2.1 APTS and knowledge spillover arguments

The knowledge spillover effect assumes that the knowledge is transferable between different departments within an audit firm, and the audit and tax services require overlapping information pertinent to clients.⁵ Such information sharing is beneficial for several reasons. First, since APTS normally are performed across the fiscal year, the communication between tax and audit partners allows the audit team to be aware of material risky transactions at an early stage, thereby enabling the audit team to detect and remedy clients' internal control weaknesses (both tax and nontax related) before the release of the financial statements (De Simone et al., 2015). Second, prior studies find a strong and positive association between financial and tax reporting aggressiveness (Frank et al., 2009). When auditors provide tax services to their audit clients, auditors will gain good understandings of the client's tax strategies. Such understandings not only facilitate auditors attesting clients' tax-related assertions but also help auditors assess clients' attitudes towards financial reporting aggressiveness. Therefore, compared with auditors who only provide audit services (i.e., less informed auditors), auditors will likely design significantly different and more effective audit procedures when they provide both audit services and APTS to their clients (Joe & Vandervelde, 2007).

More importantly, when the incumbent auditors also provide APTS, audit personnel can learn and evaluate clients' uncertain tax positions easily: the knowledge that increases audit quality. On the contrary, if firms purchase tax services from parties other than the incumbent auditors, then the latter must first detect clients' uncertain tax positions and then obtain evidence of the outcomes of those positions. By using various measures of audit quality, the majority of empirical studies find a positive relationship between audit quality and APTS, indicating the existence of knowledge spillover effects (e.g., Kinney et al., 2004; Robinson, 2008). The extant literature also considers whether or not APTS are recurring in nature (e.g., Abdul Wahab et al., 2014; Paterson & Valencia, 2011). Recurring APTS are defined as tax services purchased from the same auditors for two or more consecutive years: otherwise, they are nonrecurring APTS. Recurring NAS gives rise to economies of scope, which would contribute to auditors' cost savings (e.g., Beck et al., 1988; Chung & Kallapur, 2003), especially for some categories of NAS (e.g., APTS) (Arruñada, 1999; Francis, 2006; Gleason & Mills, 2011). If auditors pass on the part of the cost savings to their audit clients, economic bonding between the two parties is declined, with a consequent positive effect on audit quality.

Moreover, the benefits of knowledge spillover effects are generated not only from client-specific knowledge but also from industry-specific knowledge obtained through providing tax services to firms in the same industry. McGuire et al. (2012) propose that auditors have superior knowledge of the industry-specific tax planning opportunities available to their clients if auditors are tax industry specialists. Such specialisation is also found to be helpful in establishing formal or informal benchmarks for reasonable accounting practices, and hence, auditors could detect and constrain industry-specific opportunistic earnings management behaviour (Christensen et al., 2015). Thus, tax industry specialisation could enhance not only tax service effectiveness but also audit quality. These benefits even exist when clients do not purchase APTS or purchase low levels of APTS from their incumbent auditors since the knowledge could spill over to different engagements across the same audit firm/office. However, Goldman et al. (2021) argue that the improved audit quality pertinent to income tax accounts is due to auditors possessing more general tax task-specific knowledge, rather than to industry expertise, since tax issues are not necessarily industry-specific (Hux et al., 2018). Goldman et al. (2021) find that audit offices with more general tax task-specific knowledge (proxied by audit offices' exposure to complex tax issues) decrease the incidence of tax-related restatements, whereas those with industry tax expertise exacerbate such restatements. Nevertheless, both Christensen et al. (2015) and Goldman et al. (2021) show that the benefits of industry tax expertise or general tax task-specific knowledge are concentrated in firms that procure low levels of APTS from incumbent auditors, indicating a substitution effect.

2.2 APTS and the impairment of independence arguments

In contrast to the possible benefits generated from the knowledge spillover effects, auditor independence can be, or be perceived as, negatively influenced by the joint provision of audit services and NAS, including APTS. The provision of APTS could create several threats to auditor independence, such as self-interest and self-review threats. Such independence threats will affect the independence of mind, independence in appearance or both. DeFond and Zhang (2014) suggest that the impairment of auditor independence could be triggered by both demand- and supply-side factors.

The first independence threat stemming from APTS is economic bonding (i.e., self-interest). Although the economic bonds between auditors and their clients are inherent even if auditors do not provide any NAS (DeAngelo, 1981), the provision of NAS increases the client-specific current and future quasi-rents, which increase the bond with and, in turn, the fee reliance on audit clients (Francis, 2006; Simunic, 1984). From the demand side, Srinidhi and Gul (2007) document that it is easier for audit clients to influence their auditors by including excessive rents in NAS fees, which are less regulated than audit fees. Since APTS is still permitted under the current environment in many jurisdictions, the purchase of APTS could be a way, or maybe the most important way, that audit clients can influence auditor independence. Also, from the supply side, owing to the pressure on auditors' (e.g., audit offices or partners) to achieve performance targets, their incentives to acquiesce to, or even help clients with, the manipulation of earnings and/or the adoption of aggressive tax positions increases as APTS fees increase (Alsadoun et al., 2018; Causholli et al., 2014; Doty, 2011; Favere-Marchesi, 2006). Therefore, audit clients that acquire both audit services and APTS gain more opportunities to conduct opportunistic behaviour than those purchasing audit services only, or other types of NAS, from their auditors. Besides the self-interest threat generated from current quasi-rents, Causholli et al. (2014) argue that ‘a client's promise of future NAS business has potential to impair an auditor's independence’ (p. 681). Supporting this argument, Lynch et al. (2021) find that auditors will receive about 17% more (5% fewer) APTS fees in the following year from their audit clients if they stop (start or continue) issuing tax-related key audit matters (KAM) to the clients in the current year. In sum, both current and future quasi-rents related to APTS fees can create enhanced economic bonding between auditors and audit clients, thereby threatening auditor independence.

The second independence threats stemming from APTS are the self-review concerns. Francis (2006) documents that the provision of NAS may change the auditor's role from outside independent reviewer to inside adviser and decision-maker. Since the provision of APTS is more closely and directly related to clients' income and cash flow (e.g., Maydew & Shackelford, 2007; Omer et al., 2006), auditors may end up auditing their own work. In such cases, auditors are less (more) likely to challenge (rely on) the clients' treatment of complicated tax issues and tax strategies that are advised by ‘auditors themselves’ (i.e., the tax services team in the same audit firm/office).

2.3 Section summary

In this section, we discussed two theoretical perspectives on APTS. The knowledge spillover effect suggests that the provision of APTS, especially the recurring ones, helps auditors to have a better understanding of clients' operations, internal control, tax strategies and of managers' attitudes towards financial reporting. Such understanding facilitates high-quality audits. The impaired independence effect suggests that the provision of APTS increases economic bonds between auditors and clients, thus decreasing their incentives to provide high-quality audits. The net effects of APTS, therefore, have been subject to a substantial amount of academic research over the years.

3.1 United States

The SEC first issued Accounting Series Release (ASR) No. 250 in June 1978, which required firms to disclose the specific nature of NAS and total NAS as a percentage of total audit fees in their proxy statements (SEC, 1978). However, the SEC rescinded this requirement in August 1981. Parkash and Venable (1993) find that 90% of the Fortune 500 firm-year observations purchased APTS from 1978 to 1980. They also document that the demand for APTS varied with agency costs and auditor characteristics. Koh et al. (2013) use a similar dataset but do not find any evidence that the APTS fees ratio is related to impaired auditor independence. Subsequently, the SEC mandated the detailed disclosure of fees paid to registrants' auditors since 2001 to assist investors in evaluating the impact of the joint provision of audit and NAS on auditor independence. In particular, the SEC registrants must disclose their fees paid to auditors in three categories: (1) audit services fees, (2) financial information systems design and implementation (FISDI) services fees and (3) all other NAS fees (SEC, 2000). In addition, the SEC (2000) restricted certain types of NAS that would impair auditor independence when being provided by registrants' auditors, with some exceptions.

Following some accounting scandals, Section 201(a) of the Sarbanes-Oxley Act of 2002 (SOX) (hereafter, SOX, 2002) imposed stricter rules banning audit firms from simultaneously providing audit and FISDI services or some certain services classified in ‘other NAS’ (U.S. House of Representatives, 2002).⁶ However, SOX did not prohibit the provision of APTS, a category of NAS that auditors frequently provide to their clients (SEC, 2003). However, before acquiring APTS and other allowed NAS from their auditors, registrants must get pre-approval from their audit committees to assure stakeholders that providing such services will not impair the auditors' independence (Section 202 of SOX).

After SOX, the SEC (2003) amended the disclosure requirements for auditor fees. The new rules required registrants to disclose auditor fees in four separate categories: (1) audit fees, (2) audit-related fees, (3) tax fees and (4) all other fees. In regard to the addition of a ‘tax fees’ category, SEC stated that the provision of tax services needs extensive knowledge about the client, and the APTS fees are considerable in relation to other NAS fees. Therefore, SEC believes that it is appropriate and beneficial to investors if firms distinguish APTS fees from all other fee categories. The SEC (2003) rules became mandatory for years ending after 15 December 2003 and required registrants to provide such information for each of the two most recent fiscal years. As a result, registrants started to disclose the four categories of auditor fees information from 2002. Moreover, the SEC reiterated that audit firms could provide APTS (i.e., tax compliance, tax planning and tax advice) to their clients. However, some prior studies find that firms started to voluntarily dismiss, or substantially reduce, the purchase of APTS owing to concerns about the perception of impaired independence (e.g., Omer et al., 2006): a perception that imposed substantial costs on firms such as reduced tax savings (Cook et al., 2020) and low value relevance of earnings (Krishnan et al., 2013).

In 2005, the PCAOB further adopted three new rules prohibiting certain tax consulting services that are perceived to impede auditor independence. More precisely, Rule 3521 proscribed auditors from providing APTS with contingent fee arrangements, Rule 3522 prohibited auditors from promoting and providing clients with APTS for achieving aggressive tax positions to avoid tax and Rule 3523 banned auditors from providing any APTS to a person who has a role in financial reporting oversight (e.g., executives) at the client's firm or to immediate family members of such persons. On 19 April 2006, SEC (2006) approved these three PCAOB (2005) proposed rules, effective from 31 October 2006, and onward.

For new audit clients, PCAOB (2008) amended Rule 3523 to remove the restriction for tax services provided during the portion of the audit period that is completed before the beginning of the professional engagement period, since tax services provided in such a period are perceived not to impair auditor independence. The amendment of Rule 3523 would be effective immediately upon the SEC approval, which occurred on 22 August 2008 (SEC, 2008). Markelevich and Rosner (2013) find in their additional tests that APTS fees were positively associated with the likelihood of issuing fraudulent financial statements from 2000 to 2005. Also, Thornton and Shaub (2014) show that US jurors perceive a significantly lower audit quality when auditors provide audit clients with aggressive tax planning services, as compared with tax compliance services. These results support the new restrictions. Finley and Stekelberg (2016) document a significant decrease in tax avoidance for firms continuing to purchase APTS from Big 4 auditors from pre-2005 to post-2005 periods. However, Lennox (2016) examines the effects of these three new restrictions on audit quality and fails to find an increased audit quality, as proposed by the US regulators, after implementing new restrictions. Also, Nesbitt et al. (2020) show that despite complying with Rule 3522, audit firms have been advising larger clients on the use of more aggressive tax strategies increasingly over time. Moreover, some audit firms continued to provide some tax services that are not explicitly mentioned in the regulation but compromise auditor independence. For instance, KPMG was penalised 8.2 million US dollars by the SEC on 24 January 2014, for loaning tax professionals to audit clients from 2007 to 2011: a practice that violates the rule proscribing auditors from acting as an employee of clients. As a result, the SEC (2014) stated that auditors must carefully assess the independence threats of providing certain types of NAS, rather than just consider whether the proposed services fall within one of the permissible categories (e.g., tax services).

Overall, the regulatory environment of APTS in the United States consists of the following: (1) all APTS should be pre-approved by the clients' audit committee, (2) clients need to disclose APTS fees in their proxy statements separately, (3) auditors can provide only those APTS that would not jeopardise their independence and (4) SOX violations are subject to expanded criminal and civil liabilities and penalties.

3.2 EU

To make the relationships between auditors and audit clients more transparent, the EU Directive (2006) suggested and required disclosure of audit fees and the fees paid for NAS, including other assurance services, tax advisory services and other NAS in the notes of financial statements in all EU member states: requirements that are comparable with those of the US SEC (2003). In regard to auditor independence, the Directive also mentioned that auditors should not undertake any additional NAS that compromises their independence.

After the financial crisis around the world, regulators in the EU started to consider amending the audit regulations, including the prohibition of NAS, based on the presumption that NAS may compromise auditor independence as mentioned in the EU Directive (2006). According to the European Commission (EC, 2010) ‘Green Paper’, although there was no mandatory ban of NAS provision along with statutory audits in EU-wide regulations, some member states already implemented the EU Directive (2006), albeit in a very divergent manner. For instance, auditors are proscribed from providing any NAS to their audit clients in France, whereas the restrictions on NAS provision are relatively loose in other member states.

In the following year, EC (2011), Article 10, proposed a specific requirement prohibiting auditor-provided NAS in the public interest entities (PIEs). Specifically, the EC (2011) suggested banning or restricting auditors from providing services other than statutory audit services and related financial audit services to PIE audit clients. The scope of ‘related financial audit services’ includes (a) the audit or review of interim financial statements, (b) providing assurance on other statements (e.g., corporate governance, corporate social responsibility matters and regulatory reporting), (c) providing certification on compliance with tax requirements, where such attestation is required by national law and (d) other duty related to audit work imposed by the EU legislation on the statutory auditor. To enhance auditor independence, the EC (2010, 2011) even proposed banning auditors from providing any NAS to PIE clients and suggested that such prohibition might contribute to the establishment of ‘pure audit firms’, in which auditors could provide independent opinions without any business interest in their audit clients.

In 2014, Regulation (EU) No. 537/2014 amended and approved the EC (2011) proposal described above. The EU Regulation (2014) introduced a ‘blacklist’ of prohibited NAS across all EU member states. The first banned NAS in EU Regulation (2014), Article 5, is the APTS related to (i) preparation of tax forms; (ii) payroll tax; (iii) customs duties; (iv) identification of public subsidies and tax incentives; (v) support regarding tax inspections by tax authorities; (vi) calculation of direct and indirect tax and deferred tax and (vii) provision of tax advice. Other NAS included in the ‘blacklist’ are similar to those services that are banned in the United States (see Section 3.1 above). In addition, member states may further prohibit any NAS that they consider representing a threat to auditor independence. The EU Regulation (2014) also required a pre-approval from clients' audit committees if incumbent auditors want to provide some permissible NAS that is not part of the ‘blacklist’.

In addition, a fee cap was introduced on the total amount of fees to be charged for allowable NAS, assuming that the provision of NAS negatively affects auditor independence only when a certain threshold is exceeded. In particular, the total fees for those allowable NAS shall not exceed 70% of the average of the statutory audit fees in the last three consecutive fiscal years. Also, member states' regulators could opt to set stricter rules on the NAS fee cap. These new regulations were effective from 17 June 2016 onwards, and regulators expected to observe increased auditor independence after implementing those restrictions. Ratzinger-Sakel and Schönberger (2015) demonstrate that, although the EU Regulation has different impacts on the provision of NAS in different member states, the EU-wide regulation is an extension of existing restrictions in each member state. However, the new regulation seems unnecessary, considering the empirical findings in the EU, which show little impairment of independence due to APTS, using data from the pre-2016 regime (Castillo-Merino et al., 2020; Eilifsen et al., 2018; Garcia-Blandon et al., 2017, 2021; Watrin et al., 2019).

Notably, the EU Regulation (2014) provided a derogation option to member states when implementing these NAS regulations in their local legislations. That is, member states could opt to allow auditors to provide some valuation and certain tax services (i, iv, v, vi and vii contained in Article 5) that are included in the ‘blacklist’ if certain criteria are met.⁷ Therefore, EU regulators view some categories of APTS as less harmful to auditor independence than other NAS categories. According to the Audit Analytics (2020, December), most member states opted to allow the aforementioned tax services, and only eight member states decided not to use this option. In other words, APTS are totally banned in these eight member states (i.e., Croatia, Greece, France, Italy, Netherlands, Poland, Portugal and Slovenia). Regarding the fee cap on NAS, all member states have opted for the 70% except for Portugal where a 30% cap has opted. It is worth noting that there is no specific fee cap for APTS. Moreover, recent studies show that the mean or median APTS fees ratios are well below 20% or 30% in some EU countries, such as Spain (Castillo-Merino et al., 2020) and Germany (Eilifsen et al., 2018; Watrin et al., 2019). Although there are some firms with very high APTS fees, this fee cap regulation may have less or no influence on the purchase or provision of APTS for most of the listed firms in the EU.

The EU Regulation (2014) specified some tax services that should be banned without exemption, namely, (ii) and (iii) contained in Article 5, which seems to be different from the US regulation. However, although these services fall within the tax services category, providing these services would make auditors play either an accounting role or a management function role for the audit clients, posing serious threats to auditor independence. Therefore, this rule is, in general, consistent with the US SEC (2014) argument that auditors should assess both the nature and the manner of delivering some types of services to make sure auditor independence is not compromised, although EU Regulation (2014) is less ambiguous than the US regulation.

3.3 Other jurisdictions

Besides the regulatory environments in the United States and EU, which have received much attention in the previous literature, we also provide a brief review of regulations related to APTS in some other jurisdictions, based on the importance of their capital markets in the world economy. As the second-largest capital market, China mandated the disclosure of NAS fees in 2001, but there is no specific regulation related to APTS thus far. This seems reasonable because the Chinese NAS market is small, and most firms do not purchase NAS from their incumbent auditors (Chen et al., 2010; Fang et al., 2014). Thus, the provision of NAS, including APTS, is not considered to threaten auditor independence. Hong Kong has not mandated the disclosure of NAS fees yet, and therefore, it is difficult to determine whether the provision of such services is valued favourably by the stakeholders or not. However, Law (2010), using a survey among some Hong Kong auditors and financial analysts, finds that APTS are perceived as a value-added service by their clients by comparison with other types of NAS such as corporate finance services and internal audit services. Japanese auditors are banned from providing certain types of NAS, including any type of APTS. The Companies Act of 2013 in India prohibited certain types of NAS but did not prohibit APTS. If Indian listed firms purchase APTS from their incumbent auditors, they need to disclose the fees paid under ‘Payments to the auditors for taxation matters’ in their financial statements.

Although Canadian prohibitions of NAS are very similar to the US ones, there are some differences. The CPA Canada Independence Working Group (IWG) initiated a discussion in 2012 about whether to further prohibit certain types of APTS as the US SEC did in 2006 (IWG, 2012). In the following year, IWG (2013) concluded that it would support additional prohibitions of APTS with respect to personal tax services for individuals who hold financial reporting oversight roles (similar to US Rule 3523) and aggressive and confidential tax transactions (similar to US Rule 3522). IWG (2013) called for more studies on the impacts of the provision of APTS on a contingent basis (similar to US Rule 3521). So far, these suggestions have not been enforced in Canada. Moreover, CPA Canada (2016) clearly states that if auditors are already providing services related to ‘Tax calculations for the purpose of preparing accounting entries, except under certain circumstances in emergency situations [Rule 204.4 (34) (b)]’ for a listed firm, then they shall not perform audit services for that same firm. Makni et al. (2020) published the only study using Canadian data to explore the impacts of APTS and find weak evidence that the ratio of APTS fees to total auditor fees is positively associated with the firms' use of tax havens.

The Certified Public Accountant Act in South Korea mandated the disclosure of NAS fees and their components in 2001 and prohibited certain types of NAS in 2002, although such prohibitions were less restrictive compared with the United States and EU (Kang et al., 2019). The Korean National Assembly amended the Korean CPA Act on 18 February 2016 and banned additional NAS, making Korean law similar to the US SOX (2002). However, like India, South Korea has not banned any APTS because the Korean Institute of Certified Public Accountants (KICPA, 2006) suggests that the provision of APTS is unlikely to impair auditor independence. Empirical research supports this view (Choi et al., 2009). Taken together, our review of the APTS regulations reveals significant variations across jurisdictions. Table 2 provides an overview of APTS-related regulations in different jurisdictions.

4.1 Voluntary disclosure of APTS information

As discussed in Section 3, publicly listed firms were required to disclose the amount of APTS fees from 2003 in the United States (from 2006 in the EU countries), although some firms voluntarily disclosed the APTS information prior to the passage of these regulations. Omer et al. (2006) examine the factors associated with firms' decision to purchase APTS from 2000 to 2002. To control for potential selection bias resulting from purchase and disclosure decisions, they model the factors influencing such decisions in the first stage of the Heckman two-step estimation process. There are two findings from Omer et al.'s (2006) selection bias regressions. First, the decision to disclose APTS fees is positively associated with tax complexity, auditor tenure and auditor change and negatively associated with the proportion of NAS fees to total fees, thereby supporting the notion that firms tend to resist disclosing APTS fees information to reduce political costs associated with heightened regulatory scrutiny. Second, their results imply that firm size and auditor size jointly affect a client's decision to purchase APTS but not the decision to disclose APTS fees. Bedard et al. (2010) extend Omer et al. (2006) by including two important players in the corporate governance structure of a firm (i.e., the audit committee and the auditor) in explaining firm's voluntary APTS fees disclosure decisions. They confirm the negative association between APTS fees disclosure and the proportion of NAS fees to total fees. Importantly, the negative association is found to be more pronounced when firms have a strong audit committee: a finding inconsistent with the audit committee oversight theory. A stronger audit committee may perceive costs resulting from regulatory scrutiny as outweighing the additional benefits of disclosures to investors. Also, this negative association is stronger for firms audited by non-Big 4 auditors, which is consistent with the findings that clients of non-Big 4 auditors tend to disclose less information (e.g., Clarkson et al., 2003).

4.2 Selecting incumbent auditors as tax service providers

It is worthy of note that publicly listed firms are required to disclose only the tax services fees paid to their incumbent auditors but not if they purchase tax services from other providers or if their internal tax departments perform those services. Therefore, we could not find any empirical paper using publicly available data to examine the entire spectrum of tax services provider choices in publicly listed firms. However, using proprietary data from the US Internal Revenue Service (IRS) and focusing on tax compliance services only, Klassen et al. (2016) find that firms are more likely to hire incumbent auditors to prepare tax returns when they have been less tax aggressive in the past, higher-growing and smaller. Furthermore, firms having fewer foreign activities, incurring losses, engaging in R&D activities and exhibiting high other NAS fees ratios also tend to select incumbent auditors to prepare tax returns. Neuman et al. (2015) provide additional evidence using data from the US not-for-profit sector (NFPs), where the decision to choose a tax service provider is observable for all entities. They find that the NFPs are less likely to purchase tax services from the incumbent auditor as the distance between the firm and the auditor increases, indicating that tax services, similar to financial statement audit services, also require a high degree of client contact. Moreover, Neuman et al. (2015) posit and find that clients are more likely to hire other audit firms or law firms when the set of substitute service providers is greater and, thus, the purchase of tailored services is facilitated.

4.3 Retaining or dismissing incumbent auditors as tax service providers

Instead of examining the determinants of purchasing APTS directly, another stream of studies provides some evidence about the reasons audit clients decide to dismiss or retain their incumbent auditors as their tax services providers (Ahn et al., 2021; Albring et al., 2014; Finley & Stekelberg, 2016; Lassila et al., 2010). Because SOX (2002) and SEC (2000, 2003, 2006) prohibited certain types of NAS and required more granular classifications of fees disclosure, public firms substantially reduced or terminated the purchase of some NAS, including tax services, from their auditors, to reduce negative reactions from investors and regulators (Abbott et al., 2011; Finley & Stekelberg, 2016; Lennox, 2016; Maydew & Shackelford, 2007), even when the regulations clearly suggested that inv+estors would view APTS more favourably than other types of NAS (SEC, 2003). Omer et al. (2006) show that the sample median APTS fees declined from US$256,880 in 2000 to US$145,150 in 2002, and Cook et al. (2020) further find that approximately 21% of the firms in their sample period (i.e., 2002 to 2005) eliminated or significantly reduced the use of APTS. Among these firms, the median APTS fees dropped from US$326,446 to US$53,000. Therefore, firms' decisions to retain or dismiss incumbent auditors as tax services providers became an important research question.

Lassila et al. (2010) mention that ‘retaining or dismissing’ decisions appear to be a trade-off between the benefits generated from knowledge spillover and the costs related to impaired auditor independence of mind or in appearance. If benefits exceed costs, firms will retain or hire their auditors as tax services providers and, otherwise, dismiss them. Therefore, those factors that can either increase the benefits of knowledge spillover or decrease the costs of impaired auditor independence will contribute to the hiring and retaining decisions. Lassila et al. (2010) find a positive association between clients' corporate governance and the likelihood of retaining auditors as tax services providers in the period surrounding the SOX (i.e., 2001 to 2003). The association is consistent with the argument that strong corporate governance increases auditor oversight and decreases the likelihood of impaired auditor independence (e.g., Carcello & Neal, 2003). However, using a sample of the matched switch and nonswitch firms in the post-SOX period (i.e., 2003 to 2006), Albring et al. (2014) find a negative association between corporate governance attributes and the decision to retain auditors as tax services providers. Likewise, both Almaqoushi and Powell (2021) and Bédard and Paquette (2021) show that audit committee quality indices are negatively associated with the purchase and magnitude of APTS fees. These studies support the conservative behaviour of directors or audit committee members in response to the increased litigation and potential reputation risks after the passage of SOX. Furthermore, the ex-ante independence risk will increase the costs of retaining decisions. Both Ahn et al. (2021) and Lassila et al. (2010) posit and find evidence that auditors will be perceived as lacking independence if they provide high nontax NAS relative to audit fees to their audit clients and have a long relationship with them; consequently, the clients are less likely to retain incumbent auditors as tax services providers.

On the other hand, if some firm-specific characteristics increase the potential benefits through knowledge spillover effects, firms are more prone to retain the incumbent auditors as their tax service provider. Both Albring et al. (2014) and Lassila et al. (2010) find that firms are more likely to retain their auditors as tax services providers when they have high tax and operational complexity, supporting the notion that the complexity (e.g., the existence of foreign operations and the M&A activities) will generate more knowledge spillover benefits. Complexity also encourages firms to purchase more tax advice from incumbent auditors (Omer et al., 2006). Similarly, Bédard and Paquette (2021) find that the high ex-ante litigation risk of firms encourages audit committee members with accounting financial expertise to purchase more APTS. However, the moderating effects of firm litigation risk exist only in the period of 2007 to 2011 when the public scrutiny had faded away.

Finley and Stekelberg (2016) use a ‘natural experiment’ method to examine the impact on retaining decisions of external oversight imposed on tax service providers. Specifically, KPMG, one of the Big 4 firms, entered into a deferred prosecution agreement (DPA) with the US Justice Department in 2005 to resolve charges stemming from the sale of tax shelter services to its individual clients. KPMG agreed to comply permanently with several tax services rules that are stricter than those required of other tax services providers, which limited KMPG's ability to facilitate some tax avoidance strategies for clients.⁸ The authors investigate the reactions of firms that are currently purchasing APTS from KPMG in response to the potential damage to KPMG's reputation and/or the decreased expected tax benefits owing to the DPA. The results show that while all other Big 4 firms' provision of APTS declined in the post-DPA period, KPMG suffered more from the DPA than other firms, in terms of both the likelihood of purchasing APTS and the amount of purchased APTS. The authors also examine whether tax avoidance decreased significantly for the KPMG clients that decided to continue purchasing tax services from KPMG after the DPA, compared with clients of other Big 4 auditors. Surprisingly, KPMG's clients did not exhibit any difference in tax avoidance behaviour relative to clients of other Big 4 auditors. In addition, Baugh et al. (2019) find that the audit quality of KPMG did not change significantly relative to other Big 4 auditors after the DPA, even for KPMG clients that dropped KPMG as the tax service provider. Combining these two studies, we may conclude that the DPA had no or little effects on KPMG's audit and tax services quality.

Similarly, Ahn et al. (2021) examine the changes in APTS after the public disclosure of the 2007 PCAOB Part II inspection report, which identified Deloitte's deficiencies related to audits of income tax accounts. They find that, compared with other large audit firms inspected by the PCAOB, both the increased costs stemming from perceived impaired independence and the decreased benefits due to Deloitte's inability to use knowledge transferred from the tax team, prompted Deloitte's audit clients to dismiss or reduce the purchase of APTS from Deloitte. Overall, these results related to KPMG and Deloitte are supportive of Omer et al.'s (2006) finding that clients terminated or significantly reduced APTS investment around the uncertain external oversight environment.

4.4 The magnitude of APTS fees

Halperin and Lai (2015) develop a tax fee model from the clients' demand side to examine the determinants of APTS fees. In addition to unexpected audit fees (Omer et al., 2006), they find that expected audit fees are also positively associated with APTS fees, indicating a cross-selling behaviour of auditors.⁹ While Chan et al. (2012) find results similar to those of Halperin and Lai (2015), they do not develop a specific model for APTS fees. Other factors found to be associated with APTS fees are firm characteristics, executive and board characteristics (e.g., tenure, compensation, publicity, interlocks and alumni) and auditor characteristics (e.g., size, tenure and switch) (e.g., Alexander & Hay, 2013; Duan et al., 2018; Naiker et al., 2013; Omer et al., 2012; Parkash & Venable, 1993; Shi et al., 2021). Parkash and Venable (1993) find that firms with fewer agency costs (i.e., firms having high managerial ownership, high outside ownership concentration and low leverage), and firms having industry specialist auditors, tend to pay more APTS fees. Duan et al. (2018) utilise Google search volume to measure CEOs' publicity and find that firms whose CEOs have high publicity tend to avoid tax to meet investors' performance expectations. Importantly, such firms pay more fees for tax planning services to their incumbent auditors. Naiker et al. (2013) show that audit committees with former audit firm partners are less likely to purchase more tax services from firms' incumbent auditors, probably because of concerns over auditor independence. Regarding audit committee networks, Shi et al. (2021) find that purchases of APTS increase with the prior year's average of APTS purchased by board-interlocked firms via an audit committee board member. Omer et al. (2012) develop a composite measure to differentiate a group of firms having a new economy business model that ‘allows firms to exploit their operational flexibility to align with tax incentives’ and consequently reduce their tax burdens (p. 33). Therefore, if they follow the new economy business model, firms have less demand for tax planning services than their counterparts following the traditional business model.

Two more recent studies use a change specification to examine the determinants of APTS fees (Kim et al., in press; Lynch et al., 2021). Using the changes in APTS fees as the dependent variable, Lynch et al. (2021) find that auditors receive more APTS fees in the next year when they stop issuing tax-related KAM in clients' audit reports in the current year. Kim et al. (in press) explore the macroeconomic determinants of APTS fees. They argue that the benefits from investment in tax planning activities increase with an increase in optimism about future economic growth because firms expect to generate more pretax income and cash flows as well as more tax planning opportunities. Consequently, firms are likely to invest more in purchasing tax planning services. The authors find supportive results using US macroeconomic forecasts of real gross domestic growth (GDP) as a proxy for expected macroeconomic conditions. Moreover, this positive association is found to be more pronounced for firms with financial constraints and high tax rate volatility.

4.5 Section summary

In this subsection, we summarised prior studies examining four types of APTS-related firm decisions. Table 3 summarises the research questions, samples used and key findings of these papers. Although the papers examined different research questions, we can conclude that all of these decisions are related to a cost–benefit trade-off. If the expected benefits exceed the expected costs, firms are more likely to disclose APTS information, purchase or retain APTS and pay more fees for APTS. We will discuss the costs and benefits (i.e., the consequences) of APTS in the next section. However, the extant literature does find significant variations in the purchase of APTS based on firm-level and auditor-level characteristics, so it is crucial to control and correct selection bias induced by such factors before investigating the consequences of APTS.

5.1 Output-based measures

5.2 Input-based factors

In Section 5.1, we have discussed the effects of APTS on output-based measures, which could be used to infer audit quality. DeFond and Zhang (2014) also find that audit clients could choose audit quality based on some observable inputs, such as auditor-specific characteristics (e.g., auditor size and industry specialisation) and audit-client contracting features (e.g., audit fees).

Extant APTS literature has examined the association between APTS and audit fees, but the results are inconclusive as well. Audit fees could measure the audit effort levels, given the intuition that more audit efforts increase audit quality. However, over and above the audit effort exerted by auditors, audit fees are also influenced by the unit cost of audit effort and the risk premium charged to cover future losses (Houston et al., 2005; Pratt & Stice, 1994; Simunic, 1980). Thus, audit fees could be a noisy measure of auditor quality in the context of APTS for at least two reasons. First, the knowledge spillover effects of providing APTS facilitate auditors' assessments of clients' internal control and tax-related or nontax-related accounts (e.g., De Simone et al., 2015), reducing the required audit effort and the high-quality pre-audit financial information decreases the auditors' concerns about future losses. Also, the economies of scope achieved by the simultaneous provision of NAS and audit services may reduce the unit cost of audit effort (Arruñada, 1999). These factors suggest that the provision of APTS would reduce audit fees without impairing audit quality. Second, audit fees could be increased either because the reduced unit cost of both audit and NAS encourages clients to buy more services from incumbent auditors (Simunic, 1984) or because of the cross-selling behaviour of auditors (Halperin & Lai, 2015). In these cases, the provision of APTS would increase audit fees without improving audit quality. As a result, we should interpret previous papers examining the relations between APTS and audit fees with caution.

Early studies using proprietary data, or data collected through surveys, generally find a positive association between audit fees and APTS fees in both the United States (e.g., Palmrose, 1986) and the United Kingdom (Ezzamel et al., 2002). Palmrose (1986) is the first study to investigate the effects of different types of NAS on audit fees. She decomposes total NAS into three categories: tax, accounting-related and nonaccounting services. Her results show a significantly positive relationship between audit fees and APTS fees, especially for small clients. Later, Davis et al. (1993) test and attribute this audit fee premium resulting from tax services to increased audit effort, proxied by audit hours. However, O'Keefe et al. (1994) fail to find any evidence that the APTS fees ratio is related to either audit hours or audit fees. From a cross-selling perspective, Halperin and Lai (2015) use simultaneous equations to test whether APTS fees and audit fees are jointly determined. Their results suggest that the demand for tax services gives rise to a higher demand for audit services and vice versa (cross-selling behaviour). Alexander and Hay (2013) and Klumpes et al. (2016) test the associations between APTS fees and audit fees in New Zealand and the United Kingdom, respectively. While Alexander and Hay (2013) find that APTS fees are positively associated with audit fees, Klumpes et al. (2016) find negative associations between these two types of fees in the UK life insurance industry. The former suggests that there is no issue of impaired independence, whereas the latter suggests the opposite. Fleischer and Goettsche (2012) also find a positive relation between these fees in the German context.

Instead of testing the associations between the provision of APTS and audit fees directly, the research could examine the moderating effects of APTS on the relationships between some variables and audit fees. Prior auditing and taxation studies show that auditors charge higher audit fees for clients with high levels of tax aggressiveness (Donohoe & Knechel, 2014) and with high tax risk (Abernathy et al., 2021). However, both studies document that such risk premiums charged by auditors are significantly lower when audit clients also purchase a material level of APTS. Consistent with our arguments suggesting that the provision of APTS may reduce audit fees without impairing audit quality, both studies attribute the reduced audit fees to synergies between the tax and audit teams. Donohoe and Knechel (2014) also find that auditors' tax or overall industry expertise could provide high-quality audits to clients regardless of whether clients hire their incumbent auditors for tax services, which increase the audit fees, especially for tax aggressive clients. Surprisingly, Abernathy et al. (2021) show that the provision of APTS reduces approximately 48% to 68% of the audit fee premiums associated with different tax risk measures, which could be a huge benefit of hiring incumbent auditors as tax service providers (Table 5E).

5.3 Section summary

In this subsection, we reviewed studies related to the consequences of APTS. We categorised these studies into five themes: (1) misstatement and auditor communication, (2) earnings quality (both tax and nontax related), (3) tax avoidance and tax reserves, (4) market perceptions of APTS and (5) APTS and input-based audit quality proxy. In general, the APTS literature in the United States supports the knowledge spillover effect with some exceptions. However, the findings in the non-US settings are mixed, a finding that may be attributed to diverse institutional factors in different countries. The market perceptions of APTS in both the US and non-US settings seem to suggest that market participants value APTS in uncertain periods negatively and value APTS in more stable periods positively. Nonarchival studies suggest that the perceptions of APTS vary with stakeholder groups and with the type of APTS provided.

6.1 Measurement of APTS

There remain wide variations in the measurement of APTS in the existing literature. Table 6 provides an overview of APTS measures used in the surveyed papers. It is evident that APTS measures vary significantly across studies, which is not surprising given that different measures capture different aspects of APTS pertinent to different research questions. For instance, the APTS ratios either consider the importance of APTS fees to each audit client (e.g., APTS fees/total fees) or control for client size (e.g., APTS fees/total assets). The absolute value of APTS fees, on the other hand, captures the magnitude of fees auditors received from their audit clients. Conceptually, both the high levels of relative and absolute APTS fees indicate a high degree of involvement in the client's tax-related work and an increased economic bond between the incumbent auditors and their clients. It remains unclear which measure has the largest explanatory power in a specific research setting. Future studies could examine several proxies for APTS to assess the incremental explanatory power of using one measure over the others.

6.2 Replication of previous studies

Over the years, the regulators from different jurisdictions implemented a multitude of regulations and restrictions on the provision of NAS and APTS to regain investor confidence: confidence that rock bottomed because of large-scale corporate frauds. Prior studies have shown that the behaviour of key stakeholders substantially changed during those uncertain times (Cook et al., 2020; Eilifsen et al., 2018; Maydew & Shackelford, 2007; Omer et al., 2006). Several prior studies reviewed in this paper explored the effects of SOX (2002) on certain research questions in the United States (e.g., Cook et al., 2008; Krishnan & Visvanathan, 2011). However, there is a lack of research investigating the impacts of other worldwide events on APTS-related questions. First, the GFC has had extensive and major consequences for worldwide economics. Both archival and nonarchival findings related to stakeholders' perceptions of APTS may have changed after the GFC period. Therefore, some earlier findings might no longer be relevant to standard setters or academic researchers. Indeed, Eilifsen et al. (2018) show that the German investors' perceptions of APTS changed significantly during the GFC period. Thus, a revisit to some prior research questions using updated data is warranted.

Second, the EU audit and nonaudit markets have experienced significant reforms, owing to the enactment and implementation of EU Regulation (2014), which became effective on 17 June 2016. We believe that this provides a rich research setting to compare the effects of APTS on certain research questions, such as tax avoidance, in both the preregulation and postregulation periods, thereby informing regulators about the effectiveness of the new regulations. Our review of nonarchival literature also reveals that there is only one study that examined the stakeholders' perceptions of APTS in the postregulation period (van Liempd et al., 2019). It is reasonable to expect that stakeholder perceptions might have changed because of recent reform in the EU. Both archival and nonarchival evidence are needed in this area.

Third, similar to the GFC, the outbreak of COVID-19 has affected global economies significantly. Since 2020, academic researchers started to explore the consequences of COVID-19 extensively. We have not been aware of any COVID-related papers related to APTS. As discussed in Section 5, one benefit of purchasing APTS is to avoid tax. While the pandemic lasts, firms that suffer from financial constraints may choose to use APTS to avoid more tax as a survival mechanism. However, curtailed demand in many sectors also means that firms may have to scale back their investments, leading to a decreased demand for APTS. Furthermore, the effects of COVID-19 vary with industries (Qin et al., 2020). Firms from certain industries (e.g., tourism, restaurant and accommodation) are facing more severe challenges than those from others. The relations between the demand for APTS, COVID-19 and industry groups are unclear and worthy of investigation. Another interesting topic is the relation between the provision of APTS and the cost of debt capital during the COVID-19 period. Audit Analytics (2021) illustrates that the total debt offerings in the United States surged in 2020. Given the negative association between the cost of debt and the levels of APTS fees found by Fortin and Pittman (2008), it would be meaningful to examine whether such an association still exists in the US or in other non-US settings.

6.3 APTS research in other jurisdictions

In our review of the APTS literature, we find that more than half of the surveyed papers are from the United States, followed by some studies from the EU. However, the APTS-related research in other jurisdictions is relatively scarce, which makes our review more US- and EU-centric. As reviewed in Section 3, the regulations pertaining to APTS vary across jurisdictions. Although prior research has examined the effects of APTS in certain non-US jurisdictions, it is surprising to find little research on APTS in Canada and India, for example, where APTS-related data are readily available. After summarising research findings related to APTS in this paper, we call for more research employing both archival and nonarchival designs from other jurisdictions to enrich our understanding of various facets of APTS. Empirical research suggests that investors and other stakeholders do incorporate APTS-related information into their decision-making processes. We also hope that our review will provide useful information to jurisdictions that have not mandated the disclosure of APTS information (e.g., China and Hong Kong). Mandatory disclosure of APTS-related information could help various stakeholders understand and assess auditor independence more precisely.

We also encourage more research in an international (i.e., cross-country) setting to better understand the role of institutional settings across different jurisdictions in influencing firms' decision to purchase APTS from the incumbent auditors and its associated consequences. Habib et al. (2019) document that institutional factors such as political connections, investor protection and national culture determine auditor choice decisions. It is plausible to contend that such factors might also affect both the demand for and supply of APTS. Regarding the consequences of APTS, some prior studies have shown conflicting findings for the same research question in different countries. For example, APTS fees are found to be positively associated with tax avoidance in the United States (e.g., Hogan & Noga, 2015; Omer et al., 2006), whereas they are associated negatively or insignificantly in other countries (Garcia-Blandon et al., 2021; Watrin et al., 2019). Some prior studies attribute the different findings found in their papers to the different institutional environments across jurisdictions, such as book-tax conformity (Aschauer & Quick, 2018), auditor liability (Watrin et al., 2019), enforcement level (Castillo-Merino et al., 2020) and investor protection (van Liempd et al., 2019). To the best of our knowledge, there is no research using an international dataset to examine the determinants and consequences of APTS.

Given that APTS-related regulations vary across countries, it may not be productive to explore the effects of APTS regulation using a global dataset. However, the EU is a good setting for this line of research because all EU member states follow similar audit regulation with little variation (e.g., EU Regulation, 2014). Although European firms have similar financial reporting standards (i.e., IFRS), some variations in the regulatory environment do exist within Europe. For example, EU member states have different legal origins (i.e., code or common law), financial secrecy (i.e., high or low transparency) and tax filing rules (i.e., independent or IFRS-based rules). These differences could affect both the determinants and consequences of APTS: a fruitful avenue for future research in the EU setting.

6.4 Tax services fees and its components

Finally, we discuss research opportunities related to the disclosure requirements of tax services fees rather than APTS fees discussed in Section 6.3. A firm could choose to purchase tax services from several tax service providers (e.g., in-house, third party, or incumbent auditor). Under the current regulatory environment, firms need to disclose only the tax services fees paid to their incumbent auditors but not those paid to other providers. Therefore, extant studies could investigate only the APTS supplied by incumbent auditors without simultaneously considering the role played by other tax service providers. One exception is Chen et al. (2021), who hand-collected data on firms' in-house tax departments and find that firms' in-house tax investments and APTS work as substitutes. Such evidence suggests that ignoring the effects of other types of tax services provider may significantly bias research findings, especially when APTS fees decline but the total tax service revenue for audit firms increases or remains steady (Maydew & Shackelford, 2007). For example, McGuire et al. (2012) measure auditors' tax expertise using data on the market share of tax services fees charged by the incumbent auditor, rather than total fees charged by both the incumbent auditor and other tax service providers. This problem can be mitigated if regulation requires firms to disclose tax service fees paid to all providers. Such a mandatory disclosure requirement can also help future studies enrich the literature on the consequences of dismissing or switching tax services providers. For instance, how did the tax service fees change after firms' decisions to switch from incumbent auditors to others as tax service providers?

Regulators could also consider mandating the disclosure of APTS fees components, that is, tax compliance and tax planning fees. As discussed in Section 1, tax planning services are more damaging than tax compliance services. Also, prior studies show, both theoretically and empirically, that tax planning and tax compliance components of total APTS fees have differential effects on tax outcomes (e.g., Chyz et al., 2021; Klassen et al., 2016). Some prior studies use total APTS fees as the measure of tax planning (e.g., Cook et al., 2008; Francis et al., 2019; Kim et al., in press; Kubick et al., 2020). However, using hand-collected data, Chyz et al. (2021) show that total APTS fees normally represent significant fees paid to both tax compliance and tax planning services. Although Kim et al. (in press) use a change specification to mitigate this concern since tax compliance fees are likely to be relatively constant across years, mandatory disclosure of APTS fee components would enrich APTS research. It would be helpful in tax planning and tax avoidance research as well. A finer disclosure of APTS fees would help academic researchers use archival data to confirm some nonarchival evidence suggesting that perceptions of APTS are dependent on the type of tax service provided (e.g., Thornton & Shaub, 2014; van Liempd et al., 2019).