Auditor-provided tax services and long-term tax ...

The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/1475-7702.htm

Auditor-provided tax services and long-term tax avoidance

Auditorprovided tax services

Brian Hogan Department of Accountancy, University of Pittsburgh, Pittsburgh, Pennsylvania, USA, and

Downloaded by Bentley University, Doctor Tracy Noga At 11:23 04 August 2015 (PT)

Tracy Noga Department of Accountancy, Bentley University, Waltham, Massachusetts, USA

285 Received 15 October 2013 Revised 29 January 2015 Accepted 29 January 2015

Abstract Purpose – The purpose of this paper is to determine the association between auditor-provided tax services (APTS) and long-term corporate tax rates. Design/methodology/approach – The paper uses empirical data and multivariate regression models to explore the relationship between a firm’s use of APTS and their long-term effective tax rate. Findings – An economically and statistically significant long-term negative relationship was found between firm levels of APTS and taxes paid. Further, a portion of this benefit is lost for some firms when returning to their auditor for tax services even after a short break. Originality/value – This paper contributes to the debate regarding the value of APTS by providing evidence of the apparent long-term negative consequences to firms who reduce their reliance on APTS, perhaps even through the engagement of separate accounting firms for their audit and tax functions, although these consequences may be mitigated upon return with a significant increase in APTS. However, this is the first study, to our knowledge, to explore, in a long-term setting, the consequences of a firm’s return to their auditors for a non-audit service previously reduced or terminated. Additionally, further incremental contributions are made to other studies that look at APTS and tax avoidance by studying the long-term relationship which allows firms to consider the cumulative cost/benefit relationship between independence and knowledge spillover. Keywords Tax avoidance, Independence, Auditor-provided tax services, Long-term taxes paid Paper type Research paper

1. Introduction We study the long-term relationship between auditor-provided tax services (APTSs) and cash-effective tax rates (CETR) over the long term (five years). After the passage of The Sarbanes–Oxley Act of 2002 (SOX), auditor-provided services declined in part due to new rules eliminating certain services, as well as requirements involving audit committee pre-approval for permissible services, such as tax engagements (SOX, Section 201). These changes came about due to concerns about the impact of high levels of non-audit fees on auditor independence (Abbott et al., 2003; Davis and Hollie, 2008; Grant et al., 2008). However, proponents of not limiting non-audit services suggest that This study has benefited from helpful comments from Charlie Bame-Aldred, Jean Bedard, Diana Falsetta, Ronen Gal-Or, Udi Hoitash, Ganesh Krishnamoorthy, Colin Reid, Tim Rupert and anonymous reviewers and participants at the 2010 AAA Annual Meeting. All errors or omissions are the authors’ own.

Review of Accounting and Finance Vol. 14 No. 3, 2015 pp. 285-305 © Emerald Group Publishing Limited 1475-7702 DOI 10.1108/RAF-10-2013-0116

RAF 14,3


286

synergies are created by combining the audit and tax service functions within one accounting firm. In other words, a “knowledge spillover” effect may exist by which knowledge gained by performing one service “spills over” to other firm engagements which produce production efficiencies (Gleason and Mills, 2011; Kinney et al., 2004; McGuire et al., 2012). These proponents feel that the benefits of the combined services outweigh the potential risks of combined services. As a result, there is a seeming trade-off between auditor independence and knowledge spillover which must be evaluated by each company making the decision. The extant literature remains mixed as to which position is more beneficial for firms, shareholders and policymakers[1]. A number of studies document the impairment of auditor independence (Choi et al., 2010; Frankel et al., 2002) and others refute independence concerns (Knechel and Sharma, 2012; Chung and Kallapur, 2003; Ashbaugh et al., 2003; DeFond et al., 2002)[2]. As APTSs were declined post-SOX, tax revenues for accounting firms increased (Bédard and Paquette, 2011). Therefore, it appears that companies are indeed shifting tax engagements. However, if tax planning is a long-term investment (Rego and Wilson, 2012), looking at only the short-term relationship may not fully capture the effects of investing in APTS. This study contributes to this research area by examining the long-term[3] benefits that arise when firms utilize APTSs. We review a sample of APTS fees in the years following the passage of SOX and find a statistically negative association between long-term cash taxes paid and fees paid to auditors for tax services. Firms that do contract their auditors for tax services, perhaps due to the benefits of knowledge spillover, appear to reap the benefits of these services in the form of reduced cash taxes paid. We also study the tax impact on firms who return to APTS after reducing or eliminating them. Although we find that some benefits are immediately recaptured, others are not. Additional analysis shows that even one year in which APTS were not present not only removes the tax benefit but rather creates a negative impact on taxes paid. This result remains even when a portion of APTS is resumed. This research adds to the extant literature in several important ways. We contribute to the debate regarding the value of APTS by providing evidence of the apparent long-term negative consequences to firms who reduce their reliance on APTS, perhaps even through the engagement of separate accounting firms for their audit and tax functions, although these consequences may be mitigated upon return with a significant increase in APTS. Additionally, we make further incremental contributions to other studies that look at APTS and tax avoidance by studying the long-term relationship which allows firms to consider the cumulative cost/benefit relationship between independence and knowledge spillover. 2. Literature review and hypothesis development 2.1 Audit and tax fees In 2002, the passage of SOX imposed strict limitations on the types of non-audit services that audit firms could provide to their audit clients. Although tax services were not strictly prohibited, the fees paid annually for non-audit services, and separately for tax services, are again required to be disclosed annually by companies. Due to the new limitations on the provision of non-audit services, along with increases in audit fees for SOX compliance, fees paid by audit clients for audit services have grown substantially, while fees for non-audit services have declined (Markelevich et al., 2005; Omer et al.,


2006). Maydew and Shackelford (2007) find that payments to auditors for tax and audit services declined from a 1:1 ratio in 2001 to approximately 1:4 in 2004. More recently, in 2007, non-audit fees represented approximately 21 per cent of total fees paid by companies to their auditors (Audit Analytics, 2008). In 2005, certain APTS were banned by the Public Company Accounting Oversight Board (PCAOB), and today, there are additional APTS that require approval of the audit committee. Prior research has examined non-audit fee changes and fee disclosures in the period following the passage of SOX in 2002 and the subsequent PCAOB regulations. Krishnan and Yu (2011) find that increases in NAS fees are associated with reduced audit fees. Omer et al. (2006) find, perhaps in anticipation of the new disclosure requirements, that the association between tax fees and higher than expected audit fees weakened in 2002. The authors posit that firms paying high audit fees terminated the provision of tax services by their auditors. Bedard et al. (2010), studying the decisions of firms to voluntarily disclose APTS, detect a negative relationship between companies disclosing these tax fees before they were mandated in 2003 (pre-disclosers) and the ratio of non-audit fees to total fees. Bédard and Paquette (2011) observe that mean tax fees for audit sample firms decreased by 57 per cent from 2004 to 2007, while mean tax revenues for the Big Four accounting firms increased by 29 per cent over the same period. Maydew and Shackelford (2007) find similar results and show that, while APTS declined significantly, post-SOX, total tax fees collected by accounting firms remained steady. The authors attribute this finding to a shift in clients among the providers of tax services away from the audit provider. If tax fees have indeed shifted among service providers, then changes in taxes paid by companies may be related to external tax consulting arrangements, or the internal tax department, as opposed to the auditing firm. These findings suggest perhaps that additional monitoring of firms’ aggressiveness in tax planning is necessary. 2.2 Auditor independence and knowledge spillover The research on auditor independence and knowledge spillover has spanned many years but has increased substantially post-SOX. Presumably, firms would choose to engage their auditors for non-audit services if the expected tax benefits outweighed the expected benefits of independence. The proponents of limiting non-audit fees paid contend that auditor independence, or perceptions of auditor independence, may be impaired if large amounts of non-audit fees are collected by the auditing firm (Davis and Hollie, 2008; Grant et al., 2008). Favere-Marchesi (2006), in a case study administered to audit partners and senior managers, find that auditors who also performed tax services reported a significantly lower assessment of fraud risk than auditors that provided no tax services to the audit client. Elder et al. (2008) examine the association between tax services and the reporting of internal control weaknesses. They find that firms with APTS are less likely to report control weaknesses and conclude that their findings are consistent with APTS impairing auditor independence rather than knowledge spillovers arising from the tax services. Other studies argue that providing dual audit and tax service creates a “knowledge spillover” effect which improves areas of the reporting process and firm performance and does not impair auditor independence (DeFond et al., 2002; Choi et al., 2010). Kinney et al. (2004), in a study of pre-SOX firms, find that firms engaging their auditor for tax

Auditorprovided tax services 287

RAF 14,3


288

services are associated with a lower incidence of earnings restatements. Accounting risk, as measured by either misleading or fraudulent financial reporting, has also been found to be lower for companies which outsource their internal audit work to their external auditor as compared to those which keep the audit work entirely in-house or outsource the work to another accounting firm (Prawitt et al., 2012). Krishnan et al. (2013) find that APTS are associated with less earnings management, indicating improved financial reporting and audit quality. However, some findings are mixed. Cook et al. (2008) find inconsistent results in reporting quality in a post-SOX sample. If the theory of knowledge spillover is valid, synergies between the audit and tax functions may help in uncovering tax savings opportunities. Unnecessary taxes paid by firms represent an allocation of resources away from firm stakeholders; any incremental benefit received from retaining the auditor for tax services should be examined[4]. Recent post-SOX studies (Cook and Omer, 2013; Dhaliwal et al., 2013; McGuire et al., 2012) find evidence of knowledge spillover in the short term as they demonstrate that a reduction in APTS is associated with less tax avoidance. Some studies have explored the link between APTS and tax measures, although most of these studies examined data prior to SOX. Mills et al. (1998) find a negative relation between tax fees paid and tax liabilities. Gleason and Mills (2011) also note that using APTS is associated with lower US tax expenses. However, the authors, in their sample of companies from 2000-2002, do not find a positive relation between APTS and lower US taxes paid. Omer et al. (2006), studying the relation between APTS and subsequent changes in tax rates, found a negative association, although this relation weakened in their final sample year (2002)[5]. 2.3 Long-term tax effects Dyreng et al. (2008), examining a sample of firms from 1995-2004, discuss the capabilities of some corporations to consistently pay low cash taxes. The authors argue that analyzing longer time horizons helps strengthen findings by accounting for significant year-to-year variations that exist within short-term tax measures (Dyreng et al., 2008). Managers, often rated based upon short-term performance measures, are constantly faced with resource allocation decisions. However, the benefits received from tax planning expenditures are uncertain and also may not be realized in the short-term (Rego and Wilson, 2012). A longer-term analysis may provide support for why managers make long-term tax investments, even in the face of short-term resource constraints. Hanlon and Heitzman (2010), in their review of tax research, discuss long-run effective tax rates and cite many benefits of using such a measure. Cook and Omer (2013) and Dhaliwal et al. (2013) find an association between APTS and their short-term measure indicating there is indeed a short-term benefit. However, for a manager to make an informed cost-benefit analysis of choosing APTS, the full (long-term) impact should be considered. 2.4 Hypotheses Cripe and McAllister (2009) observe, in a survey of CFOs, that auditor independence concerns was the primary reason for decisions to separate audit and tax functions. Fees for non-audit services also declined because a number of non-audit services can no longer be provided by the firm’s external auditor under SOX. Although tax services were not included in this list of excluded services, anecdotally, auditors and firms


separated from these services in the interest of independence. Engaging separate accounting firms to provide audit and tax work, however, may affect the quality of services obtained (e.g. lack of knowledge spillover). Prior studies have argued that synergies are created by combining the audit and tax function within one accounting firm (Cook and Omer, 2013; Dhaliwal et al., 2013; Seetharaman et al., 2011) and result in improved services. The alternative argument to knowledge spillover is that the lack of independence will result in a reduced tax rate from the auditor’s signing off on questionable aggressive tax planning. However, under both scenarios, APTS and tax rate will have a negative relationship. Building on these prior studies, H1 investigates the effects of reductions in APTS, perhaps through shifting of tax fees to non-audit firms. If firms are losing the benefits of APTS, then cash taxes paid will be increasing as APTS decrease. However, as tax planning may be a long-term investment (Rego and Wilson, 2012), the entire benefit must be measured over time. H1, stated in alternate form: H1. Fees paid for APTS are negatively associated with long-run cash taxes paid. Seemingly, when a firm is not using APTS providers, the work is still being completed, either by another firm or internally. Separating the audit and tax functions may limit the implementation of tax service recommendations, as auditors may be less likely to sign off on projects introduced by competing firms (Gleason and Mills, 2011). If CETR benefits are obtained from APTS, then firms that move back to APTS after a hiatus should once again experience the benefits of this synergy. Therefore, we examine whether re-engaging an auditing firm for tax services, after decreasing fees post-SOX, results in lower taxes paid. Our H2, stated in the alternate form: H2. Fees paid to re-engage an auditor for tax services are associated with lower long-run cash taxes paid. 3. Method 3.1 Data The sample was collected from Audit Analytics (2003-2009)[6] and Compustat Fundamentals annual financial statement data (fiscal years 2003-2009). As shown in Table I, reductions in firm-year observations occurred due to missing tax or control data from Compustat, the elimination of observations identified as REITS, the need to have five consecutive observations from Audit Analytics to capture the long-term impact of APTS and taxes paid and the elimination of firms which changed auditors, as we are studying the long-term benefits of auditor retention for audit and tax services. Finally, similar to Dyreng et al. (2008), we eliminated firms that had pretax income of less than zero to enhance the interpretation of the cash taxes paid measure. This leaves 11,409 individual firm-year observations. Of these firms, there are 1,330 firms that have complete long-term observations from 2003 to 2007, 1,477 with observations from 2004 to 2008 and 1,366 with observations from 2005 to 2009[7]. 3.2 Models and variables Our main analysis utilizes a fixed effects regression controlling for industry (two-digit SIC code) and year. We also cluster by firm to generate robust standard errors. All variables were examined for multicollinearity[8] to ensure that our regression results


RAF 14,3


290

Table I. Sample selection

Total observations from audit analytics from 2003 to 2009 Less: Observations missing cash taxes paid, pretax income or income taxes Less: Observations identified as REITs (SIC code 6798) Less: Observations without at least five consecutive years of observations Less: Observations missing control data from Compustat Less: Observations reporting an auditor change during the sample period Total Firm Observations Less: Observations with pretax income ⬍ 0 Total single year firm observations

126,007 (82,115) (1,303) (17,463) (5,933) (4,867) 14,326 (2,917) 11,409

Five-year sample breakdown (number of unique firms with full five-year data) 2003-2007 1,330 2004-2008 1,477 2005-2009 1,366 Total five year observations

4,173

were not influenced by correlations between our explanatory variables. The models to test H1 (model one) and H2 (model two), respectively, are: CETRi,t ⫽ ␤0 ⫹ ␤1AVG_TAXFEESi,t ⫹ ␤2RESTATEMENTi,t ⫹ ␤3CAPITALi,t ⫹ ␤4LOGSALESi,t ⫹ ␤5LOGMVEi,t ⫹ ␤6N_SEGi,t ⫹ ␤7BIG4i,t ⫹ ␤8NOLi,t ⫹ ␤9LEVERAGEi,t ⫹ ␤10ACQUISi,t

(1)

⫹ ␤11PRETAXROAi,t ⫹ ␤12FOREIGNi,t ⫹ ␤13PERCENTHELDi,t ⫹ ␤14BOARDINDi,t ⫹ ␤15⫺82INDUSTRYi,t ⫹ ␤83⫺84YEARi,t ⫹ ␧i,t CETRi,t ⫽ ␤0 ⫹ ␤1SMALLSPIKEi,t ⫹ ␤2MEDIUMSPIKEi,t ⫹ ␤3LARGESPIKEi,t ⫹ ␤4RESTATEMENTi,t ⫹ ␤5CAPITALi,t ⫹ ␤6LOGSALESi,t ⫹ ␤7LOGMVEi,t ⫹ ␤8N_SEGi,t ⫹ ␤9BIG4i,t ⫹ ␤10NOLi,t ⫹ ␤11LEVERAGEi,t ⫹ ␤12ACQUISi,t ⫹ ␤13PRETAXROAi,t

(2)

⫹ ␤14FOREIGNi,t ⫹ ␤15PERCENTHELDi,t ⫹ ␤16BOARDINDi,t ⫹ ␤17⫺84INDUSTRYi,t ⫹ ␤85⫺86YEARi,t ⫹ ␧i,t Note that all variable definitions are discussed below and included in the Appendix. All variables are obtained from Compustat unless otherwise noted. 3.3 Main variables The dependent variable CETRi,t represents a firm’s cash-effective tax rate. This rate is measured in various ways. For the main testing of model one, CETRi,t is represented by FIVEYEAR_CETR and is defined, similar to Dyreng et al. (2008), as the sum of cash taxes paid over five years divided by the five-year sum of firm pretax income after deducting special items[9]. CETRi,t is also denoted as AVG_CETR in model one, which represents the five-year average annual CETR over the sample period. This is included as a dependent variable in a separate long-term analysis for comparison with the Dyreng et al. (2008) measure (FIVEYEAR_CETR) in model one. In model two, CETRi,t is


represented by AVG_CETR which is the average of the annual CETR over a three-year period. AVG_TAXFEES, our measure of long-term APTS, is measured as the average tax fees paid scaled by selling, general and administrative expenses for each sample firm, by year, then averaged over the five-year sample period[10]. Using a per cent change from year one to year five could overly influence both these individual years, as well as fail to account for large variations that occur during the middle sample years. Therefore, average scaled tax fees, a level instead of change variable, were used to smooth out any potential single-year variations. Firms with lower tax fees paid to their auditors over the sample period are predicted in H1 to have higher long-term CETRs. Following, firms that invest in APTS may be more able to realize low CETRs over longer time horizons. A negative relation is predicted as lower fees paid should be associated with higher taxes paid[11]. To test H2, we investigate three scenarios of changes representing year (t) decreases in fees followed by subsequent year (t ⫹ 1) increases, between 10 and 40 per cent (SMALLSPIKE)[12], between 41 and 70 per cent (MEDIUMSPIKE) and over 70 per cent (LARGESPIKE). For each per cent analysis, we add indicator variables for firms that reduced their services in one year (t) and then increased their fees by at least that much in the following year (t ⫹ 1)[13]. For example, a company would only be classified as a MEDIUMSPIKE if the firm decreased APTS between 41 and 70 per cent in a single year (t) followed in year (t ⫹ 1) with an increase in APTS resulting in the firm at least paying the same (or greater) tax fees than they were paying prior to the decrease year (year t ⫺ 1)[14]. To test H2, we use a fixed effects model that controls for industry (two-digit SIC code) and year. Our dependent variable is (AVG_CETR), computed similar to our five-year average cash taxes paid measure, beginning the year after the fee increase for three years (t ⫹ 2 to t ⫹ 4) as an indication of the long term, non-immediate nature of tax planning[15]. Because some short-term benefit is expected with an increase in (return to) APTS (Cook and Omer, 2013), the question becomes: Does it outweigh the longer term loss from the previous decision to reduce APTS? We examine taxes paid in the three years following the two-year fee decrease/increase. However, examining fee decreases, and then subsequent increases, immediately following the implementation of SOX may be ideal as tax fee expenditure changes are more likely to have been initiated by this new legislation as opposed to other non-regulatory reasons (Cook and Omer, 2013). Our sample should include firms which reduced their APTS and then, perhaps realizing the benefits of knowledge spillover, re-engaged their audit firm for tax services. 3.4 Control variables We include industry-specific indicator variables and a number of control variables. We winsorize our continuous control variables at the 1st and 99th percentiles by year to account for any potential extreme values. The natural log of total sales (LOGSALES) and the natural log of the market value of equity (LOGMVE) are used to proxy for firm size and complexity (Ashbaugh et al., 2003; Frankel et al., 2002; Gupta et al., 2011)[16]. A measure of firm leverage (LEVERAGE), defined as total liabilities minus current liabilities divided by total assets, is also included (Mills et al., 1998). Because the analysis is long-term, the financial control variables are computed as the sum of the first year’s end-of-year number and the last year’s end-of-year number divided by two (Dyreng et al., 2008). For example, in the five-year analysis, LEVERAGE equals the ratio of total


RAF 14,3


292

liabilities minus current liabilities divided by total assets in year t plus the same ratio for year t ⫹ 4 divided by 2[17]. We also account for the tax complexity of firms (Mills et al., 1998; Seetharaman et al., 2011; Kinney et al., 2004). An indicator variable equal to one if the firm had a restatement in the tax year and zero otherwise (RESTATEMENT) is included and summed as the total number of restatement years during the five- (or three-) year period being studied. CAPITAL, which is computed as the ratio of net property, plant and equipment to total firm assets is included. When AVG_CETR is the dependent variable, CAPITAL is calculated as the average amount of net property, plant and equipment, scaled by total assets, over the five- (three-) year period. When FIVEYEAR_CETR is the dependent variable, CAPITAL is calculated as the ratio of net property, plant and equipment, scaled by total assets in the first year plus the same ratio for the fifth year divided by 2. PRETAXROA is used to control for the progressive nature of the tax rate schedule. The longer-term measures are also calculated according to the dependent variable as just described for CAPITAL. FOREIGN is an indicator variable equal to one if foreign taxes are positive (Compustat TXFO ⬎ 0) and zero otherwise, while N_SEG is a continuous variable equal to the number of reportable business segments. For our long-term analysis, we use the average number of segments over the reporting period used (N_SEG) and the sum of the number of years in which foreign operations are reported (FOREIGN). Acquisitions are controlled for with ACQUIS, an indicator variable equal to the sum of the number of years the firm had an acquisition in the five(or three) year period being studied. PERCENTHELD, representing the percentage of shares held by institutional investors divided by total shares outstanding is included and is predicted to be positively related to taxes paid. BOARDIND, representing the percentage of board members who would be considered independent and BIG4 is the sum of the number of years in which the firm engaged a big four accounting firm (BIG4) as auditor. NOL is an indicator variable equaling one if the firm had a net operating loss (NOL) carry-forward and zero otherwise. For the long-term analysis, the sum of NOL years is utilized (NOL). INDUSTRY is measured by two-digit SIC codes and is used to control for any industry-based influences on CETRs, while YEAR represents indicator variables for each sample year. In all analyses, these variables are consistently significant but not presented in the tables due to space limitations. 4. Results 4.1 Descriptive statistics Table II reports descriptive statistics for one- and five-year periods. The mean value for CETR in the one- and five-year periods is 0.25 and 0.25, respectively. These numbers are slightly lower than the figures reported of 27 (29) per cent for 1- (five) year CETRs reported in the Dyreng et al. (2008) study[18]. Looking at CHANGECETR, the yearly per cent change in CETR, highlights the importance of looking at a long-term measurement. The one-year mean change of 1.38 clearly indicates a significant amount of variation when compared to the median of 0.07. However, the mean (median) over five years is 0.28 (0.27). Firm and tax complexity demographics in this study are comparable to other studies using similar sample periods (Bédard and Paquette, 2011). Approximately 82 per cent of firm years are audited by one of the big four accounting firms, 74 per cent of directors are considered independent and institutional investor holdings are approximately

One year


Variables Tax variables CETR CHANGECETR BTD PERMBTD PRETAXINC TAX FEES LTAX FEES AVG_TAXFEES TAX RATIO FEEPERCENTCHANGE

Five years

Mean Median (n ⫽ 11,409/9,169) 0.25 1.38 0.03 0.03 482.47 388,663 9.47 0.00175 0.11 0.33

0.24 0.07 0.02 0.02 52.55 66,000 11.09 0.0007 0.08 ⫺0.02

Mean

Median (n ⫽ 4,173)

0.25 0.28 0.02 0.02 511.87 394,887 9.58 0.00165 0.11 0.34

0.26 0.27 0.02 0.02 58.79 90,771 10.88 0.0008 0.09 ⫺0.32

Firm control variables LOGSALES LOGMVE LEVERAGE PRETAXROA BIG4 PERCENTHELD BOARDIND

6.41 6.33 0.27 0.09 0.82 0.62 0.74

6.45 6.56 0.24 0.08 1.00 0.69 0.77

6.54 6.34 0.28 0.08 4.25 0.64 0.74

6.58 6.54 0.26 0.07 5.00 0.69 0.76

Tax complexity variables CAPITAL N_SEG NOL ACQUIS FOREIGN RESTATEMENT

0.22 2.54 0.35 0.16 0.51 0.13

0.14 1.00 0.00 0.00 1.00 0.00

0.23 2.53 1.83 0.79 2.73 0.58

0.17 1.60 0.00 0.00 4.00 0.00


Notes: See Appendix for variable definitions. All continuous variables are winsorized at top and bottom 1% of their cross-sectional distribution by year. For the one-year descriptives, “change” variables (CHANGECETR, FEEPERCENTCHANGE) have only 9,169 observations as a change can not Table II. Descriptive statistics be computed for the first year. This reduces the one-year sample by 2,240 firm-year observations

62 per cent. In regards to the tax complexity variables, 51 per cent of sample firm years have foreign operations, 35 per cent of firms have net operating losses in a single firm year, 13 per cent of firm years have a restatement and 16 per cent of firm years have a merger or acquisition. Finally, the short-term ratio of net property, plant and equipment to total assets (CAPITAL) is 0.22. This is much smaller than the 0.46 mean found by Mills et al. (1998). However, a direct comparison is hard to make, as the sample used in that study is from the early 1990s, while our sample utilizes firm data from 2003-2009. Overall, these complexity measures are largely consistent between the one- and five-year observations. H2 studies firms that have reduced their APTS and then increased them the following year. Table III provides descriptive statistics on the firms that show this type of fluctuation in their utilization of APTS. The SPIKE firms are companies who

Table III. Descriptive statistics–SPIKE firms 0.15 2.00 0.00 0.00 3.00 0.00

7.03 6.89 64.01 0.27 0.22 3.00 0.77 0.78

0.261 75,030

0.22 2.84 1.12 0.46 1.73 0.32

6.45*** 6.14*** 294.90** 0.27 0.19 2.50*** 0.64*** 0.74* 0.15 1.67 0.00 0.00 3.00 0.00

6.52*** 6.43*** 28.22*** 0.25 0.18 3.00 0.70*** 0.75**

0.284** 0.299*** 392,518 156,857***

LARGESPIKE Mean Median (n ⫽ 182)

0.23 3.20 1.20 0.54 1.76 0.34

6.83 6.56 181.98** 0.30 0.20 2.71 0.74 0.77 0.15 1.67 0.00 0.00 3.00 0.00

6.95 6.79 60.75 0.28 0.20 3.00 0.76 0.80

0.24 3.06 1.34 0.51 1.58*** 0.43**

6.95 6.61 454.01 0.29 0.16** 2.70 0.73 0.76

0.16 1.83 1.00 0.00 2.00*** 0.00

6.89 6.75 51.84 0.24 0.17** 3.00 0.76 0.76

0.282* 0.283* 0.242** 0.244 288,409 147,467*** 171,698*** 58,360

MEDIUMSPIKE Mean Median (n ⫽ 125)

Notes: * , ** and *** denote two-tailed statistical significance at 10, 5 and 1%, respectively; see Appendix for variable definitions. SPIKE is an indicator variable equal to 1 if the firm experienced a reduction in tax fees of 10-40% (SMALLSPIKE), 41-70% (MEDIUMSPIKE) or greater than 70% (LARGESPIKE) followed by an increase that exceeded the prior year percent decrease for various analysis. In addition, SPIKE ⫽ 1 only if the increase year returned the firm to at least the tax fee amounts paid prior to the decrease year. CETR equals the average CETR over the three-year sample period following the SPIKE year. The non-SPIKE sample represents the firms that did not have SPIKE ⫽ 1 for any of the SPIKE calculations and includes all firms with decreases in APTS in year one of at least 10%. The SPIKE sample represents all SPIKE firms that meet the criteria for each SPIKE category. In other words, the MEDIUMSPIKE group equals only SPIKE firms that returned to at least their initial pre-decrease level in the year immediately following a decrease in fees of between 41-70%. All mean/median tests are between each individual SPIKE category and the sample of Non-SPIKE firms

0.23 2.87 1.26 0.53 1.88 0.33

7.04 6.79 630.96 0.28 0.21 2.67 0.70 0.76

0.262 360,507

Tax complexity variables CAPITAL N_SEG NOL ACQUIS FOREIGN RESTATEMENT

Firm control variables LOGSALES LOGMVE PRETAXINC LEVERAGE PRETAXROA BIG4 PERCENTHELD BOARDIND

Tax variables CETR TAX FEES

SMALLSPIKE Mean Median (n ⫽ 267)

294

Variables

NON-SPIKE SAMPLE Mean Median (n ⫽ 1,713)


RAF 14,3


increased their APTS in the year immediately following a decrease year[19]. To see if these firms differ from other firms that decreased APTS over the same period, we isolate these firms as the control sample. The firms are segregated by the amount of fluctuation they have in APTS. As can be seen in Table III, for the most part, there are few significant differences between the SPIKE and non-SPIKE firms. Consistent with our expectations, CETRs of SPIKE firms show a significant decreasing pattern as the level of subsequent year fee increase rises. Firms in the SMALLSPIKE category exhibit significantly higher CETRs (0.284) over the non-SPIKE sample in the three years following their return to APTS while the LARGESPIKE firms have the lowest CETR (0.242). Examining the SMALLSPIKE group, it is interesting to note that these firms pay the highest APTS on average (not significantly different from non-SPIKE sample) while reporting the highest mean CETR. Additionally, they are significantly smaller than the non-SPIKE group in terms of sales (LOGSALES) and are less likely to choose a Big4 provider. However, none of the tax complexity variables is significantly different than the non-SPIKE sample. Further, pre-tax income appears lower, on average, for the two smaller SPIKE groups. 4.2 Results of hypothesis tests H1 examines the relation between long-term taxes and fees paid to auditing firms for tax services. Results supporting H1 are reported in Table IV. As previously defined, two different five-year measures are used (FIVEYEAR_CETR and AVG_CETR). FIVEYEAR_CETR is used to minimize the potential effects of single-year extreme values (Dyreng et al., 2008), while AVG_CETR provides supplemental support. In this regard, AVG_CETR will confirm whether single-year variations overly influence total long-term cash taxes paid. The first column of Table IV shows results for FIVEYEAR_CETR. We find the expected strong negative relationship between APTS and CETRs to be significant at 1 per cent. Having lower APTS is associated with an increased tax burden as measured by long-term CETR. The coefficient on AVG_TAXFEES is ⫺2.68. Using the five-year average of our sample (AVG_TAXFEES ⫽ 0.00165) will result, on average, in a 0.4422 per cent lower CETR. When applied to the sample average pretax book income of $511.87 million, the tax savings result is almost $2.26 million per year with the five-year savings of $11.30 million[20]. This long-term investment in tax planning, particularly with APTS, is significantly associated with fewer taxes paid[21]. Our supplemental long-term measure, AVG_CETR, also shows a significant and expected negative relationship in Table IV. When extrapolating to the sample, these findings produce an average tax savings over five years of $5.95 million. It is important to note that, although this alternative test does provide robustness for our results, it is possible that these results are biased by extreme one-year variations in the annual measures of CETR (Dyreng et al., 2008). Overall, the analysis shows that APTS are negatively correlated with cash taxes paid over the long run. H2 examines firms that increase their APTS after having reduced those services. If firms potentially pay higher taxes following reductions in APTS, then H2 analyzes whether firms that reengage their auditors for tax services regain any benefits in the following years. The results of H2 are provided in Table V and show some interesting findings. The variable of interest, the different levels of spike, has a positive and significant coefficient in the SMALLSPIKE analysis, positive and marginally


RAF 14,3


296

Independent variablesa AVG_TAXFEES RESTATEMENT CAPITAL LOGSALES LOGMVE N_SEG BIG4 PERCENTHELD BOARDIND NOL LEVERAGE PRETAXROA ACQUIS FOREIGN Adjusted R2 F-statistic Pr ⬎ F N

Prediction

FIVEYEAR_CETR

AVG_CETR

⫺ ? ⫺ ? ? ? ? ⫹ ? ⫺ ⫺ ? ? ?

⫺2.68*** 0.02 ⫺0.01 0.03 ⫺0.06*** 0.00 0.02 ⫺0.04 ⫺0.77*** ⫺0.07*** ⫺0.66*** 0.24 ⫺0.02 0.04*** 0.77 295.40 ⬍0.0001 4,173

⫺1.41** 0.04** 0.09 0.03 ⫺0.03 0.01* 0.05*** 0.02* ⫺0.39*** ⫺0.06*** ⫺0.40*** 0.22*** ⫺0.02 0.05*** 0.84 395.89 ⬍0.0001 4,173

Notes: * , ** and *** denote statistical significance at 10, 5 and 1%, respectively; a industry-level fixed effects based on two-digit SIC code industry classifications are included in the regression analysis (omitted to conserve space). t-statistics using Huber–White corrections (Gleason and Mills, 2011) and clustered by firm are included below coefficient estimates. Regression analysis of measures of long-term measures of tax avoidance on tax services. AVG_TAXFEES represent tax fees paid to the auditing firm (APTS), scaled by selling, general and administrative costs (data XSGA), averaged over the years being Table IV. Long-term regression studied. The AVG_CETR variable is computed by summing the individual CETR measures and then analysis of tax dividing them by the number of sample periods. FIVEYEAR_CETR is computed similar to Dyreng et al. avoidance on tax (2008) (⌺ cash taxes paid/⌺ pretax income ⫺ ⌺ special items) over the five-year sample period. For a complete description of all variables, see the Appendix services

significant in the MEDIUMSPIKE and negative and significant in the LARGESPIKE analysis. The positive coefficient on SMALLSPIKE indicates that for the three years following the return to previous levels of APTS (t ⫹ 2 to t ⫹ 4), SMALLSPIKE firms have higher CETRs than the sample of firms (non-SPIKEs) that decreased fees in year one (t) and did not reengage their auditor to at least their pre-decrease level over the remaining sample period (t ⫹ 1 to t ⫹ 4)[22]. Results for the SMALLSPIKE firms exhibit a 2.0 per cent increase in three-year average CETR as compared to the other firms. The 2.0 per cent increase in CETR is quite significant when analyzed with firm pretax income. The demographics for the SPIKE sample (Table III) shows pretax income for the SMALLSPIKE firms to be $294.90 million which results in an increased tax cost on average of approximately $5.9 million per year and almost $18 million for three years over firms that do not reengage their auditors after a similar decrease. Although marginally significant, firms in the MEDIUMSPIKE also continue to exhibit higher taxes paid when compared to the control sample of firms. However, when looking at the LARGESPIKE firms, results confirm that large increases in APTS, after prior year similar decreases yield renewed benefits upon return. These greater than 70 per cent SPIKE firms reduce their CETR by an additional 1.9 per cent when compared to the


Independent variablesa INTERCEPT SMALLSPIKE MEDIUMSPIKE LARGESPIKE PRETAXROA RESTATEMENT CAPITAL LOGSALES LOGMVE N_SEG BIG4 PERCENTHELD BOARDIND NOL LEVERAGE ACQUIS FOREIGN Adjusted R2 F-statistic Pr ⬎ F N

Coefficient 0.297*** 0.020** 0.021* ⫺0.019** 0.067*** 0.012*** ⫺0.058*** 0.010*** ⫺0.002 ⫺0.002 0.006 ⫺0.015 ⫺0.025 ⫺0.011*** ⫺0.064*** ⫺0.003 ⫺0.010*** 0.08 11.00 ⬍0.0001 2,287


Notes: Coefficient results are included, while *** , ** and * indicate two-tailed statistical Table V. significance at 1, 5 and 10% levels; a industry variables omitted to conserve space; see the Appendix for Regression analysis of returning to all variable definitions. Results above are for regression analysis with three-year average CETR (AVG_CETR) as the dependent variable and indicator variables representing increases in tax fees paid auditor provided tax in the year immediately following a decrease in fees paid to the auditing firm of at least 10-40, 41-70 and services on long-term greater than 70%. Firms are not included as SPIKE firms if increase year fees do not exceed the fees paid taxes paid dependent variable ⫽ in the year prior to the decrease year. In this regard, firms are SPIKE firms if they return to pre-decrease AVG_CETR levels

subsample of firms that decreased APTS. The average tax savings for the sample would be $8.63 million per year or $25.9 million over three years. Examining the different results between the SMALLSPIKE and LARGESPIKE firms, it is important to note that, overall, when considering the coefficient on LARGESPIKE and the intercept, that the firm is still experiencing an increase in CETR. However, it is just a smaller increase than those firms which had a SMALLSPIKE or did not have any subsequent increase after their decrease. Overall, there is still a significantly smaller increase in CETR for the LARGESPIKE firms. With the SMALLSPIKE firms, it may be a simple matter of the increases being compliance oriented or less aggressive or small planning initiatives[23]. On the other hand, the LARGESPIKE firms may be buying back that synergy with more services. Our results provide support to the theory that the potential beneficial effects from knowledge spillover are mitigated upon reductions in APTS, at least to some extent. This may be surprising given that the reduction was only for one year. Although firm knowledge might have been lost, another potential explanation might be that firms, still concerned about independence perceptions, reengaged their auditors but only a small

RAF 14,3


298

group were willing to sign off on more aggressive tax planning strategies that would result in immediate tax benefits. These firms may have overreacted (large decreases) over fears of independence concerns but were quick to reestablish their previous fee levels after this initial year passed. Overall, only firms that reengage their auditor for tax services at the highest level (greater than 70 per cent) obtain lower CETRs in the immediate years following their return. In this case, some of the knowledge spillover is regained at a price. 5. Summary and conclusions This study examines the impact of tax fees paid to auditors on cash taxes paid. Results show that, over the long term, lower levels of fees paid to auditing firms are associated with higher taxes. Therefore, the long-term effects, as opposed to only short-term, should play a significant role in the cost-benefit analysis of the decision to utilize APTS. Replacing APTS with the services of another firm should still produce tax savings. However, if knowledge spillover does exist (Larcker and Richardson, 2004; Seetharaman et al., 2011), then synergies gained from combining the audit and tax function within one accounting firm might be lost. Although causality is inconclusive, our results appear to support the theory of knowledge spillover. H2 further explores the relationship between CETRs and APTS by studying those firms that return to APTS after a prior year fee reduction. Our results indicate that, although some value is gained by returning to APTS, there appears to be a detrimental effect to the firm’s long-term tax burden that corresponds to the amount of tax services dropped and then repurchased at a later date, unless that repurchase is over 70 per cent. When combined with the results in H1, it seems likely that the disruption in APTS damaged the efficiencies and synergies created through a combined audit–tax engagement. It is possible, even likely, that benefits would return in the coming years. This study contributes to the extant literature in several ways. The results contribute to the policy discussion on auditor independence. The tax savings benefit associated with APTS is likely to be a cost if further restrictions on non-audit services to audit clients are implemented. Additionally, we contribute to the growing literature on the tradeoff between independence and knowledge spillover. We are the first, to our knowledge, to look at the long-term benefits of APTS as they relate to tax savings. Further, we also are the first study to look at firms that return to APTS after a reduction. We find that even a short break from APTS can lead to a higher tax burden for most firms, unless the increase back to APTS is greater than 70 per cent. There are limitations in this study. As pointed out by Maydew and Shackelford (2007), tax fees paid to non-audit firms are not publicly available. While this study does provide evidence that APTS have value, firms that are engaging their auditor for tax services may also be paying outside firms for tax planning. Increasing APTS may be accompanied by increasing fees to outside firms and vice versa. As this additional variable is unknown, the effects accrued from each category are unknown. Further, the amount of tax fees reported are not segregated between compliance fees and planning fees. Presumably, compliance fees will be relatively stable from year to year, and most changes will be reflective of changes in planning fees. Although this is a reasonable assumption, there is no way to verify this. Additionally, as the sample solely includes companies that paid their auditors for tax services during any period in 2003-2009, the possibility exists that some companies only


paid fees to non-audit firms, or did not pay fees to either type of accounting firm and solely relied on internal tax services. If this is the case, the restrictive sample limits the generalizability of any conclusions. Further, the time frame includes the implementation of FIN 48 in 2006. It is possible that this implementation has changed firms’ level of tax services in general or those contracted with their auditor. Lastly, as mentioned by Whisenant et al. (2003), the pricing of audit and tax fees are jointly determined. Additionally, the choice to buy APTS is perhaps endogenous. Firms that choose to buy APTS presumably have tax planning opportunities available resulting in reduced CETRs. Firms with fewer opportunities choose not to buy APTS and have higher CETRs. Future research should study the complex endogeneity of this relationship. There are also opportunities for additional research. Dyreng et al. (2008) look at a 10-year time horizon for tax planning implications. Researchers can look at a longer horizon relationship between APTS and CETRs. Additional research can also examine the sample of firms who return to their auditors after reducing services. As our results provide support for the detrimental impact of APTS reductions in certain groups, research can further examine potential motivations of these firms. Notes 1. Although we do not make any commentaries on whether reduced taxes are good or bad, presumably firms and shareholders would prefer to pay less taxes while policymakers would prefer firms that are less tax aggressive. 2. It is important to note that these studies predominantly look at all non-audit services provided by the auditor and not just APTS. In general, APTS may be viewed by some as less of a threat to independence than other services. 3. The definition of long-term is largely subjective though studies often use five- or ten-year horizons (Dyreng et al., 2008). 4. An unnecessary tax does not imply tax savings from illicit or illegal activities (activities that may result in IRS audit adjustments and/or financial statement restatements). These taxes are merely those that can be eliminated or delayed through legal tax planning opportunities. 5. The Omer et al. (2006) study uses marginal tax rates (MTRs) and effective tax rates (ETRs) as their tax variables. While their ETR measure is similar to the CETR measure used in our study, we also examine the long-term impact of tax fees on tax rates, as opposed to the one year (2002) used in their study. 6. It is important to note that Fin 48 became effective for sample firms in 2006. It is unclear as to how this implementation will effect firm tax planning. 7. Results for all five-year analysis tests were qualitatively similar when analyzed with only the 2003-2007 sample firms (the time period immediately following SOX). 8. We use a VIF statistic to check for multicollinearity. All test results were well below the threshold for concerns of multicollinearity. 9. As documented by Dyreng et al. (2008), the cash-based taxes paid (CETR) is limited because the denominator (pretax income minus special items) may be affected by activities designed to manage pretax income that have no effect on cash taxes paid, such as accruals management. Those authors note that a long-term analysis of cash taxes paid should be less affected by this limitation as the long-term measure should capture the reversal of these accruals.


RAF 14,3


300

10. We scale our measure of long-term APTS by selling, general and administrative expenses similar to prior literature (Mills et al., 1998; Gleason and Mills, 2011) to ensure that high-volume and low-margin firms do not unnecessarily skew the ratio. 11. One other independent variable that has been used is the ratio of non-audit fees to total fees (Bedard et al., 2010) or tax to audit fees. We did not use this measure of auditor independence because firm ratios have also changed due to increased audit fees for compliance post-SOX. As we are more concerned about the impact of changes in tax fees on taxes paid, as opposed to the change in the overall fee structure, we did not separately test this measure. 12. Any changes less than 10 per cent are likely to be due to year-to-year variations in compliance fees paid and not necessarily due to management decisions about changing service providers for tax consulting. We chose our percent breakdown of APTS to approximate the nearest 10 per cent for a split of our firms into three groups. Further, we wanted to ensure a sufficient sample size for each group. 13. Firms in the larger SPIKE categories are not classified as SPIKE firms in lower percent groups. In other words, firms are classified in groups based upon their initial percent decrease (10-40, 41-70, greater than 70 per cent) as well as their subsequent return (SPIKE) to pre-decrease fee levels. 14. For example, a decrease of 30 per cent (t), followed by an increase of 30 per cent (t ⫹ 1) would result in the firm paying less fees as the year prior to the decrease (t ⫺ 1). Further, we only included firms as SPIKEs that had an increase in the year following a decrease rather than any subsequent year following a decrease due to constraints of the time horizon. 15. We also ran our SPIKE analysis with THREEYEAR_CETR, computed similar to our FIVEYEAR_CETR variable as the main DV and results were qualitatively similar. Further, we also examined CETRs in the year of the increase (short-term analysis) to test whether firms that reengage their auditors immediately pay lower cash taxes. This short-term SPIKE analysis showed that our two smallest SPIKE categories (SMALLSPIKE and MEDIUMSPIKE) had significantly higher CETRs in the SPIKE year (p ⬍ 0.05), while the largest SPIKE category (LARGESPIKE) was positive and insignificant. This indicates that any benefits lost due to the break in services may be immediately felt for these specific categories. In addition, as our long-term analysis is negative and significant (p ⬍ 0.05), while the short-term results were positive and insignificant for the largest SPIKE group (LARGESPIKE), implies that any benefits received for returning may not be immediate. 16. In supplemental analysis (untabulated), the natural log of total assets (LOGTA) was also used as a control variable for firm size. Results were not significantly different. 17. When FIVEYEAR_CETR is the main DV, long-term continuous control variables are computed similar to Dyreng et al. (2008). For example, LEVERAGE is the ratio of total liabilities minus current liabilities divided by total assets in year t plus the same ratio for year t ⫹ 4 divided by 2. When AVG_CETR is the main DV, average values of continuous variables are used over the five-year sample period. Long-term indicator variables (“SUM”) are computed in the same manner for both DV variable studies and are equal to the sum of all indicator years. 18. The Dyreng et al. (2008) paper used 1995-2004 data, while our sample comprised firm years 2003-2009. One potential explanation is the Dyreng et al. (2008) sample includes all firms with available taxes paid data as opposed to firms that also have fees paid to their auditors.

19. Our SPIKE sample does not appear to be driven by any particular external tax event that occurred over the sample period. Therefore, SPIKEs in a particular tax year, for example, caused by increased APTS due to the Dividend Tax Repatriation Holiday in 2004 do not appear to be driving our results.

Auditorprovided tax services


20. For purposes of estimate, we use pretax book income, as actual taxable income is not known. 21. Firms that are either increasing or decreasing APTS may have different motivations for making that decision. Treating them as a homogenous group may obscure information regarding changes in APTS and the association with CETRs. In untabulated results, we divided our sample based upon firms that increased or decreased fees over the five-year sample period. Relationships between APTS and CETRs in these tests were similar and in the predicted directions to the full sample results. 22. A five-year window is used for our SPIKE analysis. Year t represents the year of decrease for both our SPIKE sample as well as our NON-SPIKE group. In the second year (t ⫹ 1), the SPIKE sample reengages their auditor to at least the same fee level as was paid prior to the year t decrease. The NON-SPIKE sample does not return to this pre-decrease level. We then measure CETRs over the following three years (t ⫹ 2 to t ⫹ 4) for these two groups and ensure that the non-SPIKE firms in our sample never return in these years to their pre decrease levels of APTS. 23. One additional explanation might be that non-SPIKE firms, after reducing APTS, might have chosen to engage other firms for their tax services. As fees paid to non-audit firms are unknown, we could not directly test these fees versus firms that return to their auditors in whatever increment.

References Abbott, L.J., Parker, S., Peters, G. and Raghunandan, K. (2003), “An empirical investigation of audit fees, non-audit fees, and audit committees”, Contemporary Accounting Research, Vol. 20 No. 2, pp. 215-234. Ashbaugh, H., LaFond, R. and Mayhem, B. (2003), “Do non-audit services compromise auditor independence? Further evidence”, The Accounting Review, Vol. 78 No. 3, pp. 611-639. Audit Analytics (2008), Non-Audit Fees: A Five Year Trend, Audit Analytics. Bédard, J. and Paquette, S. (2011), “Perception of auditor independence, audit committee characteristics, and auditor provision of tax services”, Working Paper, Laval University, Quebec City. Bedard, J.C., Falsetta, D., Krishnamoorthy, G. and Omer, T. (2010), “Voluntary disclosure of auditor-provided tax service fees”, The Journal of the American Taxation Association, Vol. 32 No. 1, pp. 59-77. Choi, J., Kim, J. and Zang, Y. (2010), “Do abnormally high audit-fees impair audit quality?”, Auditing: A Journal of Practice and Theory, Vol. 29 No. 2, pp. 115-140. Chung, H. and Kallapur, S. (2003), “Client importance, non-audit services, and abnormal accruals”, The Accounting Review, Vol. 78 No. 4, pp. 931-955. Cook, K., Huston, G.R. and Omer, T. (2008), “Earnings management through effective tax rates: the effects of tax-planning investment and the Sarbanes-Oxley act of 2002”, Contemporary Accounting Research, Vol. 25 No. 2, pp. 447-471. Cook, K. and Omer, T. (2013), “The cost of independence: evidence from firms’ decisions to dismiss auditors as tax service providers”, Working Paper, TX Tech University, Lubbock, TX.

301

RAF 14,3


302

Cripe, B. and McAllister, B. (2009), “Determinants of audit/tax separation decisions”, American Journal of Business, Vol. 24 No. 1, pp. 47-56. Davis, S.M. and Hollie, D. (2008), “The impact of non-audit service fee levels on investors’ perception of auditor independence”, Behavioral Research in Accounting, Vol. 20 No. 1, pp. 31-44. DeFond, M., Raghunandan, K. and Subramanyam, K.R. (2002), “Do non-audit service fees impair auditor independence? Evidence from going concern audit opinions”, Journal of Accounting Research, Vol. 40 No. 4, pp. 1247-1274. Dhaliwal, D., Gal-Or, R., Naiker, V. and Sharma, D. (2013), “Auditor-provided tax services and tax avoidance”, Working Paper, University of Arizona, Tucson, AZ. Dyreng, S., Hanlon, M. and Maydew, E. (2008), “Long-run corporate tax avoidance”, The Accounting Review, Vol. 83 No. 1, pp. 61-82. Elder, R., Harris, D. and Zhou, J. (2008), “Tax consulting and reported weaknesses in internal control”, Working Paper, Syracuse University, Syracuse, NY. Favere-Marchesi, M. (2006), “The impact of tax services on auditors’ fraud-risk assessment”, Advances in Accounting, Vol. 22, pp. 149-165. Frankel, R., Johnson, M. and Nelson, K. (2002), “The relation between auditors’ fees for non-audit services and earnings management”, The Accounting Review, Vol. 77, pp. 71-105. Gleason, C. and Mills, L.F. (2011), “Do auditor-provided tax services improve the estimate of tax expense?”, Contemporary Accounting Research, Vol. 28 No. 5, pp. 1484-1509. Grant, T., Park, N. and Wheeler, S. (2008), “Post-SOX market perceptions of non-audit services and auditor independence”, Advances in Accounting, Finance and Economics, Vol. 1 No. 1, pp. 1-16. Gupta, S., Mills, L.F. and Towery, E. (2011), “Did Fin48 arrest the trend in multistate tax avoidance?”, Working Paper, MI State University, Ann Arbor, MI. Hanlon, M. and Heitzman, S. (2010), “A review of tax research”, Journal of Accounting and Economics, Vol. 50 Nos 2/3, pp. 127-178. Kinney, W., Palmrose, Z.-V. and Scholz, S. (2004), “Auditor independence, non-audit services, and restatements: was the US government right?”, Journal of Accounting Research, Vol. 42 No. 3, pp. 561-588. Knechel, W.R. and Sharma, D.S. (2012), “Auditor-provided non-audit services and audit effectiveness and efficiency: evidence from pre- and post-SOX audit report lags”, Auditing: A Journal of Practice & Theory, Vol. 31 No. 4, pp. 85-114. Krishnan, G.P. and Yu, W. (2011), “Further evidence on knowledge spillover and the joint determination of audit and non-audit fees”, Managerial Auditing Journal, Vol. 26 No. 3, pp. 230-247. Krishnan, G.P., Visvanathan, G. and Yu, W. (2013), “Do auditor-provided tax services enhance or impair the value relevance of earnings?”, Journal of American Taxation Association, Vol. 35 No. 1, pp. 1-19. Larcker, D. and Richardson, S. (2004), “Fees paid to audit firms, accrual choices, and corporate governance”, Journal of Accounting Research, Vol. 42 No. 3, pp. 625-658. McGuire, S.T., Omer, T.C. and Wang, D. (2012), “Tax avoidance: does tax-specific industry expertise make a difference?”, The Accounting Review, Vol. 87 No. 3, pp. 975-1003. Markelevich, A., Barragato, C.A. and Hoitash, R. (2005), “The nature and disclosure of fees paid to auditors: an analysis before and after the Sarbanes-Oxley Act”, The CPA Journal, Vol. 75, pp. 6-10.


Maydew, E.L. and Shackelford, D.A. (2007), “The changing role of auditors in corporate tax planning”, in Auerbach, A., Hones, J. Jr and Slemrod, J. (Eds), Taxing Corporate Income in the 21st Century, Cambridge University Press. Mills, L.F., Erickson, M.M. and Maydew, E.L. (1998), “Investments in tax planning”, The Journal of the American Taxation Association, Vol. 20 No. 1, pp. 1-20. Omer, T., Bedard, J. and Falsetta, D. (2006), “Auditor-provided tax services: the effects of a changing regulatory environment”, The Accounting Review, Vol. 81 No. 5, pp. 1095-1117. Prawitt, D.F., Sharp, N. and Wood, D. (2012), “Internal audit outsourcing and the risk of misleading or fraudulent financial reporting: did Sarbanes-Oxley get it wrong?”, Contemporary Accounting Research, Vol. 29 No. 4, pp. 1109-1136. Rego, S.O. and Wilson, R. (2012), “Equity risk incentives and corporate tax aggressiveness”, Journal of Accounting Research, Vol. 50 No. 3, pp. 775-810. Seetharaman, A., Sun, Y. and Wang, W. (2011), “Tax-related financial statements restatements and auditor-provided tax services”, Journal of Accounting, Auditing and Finance, Vol. 26 No. 4, pp. 677-698. The Sarbanes-Oxley Act of 2002, Section 201. Whisenant, S., Sankaraguruswamy, S. and Raghunandan, K. (2003), “Evidence on the joint determination of audit and non-audit fees”, Journal of Accounting Research, Vol. 41 No. 4, pp. 721-744.


RAF 14,3

304

Appendix

Variable

Description

ACQUIS

1 if the firm had a merger/acquisition in the current year (variable Sale_FN ⫽ “AA” or “AB”) 0 otherwise and summed over the time period indicated 1 If auditor_fkey ⫽ 1, 2, 3 or 4 (Audit Analytics); 0 otherwise; long-term analysis is the sum of the variable over the time period indicated From BoardEX database, represents the percentage of the total members of the firm’s board of directors who would be considered independent Net property, plant and equipment (data PPENT) scaled by total assets (data AT); long-term analysis uses the average over the time period indicated when AVG_CETR is the dependent variable or equal to first-year calculation plus last-year calculation divided by two when FIVEYEAR_CETR is the dependent variable CETR equals cash taxes paid (data TXPD) divided by (pretax income (data PI) – special items (data SPI)), AVG_CETR equals the average of each annual CETR over the time period indicated Percent change in CETR from year t ⫺ 1 to t. For long-term analysis, FIVEYEAR_CETR is computed similar to Dyreng et al. (2008) (⌺cash taxes paid/⌺pretax income – ⌺ special items) Percentage of current tax fees ⫺ prior year tax fees divided by prior year tax fees. Tax fees are scaled by SG&A expenses (data XSGA). For the long-term analysis, unlike the short-term in which changes are computed from years t ⫺ 1 to t (for example, 2004 is computed as the change from 2003 to 2004), scaled changes are computed from the first reported observation (year t) to the fifth year observation (year t ⫹ 4). For example, changes from the 2003 (t) ⫺ 2007 (t ⫹ 4) tax years is used to compute this five-year scaled change 1 if foreign taxes paid; 0 otherwise (data TXFO) Liabilities total (Compustat LT) minus current liabilities (data LCT) scaled by total assets (data AT) averaged over the time period indicated when AVG_CETR is the dependent variable or equal to first-year calculation plus last-year calculation divided by two when FIVEYEAR_CETR is the dependent variable Natural log of market value at the fiscal year end (data MKVALT) averaged over the time period indicated when AVG_CETR is the dependent variable or equal to first-year calculation plus last-year calculation divided by two when FIVEYEAR_CETR is the dependent variable Natural log of total revenue (data REVT) averaged over the time period indicated when AVG_CETR is the dependent variable or equal to first-year calculation plus last-year calculation divided by two when FIVEYEAR_CETR is the dependent variable Natural log of tax fees paid to the auditing firm (APTS) from Audit Analytics Number of business segments averaged over the time period indicated 1 if tax loss carryforward ⬎0 (data TLCF) and 0 otherwise summed over the time period indicated From Thomson Reuters Institutional Holdings (13F) database, represents the percentage of shares held by institutional investors divided by total outstanding shares Pretax Income (data PI) in millions averaged over the time period indicated (continued)

BIG4


BOARDIND CAPITAL

CETR/AVG_CETR

CHANGECETR/ FIVEYEAR_CETR FEEPERCENTCHANGE

FOREIGN LEVERAGE

LOGMVE

LOGSALES

LTAXFEES N_SEG NOL PERCENTHELD

PRETAXINC

Table AI.

Variable

Description

PRETAXROA

Pretax Income (data PI) ⫺ special items (data SPI)/total assets (data AT) averaged over the time period indicated when AVG_CETR is the dependent variable or equal to first-year calculation plus last-year calculation divided by two when FIVEYEAR_CETR is the dependent variable 1 if the firm announced a restatement in the current year and 0 otherwise summed over the time period indicated Indicator variable if the firm experienced a reduction in tax fees of between 10-40%, 41-70% or greater than 70% followed in the next tax year by an increase for various analyses; SPIKE ⫽ 1 only if increase year returns firm to at least amount of fees paid in the pre-decrease year, 0 otherwise. Tax fees paid to the auditing firm (APTS) from Audit Analytics. AVG_TAXFEES is the TAX FEES measure, scaled by selling, general and administrative costs (data XSGA), averaged over the years being studied (3 or 5 years) Ratio of tax fees paid to auditing firm to total fees reported paid to auditing firm

RESTATEMENT


SPIKE

TAX FEES/AVG_TAXFEES

TAX RATIO

Note: All data are from Compustat unless otherwise indicated

About the authors Brian Hogan is a Clinical Assistant Professor of business administration at the University of Pittsburgh. His research focuses on the effects of corporate disclosures on internal and external parties as well as studying the implications of changes in corporate and individual tax policies. He has published papers in The Journal of the American Taxation Association, The Journal of Accounting and Finance and Global Perspectives in Accounting Education. Brian Hogan is the corresponding author and can be contacted at: [email protected] Tracy Noga is an Associate Professor of accountancy at Bentley University with a specialty in researching and teaching taxation. She earned her PhD at Texas Tech University. She is currently interested in corporate book-tax differences, auditor-provided tax services and corporate lobbying. She has published in Journal of Corporate Finance and Accounting Horizons.

For instructions on how to order reprints of this article, please visit our website: www.emeraldgrouppublishing.com/licensing/reprints.htm Or contact us for further details: [email protected]


Table AI.