Auditor-Provided Tax Services and Earnings Management in Tax Expense

Abstract

Regulators do not prohibit auditors from providing tax services to their audit clients, provided that these services are preapproved by audit committees. I examine whether the association between auditor-provided tax services and earnings management in tax expense varies with audit committee effectiveness. I develop a composite proxy for audit committee effectiveness by combining six audit committee characteristics. I find that auditor-provided tax services are less positively associated with earnings management in tax expense as audit committee effectiveness increases. My findings suggest that the knowledge spillover effect of auditor-provided tax services is more likely to dominate the independence impairment effect when auditor-provided tax services are preapproved by more effective audit committees.

Keywords

audit committee corporate governance earnings management non-audit fees tax expense

Introduction

I examine whether the association between auditor-provided tax services and earnings management in tax expense varies with audit committee effectiveness. Regulators have long been concerned that non-audit services might impair auditor independence. Although the majority of empirical articles fail to find supporting evidence, the Sarbanes–Oxley Act of 2002 (SOX) and the subsequent Securities and Exchange Commission’s (SEC; 2003) independence rule nevertheless prohibit auditors from providing a variety of non-audit services for their audit clients. However, the SEC determined that it would not prohibit auditor-provided tax services, a specific type of non-audit service that auditors frequently provide to their audit clients, provided that these services are preapproved by the clients’ audit committees (SEC, 2003).¹

Despite this requirement for audit committee preapproval, regulators remain concerned that auditor-provided tax services might impair auditor independence. In 2004, the Public Company Accounting Oversight Board (PCAOB) held a round table to discuss this concern (PCAOB, 2004). In 2005, the PCAOB adopted a final rule (PCAOB, 2005) requiring auditors to discuss the potential impacts of proposed tax services on auditor independence with audit committees. Later, the PCAOB further proposed and adopted a series of rules (PCAOB, 2008, 2010, 2011) encouraging auditors to communicate with audit committees on issues concerning auditor independence, including auditor-provided tax services.

To address regulatory concerns, recent empirical studies have examined the relation between auditor-provided tax services and earnings management and provided mixed evidence. Some studies find that auditor-provided tax services are negatively associated with the incidence of financial restatements (Kinney, Palmrose, & Scholz, 2004; Seetharaman, Sun, & Wang, 2011). However, Omer, Bedard, and Falsetta (2006) do not find an association between auditor-provided tax services and discretionary accruals. Gleason and Mills (2011) suggest that companies purchasing tax services from their auditors do not manage tax reserves to meet/beat analyst forecasts more than other companies.

Prior studies have not considered the roles of audit committees. Larcker and Richardson (2004) argue that the association between non-audit services and earnings management varies with the quality of corporate governance.² As SOX requires that non-audit services, including auditor-provided tax services, be preapproved by audit committees, I extend Larcker and Richardson (2004) and expect the relation between auditor-provided tax services and earnings management to vary with audit committee effectiveness.

Because I study auditor-provided tax services, I specifically focus on earnings management in tax expense where independence impairment and knowledge spillover can most directly occur to increase the power of the test to detect earnings management. On one hand, when auditing income tax provisions, an audit firm is essentially auditing the results of its own prior work if the audit firm also provides tax services to the client (Barrett, 2004). Because of the potential impairment of auditor independence, auditor-provided tax services may be associated with greater earnings management in tax expense.³ On the other hand, knowledge spillover (Simunic, 1984) from tax to audit services provides auditors with more information regarding the tax services audit clients have received, which may assist auditors in evaluating tax expense, and therefore help them combat earnings management in tax expense. Thus, auditor-provided tax services may be associated with less earnings management in tax expense.

I argue that a more effective audit committee is less likely to preapprove auditor-provided tax services that might impair auditor independence. Even if such an audit committee approves those services, it should provide better monitoring to mitigate potential independence impairment. Consequently, when audit committees are more effective, there is a higher probability that the knowledge spillover effect of auditor-provided tax services will dominate the independence impairment effect, and therefore, auditor-provided tax services are more likely to be associated with less earnings management in tax expense. Thus, although all auditor-provided tax services are required to be preapproved by audit committees, the association between auditor-provided tax services and earnings management in tax expense depends on how effective the pre-screening approval and the post-screening monitoring processes are. Following Dhaliwal, Gleason, and Mills (2004), I consider earnings management incentives to meet/beat analyst forecasts and focus on earnings management in tax expense to avoid missing analyst forecasts.

The Blue Ribbon Committee (BRC; 1999) recommends that audit committee effectiveness be examined in a holistic manner. Thus, I follow the methodology in Gompers, Ishii, and Metrick (2003) and develop a composite proxy for audit committee effectiveness, AC_score, by summing dichotomous measures of six audit committee characteristics. I consider an audit committee as more effective if it consists of more members, meets more frequently, has a chair with management expertise, has a larger proportion of financial experts, holds a larger average number of other board positions, and has a longer average board tenure.

Using a sample of 799 Standard & Poor’s (S&P) 1500 companies in 2003, I find that, as AC_score increases, auditor-provided tax services are less positively associated with earnings management in tax expense. These results suggest that the knowledge spillover effect of auditor-provided tax services is more likely to dominate the independence impairment effect as audit committees become increasingly effective. Moreover, when I partition my sample by the median of AC_score, I find that auditor-provided tax services are positively associated with earnings management in tax expense when AC_score is below the median. This evidence suggests that the independence impairment effect of auditor-provided tax services dominates the knowledge spillover effect when audit committee effectiveness is low. In contrast, I find that auditor-provided tax services are negatively associated with earnings management in tax expense when AC_score is above median. This suggests that the knowledge spillover effect of auditor-provided tax services dominates the independence impairment effect when audit committee effectiveness is high. These contrasting results further suggest the importance of considering the role of audit committee effectiveness when assessing the association between auditor-provided tax services and earnings management in tax expense.

Additional analyses suggest that audit committee effectiveness is not predominated by any one of the six underlying attributes. My results are robust to a battery of sensitivity tests. For example, I obtain similar results after entertaining the possibility that audit committee effectiveness could vary non-linearly with certain audit committee characteristics such as size, other board memberships, and board tenure. My results are also robust to controlling for overall corporate governance and accounting for the endogeneity of audit committee effectiveness, and I rule out an alternative explanation that my proxy for earnings management in tax expense could partly reflect the economic benefits of auditor-provided tax services. Furthermore, I do not find that the association between auditor-provided tax services and discretionary accruals varies with AC_score, suggesting that in the context of auditor-provided tax services, focusing on earnings management in tax expense improves detection of earnings management as compared with focusing on aggregate accruals. Finally, I obtain similar results when I extend the sample to 250 S&P 500 companies in 2009.

My article extends Larcker and Richardson (2004) and contributes to the earnings management and corporate governance literature. I provide empirical evidence suggesting that effective audit committees help mitigate the independence impairment effect and facilitate the knowledge spillover benefit of auditor-provided tax services. These results have important policy implications. As audit committee effectiveness is of key importance to realize the net benefit of auditor-provided tax services, my findings support SOX and the SEC’s rule of not entirely banning auditor-provided tax services but requiring audit committee preapproval. Furthermore, policy makers should find my results informative to jointly consider rules governing auditor-provided tax services and rules governing audit committees. Rules aimed at improving audit committee effectiveness used to be motivated by the potential of directly enhancing financial reporting. My results suggest that promoting audit committee effectiveness has an added benefit of indirectly enhancing financial reporting through boosting the value of auditor-provided tax services.

The remainder of this article is organized as follows: “Background, Literature Review, and Hypothesis Development” section provides the background, literature review, and hypothesis development. “Research Design” section presents the research design. “Sample Description and Empirical Results” section describes the sample and discusses the empirical results. “Supplemental and Sensitivity Analyses” section contains supplemental and sensitivity analyses, and the final section concludes.

Background, Literature Review, and Hypothesis Development

Regulators were concerned that non-audit services overall might impair auditor independence, although most empirical studies do not find supporting evidence.⁴ Section 201 of SOX further prohibits auditors from providing many types of non-audit services to their audit clients.⁵ Fulfilling the SOX mandate, the SEC adopted a rule strengthening auditor independence in 2003 (SEC, 2003). The SEC debated whether to proscribe auditor-provided tax services, but ultimately it determined not to do so, provided that such services are preapproved by audit committees. However, the SEC cautions audit committees to scrutinize carefully the nature of auditor-provided tax services before approving them.

Despite the preapproval requirement, regulators’ concerns over auditor-provided tax services’ potential impairment of auditor independence persist. In 2004, the PCAOB held the Auditor Independence and Tax Services Roundtable (PCAOB, 2004) to discuss the impact of auditor-provided tax services on auditor independence. It also adopted a final rule in 2005 (PCAOB, 2005) increasing auditors’ responsibilities in seeking audit committees’ preapproval of auditor-provided tax services. In 2008, the PCAOB passed another rule (PCAOB, 2008) requiring the auditor to discuss with the audit committee any issues concerning auditor independence, including auditor-provided tax services. Recognizing the importance of this discussion, the PCAOB proposed an auditing standard in 2010 (PCAOB, 2010) and reproposed the same standard again in 2011 (PCAOB, 2011) encouraging auditor communication with the audit committee.

In response to regulatory concerns, recent empirical articles have examined the association between auditor-provided tax services and earnings management and have provided inconclusive evidence. Although Kinney et al. (2004) document a negative relation between auditor-provided tax services and the incidence of general financial restatements, Seetharaman et al. (2011) do not find a significant association. Instead, Seetharaman et al. (2011) find that auditor-provided tax services are only related to a lower incidence of tax-related financial restatements. Omer et al. (2006) also do not find that auditor-provided tax services are associated with discretionary accruals and Gleason and Mills (2011) do not find that companies that purchase tax services from their auditors manage tax reserves to meet/beat analyst forecasts more than other companies.

Extant studies have not considered the roles of audit committees. Larcker and Richardson (2004) find that non-audit services, which include auditor-provided tax services, have different impact on earnings management depending on the quality of corporate governance. Because SOX and the SEC require that auditor-provided tax services be scrutinized and preapproved by audit committees, I extend the logic in Larcker and Richardson (2004) and argue that the relation between auditor-provided tax services and earnings management varies with audit committee effectiveness, that is, the quality of audit committees’ screening and monitoring of auditor-provided tax services.

I focus on earnings management in tax expense where the independence impairment and knowledge spillover effects of auditor-provided tax services can most directly occur.⁶ Auditor-provided tax services can be positively or negatively associated with earnings management in tax expense for the following two contrasting reasons.⁷ On one hand, when auditing income tax provisions, audit partners are less likely to challenge the book treatment of the tax services provided by their tax colleagues in the same audit firm (Maydew & Shackelford, 2007). This independence impairment effect predicts that auditor-provided tax services are positively associated with earnings management in tax expense.⁸ On the other hand, evaluating tax expense is difficult because it involves an understanding of complex tax plans and judgment in estimating tax-related reserves.⁹ Arguably, when auditors also provide tax services to their audit clients, information regarding the tax plans and tax positions is more readily available to the auditors, which assists them in evaluating tax expense and combating earnings management. Therefore, this knowledge spillover effect predicts a negative association between auditor-provided tax services and earnings management in tax expense.

One of the primary functions of audit committees is to safeguard auditor independence (Carcello & Neal, 2003). Thus, a more effective audit committee is more likely to preapprove those auditor-provided tax services that are less likely to impair auditor independence. Gaynor, McDaniel, and Neal (2006) also argue that audit committees make the preapproval decisions of auditor-provided non-audit services based on a cost–benefit analysis. When the benefit (i.e., knowledge spillover) is greater than the cost (i.e., independence impairment), they are more likely to preapprove those services. Gaynor et al. (2006) present experimental evidence consistent with this argument.

Gaynor et al. (2006) do not manipulate the level of audit committee effectiveness, that is, they assume that audit committees are all effective. I, however, argue that the level of audit committee effectiveness varies. A more effective audit committee arguably performs a better cost–benefit analysis and is more likely to preapprove those auditor-provided tax services for which the independence impairment effect is dominated by the knowledge spillover effect. Even if an audit committee approves some auditor-provided tax services that might impair auditor independence, a more effective committee should provide better monitoring to mitigate independence impairment. After interviewing 42 public company audit committee members, Beasley, Carcello, Hermanson, and Neal (2009) present survey evidence suggesting that audit committees consider tax-related accounts as one of the key financial reporting risk areas and are often extensively involved in reviewing management estimates, judgments, and assumptions. Anticipating that more effective audit committees are better at evaluating the reasonableness of tax expense, auditors are less likely to compromise their independence to allow more earnings management in tax expense. Thus, I expect that auditor-provided tax services are less likely to be positively associated with earnings management in tax expense as audit committee effectiveness increases.

Skinner and Sloan (2002) suggest that companies have strong incentives to avoid missing analyst forecasts because the capital market rewards companies that meet/beat analyst forecasts and punishes companies that miss these forecasts. Dhaliwal et al. (2004) further suggest that companies manage tax expense to avoid missing analyst forecasts. I examine whether the association between auditor-provided tax services and earnings management in tax expense to avoid missing analyst forecasts varies with audit committee effectiveness. I state the following hypothesis in its alternative form:

Hypothesis 1: As audit committee effectiveness increases, auditor-provided tax services are less positively associated with earnings management in tax expense to avoid falling short of analyst forecasts.

Research Design

I examine whether the relation between auditor-provided tax services and earnings management in tax expense varies with audit committee effectiveness using the following ordinary least squares (OLS) regression Model 1. The model is adapted and augmented from the regression models used in Dhaliwal et al. (2004). To reduce the influence of outliers, all continuous variables are winsorized at their 1st and 99th percentiles, respectively. All test statistics are based on the White robust standard errors.

\begin{matrix} E T R Q 4_E T R Q 3_{i t} = β_{0} + β_{1} I n d u c e d_c h g_E T R_{i t} + β_{2} T a x_o w e d_{i t} + β_{3} E T R Q 3 + β_{4} M i s s_{i t} \\ + β_{5} l n T a x f e e_{i t} + β_{6} M i s s_{i t} \times l n T a x f e e_{i t} + β_{7} l n T o t a l f e e_{i t} + β_{8} M i s s_{i t} \times l n T o t a l f e e_{i t} \\ + β_{9} A C_s c o r e_{i t} + β_{10} M i s s_{i t} \times A C_s c o r e_{i t} + β_{11} l n T a x f e e_{i t} \times A C_s c o r e_{i t} \\ + β_{12} M i s s_{i t} \times l n T a x f e e_{i t} \times A C_s c o r e_{i t}, \end{matrix}

where ETRQ4_ETRQ3 = the fourth-quarter effective tax rate (ETR; ETRQ4) minus the third-quarter ETR (ETRQ3), where ETR is defined as accumulated tax expense (Compustat quarterly #6) divided by accumulated pre-tax income (quarterly #23). Induced_chg_ETR = induced tax change divided by pre-tax income, where induced tax change equals (35% −ETRQ3) × unexpected pre-tax income. The unexpected pre-tax income is estimated as (I/B/E/S actual EPS − I/B/E/S most recent median consensus forecast) × common shares outstanding (annual #25)/(1 − 35%). Tax_owed = taxes payable (annual #71) net of refunds (annual #161), scaled by pre-tax income. ETRQ3 = the third-quarter estimate of annual ETR, calculated as accumulated tax expense divided by accumulated pre-tax income up to the third quarter. Miss = 1 if earnings absent fourth-quarter ETR adjustment [pre-tax income (annual #170) × (1 −ETRQ3) / common shares outstanding (annual #25)], rounded to the nearest cent, is lower than the most recent median consensus analyst forecast before earnings announcement, and 0 otherwise. lnTaxfee = natural logarithm of Taxfee, where Taxfee is fees paid to the auditor for tax services, in millions. lnTotalfee = natural logarithm of Totalfee, where Totalfee is total fees paid to the auditor, in millions, including audit fees, audit-related fees, tax fees, and other fees. AC_score = summary index of audit committee effectiveness that is equal to the sum of the values of the following six dichotomous variables: ACsize_D = 1 if the number of members on the audit committee is above the sample median, and 0 otherwise. ACmtg_D = 1 if the number of meetings held annually by the audit committee is above the sample median, and 0 otherwise. ACchairmgt = 1 if the audit committee chair is an individual with experience as a CEO or president of a company, and 0 otherwise. ACfin_D = 1 if the proportion of financial experts on the audit committee is above the sample median, and 0 otherwise. A financial expert is an individual with experience as a public accountant, auditor, principal or chief financial officer, controller, principal or chief accounting officer, CEO or president of a for-profit corporation, investment banker, venture capitalist or financial analyst. ACgov_D = 1 if the average number of other board positions audit committee members hold is above the sample median, and 0 otherwise. ACtenure_D = 1 if the average board tenure of audit committee members is above the sample median, and 0 otherwise.

The dependent variable, ETRQ4_ETRQ3, is a measure of earnings management in tax expense, which is defined as the difference between the annual ETR at year end and the estimated annual ETR in the third quarter.¹⁰ A more negative ETRQ4_ETRQ3, that is, a larger decrease in ETR from the third quarter to the fourth quarter, indicates greater earnings management in tax expense. Induced_chg_ETR controls for the variation in ETRQ4_ETRQ3 that is due to unexpected earnings surprise not related to earnings management.¹¹ I expect its coefficient to be positive. Dhaliwal et al. (2004) suggest a public data proxy for the extent of over-or-underpayment of estimated taxes, Tax_Owed, which controls for unexpected changes in ETR due to fourth-quarter misestimation. I predict its coefficient to be positive. ETRQ3 controls for possible mean reversion and I expect its coefficient to be negative. Miss is an indicator variable, which takes the value 1 if pre-managed earnings (i.e., pro forma earnings using ETRQ3) would have missed analyst forecasts, and 0 otherwise. A negative coefficient would corroborate the finding in Dhaliwal et al. (2004), suggesting that companies manage tax expense to avoid missing analyst forecast.

When I further include the interaction between Miss and lnTaxfee, where lnTaxfee is the natural logarithm of fees paid to the auditor for tax services, its coefficient indicates the association between auditor-provided tax services and earnings management in tax expense. A negative (positive) coefficient suggests that auditor-provided tax services are positively (negatively) associated with earnings management in tax expense. I do not predict the sign on Miss × LnTaxfee for the overall sample as it is an empirical question whether the independence impairment effect or the knowledge spillover effect of auditor-provided tax services dominates. As Kinney and Libby (2002) suggest that economic bonding may be created by total fees, I control for Miss × lnTotalfee.¹² Although DeAngelo (1981) predicts that auditors have incentives to compromise independence for larger clients, Reynolds and Francis (2001) find that, due to reputation and litigation concerns, auditors report more conservatively for larger clients. Thus, I do not predict the sign on Miss × lnTotalfee.

Finally, to test my hypothesis, I include a three-way interaction term Miss × lnTaxfee × AC_score, where AC_score is a summary index to measure audit committee effectiveness, and predict a positive sign. Because of the three-way interaction, all the two-way interactions and main effects are also included. However, I caution that the signs on the two-way interactions and main effects can no longer be easily interpreted after the inclusion of the three-way interaction, that is, the signs on Miss and Miss × lnTaxfee in the full model cannot be predicted or interpreted as discussed above.

Following the recommendation of BRC (1999), I examine audit committee effectiveness in a holistic manner. I construct AC_score based on economic and organizational theories by combining the following six audit committee characteristics: committee size, meeting frequency, whether the committee chair has management experience, the proportion of financial experts on the committee, the average number of other board positions committee members hold, and the average board tenure of committee members.¹³

Audit committee size in general is found to be positively associated with audit committee effectiveness. A larger audit committee is likely associated with a larger monitoring capacity. Reflecting concerns over understaffed audit committees (Vafeas, 2005), the BRC (1999) recommends listed companies include at least three directors on the audit committees.¹⁴ This suggests that regulators believe larger committee size improves audit committee effectiveness. Consistent with this, the literature finds that audit committee size is negatively associated with the cost of debt financing (Anderson, Mansi, & Reeb, 2004) and the incidence of restatements (Carcello, Neal, Palmrose, & Scholz, 2011).

The BRC (1999) argues that the complex nature of accounting and financial matters requires a significant amount of time for audit committee members to devote to committee matters. Kalbers and Fogarty (1993) and DeZoort, Hermanson, Archambeault, and Reed (2002) both emphasize that diligence fundamentally contributes to audit committee effectiveness. More frequent audit committee meetings suggest that committee members are more diligent in performing their monitoring functions and, thus, should improve audit committee effectiveness. Consistent with this, prior research has shown that more frequent audit committee meetings are associated with a reduction in accruals (Xie, Davidson, & DaDalt, 2003), fraud (Farber, 2005), restatements (Abbott, Parker, & Peters, 2004), and the cost of debt financing (Anderson et al., 2004).

Kolodny and Kiggundu (1980) emphasize the role of leadership on group performance. The empirical audit committee literature has largely ignored this leadership issue and therefore has not investigated the roles of audit committee chairs. Gendron and Bédard (2006) report that well-structured agenda and committee meetings improve audit committee effectiveness. Because the committee chair usually determines the agenda and leads the meeting (Beasley et al., 2009), I argue that the audit committee is more effective when the chair has management experience as a group leader such as a CEO or president of a company.

SOX Section 407 requires a company to disclose whether there is at least one financial expert on the audit committee and, if not, to explain why not. Financial expertise improves individuals’ ability to understand financial statements, and thus helps improve their monitoring quality (Kalbers & Fogarty, 1993). Consistent with this, prior empirical studies find that financial expertise is associated with an increase in audit fees (Abbott, Parker, Peters, & Raghunandan, 2003) and accounting conservatism (Krishnan & Visvanathan, 2008), and a reduction in earnings restatements (Abbott et al., 2004; Agrawal & Chadha, 2005), fraud (Farber, 2005), internal control problems (Hoitash, Hoitash, & Bedard, 2009; Krishnan, 2005), accruals (Bédard, Chtourou, & Courteau, 2004), and the cost of debt financing (Anderson et al., 2004). Following DeFond, Hann, and Hu (2005) and Hoitash et al. (2009), I define a financial expert as an individual with experience as a public accountant, auditor, principal or chief financial officer, controller, principal or chief accounting officer, CEO or president of a for-profit corporation, investment banker, venture capitalist or financial analyst.¹⁵

Audit committee effectiveness should also increase with committee members’ governance expertise as measured by the number of other board positions they hold. Srinivasan (2005) finds that, after restatements, audit committee members frequently lose not only their positions in the restatement companies but also their other board positions. I thus expect that audit committee members’ monitoring incentives increase with the number of other board positions they hold. Furthermore, audit committee members’ experience in performing monitoring functions should also help improve audit committee effectiveness. Consistent with these arguments, studies find that audit committee members’ governance expertise is negatively associated with auditor dismissal after issuing first-time going concern opinions (Carcello & Neal, 2003) and abnormal accruals (Bédard et al., 2004).

Firm-specific knowledge contributes to audit committee members’ understanding of the company’s operations and thus their effective monitoring of managers’ financial reporting choices (Vafeas, 2005). Because committee members likely gain firm-specific knowledge through experience serving the firm, I expect that audit committee effectiveness increases with their board tenure. Consistent with this, Dhaliwal, Naiker, and Navissi (2010) find that the average board tenure of audit committee members is positively associated with their proxy for accruals quality.

Following the methodology in Gompers et al. (2003), I create dichotomous measures of each of the six above-discussed audit committee characteristics based on the sample median (except for the characteristic of whether the committee chair being a management expert, which is already a dichotomous measure). Each sample observation takes the value 1 if the observation is above the sample median and 0 otherwise. I construct AC_score by summing the values of the six dichotomous measures.¹⁶ Although the construction of this index may seem ad hoc and may not accurately reflect the relative importance of and the complementary/substitutive relations among different audit committee characteristics, it is a simple first step attempting to measure overall audit committee effectiveness. I also validate AC_score as an empirical proxy for audit committee effectiveness by showing in an untabulated analysis that companies with AC_score above sample median tend to have lower incidences of financial restatements and internal control material weaknesses than companies with AC_score below sample median. Furthermore, I find that AC_score is negatively associated with earnings management in tax expense as a main effect (please see Note 20 for details). In “Supplemental and Sensitivity Analyses” section, I show that AC_score is positively correlated with my proxy for overall corporate governance. Collectively, these findings suggest that AC_score is a reasonable empirical proxy for audit committee effectiveness. My hypothesis predicts a positive coefficient on Miss × lnTaxfee × AC_score.

Sample Description and Empirical Results

The SEC (2003) requires public companies to disclose auditor fees in four categories: audit, audit-related, tax-related, and all other fees.¹⁷ I hand collect these fee data and audit committee information from companies’ proxy statements. I obtain analyst forecasts and actual EPS figures from the I/B/E/S Unadjusted Detail datasets. All other accounting variables are from Compustat.

Due to intensive data collection requirements, I constrain the sample to S&P 1500 companies at the end of 2003, the first year after SOX became effective, and collect data for fiscal year 2003. Following Engel, Hayes, and Wang (2010), I exclude 336 financial institutions and utility firms because they tend to have different corporate governance structures than companies in non-regulated industries. I also exclude 15 companies that changed auditors during the year and 118 companies that have missing auditor fee or audit committee information from their proxy statements. Following Dhaliwal et al. (2004), I further exclude 168 companies with negative pre-tax income or negative fourth-quarter ETRs. Finally, I exclude 64 companies that do not purchase any tax services from their auditors. My final sample consists of 799 companies. The sample selection procedure is presented in Table 1.

Table 1.

Sample Selection Procedure.

S&P 1500 companies as of December 31, 2003	1,500
Less
Financial institutions and utilities	(336)
Auditor switches	(15)
Missing auditor fee or audit committee information from the proxy statements	(118)
Negative pre-tax income or negative fourth-quarter ETR	(168)
No purchase of tax services from auditors	(64)
Final Sample	799

Note. S&P = Standard & Poor’s; ETR = effective tax rate.

Table 2 reports descriptive statistics. Largely consistent with existing literature (Hoitash et al., 2009; Vafeas, 2005; Zhang, Zhou, & Zhou, 2007), on average, an audit committee has 3.727 members and holds 7.65 meetings annually. A total of 48.7% of the committee chairs have management experience as CEOs or presidents, and 74.9% of committee members are financial experts. Committee members on average hold 1.413 other board positions and their average board tenure is 8.101 years. The mean of Miss is 0.388, suggesting that 38.8% of the sample would have missed consensus analyst forecasts absent fourth-quarter ETR adjustments. The means of Taxfee and Totalfee are 0.897 million and 3.391 million, respectively. The distributions of both variables are skewed (skewness = 5.979 and 7.057, respectively). Thus, I use natural logarithm of both variables in the regression analyses.

Table 2.

Descriptive Statistics.

	M	SD	Q1	Median	Q3
ACsize	3.727	0.969	3	4	4
ACmtg	7.65	2.861	5	7	9
ACchairmgt	0.487	0.5	0	0	1
ACfin	0.749	0.239	0.667	0.75	1
ACgov	1.413	0.902	0.667	1.333	2
ACtenure	8.101	4.118	5	7.333	10.333
AC_score	2.577	1.314	2	3	3
ETRQ4	0.337	0.092	0.302	0.351	0.38
ETRQ3	0.337	0.091	0.31	0.355	0.38
ETRQ4_ETRQ3	−0.001	0.071	−0.009	0	0.001
Induced_chg_ETR	0	0.017	−0.001	0	0.001
Tax_owed	0.134	0.246	0	0.057	0.182
Miss	0.388	0.488	0	0	1
Taxfee (in millions)	0.897	2.062	0.112	0.297	0.796
lnTaxfee	−1.265	1.592	−2.191	−1.213	−0.228
Totalfee (in millions)	3.391	7.047	0.713	1.379	3.164
lnTotalfee	0.464	1.113	−0.338	0.321	1.152

Note. See the appendix for variable definitions.

Table 3 provides the correlation matrix. AC_score is positively correlated with each of its six underlying attributes by construction. Miss is negatively correlated with ETRQ4_ETRQ3, suggesting that companies manage tax expense to avoid missing analyst forecasts, consistent with the finding in Dhaliwal et al. (2004). I note some high pairwise correlations between the independent variables. However, multicollinearity should not be a serious concern in my later regression analyses because the highest variance inflation factors (VIFs) in all regressions are less than 10.¹⁸

Table 3.

Correlation Matrix.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
AC_score (1)	1	.4250*	.3907*	.5184*	.4117*	.4435*	.2412*	.0218	.0105	.0742*	−.0996*	−.0373	.2339*	.3111*
ACsize (2)	.4548*	1	.1015*	.1832*	−.0093	.1735*	.0258	−.0483	−.0279	.0897*	−.0854*	−.007	.2380*	.3720*
ACmtg (3)	.3677*	.1057*	1	.0245	.0806*	.0873*	−.0585*	−.0093	−.0057	.0810*	.0028	.0137	.1176*	.2141*
ACchairmgt (4)	.5161*	.1789*	.0178	1	.1994*	.1632*	−.0199	.0229	.0103	.0088	−.0466	−.0459	.1549*	.1628*
ACfin (5)	.3995*	.0176	.0861*	.2151*	1	.0615*	−.1177*	−.0045	−.0195	−.0205	−.004	−.0074	−.0387	−.0269
ACgov (6)	.4004*	.1333*	.0624*	.1594*	.0768*	1	−.0843*	.0451	.0074	.0717*	−.0900*	.02	.1964*	.2948*
ACtenure (7)	.1999*	−.0146	−.053	−.027	−.1501*	−.0704*	1	.0346	.0589*	.0042	−.0594*	−.0796*	.0234	.0034
ETRQ4_ETRQ3 (8)	−.0299	−.0845*	−.0518	−.0458	.0107	.0046	.0368	1	.0686*	−.0292	−.0787*	−.1623*	−.0289	−.0614*
Induced_chg_ETR (9)	−.0174	−.0616*	.0084	−.0236	.0264	.006	.0351	.1984*	1	.0019	−.4132*	.01	.1198*	.0883*
Tax_owed (10)	.0411	.0409	.0621*	−.0413	−.0018	.0598*	.0256	.0751*	−.0923*	1	−.2152*	.0717*	.2167*	.2775*
ETRQ3 (11)	−.0479	−.0089	.0131	−.003	−.0396	−.0612*	−.0269	−.4497*	−.3850*	−.0920*	1	−.0335	−.2879*	−.2675*
Miss (12)	−.0447	−.017	.0006	−.0459	−.0048	.0269	−.0896*	−.0994*	−.0921*	.1296*	.0314	1	.0566	.1188*
lnTaxfee (13)	.2441*	.2304*	.1050*	.1716*	−.0388	.1688*	.0056	.0242	.0359	.1338*	−.2100*	.0525	1	.7811*
lnTotalfee (14)	.3256*	.3709*	.2069*	.1757*	−.018	.2604*	−.0149	−.0408	−.0051	.1444*	−.1584*	.1022*	.7772*	1

Note. n = 799. Pearson (Spearman) correlations in the lower (upper) diagonal. See the appendix for variable definitions.

Significant at the 10% level.

Table 4 presents the OLS multiple regression results. In column 1, I first corroborate the finding in Dhaliwal et al. (2004) and report a negative coefficient of −0.013 on Miss (t = −2.67). This means that firms lower their fourth-quarter ETRs from third-quarter ETRs by −0.013 more when they would have missed analyst forecasts absent the adjustments than when they would have met/beaten analyst forecasts absent the adjustments. The magnitude is economically significant because the mean of ETRQ4_ETRQ3 is merely −0.001. This evidence is consistent with companies managing tax expense to avoid missing analyst forecasts.

Table 4.

OLS Regression Analyses.

		(1)	(2)	(3)
	Prediction	ETRQ4_ETRQ3	ETRQ4_ETRQ3	ETRQ4_ETRQ3
Intercept	?	0.117***	0.129***	0.142***
		(4.67)	(4.81)	(4.56)
Induced_chg_ETR	+	0.123	0.098	0.096
		(0.27)	(0.21)	(0.21)
Tax_owed	+	0.014	0.018	0.019
		(0.80)	(1.00)	(1.05)
ETRQ3	−	−0.339***	−0.354***	−0.356***
		(−4.98)	(−5.06)	(−5.11)
Miss	−	−0.013***	−0.017*	−0.040***
		(−2.67)	(−1.77)	(−2.59)
lnTaxfee	?		0.003	0.007**
			(1.08)	(2.03)
Miss × lnTaxfee	?		−0.003	−0.014**
			(−0.77)	(−2.31)
lnTotalfee	?		−0.010**	−0.008*
			(−2.37)	(−1.90)
Miss × lnTotalfee	?		0.003	−0.000
			(0.47)	(−0.08)
AC_score	?			−0.005*
				(−1.70)
Miss × AC_score	?			0.009**
				(2.12)
lnTaxfee × AC_score	?			−0.002
				(−1.56)
Miss × lnTaxfee × AC_score	+			0.004***
				(2.65)
n		799	799	799
Adjusted R²		.208	.218	.219

Note. t statistics in parentheses based on robust standard errors. See the appendix for variable definitions. OLS = ordinary least squares.

,**,***Significant at the 10%, 5%, and 1% levels, respectively.

Next, I examine whether auditor-provided tax services are associated with earnings management in tax expense in the overall sample. In column 2, I find an insignificant coefficient on Miss × lnTaxfee (t = −0.77), providing no evidence that auditor-provided tax services is related to earnings management in tax expense.

Finally, I run Model 1 to examine whether the association between auditor-provided tax services and earnings management in tax expense varies with audit committee effectiveness. In column 3, I find a positive coefficient of 0.004 on Miss × lnTaxfee × AC_score (t = 2.65).¹⁹ This result is consistent with my hypothesis, suggesting that auditor-provided tax services are less positively associated with earnings management in tax expense (i.e., more auditor-provided tax services are associated with less earnings management) as audit committee effectiveness increases. The evidence indicates that the knowledge spillover effect of auditor-provided tax services is more likely to dominate the independence impairment effect as audit committee effectiveness improves.

Furthermore, in untabulated analyses, I partition the sample into two sub-samples based on the median of AC_score and rerun the second column regressions. I find a negative coefficient of −0.013 on Miss × lnTaxfee (t = 1.96) when AC_score is below median (n = 379), suggesting a positive relation between auditor-provided tax services and earnings management in tax expense for firms with less effective audit committees. This evidence indicates that the independence impairment effect of auditor-provided tax services dominates the knowledge spillover effect when audit committee effectiveness is low. In contrast, I find a positive coefficient of 0.007 (t = 1.92) when AC_score is above or equal to median (n = 408), suggesting a negative relation between auditor-provided tax services and earnings management in tax expense for firms with more effective audit committees. This evidence indicates that the knowledge spillover effect of auditor-provided tax services dominates the independence impairment effect when audit committee effectiveness is high. These contrasting results highlight the importance of considering the role of audit committee effectiveness when assessing the association between auditor-provided tax services and earnings management in tax expense. When all firms are pooled together without considering audit committee effectiveness, the positive and negative association sub-samples are mixed together, and this explains my earlier finding regarding the lack of overall association between auditor-provided tax services and earnings management in tax expense. A Chow test suggests a significant increase in the coefficient on Miss × lnTaxfee (t = 2.36) from the sub-sample with AC_score below median to the sub-sample with AC_score above median. This evidence is consistent with the three-way interaction result on the overall sample reported in Table 4. Both results provide supporting evidence for my hypothesis that that auditor-provided tax services are less positively associated with earnings management in tax expense as audit committee effectiveness increases.

Supplemental and Sensitivity Analyses

Six Underlying Attributes of AC_score

I seek to shed light on the relative importance of the six underlying attributes of AC_score in explaining the association between auditor-provided tax services and earnings management in tax expense. First, I reconstruct AC_score after dropping each of the six attributes one at a time and reexamine Model 1 regression results. I find that, in each case, the coefficient on Miss × lnTaxfee × AC_score remains significantly positive. These results suggest that no single attribute is predominant in determining audit committee effectiveness. Second, I reconstruct AC_score using each of the six attributes one at a time and reexamine Model 1 regression results. In each case, the coefficient on Miss × lnTaxratio × AC_score is insignificant. These untabulated results suggest that audit committee effectiveness should be examined in a holistic, rather than piecemeal, manner, consistent with the recommendation of the BRC (1999).

One can argue that audit committee effectiveness may vary non-linearly with some audit committee characteristics such as audit committee size, committee members’ governance expertise, and committee members’ board tenure. When these measures first start to increase, audit committee effectiveness improves due to increased monitoring capacity or improved experience. However, when audit committee size gets too large, audit committee effectiveness will likely decrease due to the free-rider problem. Similarly, when audit committee members hold too many other board positions, they might be too busy to spend sufficient amount of time in each company, thus decreasing audit committee effectiveness. Finally, as board tenure gets too long, audit committee members might develop an overly friendly relationship with the management, thus decreasing audit committee effectiveness. Hence, audit committee effectiveness could first increase in these three characteristics, and then decrease.

The literature is silent on how to treat these non-linearities, which I attempt to incorporate into the research design. In this robustness test, I re-define each of ACsize_D, ACgov_D, and ACtenure_D as 1 if the observation falls between the 25th and the 75th percentile, and 0 otherwise.²⁰ACmtg_D, ACchairmgt, and ACfin_D remain as defined earlier. I then sum the values of the six dichotomous variables to reconstruct AC_score. I rerun Model 1 with the new AC_score measure and continue to find a positive coefficient on Miss × lnTaxfee × AC_score (t = 2.26). Thus, my results are robust to taking into consideration that audit committee effectiveness may vary non-linearly with audit committee size, committee members’ governance expertise, and committee members’ board tenure.

Controlling for Corporate Governance

Although not directly responsible for preapproving auditor-provided tax services, a high-quality board and a strong overall corporate governance system might also help improve financial reporting quality. Thus, it is informative to examine whether audit committees provide effective incremental monitoring of auditor-provided tax services beyond the overall corporate governance system.

Prior literature suggests that corporate governance is stronger when the board is more independent (Beasley, 1996), smaller (Yermack, 1996), meets more frequently (Conger, Finegold, & Lawler, 1998), holds more board positions (Srinivasan, 2005), and has longer tenure (Beasley, 1996); it is weaker when the CEO is also the chair of the board (Imhoff, 2003) and when the G-index is larger (Gompers et al., 2003). Similar to AC_score, I construct a summary index, Gov_score, to proxy for the strength of corporate governance by combining the above seven characteristics. I begin Gov_score with a value of 0, and add 1 if the proportion of independent directors on the board is above the sample median; add 1 if the board size is below the sample median; add 1 if the board meeting frequency is above the sample median; add 1 if the average number of other board memberships directors hold is above the sample median; add 1 if the average board tenure of directors is above the sample median; add 1 if the CEO is not the chair of the board; and add 1 if the G-index is below the sample median. I construct the following OLS multiple regression Model 2 to examine whether my Model 1 results are robust to controlling for corporate governance:

\begin{matrix} {E T R Q 4_E T R Q 3}_{i t} = γ_{0} + γ_{1} {I n d u c e d_c h g_E T R}_{i t} + γ_{2} {T a x_o w e d}_{i t} + γ_{3} {E T R Q 3 + γ}_{4} {M i s s}_{i t} \\ + γ_{5} {l n T a x f e e}_{i t} + γ_{6} {M i s s}_{i t} \times {l n T a x f e e}_{i t} + γ_{7} {l n T o t a l f e e}_{i t} + γ_{8} {M i s s}_{i t} \times {l n T o t a l f e e}_{i t} \\ + γ_{9} {A C_s c o r e}_{i t} + γ_{10} {M i s s}_{i t} \times {A C_s c o r e}_{i t} + γ_{11} {l n T a x f e e}_{i t} \times {A C_s c o r e}_{i t} \\ + γ_{12} {M i s s}_{i t} \times {l n T a x f e e}_{i t} \times {A C_s c o r e}_{i t} \\ + γ_{13} {G o v_s c o r e}_{i t} + γ_{14} {M i s s}_{i t} \times {G o v_s c o r e}_{i t} + γ_{15} {l n T a x f e e}_{i t} \times {G o v_s c o r e}_{i t} \\ + γ_{16} {M i s s}_{i t} \times {l n T a x f e e}_{i t} \times {G o v_s c o r e}_{i t} . \end{matrix}

I obtain board information and G-index from the Investor Responsibility Research Center (IRRC) database. Additional data requirements reduce the sample to 785 firms. Untabulated descriptive statistics show that, on average, 66.2% of directors on the board are independent, there are 9.259 members on the board, and they hold 6.848 meetings annually. Directors on average sit on 0.857 other boards. The average board tenure is 9.740 years. A total of 67.8% of the board chairs are the CEOs of the same companies and the average G-index is 9.461. These statistics are largely consistent with the existing literature (Gompers et al., 2003; Vafeas, 2005; Zhang et al., 2007). As expected, Gov_score is positively correlated with AC_score at the 5% level (Pearson correlation = .081). Table 5 reports Model 2 OLS multiple regression results. The coefficient on Miss × lnTaxfee × AC_score remains significantly positive (t = 2.71), suggesting that my prior results are robust to controlling for corporate governance.

Table 5.

Controlling for Corporate Governance.

	Prediction	ETRQ4_ETRQ3	t
Intercept	?	0.143***	4.33
Induced_chg_ETR	+	0.093	0.20
Tax_owed	+	0.019	1.07
ETRQ3	−	−0.359***	−5.14
Miss	−	−0.031*	−1.70
lnTaxfee	?	0.006	1.39
Miss × lnTaxfee	?	−0.013*	−1.72
lnTotalfee	?	−0.008*	−1.88
Miss × lnTotalfee	?	−0.001	−0.09
AC_score	?	−0.005*	−1.69
Miss × AC_score	?	0.009**	2.15
lnTaxfee × AC_score	?	−0.002	−1.55
Miss × lnTaxfee × AC_score	+	0.004***	2.71
Gov_score	?	0.000	0.04
Miss × Gov_score	?	−0.003	−0.62
lnTaxfee × Gov_score	?	0.000	0.35
Miss × lnTaxfee × Gov_score	?	−0.000	−0.18
n		785
Adjusted R²		.216

Note. t statistics in parentheses based on robust standard errors. See the appendix for variable definitions.

,**,***Significant at the 10%, 5%, and 1% levels, respectively.

Endogeneity of Audit Committee Effectiveness

Audit committee effectiveness may be endogenous. If some firm characteristics explaining audit committee effectiveness are omitted from Model 1, my prior results might be spurious.²¹ To address this potential endogeneity concern, I follow Nikolaev (2010) and employ a two-step procedure. In the first stage, I construct a model to predict AC_score using log of total assets, leverage, return on assets, market-to-book ratio, and Gov_score.²² I obtain the residuals from this first-stage model as Abn_AC_score.

In the second stage, I replace AC_score with Abn_AC_score in the main effect and all the two- and three-way interaction terms in Model 1. Intuitively, one can think of Abn_AC_score as an instrumental variable for AC_score; while it is correlated with AC_score, by construction, it is uncorrelated with other variables that endogenously determine AC_score. Consistent with my prior results, I find a positive coefficient of 0.005 on Miss × lnTaxfee × Abn_AC_score (t = 2.60). This untabulated result suggests that my prior result is robust to accounting for the endogeneity of audit committee effectiveness.²³

Discretionary Accruals

I focus on earnings management in tax expense because this is where the independence impairment and knowledge spillover from auditor-provided tax services can most directly occur. To provide a comparison, I also examine whether auditor-provided tax services are associated with general accruals management, with a particular focus on whether this association varies with audit committee effectiveness. I construct the following OLS regression Model 3.

\begin{matrix} D A C C = δ_{0} + δ_{1} L a g T A C C + δ_{2} L e v + δ_{3} M T B + δ_{4} L o s s + δ_{5} C F O \\ + δ_{6} M i s s_D A C C + δ_{7} ln T a x f e e + δ_{8} M i s s_D A C C \times ln T a x f e e \\ + δ_{9} ln T o t a l f e e + δ_{10} M i s s_D A C C \times ln T o t a l f e e + δ_{11} A C_s c o r e \\ + δ_{12} M i s s_D A C C \times A C_s c o r e + δ_{13} ln T a x f e e \times A C_s c o r e \\ + δ_{14} M i s s_D A C C \times ln T a x f e e \times A C_s c o r e + ε . \end{matrix}

Following Kothari, Leone, and Wasley (2005), I estimate the cross-sectional Jones model discretionary accruals adjusted for firm performance (DACC) as the dependent variable. Similar to the Miss variable used earlier, I define Miss_DACC, which takes the value 1 if earnings absent discretionary accrual adjustments would have missed analyst forecasts, and 0 otherwise. I first regress DACC on Miss_DACC after controlling for lagged total accruals (LagTACC), leverage (Lev), market-to-book ratio (MTB), an earnings loss dummy (Loss), and cash flow from operations (CFO). My choices of control variables follow Frankel, Johnson, and Nelson (2002). Additional data requirements reduce the sample size to 786 companies. I report results of this base model in the first column of Table 6. I find a positive coefficient on Miss_DACC (t = 13.72), suggesting that companies manage discretionary accruals upwards to avoid missing analyst forecasts.

Table 6.

Discretionary Accruals.

	Prediction	ETRQ4_ETRQ3	ETRQ4_ETRQ3	ETRQ4_ETRQ3
Intercept	?	−0.074***	−0.095***	−0.126***
		(−3.78)	(−3.50)	(−3.35)
LagTACC	+	−0.000	−0.000	−0.000
		(−0.90)	(−1.31)	(−1.24)
Lev	+	0.026	0.031	0.029
		(0.94)	(1.19)	(1.09)
MTB	+	0.000	0.000	0.000
		(1.03)	(1.36)	(1.42)
Loss	−	0.008	0.009	0.011
		(0.49)	(0.50)	(0.63)
CFO	−	−0.182**	−0.198**	−0.207**
		(−2.20)	(−2.42)	(−2.55)
Miss_DACC	+	0.139***	0.152***	0.184***
		(13.72)	(7.43)	(5.24)
lnTaxfee	?		−0.015**	−0.021
			(−2.04)	(−1.64)
Miss_DACC × lnTaxfee	?		0.011	0.010
			(1.32)	(0.75)
lnTotalfee	?		0.006	0.001
			(0.52)	(0.12)
Miss_DACC × lnTotalfee	?		−0.001	0.003
			(−0.11)	(0.23)
AC_score	?			0.013
				(1.49)
Miss_DACC × AC_score	?			−0.013*
				(−1.65)
lnTaxfee × AC_score	?			0.002
				(0.57)
Miss_DACC × lnTaxfee × AC_score	?			0.000
				(0.10)
n		786	786	786
Adjusted R²		.247	.252	.254

Note. t statistics in parentheses based on robust standard errors. See the appendix for variable definitions.

,**,***Significant at the 10%, 5%, and 1% levels, respectively.

I then expand the above model by including lnTaxfee, Miss_DACC × lnTaxfee, lnTotalfee, and Miss_DACC × lnTotalfee, and find an insignificant coefficient on Miss_DACC × lnTaxfee (t = 1.32). Finally, I further expand the model to Model 3. I find an insignificant coefficient on Miss_DACC × lnTaxfee × AC_score (0.001, t = 0.10). My findings suggest that, in the context of auditor-provided tax services, focusing on earnings management in tax expense helps improve the power of the test to detect earnings management over focusing on aggregate accruals. This is consistent with the suggestion in Healy and Wahlen (1999) to focus on specific accruals when studying earnings management.

Economic Benefits of Auditor-Provided Tax Services

Cook, Huston, and Omer (2008) argue that, in addition to earnings management in tax expense, ETRQ4_ETRQ3 might also reflect the economic benefit of effective tax planning occurring in the fourth quarter. They use the amount of auditor-provided tax services as a proxy for companies’ investment in their fourth-quarter tax planning. Gaynor et al. (2006) suggest that audit committees consider economic benefits of non-audit services when performing a cost–benefit analysis prior to the approval of these services. As more effective audit committees arguably have better skills in identifying cost–beneficial tax planning, the economic benefits explanation of auditor-provided tax services would predict that the coefficient on Miss × lnTaxfee × AC_score in Model 1 is negative. Contrary to this prediction, I find the coefficient to be positive. Thus, my findings are inconsistent with the economic benefit interpretation of ETRQ4_ETRQ3.

Generalizability

I evaluate whether my results can be generalized to other samples and other periods. Due to intensive data collection requirements, I randomly select a recent sample period—fiscal year 2009 and collect audit committee information for S&P 500 companies. I follow the same sample selection procedures as detailed in Table 1 and my final sample consists of 250 companies. I construct AC_score based on the medians of the six underlying audit committee attributes in this sample.

Untabulated descriptive statistics show that, for this sample, the mean audit committee size is 4.208 and meeting frequency is 8.612. A total of 58.8% of the committee chairs have management experience, and 87% of committee members are financial experts. Committee members on average hold 1.209 other board positions and their average board tenure is 7.813 years. A total of 46% of the sample would have missed consensus analyst forecasts absent fourth-quarter ETR adjustments. The means of Taxfee and Totalfee are 1.129 million and 9.105 million, respectively. Although this sample appears to be systematically different from the sample of my main analyses, Table 7 shows that my main results still hold in this sample. I continue to find a positive coefficient on Miss × lnTaxfee × AC_score (0.006, t = 2.05). This result adds confidence to the external validity of my primary findings.

Table 7.

S&P 500 Companies for Fiscal Year 2009.

	Prediction	ETRQ4_ETRQ3	ETRQ4_ETRQ3	ETRQ4_ETRQ3
Intercept	?	0.016	0.005	0.008
		(1.18)	(0.31)	(0.45)
Induced_chg_ETR	+	2.127***	2.071***	2.054***
		(3.78)	(3.67)	(3.78)
Tax_owed	+	−0.040	−0.034	−0.032
		(−1.01)	(−0.84)	(−0.80)
ETRQ3	−	−0.057	−0.067*	−0.074*
		(−1.46)	(−1.66)	(−1.82)
Miss	−	−0.011**	−0.009	−0.013*
		(−2.01)	(−1.40)	(−1.70)
lnTaxfee	?		−0.006**	−0.000
			(−2.39)	(−0.09)
Miss × lnTaxfee	?		0.005	−0.011*
			(0.87)	(−1.71)
lnTotalfee	?		0.004	0.004
			(0.81)	(0.81)
Miss × lnTotalfee	?		−0.008	−0.006
			(−0.84)	(−0.65)
AC_score	?			−0.000
				(−0.09)
Miss × AC_score	?			0.004*
				(1.82)
lnTaxfee × AC_score	?			−0.002
				(−1.09)
Miss × lnTaxfee × AC_score	+			0.006**
				(2.05)
n		250	250	250
Adjusted R²		.350	.356	.366

Note. t statistics in parentheses based on robust standard errors. See the appendix for variable definitions. S&P = Standard & Poor’s.

,**,***Significant at the 10%, 5%, and 1% levels, respectively.

Other Sensitivity Analyses

In other sensitivity tests, I use the ratio of tax fees to total fees paid to the auditor as the primary measure of auditor-provided tax services in place of lnTaxfee in the stand-alone term and all interaction terms; include the 64 companies that do not purchase any tax services from their auditors; define an indicator variable Zerotax = 1 for these 64 companies and include Zerotax and Miss × Zerotax as additional variables²⁴; include the natural logarithm of total assets, its interaction with Miss, and a three-way interaction with Miss and AC_score to control for firm size effect; exclude the 19 companies that are not audited by Big 4 auditors; and exclude outliers with externally studentized residuals larger than 2 or less than −2; my primary results are robust to all of these alternative specifications.

Conclusion

SOX and the SEC prohibit auditors from providing a variety of non-audit services to their audit clients. Tax services remain the largest type of non-audit services permitted (Beasley et al., 2009), provided that these services are preapproved by clients’ audit committees. However, the PCAOB remains concerned that auditor independence may be impaired, and hence increases the audit committees’ involvement in the preapproval process. This article empirically examines whether the relation between auditor-provided tax services and earnings management (particularly with regard to the amount recorded as tax expense) varies with audit committee effectiveness.

I develop a composite proxy for audit committee effectiveness by combining six audit committee characteristics: committee size, meeting frequency, whether the committee chair has management expertise, the proportion of financial experts on the committee, the average number of other board positions the committee members hold, and the average board tenure of committee members. I find that auditor-provided tax services are less positively associated with earnings management in tax expense as my proxy for audit committee effectiveness increases. This suggests that the knowledge spillover effect of auditor-provided tax services is more likely to dominate the independence impairment effect as audit committee effectiveness improves. Moreover, I find a positive association between auditor-provided tax services and earnings management in tax expense when audit committee effectiveness is below the sample median, and a negative association when audit committee effectiveness is above the sample median. These contrasting results further highlight the importance of considering the role of audit committee effectiveness when assessing the association between auditor-provided tax services and earnings management in tax expense.

I further show that my proxy for audit committee effectiveness is not predominated by any one of the six underlying attributes. My results are robust to controlling for overall corporate governance and addressing the endogeneity of audit committee effectiveness. I also rule out an alternative explanation that my proxy for earnings management in tax expense could partly reflect the economic benefits of auditor-provided tax services. In addition, I find that, in the context of auditor-provided tax services, focusing on earnings management in tax expense improves the power of the test to detect earnings management over focusing on aggregate accruals.

I acknowledge some limitations of my study and suggest avenues for future research. First, due to intensive data collection requirements, I constrain the sample to S&P 1500 companies and only investigate fiscal year 2003 in the main analyses. Although I find in the additional analyses section that my results can be extended to a sample of S&P 500 companies in fiscal year 2009, whether my results can be further generalized to other samples and other periods is an empirical question for future research. Second, I follow the methodology in Gompers et al. (2003) to construct a summary index for audit committee effectiveness by summing the values of the dichotomous measures of six audit committee characteristics. Although I strive to validate this empirical proxy for audit committee effectiveness via different tests, I acknowledge that the construction of this index may seem ad hoc. Future research could explore other empirical proxies to improve the construct validity of audit committee effectiveness.

Footnotes

Appendix

Variable Definitions.

ETRQ4_ETRQ3	=	the fourth-quarter effective tax rate (ETR; ETRQ4) minus the third-quarter ETR (ETRQ3), where ETR is defined as accumulated tax expense (Compustat quarterly #6) divided by accumulated pre-tax income (quarterly #23).
ETRQ4	=	the fourth-quarter ETR, calculated as accumulated tax expense divided by accumulated pre-tax income up to the fourth quarter.
ETRQ3	=	the third-quarter estimate of annual ETR, calculated as accumulated tax expense divided by accumulated pre-tax income up to the third quarter.
Miss	=	1 if earnings absent fourth-quarter ETR adjustment [pre-tax income (annual #170) × (1 −ETRQ3)/common shares outstanding (annual #25)], rounded to the nearest cent, is lower than the most recent median consensus analyst forecast before earnings announcement, and 0 otherwise.
Induced_chg_ETR	=	induced tax change divided by pre-tax income, where induced tax change equals (35% −ETRQ3) × unexpected pre-tax income. The unexpected pre-tax income is estimated as (I/B/E/S actual EPS − I/B/E/S most recent median consensus forecast) × common shares outstanding (annual #25)/(1 − 35%).
Tax_owed	=	taxes payable (annual #71) net of refunds (annual #161), scaled by pre-tax income.
Taxfee	=	fees paid to the auditor for tax services, in millions.
lnTaxfee	=	natural logarithm of Taxfee.
Totalfee	=	total fees paid to the auditor, in millions, where total fees include audit fees, audit-related fees, tax fees, and other fees.
lnTotalfee	=	natural logarithm of Totalfee.
ACsize	=	number of members on the audit committee.
ACmtg	=	number of meetings held annually by the audit committee.
ACchairmgt	=	1 if the audit committee chair is an individual with experience as a CEO or president of a company, and 0 otherwise.
ACfin	=	proportion of financial experts on the audit committee. A financial expert is an individual with experience as a public accountant, auditor, principal or chief financial officer, controller, principal or chief accounting officer, CEO or president of a for-profit corporation, investment banker, venture capitalist or financial analyst.
ACgov	=	average number of other board memberships audit committee members hold.
ACtenure	=	average board tenure of audit committee members.
AC_score	=	summary index of audit committee effectiveness that is equal to the sum of the values of the following six dichotomous variables: ACsize_D = 1 if ACsize is above the sample median, and 0 otherwise. ACmtg_D = 1 if ACmtg is above the sample median, and 0 otherwise. ACchairmgt = 1 if the audit committee chair is an individual with experience as a CEO or president of a company, and 0 otherwise. ACfin_D = 1 if ACfin is above the sample median, and 0 otherwise. ACgov_D = 1 if ACgov is above the sample median, and 0 otherwise. ACtenure_D = 1 if ACtenure is above the sample median, and 0 otherwise.
BDind	=	proportion of independent directors on the board. An independent director is identified by IRRC as a director who has no significant connections with the company, that is, a director who is not a current or former employee of the company, who (and whose employer) does not provide professional services to the company and is not a major customer, who is not a relative of an executive of the company, and who is not affiliated with the company in other manners.
BDsize	=	number of directors on the board.
BDmtg	=	number of board meetings held annually.
BDgov	=	average number of other board memberships directors hold.
BDtenure	=	average board tenure of directors.
CEO_chair	=	1 if the CEO is also the chair of the board, and 0 otherwise.
G-index	=	governance index reported by IRRC, which is constructed following the methodology in Gompers, Ishii, and Metrick (2003).
Gov_score	=	summary index of the strength of corporate governance that is equal to the sum of the values of the following seven dichotomous variables: BDind_D = 1 if BDind is above the sample median, and 0 otherwise. BDsize_D = 1 if BDsize is below the sample median, and 0 otherwise. BDmtg_D = 1 if BDmtg is above the sample median, and 0 otherwise. BDgov_D = 1 if BDgov is above the sample median, and 0 otherwise. BDtenure_D = 1 if BDtenure is above the sample median, and 0 otherwise. NonCEO_chair = 1 −CEO_chair. G-index_D = 1 if the G-index is below the sample median, and 0 otherwise.
DACC	=	cross-sectional Jones model discretionary accruals adjusted for firm performance following Kothari, Leone, and Wasley (2005).
LagTACC	=	prior year’s total accruals, calculated as income before extraordinary items (annual #123) minus the difference between cash flow from operations (annual #308) and extraordinary items and discontinued operations (annual #124), deflated by prior year’s total assets (annual #6).
Lev	=	the sum of current debt (annual #34) and long-term debt (annual #9) deflated by total assets (annual #6).
MTB	=	market-to-book ratio, calculated as common stock price (annual #199) times common shares outstanding (annual #25) divided by book value of equity (annual #60).
Loss	=	1 if net income (annual #172) is negative, and 0 otherwise.
CFO	=	cash flow from operations (annual #308) deflated by total assets (annual #6).
Miss_DACC	=	1 if earnings absent discretionary accrual adjustments would have missed the most recent median consensus analyst forecast, and 0 otherwise.

Acknowledgements

I am grateful for the guidance of my committee members, Shanta Hegde, John Phillips, and Mike Willenborg (chair). I also thank an anonymous reviewer, Larry Abbott, JK Aier, Peter Clarkson, Patricia Dechow, Amy Dunbar, Angela Han, Alfred Liu, Lil Mills, Tom Omer, Dave Papandria, Ray Pfeiffer, Richard Sansing, Bharat Sarath (editor in chief), Partha Sengupta, Dino Silveri, Stan Veliotis, Gnanakumar Visvanathan, Dave Weber, Scott Whisenant, and workshop participants at the AAA Auditing Section Midyear Conference, AAA Annual Meeting, Conference on Financial Economics and Accounting, Binghamton University, Fordham University, George Mason University, University of Houston, Lehigh University, University of Massachusetts at Amherst, University of Massachusetts at Boston, Miami University, University of New Hampshire, Santa Clara University, University of Vermont, and Virginia Commonwealth University for helpful comments and suggestions.

Author’ Note

This article is based on my dissertation completed at the University of Connecticut.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abbott

L. J.

Parker

Peters

G. F.

(2004). Audit committee characteristics and restatements. Auditing: A Journal of Practice & Theory, 23(1), 69-87.

Abbott

L. J.

Parker

Peters

G. F.

Raghunandan

(2003). The association between audit committee characteristics and audit fees. Auditing: A Journal of Practice & Theory, 22(2), 17-32.

Agrawal

Chadha

(2005). Corporate governance and accounting scandals. Journal of Law and Economics, 48, 371-406.

Anderson

R. C.

Mansi

S. A.

Reeb

D. M.

(2004). Board characteristics, accounting report integrity, and the cost of debt. Journal of Accounting & Economics, 37, 315-342.

Ashbaugh

LaFond

Mayhew

B. W.

(2003). Do non-audit services compromise auditor independence? Further evidence. The Accounting Review, 78, 611-639.

Barrett

M. J.

(2004). “Tax services” as a Trojan horse in the auditor independence provisions of Sarbanes-Oxley. Michigan State Law Review, 2004, 463-504.

Beasley

M. S.

(1996). An empirical analysis of the relation between the board of director composition and financial statement fraud. The Accounting Review, 71, 443-465.

Beasley

M. S.

Carcello

J. V.

Hermanson

D. R.

Neal

T. L.

(2009). The audit committee oversight process. Contemporary Accounting Research, 26, 65-122.

Bédard

Chtourou

S. M.

Courteau

(2004). The effect of audit committee expertise, independence, and activity on aggressive earnings management. Auditing: A Journal of Practice & Theory, 23(2), 13-35.

10.

Blue Ribbon Committee. (1999). Report and recommendations of the blue ribbon committee on improving the effectiveness of corporate audit committees: New York Stock Exchange and National Association of Securities Dealers, New York, NY.

11.

Carcello

J. V.

Neal

T. L.

(2003). Audit committee characteristics and auditor dismissals following “new” going-concern reports. The Accounting Review, 78, 95-117.

12.

Carcello

J. V.

Neal

T. L.

Palmrose

Z.-V.

Scholz

(2011). CEO involvement in selecting board members, audit committee effectiveness, and restatements. Contemporary Accounting Research, 28, 396-430.

13.

Chung

Kallapur

(2003). Client importance, nonaudit services, and abnormal accruals. The Accounting Review, 78, 931-955.

14.

Conger

J. A.

Finegold

Lawler

E. E.

III. (1998). Appraising boardroom performance. Harvard Business Review, 76(1), 136-148.

15.

Cook

K. A.

Huston

R. G.

Omer

T. C.

(2008). Earnings management through effective tax rates: The effects of tax planning investment and the Sarbanes-Oxley Act of 2002. Contemporary Accounting Research, 25, 447-471.

16.

DeAngelo

L. E.

(1981). Auditor independence, “low balling,” and disclosure regulation. Journal of Accounting & Economics, 3, 113-127.

17.

DeFond

M. L.

Hann

R. N.

(2005). Does the market value financial expertise on audit committees of boards of directors?Journal of Accounting Research, 43, 153-193.

18.

DeFond

M. L.

Raghunandan

Subramanyam

K. R.

(2002). Do non-audit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of Accounting Research, 40, 1247-1274.

19.

DeZoort

F. T.

Hermanson

D. R.

Archambeault

D. S.

Reed

S. A.

(2002). Audit committee effectiveness: A synthesis of the empirical audit committee literature. Journal of Accounting Literature, 21, 38-75.

20.

Dhaliwal

D. S.

Gleason

C. A.

Mills

L. F.

(2004). Last-chance earnings management: Using the tax expense to meet analysts’ forecasts. Contemporary Accounting Research, 21, 431-459.

21.

Dhaliwal

D. S.

Naiker

Navissi

(2010). The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees. Contemporary Accounting Research, 27, 787-827.

22.

Engel

Hayes

R. M.

Wang

(2010). Audit committee compensation and the demand for monitoring of the financial reporting process. Journal of Accounting and Economics, 49, 136-154.

23.

Farber

D. B.

(2005). Restoring trust after fraud: Does corporate governance matter?The Accounting Review, 80, 539-561.

24.

Frankel

R. M.

Johnson

M. F.

Nelson

K. K.

(2002). The relation between auditors’ fees for non-audit services and earnings management. The Accounting Review, 77(Suppl.), 71-105.

25.

Gaynor

L. M.

McDaniel

L. S.

Neal

T. L.

(2006). The effects of joint provision and disclosure of non-audit services on audit committee members’ decisions and investors’ preferences. The Accounting Review, 81, 873-896.

26.

Gendron

Bédard

(2006). On the constitution of audit committee effectiveness. Accounting, Organizations and Society, 31, 211-239.

27.

Gleason

C. A.

Mills

L. F.

(2011). Do auditor-provided tax services improve the estimate of tax reserves? Contemporary Accounting Research, 28, 1484-1509.

28.

Gompers

Ishii

Metrick

(2003). Corporate governance and equity prices. The Quarterly Journal of Economics, 118, 107-155.

29.

Healy

P. M.

Wahlen

J. M.

(1999). A review of the earnings management literature and its implications for standard setting. Accounting Horizons, 13, 365-383.

30.

Hoitash

Bedard

J. C.

(2009). Corporate governance and internal control over financial reporting: A comparison of regulatory regimes. The Accounting Review, 84, 839-867.

31.

Imhoff

E. A. Jr.

(2003). Accounting quality, auditing, and corporate governance. Accounting Horizons, 17(Suppl.), 117-128.

32.

Kalbers

L. P.

Fogarty

T. J.

(1993). Audit committee effectiveness: An empirical investigation of the contribution of power. Auditing: A Journal of Practice & Theory, 12(1), 24-48.

33.

Kennedy

(1998). A guide to econometrics (4th ed.). Cambridge, MA: The MIT Press.

34.

Kinney

W. R. J.

Libby

(2002). Discussion of the relation between auditors’ fees for nonaudit services and earnings management. The Accounting Review, 77(Suppl.), 107-114.

35.

Kinney

W. R.

Palmrose

Z.-V.

Scholz

(2004). Auditor independence, non-audit services, and restatements: Was the U.S. government right?Journal of Accounting Research, 42, 561-588.

36.

Koh

Rajgopal

Srinivasan

(2013). Non-audit services and financial reporting quality: Evidence from 1978-1980. Review of Accounting Studies, 18, 1-33.

37.

Kolodny

H. F.

Kiggundu

M. N.

(1980). Towards the development of a sociotechnical systems model in woodlands mechanical harvesting. Human Relations, 33, 623-645.

38.

Kothari

S. P.

Leone

A. J.

Wasley

C. E.

(2005). Performance matched discretionary accrual measures. Journal of Accounting & Economics, 39, 163-197.

39.

Krishnan

Visvanathan

(2008). Does the SOX definition of an accounting expert matter? The association between audit committee directors’ accounting expertise and accounting conservatism. Contemporary Accounting Research, 25, 827-857.

40.

Krishnan

Visvanathan

(2009). Do auditors price audit committee’s expertise? The case of accounting vs. non-accounting financial experts. Journal of Accounting, Auditing & Finance, 24, 115-144.

41.

Krishnan

(2005). Audit committee effectiveness and internal control: An empirical analysis. The Accounting Review, 80, 649-675.

42.

Larcker

D. F.

Richardson

S. A.

(2004). Fees paid to audit firms, accrual choices, and corporate governance. Journal of Accounting Research, 42, 625-658.

43.

Maydew

E. L.

Shackelford

D. A.

(2007). The changing role of auditors in corporate tax planning. In Auerbach

A. J.

Hines Jr

J. R.

Jr. Slemrod

J. B.

(Eds.), Taxing corporate income in the 21st century (pp. 307-337). Cambridge, UK: Cambridge University Press.

44.

Nam

Ronen

(2012). The impact of nonaudit services on capital markets. Journal of Accounting, Auditing & Finance, 27(1), 32-60.

45.

Nikolaev

(2010). Debt covenants and accounting conservatism. Journal of Accounting Research, 48, 137-175.

46.

Omer

T. C.

Bedard

J. C.

Falsetta

(2006). Auditor-provided tax services: The effects of a changing regulatory environment. The Accounting Review, 81, 1095-1117.

47.

Public Company Accounting Oversight Board. (2004, July14). Auditor independence and tax services roundtable (Briefing paper). Washington, DC: Author.

48.

Public Company Accounting Oversight Board. (2005, July26). Ethics and independence rules concerning independence, tax services, and contingent fees (PCAOB Release No. 2005–014). Washington, DC: Author.

49.

Public Company Accounting Oversight Board. (2008, April22). Ethics and independence rule 3526, communication with audit committees concerning independence (PCAOB Release No. 2008–003). Washington, DC: Author.

50.

Public Company Accounting Oversight Board. (2010, March29). Proposed auditing standard related to communication with audit committees (PCAOB Release No. 2010–001). Washington, DC: Author.

51.

Public Company Accounting Oversight Board. (2011, December20). Proposed auditing standard related to communication with audit committees (PCAOB Release No. 2011–008). Washington, DC: Author.

52.

Reynolds

J. K.

Francis

J. R.

(2001). Does size matter? The influence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics, 30, 375-400.

53.

Securities and Exchange Commission (SEC). (2003). Final rule: Strengthening the Commission’s requirements regarding auditor independence (SEC Release No. 33–8183, Washington, DC).

54.

Seetharaman

Sun

Wang

(2011). Tax-related financial statement restatements and auditor-provided tax services. Journal of Accounting, Auditing & Finance, 26, 677-698.

55.

Simunic

D. A.

(1984). Auditing, consulting, and auditor independence. Journal of Accounting Research, 22, 679-702.

56.

Skinner

D. J.

Sloan

R. G.

(2002). Earnings surprises, growth expectations, and stock returns or don’t let an earnings torpedo sink your portfolio. Review of Accounting Studies, 7, 289-312.

57.

Srinivasan

(2005). Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit committee members. Journal of Accounting Research, 43, 291-334.

58.

StataCorp. (2005). Stata Statistical Software: Release 9. College Station, TX: Author.

59.

Vafeas

(2005). Audit committees, boards, and the quality of reported earnings. Contemporary Accounting Research, 22, 1093-1122.

60.

Xie

Davidson

W. N. III

DaDalt

P. J.

(2003). Earnings management and corporate governance: The role of the board and the audit committee. Journal of Corporate Finance, 9, 295-316.

61.

Yermack

(1996). Higher market valuation of companies with a small board of directors. Journal of Financial Economics, 40, 185-211.

62.

Zhang

Zhou

(2007). Auditor quality, audit committee quality, and internal control weakness. Journal of Accounting and Public Policy, 26, 300-327.