Quality? Further Evidence
Abstract: With the intention to increase audit quality, the PCAOB Rules on Ethics, Independence,
and Tax Services prohibited accounting firms from providing aggressive tax services to their audit
clients. Lennox (2016) found the restrictions resulted in a significant drop in auditor-provided tax
services (APTS), but no change in audit quality for companies with significant reductions in APTS
fees vis-à-vis those without. This study seeks to provide further evidence on the consequences of
the PCAOB restrictions. Using tax accrual quality as a proxy for audit quality specific to the tax
account, we find robust evidence that audit quality increased among companies that significantly
reduced APTS fees after the restrictions were imposed. Furthermore, the increase in audit quality
stems primarily from companies potentially targeted by the PCAOB restrictions on aggressive tax
services. Overall, our evidence suggests net benefits from restricting auditors’ ability to provide
aggressive tax avoidance schemes to audit clients.
Keywords: PCAOB; audit quality; auditors’ tax services; tax accrual quality
Data availability: The data used in this study is available from public sources indicated in the paper.
1
Did PCAOB Rules Concerning Independence and Tax Services Improve Audit Quality? Further Evidence
I. INTRODUCTION
The Public Company Accounting Oversight Board (PCAOB) Rules on Ethics,
Independence, and Tax Services are intended to increase auditor independence by prohibiting
registered public accounting firms from providing contingent fee-based and highly aggressive tax
services to their Securities and Exchange Commission (SEC) regulated audit clients (PCAOB
2005, 2006). The PCAOB’s concern that aggressive auditor provided tax services (APTS)
compromise auditor independence is based on the premise that an auditor’s incentive to
compromise independence with respect to a client depends on the client’s economic importance
(i.e., quasi-rents) (DeAngelo 1981). Regulators had hoped the restrictions on APTS to reduce the
economic bond between the auditor and the auditee and lessen the potential impairment of auditor
independence and audit quality. However, opponents claimed that the restrictions could have
negative consequences because APTS provide auditors with superior client-specific knowledge
that leads to more efficient audits (e.g., Simunic 1984). In general, prior evidence has suggested
both a positive (e.g., Kinney, Palmrose, and Scholz 2004; Gleason and Mills 2011; De Simone,
Ege, and Stomberg 2015) and negative or mixed (e.g., Cook, Huston, and Omer 2008; Alsadoun,
Naiker, Navissi, and Sharma 2018) association between APTS fees and audit quality. However,
the literature has provided little evidence on the effects to audit quality from legislative efforts to
restrict certain types of APTS, such as in the case of the PCAOB rules.
The purpose of this study is two-fold. First, we build on and extend the empirical
framework of Lennox (2016) on the effectiveness of the PCAOB restrictions by examining
whether the reductions in APTS fees in response to the restrictions affect audit quality. Lennox
(2016) documents that the PCAOB restrictions had a chilling effect on non-audit tax services
2
provided by audit firms. Using restatements and going concern opinions to proxy for audit quality,
he finds no change in audit quality for companies with significant reductions in APTS fees vis-à-
vis companies without. We focus on the quality of financial reporting specific to the tax account
to provide a more direct test of the link between changes in APTS and related financial statement
effects. Second, given the specific target of the PCAOB restrictions has much to do with auditors’
aggressive tax services to audit clients, we investigate whether companies that practice aggressive
tax avoidance are more likely to experience an improvement in audit quality following the
reductions in APTS fees. The use of different measures and subsets of companies for evaluating
audit quality can provide additional evidence as to how audit quality changes with legislative
efforts to restrict certain APTS. The evaluation is highly necessary because stakeholders in the
capital markets are increasingly concerned that misconduct in an audit firm’s nonaudit tax practice
can adversely impact its audit practice, a debate which has reemerged in recent years (e.g., Brown,
Shu, Soo, and Trompeter 2013; Baugh, Boone, Khurana, and Raman 2019).1
The Sarbanes-Oxley Act of 2002 prohibited most non-audit services (NAS) to audit clients
to improve auditor independence by reducing the financial alignment between auditors and their
audit clients. Nonetheless, non-audit tax services in general are not prohibited under the SOX. The
PCAOB 2006 restrictions are the first to specifically restrict APTS. Prior to the restrictions and
during the investigation into accounting firm tax shelter abuses, regulators learned that some audit
firms were selling aggressive tax services, often for contingent fees, and selling personal tax
services to audit client executives responsible for overseeing the relationship between the client
1 One example is the U.S. Senate Permanent Subcommittee on Investigations (PSI 2014) report on Caterpillar’s tax invasion. Caterpillar paid over $55 million to its independent auditor PwC to devise an abusive tax shelter that generated $2.4 billion in U.S. tax benefits. The PSI report criticized that PwC’s roles as auditor and tax consultant represented a conflict of interest. Another example is a full-scale break-up of KPMG, EY, PwC and Deloitte commanded by U.K. regulators to legally separate audit and non-audit services (Beardsworth and Browning 2019).
3
and auditor (General Accounting Office, GAO 2005; PSI 2005). The potential economic
dependence between the auditor and client resulting from these transactions heightened regulator
concerns about auditor independence and contributed to the PCAOB 2006 restrictions. Rule 3521
of the PCAOB restrictions prohibits regulated accounting firms from using a contingent fee
arrangement in any transaction with a SEC audit client. Rule 3522 prohibits SEC regulated
accounting firms from marketing, selling or opining a tax treatment or transaction which is
confidential or tax aggressive to their SEC regulated audit client, including recommending, either
directly or indirectly, a tax aggressive position transaction. Rule 3523 prohibits a registered public
accounting firm from providing any tax service to a person in a financial reporting oversight role
at the issuer audit client, or an immediate family member of such person. The PCAOB announced
the rules on July 26, 2005 with the latest effective date of October 31, 2006.
The PCAOB restrictions represent a significant change to the tax service industry as it
realigned clients and tax service providers. However, it is unclear whether the PCAOB restrictions
achieved its objectives in improving audit quality and, by extension, financial reporting quality
(PCAOB 2004). For companies reducing their APTS, they can choose another tax service provider
or their internal tax department for tax advice. When a client eliminates APTS, it mitigates
concerns about the negative effects of economic bonding and thus, theory suggests increases in
auditor independence and audit quality. However, the client also eliminates the benefits of
knowledge spillover from APTS and thus, potentially reduces audit quality. The change in the
offsetting effects of economic bonding and knowledge spillover, as a result of the separation of
jointly provided tax and audit services, can increase, decrease, or have no effect on audit quality—
making it important for us to test which effects are strongest.
4
We measure financial reporting quality specific to the tax account to determine whether
the APTS restrictions resulted in a change in audit quality. Specifically, we use a tax-specific
proxy, tax accrual quality, developed following the approach in Choudhary, Koester, and Shevlin
(2016). Several considerations motivate our focus on this tax-specific audit quality measure. First,
the income tax accrual (the difference between income tax expense and income tax cash flow) is
economically large, complex, involves complicated judgments related to the financial reporting of
income taxes and, therefore, are amenable to both intentional and unintentional estimation errors.
Choudhary et al. (2016) show that tax accrual quality predicts future tax-related restatements and
internal control material weaknesses, which are more extreme forms of poor financial reporting
quality. Furthermore, the tax account provides a better representation of audit quality when
investigating its association with APTS (reductions). On the one hand, the audit firm’s assessment
of the tax account reflects the product of joint work by audit and tax professionals, as the task
requires a deep understanding of both tax-related GAAP and the tax law that surround firms’
various tax strategies (McGuire, Omer, and Wang 2012). On the other hand, there were concerns
that conflicts of interest may compromise auditors’ judgment over the tax account when auditors
provide aggressive tax services to audit clients (e.g., PCAOB 2004).2 Either way a direct link
between APTS (restrictions) and tax accrual quality is expected.
We identify companies that significantly reduced APTS during the transition prior to when
the restrictions were given effect (from July 26, 2005 through October 31, 2006). Using a
difference-in-difference research design, we find that companies that significantly reduced APTS
when the restrictions were imposed (i.e., the treatment group) experienced a larger improvement
2 For example, providing aggressive tax services may put the auditor in a position to act as management (e.g., to implement a tax strategy), audit their own firm’s work (the financial statement effects of a tax strategy the audit firm sold to the client), or perform as an advocate for the client (e.g., to withstand IRS scrutiny) (Harris 2014).
5
in tax accrual quality after the change compared to companies that did not significantly reduce
APTS (i.e., the control group). Furthermore, the increase in tax accrual quality is driven by the
target of the restrictions—clients associated with aggressive APTS. Specifically, for tax-
aggressive companies, those that reduced APTS experienced a significant increase in tax accrual
quality after the change compared to tax aggressive companies that did not significantly reduce
APTS. By contrast, companies not considered tax aggressive but also reduced APTS did not
experience the same increase in tax accrual quality. Our results are robust to matching of treated
and control firms based on their propensity scores, using alternative post-event windows, using
alternative measures of tax aggressiveness, and controlling for audit efforts. Overall, our results
indicate that the PCOAB restrictions were effective in lessening the potential impairment of
auditor independence, thereby leading to improved audit quality, especially among tax-aggressive
companies that were likely to be targeted by the rules.
Our study provides several contributions to the literature. First, our study extends Lennox
(2016) by providing further evidence on the impact of the initial legislative limit on certain types
of APTS. We focus on financial reporting quality specific to the tax account as a key accounting
estimate to directly address the regulatory concern about auditors’ conflicts of interest in this area.
Our study complements Lennox (2016), suggesting that the PCAOB restrictions not only had a
chilling effect over the provision of APTS, but also to some extent lessened concerns over auditors’
objectivity in evaluating the tax account.
Second, our study extends prior studies that examine the effect on firms’ accounting / audit
quality from APTS. Prior studies produce mixed findings regarding the economic bonding vis-à-
vis knowledge spillover effect of APTS fees on audit quality. Using an exogenous regulatory shock
to the tax service industry, we examine the link between a significant reduction in APTS fee and
6
the change in audit quality. Our evidence suggests net benefits from restricting auditors’ aggressive
tax services to audit clients, evidence consistent with the economic bonding theory.
Third, our study also adds to the line of research investigating the relationship between
aggressive tax avoidance and financial reporting quality (e.g., Frank, Lynch, and Rego 2009). We
find that tax-aggressive companies experienced greater improvement in tax accrual quality
following the restrictions on auditors’ aggressive tax services. These findings are broadly
consistent with aggressive tax strategies adversely impacting financial statements.
The next section develops the study’s hypotheses. Section III presents the research design,
while Section IV discusses sample selection. Section V reports main empirical results and
additional analyses, and Section VI concludes.
II. BACKGROUND AND HYPOTHESES DEVELOPMENT
Prior Research
In a seminal paper, DeAngelo (1981) argues that the bilateral audit-client relationship
creates client-specific quasi-rents that provide an incentive for auditors to compromise their
independence leading to lower audit quality. Compromised independence arises because the quasi-
rents lower the likelihood the auditor will report an audit failure in order to maintain future business
from the client. Greater future quasi-rents lead to lower independence and hence lower audit
quality. Extending DeAngelo’s (1981) theory to the non-audit tax service suggests that an
economically meaningful amount of APTS fee paid by an audit client increases the audit firm’s
economic dependence on the client thereby compromising auditor independence. For example, the
auditor may allow greater managerial discretion in the client’s accounting choices to secure future
quasi-rents from the client. Alternatively, practitioners and some scholars assert that knowledge
7
generated from the provision of tax services may lead to synergy, knowledge sharing, and more
efficient audits (e.g., Simunic 1984; Kinney et al. 2004). Simunic (1984) describes knowledge
spillover as the efficiencies that can occur when joint services, with interdependent production
functions, are provided by the same accounting firm. Because the production functions overlap,
there is also an overlap of client specific knowledge. When this knowledge is shared, a synergy is
created that increases the quality of one or both services. Tax services provided by audit firms may
foster expertise in a specialty area essential for conducting quality audits. The tax team may help
the audit team understand the client’s tax issues or serve as a tax specialist during the audit
(McGuire et al. 2012).
Using proprietary fee data in the pre-SOX period, Kinney et al. (2004) find a significant
negative association between tax services fees and financial statement restatements, consistent
with tax service improving audit effectiveness through a knowledge spillover. In a follow-up study
based on post-SOX data (2003-2005), Seetharaman, Sun, and Wang (2001) find a negative
association between APTS and the incidence of accounting restatements but only when they are
tax-related. Robinson (2008) documents a positive association between APTS fees and issuance
of going-concern opinions prior to bankruptcy filings for the period 2001–2004. Christensen,
Olson, and Omer (2015) find that providing APTS constrains last-minute earnings management
via income tax expense. Using IRS disputes data between 2000-2002, Gleason and Miller (2012)
find that companies that purchased APTS had more accurate and adequate reserves for IRS
disputes or settlements relative to companies that did not purchase APTS. Using data from 2004
to 2012, De Simone et al. (2015) find that companies purchasing tax NAS are significantly less
likely to disclose a material weakness consistent with communication between the audit and tax
side leading to timely detection and remediation of internal control weaknesses.
8
However, Cook et al. (2008) find a positive association between APTS fees and the
propensity to meet or beat analysts’ forecasts by managing the tax expense in both the pre and
post-SOX regimes. By contrast, Krishnan and Visvanathan (2011) find a positive (negative)
association between APTS and the propensity to meet or beat the loss-avoidance benchmark in the
pre-SOX (post-SOX) period. Alsadoun et al. (2018) suggest that investors seem to perceive APTS
negatively because of increased tax risk rather than increased financial reporting risk. Collectively,
the prior literature has used different measures and subsets of companies for evaluating the effect
of APTS on audit quality but suggests only limited evidence that non-audit tax services are
associated with impaired independence.
The Effectiveness of the PCAOB Restrictions
While the PCAOB 2006 rules extended SOX by improving auditor independence
standards, these rules are the first to specifically restrict APTS. The concern about independence
impairment has been renewed following the increasing awareness of regulators and investors that
many audit firms were supplying highly aggressive tax services on a contingent fee basis to audit
clients and their top executives (GAO 2005; PSI 2005). To address potential threats that these tax
services may pose to audit quality and auditor independence, the SEC approved the PCAOB
proposed rules limiting tax services and prohibiting contingent fees in 2006. More specifically,
Rule 3522 states that a “registered public accounting firm is not independent of its audit client if
the firm, or any affiliate of the firm, during the audit and professional engagement period, provides
any non-audit service to the audit client related to marketing, planning, or opining in favor of the
tax treatment of either a confidential or aggressive tax position transaction” and the aggressive tax
9
position cannot be a recommendation, directly or indirectly (PCAOB 2006).3 The PCAOB also
considers that an accounting firm lacks independence from a client if the firm charges that client
contingent fees (Rule 3521) or provides tax services to executives, or their family, who have
financial statement oversight responsibility (Rule 3523).
There are very few studies that have investigated the effectiveness of the PCAOB
restrictions of APTS on audit quality. Whereas the PCAOB restrictions resulted in a significant
drop in APTS when the new rules were introduced, Lennox (2016) suggests that such reductions
did not influence the incidence of accounting restatements or auditors’ going-concern opinions.
We focus on auditors’ ability to evaluate complex accounting estimates associated with the tax
account. More specifically, we examine whether a substantial reduction in APTS following the
PCAOB restrictions leads to a change in the quality of the income tax accrual.
Actions of the PCAOB and others to restrict APTS are consistent with the empirical
predictions about the association between economic bonding and lower quality financial reporting.
Furthermore, much of the debate about Rule 3521 (contingent tax fees) and Rule 3522 (aggressive
tax services) centered on the quality of the tax account (PCAOB 2004; SEC 2006). If a company’s
tax obligations rely on a tax strategy sold by its audit firm, the provision of tax services could place
the auditor in a position of auditing its own work (i.e., a self-review threat). When a company
separate its audit and tax service providers, the economic bond between the auditor and the auditee
is reduced which may lessen the potential impairment of auditor independence and audit quality.
Therefore, we would expect a positive association between a substantial reduction in APTS and
an improvement in tax accrual quality.
3 Rule 3522 defines an “aggressive tax position” as one where tax avoidance is the main purpose of the plan and is less likely than not to be allowed under applicable tax law.
10
However, there are at least three bases suggesting a no association or negative association
would be expected. First, APTS may improve audit effectiveness through a knowledge spillover.
Financial reporting for taxes requires technical knowledge regarding how GAAP-based income
and taxable income articulate (Choudhary et al 2016). Knowledge generated from the provision of
tax services (such as client-specific knowledge and knowledge of the tax laws in specific
jurisdictions) may benefit the auditor’s assessment of the tax account (McGuire et al. 2012).
Conversely, when a company separate its audit and tax service providers, the benefit of knowledge
spillover would be missing. Second, the tax account comprises a significant source of financial
reporting risk (Graham, Raedy, and Shackelford 2012) and the combination of tax aggressiveness
and tax uncertainty may lead to an additional audit risk (Donohoe and Knechel 2014). Therefore,
the incumbent auditor has incentives to enhance the quality of the tax account to reduce audit risk.
However, when companies choose another tax service provider, the new firm is not responsible
for a thorough consideration of the ramifications of their tax planning and positions on their clients’
financial statements. Finally, auditors are required to implement quality controls to address the
threats to auditor independence (PCAOB 2008). If any of these scenarios applies, a substantial
reduction in APTS will suggest a decrease or no change in tax accrual quality, that is, the goal of
the PCOAB restrictions of increased audit quality may not occur. Given the arguments for and
against the effects of PCAOB restrictions on the reporting quality of the tax account, we state our
first hypothesis (in null form) as follows:
H1: There is no change in tax accrual quality after companies reduce purchasing APTS
following the new rules.
The Target of the PCAOB Restrictions
11
Our second research question addresses whether the effect of the PCAOB restrictions on
the reporting quality of the tax account is most pronounced for the target of the restrictions—
companies aggressively avoiding taxes with the assistance of APTS. There are two groups of
companies that significantly reduced APTS surrounding the PCAOB restrictions. The first group
is not concerned with aggressive tax services because they do not practice that level of tax
avoidance. Due to close scrutiny by the SEC and the PCAOB on independence violations in the
auditor-client relationship, and by the IRS over the tax shelters sold by KPMG and other audit
firms (e.g., GAO 2005), companies in the first group wanted to distance themselves not only from
APTS, but particularly from KPMG and other tax providers of potentially abusive or illegal tax
services. Finley and Stekelberg (2016) report a significantly larger decline in APTS services for
KPMG than for other Big 4 auditors after KPMG’s acknowledgement of “unlawful conduct”
connected with tax shelter fraud and the related deferred prosecution agreement. This finding is
consistent with the view that audit firm reputation could be a major consideration in a client’s
decision to change or continue with the incumbent service provider (Weber, Willenborg, and
Zhang 2008; Boone et al. 2019). In investigating the effectiveness of PCAOB restrictions, Lennox
(2016) reports 46 percent of companies reduced their APTS purchases. At least part of this decline
could be explained by a greater reputation concern. While the reduction or elimination of APTS
mitigates this concern, it does not necessarily enhance auditor independence in a way that
positively affects audit quality.
The second group that significantly reduced APTS consists of companies that practice
various levels of aggressive tax avoidance and reduced their APTS due to the ban on aggressive
tax services. The combination of contingent fee based and aggressive tax services may create a
strong economic bond between the client and the audit firm than for non-aggressive tax services.
12
Tax avoidance, or a low level of aggressive tax planning, can equate to legal tax savings. As the
level of aggressiveness increases, so does the tax savings. Conversely, the accompanying risk of
repaying tax savings (including penalties and interest) to the IRS and the costs of restating financial
statements also increases with aggressive tax avoidance. Often, for clients to take this risk, they
agree to pay a high fee only if the plan is successful (a contingent fee). The lucrative contingent
tax fees may create a strong economic bond between the auditor and the client. Furthermore, the
PCAOB is also concerned that the reason behind a particular aggressive tax strategy or abusive
tax shelter is to obtain a certain financial statement outcome. Combined with economic bonding,
if accounting firms are auditing their own work, there is a greater risk of independence impairment
supporting the regulation to limit economic bond by reducing aggressive tax fees.
However, the loss of knowledge spillover due to the switch of tax service providers could
also be greater for companies practicing aggressive tax avoidance due to greater complexity in
estimating the various components of the tax expense of such companies (Donohoe and Knechel
2014). In addition, because of increased regulatory scrutiny over aggressiveness tax services,
auditing firms may have already taken stronger measures to protect their reputation and avoid
litigation (e.g., Finley and Stekelberg 2016). These factors suggest no change or even a decrease
in audit quality after tax aggressive companies switch their tax service providers. In view of the
conflicting arguments, we state our second of hypothesis (in null form) as follows:
H2: There is no change in tax accrual quality after tax-aggressive companies reduce
repurchasing APTS following the new rules relative to similar actions taken by non-tax
aggressive companies.
III. RESEARCH DESIGN
13
Measure of Tax Accrual Quality
We follow the approach of Choudhary et al. (2016) to create a tax-specific measure of audit
quality. Choudhary et al. (2016) develop a measure of tax accrual quality to capture variation in
the extent to which the income tax accrual (difference between income tax expense and tax cash
flow) maps into income tax-related cash flows. Following Dechow and Dichev’s (2002) approach
for estimating working capital accruals, the tax accrual quality measure is based on the precision
in which the income tax accrual (the estimate) maps into past, current, and future income taxes
paid (the realization) after controlling for long-term temporary book-tax differences. Low tax
accrual quality may capture intentional and unintentional management estimation error in the tax
account for several reasons. First, the tax accrual is estimated for financial statement purposes prior
to the completion of the tax return. The estimation often requires significant knowledge and
involves complex judgment. Second, management can exploit the intricacy of tax expense
computations and use discretion estimating tax accruals to manage earnings (e.g., Hanlon 2005;
Hanlon, Krishnan, and Mills 2012; Donohoe and Knechel 2014).
Following Choudhary et al. (2016), we begin with the following regression model of tax
accrual (Eq. (1); see Appendix for variable definitions):
TAXACCi,t= α0 + α1CTPi,t-1 + α2CTPi,t + α3CTPi,t+1 + α4∆DTLi,t + α5∆DTAi,t + εi,t (1)
where tax accrual (TAXACC) is calculated by subtracting cash taxes paid (CTP) from total tax
expense (TTE). TAXACC is equivalent to the balance sheet tax accrual measure, which is calculated
as the sum of changes in income taxes payable, deferred tax assets, deferred tax liabilities, the
valuation allowance, and unrecognized tax benefits.4 Cash taxes paid (CTP) for t - 1, t and t + 1
4 The components of the tax accrual are different from those of working capital accruals, with only income taxes
payable and short-term deferred tax assets and liabilities being included in both accruals (Choudhary et al. 2016).
14
are used to assist in the mapping between the estimate and the realization. Changes in the long-
term deferred tax liability (∆DTL) and long-term deferred tax asset (∆DTA) are included to control
for the temporary differences between tax and financial reporting that do not represent
management estimation errors and will often reverse over the long term, i.e., outside of t - 1 and t
+ 1. All variables are scaled by total assets. A high quality tax accrual should map into its related
cash flows. Therefore, we expect a positive relation of TAXACC with the lag (CTPt-1) and lead
values of cash taxes paid (CTPt+1), and the increase in long-term deferred tax liability (∆DTL),
whereas we expect a negative relation for cash taxes paid (CTP) and the increase in long-term
deferred tax asset (∆DTA).
We estimate Eq. (1) using annual industry cross-sectional regressions over rolling windows
from t - 2 through t.5 Tax accrual quality (TAXAQ) is the standard deviation of the residual values
over the rolling three-year window, multiplied by -1. As such, a higher TAXAQ indicates better tax
accrual quality, i.e., lower variation in a firm’s tax accrual (an estimate) and its related cash flows
(a realization) mapping.
By variable construction, a low TAXAQ could reflect management estimation errors related
to accounting for transactions that affect current tax expense, short-term deferred tax assets and
liabilities, short-term valuation allowances, and unrecognized tax benefits. It could also reflect
GAAP-induced mismapping, that is, book-tax differences that will not be resolved over time
through the reversal of deferred tax assets and liabilities (not captured by ∆DTA and ∆DTL). Since
we focus on the management error aspect of TAXAQ, we follow Choudhary et al. (2016) and
5 Industry is defined as two-digit Standard Industry Classification code and we require a minimum of 20 observations per industry year to estimate Eq. (1).
15
control for the GAAP-induced mismapping between tax accrual and related cash flow in our
research design (see Section of “Regression Model”).
Treatment and Control Groups
We intend to capture the change in tax accrual quality due to the reduction in APTS in
response to the PCAOB restrictions. The PCAOB announced the three rules on July 26, 2005 and
the SEC approved them on April 19, 2006. Rule 3521 and Rule 3522 became effective as of June
18, 2006, whereas Rule 3523 became effective after October 31, 2006 (PCAOB 2006). Following
Lennox (2016), we expect reductions in APTS fees likely occur during the transition period from
July 26, 2005 to October 31, 2006 (referred to as the “event window”). Accordingly, the pre-event
period consists of fiscal years ended prior to July 26, 2005, whereas the post-event period consists
of fiscal years beginning after October 31, 2006. We compare tax accrual quality over the one-
year period before the event window to the one-year period after the event window (see Figure 1
for the timeline for measuring tax accrual quality relative to reductions in APTS fees).
<Insert Figure 1 about here>
To test the effect of the PCAOB 2006 restrictions, we compare the change in tax accrual
quality for the group of companies that significantly reduced APTS (i.e., the treatment group) to
the group of companies that did not significantly reduce APTS (i.e., the control group) when the
restrictions were introduced. Consistent with Lennox (2016), we define a significant reduction in
APTS as at least a 75 percent reduction during the event window, i.e., from July 25, 2005 to
16
October 31, 2006. We employ both the full sample and a propensity-score matched sample of the
treatment and control firms for testing our hypotheses.6
To implement the propensity-score matching procedure, we estimate a logit model for the
probability of reducing APTS fees conditional on a number of observable dimensions.
Specifically, we use the following logistic regression (Eq. (2)), including factors affecting firms’
decision to reduce APTS (Lennox 2016) and tax accrual quality (second-stage regression):
REDUCEi = β0 + β1APTS_LASTi,t + β2∆OTHERi,t + β3SIZEi,t + β4EXCHANGEi,t +
β5BIG4i,t + β6KPMGi,t + β7LOSS + ΣPROXY_CONTROLS + Σindustries + µi
(2)
where REDUCE is equal to one if the reduction in APTS is at least 75 percent, and zero otherwise;
APTS_LAST is the logarithm of APTS fees from the last financial statement in the pre-event
window and is used to control for the magnitude of APTS fees prior to the restriction taking effect;
∆OTHER is the absolute percentage reduction of other non-audit services and controls for the
possibility that companies eliminated all NAS to signal high quality financial statements (Maydew
and Shackelford 2007); SIZE is the logarithm of total assets; EXCHANGE is an indicator variable
that equals one if the stock is traded on an exchange, and zero otherwise; BIG4 is an indicator
variable equal to one if the auditor is one of the Big 4 audit firms (PwC, EY, Deloitte, or KPMG),
and zero otherwise; KPMG is an indicator variable equal to one is the auditor is KPMG, zero
otherwise; and LOSS is an indicator variable equal to one if the company reports a loss, and zero
otherwise. Industry fixed effects are based on two-digit SIC code. We also include all control
6 On one hand, the propensity-score procedure allows an efficient matching along multiple dimensions and is more robust to a partial-matched econometric method using a small set of variables (e.g., Armstrong, Jagolinzer, and Larker 2009; Lawrence, Minutti-Meza, and Zhang 2011). On the other hand, matching reduces the size of the control group and hence the testing power, is ineffective in mitigating unobservable, and the resulting inference could be sensitive to matching criteria (DeFond, Erkens, and Zhang 2014; Lennox, Francis, and Wang 2012). The use of alternative samples provides a robustness check on our results.
17
variables (ΣPROXY_CONTROLS) in our tax accrual quality analyses (see Eq. (3)) except for those
redundant to the variables in predicting APTS reductions. We then identify matched-pairs with the
smallest propensity-score differences and examine the covariate balance between the treatment
and control groups (see Section of “Empirical Results”).
Regression Model
For testing our hypotheses, we implement a difference-in-difference research design to
measure the tax accrual quality for companies that significantly reduced APTS at the time periods
before and after the restrictions against companies that did not significantly reduce APTS.
Specifically, we estimate the following OLS model using observations from the treatment and
control groups one year before and after the test event:
TAXAQi,t = δ0 + δ1REDUCEi + δ2POSTi,t + δ3REDUCEi*POSTi,t + δ4 UTB_ESTi,t + δ5
ESO_INDi,t + δ6 DISCi,t + δ7FOREIGNi,t + δ8PTBI_VOLi,t + δ9TAX_LOSSi,t +
δ10SIZEi,t + δ11BIG4i,t + δ12 KPMGi,t + Σindustries + εi,t
(3)
where POST is an indicator variable equal to 1 for the one-year period in the post-event window,
and zero for the one year period in the pre-event window; UTB_EST is the predicted value of
unrecognized tax benefits (UTB), calculated based on coefficient estimates from Rego and Wilson
(2012);7 ESO_IND is an indicator variable equal to one if the client operates in industries likely to
issue employee stock options, and zero otherwise; DISC is an indicator variable equal to one if the
absolute value of discontinued operations and extraordinary items is greater than one percent of
sales, and zero otherwise; FOREIGN is an indicator variable equal to one for the presence of
7 UTB_EST is used to proxy for the magnitude of uncertain tax positions. The amount of UTB is estimated because
actual data is not available until 2007.
18
foreign operations, and zero otherwise; and PTBI_VOL is the standard deviation of pre-tax book
income scaled by total assets, measured from years t - 5 through t.
The coefficient on REDUCE (δ1) represents the difference in tax accrual quality (TAXAQ)
between the treatment and control groups during the pre-event window. The coefficient on POST
(δ2) captures the change in audit quality for the control group from the pre-event window to the
post-event window, whereas the interaction term, REDUCE* POST (δ3), captures the incremental
change in audit quality for the treatment group from the pre-event window to the post-event
window. Consistent with the discussion leading to Hypothesis 1, we provide no prediction for δ3
given the simultaneous influence of many factors affecting audit quality in opposite directions. A
positive (negative) coefficient would suggest the reductions in APTS increased (decreased) audit
quality due to the lessening of economic bonding (the loss of knowledge spillover). Similarly, we
leave the coefficients on δ1 and δ2 unpredicted.
Following Choudhary et al. (2016), we control for items that reflect significant judgment
in the application of tax-related GAAP. We include reserves for uncertain tax positions
(UTB_EST), which involve uncertainty about the ultimate treatment of a specific transaction as
well as judgment about how the tax law applies. Examples of such circumstances may include
whether an expense qualifies for a tax deduction, and whether a tax credit can be claimed.
Moreover, the application of tax-related GAAP can also induce differences between income tax
expense and income tax cash flows not captured by deferred tax assets and liabilities. For
example, firms are not permitted to deduct the expense associated with stock options and will
receive excessively large tax deductions from the exercise of options. Tax expense does not
reflect the tax effects of discontinued operations (reported net of tax on the income statement)
while cash tax payments do. We control for the GAAP-induced mismapping by including an
19
indicator for industries with potentially large tax deductions from ESOs (ESO_IND), and an
indicator equaling one if the absolute value of discontinued operations and extraordinary items is
greater than 1 percent of revenues (DISC). We also control for firm-level factors that may
increase complexity in transactions that affect tax-related accounts and hence lead to greater
estimation error: foreign operations (FOREIGN), earnings volatility (PTBI_VOL), and the
presence of a tax loss (TAX_LOSS), while controlling for SIZE to capture the magnitude of
available resources that potentially reduce management estimation error. We also control for the
effect of audit firm reputation by including the indicator variables BIG4 and KPMG (e.g.,
DeFond and Zhang 2014). Finally, we include year and industry fixed effects and estimate robust
standard errors clustered by firm (Petersen 2009; Gow, Ormazabal, and Taylor 2010).
Hypothesis 2 examines whether the effectiveness of the PCAOB restrictions is most likely
to be observed among the target of the restrictions—companies that take aggressive take positions
via the assistance of APTS. We identify tax aggressive companies (AGGRESSIVE) based on below
(vis-à-vis above) median book effective tax rates (ETR). We view a firm’s book ETR as the most
appropriate measure to test our hypothesis as it is a highly visible measure of tax planning
performance that boards and policy-makers can monitor and evaluate (e.g., Armstrong et al. 2015;
Kubick et al. 2017). For example, many SEC comment letters relate specifically to companies’
effective tax rates (e.g., Kubick et al. 2016). To test whether the change in tax accrual quality is
most pronounced among tax aggressive companies, we use Eq. (2) by estimating the model among
the two levels of tax aggressiveness (AGGRESSIVE). Hypothesis 2 suggests the improvement in
tax accrual quality after reducing APTS fees is not significantly different between tax aggressive
and non-tax aggressive companies. Consistent with Hypothesis 2, we provide no prediction for the
sign of the difference in the coefficient on REDUCE*POST between the two groups.
20
IV. SAMPLE
Sample Selection and Data
Table 1 delineates the sample section for this study. For computing TAXAQ, we start with
all firm-years in the Compustat universe that have required data for estimating Eq. (1) over the
period 2003-2009. We chose to begin the data period with fiscal year 2003 because the SEC (2003)
required companies to disclose tax-related fees paid to a company’s auditing firm for fiscal years
ending after December 15, 2003. We end data collection by July 31, 2009 to ensure the time
periods before and after the event window (July 26, 2005 to October 31, 2006) are approximately
the same length. Since TAXAQ is based on the variation in the residuals from Eq. (1) over t - 2 to
t, we eliminate industry-years with fewer than twenty observations for estimating Eq. (1) and firms
with fewer than three years of consecutive data for computing the residual variation (see Section
of ‘Estimating tax accrual quality’). Separately, we collect audit fee and APTS fee data from Audit
Analytics and the intersection of Compustat and Audit Analytics produces 31,084 firm-year
observations. We further exclude observations with missing data for computing other control
variables in the regression models (Eqs. (2) and (3)). Since we require one year of data both pre
and post-event, we also eliminate observations without corresponding data in both the pre-event
and post-event windows.
The above sample selection procedures provide us with 565 and 1,809 unique firms in the
treatment and unmatched control groups, respectively. The resulting full sample includes 4,748
firm-year observations with a pre-event and post-event observation for each firm. To obtain a
matched control group, we estimate the propensity-score model (Eq. (2)) and match a treatment
observation with a control observation that has the closest predicted value. We further eliminate
1,244 firms (2,488 firm-year observations) that are not identified as the closest match to the
21
treatment group based on propensity scores. The resulting propensity-score matched sample
contains 2,260 matched treatment and control firm-year observations (i.e., 565 unique firms * 2
groups * 2 years) available for analyses.
<Insert Table 1 about here>
Table 2 Panel A reports the mean fees paid to auditors for tax services, other non-audit
services, and audit services between 2002 and 2009 according to Audit Analytics. Although SOX
did not restrict APTS, there is a decline in APTS fees in both 2003 and 2004, consistent with
companies facing increasing pressure from boards and other interested parties to voluntarily reduce
or eliminate APTS (e.g., Maydew and Shackelford 2007; Omer, Bedard, and Falsetta 2006).
However, the largest reduction in APTS fees (24 percent) occurred in 2005 and continued in early
2006, coinciding with the event window during which the PCAOB restrictions were imposed. By
contrast, other NAS fees declined by 54 percent in 2003, due to the prohibition of NAS by SOX,
whereas the change is relatively small during 2005 and 2006 (a 12 percent decrease followed by a
9 percent increase). The decline in other NAS fees in 2005 suggests that some companies had
reduced APTS and other NAS concurrently. Meanwhile, audit fees increased by 74 percent in 2004
and continued to increase throughout 2008, consistent with the increase in internal control
reporting requirements.
<Insert Table 2 about here>
Panel B further reports the annual percentage change in APTS fees, NAS fees, and audit
fees by the pre-event window, event window, and post–event window. The computation of the
annual percentage change in fees reduces the number of observations to 36,809 firm-year
observations. Consistent with the observed trend in Panel A, firms reduced APTS fees more
frequently in the event window relative to the pre-event and post-event windows. During the event
22
window, 40.14 percent of observations exhibit a reduction in APTS fees, and the mean annual
reduction is 21.01 percent, as compared to reductions of 17.70 percent and 16.22 percent for the
pre and post-event periods, respectively. We do not observe a similar pattern for other NAS and
audit fees. Overall, consistent with Lennox (2016), we find evidence that suggests the restrictions
had a chilling effect on the provision of tax service by audit firms.
Descriptive Statistics
Table 3 presents descriptive statistics for the treatment, unmatched, and matched control
groups, along with t-statistics for differences in means. Panels A, B, and C aggregate the
observations into the pre-event, event, and post-event periods, respectively. In the pre-event period
(Panel A), significant differences between the treatment and control groups exist in the variable of
interest, TAXAQ. The treatment group has a significantly lower tax accrual quality (TAXAQ)
compared to both control groups (p-values < 0.05 or better). By contrast, the differences between
treatment and control groups for TAXAQ become smaller during the event window in Panel B (p-
value < 0.10 only for the unmatched control group) and became statistically insignificant in the
post-event period in Panel C. These statistics suggest that the firms that significantly reduced
APTS in response to the PCAOB restrictions (i.e., the treatment group firms) experienced a larger
improvement in tax accrual quality after the restrictions were imposed relative to those firms that
did not significantly reduce APTS.
Additionally, Panels A, B, and C of Table 3 show significant differences in company
characteristics (from APTS_LAST to TAX_LOSS) exist between the treatment and unmatched
control groups during the pre-event, event, and post-event windows.8 For the treatment group that
8 The variables are measured in the last year of the pre-event window (t – 1), the first year of the event window (t), and the first year of the post-event window (t + 1).
23
significantly reduced APTS fee when the PCAOB restrictions were introduced, the mean pre-event
APTS fee (APTS_LAST) was approximately $73,865 compared with $6,974 for the unmatched
control group indicating a concern for regulators. The treatment group also pays higher other NAS
fees (ΔOTHER), is more likely to trade on an exchange (EXCHANGE), engage KPMG as auditor
(KPMG), report uncertain tax positions (UTB_EST), discontinued operations (DISC), foreign
taxable income (FOREIGN), and tax losses (TAX_LOSS).
These differences are lessened after implementing the propensity score matched-pair
approach. As shown in Panels A, B, and C of Table 3, the treatment and matched control groups
are similar in most respects.9 Although the differences in APTS_LAST and KPMG are statistically
significant, they are economically small. The treatment group also tends to be smaller in size than
the matched control group, while they are slightly larger compared to the unmatched control group.
The variation in size might not be a problem as our sample is comprised of large firms. For all
analyses reported in the paper, we use the full sample as well as the matched sample and obtained
similar inferences.
<Insert Table 3 about here>
Implementing the Propensity Score Matched-Pair Approach
As reported in Table 3, we create a matched control group that is comparable to the
treatment group based on observables at the time the PCAOB rules were introduced (i.e., t). We
present the results from the propensity-score model (Eq. (2)) for the probability of a significant
APTS reduction during the event period in Table 4. APTS_LAST is pre-event APTS fee and other
9 As discussed in the next section, we match treatment and control companies based on covariates measured at time t and as such firms look more similar during the event time frame. Our inferences remain similar when we implement the matching algorithm based on covariates measured at time t - 1.
24
variables are measured at time period t. Model 1 reports the results of the logistic regression using
the indicator variable REDUCE as the dependent variable. Model 2 reports an OLS regression
using the continuous variable for the magnitude of the reduction %APTS_DOWN as the dependent
variable. Consistent with Lennox (2016),10 we find that companies that spent more on APTS in
the pre-event period are more likely to significantly reduce APTS purchasing by a greater
magnitude during the transition window. Further, the likelihood and magnitude of the APTS
reduction is positively associated with the decline in other NAS fees, exchange listing, and KPMG,
and negatively associated with (other) Big 4 auditors. Based on the predicted probability of
REDUCE from the logistic regression, we identify matched-pairs with the smallest propensity-
score differences and combine the event period data with pre and post-event data of the same pair
of companies (as shown previously in Table 3).
<Insert Table 4 about here>
V. RESULTS
Difference-in-Difference Descriptive Statistics
Table 5 presents the differences between treatment and control groups in tax accrual quality
across the pre and post-event periods. Specifically, the treatment group is compared to the
unmatched (matched) control group in Panel A (Panel B). The tax accrual quality for the treatment
group increased 0.0027 while the increase for the unmatched (matched) control group was 0.0006
(0.0009). The between-group difference in the tax accrual quality change is significant at the 5
10 We note Lennox (2016), Table 3, is based on the signed percentage reduction in APTS, but we examine both the likelihood and the magnitude of a significant APTS reduction and therefore the predicted signs for the coefficients are the opposite of those reported in Lennox (2016). With the exception of ASSETS, all estimated coefficients are consistent with Lennox (2016) in predicted directions. Our focus on tax accrual quality may potentially bias our sample to larger companies with longer time series of data which might explain the difference in the estimated coefficient on ASSETS.
25
percent level for the unmatched control group and at the 10 percent level for the matched control
group. These results are consistent with the descriptive statistics in Table 3. These findings are
contrary to the null hypothesis, suggesting that the treatment group had a larger improvement in
tax accrual quality over the sample period when compared to the two control groups.
<Insert Table 5 about here>
The Impact of the PCAOB Restrictions on Tax Accrual Quality
Table 6 presents the results from the difference-in-difference model (Eq. (3)) using the full
sample and matched samples. H1 states the null hypothesis that there is no difference in audit
quality changes following the PCAOB 2006 restrictions between companies that significantly
reduced APTS (treatment group) compared to companies that did not significantly reduce APTS
(control group). Therefore, the interaction term on REDUCE*POST is expected to be insignificant.
We find that tax accrual quality (TAXAQ) is significantly lower for the treatment group in the pre-
event window than the control groups. The coefficient on REDUCE is negative and significant in
both the full sample and the matched sample (p-values < 10% or better). Furthermore, the
interaction term REDUCE*POST is positive and significant in both samples (p-values < 10% or
better), whereas the coefficient on POST is not statistically significant. These results suggest that
the tax accrual quality of companies that significantly reduced APTS increased significantly from
the pre- to the post-event window compared to no change for companies that did not significantly
reduce APTS. Therefore, the null hypothesis is rejected.
Based on the mean TAXAQ (-0.0119) for the treatment group in the pre-event period (Table
3, Panel A), the treatment group increased tax accrual quality by approximately 17.65 (0.0021 /
26
|0.0119| × 100) percent compared to the unmatched control group. Inferences are similar using the
matched control group for comparison.
All other significant coefficients in both the full sample and the matched sample are in the
predicted direction. As expected with the controls, the presence of employee stock options
(ESO_IND) and a tax loss (TAX_LOSS) are negatively related to TAXAQ, consistent with
complexity and judgment involved in the application of tax-related GAAP (Choudhary et al. 2016).
Overall, the results from the full and matched samples suggest an improvement in tax-
related audit quality for companies that significantly reduced APTS following the PCAOB
restrictions. The findings are consistent with the PCAOB premise that the reduction in APTS
mitigates the impact of economic dependence on audit quality.
<Insert Table 6 about here>
The Target of the PCAOB Restrictions
Panels A and B of Table 7 present the results from the Eq. (2) regressions for the
subsamples partitioned based on tax aggressiveness. In Panel A with tax-aggressive observations,
the interaction term REDUCE*POST assumes a positive and significant coefficient (0.0029; p-
value < 0.05) for the full sample. The coefficient is positive (0.0020) but not significant for the
matched sample. By contrast, in Panel B with non-tax aggressive observations, the coefficient on
REDUCE*POST is of smaller magnitude and also not statistically significant in either sample.
This provides some support that, within the group of high ETR companies who are high-potential
target companies of the PCAOB new rules, those significantly reduced APTS experienced an
improvement in tax accrual quality relative to those did not significantly reduce APTS.
27
In the bottom of Panel B, we test for a difference in the coefficient on REDUCE*POST
between tax-aggressive and non-tax aggressive observations in Panels A and B, respectively. The
difference in coefficients is statistically significant for both the full sample and the matched sample
(p-values < 0.05 or better). Compared to non-tax aggressive companies, tax-aggressive companies
experienced a greater improvement in tax accrual quality, by approximately 8.40 (0.0010 / |0.0119|
× 100) percent, following a significant reduction in APTS fees. The economic interpretation is
identical using the matched control group for comparison.
Overall, the null hypothesis for Hypothesis 2 is also rejected. The results in Table 7 suggest
that tax aggressive companies that reduced APTS experienced a significant increase in tax accrual
quality. However, no tax accrual quality change was observed for non-tax aggressive companies
that also reduced APTS. These findings are largely consistent with economic bonding, suggesting
the APTS restrictions weakened the economic bond between the auditor and client, especially in
regard to companies that likely purchased aggressive APTS. These results suggest that effects of
the PCAOB restrictions on the reporting quality of the tax account are most pronounced for the
companies that are the target of the restrictions – companies that aggressively avoid taxes with the
assistance of APTS.
<Insert Table 7 about here>
Additional Analyses
An alternative explanation for the improvement in audit quality following APTS reduction
is increased audit effort. As shown in Panel B of Table 2, the percentage of observations reporting
an increase in audit fees and the mean percentage increase in audit fees during the transition period
were 65.07 percent and 53.29 percent, respectively. The increase in audit fees coincides with the
28
changes in APTS and tax accrual quality. To examine whether our primary results are explained
by an overall increase in auditor effort to increase audit quality for all companies, we control for
the logarithm of audit fees in Eq. (3) in Table 8. Results for the full and matched sampled are
similar to those reported in Table 6.
<Insert Table 8 about here>
We conduct several additional robustness checks (results untabulated), and our inferences
remain intact. First, we delay the post-event window by one year. This sensitivity analysis serves
two purposes. Firstly, the restrictions become effective after October 31, 2006 and the one-year
lag takes into account the time needed for the new rules to influence practice. Secondly, we
compute tax accrual quality using three-year residuals from Eq. (1) and lagging by one year ensures
TAXAQ is based primarily on post-event residuals which potentially mitigates noise in our data. In
a similar vein, we also use current year residuals from estimating Eq. (1) as an alternative measure
of tax accrual quality. Second, we examine alternative tax aggressiveness measures in testing for
H2. Specifically, we combine our main tax aggressive measure (low book ETR) with either low
cash ETR, high permanent book-tax difference, or both (defined based on median values).
VI. CONCLUSION
With the intention to increase audit quality, the PCAOB (2006) Rules on Ethics,
Independence, and Tax Services prohibited accounting firms from providing aggressive tax
services to their audit clients. The PCAOB was concerned that providing aggressive tax service
places an audit firm in a position of auditing the financial statement effects of its own tax advice.
We examine the effect of the PCAOB restrictions on audit quality by focusing on the quality of
the income tax accrual as a key accounting estimate. We investigate whether companies that
29
significantly decreased APTS during the implementation of the restrictions had an improvement
in audit quality after the change compared to companies that did not significantly reduce APTS.
We also investigate whether companies associated with aggressive tax services are more likely to
experience an improvement in audit quality following the reductions in APTS.
Our findings indicate that audit quality increased following the PCOAB 2006 restrictions
consistent with a reduction in economic bonding between the auditor and client. Companies that
reduced APTS experienced an improvement in tax accrual quality after the change compared to
companies that did not significantly reduce APTS. The improvement in audit quality stems
primarily from companies associated with aggressive tax services. Specifically, tax-aggressive
companies that reduced APTS experienced an increase in tax accrual quality after the change, but
non-tax aggressive companies that also reduced APTS did not experience an increase in tax accrual
quality after the change. Our results are both statistically and economically significant.
Overall, our results suggest that the restrictions on auditors’ ability to devise and offer
aggressive tax avoidance schemes to their clients improved financial reporting quality specific to
the tax account. The effectiveness of the PCAOB restrictions on audit quality is investigated by
few studies. Therefore, our study fills this void by using a tax specific measure of audit quality,
tax accrual quality, to specifically examine the target of the PCAOB restrictions—audit clients that
are associated with aggressive tax services. Our study confirms and contributes to the research on
economic bonding, audit quality, tax accrual quality, and tax aggressiveness, and is useful in
current policy debates.
30
APPENDIX: Variable Definitions
Variable Definition TAXAQ Tax accrual quality, measured as the standard deviation of the residuals for
t-2 to t obtained from estimating Eq. (1) by industry and year, multiplied by -1, so larger values indicate better tax accrual quality. A minimum of 20 observations per industry-year is required for estimating Eq. (1). TAXACCi,t= α0 + α1CTPi,t-1 + α2CTPi,t + α3CTPi,t+1 + α4∆DTLi,t +
α5∆DTAi,t + εi,t (1) where:
TAXACC Total tax accrual, defined as TTE – CTP
TTE Total tax expense, scaled by total assets CTP Cash taxes paid related to income taxes, scaled by total
assets C_DTL_LT Change in long-term deferred tax liability, scaled by total
assets C_DTA_LT Change in long-term deferred tax asset, scaled by total
assets %APTS_DOWN The absolute percent reduction in APTS during the transition period,
measured as│(APTSi,t+1 – APTSi,t-1)/APTSi,t-1 │if APTSi,t+1 < APTSi,t-1, and 0 if APTSi,t+1 ≥ APTSi,t-1, where t + 1 denotes the post-event period and t - 1 denotes the pre-event period
REDUCE 1 if %APTS_DOWN is at least 75%, and zero otherwise APTS_LAST The logarithm of APTS fees in the pre-event period ∆OTHER The absolute percentage reduction in NAS fees less tax service fees during
the transition period SIZE The nature logarithm of total assets EXCHANGE 1 if the company is listed on a stock exchange and zero otherwise BIG4 1 if the company is audited by one of the Big 4 accounting firms (PwC,
EY, Deloitte, or KPMG), and zero otherwise KPMG 1 if the company is audited by KPMG, and zero otherwise LOSS Loss is an indicator variable where 1 = net loss after taxes, and 0 otherwise. TAX_LOSS Indicator variable for the presence of a tax loss, coded as 1 if tax expense <
0, and zero otherwise POST 1 for the one-year period in the post-event window (fiscal years beginning
after October 31, 2006), and zero for the one year period in the post-event window (fiscal years ended before July 26, 2005)
UTB Predicted value of unrecognized tax benefits, based on coefficient estimates from Rego and Wilson’s (2012) Equation (1): UTB_EST = -.00010072 + (.00648 * PTROA) + (.00078288 * SIZE) +
(.00601 * FOREIGN) + (.06494 * RD) + (.00080232 * LEV)
31
+ (.00521 * SGA) + (.0000003193495 * MTB) - (.00176 * SALES_GR)
where:
PTROA Pre-tax return on assets, calculated as pretax income
scaled by prior year total assets
FOREIGN Indicator variable for the presence of foreign operations, coded as 1 for non-zero foreign tax expense, and zero otherwise
RD Research and development expenses, scaled by prior year
total assets
LEV Total debt divided by total assets
SGA Selling, general and administrative expenses, scaled by
prior year total assets
MTB Market to book ratio
SALES_GR Annual percent change in sales ESO_IND Indicator variable for industries likely to issue ESOs, coded as 1 if firm
operates in an industry with potentially large tax deductions from the exercise of options (Industry SIC codes 30-39 and 70-89), and zero otherwise
DISC Indicator variable for the presence of large discontinued operations or extraordinary items, coded as 1 if the absolute value of discontinued operations and extraordinary items > 1 percent of sales, and zero otherwise
PTBI_VOL Standard deviation of pre-tax book income scaled by total assets for t - 5 to t
AGGRESSIVE 1 if a company has either a below-median book effective tax rate (ETR), and zero otherwise
ETR Effective tax rate, defined as total tax expense (TXT) divided by pre-tax book income (PI) less special items (SPI). ETRs with negative denominators are deleted. The remaining non-missing ETRs are winsorized (reset) so that the largest observation is equal to 1 and the smallest is equal to 0
32
REFERENCES
Alsadoun, N., V. Naiker, F. Navissi, and D. Sharma. 2018. Auditor-provided tax nonaudit services and the implied cost of equity capital. Auditing: A Journal of Practice & Theory 37 (3): 1-24.
Armstrong, C., A. Jagolinzer and D. Larker. 2009. Chief executive officer equity incentives and accounting irregularities. Journal of Accounting Research 48: 225-271.
Baugh, M., J. Boone, I. Khurana, and K. Raman. 2019. Did the 2005 Deferred Prosecution Agreement adversely impact KPMG’s audit practice? Auditing: A Journal of Practice & Theory 38 (1): 77-102.
Beardsworth, T., and J. Browning. 2019. U.K. lawmakers demand breakup of Big Four accounting firms. Bloomberg News (April 3). Available at: https://www.accountingtoday.com/articles/uk-lawmakers-demand-breakup-of-big-four-accounting-firms. Last accessed June 9, 2019
Brown, D., S. Shu, B. Soo, and G. Trompeter. 2013. The Insurance Hypothesis: An examination of KPMG’s audit clients around the investigation and settlement of the tax shelter case. Auditing: A Journal of Practice & Theory 32 (4): 1-24.
Choudhary, P., A. Koester, and T. Shevlin. 2016. Measuring income tax accrual quality. Review of Accounting Studies 21 (1): 89-139
Christensen, B., A. Olson, and T. Omer. 2015. The role of audit firm expertise and knowledge spillover in mitigating earnings management through the tax accounts. The Journal of the American Taxation Association 37 (1): 3-36.
Cook, K., R. Huston, and T. Omer. 2008. Earnings management through effective tax rates: The effects of tax planning investment and the Sarbanes-Oxley Act of 2002. Contemporary Accounting Research 25 (2): 447-471.
DeAngelo, L. 1981. Auditor Independence. “Low Balling” and disclosure regulation. Journal of Accounting and Economics 3 (2): 113-127.
DeFond, M., and J. Zhang. 2014. A review of archival auditing research. Journal of Accounting and Economics 58 (2-3). 275–326.
DeFond, M., D. H. Erkens, and J. Zhang. 2016. Do client characteristics really drive the Big N audit quality effect? New evidence from propensity score matching. Management Science 63 (11): 3628-3649.
De Simone, L., M. Ege, and B. Stomberg. 2015. Internal control quality: the role of auditor provided tax services. The Accounting Review 90 (4): 1469-1496.
Donohoe, M., and W.R. Knechel. 2014. Does corporate tax aggressiveness influence audit pricing? Contemporary Accounting Research 31 (1): 284-308.
Finley, A., and J. Stekelberg. 2016. The economic consequences of tax service provide sanctions: Evidence from KPMG’s deferred prosecution agreement. Journal of the American Taxation Association 38 (1): 57-78.
Frank, M., L. Lynch, L., and S. Rego. 2009. Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review 84 (2): 467-496.
33
Gleason, C., and L. Mills. 2011. Do auditor-provided tax services improve the estimate of tax reserves? Contemporary Accounting Research 28 (5): 1484–1509.
Government Accountability Office (GAO). 2005. Tax shelters: Services provided by external auditors. GAO-05-171. Washington, D.C.: GAO.
Gow, I. D., G. Ormazabal, and D. J. Taylor. 2010. Correcting for cross-sectional and time-series dependence in accounting research. The Accounting Review 85: 483-512.
Hanlon, M., and S. Heitzman. 2010. A review of tax research. Journal of Accounting and Economics 50: 127-178.
Hanlon, M., G. Krishnan, and L. Mills. 2012. Audit fees and book-tax differences. Journal of the American Taxation Association 34 (1): 55–86.
Harris, S. 2014. The rise of advisory services in audit firms. Public Company Accounting Oversight Board at the Practicing Law Institute 12th Annual Directors’ Institute on Corporate Governance (November 24). Available at: https://pcaobus.org/News/Speech/Pages/11242014_Harris.aspx.
Kinney W., Z. Palrose, and S. Scholz. 2004. Auditor independence, non-audit services, and restatements: Was the U.S. government right? Journal of Accounting Research 42 (3): 561-588.
Krishnan, G. and G. Visvanathan. 2011. Is there an association between earnings management and auditor-provided tax services? The Journal of the American Taxation Association 33 (2): 111-135.
Kubick, R., D. Lynch, M. Mayberry, and T. Omer. 2016. The effects of regulatory scrutiny on tax avoidance: An examination of SEC comment letters. The Accounting Review 91 (6): 1751-1780.
Kubick, T. R., G. B. Lockhart, L. F. Mills, and J. R. Robinson. 2017. IRS and corporate taxpayer effects of geographic proximity. Journal of Accounting and Economics 63 (2): 428-453.
Lawrence, C., and M. Minutti-Meza, and Z. Zhang. 2011. Can big 4 and non-big 4 differences in audit-quality proxies be attributed to client characteristics? The Accounting Review 86: 259-288.
Lennox, C., 2016. Did the PCAOB’s restrictions on auditor’s tax services improve audit quality? The Accounting Review 91 (5):1493-1512.
Lennox, C., J. R. Francis, and Z. Wang. 2012. Selection models in accounting research. The Accounting Review 27: 209-247.
Maydew, E., and D. Shackelford. 2007. The changing role of auditors in corporate tax planning. In Taxing Corporate Income in the 21st Century, 307–337. Cambridge, UK: Cambridge University Press.
McGuire, S., T. Omar, and D. Wang. 2012. Tax avoidance: Does tax-specific industry expertise make a difference? The Accounting Review 87: 975-1003.
Omer, T., J. Bedard, and D. Falsetta. 2006. Auditor-provided tax services: The effects of a changing regulatory environment. The Accounting Review 81 (5): 1095-1117.
34
Petersen, M. A. 2009. Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies 22 (1): 435-480.
Permanent Subcommittee on Investigations (PSI). 2005. The Role of Professional Service Firms in the U.S. Tax Shelter Industry. Prepared by United States Senate The Permanent Subcommittee on Investigations of the Committee on Homeland Security and Government Affairs. Washington, D.C.: PSI.
Permanent Subcommittee on Investigations (PSI). 2014. Caterpillar’s Offshore Tax Strategy. Prepared by United States Senate The Permanent Subcommittee on Investigations of the Committee on Homeland Security and Government Affairs. Washington, D.C.: PSI.
Public Company Accounting Oversight Board (PCAOB). 2004. Auditor Independence and Tax Services Roundtable Transcript. (July 14). Available at: http://pcaobus.org/Rules/Rulemaking/Docket017/2004-07-14_Roundtable_Transcript.pdf
Public Company Accounting Oversight Board (PCAOB). 2005. Ethics and Independence Rules Concerning Independence, Tax Services, and Contingent Fees. PCAOB Release No. 2005-014. PCAOB Rulemaking Docket Matter No. 017. (July 26). Available at: https:// pcaobus.org//Rulemaking/Docket017/2005-07-26_Release_2005-014.pdf
Public Company Accounting Oversight Board (PCAOB). 2006. SEC approves PCAOB rules on auditor ethics, independence and tax services. News releases. (April 21). Available at: https://pcaobus.org/News/Releases/Pages/04212006_SECApprovesPCAOBRules.aspx
Rego, S., and R. Wilson. 2012, Equity risk incentives and corporate tax aggressiveness. Journal of Accounting Research 50 (3): 775-810.
Robinson, D. 2008. Auditor independence and auditor-provided tax service: Evidence from going-concern audit opinions prior to bankruptcy filings. Auditing: A Journal of Practice & Theory 27 (2): 31-54.
Securities and Exchange Commission (SEC). 2003. Strengthening the Commission’s Requirements Regarding Auditor Independence. Release No. 33-8183; 34-47265; 35-27642; IC-25915; IA-2103, FR-68, File No. S7-49-02. Washington, D.C.: SEC.
Seetharaman, A., Y. Sun, and W. Wang. 2011. Tax related financial statement restatements and auditor-provided tax services. Journal of Accounting, Auditing & Finance 26 (4): 677-698.
Simunic, D. 1984. Auditing, consulting, and auditor independence. Journal of Accounting Review 82 (4): 679-702.
Weber, J., M. Willenborg, and J. Zhang. 2008. Does auditor reputation matter? The case of KPMG Germany and ComROAD AG. Journal of Accounting Research 46: 941-972.
35
Figure 1: Timeline
Dec 16, 2003 July 26, 2005 Oct 31, 2006
PCAOB announced its adoption of Rules
3521, 3522, and 3523.
Rules 3521, 3522, and 3523 became effective
for SEC registrants.
t-1 Pre-event window: fiscal years
ending before July 26, 2005
t Event window:
fiscal years beginning or
ending July 26, 2005 through Oct 31, 2006
t+1 Post-event window: fiscal years
beginning after October 31, 2006
Mandatory disclosure of APTS fees
36
TABLE 1 Sample Selection
Audit Analytics Compustat Merged
Firm-year observations with Compustat data for estimating Eq. (1) (i.e., TAXACC, CTPt-1, CTP, CTPt+1, C_DTL_LT, and C_DTA_LT) between 2003 and 2009 50,764 Less: Firm-years with less than 3 years of consecutive data
(between [t-2, t]) and industry-years with less than 20 observations for estimating Eq. (1)
(10,505) Firm-years with TAXAQ available for analysis 40,259 Firm-year observations with Audit Analytics audit fee and APTS data between 2002 and 2009 (Table 2, Panel A) 63,891 Less: Fiscal Year 2002 firm-years (6,753)
Firm-years with Audit Analytics audit fee and APTS data between 2003 and 2009 57,138 Less: Firm-years not in the intersection of Compustat and
Audit Analytics (26,054) (9,175) Firm-years with TAXAQ and required fee data 31,084 Less: Firm-years with missing data for control variables in
Eqs. (2) & (3) (9,733)
Firm-years not in the one-year period before the event window or the one-year period after the event window (14,658)
Firm-years without corresponding data in both the pre-event and post-event windows (1,945)
Firm-years in the treatment sample (565 unique firms * 2 periods) 1,130 Firm-years in the control sample (1,809 unique firms * 2 periods) 3,618 Total firm-years in the full sample 4,748 Less: Firm-years not identified as the closest match for the
treatment group using propensity-score matching (1,244 unique firms * 2 periods) (2,488)
Total firm-years in the propensity-score matched-pair sample (565 unique firms * 2 groups * 2 periods) 2,260
37
TABLE 2 Fees By Year and Window
Panel A: Changes in APTS Fees, Other NAS Fees, and Audit Fees by Year (n = 63,891)
Year Fees for auditor provided
tax services ($000) Fees for other non-audit
services ($000) Fees for audits ($000)
%change %change %change
2002 246.46 327.10 598.88
2003 197.87 -20% 151.57 -54% 562.06 -6%
2004 173.36 -12% 140.40 -7% 975.18 74%
2005 131.88 -24% 123.04 -12% 1020.06 5%
2006 123.00 -7% 134.12 9% 1094.48 7%
2007 140.58 14% 150.17 12% 1135.31 4%
2008 141.33 1% 136.64 -9% 1163.07 2%
2009 149.16 6% 125.08 -8% 1146.55 -1%
38
TABLE 2 (continued)
Panel B: Changes in APTS Fees, Other NAS Fees, and Audit Fees by Event Window (n = 36,809)
Percentage of observations
with a reduction in
fees
Mean percentage
reduction in fees
Observations APTS Fees Pre-event data period 38.32% -17.70% 13,780 Event window 40.14% -21.01% 9,262 Post-event data period 31.87% -16.22% 13,767
36,809
Other NAS Fees Pre-event data period 38.59% -22.98% 13,780 Event window 36.40% -22.84% 9,262 Post-event data period 33.04% -20.17% 13,767
36,809
Percentage of observations
with an increase in
fees
Mean percentage increase in
fees
Audit Fees Pre-event data period 76.95% 66.59% 13,780 Event window 65.07% 53.29% 9,262 Post-event data period 60.14% 34.50% 13,767
36,809 Note: Panel A includes all observations in Audit Analytics with non-missing audit fee and APTS data between 2002 and 2009. Panel B includes only observations that fall within the pre-event period, event window, and post-event period. The pre-event period comprises fiscal years ending December 16, 2003 through July 25, 2005. The event window comprises fiscal years beginning or ending in the period from July 26, 2005 through October 31, 2006. The post event period comprises fiscal years beginning after October 31, 2006 (ending October 31, 2007 through July 31, 2009).
39
TABLE 3 Descriptive Statistics
Panel A: The Pre-Event Period
Treatment Group
(N = 565)
Unmatched Control Group
(N = 1,809)
Matched Control Group
(N = 565)
Treatment vs. Unmatched
Control Treatment vs.
Matched Control Variable Mean Std Dev Mean Std Dev Mean Std Dev t-statistics t-statistics
TAXAQ -0.0138 0.0159 -0.0115 0.0148 -0.0121 0.0138 -3.09 *** -2.64 ** %APTS_DOWN 0.9593 0.0720 0.1270 0.2226 0.1836 0.2490 137.66 *** 71.15 *** APTS_LAST 11.2188 1.6250 8.8545 4.8953 11.5068 2.0200 17.66 *** -2.72 *** ΔOTHER 0.5451 0.4984 0.4638 0.4988 0.5204 0.5000 3.39 *** 0.83 SIZE 5.8422 2.0591 5.6587 2.4267 6.1424 2.4583 1.77 * -2.22 ** EXCHANGE 0.8496 0.3578 0.7927 0.4055 0.8531 0.3543 3.19 *** -0.17
BIG4 0.7681 0.4224 0.6639 0.4725 0.7062 0.4559 4.97 *** 2.37 ** KPMG 0.2071 0.4056 0.1443 0.3515 0.1894 0.3922 3.31 *** 0.75
LOSS 0.2726 0.4457 0.2388 0.4265 0.2212 0.4154 1.59
2.00 ** UTB_EST 0.0116 0.0067 0.0109 0.0069 0.0114 0.0064 2.03 ** 0.35
ESO_IND 0.4903 0.5003 0.4798 0.4997 0.4832 0.5002 0.43
0.24
DISC 0.0673 0.2507 0.0453 0.2081 0.0513 0.2209 2.08 ** 1.13
FOREIGN 0.4442 0.4973 0.3836 0.4864 0.4531 0.4982 2.57 *** -0.30
PTBI_VOL 0.1651 0.3504 0.2017 0.6326 0.1651 0.6995 -1.75 * -0.14
TAX_LOSS 0.1080 0.3106 0.0697 0.2546 0.0850 0.2791 2.67 *** 1.31
40
Table 3 - Continued
Panel B: The Event Period
Treatment Group
(N=565)
Unmatched Control Group
(N=1,809)
Matched Control Group
(N=565)
Treatment vs Unmatched
Control
Treatment vs
Matched Control
Variable Mean Std Dev Mean Std Dev Mean Std Dev t-statistics t-statistics TAXAQ -0.0119 0.0131 -0.0109 0.0130 -0.0117 0.0132 -1.78 * -0.38
%APTS_DOWN 0.9593 0.0720 0.1270 0.2226 0.1836 0.2490 137.66 *** 71.15 *** APTS_LAST 11.2188 1.6250 8.8545 4.8953 11.5068 2.0200 17.66 *** -2.72 *** ΔOTHER 0.5451 0.4984 0.4638 0.4988 0.5204 0.5000 3.39 *** 0.83 SIZE 6.0373 2.0557 5.8540 2.4361 6.1432 2.4321 2.53 ** -1.05
EXCHANGE 0.8620 0.3451 0.8016 0.3988 0.8459 0.3612 5.01 *** 1.02
BIG4 0.6652 0.4721 0.6334 0.4819 0.6479 0.4747 2.00 ** 0.35
KPMG 0.1976 0.3983 0.1327 0.3393 0.1639 0.3704 5.01 *** 1.99 ** LOSS 0.2528 0.4348 0.2318 0.4220 0.2383 0.4262 1.45
0.77
UTB_EST 0.0120 0.0067 0.0114 0.0072 0.0116 0.0067 2.58 ** 1.23
ESO_IND 0.4935 0.5002 0.4857 0.4999 0.5005 0.5003 0.46
-0.32
DISC 0.0638 0.2446 0.0499 0.2178 0.0579 0.2337 1.74 * 0.56
FOREIGN 0.4823 0.4999 0.4187 0.4934 0.4667 0.4992 3.80 *** 0.71
PTBI_VOL 0.1960 1.2017 0.2740 1.7807 0.2021 1.3926 -1.70 * -0.11
TAX_LOSS 0.0915 0.3150 0.0499 0.2296 0.0820 0.2745 3.85 *** 0.76
41
Table 3 - Continued
Panel C: The Post-Event Period
Treatment Group (N=545)
Unmatched Control Group
(N=1,809)
Matched Control Group
(N=565)
Treatment vs Unmatched
Control
Treatment vs Matched Control
Mean Std Dev Mean Std Dev Mean Std Dev t-statistic t-statistic TAXAQ -0.0113 0.0121 -0.0109 0.0119 -0.0112 0.0104 -0.63
0.14
%APTS_DOWN 0.9593 0.0720 0.1270 0.2226 0.1836 0.2490 137.66 *** 71.15 *** APTS_LAST 11.2188 1.6250 8.8545 4.8953 11.5068 2.0200 17.66 *** -2.72 *** ΔOTHER 0.5451 0.4984 0.4638 0.4988 0.5204 0.5000 3.39 *** 0.83 SIZE 6.1027 2.1138 5.9897 2.4392 6.3957 2.5198 1.07
-2.12 **
EXCHANGE 0.8496 0.3578 0.7927 0.4055 0.8531 0.3543 3.19 *** -0.17
BIG4 0.6159 0.4868 0.6075 0.4884 0.6496 0.4775 0.36
-1.17
KPMG 0.1841 0.3879 0.1244 0.3301 0.1805 0.3850 3.30 *** 0.15
LOSS 0.2920 0.4551 0.2631 0.4405 0.2885 0.4535 1.33
0.13
UTB_EST 0.0120 0.0063 0.0115 0.0068 0.0112 0.0064 1.65 * 0.03
ESO_IND 0.4903 0.5003 0.4798 0.4997 0.4832 0.5002 0.43
0.24
DISC 0.0832 0.2764 0.0569 0.2318 0.0885 0.2843 2.04 ** -0.32
FOREIGN 0.4938 0.5004 0.4411 0.4967 0.5168 0.5002 2.19 ** -0.77
PTBI_VOL 0.1641 0.9273 0.2143 1.2193 0.1378 0.5135 -1.04
0.06
TAX_LOSS 0.1115 0.3150 0.0691 0.2537 0.1044 0.3061 2.92 ** 0.38 Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles. ***, **, and * Denote significant differences across firms at the 1 percent, 5 percent, and 10 percent levels (two-tailed), respectively.
42
TABLE 4 Determinants of Significant Reductions in APTS (n = 2,374 companies)
REDUCEi = β0 + β1APTS_LASTi,t + β2∆OTHERi,t + β3SIZEi,t + β4EXCHANGEi,t + β5BIG4i,t + β6KPMGi,t + β7LOSS +
ΣPROXY_CONTROLS + Σindustries + µi
(2)
Model 1: REDUCE Model 2:
%APTS_DOWN Variable Prediction Coefficient p-value Coefficient p-value
INTERCEPT ? -3.337 *** <.0001 -0.032 0.3642 APTS_LAST + 0.245 *** <.0001 0.037 *** <.0001 ∆OTHER + 0.244 ** 0.0176 0.049 ** 0.0016 SIZE + -0.098 ** 0.0088 -0.002 0.7114 EXCHANGE + 0.529 ** 0.0012 0.051 ** 0.0251 BIG4 - -0.695 *** <.0001 -0.099 *** <.0001 KPMG + 0.685 *** <.0001 0.112 *** <.0001 LOSS + 0.191 0.1487 0.053 ** 0.0075 UTB_EST ? 5.676
0.5789 1.067 0.4692
ESO_IND ? -0.162
0.3640 -0.058 ** 0.0291 DISC ? 0.367 * 0.0670 0.066 ** 0.0370 FOREIGN ? -0.146
0.3174 0.002 0.9410
PTBI_VOL ? 0.008
0.8609 0.009 0.1295 TAX _LOSS ? 0.487 ** 0.0068 0.048 0.1014 Industry FE YES ` YES p-Value <.0001 <.0001 Pseudo R2 / R2 15.60% 17.20% Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles. Model 1 reports the results of the logistic regression using the indicator variable, REDUCE, as the dependent variable, whereas Model 2 reports an OLS regression using the continuous variable, %APTS_DOWN, as the dependent variable. ***, **, and * Denote significance at the 1 percent, 5 percent, and 10 percent levels (two-tailed), respectively.
43
TABLE 5
Difference-in-Differences Analysis of the Change in Tax Accrual Quality (TAXAQ) surrounding the PCAOB Restrictions
Panel A: Treatment Group vs. Unmatched Control Group
Treatment Group Unmatched Control Group Difference (A) (B) (A) – (B) Pre-Event (I) -0.0138 -0.0115 0.0023 ** Post-Event (II) -0.0111 -0.0109 0.0002 Difference (II) – (I) 0.0027 *** 0.0006 0.0021 **
Panel B: Treatment Group vs. Matched Control Group
Treatment Group Matched Control Group Difference
(A) (B) (A) – (B) Pre-Event (I) -0.0138 -0.0121 0.0017 * Post-Event (II) -0.0111 -0.0112 -0.0001 Difference (II) – (I) 0.0027 *** 0.0009 0.0018 * Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles. ***, **, and * Denote significance at the 1 percent, 5 percent, and 10 percent levels (two-tailed), respectively.
44
TABLE 6 Regressions of Changes in Tax Accrual Quality Following PCAOB Restrictions
(Dependent Variable = TAXAQ)
TAXAQi,t = δ0 + δ1REDUCEi + δ2POSTi,t + δ3REDUCEi*POSTi,t + δ4 UTB_ESTi,t + δ5
ESO_INDi,t + δ6 DISCi,t + δ7FOREIGNi,t + δ8PTBI_VOLi,t + δ9TAX_LOSSi,t
+ δ10SIZEi,t + δ11BIG4i,t + δ12 KPMGi,t + Σindustries + εi,t
(3)
Full Sample (N=4,748)
Matched Sample (N=2,280)
Variable Prediction Coefficient t value Coefficient t value
Intercept ? -0.0124 *** -13.62 -0.0142 *** -10.28 REDUCE - -0.0017 ** -2.43 -0.0016 * -1.95 POST + 0.0007 1.60 0.0011 1.57 REDUCE*POST ? 0.0021 ** 2.36 0.0019 * 1.81 UTB_EST - 0.1028 ** 2.14 0.0791 0.95 ESO_IND - -0.0025 ** -2.65 -0.0022 ** -2.12 DISC - -0.0008 -0.85 -0.0011 -0.94 FOREIGN - -0.0008 -1.40 -0.0001 -0.07 PTBI_VOL - 0.0001 0.23 -0.0006 -0.62 TAX_LOSS - -0.0053 *** -6.65 -0.0054 *** -5.53 SIZE + 0.0000 0.23 0.0003 1.43 BIG4 ? 0.0014 ** 2.38 0.0017 ** 2.13 KPMG ? 0.0000 -0.03 0.0004 0.59 Industry FE YES YES p-Value <.0001 <.0001 R2 11.25% 13.18% Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles. ***, **, and * Denote significance at the 1 percent, 5 percent, and 10 percent levels (two-tailed), respectively.
45
TABLE 7 Regressions of Changes in Tax Accrual Quality Following PCAOB Restrictions:
The Targeted Companies (Dependent Variable = TAXAQ) Panel A: Aggressive = 1
Full Sample Matched Sample
N = 2,374 N = 1,130
Variable Prediction Coefficient t value Coefficient t value Intercept ? -0.0093 *** -8.00 -0.0121 *** -6.40 REDUCE - -0.0025 ** -2.43 -0.0017 -1.34 POST + 0.0008 1.19 0.0017 1.61 REDUCE*POST ? 0.0029 ** 2.14 0.0020 1.24 UTB_EST - 0.1798 ** 3.38 0.1193 1.22 ESO_IND - -0.0009 -0.74 0.0000 0.01 DISC - -0.0006 -0.52 -0.0007 -0.47 FOREIGN - -0.0018 ** -2.21 -0.0007 -0.60 PTBI_VOL - -0.0001 -0.31 -0.0008 -0.78 TAX_LOSS - -0.0053 *** -5.88 -0.0061 *** -5.45 SIZE + -0.0004 ** -2.25 -0.0001 -0.55 BIG4 ? 0.0015 * 1.79 0.0020 1.58 KPMG ? -0.0001 -0.12 0.0002 0.22
Industry FE YES YES p-Value <.0001 <.0001 R2 9.32% 11.71%
46
TABLE 7 (continued)
Panel B: Aggressive = 0
Full Sample Matched Sample N = 2,374 N = 1,130
Variable Prediction Coefficient t value Coefficient t value Intercept ? -0.0165 *** -11.98 -0.0161 *** -8.01 REDUCE - -0.0007 -0.78 -0.0013 -1.24 POST + 0.0005 0.96 0.0009 1.03 REDUCE*POST ? 0.0010
0.88 0.0010 0.75
UTB_EST - -0.5969 *** -4.30 -0.3155 * -1.70 ESO_IND - -0.0046 ** -3.14 -0.0042 ** -2.77 DISC - -0.0024 -1.37 -0.0034 -1.62 FOREIGN - 0.0050 *** 4.57 0.0035 ** 2.50 PTBI_VOL - -0.0098 ** -3.08 -0.0108 ** -2.43 TAX_LOSS - -0.0054 ** -3.08 -0.0041 * -1.75 SIZE + 0.0013 *** 6.72 0.0011 *** 4.13 BIG4 ? 0.0012 1.59 0.0013 1.23 KPMG ? -0.0000 0.00 0.0003 0.48
Industry FE YES YES
p-Value <.0001 <.0001
R2 23.86% 23.85% Difference in Coefficients: Aggressive = 1 vs. Aggressive = 0:
REDUCE*POST 0.0019 ** 0.0010 ** F-Value 5.23 4.39 Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles.
***, **, and * Denote significance at the 1 percent, 5 percent, and 10 percent levels (two-tailed), respectively.
47
TABLE 8 Regressions of Changes in Tax Accrual Quality Following PCAOB Restrictions:
Controlling for Audit Effort (Dependent Variable = TAXAQ)
Full Sample
Matched Sample
Variable Prediction Coefficient t value Coefficient t value
Intercept ? 0.0096 *** 3.65 -0.0068 -1.46 REDUCE - -0.0017 ** -2.43 -0.0016 * -1.92 POST + 0.0008 * 1.94 0.0012 * 1.71 REDUCE*POST ? 0.0021 ** 2.25 0.0019 * 1.81 UTB_EST + 0.0015 *** 9.45 0.0006 ** 1.99 ESO_IND + 0.0006
1.06 0.0019 ** 2.37
DISC ? 0.0006
1.12 0.0004 0.58 FOREIGN - 0.1143 ** 2.37 0.0910 1.08 PTBI_VOL - -0.0039 *** -5.94 -0.0021 ** -2.04 TAX_LOSS - -0.0013 ** -2.21 0.0002 0.21 SIZE - 0.0003
0.70 -0.0005 -0.49
BIG4 - -0.0053 *** -6.68 -0.0054 *** -5.47 KPMG - -0.0007
-0.74 -0.0010 -0.86
LNAU_FEE ? -0.0021 *** -7.71 -0.0008 -1.59 Industry FE YES YES p-Value <.0001 <.0001 R2 9.10% 13.30%
Note: All variables are defined in Appendix and continuous variables are winsorized at the 1st and 99th percentiles. ***, **, and * denote significance at the 1%, 5%, and 10% level (two-tailed), respectively.