The influence of auditor provided tax planning and tax

compliance services in the “tax transparency era”

James A. Chyz* University of Tennessee

Ronen Gal-Or

Northeastern University

Vic Naiker Monash University

Divesh Sharma

Kennesaw State University

ABSTRACT This study closely examines the engagement between companies and their auditors in a period where the tax function has evolved into a risk management center. Our empirical analysis focuses on the associations between auditor-provided tax compliance and tax planning services and three primary tax outcomes noted in a recent survey by the Tax Council Policy Institute and Ernst & Young 2006 Global Tax Risk Survey as particularly important in the current tax environment: effective tax rate reductions, cash tax savings, and tax risk management. Based on an extensive hand-collected sample of firms voluntarily disclosing the tax planning and compliance components of tax NAS fees, we find that tax planning and tax compliance services, and the overlap of the two services, capture different company goals and different costs faced by audit firms in the current tax environment. Collectively, our results suggest that companies paying their auditors for tax planning advice are more effective tax planners (in terms of both reductions in effective tax rates and higher cash tax savings, and, lower risk) than firms who do not engage their auditor for tax work. However, our tax rate minimization and cash tax savings results hold only when firms also engage their auditors for tax compliance work consistent with auditors seeking to minimize reputation threats. We also find evidence suggesting tax risk management can be an effective stand-alone strategy that is perceived by auditors as posing less reputation threats. Our study provides a more precise and longitudinal perspective on the role auditors play in designing and overseeing tax outcomes, and informs policy debates regarding auditor provided tax services. Keywords: tax non-audit services; tax avoidance; tax risk ____________

*Corresponding author. E-mail address: jchyz@utk.edu Phone: 865.974.1701

1

1. INTRODUCTION

Our study examines how the engagement of companies’ auditors in tax planning and tax

compliance services help to shape corporate tax outcomes. Unlike prior related research (e.g.

Cook, Huston, and Omer, 2008; McGuire et al., 2012, Cook and Omer, 2013), our study uses a

unique hand-collected dataset that is able to separate these two components from the total of

auditor provided tax service fees (APTS). This is an essential innovation because tax compliance

and tax planning are a distinct set of services offered by different professionals with unique aims

(AICPA 2011). Separating planning and compliance is especially important during our period of

study, also unique to the related research, because it coincides with unprecedented transparency

around the corporate tax function (Donohoe, McGill, and Outslay, 2014; Tax Council Policy

Institute, 2006). The so-called “age of tax transparency” (Tax Council Policy Institute, 2006) has

forced the corporate tax function to evolve from an endeavor that was solely profit seeking to one

that seriously considers risk management (Donohoe et al., 2014). This evolution is reflected in the

three tax outcomes examined in our study which, according to the Tax Council Policy Institute

and Ernst & Young 2006 Global Tax Risk Survey, are particularly important in the age of tax

transparency: effective tax rate reductions, cash tax savings, and tax risk management.1,2

Because of their unique position and relationship with companies, the involvement of a

company’s auditor in the corporate tax function can be an important determinant of tax outcomes

1 In a 2006 survey by Tax Council Policy Institute survey 50% (39%) of senior tax executives at Fortune 500 companies responded that reduced effective tax rate reduction (cash tax savings) was their most important tax management objective, while the Ernst & Young 2006 Global Tax Risk Survey suggests that a leading measures of tax department performance is tax risk management. 2 We capture the first two tax outcomes with GAAP and cash effective tax rates respectively. As of yet, there is no commonly agreed upon definition or measure of tax risk. Consistent with recent and concurrent research, we capture tax risk management through the volatility in future tax savings, i.e. “tax volatility” (Guenther, Matsunaga, and Williams, 2017; Hutchens and Rego, 2015) and “tax uncertainty” using uncertain tax benefits (UTBs) as a proxy (Dyreng, Hanlon, and Maydew, 2017; Guenther et al., 2017; Hutchens and Rego, 2015; Klassen et al., 2015; Rego and Wilson, 2012).

2

(Klassen, Lisowsky, and Mescall, 2015). It is not immediately clear however, whether effective

tax rate reductions, cash tax savings, and tax risk management are mutually exclusive or whether

they can be achieved simultaneously. It is also not clear whether these tax outcomes would be

related to auditor provided tax planning services, tax compliance services, or total APTS. Klassen

et al. (2015) postulate that threats to auditor reputation are greater when they provide substantial

tax compliance related services. This suggests that audit firms providing higher levels of tax

compliance services could be less inclined to support greater effective tax rate reductions and cash

tax savings (Klassen et al., 2015). While reputational threats are also likely relevant for auditors

providing higher levels of tax planning, auditors must manage the desire of companies that are

probably engaging their auditors to help reduce effective tax rates and generate cash tax savings.

Because reputational threats stem from auditors’ heightened sensitivity to having tax positions

overturned (Klassen et al., 2015), it is possible that auditors providing tax planning services seek

to reduce both tax levels and tax risk. How these tradeoffs manifest in tax outcomes and tax

services is an important empirical question.

Our empirical analysis is generally consistent with our predictions while also yielding a

number of interesting insights on auditors’ roles in tax outcomes deemed to be important in the

age of tax transparency. We find that after holding constant the variation in tax planning,

companies with audit firms providing higher levels of tax compliance services are no different in

effective tax rate reductions and cash tax savings than companies that do not engage their auditors

for tax services of any kind. This is consistent with reputational threats argument postulated by

Klassen et al. (2015). Turning to tax planning we find that after holding constant the variation in

tax compliance, companies with audit firms that provide higher levels of tax planning services are

associated with greater cash tax savings and effective tax rate reductions than companies that do

3

not engage their auditors for tax services of any kind. This finding suggests that in the context of

tax planning, audit firm tradeoffs favor their client’s tax burden over reputational threats. However,

as predicted above, we find some evidence that auditors mitigate reputational threats associated

with tax planning by advising on tax strategies that reduce both tax levels and tax risk. Specifically,

we find that companies with audit firms providing higher levels of tax planning services are

associated with lower tax volatility and lower uncertain tax benefit (UTB) levels than companies

that do not engage their auditors for tax services of any kind. Similar to our findings with respect

to effective tax rate reductions and cash tax savings, tax compliance services are not related to tax

risk. Overall our results suggest that companies paying their auditors for tax planning advice are

more effective tax planners (in terms of both reductions in effective tax rates and higher cash tax

savings, and, lower tax risk) than companies who do not engage their auditor for tax work. These

findings are timely because recent research has highlighted the importance of tax risk to audit

firms, companies, and taxing authorities (Guenther et al., 2017; Neuman, Omer, and Schmidt,

2016).

In additional analysis we explore whether the joint provision of tax compliance and tax

planning impacts our three sets of results. This analysis is motivated by our conversations with

Big 4 tax partners who confirm that large audit firms are less willing in the age of tax transparency

to provide advice on tax transactions having material financial reporting implications that generate

GAAP and cash tax savings unless they are also involved in the company’s compliance work. This

suggests that firms are likely to realize greater effective tax rate reductions and higher cash tax

savings when they purchase auditor provided tax planning services in conjunction with auditor

provided tax compliance services. Consistent with this prediction, we find that the effect of tax

planning on cash tax savings and effective tax rate reductions is concentrated in the set of

4

companies that purchase both tax planning and tax compliance services from their auditor. The

apparent dependence on tax planning in the presence of tax compliance is consistent with the view

that compliance services help to improve financial reporting outcomes (Chyz, Gal-Or, and Naiker,

2016) thus minimizing auditors’ reputational threats from providing tax planning services. It also

conforms to our conversations with partners at Big 4 accounting firms about the importance of tax

compliance in concert with tax planning in the age of tax transparency.

When we examine whether the provision of tax planning with and without the provision of

tax compliance is important with respect to the association between tax planning and tax risk we

find that our result emanates from the set of companies that appear to purchase auditor provided

tax planning services in isolation. Combined with the finding that tax planning services purchased

in isolation do not reduce effective tax rates or increase cash tax savings in our setting, our results

could suggest that tax planning in isolation reflects services specifically designed to minimize tax

risk for clients. The most likely candidate for such tax planning APTS is tax risk management

plans that have been developed by audit firms (Neuman, Omer, and Schmidt, 2016), which are

less likely to focus on effective rate minimization or cash tax savings.

Our study contributes to the literature in a number of ways. From a methodological

standpoint we contribute to prior studies that highlight the importance of refining proxies to better

understand the true significance of the underlying associations in the accounting and auditing

literature (e.g., Doyle, Ge, and McVay, 2007; Hennes, Leone, and Miller, 2008).3 Quantifying tax

planning and tax compliance also helps us to extend the work done by Klassen et al. (2015) by

testing the relation between tax outcomes and explicit variables that capture separate externally

3 Hennes, Leone, and (2008) provide evidence that while all restatements are associated with class action lawsuits and managerial turnover, the relationship is driven by certain types of restatements, i.e., irregularities rather than errors. Doyle et al. (2007) reveal that “company-level” rather than “account-specific” internal control weaknesses are associated with poor accruals quality.

5

sourced investments made by firms’ into their tax compliance and tax planning functions.

Accordingly we answer the call from Klassen et al. (2015) for research that sharpens our inferences

regarding links between tax practitioners and corporate tax outcomes.

Our study also helps to shed light on a somewhat unexpected set of findings in the recent

literature. A maintained view in much of the contemporary tax literature has been that tax

avoidance potentially increases a firm’s tax related risk (Rego and Wilson, 2012; Badertscher,

Katz, and Rego, 2013; Hasan, Hoi, Wu, and Zhang, 2014; Ciconte, Donohoe, Lisowsky, and

Mayberry, 2016). Contrary to this prediction Guenther et al., (2017) find little evidence of a

positive association between tax avoidance and tax volatility across a number of empirical proxies

for tax avoidance. Instead, the only statistically significant set of results in Guenther et al. (2017)

supports a negative relation between empirical proxies for tax avoidance and tax risk. This could

suggest that tax avoidance as commonly measured in the literature is comprised largely of

relatively “benign tax-advantaged investments” such as tax-exempt municipal bonds that do not

increase tax risk (Hanlon and Heitzman, 2010; Dyreng, Hanlon, and Maydew, 2016). By showing

that tax planning APTS reduce effective tax rates and improve cash tax savings while also reducing

tax risk, our results suggest that tax planning APTS is one avenue through which firms can realize

tax savings while reducing tax risk. Because firms are unlikely to pay for APTS that provide benign

tax planning, overall our results suggest that at least part of the finding in Guenther et al. (2017)

and Dyreng, Hanlon, and Maydew (2008) that tax avoidance is accomplished using strategies that

are persistent and do not increase tax risk appear attributable to the auditor’s participation in

corporate tax planning and not necessarily the use of benign tax avoidance.

The findings in our paper are also important in the context of the current regulatory

environment that has cast tax planning by auditors in a generally unfavorable light (PCAOB 2004a;

6

2004b; SEC 2006; US Treasury 2008).4 Our findings with respect to corporate tax outcomes,

combined with research that has documented generally benign impacts of tax planning in non-tax

financial reporting contexts (Chyz, et al., 2016) raises the possibility that before proposing any

further prohibitions of tax planning services, regulators should first consider the possibility that

auditor proposed tax strategies could minimize taxes in a non-abusive manner while not adversely

impacting other financial reporting outcomes.

The remainder of the paper is organized as follows. Section 2 provides the background to

the study and develops our hypotheses and Section 3 explains our sample selection. Section 4

discusses the research design. Section 5 presents the empirical results. Section 6 discusses the

additional analyses and robustness tests. Finally, Section 7 summarizes and concludes.

2. BACKGROUND AND HYPOTHESES

While Section 201 of SOX banned most types of auditor-provided non-audit services

(NAS), audit committee approved APTS remain permissible. After the enactment of SOX,

regulators (e.g., PCAOB 2004b, US Treasury 2008) continued to propose further restrictions or a

complete ban on all NAS. Some detractors argue that APTS impairs auditor independence and

negatively influences financial reporting quality. Others are concerned about the potential for risky

and abusive tax strategies. Practitioners contend that efficiencies and knowledge spillover benefits

arise when audit services and APTS are jointly provided. The controversy over the provision of

APTS has motivated several studies on the effect of these services on the tax avoidance activities

of clients. Mills et al. (1998) find that effective tax rates are decreasing in tax-related fees paid to

4 See Beale (2004) for a discussion of the role of Congress, the SEC and the media in the introduction of the Public Company Accounting Oversight Board rules to limit abusive tax strategies and examples of auditor involvement in tax shelter transactions.

7

attorneys, accountants and consultants. Omer et al. (2006) use voluntarily-disclosed auditor-

provided tax fee data from 2000 to 2002 to document an association between tax fees paid to

auditors and subsequent reductions in tax rates. Evidence also suggests that firms continue to use

tax NAS to reduce taxes in the post-SOX era (Cook and Omer 2013; Hogan and Noga 2015;

McGuire et al. 2012).

Our study is unique relative to this earlier work for two important reasons. First, we

distinguish between auditor-provided tax planning and tax compliance service fees while prior

studies are unable to separate the two. We argue that it is important to account for the fundamental

differences between tax planning and tax compliance NAS when assessing the implications of

auditor-provided tax services, because we do not know from prior studies whether tax planning

and compliance services have differential effects on effective rate reductions and cash tax savings.

The accounting profession argues that routine tax compliance services provided by the audit firm

do not promote tax avoidance (PCAOB 2004b).5 On the other hand, tax planning NAS are aimed

at maximizing after-tax wealth. Examples of legitimate tax planning activities include taking

advantage of loss carryforwards, setting up operations in low-tax foreign countries, utilizing

foreign and R&D tax credits, engaging in tax-efficient spin-offs, and minimizing merger and

acquisition tax consequences.6 Examples of somewhat aggressive or even abusive tax planning

5 According to the American Institute of Certified Public Accountants (AICPA 2011) and PCAOB (2004a), tax compliance services include the preparation of original and amended tax returns, estimation and transmittal of any related tax payments to the taxing authority at the federal, state, and local levels, signing and filing tax returns and extensions, authorized representation of clients in administrative proceedings before a taxing authority, preparing and filing payroll and sales tax returns, and preparing the returns for employee benefit and similar plans. 6 The tax planning services are wide ranging and ever expanding. The PCAOB (2004a) considers tax planning to include advice relating to executive compensation, employee benefit plans, proposed or pending tax legislation, and international tax requirements (such as trade and customs duties), advice on how to minimize liability for state and local taxes, tax consulting services to obtain tax refunds, tax planning for state and local tax credits and incentives, document searches for unclaimed property, and tax advice on tax assessments for property.

8

that potentially serves little or no business purpose can include abusive strategies such as offshore

tax sheltering and corporate inversion transactions (Lisowsky, Robinson, and Schmidt, 2013).

Second, during our sample period, companies are subjected to significantly greater

transparency with respect to the tax function. This includes: the implementation of provisions in

the Sarbanes-Oxley Act (SOX), mandated risk factor disclosure consistent with SEC Regulation

S-K Item 305(c), the detailed presentation of book-tax-differences in IRS Schedule M-3, and

additional standardized financial statement presentation and disclosure of tax uncertainty

mandated by FIN48. The complexities of APTS have arguably increased most dramatically

following the enactment of FIN 48 in 2007. FIN 48 requires firms to analyze and disclose more

information on income tax risks associated with auditor-provided tax strategies, including a tabular

reconciliation of total unrecognized tax benefits at the beginning and the end of the fiscal year.

Given the intention of the IRS to use FIN 48 disclosures as a roadmap for tax audits (KPMG 2007;

SEC 2007), larger tax reserves would attract greater regulatory scrutiny and reflect unexpected

cash outflows, interest and penalties from disallowed tax positions. Larger tax reserves could also

have a negative reputation effect on firms as they may be viewed as poor corporate citizens

unwilling to pay their fair share of taxes (Hanlon and Slemrod 2009). While the Professional Ethics

Executive Committee has determined that audit firms can assist their clients with the application

of FIN 48 without impairing independence (AICPA 2007), audit firms would still be concerned

about being held responsible for financial reporting deficiencies relating to challenged tax

positions (Donohoe and Knechel 2014). Such concerns are likely to be heightened in the presence

of tax strategies contrived through auditor-provided tax planning services. This is expected to

result in audit firms increasing their efforts in reviewing and documenting the evidence to support

their clients’ tax positions.

9

We also do not know from prior studies how tax NAS is related to tax risk. The SEC and

FASB issued regulations requiring more tax risk disclosures (e.g., Regulation S–K, Item 305(c),

SEC 2005 and FIN 48). Based on these regulatory changes and the abundance of tax-related

internal control weakness over the same period, accounting firms have since proscribed tax risk as

a top priority for audit committees (e.g., Deloitte 2011; KPMG 2013). Given the importance and

sensitivity of tax risk in recent years, KPMG (2013) recommends audit committees ensure tax

positions taken by the firm do not threaten the firm’s tax risk. As such, our study examines whether

the tax reducing initiatives proposed by the audit firm are successfully implemented without

increasing tax volatility.

Auditor-Provided Tax Services and Effective Tax Rate Reductions and Cash Tax Savings

Although there are several types of tax providers (e.g., in-house, third-party, auditor-

provided), audit firm tax providers may be able to develop more effective tax strategies because

they accumulate significant knowledge about their client’s business, internal processes, systems,

industry, and have access to extensive internal financial information. Indeed, anecdotal evidence

suggests audit firms leverage on a client’s “inside” information to provide APTS (Beale 2004;

PCAOB 2005). Cripe and McAllister (2009) report that CFOs hire their auditor as tax consultants

because they possess considerable knowledge about the client’s tax structure and are able to bring

cost savings and efficiencies to the client.

Moreover, audit firms as tax providers can draw on much broader tax specific knowledge

as they provide APTS to a relatively larger number of clients than do other providers such as large

law firms (Maydew and Shackelford 2007). To provide such services in a competitive market,

audit firms continuously invest in cutting-edge accounting, audit, and tax technologies. Coupled

with the large number of clients serviced, an audit firm possessing current tax technology may be

10

able to market itself as a tax services specialist. Maydew and Shackelford (2007) provide anecdotal

evidence to suggest that “it would take years to develop the firm-specific tax and business expertise

that currently resides among the tax consultants at the accounting firm that has conducted its audit

for decades”. As a result of being heavily involved in providing cross-border accounting, audit,

consulting and tax NAS, large audit firms also have more extensive global reach than other tax

consultants such as law firms and in-house providers (Maydew and Shackelford 2007). This may

allow large audit firms to devise more effective tax reducing strategies for their clients because

such strategies often involve international and segmental transfer of funds and assets. These

arguments would suggest a positive association between tax planning NAS and important tax

outcomes such as reduction in effective tax rates and increased cash tax savings. Thus, we propose

the following hypotheses, stated in the alternative form:

H1a: There is a negative association between tax planning NAS and effective tax rates. H1b: There is a negative association between tax planning NAS and cash tax savings.

It has been argued by the accounting profession that routine tax compliance services

provided by the audit firm do not promote tax avoidance (PCAOB 2004b). Instead, such services

are most likely to represent routine tax related tasks including the preparation of tax returns and

tax payment schedules. Furthermore, Klassen et al. (2015) suggests that audit firms providing

significant tax compliance services could be more reluctant to devise aggressive strategies because

of reputational concerns. Together these imply either no association or a positive association

between auditor-provided tax compliance services and both effective tax rate rates and cash tax

savings. Companies that engage their auditor for tax compliance services could be particularly

complex having filing requirements in multiple taxing jurisdictions. Having the auditor help

navigate complex filing requirements could reveal tax planning opportunities and help clients tax

plan more efficiently even if they are not formally engaging their auditor in tax planning services.

11

Because it is unclear whether tax compliance services are associated with effective tax rate

reductions or cash tax savings we do not make a formal prediction. Nevertheless, our tests include

the effect of tax compliance services on effective tax rate reductions and cash tax savings.

Auditor-Provided Tax Services and Tax risk

Relative to the more thoroughly examined construct of tax avoidance or aggressiveness,

the academic literature has yet to coalesce around a uniform definition of tax risk (Blouin, 2014,

Hutchens and Rego, 2015). Furthermore, Omer, Nueman and Schmidt (2016) suggest that tax risk

as defined by academics does not necessarily equate to tax risk as defined by practitioners. For the

purpose of our study, we define tax risk somewhat broadly as the relative uncertainty of future tax

outcomes. By this definition at least, we believe that, on average, firms have incentives to limit

their tax risk. However, it is unclear, ex-ante, whether the strategies devised by tax NAS are

associated with tax risk either positively or negatively.

While there have been some well publicized cases of accounting firms promoting

aggressive tax strategies, the same audit firms are also best positioned to provide effective tax

strategies that are associated with low levels of tax risk. Audit firms are not only privy to the latest

tax technology at competitor firms, their global reach allows them to more effectively coordinate

tax services across offices in different countries. With experts in both financial reporting and tax

planning, this could result in audit firms reserving their most effective tax strategies (those that

save taxes at low risk) for clients that recruit them for both audit and tax services (Maydew and

Shackelford 2007). This perspective was corroborated by a tax partner we interviewed at a Big 4

accounting firm who said:

“The tax partner gains an understanding of the business perspective behind any tax strategy based on discussion with the audit partner. The transactions accompanying the tax position must have legitimate business purposes other than tax avoidance in order to be sustained

12

upon audit. The audit partner helps the tax partner support the position when grey areas arise.” In the period of our study, auditors providing tax services face greater scrutiny by firms’

by regulators and firms’ boards who have to approve certain tax services. This provides auditor-

tax service providers additional incentives to minimize tax risk relative to outside providers. In

addition, because every tax decision has a financial reporting consequence it is possible that higher

tax risk also increases the chances of a financial reporting restatement. Auditor-tax planners face

adverse litigation and reputational consequences with respect to financial restatements (Lennox,

2016) and are thus implicitly motivated to provide tax planning advice that minimizes tax risk.

This is less likely to be the case for non-auditor tax planners because they are not as sensitive to

the possibility that a risk tax position would lead to a financial restatement (PCAOB, 2004). This

leads us to the following alternative-form hypothesis:

H2: There is a negative association between tax planning NAS and the level of tax risk. Congress and regulators have accused accounting firms of flouting tax laws to help their

clients avoid billions of dollars of taxes through improper tax shelters (Bryan-Low 2003). For

example, the IRS pursued an enforcement action against KPMG in 2003, alleging that KPMG

partners sold two different tax strategies that deprived Treasury of at least $1.28 million in taxes.

The IRS identified both strategies as being potentially abusive strategies.7 While these risky and

abusive plans were provided to non-audit clients, it is possible that similar strategies permeated to

7 The IRS crackdown on tax shelters has extended to include tax strategies devised by Andersen, BDO Seidman, Deloitte & Touche, Ernst & Young, and PWC. More than 50% of BDO Seidman's tax revenues in 2000 came from the "wolf pack" tax team that provided tax NAS. Deloitte also had a specialist tax team known as the "Creative Tax NAS" group that marketed tax strategies to wealthy clients. For further details, see Beale (2004).

13

audit clients as well. To the extent that such abusive tax strategies increase tax risk, then we would

expect to find a positive association between auditor-provided tax planning services and tax risk.8

Similar arguments discussed above with respect to auditor provided tax compliance fees

apply here. Specifically, a desire for auditors to reduce reputational costs could lead to a negative

association between tax compliance NAS and tax risk. If tax compliance NAS are not associated

effective tax rates or cash tax savings as we predict above, then it is not clear why we would find

an association with tax risk. Although we do not make a formal hypothesis, we nevertheless

examine the relation between tax compliance NAS and tax volatility.

3. SAMPLE

Our initial sample is obtained by merging the Audit Analytics, COMPUSTAT and

BoardEx databases over fiscal years 2007 to 2012. After excluding firm-year observations without

the required SOX 201 auditor fee disclosures in Audit Analytics and missing data in

COMPUSTAT and BoardEx, our initial sample consists of 11,324 firm-year observations.

Consistent with prior tax avoidance literature, we eliminate 3,997 observations with negative

pretax income.9 We then eliminate 294 observations for firms not incorporated in the US and “non-

corporate firms” including REITs, limited partnerships, and other firms designated as “trusts.”10

Finally, we eliminate 734 observations from the financial industry resulting in an extended sample

of 6,299 observations. 11 This sample consists of three types of firm-year observations: 1)

8 Implicit in our tests is the notion that a firm can have a tax risk profile that is distinct from its underlying operating risk (Blouin, 2014). If this is not true and the control variables in our regressions do a reasonable job at capturing variation in operating risk, then we would be less likely to document any support for H2. 9 As expected, the sample of loss firm observations is much larger in the financial crisis period (2008 and 2009). Thus, the firms included in these years may differ from those in the non-recessionary years. Additional analyses reveal that eliminating all 2008 and 2009 observations does not influence any of our findings and conclusions. 10 Specifically, we exclude firms with SIC code of 6798 (Real Estate Investment Trusts), firms with names ending in “-LP” or containing the word “Trust”, and firms with a six-digit CUSIP ending in “Y” or “Z”. 11 We assess the representativeness of our sample firms by comparing the mean and median values of the dependent and independent variables between the populations of non-negative pre-tax income firms on the COMPUSTAT

14

observations where audit clients purchase tax services but do not disclose the breakdown of tax

planning and compliance fees (3,556 observations), 2) observations where clients purchase tax

services and disclose the breakdown of tax planning and compliance fees (1,133 observations),

and 3) observations where clients purchase no tax services from the auditor (1,610 observations).

Due to data requirements, we retain smaller extended samples for our cash ETR, and uncertain tax

benefit tests (6,227 and 5,733 respectively). We use these extended samples to compare the

association between tax outcomes and total auditor provide tax fees for the population of firms

with available data. This helps somewhat to increase the generalizability of our findings while also

allowing us to compare to prior research that has examined relations between total tax NAS and

tax outcomes (Mills, Erickson, and Maydew, 1998; Omer, Bedard, and Falsetta, 2006; Cook and

Omer 2013, Hogan and Noga 2015);

Our hypothesis testing requires that we separate tax planning NAS from tax compliance

NAS. To do so we manually collect the planning and compliance fee data from the firm’s proxy

statement or 10K (See Appendix A for examples of tax fee disclosures).12 Our primary sample

includes firms that do not purchase tax NAS of any kind (i.e. Group 3) because the breakdown of

APTS components is observable (i.e. it is zero in both cases) for these firms. We exclude firms

that purchase auditor-provide tax services but do not disclose the breakdown between planning

and compliance (i.e. Group 1) because tax NAS components are not observable for these firms.

database and our sample firms (which have complete data for all variables). The results from this analysis (not reported) indicate that our sample firms are comparable with the population. 12 In 2012, Audit Analytics began collecting these tax fee breakdowns and making the information available in machine-readable form. At present, Audit Analytics has coded the tax planning and compliance variables from 2009 to 2012 for a sub-sample of firms that provide this breakdown. Given that our hand-collected sample spans from 2003 to 2012, we are able to assess whether our bifurcation methodology is similar to that of Audit Analytics for the overlapping period. Of the 2,101 observations we collected with positive tax fees and a breakdown of planning and compliance fee between 2009 and 2012, Audit Analytics provides the breakdown for 493 observations. The correlation between Audit Analytics’ and our breakdown of tax planning and compliance is 99%. The few discrepancies are Audit Analytics’ coding errors for overall tax fees. This provides some assurance that our categorization of planning and compliance fees is rigorous and accurate.

15

This reduces our main analysis sample to 2,743, 2,718, and 2,225 observations for our effective

tax rate reductions, cash tax savings, and uncertain tax benefit tests respectively.

Consistent with Guenther et al. (2017) our proxy for tax volatility requires five years of

sequential data in the COMPUSTAT files, further reducing our extended sample size to 2,104, and

our primary sample size to 899 observations. Table 1 describes the sample selection process for

our analysis.

4. RESEARCH DESIGN

Because the choice to purchase tax NAS is not random (Lassila, Omer, Shelley, and Smith,

2010) our regression analyses could produce biased coefficients on our variables of interest if we

do not attempt to control for this selection issue. To try and remove, or at least minimize the impact

of this potential confound on our regression results, we first model the decision to purchase tax

NAS consistent with the model in Lassila et al. (2010) and McGuire et al. (2012). The results of

this model are summarized in Table 2. We find that the area under the Receiver Operator Curve

(ROC) is 0.7444 suggesting that our model has reasonable discriminatory power (Hosmer and

Lemeshow 2000, p. 162). We then implement a Heckman (1979) correction procedure by

extracting the inverse Mills ratio (IMR) from this selection model and including it as an additional

control variable in our models described below.

Tests of H1a and H2b

We test H1a and H1b by analyzing whether the planning component of tax NAS drives the

positive association between tax NAS and effective tax rate reductions, and cash tax savings,

respectively. For comparability with recent and contemporaneous studies including Cook and

Omer (2013) and Hogan and Noga (2015), we examine the impact of total tax NAS on effective

16

tax rate reductions, and cash tax savings, in our extended and primary samples using the following

OLS regression model (1a):

GETR/CETR ={TAX_FEE_RATIO, ACCT_EXP, AUD_CHG, OTH_NAF_RATIO, AUD_FEE, BIG4, SIZE, ROA, FOR_D, NOL, LEV, BM, DSALES, RD, CAPX, IMR, Industry, Year}

We then consider the separate effects of the tax planning and tax compliance NAS for the

primary sample of firms disclosing the breakdown of tax NAS coupled with firms that do not

purchase any NAS using the following OLS regression model (1b):

GETR/CETR ={TAX_PLAN_RATIO, TAX_COMP_RATIO, ACCT_EXP, AUD_CHG, OTH_NAF_RATIO, AUD_FEE, BIG4, SIZE, ROA, FOR_D, NOL, LEV, BM, DSALES, RD, CAPX, IMR, Industry, Year}

Following prior studies (e.g., Gupta and Newberry 1997; Mills et al. 1998; Rego 2003;

Chen et al. 2010, Chyz and Gaertner, 2017), our first dependent variable in models (1a) and (1b)

capture GAAP effective tax rates. Besides researchers, tax policy advocacy groups such as Citizens

for Tax Justice (CTJ) also use the ETR measure to identify US firms with relatively low tax

burdens. Testimony from Brunswick Corporation before the House Ways and Means Committee

(Zelisko, 2014) provides anecdotal evidence that corporations consider the reduction in ETR when

devising tax strategies13. GAAP effective tax rates are expected to capture tax minimization

strategies stemming from permanent book-tax difference and reflect the use of tax-related accruals

that affect accounting earnings (Hanlon and Heitzman 2010). We follow prior studies and define

GAAP effective tax rates as total tax expense scaled by pre-tax book income less special items.

One limitation of GETR is that it does not capture tax strategies that defer taxes (e.g., use of

accelerated depreciation for tax purposes). Our second proxy of effective tax rates does not suffer

13 This anecdotal evidence is supported by a recent survey conducted by Graham et al (2014). The authors find that 47% of the tax executives in publicly traded firms state that top management values the GAAP ETR more than cash taxes paid, and in another 37 percent of public firms, the two metrics are equally valued by top management.

17

from this problem as it is based on cash taxes paid (i.e., cash tax savings) and is not affected by

changes in tax accounting accruals. This measure is defined as the ratio of cash taxes paid to pretax

income (excluding special items).

Because strategies are likely to be more effective if deployed in periods prior to when tax

payments are due, our measures of tax NAS are lagged (prior year) tax fee ratio

(TAX_FEE_RATIO) in (1a) and the planning (TAX_PLAN_RATIO) and compliance

(TAX_COMP_RATIO) components in (1b).14 We include the percentage of AC members with

accounting financial expertise when the tax services were purchased (ACCT_EXP), because prior

studies suggest that AC accounting experts influence corporate tax planning (Robinson et al.

2012), and the decision to purchase tax NAS (Naiker et al. 2013). We include an indicator variable

capturing whether the company switched auditors (AUD_CHG) in the current or prior year,

because the knowledge used to enhance the tax plan is likely gained over time.15 We control for

other non-audit fees 16 (i.e., non-tax NAS) divided by total fees paid to the auditor

(OTH_FEE_RATIO), and the natural logarithm of audit fees (AUD_FEE) to ensure that the tax

planning NAS rather than other auditor services are driving our results. We control for auditor

size because bigger accounting firms are more likely to provide higher quality services (DeAngelo

1981) and draw on a larger base of tax knowledge (Maydew and Shackelford 2007). Our audit

firm size variable, BIG4, is set to 1 if the auditor is one of the Big-4 auditors (E&Y, D&T, PWC

and KPMG), and 0 otherwise.

14 The results are similar but weaker if we use the contemporaneous version. This supports the argument presented in Hogan and Noga (2015) that tax NAS is associated with the long term tax strategy of the firm. 15 The results are unchanged when we use a continuous version of auditor tenure. 16 These fees capture audit-related fees and all other non-audit fees. Audit-related fees consist of fees billed for assurance and related services such as due diligence in connection with mergers and acquisitions, internal control reviews and assistance, and financial statement audits of employee benefit plans. It is less clear what types of services are classified as “other non-audit fees”. However, such fees may capture billings for risk management advisory services, assistance with continuing education and training, and provision of information relating to market conditions.

18

We control for firm size captured by the log of total assets (SIZE) and firms reporting

foreign income (FOR_D), because Mills et al. (1998) suggest that larger multinational firms have

more tax-planning opportunities. Consistent with Mills et al. (1998), we also control for leverage

(LEV) and capital expenditure (CAPX) due to interest and depreciation deductions. Further, we

control for the change in sales (DSALES) because growing firms are more likely to make larger

investments in depreciable assets. We control for research and development intensity (RD)

because of the associated investment-tax shield (Gupta and Newberry 1997). Prior studies also

show that profitability influences tax avoidance (Gupta and Newberry 1997; Rego 2003; Dyreng

et al. 2008). Thus, we control for return on assets (ROA). Because the presence of net operating

loss carryforwards and growth opportunities influence incentives to avoid taxes (Rego 2003; Frank

et al. 2009), we include an indicator variable capturing the presence of net operating losses (NOL),

and the book to market ratio (BM). IMR is the inverse Mills ratio derived from the decision to

purchase tax NAS model described above and summarized in Table 2. Finally, we include industry

(two digit SIC) and year fixed effects in order to capture differences in tax avoidance across

industries and time.

Tests of H2

We test H2 by examining the association between tax planning NAS and tax risk. We

measure tax risk using both tax volatility and uncertain tax benefits (UTBs). Consistent with

Guenther et al. (2017) we measure tax volatility as the standard deviation of annual cash ETRs

(VOL_OF_CETR) over the current and future four years, where a greater standard deviation (i.e.,

larger values of VOL_OF_CETR) implies greater tax risk. The use of this tax risk proxy limits the

sample period to the years 2007 to 2009 because our hand-collected data on APTS components

19

ends in 2013. We measure UTBs as the ending balances of uncertain tax benefits scaled by total

assets (Guenther et al., 2017; Hutchens and Rego, 2015; Rego and Wilson, 2012).17

As we do in our effective tax rate tests, we first examine the relation between total tax NAS

and tax risk in our extended and primary samples using the following OLS regression model (2a):

VOL_OF_CETR/UTB ={TAX_FEE_RATIO, ACCT_EXP, AUD_CHG, OTH_FEE_RATIO, AUD_FEE, BIG4, SIZE, ROA, FOR_D, NOL, LEV, DSALES, RD, CAPX, VOL_OF_PTBI, ESO_BENEFITS, IMR, Industry, Year}

For tests of H2 that examine the separate associations of tax planning and tax compliance NAS on

tax risk, we use the following model (2b) with our primary sample18:

VOL_OF_CETR/UTB ={TAX_PLAN_RATIO, TAX_COMP_RATIO, (GETR or CETR), ACCT_EXP, AUD_CHG, OTH_FEE_RATIO, AUD_FEE, BIG4, SIZE, ROA, FOR_D, NOL, LEV, DSALES, RD, CAPX, VOL_OF_PTBI , ESO_BENEFITS, IMR, Industry, Year}

The control variables also remain similar to those employed in model (1) as the factors

influencing firms’ tax payments influence both effective tax rates cash tax savings and tax risk.

Consistent with Guenther et al. (2017), we also control for the standard deviation of pre-tax book

income over the current and future four years (VOL_OF_PTBI) and the excess tax benefit of stock

options scaled by lagged total assets (ESO_BENEFITS) when our dependent variable is

VOL_OF_CETR. We omit these additional controls when our dependent variable is UTB. Finally,

we control for effective tax rate reductions (GETR) or cash tax savings (CETR) given that higher

tax avoidance could lead to higher tax risk.19

17 Although various measures of UTBs have been used in prior literature to capture tax aggressiveness (e.g. Klassen et al., 2015; Lisowsky, Robinson, and Schmidt, 2013) UTBs are also thought to be related to tax risk. Rego and Wilson (2012) use UTB ending balances to capture “risky” tax avoidance and Blouin (2014) notes that UTBs should reflect some “scope of uncertainty inherent in a tax position”. 18 Unlike model (2), we do not capture the lagged versions of the main independent variables or interest, because in the case of tax volatility the variable construction induces the lead-lag analysis. VOL_OF_PTBI capture the sustainability of current positions for five future years. Unlike tax strategies which could take time to achieve results, we expect tax NAS to have a more immediate impact on the recognition of tax uncertainty. However, the results are robust to the lagged auditor fee variables for either dependent variable. 19 Our inferences do not change if we omit effective tax rate controls.

20

While we expect that our two tax risk measures individually capture at least some element

of tax risk, it is not clear, ex-ante, how closely they will be related or how much noise each measure

will possess. 20 Generally speaking, the volatility of future cash tax payments is an outcome

variable that captures, to some extent, the realization of tax risk. UTBs are an accounting accrual

meant to capture, to some extent, the expected (i.e. ex-ante) realizability of future tax benefits

(Ciconte et al., 2016). The findings in De Simone, Robinson, and Stomberg (2014) and the

discussion in Blouin (2014) underscore the fact that UTBs reflect substantial managerial

judgement and financial reporting incentives that could reduce their ability to reliably capture

cross-sectional variation in tax risk in all settings. If our tax NAS variables of interest are

negatively (positively) related to tax volatility but positively (negatively) related to UTBs, this

could suggest that employing an auditor to provide tax services is associated with conservative

(aggressive) financial reporting.

5. EMPIRICAL RESULTS

Descriptive Statistics

Table 3, Panel A through D presents descriptive statistics for the samples employed in our

tests of H1a, H1b, and H2 respectively. We present the summary statistics for our full sample and

for two subsamples representing observations with zero and positive values of tax planning NAS

(TAX_PLAN_RATIO).

Table 3, Panel A through D presents descriptive statistics for the samples employed in our

tests of H1a, H1b, and H2 respectively. We present the summary statistics for our full sample and

20 Ciconte, Donohoe, Lisowsky, and Mayberry (2016) find that UTBs are generally predictive of future cash outflows. Given this finding, Guenther et al. (2017) note that UTBs may not actually reflect the degree of uncertainty, or risk in the firm’s tax payments. Somewhat contrary to these inferences, Robinson, Stomberg, and Towery (2016) find no evidence that UTBs are predictive future tax cash flows required to settle uncertain tax positions.

21

for two subsamples representing observations with zero and positive values of tax planning NAS

(TAX_PLAN_RATIO).

The statistics reported in Panel A reveal a mean (median) GAAP effective tax rate (GETR)

of around 31.0% (32.7%), which is consistent the values reported in prior studies utilizing this

measure (e.g., Chen et al. 2010; Cook and Omer 2013). Summary statistics for the two subsamples

in Panel A indicate that firms that purchase tax planning NAS (TAX_PLAN_RATIO > 0) have

significantly lower effective tax rates (mean =28.3%) in comparison to firms that do not purchase

any tax planning NAS (mean = 31.9% in the TAX_PLAN_RATIO = 0 subsample). This difference

is statistically significant (p-value < 0.01) and is in line with our expectations.

The summary statistics reported in Panel B, show that the mean (median) cash effective

tax rates equates to around 24.0% (23.1%) percent in our sample. These statistics is similar to those

reported in prior studies (Dyreng et al. 2008). While we find that the mean value of CETR is

marginally smaller in firms that purchase tax planning NAS (mean = 23.9%) in comparison to

firms that do not purchase tax planning NAS (mean = 24.1%), this difference is not statistically

significant. It is important to note that this finding is obtained without controlling for other

determinants of tax avoidance.

The summary statistics in Panel C indicate that tax planning NAS purchasing firms have

significantly (p < 0.01) lower values of VOL_OF_CETR (mean = 0.148) in comparison to firms

without tax planning NAS (mean = 0.247). Finally, the summary statistics reported in Panel D

indicate that tax planning NAS purchasing firms have significantly (p < 0.01) higher values of

UTB (mean = 0.013) in comparison to firms without tax planning NAS (mean = 0.010). The

univariate results in Table 2 – Panels C and D are consistent with the predictions in H2.

22

In addition to the statistics reported in Panels A through D of Table 3, we examined how

the values of our effective tax rates, cash tax savings and tax risk measures varies across firms

with high and low levels of tax planning NAS (based on the median value of TAX_PLAN_RATIO)

within firms that purchase tax planning NAS. Our findings from these tests (untabulated) indicate

that firms with high levels of tax planning NAS have significantly lower effective tax rates (p <

0.10), higher cash tax savings (p < 0.10), lower volatility of Cash ETR (p < 0.05), and higher

uncertain tax benefits (p < 0.01) in comparison to firms with low levels of tax planning NAS.

While these summary statistics support the view that tax planning NAS is associated with

higher tax avoidance and lower tax risk, these findings should be interpreted with caution as they

are obtained without controlling for other factors influencing tax avoidance and tax risk. Indeed,

the summary statistics reported across all Panels in Table 3 reveal statistically significant

differences between many control variables for firms with high and low levels of tax planning

NAS. We control for these factors in the multivariate analyses.21

Effective Tax Rate Reductions and Cash Tax Savings Multivariate Results

Table 4 reports the regression results for tests of H1a and H1b. Columns (1) through (3)

reports results examining associations between APTS and effective tax rates. Columns (4) through

(6) reports results examining associations between APTS and cash tax savings. The analysis

reported in columns (1) and (4) considers the effect of total tax fees (TAX_FEE_RATIO) in an

extended sample, which includes firms in the main sample (firms that do not purchase tax NAS

21 Correlation statistics (untabulated) indicate that the three tax fee variables (TAX_FEE_RATIO, TAX_PLAN_RATIO, and TAX_COMP_RATIO) are negatively and significantly (p-value < 0.01) correlated with our measure of effective tax rates (GETR). We also find that TAX_FEE_RATIO and TAX_PLAN_RATIO are negatively (positively) and significantly (p-value < 0.01) correlated with VOL_OF_CETR (UTB). While the correlation coefficients between the independent variables reveal a number of significant correlations, none of these correlation coefficients are sufficiently large to affect the study’s conclusions. The highest correlation was between SIZE and AUDFEE (0.83). Our findings for the tax NAS variables remain robust when either SIZE or AUDFEE is excluded from the analyses. All variance inflation factors in our models (not tabulated) are below the threshold of 10 beyond which multicollinearity may be a problem (Kennedy 1992).

23

from their auditor and those that breakdown the planning and compliance components of their total

tax fees) as well as firms that purchase but do not disclose the breakdown of their tax NAS fees.

The analysis in Columns (2) and (5) evaluates the impact of total tax fees (TAX_FEE_RATIO) in

the main sample of firms that either do not purchase APTS or purchase APTS and provide the

breakdown. Columns (3) and (6) considers the effects of tax planning NAS (TAX_PLAN_RATIO)

and tax compliance NAS (TAX_COMP_RATIO) in the main sample.

Focusing first on the effective tax rate analysis (GETR) and tests of H1a we find a negative

and significant (p-value < 0.10 or better) coefficient on TAX_FEE_RATIO in Columns (1) and (2).

These indicates lower effective tax rates for firms that procure greater tax NAS from their auditor

and support the findings of Cook and Omer (2013) and Hogan and Noga (2015) who find that

changes in total tax fees (as a proxy for tax planning fees) is positively associated with changes in

tax avoidance. The positive and significant (p-value < 0.01) coefficient on TAX_PLAN_RATIO

and the insignificant coefficient on TAX_COMP_RATIO in Column (3) verify that the tax planning

portion of tax NAS is indeed responsible for the results presented in Columns (1) and (2), thus

supporting H1a.

Turning to our findings for the control variables, the results reported in Columns (1)

through (3) reveal negative and significant (p < 0.10) associations between effective tax rate

reductions and purchase of other NAS (OTH_NAF_RATIO) in our primary sample, audit fees

(AUD_FEE), firm size (SIZE), foreign operations (FOR_D), net operating losses (NOL) and

research and development expenditure (R&D). We also find positive and significant associations

between effective tax rate reductions and the presence of Big 4 auditors (BIG4) but only in the

extended sample, and firm profitability (ROA) across all three specifications. Further, the

coefficients on the statistically significant control variables are all of the predicted sign with the

24

exception of BIG4. We find that the inverse Mills ratio derived from the decision to purchase

APTS model is not statistically significant in any of the first three columns.

Turning to the cash tax savings regressions (CETR) and tests of H1b we find no evidence

of associations between total tax NAS (TAX_FEE_RATIO) and cash tax savings in either Columns

(4) or (5). In Column (6) we see that separating tax planning fees from compliance reveals an

association between APTS and cash tax savings that is driven by tax planning fees. This is

evidenced by the negative and significant (p < 0.10) on (TAX_PLAN_RATIO). As in the effective

tax rate regressions, variation in tax compliance NAS are not related to cash tax savings as

evidenced by the insignificant coefficient on TAX_COMPLY_RATIO. In fact, this lack of

association with cash tax savings appears to offset the tax planning effect when total tax NAS are

considered. This highlights a methodological contribution of our study.

We find that many of our control variables are consistently negative and statistically

significant (presence of accounting experts on the audit committee [ACCTG_EXP], firm size

[SIZE], leverage [LEV], research and development expenditure [R&D], and capital expenditure

[CAPX]). A few of our control variables are positive and significantly related to cash tax savings

at the 10 percent (or better) level (audit fees [AUD_FEE], change in auditor [AUD_CHG] firm

profitability [ROA])– but not necessarily across all columns. The inverse Mills ratio (IMR) is

consistently negative, but only significant in our extended sample.

Tax Risk Multivariate Results

Table 5 Panel A reports the results for tests of H2 that consider the influence of tax planning

NAS on tax volatility (VOL_OF_CETR). We report the results based on controlling for either

effective tax rate reductions (GETR) in columns (1) through (3) or cash tax savings (CETR) in

columns (4) through (6). In both cases we commence by examining the impact of total tax NAS

25

on tax risk using an extended sample (Columns (1) and (4)) and our main sample (Columns (2)

and (4)), before examining the separate effects of tax planning NAS and tax compliance NAS on

tax risk (Columns (3) and (6)).

Similar to our findings for cash tax savings, the results reported in Table 5 indicate that

while total tax NAS (TAX_FEE_RATIO) is not significantly associated with tax risk in our

extended and main samples. However, the results reported in Columns (3) and (6) indicate that tax

planning NAS (TAX_PLAN_RATIO) is negatively and significantly (p < 0.05) related to tax risk

after controlling for either GETR or CETR.

The results reported in Table 5 Panel A reveal significant (p < 0.10) associations between

tax risk (when using tax volatility as our proxy) and several control variables. Specifically, we find

that tax risk is consistently positively associated with audit fees (AUD_FEE), volatility of pre-tax

book income (VOL_OF_PTBI), and consistently negatively related to firm size (SIZE), firm

profitability (ROA), and research and development expenditure (R&D). Consistent with Guenther

et al. (2017), we also find that tax risk is positively associated with both GAAP (GETR) and Cash

(CETR). In other words, firms paying lower tax rates tend to have lower volatilities in tax rates.

Taken together with our results for tax planning NAS in Tables 3 and 4, this suggests that one way

through which firms can realize a negative correlation between tax risk and tax savings is through

the tax planning services of their auditor.

Our uncertain tax benefits results summarized in Table 5 Panel B tell a similar story. While

total tax NAS (TAX_FEE_RATIO) is not significantly associated with tax risk in our extended and

main samples, the results reported in Columns (3) and (6) indicate that tax planning NAS

(TAX_PLAN_RATIO) is negatively and significantly (p < 0.05) related to tax risk (as measured

using UTBs) after controlling for either GETR or CETR.

26

The results reported in Table 5 Panel B reveal significant (p < 0.10) associations between

tax risk (when using UTBs as a proxy) and several control variables. Specifically, we find that tax

risk is consistently positively associated with audit fees (AUD_FEE), change in auditor

(AUD_CHG), firm profitability (ROA), research and development expenditure (R&D), having

foreign source income (FOR_D). Although many of our control variable coefficient estimates are

negative, the only coefficient that is consistently statistically significant is that on cash effective

tax rates (CETR). Many of our control variables yield coefficient estimates that are different in

sign from those we achieve when tax volatility is our dependent variable. This suggests that while

UTBs and tax volatility potentially capture tax risk, their sources of variation (with the exception

of tax NAS) appear to be substantially different.

Taken together with our results for tax planning NAS in Table 4, the results in Table 5

suggests that one way through which firms can realize a negative correlation between tax risk and

tax savings is through the tax planning services of their auditor.

6. ADDITIONAL ANALYSES AND ROBUSTNESS TESTS

Tax Planning With and Without Tax Compliance NAS

With our main analyses we document that effective tax rates and cash tax savings are

significantly influenced by tax planning, holding constant the variation in tax compliance. While

this approach helps to isolate the effect of the variation in tax planning from the variation in tax

compliance it is possible that the effect of tax planning NAS can vary depending on whether firms

procure tax planning NAS in isolation or in conjunction with tax compliance NAS. One view is

that external auditors are hesitant to assist their clients generate tax savings unless they are also

involved in the company’s compliance work. This could be due to the external auditor being

27

concerned of their promoted tax strategies being subject to greater scrutiny from a third-party

provider of tax compliance services. Auditors may also be cognizant of third party tax services

providers gaining access to more information on proprietary tax strategies auditors provide to their

clients. This suggests that external audit firms are likely to help their clients realize greater

effective tax rate reductions and higher cash tax savings when they provide tax planning NAS in

conjunction with tax compliance NAS.22

It is also possible that the impact of tax planning NAS on tax risk is influenced by the

presence or absence of tax compliance NAS. To the extent that external auditors are more willing

to assist their clients realize greater tax savings when they provide tax planning NAS in

conjunction with tax compliance NAS, it is possible that the promoted tax strategies may be more

aggressive in nature too, which would increase their clients’ exposure to tax risk. However, this

finding would be inconsistent with the negative relation between tax planning NAS and tax risk

documented in our main analysis unless that result is driven by tax planning NAS purchased in

isolation. This is possible as planning NAS can involve audit firms offering tax risk management

plans to reduce the tax risk of their clients (Neuman et. al., 2016). Since these tax planning NAS

are less likely to directly focus on tax savings, there may be a lower need for firms to purchase and

auditors to provide tax compliance services in tandem with these specific types of tax planning

services to minimize tax risk.

To formally evaluate the above possibilities, we repeat our regression analyses in Tables 4

and 5 after replacing our test variable capturing tax planning NAS (TAX_PLAN_RATIO) with two

variables representing the provision of tax planning NAS in isolation

22 Untabulated analysis shows that the proportion of companies with non-zero APTS that purchase tax planning over our sample period varies between 42% and 47%. The proportion that purchase tax compliance varies between 76% and 81%. The proportion that purchase only tax planning, only tax compliance, or boht varies between 10% and 21%, 50% and 60%, and 30% and 22% respectively.

28

(TAX_PLAN_RATIO_WOCOMP) and in conjunction with tax compliance NAS

(TAX_PLAN_RATIO_WCOMP).23 The results reported in Table 6, Panel A indicate a negative

and significant (p < 0.10 or better) association between tax planning NAS in conjunction with tax

compliance NAS (TAX_PLAN_RATIO_WCOMP) and effective tax rate reductions and cash tax

savings, respectively. Conversely, we do not find that tax planning NAS affects these tax outcomes

when purchased in isolation (TAX_PLAN_RATIO_WOCOMP). The comparative results from tax

risk analyses, reported in Panel B and Panel C of Table 6 show that tax risk (when measured with

tax volatility or UTBs) is negatively and significantly associated with tax planning purchased in

isolation (p < 0.05) but is not associated tax planning purchased in conjunction with tax

compliance.

Collectively, these results suggest that while tax planning NAS purchased in conjunction

with tax compliance NAS is likely to promote tax minimization strategies in clients, these tax

savings are realized with marginally lower tax risk in comparison to firms that do not procure tax

services from their auditor. On the other hand, the results for tax planning NAS purchased in

isolation indicate that while these services do not produce tax savings for clients they significantly

lower clients’ exposure to tax risk.

For the sake of consistency, we also undertake comparative analyses to assess the separate

effects of tax compliance NAS in isolation (TAX_COMP_RATIO_WOPLAN) and in conjunction

with tax planning NAS (TAX_COMP_RATIO_WPLAN). The results from these analyses, reported

23 These variables allow the association between tax planning and our outcomes of interest to vary differently when the client also procures tax compliance and when the company only purchases tax planning. Thus, TAX_PLAN_RATIO_WCOMP is essentially TAX_PLAN_RATIO multiplied by an indicator variable that equals 1 when the company also purchases tax compliance. Similarly, TAX_PLAN_RATIO_WOCOMP is essentially TAX_PLAN_RATIO multiplied by an indicator variable that equals 1 when the company only purchases tax planning. Because we omit the main effects of the indicator variables denoting compliance, or no compliance fees purchased, the main effects of purchasing compliance services, or purchasing tax planning services only, are captured by TAX_PLAN_RATIO_WCOMP and TAX_PLAN_RATIO_WOCOMP respectively. This specification makes it easier to interpret our results.

29

in Table 6 Panels A, B, and C indicate that tax compliance NAS remain insignificantly associated

with effective tax rates, cash tax savings, and tax risk, regardless of whether they are purchased

with tax planning NAS or not. As is evident from the results in Table 6, the positive effect of tax

planning NAS in conjunction with tax compliance NAS (TAX_PLAN_RATIO_WCOMP) on

effective tax rates and cash tax savings and the negative effect of tax planning NAS in isolation

(TAX_PLAN_RATIO_WOCOMP) on tax risk are robust to controlling for

TAX_COMP_RATIO_WOPLAN and TAX_COMP_RATIO_WPLAN separately.

Industry-Size Adjusted of Effective Tax Rates

Armstrong , Blouin, Jagolinzer and Larker (2015) and Chyz and Gaertner (2017) suggest

that boards and managers have incentives to benchmark effective tax rates against peer firms. As

a result, it is possible that benchmarked rates will better capture the alignment of incentives

between auditors and their clients. In an attempt to capture this incentive alignment, in untabulated

results we follow Armstrong, Blouin, Jagolinzer, and Larker (2015) and Chyz and Gaertner (2017)

and benchmark each annual effective tax rate measure by Fama-French 48 industry-size peers.

These alternative effective tax rate measures capture cross-sectional variation in firms’ tax

planning and benchmarks a given firm’s effective tax rate reductions and cash tax relative to that

of similar-sized firms in the same industry and year. We find that our results continue to hold, and

in most cases are stronger, using benchmarked effective tax rates.

Alternate Definitions of UTBs

Our primary measure of uncertain tax benefits scales total end of year uncertain benefits

by total assets. Some recent and contemporaneous research, including Klassen et al. (2015) and

Dyreng et al. (2017) focuses on the current-year additions to the UTB balance. Other

contemporaneous research focusses on the portion of UTBs that affect the ETR (Neuman, et al.,

30

2016). The inferences gleaned from our hypothesis testing are very similar using the alternative

specifications of UTBs.

Controlling for the Decision to Disclose the Tax Planning and Compliance Components of APTS

As noted above, the composition of our sample raises the potential selection issue that the

decision to purchase APTS is related to the tax outcomes we examine. We control for this

possibility using a Heckman (1979) correction procedure. There is another potential selection issue

that is not addressed in our primary analysis. Specifically, that the decision to disclose the

breakdown of tax NAS components is related to the tax outcomes we examine. To begin to address

this unique selection issue, we model the decision to disclose the components of tax NAS

(conditional on purchasing tax NAS) using the determinants described in Gal-Or, Harding, and

Naiker (2017). To our knowledge, Gal-Or et al. (2017) is the only paper to model the decision to

disclose the components of tax NAS. In untabulated analysis, we find that the area under the ROC

(0.6279) suggests a very modest ability to predict the disclosure of APTS components. In the

context of our study, we believe this provides some evidence that Group 1 and Group 2 firms (the

only groups of firms for whom the decision to disclose tax NAS components is relevant) are similar

thus increasing the generalizability of our hypothesis testing that necessitates the exclusion of

Group 1 firms.

We find that ten of the decision model’s covariates are statistically significant at

conventional or marginal levels (i.e. p-values < 0.20). Because we find evidence that these

covariates are related to the decision to purchase tax NAS, as an alternative to a Heckman (1979)

correction approach, we ensure that these covariates are included in our primary regression models

to control for the possibility that they are also related to our tax outcomes of interest.24 After doing

24 Because the test of our hypotheses require that we exclude Group 1 firms, we cannot implement a traditional Heckman (1979) procedure because Group 1 firms would be included in first stage but excluded from the second

31

so, we continue to document consistent results across all of our tests.

7. CONCLUSION

This study provides insight into the role of two different types of tax NAS that auditors

provide to their clients to help them realize tax savings and lower tax risk. Specifically, we use

hand-collected tax NAS fee data decomposed into its tax planning and compliance components to

investigate the separate effects of these two types of services on effective tax rates, tax savings and

tax risk. We find that tax planning NAS is positively related to effective tax rate minimization and

cash tax savings. These findings are driven by instances where tax planning NAS is purchased in

conjunction with tax compliance NAS. We also find that tax planning NAS has a negative impact

on tax risk but this finding is most pronounced in firms that procure tax planning NAS in isolation.

In contrast, we document insignificant relationships between tax compliance fees and effective tax

rates, tax savings and tax risk. These findings are robust to research design choices meant to control

for the decision to purchase tax NAS and the decision to disclose the components of tax NAS.

They are also robust to alternative proxy constructions for tax risk and effective tax rates.

Overall, our results are consistent with the view that audit firms use their tax planning NAS

to promote effective and less-risky tax avoidance plans, and that further regulatory bans or

limitations on auditor-provided tax services may diminish such benefits. Our results also imply

that the disclosure of information on the nature of tax services acquired from the auditor can

provide useful information to stakeholders on the potential impact of various tax NAS on the tax

stage. Another issue is that Group 3 firms would be excluded from the first stage and included in the second stage, and as a result, we would not be able to compute an inverse Mills ratio for these firms.

32

avoidance strategies of firms. This finding should be of interest to regulators who are considering

revisions to the disclosure requirements relating to auditor-provided services.25

One caveat to our findings is that, like most archival studies, we document association, not

causation. Second, our ability to separately examine the effects of fees paid for auditor-provided

tax planning and tax compliance services is only as effective as the accuracy of the disclosures

firms provide to distinguish between these services within their portfolio of total auditor-provided

tax services.26 Further, due to unavailability of data, we are unable to directly test the effects of

tax planning fees paid to a third party who is not the firm auditor, nor are we able to consider

differences in tax reporting implications between audit clients and non-audit clients purchasing tax

planning services from the same audit firm provider. Next, we note that the firms in our sample

voluntarily disclose their tax fee components. As such, we are unable to ascertain whether our

results can be extrapolated to clients purchasing tax NAS but not disclosing the tax fee breakdown.

Additionally, we do not test how characteristics and incentives differentiate firms that procure tax

planning NAS only from those that purchase both tax planning and tax compliance NAS. These

present potential opportunities for future research.

25 In 2015, the SEC published a concept release exploring and seeking advice on possible revisions to audit committee disclosures (SEC 2015). Among other key areas of concern, the SEC solicited advice about the following question: “Should the audit committee provide additional disclosure about the nature and extent of non-audit services...?” 26 According to a former audit partner and current AC member, the disclosures in the proxy statement are not audited, but are reviewed for reasonableness. In most cases, the auditor provides the breakdown of tax services during the first year of the audit. The client keeps tracks of fees for each type of service in subsequent audits. The auditor checks these numbers for reasonableness on an ongoing basis.

33

REFERENCES

American Institute of Certified Public Accountants (AICPA). 2007, Have you Impaired your Independence by Performing Nonattest Services for your Attest Clients?

_____ (AICPA). 2011, Code of Professional Conduct Rule No. 101-3, Performance of Nonattest Services, New York, NY

Armstrong, C. S., Blouin, J. L., Jagolinzer, A. D., and D. F Larcker. 2015. Corporate governance, incentives, and tax avoidance. Journal of Accounting and Economics 60 (2015):1-17.

Badertscher, B., S. Katz, and S. Rego, 2013. The separation of ownership and control and corporate tax avoidance. Journal of Accounting and Economics. 56(203): 228-250.

Beale, L. M. 2004, Putting SEC Heat on Audit Firms and Corporate Tax Shelters: Responding to Tax Risk with Sunshine, Shame and Strict Liability, Journal of Corporation Law 29 (2): 219-266.

Blouin, J. 2014. Defining and Measuring Tax Planning Agressiveness. The National Tax Journal. 67 (4): 875-900.

Bryan-Low, C. 2003, Unhappy returns: Accounting Firms Face Backlash over the Tax Shelters they Sold, Wall Street Journal (February): A1.

Chen, S., X. Chen, Q. Chiang, and T. Shevlin. 2010, Are Family Firms More Tax Aggressive than Non-Family Firms?, Journal of Financial Economics 95 (1): 41-61.

Chyz, J, R. Gal-Or, and V. Naiker. 2016. Separating Auditor Provided Tax Compliance and Tax Planning Services: Audit Quality Implications. Working Paper.

Chyz, J., and F. Gaertner. 2017. Can Paying "Too Much" or "Too Little" Tax Contribute to Forced CEO Turnover?. The Accounting Review In-Press.

Ciconte, W., M. Donohoe, P. Lisowsky, and M. Mayberry 2016. Predictable uncertainty: The relation between unrecognized tax benefits and future income tax cash outflows. Working Paper.

Cook, K. A., R. G. Huston, and T. C. 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.

____, K. A., and T. C. Omer. 2013, The Cost of Independence: Evidence From Firms’ Decisions to Dismiss Auditors as Tax-Service Providers, Texas Tech and University of Nebraska at Lincoln Working Paper.

Cripe, B., and B. McAllister. 2009. Determinants of audit/tax separation decisions. American Journal of Business 24 (1): 47-56.

DeAngelo, L. E. 1981, Auditor Size and Audit Quality. Journal of Accounting and Economics, 3 (3): 183-199.

Deloitte. 2011, Audit Committee Brief: Tax complexities drive audit committee oversight.

Donohoe, M., G. McGill, and E. Outslay. 2014, Risky Business: The Prosopography of Corporate Tax Planning, National Tax Journal, 67 (4): 851-874

34

Donohoe, M., and R. Knechel. 2014, Does Corporate Tax Aggressiveness Influence Audit Pricing?, Contemporary Accounting Research, 31 (1): 284–308.

Doyle, J., W. Ge, and S. McVay. 2007, Accruals Quality and Internal Control over Financial Reporting. The Accounting Review, 82 (5): 1141–1170.

Dyreng, S, M. Hanlon, and E. Maydew. 2008, Long‐Run Corporate Tax Avoidance, The Accounting Review, 83 (1): 61-82

Dyreng, S., M. Hanlon, and E. Maydew 2017. When does tax avoidance result in tax uncertainty? Working Paper.

Ernst & Young LLP, 2006. Global Tax Risk Survey. Ernst & Young LLP, Washington, DC.

Frank, M., L. Lynch, and S. Rego. 2009, Tax Reporting Aggressiveness and its Relation to Aggressive Financial Reporting. The Accounting Review 84 (2): 467-496.

Gal-Or, R., M. Harding, V. Naiker, 2017, The Determinants and Value Relevance of Voluntarily Disclosing the Tax Planning and Compliance Components of Auditor Provided Tax Services, Working paper

Graham, J. M. Hanlon, T. Shevlin, N. Shroff. 2014, Incentives for Tax Planning and Avoidance: Evidence from the Field, The Accounting Review, 89 (3): 991-1023

Guenther D, Matsunaga S, Williams B. 2017, Is Tax Avoidance Related to Firm Risk. The Accounting Review, 92(1): 115-136.

Gupta, S., and K. Newberry. 1997, Determinants of the Variability in Corporate Effective Tax Rates: Evidence from Longitudinal Data, Journal of Accounting and Public Policy, 16 (1): 1-34

Hanlon, M., and Heitzman, S. 2010. A review of tax research. Journal of Accounting and Economics 50 (2/3): 127-178.

Hanlon, M., and J. Slemrod. 2009, What Does Tax Aggressiveness Signal? Evidence from Stock Price Reactions to News about Tax Shelter Involvement, Journal of Public Economics 93(1-2): 126-141.

Hasan, I., C.K. Hoi, Q. Wu, and H. Zhang, 2014. Beauty is in the eye of the beholder: The effect of corporate tax avoidance on the cost of bank loans. Journal of Financial Economics. 113(1): 109-130.

Heckman, J. 1979. Sample Selection Bias as a Specification Error. Econometrica. 47(1): 153-161.

Hennes, K., A. Leone, and B. Miller 2008. The Importance of Distinguishing Errors from Irregularities in Misstatement Research: The Case of Misstatements and CEO/CFO Turnover. The Accounting Review: November 2008, Vol. 83, No. 6, pp. 1487-1519

Hogan, B. and T. Noga. 2015, Auditor-provided tax services and long-term tax avoidance, Review of Accounting and Finance, 14 No. 3, 2015pp. 285-305

Hosmer, D. and S. Lemeshow 2000. Applied Logistic Regression, Second Edition. John Wiley and Sons.

Hutchens, M., and S. Rego. 2015, Does Greater Tax Risk Lead to Increased Firm Risk?, Indiana University Working Paper

35

Kennedy, P. 1992, A Guide to Econometrics. Cambridge: MIT Press.

Kenneth J. Klassen, Petro Lisowsky, and Devan Mescall (2016) The Role of Auditors, Non-Auditors, and Internal Tax Departments in Corporate Tax Aggressiveness. The Accounting Review: January 2016, Vol. 91, No. 1, pp. 179-205.

KPMG 2003, U.S. Tax Shelter Industry: The Role of Accountants, Lawyers and Financial Professionals, Four KPMG Case Studies: FLIP, OPIS, BLIPS, and SC2. Report prepared by the Minority Staff of the Senate Permanent Subcommittee on Governmental Affairs.

KPMG 2007, FIN 48: The first quarter experience. KPMG Tax Governance Institute. May 9.

KPMG 2013, Audit committee priorities for 2014. KPMG Audit Committee Institute.

Lassila, D., T. Omer, M. Shelley, and L. Smith. 2010. Do complexity, governance, and auditor independence influence whether firms retain their auditors for tax services? Journal of the American Taxation Association 32(1): 1-23.

Lisowsky, P., L. Robinson, and A. Schmidt. 2013. Do publicly disclosed tax reserves tell us about privately disclosed tax shelter activity? Journal of Accounting Research 51 (3): 583-629.

Maydew, E., and D. Shackelford. 2007, The Changing Role of Auditors in Corporate Tax Planning. Taxing Income in the 21st Century. Joel Slemrod and Alan Auerbach, editors, University of Michigan Office of Tax Policy Research / University of California at Berkeley Burch Center, Cambridge University Press: 307-337.

McGuire, S. T., T. C. Omer, and D. Wang. 2012, Tax Avoidance: Do Industry Experts Make a Difference?, The Accounting Review, 87 (3): 975-1003.

Mills, L., M. Erickson, and E. Maydew. 1998, Investments in Tax Planning, Journal of American Taxation Association, 20 (1): 1-20.

Naiker, V., and D. Sharma. 2009, Former Audit Partners on the Audit Committee and Internal Control Deficiencies. The Accounting Review, 84 (2): 559-587.

____________., ____________., and V. Sharma. 2013, Do Former Audit Firm Partners Procure Greater Non-Audit Services from the Auditor?, The Accounting Review, 88 (1): 297-326.

Neuman, S., T. Omer, M. Shelley 2013, Corporate Transparency, Uncertain Tax Activities, and Sustainable Tax Strategies, University of Nebraska at Lincoln Working Paper

Neuman, S., T. Omer, and A. Schmidt. 2016. Assessing Tax Risk: Practitioner Perspectives. Working Paper

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.

Public Company Accounting Oversight Board (PCAOB). 2004a, Auditor Independence and Tax Services Roundtable. PCAOB Briefing Paper, July 14, 2004.

____________. 2004b, Auditor Independence and Tax Services Roundtable Transcript. July 14, 2004.

36

____________. 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.

Rego, S., 2003, Tax Avoidance Activities of U.S. Multinational Corporations, Contemporary Accounting Research, 20 (4): 805–833

Rego, S., and R. Wilson, 2012, Equity Risk Incentives and Corporate Tax Aggressiveness, Journal of Accounting Research, 50 (3): 775-810

Robinson, J. R., Y. Xue, and M. H. Zhang. 2012, Tax Planning and Financial Expertise in the Audit Committee, The University of Texas at Austin, George Washington University and Fordham University Working Paper

Securities and Exchange Commission (SEC). 2005, Regulation S–K, Item 305(c), http://www.sec.gov/rules/final/33-8591.pdf

_______ 2006, Public Company Accounting Oversight Board; Order Approving Proposed Ethics and Independence Rules Concerning Independence, Tax Services, and Contingent Fees and Notice of Filing and Order Granting Accelerated Approval of the Amendment Delaying Implementation of Certain of these Rules. Release No. 34-53677; File No. PCAOB-2006-01, April 19

_______ 2007, Speech by SEC Staff: The Economics of FIN 48: Accounting for Uncertainty in Income Taxes, Chester S. Spatt, Chief Economist and Director of the Office of Economic Analysis, U.S. Securities and Exchange Commission, Washington, D.C., March 8

_______ 2015, Possible Revisions to Audit Committee Disclosures. Release No. 33-9862; 34-75344, July 1, https://www.sec.gov/rules/concept/2015/33-9862.pdf

Tax Council Policy Institute, 2006. “The TCPI 7th Annual Tax Policy and Practice Symposium – The Corporate Tax Practice in the Age of Transparency: A Path Forward.” Taxes — The Tax Magazine (June), 27–40

US Treasury. 2008. Final Report of the Advisory Committee on the Auditing Profession to the U.S. Department of the Treasury. U.S. Treasury: Washington D.C.

Zelisko, J., 2014. Statement of Brunswick Corporation. For the Hearing Record of the Committee on Ways and Means, U.S. House of Representatives. April 8, 2014. http://waysandmeans.house.gov/uploadedfiles/040814zelisko_testimony.pdf

37

Appendix A Examples of Tax Fee Disclosures

General Electric Proxy Statement (2010) (No breakdown of planning and compliance services) “Tax fees ($8 Million) are fees for tax compliance, tax advice and tax planning”

Categorization: Tax_Fee_Breakdown = 0 Tax_Comply = N/A TAX_PLAN = N/A ExxonMobil Proxy Statement (2009) (Only Compliance disclosed)

“PwC assisted various ExxonMobil affiliates with the preparation of local tax filings and related tax services. These fees were $1.6 million for 2009 (versus $1.4 million in 2008). PwC also assisted in preparing tax returns for individual ExxonMobil expatriate employees. These fees were $0.3 million for 2009 (versus $2.6 million for 2008).”

Categorization: Tax_Fee_Breakdown = 1 Tax_Comply = $1.9 million27 TAX_PLAN = $0 Albemarle Proxy Statement (2006) (Dollar breakdown of planning and compliance services)

“Tax fees include original and amended tax returns, studies supporting tax return amounts as may be required by Internal Revenue Service regulations, claims for refunds, assistance with tax audits and other work directly affecting or supporting the payment of taxes (“compliance”), and planning, research and advice supporting the Company’s efforts to maximize the tax efficiency of its operations (“planning”). For the fiscal years ended December 31, 2006 and December 31, 2005, payments for compliance totaled $460,000 and $2,139,000, respectively, and payments for planning were $198,000 and $201,000, respectively.”

Categorization: Tax_Fee_Breakdown = 1 Tax_Comply = $460 thousand TAX_PLAN = $198 thousand Ford Proxy Statement (2010) (Percentage breakdown of planning and compliance services)

“The Company paid PricewaterhouseCoopers $3.6 million and $4.1 million for tax services for the years ended December 31, 2010 and 2009, respectively. The types of tax services provided included assistance with tax compliance and the preparation of tax returns, tax consultation, planning and implementation services, assistance in connection with tax audits, and tax advice related to mergers, acquisitions and divestitures. Of the fees paid for tax services, the Company paid 72% and 59% for tax compliance and the preparation of Company tax returns in 2010 and 2009, respectively.”

Categorization: Tax_Fee_Breakdown = 1 Tax_Comply = $2.592 million (72% of $3.6 million) TAX_PLAN = $1.008 million (28% of $3.6 million)

27 It is common for the entity level tax preparer to assist with the returns of expatriated employees. Thus, per the AICPA definition of tax compliance (AICPA 2011), expatriate tax services are considered tax compliance.

38

Appendix B Variable Definitions

Variable Name Variable Definition [source] Dependent Variables GETR The effective tax rate calculated as total tax expense [COMPUSTAT

data TXT] divided by pre-tax book income [PI] less special items [SPI]. CETR The cash effective tax rate calculated as cash taxes paid [TXPD] divided

by pre-tax book income [PI] less special items [SPI]. VOL_OF_CETR The standard deviation of one-year CASH_ETR over the time period t to

t+4. UTB Total uncertain tax benefits [TXTUBEND] scaled by total assets [AT].

Independent Variables of Interest TAX_FEE_RATIO Fees paid for auditor provided tax services divided by total fees paid to

the auditor. [Audit Analytics] TAX_PLAN_RATIO Auditor provided tax planning services fees divided by total fees paid to

the auditor. TAX_PLAN_RATIO_WCOMP Auditor provided tax planning services fees divided by total fees paid to

the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP Auditor provided tax planning services fees divided by total fees paid to

the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO Auditor provided tax compliance services fees divided by total fees paid

to the auditor. TAX_COMP_RATIO_WPLAN Auditor provided tax compliance services fees divided by total fees paid

to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO_WOPLAN Auditor provided tax compliance services fees divided by total fees paid

to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. Control Variables ACCT_EXP The percentage of audit committee members with accounting financial

expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). [BoardEx]

OTH_FEE_RATIO Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. [Audit Analytics]

AUD_FEE Log of audit fees. [Audit Analytics]

AUD_CHG Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. [Audit Analytics]

BIG4 Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. [Audit Analytics]

SIZE Log of total assets [AT].

ROA Return on assets calculated as pretax book income [PI] divided by prior year total assets [AT].

FOR_D Indicator variable equal to 1 if firm reports foreign income [PIFO], 0 otherwise.

NOL Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards [TLCF] in previous year, 0 otherwise.

LEV Leverage calculated as long term debt [DLTT] plus debt in current liabilities [DLC] divided by log of prior year total assets [AT].

BM Book to market ratio calculated as the book value of stockholders equity [SEQ] divided by the market value of stockholders equity [CSHO*PRCC_F].

39

RD Research and development expense [XRD] scaled by prior year total assets [AT].

CAPX Capital expenditures [CAPX] scaled by prior year total assets [AT].

DSALES Change in total sales [SALE] from year t-1 to year t.

VOL_OF_PTBI The standard deviation of the ratio of pretax book income [PI] to lagged total assets [AT] over the time period t to t+4.

ESO_BENEFITS The excess tax benefit of stock options [TXBCOF] scaled by lagged total assets. Stock Option Tax Benefit is set to 0 if missing.

IMR Inverse Mills ratio from the decision to purchase tax NAS using the model consistent with Lassila et al. (2010) and McGuire et al. (2012).

Additional Variables for Selection Model APTS_INDICATOR Indicator variable equal to 1 if the firm purchased tax services from their

external auditor; 0 otherwise. ESO_INDUSTRY Indicator variable set equal to one if a firm operates in an industry with

potentially large tax deductions from the exercise of options (defined as industry SIC codes 30–39 and 70–89); 0 otherwise.

AUD_TENURE Length of the audit firm’s tenure with its client. [Audit Analytics]

ACQUIRE Indicator variable equal to 1 if firm participated in any merger activity during year t or year t-1; 0 otherwise.

OPPORTUNITY Market value of a client divided by the sum of the market value of all clients in the same industry at the same MSA city.

TAXLOSS Tax net operating loss carryforward reported at the end of year t, scaled by total assets at the end of year t [TLCF/AT].

DISC_EXTRA Indicator variable set equal to one when a firm reports a large discretionary/extraordinary item [defined as discontinued and extraordinary items from the statement of cash flows [XIDOC][1 % of revenue [REVT]]; 0 otherwise

DNOL Change in net operating loss carryforwards [TLCF] from year t-1 to year t.

LTD Long Term Debt scaled by total assets [DLTT/AT].

BTM Book-to-market ratio for the end of year t, measured as book value of equity divided by market value of equity [CEQ/PRCC_F*CSHO].

TIER2 Indicator variable equal to 1 if auditor is BDO Seidman, or Grant Thornton; 0 otherwise.

PPE Net PPE for year t scaled by total assets [PPENT/AT].

CASH Cash holding at the end of year t divided by total assets [CH/AT].

DEP Depreciation expense for year t divided by total assets [DEP/AT].

40

Table 1 - Sample SelectionObservations

Initial firm-year observations without missing Audit Analytics, Compustat and BoardEx variables 11,324 Firm-year observations with negative pretax income (3,997) Firms not incorporated in the US and “non-corporate firms” (294) Financial Firms (734)

Base Sample (2007-2012) 6,299

1) No Disclosure ofTax Fee Breakdown

2) Disclose Tax Fee Breakdown

3) Firm-year observations with no Tax NAS

3,556 1,133 1,610

Observations

Observations2,743

5 year sequential year requirement (1,844) Final Tax Risk Sample 899

Observations2,743

UTB Disclosure requirement (518) Final UTB Sample 2,225

UTB Model

Volatility of Cash ETR Model

Firm-year observations with positive Tax NAS

ETR Model

2,743

41

Table 2 –Decision to Purchase APTS DV=APTS_INDICATOR Hypothesized

Directional

Expectation OTH_FEE_RATIO + 0.071 (0.33) AUD_FEE + 0.124*** (2.83) ESO_INDUSTRY ? -0.838** (-2.49) AUD_TENURE + 0.017*** (5.08) ACQUIRE + 0.081^ (1.41) OPPORTUNITY + 0.414*** (2.74) TAXLOSS ? -0.061 (-1.16) DISC_EXTRA ? -0.026 (-0.23) SIZE + 0.055* (1.86) NOL + 0.031 (0.63) DNOL + -0.041 (-0.85) FOR_D + 0.193*** (3.17) RD + -0.022 (-0.03) LTD + -0.129 (-0.85) BTM + -0.051^ (-1.30) BIG4 + 0.193** (1.98) TIER2 - -0.238** (-2.35) ROA + 0.140 (1.15) PPE + 0.010 (0.05) CASH + 0.002 (0.01) DEP + 0.356 (0.37) Constant -1.405*** (-4.56) Industry and Year Fixed Effects Yes Observations 11733 Pseudo R2 0.126 Area Under ROC 0.7444

42

Numbers in parentheses are test statistics based on robust standard errors clustered at the firm level. *, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Variables are defined as follows: APTS_INDICATOR = Indicator variable equal to 1 if the firm purchased tax services from their external auditor; 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. ESO_INDUSTRY= Indicator variable set equal to one if a firm operates in an industry with potentially large tax deductions from the exercise of options (defined as industry SIC codes 30–39 and 70–89) and set equal to zero otherwise. AUD_TENURE = Length of the audit firm’s tenure with its client. ACQUIRE = Indicator variable equal to 1 if firm participated in any merger activity during year t or year t-1; 0 otherwise. OPPORTUNITY= Market value of a client divided by the sum of the market value of all clients in the same industry at the same MSA city. TAXLOSS=tax net operating loss carryforward reported at the end of year t, scaled by total assets at the end of year t. DISC_EXTRA= Indicator variable set equal to one when a firm reports a large discretionary/extraordinary item defined as discontinued and extraordinary items from the statement of cash flows (1 % of revenue); 0 otherwise. SIZE = Log of total assets. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. DNOL = Change in net operating loss carryforwards from year t-1 to year t. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. RD=Research and development expense scaled by prior year total assets. LTD=Long Term Debt scaled by total assets. BTM=Book-to-market ratio for the end of year t, measured as book value of equity divided by market value of equity. BIG4=Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. TIER2=Indicator variable equal to 1 if auditor is BDO Seidman, or Grant Thornton, 0 otherwise. ROA= Return on assets calculated as pretax book income divided by prior year total assets. PPE=Net PPE for year t scaled by total assets at the beginning of the year. CASH= Cash holding at the end of year t divided by total assets at the beginning of the year. DEP= Depreciation and amortization expense for year t divided by total assets at the beginning of the year. See Appendix B for detailed variable definitions included data sources.

43

Table 3 Descriptive Statistics

Panel A - Effective tax rate reduction (GETR) Model Descriptive Statistics

Firms Disclosing Tax Fee

Breakdown and firms purchasing no APTS

TAX_PLAN_RATIO > 0 TAX_PLAN_RATIO = 0 Expected

Sign Test of

Differences/Proportions Mean Median Std Dev Mean Median Std Dev Mean Median Std Dev TAX_FEE_RATIO 0.056 0.000 0.100 0.166 0.155 0.124 0.020 0.000 0.054 TAX_PLAN_RATIO 0.023 0.000 0.059 0.094 0.068 0.087 0.000 0.000 0.000 TAX_PLAN_RATIO_WCOMP 0.015 0.000 0.047 0.061 0.027 0.077 0.000 0.000 0.000 TAX_PLAN_RATIO_WOCOMP 0.007 0.000 0.034 0.030 0.000 0.064 0.000 0.000 0.000 TAX_COMP_RATIO 0.032 0.000 0.066 0.071 0.043 0.085 0.020 0.000 0.053 TAX_COMP_RATIO _WPLAN 0.017 0.000 0.051 0.071 0.043 0.083 0.000 0.000 0.000 TAX_COMP_RATIO _WOPLAN 0.014 0.000 0.044 0.000 0.000 0.000 0.019 0.000 0.050 GETR 0.310 0.327 0.115 0.283 0.297 0.104 0.319 0.336 0.117 - -7.11*** ACCT_EXP 0.408 0.333 0.249 0.421 0.333 0.254 0.404 0.333 0.247 + 1.61 AUD_CHG 0.102 0.000 0.303 0.034 0.000 0.181 0.125 0.000 0.331 ? -6.79*** OTH_FEE_RATIO 0.057 0.024 0.086 0.056 0.030 0.074 0.058 0.021 0.090 ? -0.39 AUD_FEE 13.851 13.830 1.206 14.573 14.395 1.022 13.615 13.628 1.168 + 19.10*** BIG4 0.740 1.000 0.439 0.957 1.000 0.202 0.669 1.000 0.471 + 14.86*** SIZE 6.710 6.616 1.832 7.737 7.681 1.583 6.374 6.234 1.783 + 17.76*** ROA 0.095 0.088 0.100 0.104 0.091 0.083 0.093 0.086 0.105 - 2.70*** FOR_D 0.515 1.000 0.500 0.720 1.000 0.449 0.447 0.000 0.497 + 12.34*** NOL 0.427 0.000 0.495 0.538 1.000 0.499 0.391 0.000 0.488 ? 6.70*** LEV 0.459 0.446 0.220 0.508 0.503 0.200 0.442 0.426 0.224 - 6.86*** BM 0.605 0.505 0.445 0.500 0.435 0.345 0.639 0.534 0.468 - -7.13*** DSALES 192.240 31.501 936.895 357.555 86.200 1176.600 138.118 21.716 837.029 + 5.32*** RD 0.025 0.000 0.044 0.029 0.007 0.043 0.024 0.000 0.045 ? 2.60*** CAPX 0.053 0.035 0.058 0.044 0.033 0.044 0.056 0.036 0.062 ? -4.67*** Observations 2,743 679 2,064

*, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Test of Differences (Proportions) are reported for continuous (binary) variables. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. TAX_PLAN_RATIO_WCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO _WPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO _WOPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. See Appendix B for detailed variable definitions included data sources.

44

Panel B – Cash tax savings (CETR) Model Descriptive Statistics

Firms Disclosing Tax Fee

Breakdown and firms purchasing no APTS

TAX_PLAN_RATIO > 0 TAX_PLAN_RATIO = 0 Expected

Sign Test of

Differences/Proportions Mean Median Std Dev Mean Median Std Dev Mean Median Std Dev TAX_FEE_RATIO 0.056 0.000 0.100 0.166 0.153 0.125 0.020 0.000 0.054 TAX_PLAN_RATIO 0.023 0.000 0.059 0.094 0.068 0.086 0.000 0.000 0.000 TAX_PLAN_RATIO_WCOMP 0.015 0.000 0.047 0.061 0.027 0.078 0.000 0.000 0.000 TAX_PLAN_RATIO_WOCOMP 0.007 0.000 0.034 0.029 0.000 0.063 0.000 0.000 0.000 TAX_COMP_RATIO 0.032 0.000 0.066 0.071 0.043 0.085 0.020 0.000 0.053 TAX_COMP_RATIO _WPLAN 0.017 0.000 0.051 0.071 0.043 0.084 0.000 0.000 0.000 TAX_COMP_RATIO _WOPLAN 0.015 0.000 0.044 0.000 0.000 0.000 0.019 0.000 0.050 CETR 0.240 0.231 0.179 0.239 0.232 0.154 0.241 0.230 0.186 - -0.25 ACCT_EXP 0.409 0.333 0.250 0.422 0.333 0.255 0.405 0.333 0.248 + 1.57 AUD_CHG 0.102 0.000 0.303 0.034 0.000 0.182 0.125 0.000 0.331 ? -6.72*** OTH_FEE_RATIO 0.057 0.024 0.086 0.056 0.030 0.075 0.057 0.021 0.090 ? -0.31 AUD_FEE 13.858 13.834 1.202 14.570 14.393 1.023 13.624 13.634 1.164 + 18.82*** BIG4 0.739 1.000 0.439 0.957 1.000 0.203 0.668 1.000 0.471 + 14.80*** SIZE 6.712 6.615 1.833 7.735 7.666 1.591 6.377 6.235 1.783 + 17.59*** ROA 0.096 0.089 0.100 0.105 0.092 0.083 0.093 0.086 0.105 - 2.70*** FOR_D 0.516 1.000 0.500 0.720 1.000 0.449 0.450 0.000 0.498 + 12.18*** NOL 0.427 0.000 0.495 0.534 1.000 0.499 0.392 0.000 0.488 ? 6.49*** LEV 0.459 0.446 0.220 0.508 0.501 0.201 0.444 0.428 0.223 - 6.61*** BM 0.604 0.505 0.443 0.500 0.433 0.346 0.638 0.534 0.465 - -7.08*** DSALES 194.184 32.200 941.486 361.985 86.250 1181.180 139.205 22.372 841.448 + 5.35*** RD 0.025 0.000 0.044 0.029 0.007 0.043 0.023 0.000 0.044 ? 2.90*** CAPX 0.053 0.035 0.057 0.044 0.032 0.043 0.056 0.037 0.061 ? -4.81*** Observations 2,718 672 2,046

*, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Test of Differences (Proportions) are reported for continuous (binary) variables. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. TAX_PLAN_RATIO_WCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO _WPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO _WOPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. See Appendix B for detailed variable definitions included data sources.

45

Panel C – Tax Risk (VOL_OF_CETR) Model Descriptive Statistics

Firms Disclosing Tax Fee Breakdown and firms not

purchasing Tax NAS TAX_PLAN_RATIO > 0 TAX_PLAN_RATIO = 0

Expected Sign

Test of Differences/ Proportions Mean Median Std Dev Mean Median Std Dev Mean Median Std Dev

TAX_FEE_RATIO 0.057 0.000 0.101 0.163 0.157 0.125 0.021 0.000 0.057 TAX_PLAN_RATIO 0.022 0.000 0.056 0.088 0.060 0.082 0.000 0.000 0.000 TAX_PLAN_RATIO_WCOMP 0.015 0.000 0.046 0.059 0.027 0.076 0.000 0.000 0.000 TAX_PLAN_RATIO_WOCOMP 0.007 0.000 0.031 0.027 0.000 0.058 0.000 0.000 0.000 TAX_COMP_RATIO 0.034 0.000 0.070 0.073 0.043 0.092 0.021 0.000 0.055 TAX_COMP_RATIO _WPLAN 0.018 0.000 0.055 0.072 0.043 0.089 0.000 0.000 0.000 TAX_COMP_RATIO _WOPLAN 0.015 0.000 0.046 0.000 0.000 0.000 0.020 0.000 0.053 VOL_OF_CETR 0.224 0.093 0.452 0.149 0.077 0.267 0.249 0.102 0.497 - -2.92*** GETR 0.313 0.335 0.109 0.294 0.308 0.093 0.319 0.340 0.114 - -3.07*** ACCT_EXP 0.415 0.333 0.253 0.423 0.400 0.268 0.412 0.333 0.248 + 0.55 AUD_CHG 0.078 0.000 0.268 0.035 0.000 0.184 0.092 0.000 0.290 ? -2.80*** OTH_FEE_RATIO 0.051 0.023 0.074 0.049 0.027 0.064 0.052 0.021 0.077 ? -0.65 AUD_FEE 13.968 13.903 1.183 14.653 14.472 1.017 13.734 13.711 1.144 + 10.79*** BIG4 0.762 1.000 0.426 0.969 1.000 0.173 0.692 1.000 0.462 + 8.52*** SIZE 6.783 6.670 1.827 7.759 7.652 1.551 6.450 6.273 1.796 + 9.85*** ROA 0.112 0.096 0.078 0.117 0.102 0.073 0.111 0.094 0.080 - 1.06 FOR_D 0.502 1.000 0.500 0.694 1.000 0.462 0.436 0.000 0.496 + 6.75*** NOL 0.394 0.000 0.489 0.437 0.000 0.497 0.379 0.000 0.486 ? 1.53 LEV 0.466 0.460 0.219 0.519 0.508 0.208 0.448 0.440 0.220 - 4.28*** DSALES 108.943 17.823 1040.510 132.783 61.983 1269.420 100.818 14.444 950.971 + 0.40 RD 0.023 0.000 0.041 0.025 0.004 0.039 0.022 0.000 0.042 ? 0.78 CAPX 0.054 0.039 0.052 0.046 0.036 0.040 0.057 0.040 0.056 ? -2.76*** VOL_OF_PTBI 0.048 0.034 0.046 0.037 0.029 0.032 0.052 0.036 0.049 - -4.32*** ESO_BENEFITS 0.003 0.000 0.006 0.003 0.001 0.005 0.003 0.000 0.006 - 0.49 Observations 899 229 670

*, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Test of Differences (Proportions) are reported for continuous (binary) variables. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. TAX_PLAN_RATIO_WCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO _WPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO _WOPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. VOL_OF_PTBI = standard deviation of the ratio of pretax book income divided by lagged total assets over the time period t to t+4. ESO_BENEFITS = excess tax benefit of stock options scaled by lagged total assets. Stock Option Tax Benefit is set to 0 if missing. See Appendix B for detailed variable definitions included data sources.

46

Panel D – Tax Risk (UTB) Model Descriptive Statistics

Firms Disclosing Tax Fee Breakdown and firms not

purchasing Tax NAS TAX_PLAN_RATIO > 0 TAX_PLAN_RATIO = 0

Expected Sign

Test of Differences/ Proportions Mean Median Std Dev Mean Median Std Dev Mean Median Std Dev

TAX_FEE_RATIO 0.057 0.000 0.104 0.173 0.164 0.126 0.018 0.000 0.054 TAX_PLAN_RATIO 0.024 0.000 0.060 0.095 0.068 0.087 0.000 0.000 0.000 TAX_PLAN_RATIO_WCOMP 0.016 0.000 0.047 0.062 0.030 0.076 0.000 0.000 0.000 TAX_PLAN_RATIO_WOCOMP 0.008 0.000 0.035 0.030 0.000 0.064 0.000 0.000 0.000 TAX_COMP_RATIO 0.033 0.000 0.069 0.076 0.045 0.089 0.018 0.000 0.052 TAX_COMP_RATIO _WPLAN 0.019 0.000 0.055 0.076 0.045 0.087 0.000 0.000 0.000 TAX_COMP_RATIO _WOPLAN 0.013 0.000 0.044 0.000 0.000 0.000 0.017 0.000 0.051 UTB 0.011 0.005 0.015 0.013 0.008 0.014 0.010 0.004 0.015 + 3.48*** GETR 0.302 0.324 0.120 0.277 0.291 0.102 0.311 0.333 0.124 + -5.91*** ACCT_EXP 0.414 0.333 0.248 0.414 0.333 0.247 0.414 0.333 0.249 ? -0.05 AUD_CHG 0.090 0.000 0.287 0.025 0.000 0.156 0.113 0.000 0.316 ? -6.28*** OTH_FEE_RATIO 0.054 0.023 0.081 0.057 0.029 0.074 0.053 0.020 0.083 + 1.07 AUD_FEE 13.961 13.911 1.100 14.585 14.382 0.972 13.749 13.715 1.060 + 16.51*** BIG4 0.775 1.000 0.417 0.961 1.000 0.194 0.712 1.000 0.453 + 12.23*** SIZE 6.715 6.606 1.708 7.631 7.539 1.528 6.404 6.301 1.654 - 15.52*** ROA 0.109 0.090 0.078 0.109 0.091 0.072 0.109 0.089 0.080 + 0.08 FOR_D 0.568 1.000 0.496 0.762 1.000 0.426 0.502 1.000 0.500 ? 10.80*** NOL 0.472 0.000 0.499 0.562 1.000 0.497 0.442 0.000 0.497 - 4.93*** LEV 0.460 0.452 0.216 0.490 0.490 0.191 0.450 0.432 0.223 + 3.75*** DSALES 256.492 45.025 978.444 412.519 100.447 1168.470 203.513 33.623 899.155 ? 4.40*** RD 0.028 0.001 0.048 0.031 0.009 0.045 0.027 0.000 0.050 ? 1.66* CAPX 0.050 0.033 0.053 0.041 0.029 0.040 0.053 0.035 0.056 ? -4.87*** Observations 2,225 564 1,661

*, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Test of Differences (Proportions) are reported for continuous (binary) variables. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. TAX_PLAN_RATIO_WCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO _WPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO _WOPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. UTB = Total uncertain tax benefits scaled by total assets. ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. VOL_OF_PTBI = standard deviation of the ratio of pretax book income divided by lagged total assets over the time period t to t+4. ESO_BENEFITS = excess tax benefit of stock options scaled by lagged total assets. Stock Option Tax Benefit is set to 0 if missing. See Appendix B for detailed variable definitions included data sources.

47

Table 4 Tests of H1a and H1b

Dependent Variable GETR CETR (1) (2) (3) (4) (5) (6) TAX_FEE_RATIO -0.029* -0.069*** 0.008 -0.010 (-1.76) (-2.65) (0.32) (-0.25) TAX_PLAN_RATIO -0.133*** -0.083* (-3.44) (-1.66) TAX_COMP_RATIO -0.019 0.051 (-0.49) (0.79) ACCT_EXP -0.006 -0.007 -0.008 -0.013 -0.037** -0.037** (-0.81) (-0.70) (-0.74) (-1.18) (-2.28) (-2.32) AUD_CHG -0.007 0.003 0.002 0.014 0.025* 0.025* (-1.04) (0.29) (0.25) (1.49) (1.82) (1.80) OTH_NAF_RATIO -0.010 -0.067** -0.068** -0.027 -0.055 -0.056 (-0.50) (-2.25) (-2.28) (-0.94) (-1.43) (-1.45) AUD_FEE -0.010*** -0.011** -0.011** 0.005 0.003 0.003 (-2.86) (-2.08) (-2.10) (0.98) (0.44) (0.43) BIG4 0.017** 0.013 0.014 0.007 0.017 0.018 (2.33) (1.35) (1.44) (0.69) (1.19) (1.26) SIZE -0.006*** -0.006* -0.006* -0.016*** -0.017*** -0.017*** (-2.61) (-1.70) (-1.65) (-4.42) (-3.31) (-3.26) ROA 0.159*** 0.112*** 0.112*** 0.183*** 0.157*** 0.156*** (6.20) (2.99) (3.00) (4.84) (2.98) (2.97) FOR_D -0.014*** -0.015** -0.013* 0.010 0.013 0.014 (-2.76) (-2.07) (-1.90) (1.28) (1.14) (1.26) NOL -0.010** -0.010* -0.010* -0.024*** -0.029*** -0.029*** (-2.56) (-1.84) (-1.83) (-4.23) (-3.36) (-3.35) LEV 0.008 0.000 -0.000 -0.057*** -0.058** -0.059** (0.76) (0.03) (-0.00) (-3.57) (-2.49) (-2.52) BM 0.006 0.007 0.006 -0.006 -0.017 -0.018 (1.05) (0.77) (0.74) (-0.71) (-1.41) (-1.43) DSALES 0.000 -0.000 -0.000 -0.000 0.000 0.000 (0.13) (-0.01) (-0.08) (-1.13) (0.30) (0.25) R&D -0.267*** -0.360*** -0.369*** -0.530*** -0.528*** -0.538*** (-4.24) (-3.66) (-3.73) (-6.34) (-4.25) (-4.31) CAPX 0.009 0.066 0.067 -0.368*** -0.342*** -0.342*** (0.20) (0.97) (0.98) (-6.21) (-3.68) (-3.69) IMR -0.005 -0.020 -0.017 -0.083** -0.078 -0.075 (-0.24) (-0.74) (-0.65) (-2.47) (-1.53) (-1.48) Constant -0.472*** -0.357*** -0.359*** 0.326*** 0.163* 0.158* (-11.18) (-6.02) (-6.04) (3.87) (1.80) (1.75) Observations 6276 2743 2743 6227 2718 2718 Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.132 0.127 0.128 0.115 0.142 0.143

48

Table 4 Tests of H1a and H1b, continued

Numbers in parentheses are test statistics based on robust standard errors clustered at the firm level. *, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. UTB = Total uncertain tax benefits [TXTUBEND] scaled by total assets [AT]. ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. IMR is the inverse Mills ratio from a first stage decision to purchase tax NAS model (see Table 2). See Appendix B for detailed variable definitions included data sources.

49

Table 5 Tests of H2

Panel A – Tax Volatility Dependent Variable VOL_OF_CETR (1) (2) (3) (4) (5) (6) TAX_FEE_RATIO -0.099 -0.110 -0.118 -0.119 (-1.07) (-0.72) (-1.26) (-0.78) TAX_PLAN_RATIO -0.432** -0.438** (-2.42) (-2.45) TAX_COMP_RATIO 0.121 0.112 (0.47) (0.43) GETR 0.125 0.333* 0.318* (1.13) (1.76) (1.70) CETR 0.371*** 0.452*** 0.448*** (4.70) (3.08) (3.06) ACCT_EXP -0.014 0.042 0.041 -0.013 0.048 0.048 (-0.36) (0.53) (0.52) (-0.34) (0.63) (0.62) OTH_NAF_RATIO -0.021 0.061 0.057 -0.051 0.016 0.013 (-0.16) (0.26) (0.25) (-0.41) (0.07) (0.06) AUD_FEE 0.036* 0.058** 0.058** 0.033* 0.056** 0.056** (1.96) (2.04) (2.05) (1.79) (2.03) (2.04) AUD_CHG 0.005 0.044 0.045 0.008 0.044 0.044 (0.14) (0.67) (0.67) (0.23) (0.68) (0.68) BIG4 -0.064 -0.059 -0.054 -0.056 -0.059 -0.054 (-1.12) (-0.66) (-0.60) (-1.01) (-0.69) (-0.63) SIZE -0.066*** -0.092*** -0.091*** -0.060*** -0.084*** -0.084*** (-3.73) (-3.51) (-3.51) (-3.45) (-3.38) (-3.38) ROA -1.091*** -0.936*** -0.930*** -1.095*** -0.936*** -0.931*** (-8.08) (-4.74) (-4.74) (-8.19) (-4.80) (-4.80) FOR_D 0.005 -0.011 -0.004 0.000 -0.022 -0.015 (0.14) (-0.25) (-0.09) (0.01) (-0.52) (-0.34) NOL -0.008 0.026 0.025 -0.002 0.036 0.035 (-0.38) (0.72) (0.69) (-0.10) (0.98) (0.94) LEV -0.122* -0.104 -0.116 -0.094 -0.041 -0.053 (-1.93) (-1.00) (-1.11) (-1.54) (-0.43) (-0.56) DSALES 0.000** 0.000 0.000 0.000** 0.000 0.000 (2.35) (0.72) (0.84) (2.55) (1.02) (1.13) R&D -0.828*** -1.201** -1.208** -0.655** -1.037** -1.039** (-2.82) (-2.41) (-2.43) (-2.30) (-2.27) (-2.28) CAPX 0.926* 0.690 0.726 0.993** 0.824 0.859 (1.85) (0.80) (0.85) (2.05) (1.04) (1.08) VOL_OF_PTBI 2.405*** 2.044*** 2.053*** 2.385*** 2.021*** 2.029*** (6.44) (4.51) (4.52) (6.39) (4.54) (4.55) ESO_BENEFITS -2.942** -3.023 -3.217 -1.218 -0.583 -0.828 (-2.26) (-1.38) (-1.47) (-0.90) (-0.27) (-0.38) IMR -0.303* -0.430* -0.422* -0.013 0.048 0.048 (-1.78) (-1.80) (-1.76) (-1.58) (-1.62) (-1.57) Constant 0.358 0.055 0.040 0.262 -0.313 -0.328 (1.26) (0.15) (0.11) (0.97) (-0.89) (-0.93) Observations 2104 899 899 2104 899 899 Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.129 0.103 0.104 0.148 0.122 0.124

50

Table 5

Tests of H2, continued Panel B – Uncertain Tax Benefits

UTB (1) (2) (3) (4) (5) (6) TAX_FEE_RATIO -0.002 -0.006 -0.002 -0.006

(-0.86) (-1.62) (-0.94) (-1.62) TAX_PLAN_RATIO -0.014** -0.013** (-2.11) (-2.04) TAX_COMP_RATIO 0.000 -0.000 (0.07) (-0.02) GETR -0.005* -0.005 -0.005 (-1.93) (-1.20) (-1.20) CETR -0.004*** -0.004* -0.004* (-2.69) (-1.82) (-1.83) ACCT_EXP -0.002 -0.001 -0.001 -0.002 -0.001 -0.002 (-1.55) (-0.71) (-0.77) (-1.58) (-0.85) (-0.90) OTH_NAF_RATIO -0.006** -0.004 -0.004 -0.006** -0.004 -0.004 (-2.05) (-1.03) (-1.07) (-2.05) (-0.85) (-0.89) AUD_FEE 0.004*** 0.004*** 0.004*** 0.004*** 0.004*** 0.004*** (5.97) (3.77) (3.78) (5.96) (3.74) (3.75) AUD_CHG 0.002* 0.003** 0.003** 0.002* 0.003** 0.003** (1.73) (2.07) (2.05) (1.89) (2.05) (2.03) BIG4 -0.003** -0.003 -0.003 -0.002* -0.002 -0.002 (-2.01) (-1.64) (-1.60) (-1.80) (-1.51) (-1.47) SIZE -0.001* -0.000 -0.000 -0.001 -0.000 -0.000 (-1.66) (-0.56) (-0.54) (-1.58) (-0.52) (-0.51) ROA 0.023*** 0.025*** 0.025*** 0.022*** 0.024*** 0.024*** (4.04) (3.22) (3.24) (4.03) (3.04) (3.07) FOR_D 0.002*** 0.003*** 0.004*** 0.002*** 0.003*** 0.003*** (2.81) (2.80) (2.87) (2.88) (2.68) (2.74) NOL -0.000 0.001 0.001 -0.000 0.001 0.001 (-0.11) (0.75) (0.74) (-0.22) (0.69) (0.69) LEV 0.004** 0.002 0.002 0.004* 0.001 0.001 (2.00) (0.53) (0.50) (1.85) (0.44) (0.42) DSALES -0.000* -0.000 -0.000 -0.000* -0.000 -0.000 (-1.77) (-0.49) (-0.52) (-1.82) (-0.48) (-0.52) R&D 0.089*** 0.088*** 0.087*** 0.091*** 0.089*** 0.088*** (7.85) (5.36) (5.27) (7.81) (5.42) (5.32) CAPX -0.012* 0.003 0.004 -0.013** 0.001 0.002 (-1.93) (0.33) (0.40) (-2.04) (0.11) (0.17) IMR -0.004 -0.003 -0.003 -0.004 -0.003 -0.003 (-1.58) (-1.18) (-1.17) (-1.54) (-1.21) (-1.20) Constant -0.035*** -0.016 -0.016 -0.036*** -0.045*** -0.045*** (-3.77) (-0.66) (-0.67) (-3.83) (-4.22) (-4.24) Observations 5733 2225 2225 5668 2196 2196 Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.212 0.231 0.232 0.215 0.230 0.231

51

Table 5 Tests of H2, continued

Numbers in parentheses are test statistics based on robust standard errors clustered at the firm level. *, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. Variables are defined as follows: TAX_FEE_RATIO = Fees paid for auditor provided tax services divided by total fees paid to the auditor; TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. UTB = Total uncertain tax benefits [TXTUBEND] scaled by total assets [AT].ACCT_EXP = Percentage of audit committee members with accounting financial expertise (i.e., certified public accountant, chief financial officer, auditor, chief accounting officer, controller, treasurer or vice president-finance). AUD_CHG = Indicator variable equal to 1 if the company switched auditors in the current or prior year, 0 otherwise. OTH_FEE_RATIO = Non-audit fees other than tax (total fees - audit fees - tax non-audit fees) divided by total fees paid to the auditor. AUD_FEE = Log of audit fees. BIG4 = Indicator variable equal to 1 if auditor is Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers or KPMG, 0 otherwise. SIZE = Log of total assets. ROA = Return on assets calculated as pretax book income divided by prior year total assets. FOR_D = Indicator variable equal to 1 if firm reports foreign income, 0 otherwise. NOL = Indicator variable equal to 1 for firms with positive unused net operating loss carryforwards in previous year, 0 otherwise. LEV = Leverage calculated as long term debt plus debt in current liabilities divided by log of prior year total assets. BM = Book to market ratio calculated as the book value of stockholders equity divided by the market value of stockholders equity. DSALES = Change in total sales from year t-1 to year t (results remain unchanged when it is scaled by total assets). RD = Research and development expense scaled by prior year total assets. CAPX = Capital expenditures scaled by prior year total assets. VOL_OF_PTBI = standard deviation of the ratio of pretax book income divided by lagged total assets over the time period t to t+4. ESO_BENEFITS = excess tax benefit of stock options scaled by lagged total assets. Stock Option Tax Benefit is set to 0 if missing. IMR is the inverse Mills ratio from a first stage decision to purchase tax NAS model (see Table 2). See Appendix B for detailed variable definitions included data sources.

52

Table 6 Additional Analysis: Joint Provision of Tax Services

Panel A – Effective Tax Rate and Cash Tax Savings

Dependent Variable ETR CETR (1) (2) (3) (4) (5) (6) TAX_PLAN_RATIO -0.121*** -0.088* (3.07) (-1.73) TAX_PLAN_RATIO_WCOMP -0.205*** -0.199*** -0.119* -0.139** (-3.91) (-3.73) (-1.87) (-2.18) TAX_PLAN_RATIO_WOCOMP -0.044 -0.045 -0.027 -0.029 (-0.72) (-0.72) (-0.31) (-0.34) TAX_COMP_RATIO 0.003 0.063 (0.08) (0.95) TAX_COMP_RATIO_WPLAN -0.049 -0.006 0.062 0.090 (-1.03) (-0.12) (0.72) (1.00) TAX_COMP_RATIO_WOPLAN 0.001 0.003 0.033 0.035 (0.02) (0.05) (0.38) (0.40) Constant 0.384*** 0.392*** 0.384*** 0.151* 0.162* 0.157* (5.39) (5.47) (5.36) (1.67) (1.75) (1.70) Observations 2743 2743 2743 2718 2718 2718 Controls (Including IMR) Yes Yes Yes Yes Yes Yes Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.129 0.128 0.129 0.143 0.142 0.142

Panel B – Tax Volatility

Dependent Variable VOL_OF_CETR (1) (2) (3) (4) (5) (6) TAX_PLAN_RATIO -0.371* -0.374* (-1.86) (-1.89) TAX_PLAN_RATIO_WCOMP -0.315 -0.196 -0.367 -0.251 (-1.35) (-0.81) (-1.54) (-1.02) TAX_PLAN_RATIO_WOCOMP -0.709*** -0.690** -0.640** -0.620** (-2.60) (-2.47) (-2.44) (-2.31) TAX_COMP_RATIO 0.083 0.088 (0.30) (0.32) TAX_COMP_RATIO_WPLAN 0.003 -0.086 -0.010 -0.074 (0.02) (-0.48) (-0.06) (-0.44) TAX_COMP_RATIO_WOPLAN 0.323 0.312 0.321 0.313 (0.63) (0.60) (0.62) (0.61) GETR 0.321* 0.310* 0.312* (1.71) (1.66) (1.67) CETR 0.447*** 0.446*** 0.444*** (3.05) (3.04) (3.03) Constant 0.069 0.017 0.050 -0.307 -0.353 -0.326 (0.19) (0.05) (0.14) (-0.87) (-0.99) (-0.91) Observations 899 899 899 899 899 899 Controls (Including IMR) Yes Yes Yes Yes Yes Yes Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.103 0.104 0.103 0.123 0.123 0.123

53

Table 6 Additional Analysis: Joint Provision of Tax Services, continued

Panel C – UTB

Dependent Variable UTB (1) (2) (3) (4) (5) (6) TAX_PLAN_RATIO -0.014** -0.014** (-2.04) (-2.01) TAX_PLAN_RATIO_WCOMP -0.007 -0.006 -0.006 -0.005 (-0.87) (-0.68) (-0.72) (-0.60) TAX_PLAN_RATIO_WOCOMP -0.027** -0.027** -0.027** -0.027** (-2.47) (-2.45) (-2.52) (-2.50) TAX_COMP_RATIO -0.002 -0.003 (-0.36) (-0.48) TAX_COMP_RATIO_WPLAN 0.001 -0.004 0.001 -0.004 (0.13) (-0.40) (0.13) (-0.41) TAX_COMP_RATIO_WOPLAN 0.000 -0.000 -0.001 -0.001 (0.06) (-0.00) (-0.09) (-0.15) GETR -0.005 -0.005 -0.005 (-1.16) (-1.20) (-1.16) CETR -0.004* -0.004* -0.004* (-1.82) (-1.83) (-1.81) Constant -0.035*** -0.016 -0.016 -0.044*** -0.045*** -0.044*** (-3.77) (-0.66) (-0.67) (-4.08) (-4.13) (-4.03) Observations 2225 2225 2225 2196 2196 2196 Controls (Including IMR) Yes Yes Yes Yes Yes Yes Industry and Year Fixed Effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.233 0.232 0.233 0.232 0.230 0.231

Numbers in parentheses are test statistics based on robust standard errors clustered at the firm level. *, **, and *** denote two-tailed statistical significance at 10%, 5%, and 1% respectively. TAX_PLAN_RATIO = Auditor provided tax planning services fees divided by total fees paid to the auditor. TAX_COMP_RATIO = Auditor provided tax compliance services fees divided by total fees paid to the auditor. TAX_PLAN_RATIO_WCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO > 0, and equal to 0 otherwise. TAX_PLAN_RATIO_WOCOMP = Auditor provided tax planning services fees divided by total fees paid to the auditor when TAX_COMP_RATIO = 0, and equal to 0 otherwise. TAX_COMP_RATIO _WPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO > 0, and equal to 0 otherwise. TAX_COMP_RATIO _WOPLAN = Auditor provided tax compliance services fees divided by total fees paid to the auditor when TAX_PLAN_RATIO = 0, and equal to 0 otherwise. GETR = Effective tax rate calculated as total tax expense divided by pre-tax book income less special items. CETR = Cash effective tax rate calculated as cash taxes paid divided by pre-tax book income less special items. VOL_OF_CETR = Standard deviation of one-year CASH_ETR over the time period t to t+4. UTB = Total uncertain tax benefits [TXTUBEND] scaled by total assets [AT]. See Appendix B for detailed variable definitions included data sources.