Auditors’ self-assessment of engagement quality and the role of stakeholder priority

Abstract

This study investigates auditors’ assessment of the quality of their own audit engagements, utilising survey data gathered from a Big Four audit firm in Sweden. We first examine to what extent auditors’ self-reported audit quality threatening behaviours (AQTBs) in the audit process are reflected in their assessment of overall audit quality (OAQ). The results indicate that AQTBs overall and all individual AQTBs are associated with quality assessment, though with variations in their significances. Second, we examine whether AQTBs and OAQ are associated with an auditor’s stakeholder priority, i.e. which stakeholder the auditor considers as her highest priority in the audit work. We find that auditors who consider the employer as the highest priority report more AQTBs. However, priorities are not related to OAQ. Furthermore, auditors prioritising the client or employer tend to assess the overall audit quality as being higher than what the AQTBs would suggest (i.e. they over-assess the quality). Interestingly, the findings regarding priorities are only evident among partners. In sum, the findings of this study provide important insights on how auditors themselves assess their audit quality, and on the role of auditors’ stakeholder priorities.

Keywords:

• audit quality
• self-assessment
• auditor behaviour
• audit process
• stakeholder priority

1. Introduction

Investigating auditors’ views on audit quality is important because auditors are the ones who conduct the audits and have a more holistic view of the audit process compared to other stakeholders of audit. More importantly, how an auditor understands audit quality is very likely to have a major influence on her behaviours in the audit process, and those behaviours will determine the audit outcome (Brivot et al. 2018). As pointed out by Likierman in the Brydon report (Brydon 2019, p. 124), ‘In audit, as in other professional activities, the quality of what is understood is critical to developing a professional judgement’. However, prior research on the self-assessment of audit engagement quality by auditors, as inside experts, is scarce.¹ This study aims to investigate auditors’ understanding of audit quality and behaviours in the audit process using auditors’ self-assessments of their own specific engagements.²

When considering auditors’ understanding of audit quality, we must recognise that an auditor is subject to the interests and expectations of various audit stakeholders, such as regulators, users of financial statements, client companies and audit firms. Although the stakeholders may share some common expectations and interests (Knechel et al. 2013a), there are unavoidable conflicts of interests – some accentuating professional and some commercial aspects of auditing (e.g. Sweeney and McGarry 2011; Cooper and Robson 2006; Suddaby et al. 2007; McNair 1991). The auditor’s professional viewpoint is developed via education, training and regulation (Li et al. 2017; Che et al. 2018; DeFond and Lennox, 2011), and it emphasises the auditing standards. The auditor’s commercial viewpoint emphasises efficiency and cost savings as audit firms are profit-making organisations (e.g. Broberg et al. 2018; McNair 1991). The auditor may also be receptive to the pressures of the client through their communication and negotiation (Salterio et al. 2012). Consequently, an individual auditor is surrounded by the inherent dilemma of conflicting interests (McNair 1991) that could shape her understanding of audit quality (Brivot et al. 2018).³

In our first research question (RQ1), we examine whether an auditor’s assessment of overall audit quality (OAQ) of a specific audit engagement is associated with her reported audit quality threatening behaviours (AQTBs) during the audit process.⁴ AQTBs are well documented in the audit literature,⁵ and they are all expected to negatively impact audit quality (e.g. Herda et al. 2019; Sweeney and Pierce, 2015; Sweeney et al. 2013).⁶ Finding highly significant associations between OAQ and AQTBs would support the argument that auditors’ quality assessments reflect key behaviours in the audit process that influence the sufficiency and appropriateness of audit evidence on which the auditor bases her conclusions, as highlighted by regulators in the auditing standards (see ISA 320, 330, 500, 505, 530, 700).⁷ However, since auditors essentially follow their own understanding of what high-quality audit work is, this understanding may be influenced by commercial views, and thus, auditors may or may not consider AQTBs very relevant for audit quality. It is unknown to what extent auditors’ views on audit quality correspond with requirements, assumptions and general quality frameworks presented by regulators (IAASB 2013; ISA ISQC 1).

In the second research question (RQ2), we examine whether AQTBs and OAQ are determined by auditors’ stakeholder priority (i.e. whether the auditor prioritises shareholders, regulators, society, the employer or client companies). Whether an auditor’s view on auditing is more professional or more commercial can manifest in which stakeholder she prioritises in her audit work. We expect that auditors’ stakeholder prioritisation influences both the reported behaviours in the audit process (i.e. AQTBs) and overall engagement quality assessment (i.e. OAQ). Finally, in the third research question (RQ3), we examine whether the stakeholder prioritisation influences how aligned an auditor’s assessment of OAQ is with the reported AQTBs.

We utilise proprietary survey data conducted in a Swedish Big 4 audit firm. The survey-based measures represent direct considerations of audit engagements by audit team members. The survey questionnaire was sent to 776 auditors in 217 audit engagement teams. The respondents are from different ranks (e.g. partners, managers and associates). We received responses from 335 auditors in 185 teams. Hence, the response rate is 85% (=185/217) at the team level and 43% (=335/776) at the auditor level. The final sample used in the main analyses consists of 251 auditors from 101 teams since we require each team to have at least two responding auditors to allow for within-team analyses and to use team fixed effects.

To examine RQ1, we regress OAQ on the AQTBs. The results indicate that auditors’ assessment of OAQ is significantly related to the mean of all AQTBs. This finding holds regardless of the auditor rank (i.e. partner, manager or associate). Moreover, OAQ is associated with all of the eight AQTBs, but with variations in terms of the level of significance. This suggests that auditors account for many of the factors influencing the sufficient appropriate audit evidence, emphasised in the auditing standards. Specifically, the two most strongly associated AQTBs are ‘The extent audit team members sign off an audit-program step without completing the work or noting the omission’ and ‘The extent audit team members reduce the amount of work performed on an audit step below a reasonable level’.

However, the AQTB ‘The extent audit team members have a greater than appropriate reliance on client work’ is only marginally significantly associated with OAQ. This AQTB is restricted by the ISAs as it deteriorates the appropriateness of audit evidence,⁸ but is still not considered very relevant for audit quality by the respondents. It might be that, for auditors, the threshold to conclude that reliance on client work has affected audit quality is quite high, since in auditing there is ‘fundamental dependence on information supplied by the auditee’ (Jeppesen 1998, p. 530). Also, the AQTB ‘The extent audit team members under-report audit time’ is only marginally significantly related to OAQ, indicating that auditors consider it less relevant for audit quality, although some research has linked under-reporting to audit quality (Kelley and Margheim 1987, Pierce and Sweeney 2004, McNamara and Liyanarachchi 2008). Underreporting may reflect the auditors’ attempt to satisfy the employer, as this behaviour is motivated by time budget pressure and performance evaluations, for example (Otley and Pierce 1996b; Pierce and Sweeney 2004).

To examine RQ2 and RQ3, we use the survey questions about how auditors prioritise different audit stakeholders, and we focus on auditors’ highest prioritised stakeholders. The different priorities are divided into the following categories: (i) regulators; (ii) shareholders; (iii) the wider society; (iv) client company; (v) employer (including audit firm, audit office and audit team); (vi) auditor herself; and (vii) others. The first three categories are assumed to represent more professional views on auditing, while the client or employer as the first priorities are assumed to represent more commercial views on auditing. In the regressions for RQ2 and RQ3, auditors considering the regulator as first priority are used as the benchmark, because those auditors could be expected to consider auditing standards the most.

The results indicate that when an auditor prioritises the employer, she reports more AQTBs. Moreover, auditors prioritising client, shareholders or themselves report more client-related AQTBs than auditors prioritising regulators. Consequently, emphasising the regulators in audit work is associated with fewer AQTBs. To compare to the results of Svanberg and Öhman (2015), they find that auditors’ identification with their clients is associated with more AQTBs. We then explore whether priorities are related to auditors’ overall quality assessment (OAQ). However, unlike the AQTB results, we do not find any significant associations between priorities and OAQ, arguably since audit quality is a multifaceted construct.

Finally, we examine whether the priorities are associated with the (level of) match between OAQ and AQTB. The match (OAQ_AQTB) is the difference between the OAQ and the reversed (mean) AQTB and represents auditors’ quality assessment (i.e. OAQ) relative to the reported behaviours in the audit process (i.e. AQTBs). Assessing OAQ above (below) reported AQTBs suggests over-assessments (under-assessments). Our results indicate that when auditors consider employers or clients as their highest priority in the audit work, they are more likely to over-assess engagement quality. This result implies that these auditors do not consider AQTBs to have a strong negative impact on audit quality, but rather, they may weigh other factors (e.g. meeting time budgets, delivering services valued by the client) in their overall quality assessment, which they consider that they have achieved.

Interestingly, our additional analyses indicate that significant results for RQ2 and RQ3 can be found only within the sub-sample of partners, but not managers or associates. This evidence underlines the influential role of partners’ views on the behaviours in the audit process,and suggest that different priorities could explain why audit partners express varying views on audit quality as documented in Brivot et al. (2018).

Overall, this study adds the following contributions to the audit literature. First, this study investigates auditors’ self-assessment of their own engagements. While prior studies on perceived audit quality investigate auditors’ conceptual or ideal view on audit quality (e.g. Brivot et al. 2018; Gonthier-Besacier et al. 2016; Warming-Rasmussen and Jensen 1998; Carcello et al. 1992; Schroeder et al. 1986), we do not know to what extent such general views correspond to self-assessments of actual audit engagements. In general, our findings corroborate prior studies underscoring that audit quality is a multifaceted construct, which cannot be captured by a single indicator (e.g. Christensen et al. 2016). Exploring the self-assessment of auditors’ engagements captures the ‘actual’ practice surrounding audit quality. From this perspective, the best way to understand how audit quality is performed on the field is to ask auditors themselves.

Second, this study provides evidence on how audit quality assessments are associated with auditors’ behaviours in the audit process. Prior studies on AQTBs (e.g. Herda et al. 2019; Svanberg and Öhman 2015; Coram et al. 2008; Donnelly et al. 2003; Malone and Roberts 1996; Kelley and Margheim 1990) have not addressed this question. The extent to which key behaviours in the process are related to self-assessment of quality is informative to understand whether auditors are indeed considering the regulators’ perspective regarding sufficient appropriate audit evidence in assessing their own work.

Third, this paper provides new insights on how auditors’ stakeholder priorities are associated with AQTBs (overall and individual AQTBs) and auditors’ overall quality assessments.⁹ Moreover, we show that a high proportion of auditors consider clients as their highest priority while a lower proportion of auditors view shareholders, society and regulators as the highest priority. Our findings suggest that regulators, educational institutions and audit firms could provide more or different types of training to address and discuss the issue of auditors’ stakeholder prioritisation. Regulators communicate with auditors in developing new regulations and ensuring adequate implementation, and in doing this, they may benefit from insights rendered in this study.

Fourth, prior audit literature has emphasised the role of individual auditor characteristics in audit quality (DeFond and Zhang 2014), and this study provides evidence on how views on audit quality differ across members of the same audit team. An individual auditor’s understanding of audit quality can constitute an important ‘input factor’ (i.e. auditor attribute) in the audit quality framework (Knechel et al. 2013a).¹⁰ While prior studies usually focus on the perceptions of audit partners (Carcello et al. 1992) or the characteristics of signing auditors (e.g. Chin and Chi, 2009, Zerni 2012, Gul et al. 2013), we consider the views and characteristics of auditors in different ranks. This is important because non-partners perform most of the evidence gathering in the audit process (Contessotto et al. 2019), the extent of commercial and professional views may depend on the auditor rank (e.g. Suddaby et al. 2009), and for example, views on earnings management may differ between auditors of different ranks (Nelson et al. 2002).

The remainder of the paper is structured as follows. Section 2 discusses the background and the research questions of the study. Section 3 describes the sample and the models. Section 4 presents the results, while section 5 concludes the paper.

2. Background literature and research questions

Audit quality has been widely examined and discussed in the prior literature (e.g. Aobdia 2019, Cameran et al. 2018, Christensen et al. 2016, Tepalagul and Lin 2015, DeFond and Zhang, 2014, Knechel et al. 2013a, Martin 2013, Francis 2011, Francis 2004). For several decades, researchers have attempted to investigate audit quality and its determinants. One conclusion that can be made from reviewing prior research is that there is little consensus on how to define this concept (Aobdia 2019, Christensen et al. 2016, DeFond and Zhang 2014, Knechel et al. 2013a). There are various perspectives on audit quality, and depending on the selected view (e.g. users, regulators, society, client, audit firms), audit quality is likely to be defined, described and assessed differently (Knechel et al. 2013a).¹¹

Interestingly, the perspectives of individual auditors on audit quality have been less explored. Some prior studies have investigated auditors’ views on attributes of audit quality and compared their views with those of other stakeholders to reveal any differences (e.g. Gonthier-Besacier et al. 2016; Warming-Rasmussen and Jensen 1998; Carcello et al. 1992; Schroeder et al. 1986). These differences have been related to the audit expectation gap or role perception gap that exist due to different expectations about what an auditor should do in the audit (Gonthier-Besacier et al. 2016; Zhang 2007; Humphrey 1997).¹²

Next, in section 2.1, we position our study within the literature on AQTBs. In section 2.2, we develop the theoretical perspectives that support the three research questions presented and developed in section 2.3.

2.1. Quality of the audit process and AQTBs

Although different stakeholders’ views on the definition of audit quality vary, Gaynor et al. (2016) point out that all stakeholders seem to agree that high-quality inputs and a high-quality process will increase the probability of a high-quality outcome [see also the audit quality framework by Knechel et al. (2013a)].¹³ One key assumption here is that what goes on during the audit process (i.e. type of behaviours, judgements and procedures) has a direct effect on the audit outcomes. For example, how audit evidence is collected and evaluated (see ISA 500) will affect financial reporting quality and the quality of audit reporting. The likelihood of auditors discovering material misstatements in the client company’s financial statements is affected by the sufficiency of audit procedures in the audit process. The aim of the procedures should be to obtain sufficient appropriate evidence (ISA 500, p. 6) that generate support for the auditor’s conclusions (ISA 500, p. A.1).

Based on auditing standards, which are complemented by audit firms’ policy documents for audit conduct, numerous tests and procedures of gathered audit evidence need to be undertaken during the audit process. However, as indicated by various empirical studies, there are sometimes significant differences between the standards and audit practice (Christensen et al. 2016). Such differences may exist because of inconsistent use of audit firm guidelines (Martinov and Roebuck 1998; Salterio and Denham 1997),¹⁴ time budget pressure (Pierce and Sweeney 2004) or because auditors are susceptible to common heuristics that cause biases in auditor judgment (see Knechel et al. 2013a for a review).

AQTBs are commonly used survey measures in the audit literature to indicate an auditor’s (in)sufficient audit conduct in the audit process (e.g. Herda et al. 2019; Svanberg and Öhman 2019; Broberg et al. 2017; Svanberg and Öhman 2015, Sweeney et al. 2013; Peytcheva and Gillett 2012; Malone and Roberts 1996; Otley and Pierce 1996b). AQTBs cover a wide range of intentional activities (such as inappropriate reliance on client work, superficial reviews and premature signoffs) that threaten the quality of the audit, thus causing serious consequences for the audit firms and profession (Malone and Roberts 1996; Coram et al. 2008; Sweeney et al. 2013). These behaviours negatively impact audit procedures and the sufficiency and appropriateness of audit evidence, and performing AQTBs are, in general, not in line with the auditing standards or the regulators’ view. The AQTBs capture key elements of gathering and evaluating audit evidence, hence they reflect the quality of the audit process.

Compared to any other audit stakeholder, auditors have the most holistic view of the entire audit process. One would therefore expect that their considerations of the audit engagement quality would reflect the perceived accuracy, completeness and sufficiency of relevant tests and procedures performed during the audit process. But do auditors view the AQTBs as harmful for the overall audit quality? When exploring auditor’s understanding of audit quality, we need to consider that auditors are subject to various expectations regarding audit quality, and different stakeholders’ expectations might influence auditors’ perspectives on audit work and audit quality. We broadly categorise these expectations into those representing professional views and those representing commercial views and discuss them in the next section.

2.2 Theoretical perspectives

By far, the most prevailing role of auditors discussed by regulators and researchers stems from the professional perspective that an auditor serves the users of financial statement by verifying the information disclosed in the financial statements – whether there exist material misstatements or not (ISA 200, p. 9). This verification is in the interest of shareholders and other financial statement users, and therefore, in the interest of regulators. One important factor in audit work, as emphasised in the auditing standards (ISA320, A.3), is that auditors should consider the interest of users (e.g. in deciding on the materiality level).¹⁵ Although the regulators emphasise the user perspective in developing the standards, the user may still expect different quality outcomes from the audit compared with the regulators. An auditor’s interests are attempted to align with those of these stakeholder groups via education (Li et al. 2017; Gramling et al. 1996; Monroe and Woodliff 1993), training (Che et al. 2018), regulation (DeFond and Zhang 2014; DeFond and Lennox 2011; Feldmann and Read, 2010) and oversight (Westermann et al. 2019; Krishnan et al. 2017). The professional views on auditing are expected to become internalised by the auditor, to shape the auditor’s professional identity and to form personal work priorities.

However, even with the knowledge of the auditing standards, ‘seeking to get the biggest bonus possible or striving to satisfy auditees rather than shareholders are goals that often come to occupy a prominent place in the mind of audit team members’ (Guénin-Paracini et al. 2014, p. 282). Auditors do not necessarily regularly engage with external shareholders and regulators, and there is a risk that auditors do not fully consider the professional perspective (e.g. Bauer 2015). Auditors are surrounded by commercial perspectives (Malsch and Gendron 2013; Wyatt 2004; Zeff 2003) thus leading the discussion to the interests and expectations of employers (audit firm) and client companies.

An auditor is continuously affected by the (financial) interests of the audit firm, that is, the employer. Inevitably, audit firms follow the requirements of the profession and develop their own internal systems, guidelines and routines to comply with rules and standards. Cooper and Robson (2006) suggest that accounting rules and standards are translated into practice in large audit firms, and these firms play a crucial role in professionalisation. At the same time, audit firms are profit-making organisations, and it is inherent in their operations to consider the profitability of the business. As debated during the last two decades, there is the risk that audit firms’ strong emphasis on commercialism comes at the expense of professionalism (Wyatt 2004; Zeff 2003). If the commercial interests dominate, auditors may perceive that investments in developing professional skills and abilities in auditing through learning and education are valued less than investments in growing a lucrative consulting business (e.g. Suddaby et al. 2007; Wyatt 2004). Cost-quality trade-offs are unavoidable in the audit setting, and there may be profitability considerations, such as reducing audit cost (e.g. time budgets) and thereby risking more AQTBs such as not performing all of the relevant audit procedures to ensure high audit quality (Pierce and Sweeney, 2004; McNair 1991).

The procedures, routines, systems and methodologies guiding audit work are typically established at the (global) audit firm level, and auditors may wish to complete the job in correspondence with firm expectations (Knechel et al., 2013a). McNair (1991) discusses the cost/quality dilemma in auditing, and reports that this dilemma is passed on to the individual auditors through the use of time budgets and norms of efficiency. Adherence and compliance with firm policies are likely important for individual auditors’ prospects at the audit firm. Hence, an auditor’s interest could be expected to be highly aligned with those of the audit firm (Knechel et al. 2013a).

Audit firms are incentivised to balance the different perspectives, such as both the regulators’ and clients’ interests, to avoid costly audit failures while at the same time ensuring a good relationship with clients (Knechel et al. 2013a; DeFond and Zhang 2014; Guénin-Paracini et al. 2015).¹⁶ In their audit work, auditors communicate and negotiate extensively with client management (see Salterio 2012 for a review), and may even align with the wishes of client management because of the (close) connection with the client and the financial interest in keeping the client. Auditors may show customer orientation, and they are possibly influenced by client preferences and explanations (Broberg et al. 2018; Svanberg and Öhman 2015), especially if the client is a former auditor (Daoust and Malsch 2020).¹⁷ Importantly, a client company’s demands for auditing might substantially deviate from those of financial statement users. Still, closeness to the client might make the auditor emphasise the client’s needs and those services valued by the client – such as consulting (Brivot et al. 2018; Bauer 2015).¹⁸ Guénin-Paracini et al. (2014) provide an interesting perspective by examining fear and risk in the audit process. Based on their discussion, it may be that auditor’s commercial views partly result from auditor’s defence strategies against anxiety caused by risk/fear of failing in terms of professional goals.

There are however important factors that are supposed to work against client orientation in audit work. These factors include the risk of litigation and reputational loss from sub-standard audits (DeFond and Zhang 2014), disciplinary sanctions (Westermann et al. 2019; Sundgren and Svanström 2017) and requirements to, for example, exercise professional scepticism in questioning and probing of a client and in examining of audit evidence (Hurtt et al. 2013; ISA 200, p. 7). It is however largely unknown to what extent these factors can reduce commercial views and client orientation, and hinder AQTBs.

2.3. Development of research questions

We explore actual audit engagements and the self-assessments of quality by auditors of various ranks who conducted these engagements. Specifically, we obtain auditors’ considerations about behaviours during the audit process, and their quality assessment of the audit engagements. Examining the association between AQTBs and auditors’ assessments of the overall audit engagement quality (OAQ) allows us to infer whether auditors themselves consider AQTBs relevant for audit quality. If the auditors’ assessments of audit quality reflect behaviours in the audit process, one would expect a higher (lower) overall quality assessment to be associated with fewer (more) AQTBs. Documenting a strong association would confirm that auditors regard insufficient behaviours in the audit process as harmful for the overall audit quality (assessment), which would then support the argument that they, in general, consider the auditing standards (the perspective of regulators) in their quality assessment. Christensen et al. (2016) report that audit professionals define audit quality in terms of compliance with auditing standards while results in Aobdia (2019) indicate that a poorly conducted audit as per applicable standards is associated with worse reporting outcomes.

However, as discussed in sections 2.1 and 2.2, auditors’ understanding of audit quality may be influenced not only by professional views but also commercial views on auditing. For example, the quality assessments may reflect consideration of whether the team was able to meet time budgets and time deadlines. While meeting budgets is essential to the audit firm, achieving this may also limit the possibility to conduct relevant audit procedures adequately (Pierce and Sweeney 2004), thus causing a weak association between the reported AQTBs and their assessment of OAQ.

Moreover, as proposed by Coram et al. (2008), auditors may perceive the various AQTBs differently. In particular, it may be that auditors do not perceive certain directly client-related AQTBs, such as over-relying on explanations and evidence from client management, to negatively influence audit quality (Coram et al. 2008).¹⁹ Assessing the negative influence of these AQTBs may be more challenging for the auditors, since their work is dependent on the information supplied by the client (Power 1997; Daoust and Malsch 2020). For example, Guénin-Paracini et al. (2015) find that auditor constantly attempt to balance with working diligently and efficiently (by means of many relational strategies to maintain the client’s desire to cooperate), and at the same time maintaining professional values. Interestingly, prior qualitative field-based research has documented that client is the most common source consulted in searching for explanations for unexpected findings (Hirst and Koonce 1996; Trompeter and Wright 2010).

Therefore, we examine which behaviours auditors themselves consider relevant for audit quality. Hence, our first research question is the following:

RQ1: To what extent are auditors’ reported audit quality threatening behaviours (AQTBs) in the audit process associated with their self-assessment of the overall audit quality (OAQ)?

Next, we investigate whether individual auditors’ stakeholder priority, i.e. how auditors prioritise various audit stakeholders, are associated with AQTBs and OAQ. The priorities could be expected to vary between individual auditors that are more or less receptive to different pressures or influences from various stakeholders.²⁰ To make sense of the world under pressures from various stakeholders, moral reasoning may guide the auditor and impact her judgments and behaviour. Moral reasoning is the formation of judgments in accordance with principles concerning others’ welfare, rights, and fairness (Dahl and Killen 2018).²¹ There are reasons to believe that considerations of others (i.e. different stakeholders) impact auditors’ understanding of various (ethical) dilemmas in audit situations and the quality of auditor judgment and behaviour (Schatzberg et al. 2005). AQTBs can be considered as moral issues (Coram et al. 2008) and unethical acts in audit practice (Herda et al. 2019).²²

Social identity theory (Tajfel and Turner 1985) suggests that a person’s identification with a certain group increases the person’s receptivity to the influence from that group, and further, decreases the person’s receptivity to influence from ‘non-group members’ (Hekman et al. 2009; Turner et al. 1987). Auditors’ individual considerations of whether actions are good or not may be dependent upon the social consensus in the group (Jones 1991, p. 375) with which the auditor identifies. Moral ambiguity – that is, the problem of not knowing what good ethics prescribes in a situation – is reduced when there is a high degree of social consensus about what good practice is (Jones 1991, Coram et al. 2008).

An auditor’s highest prioritised stakeholder could be expected to represent the group with which the auditor identifies the most, and where social consensus underlies the auditor’s assessments of the audit engagement quality. We explore whether and how auditors’ stakeholder prioritisation affects their self-assessments of engagement quality in two steps. First, the direct association between stakeholder priority and AQTB and OAQ, respectively, is analysed. Second, we test how priority is associated with the alignment (‘match’) between AQTB and OAQ.

For the first step, we expect that more professional views on auditing (in particular, the regulators as the highest priority) are associated with fewer AQTBs, while more commercial views on auditing (i.e. the employer or client as the highest priority) are associated with more AQTBs due to attempts to meet time deadline, for example. Strictly following the standards and thereby avoiding conducting AQTBs can be assumed to be more important for auditors with a professional view than for those with a commercial view. It may also be that prioritising client is specifically associated with those AQTBs that are directly related to the client. Prioritising the client may emphasise more heavily the importance of access to client information and keeping a good relationship with the client (Guénin-Paracini et al. 2015). Svanberg and Öhman (2015) document that an auditor who identifies relatively more with a client is more likely to adhere to client preferred treatment and to commit AQTBs, whereas they do not find a significant association between professional identification and AQTBs.²³

For OAQ, it is empirically unknown how stakeholder priorities may impact auditors’ assessment of audit engagement quality. However, if we assume that AQTBs are (strongly) reflected in auditors’ overall quality assessment (see arguments for RQ1), then we can expect that associations between priorities and OAQ are similar to those between priorities and AQTBs. For example, compared to an auditor who prioritises the employer or client, an auditor that prioritises the regulators is likely to perform fewer AQTBs and assess OAQ higher, because such an auditor considers that quality is about following the standards. It is however also possible that priorities impact the frequency of AQTBs but not the OAQ assessment, if auditors’ understanding of quality does not reflect the regulator/standards perspective of auditing. Based on the discussion above, we formulate our second research question:

RQ2: Is an auditor’s stakeholder priority associated with her reported audit quality threatening behaviours (AQTBs) in the audit process and her overall self-assessment of audit quality (OAQ)?

Finally, we investigate in the second step whether the alignment between individual auditors’ assessment of AQTBs and OAQ can be explained by auditors’ stakeholder priority. For example, if an auditor prioritises the regulator, it may be that OAQ and AQTB are strongly aligned. On the other hand, an auditor prioritising the client may value other non-AQTB-related factors in the OAQ assessment, and hence, there may be many AQTBs, but a low alignment between OAQ and AQTB assessments. An auditor prioritising client may underestimate the quality risks related to accepting weak client explanations and relying on client-generated audit evidence and still assess the quality as high, particularly if the client is satisfied with the service(s) provided (Brivot et al. 2018). Furthermore, prioritising the employer may lead to deviations between OAQ and AQTB, for example, if the audit firm is emphasising other (quality) aspects of the audits (e.g. efficiency gains) over those highlighted by regulators in the auditing standards, or if the firm emphasises the importance of keeping this highly profitable or reputational client at the expense of conducted an independent audit of high quality. Finally, regarding auditors who have a strong professional view and prioritise different users of financial statements (e.g. stakeholders, society), their assessment of OAQ may also deviate from AQTBs if they, for example, believe that the engagement team could have done more (e.g. to detect fraud) than only avoiding conducting AQTBs (Litjens et al. 2015). Based on the discussion above, we formulate our third research question:

RQ3: Is an auditor’s stakeholder priority associated with the alignment between her overall self-assessment of audit quality (OAQ) and her reported audit quality threatening behaviours (AQTBs) in the audit process?

3. Research design

3.1 Data and sample

This study uses survey data gathered in one of the Big 4 audit firms in Sweden. Before the survey was distributed, the Big 4 firm first provided us with data for 909 teams conducting 909 engagements (in both public and private clients) for the audit cycle from July 2015 to June 2016. We use this dataset as a starting point for the sample selection. As one auditor could be involved in multiple teams, each auditor is required to respond to the survey for one engagement only, and we choose the engagement the auditor has spent the most time on, compared to other engagements of the same auditor. We strive to gather responses from multiple auditors for each specific engagement team. A minimum of three team members (always including one partner, one manager and one associate) are included for each engagement. As a result, 776 auditors from 217 audit teams are invited to participate in this survey. Before distribution of the final survey, we performed a pilot test of the questionnaire with 17 responding auditors.

The questionnaire captures team members’ perceptions on the overall audit quality of the engagement and audit quality threatening behaviours, among others. In addition, the survey data contain information on auditors’ attributes such as gender, education, experience etc.²⁴ An online survey instrument was used to distribute the questionnaire and a letter from the researchers that informed the participants about the project, and their confidentiality and anonymity (Gibbins 2001; Nelson et al. 2002).²⁵ The survey instrument was reviewed and approved by partners in the Big 4 firm under study, and the ethical aspects regarding the questionnaire were confirmed against the current requirements at the relevant University from which the survey was sent out.²⁶ The relevant survey questions for this paper are available in the online Appendix.

Finally, 335 team members from 185 teams responded to the survey. The response rate is 43% at the auditor level (43%=335/776) and 85% at the team level (85%=185/217).²⁷ Hence, the full sample contains 335 auditors in 185 different audit teams.²⁸ The overall response rate of 43% is high relative to prior survey work (e.g. 20.7% in Abbott et al. 2016). Our high response rate can be attributed to the strong support of the Big 4 firm, and there is some precedent for such high response rates among auditor respondents, such as 56.7% response rate in Bobek et al. (2015).

To examine auditors’ self-assessment of their own audit engagement and behaviours, we compare auditors that have worked on the same engagement team, because client and team characteristics are heterogeneous. We require that a team has at least two responding auditors to be included in the analysis and use team fixed effect to control for team/client-specific effects. Hence, the final sample consists of 251 auditors from 101 teams.²⁹

3.2 Swedish setting

Auditors in Sweden follow the International Standards of Auditing (ISAs) on all their private and public audit engagements³⁰ and national legislation in the form of the Auditors Act (Sw. Revisorslag) and the Company Act (Sw. Aktiebolagslagen). The Auditors Act does not include any guidance on how to perform the audit but instead outlines some general requirements about certification of auditors, auditor oversight and disciplinary sanctions, auditor independence and registration and responsibilities of audit firms. Chapter 9 in The Company Act deals with audits and informs about the tasks of the auditor, formal competence and independence requirements for the auditor, audit report and the requirement to report (to authorities) on suspicion of a crime. These Acts establish general requirements for the auditor and the audit engagements without explaining how the audit should be carried out in practice. One important aspect that is specific to Sweden is the requirement for the auditor to investigate whether or not the auditee has followed laws and the articles of association (Company Act, 9:3). Sweden follows EU Regulation (537/2014) and has implemented the EU Directive (2014/56/EU) through provisions in the Auditors Act, Company Act and other statutes.

All auditors in Sweden are subject to external oversight from the Swedish Inspectorate of Auditors (SIA). After having conducting quality control investigations, SIA issues disciplinary sanctions against certified auditors who perform below the standards. There are regular quality inspections, risk-based inspections (e.g. complaints), and thematic inspections focusing on how the audit industry deals with specific audit issues. In the 2012–2017 period, 314 disciplinary sanctions (i.e. withdrawal of certificate, warning and reprimand) were issued against approximately 9% of the certified auditors. 42 auditors had their certificate withdrawn in this period.

One important difference between auditors that have public clients and those that only audit private clients relate to the regulatory inspections by SIA. Auditors with publicly listed clients are inspected every third year by SIA while auditors without public clients are inspected every sixth year. Another difference is that inspections of auditors without public assignments have partly been delegated to professional institute for the accountancy profession, FAR, for practical reasons. Auditors that have clients that are listed in the U.S. are in addition subject to the quality inspections performed by the Public Company Accounting Oversight Board (PCAOB). Research has documented pervasiveness of regulator influence in the U.S. (Johnson et al. 2019; Westermann et al. 2019) but to a lesser extent in other settings (Malsch & Gendron, 2011; Peecher et al. 2013). The regulatory influence can be expected to be lower in the Swedish setting compared with in the U.S., which may impact auditors’ priorities and perceptions.³¹

3.3 Regression models and variables

3.3.1 Regression model for RQ1

To examine whether auditors’ self-assessments of overall audit quality are related to quality-driving behaviours in the audit process, we conduct the following ordered logistic regression: (1) $OAQ=a+b_{1}AQBR\ast +b_{2}Female+b_{3}Partner+b_{4}Manager+b_{5}HigherEducation+b_{6}EngExperience+FE\mathrm{\_}Team+e$(1) where the dependent variable is auditors’ assessment of overall audit quality (OAQ) based on the question ‘Relative to your other engagements with similar client characteristics, how would you rate the overall audit quality of this engagement, from 1 (low) to 5 (high)?’ Hence, OAQ has a Likert scale between 1 and 5, where higher values indicate higher assessed quality.

The test variable for RQ1 is auditors’ audit quality threatening behaviours (AQTBs), which are measured by eight questions (AQTB1–AQTB8).³² All the eight AQTB measures also have a Likert scale between 1 and 5, and a low (high) value indicates fewer (more) threatening behaviours, and therefore likely high (low) audit quality. To make it easier to compare OAQ and AQTB, we reverse the scales of all the AQTB measures, so one would expect a positive association between our measures of OAQ and the reversed AQTB (denoted AQTBr). AQTBr* in Equation (1) refers to the following:

1. AQTBr_All: the mean of all eight AQTBr,

2. AQTBr_Client: the mean of the AQTB measures that explicitly mention audit clients, i.e. AQTB5r, AQTB6r, and AQTB7r,

3. AQTBr_Others: the mean of the AQTB measures that do not explicitly mention audit clients, i.e. AQTB1r, AQTB2r, AQTB3r, AQTB4r, and AQTB8r, or

4. each individual AQTB measure (AQTB1r, … , AQTB8r), separately.

Ordered logistics regression is used for all the regressions as OAQ has a Likert scale between 1 and 5.³³

We control for individual auditor attributes in Equation (1). First, considering that men are more overconfident than women (Bengtsson et al. 2005), one would expect that male auditors are more likely to be (over)optimistic regarding the overall audit quality of the engagement. Hence, in the model, we control for the auditor’s gender with the variable Female, which equals 1 if the respondent is female and 0 otherwise.³⁴ We further control for auditors’ rank by variables Partner and Manager. The variable Partner is equal to 1 if the auditor is a partner or a director, and 0 otherwise.³⁵ The variable Manager is equal to 1 if the auditor is a senior manager, manager or associate manager. Associates are treated as the benchmark.³⁶

Moreover, we control for the level of an auditor’s education and expect that auditors with higher education are better able to perform audits with high quality. HigherEducation is an indicator variable, which equals 1 if the respondent has a master’s degree or above, and 0 if she has a bachelor’s degree. Finally, a longer engagement tenure increases an auditor’s understanding of the client, which is important for audit conduct and audit quality (e.g. ISA 315; Langli and Svanström 2014; DeFond and Zhang, 2014). Auditors that have worked on the engagement over several years may have good knowledge of the client. Hence, we control for auditor tenure. EngExperience is the respondent’s experience of the audit engagement in years. Team fixed effects are included to control for team- and client-specific effects in all the regressions. All the variables are defined in Appendix 1.

3.3.2 Regression model for RQ2

To examine whether an auditor’s stakeholder priority is associated with her reported AQTBs or OAQ (RQ2), we use the following regression: (2) $\begin{array}{rl}& AQTBr\ast orOAQ=a+b_{1}PriorityClient+b_{2}PriorityEmployer+b_{3}PriorityShareholder\\ & +b_{4}PrioritySociety+b_{5}PrioritySelf+b_{6}PriorityOther+b_{7}Female\\ & +b_{8}Partner+b_{9}Manager+b_{10}HigherEducation+b_{11}EngExperience+FE\mathrm{\_}Team+e\end{array}$(2) The dependent variable is either AQTBr* or OAQ (as defined earlier). The independent variables of interest are based on the respondents’ answers to the question ‘Please rank what you consider to be your highest (1) to lowest (9) priority while working as an auditor? (Every option should be given a number)’. The available options to respond to this question are as follows: The Client/Customer; The Shareholder; The Audit Firm; The Audit Office; The Regulators; The Wider Society; The Audit Team; Yourself; and Other. PriorityClient is equal to 1 if the auditor considers the audit clients as the highest priority (i.e. ranked as 1), and 0 otherwise. PriorityEmployer is equal to 1 if the auditor considers the audit firm, audit office, or audit team as the highest priority, and 0 otherwise. PriorityShareholder is equal to 1 if the auditor considers shareholders as the highest priority, and 0 otherwise. PrioritySociety is equal to 1 if the auditor considers the wider society as the highest priority, and 0 otherwise. PrioritySelf is equal to 1 if the auditor considers herself as the highest priority, and 0 otherwise. PriorityOthers is equal to 1 if the auditor answered ‘Other’ or left the question unanswered, and 0 otherwise. The reference category includes responses indicating the regulators as the highest priority (PriorityRegulator).

Auditor-specific control variables are as defined earlier. Team fixed effects are included in all the regressions. All the variables are defined in Appendix 1.

3.3.3 Regression model for RQ3

To examine whether an auditor’s stakeholder priority is associated with the match between her reported AQTBs and the OAQ (RQ3), we use the following regression: (3) $\begin{array}{rl}OAQ\mathrm{\_}AQTBr& =a+b_{1}PriorityClient+b_{2}PriorityEmployer+b_{3}PriorityShareholder\\ & +b_{4}PrioritySociety+b_{5}PrioritySelf+b_{6}PriorityOther+b_{7}Female+b_{8}Partner\\ & +b_{9}Manager+b_{10}HigherEducation+b_{11}EngExperience+FE\mathrm{\_}Team+e\end{array}$(3) The dependent variable is OAQ_AQTBr, which is the difference between OAQ and the mean value of AQTBr, representing the self-assessment of OAQ in relation (i.e. relative) to reported behaviours in the audit process. Put differently, our dependent variable proxies whether auditors’ assessment of overall quality is consistent (or matched) with their reported AQTBs. If OAQ is lower than (reversed) AQTBs, and hence the OAQ_AQTBr is negative, we consider this as under-assessment of audit quality. In contrast, if OAQ is higher than (reversed) AQTBs, we consider this as over-assessment of audit quality.

We analyse OAQ_AQTBr in three ways: (1) We use the signed variable that ranges from under-assessments to over-assessments; (2) We use a sub-sample of observations that range from zero (i.e. match) to positive values (over-assessment); and (3) We use a sub-sample of observations that range from zero (i.e. match) to negative values (under-assessment). We use truncated OLS to analyse sub-samples of observations.

Overall, regression model 3 examines whether the alignment between auditors’ assessment of AQTBs and OAQ (measured by OAQ_AQTBr) can be explained by auditors’ stakeholder priority. The construct of OAQ_AQTBr captures a different dimension of auditors’ assessment of audit quality from the two separate variables OAQ and AQTBr tested in models 1 and 2, and this test has therefore a different focus.

Variables of auditor priorities and auditor-specific control variables are as defined earlier. Team fixed effects are included in all the regressions. All the variables are defined in Appendix 1.

3.4 Descriptive statistics

Panel A of Table 1 presents the distribution of OAQ and each of the eight AQTBr measures.³⁷

Table 1: Descriptive Statistics.

Panel A: Distribution of overall audit quality (OAQ) and (reversed) audit quality threatening behaviours (AQTB1r-AQTB8r)

	OAQ	AQTB1r	AQTB2r	AQTB3r	AQTB4r	AQTB5r	AQTB6r	AQTB7r	AQTB8r
1	0	1	6	0	1	1	1	1	1
2	6	11	14	5	5	18	16	16	10
3	31	38	33	22	20	52	39	50	23
4	143	135	110	106	107	132	140	128	126
5	71	66	88	118	118	48	55	56	91
Sum	251	251	251	251	251	251	251	251	251

Panel B: Descriptive statistics of dependent, test and control variables
	Mean	SD	Min	p5	p25	p50	p75	p95	Max
OAQ	4.11	0.7	2	3	4	4	5	5	5
AQTBr_All	4.07	0.55	2.13	3	3.88	4	4.5	5	5
AQTBr_Client	3.88	0.67	2	2.67	3.67	4	4.33	5	5
AQTBr_Others	4.18	0.56	2	3.2	4	4.2	4.6	5	5
OAQ_AQTBr	0.04	0.72	−1.88	−1	−0.38	0	0.5	1.25	1.88
Female	0.44	0.5	0	0	0	0	1	1	1
Partner	0.29	0.45	0	0	0	0	1	1	1
Manager	0.39	0.49	0	0	0	0	1	1	1
HigherEducation	0.94	0.24	0	0	1	1	1	1	1
EngExperience	4.67	5.47	0	1	2	3	5	15	25
N	251

Panel C: Distribution of auditors’ stakeholder priorities
	Main sample			Full sample
	0	1	Total	0	1	Total
PriorityClient	172	79	251	232	103	335
PriorityEmployer	228	23	251	303	32	335
PriorityShareholder	230	21	251	303	32	335
PriorityRegulator	239	12	251	319	16	335
PrioritySociety	243	8	251	324	11	335
PrioritySelf	232	19	251	308	27	335
PriorityOthers	162	89	251	221	114	335

Panel D: Correlation Matrix
		1	2	3	4	5	6	7	8	9	10	11	12
1	PriorityClient	1
2	PriorityEmployer	−0.22#	1
3	PriorityShareholder	−0.20#	−0.1	1
4	PrioritySociety	−0.12*	−0.06	−0.05	1
5	PriorityRegulator	−0.15*	−0.07	−0.07	−0.04	1
6	PrioritySelf	−0.19#	−0.09	−0.09	−0.05	−0.06	1
7	PriorityOthers	−0.50#	−0.24#	−0.22#	−0.13*	−0.17#	−0.21#	1
8	Female	0.02	0.05	−0.07	−0.07	0.06	−0.1	0.04	1
9	Partner	−0.07	0.01	0.03	0.19#	−0.02	−0.08	0.03	−0.12*	1
10	Manager	0.09	0	0.05	−0.1	−0.03	−0.14*	0.00	−0.01	−0.51#	1
11	HigherEducation	−0.02	−0.04	−0.05	0.04	−0.03	0.07	0.03	0.04	0.04	−0.02	1
12	EngExperience	−0.03	0.01	−0.01	−0.01	−0.05	0.04	0.04	−0.04	0.18#	−0.07	−0.01	1

This table presents various statistics of the dependent, test, and control variables. Panel A reports the distribution of the overall audit quality (OAQ) and the eight reversed individual audit quality threatening behaviours (AQTB1r-AQTB8r). Panel B reports the descriptive statistics of the dependent, test and control variables. Panel C reports the distribution of auditors’ different stakeholder priorities. Panel D reports the correlation matrix between auditors’ stakeholder priorities and control variables. * (#) indicates significance at the 5% (1%) level. All the variables are defined in Appendix 1.

Recall that OAQ and AQTBrs have scales between 1 and 5. Among the 251, 143 auditors assess the OAQ of their engagement as 4 out of 5, and 71 perceive they have done an excellent job (rank 5). Only 6 and 31 auditors rate their OAQ as 2 and 3, respectively, and no one perceives their OAQ as the lowest rate (1). Put differently, 85% (= (143 + 71)/251) of auditors rate their audit quality as 4 or 5 while 15% assess their audit quality below 4.

Compared to the OAQ measure, there is more variation in the distributions of the eight AQTBr measures. In general, the majority of auditors have rated them 4 or 5 on the reversed scale from ‘Always’ to ‘Never’ (the original scale is from ‘Never’ to ‘Always’, see Appendix 1). For AQTB3r (premature signoffs) and AQTB4r (unauthorised reductions of sample size), 118 auditors have rated five, and relatively fewer (106/107) auditors have rated 4. In contrast, for the three client-related AQTB measures (AQTB5r, AQTB6r, and AQTB7r), there are the lowest numbers of auditors that have rated 5 (48, 55, and 56, respectively). When we focus on the lower scales of AQTBr (i.e. more frequent quality threatening behaviours), only 26 (27) auditors have rated AQTBr3 (AQTBr4) below 4, while 71, 56 and 67 auditors have rated AQTBr5, AQTBr6 and AQTBr7 below 4, respectively. These statistics show that there are more AQTBs that explicitly mention the client than other types of AQTBs. While no auditor rated their OAQ to the lowest scale, seven AQTBr measures have been rated to the lowest scale by at least one auditor. All AQTBs are related to the various requirements communicated in the ISAs (ISA 320, 330, 500, 505, 530, 700) with those specifically relating to audit evidence gathered from the client being specified in ISA 500 (A.30, A.21).

These descriptive findings generally align with prior studies that have shown that false signoffs (similar to our AQTB3r regarding premature signoff) are less likely AQTBs to occur (e.g. Malone and Roberts 1996; Herrbach 2001), whereas accepting weak explanations from clients (our AQTB6r) and superficial review of documents (our AQTB7r) are somewhat more frequent (Otley and Pierce 1996b; Herrbach 2001).

Panel B of Table 1 presents the descriptive statistics of the dependent, test and control variables. The average of all the (reversed) AQTBs, AQTBr_All, has a mean of 4.07 and a standard deviation of 0.55. AQTBr_Client has a lower mean value (3.88) and a larger standard deviation (0.67) while AQTBr_Others has a higher mean (4.18) and lower standard deviation (0.56). These statistics indicate that auditors more frequently conduct AQTBs that explicitly mention the client, compared to other behaviours. It may be that the close connection with the client during audit work can make it difficult for the auditor to remain sceptical and critical of audit evidence that originates directly from the client. The mean value of OAQ_AQTBr is 0.04, suggesting that, on average, OAQ is assessed higher than what can be expected from the AQTBs. There are 44% female auditors and 94% of auditors have a master’s degree. Among all the auditors, 28.7% and 39% of auditors are partners and managers, respectively. The average auditor’s engagement experience is 4.67 years (EngExperience).

Panel C of Table 1 reports the frequency of stakeholder priorities; that is, how many auditors have indicated that their highest priority is the client, employer, shareholders, regulators, society, themself or others. The first column presents frequencies when using the sample used in the main analysis (251 observations). In our sample, 40.6% [=(79 + 23)/251] of the auditors consider the client or the employer as the highest priority, while only 17.6% [= (21 + 12 + 8)/251] of the auditors consider the shareholders, regulators or society as their highest priority. Moreover, 7.6% of the auditors consider themselves (PrioritySelf) as the highest priority, and 35% (=89/251) responded ‘other’ or no answer, which implies that these auditors are not sure about whom they serve or do not think the shareholders, society, client, employer, etc. are the most important stakeholders.³⁸ The second column reports the statistics for the full sample (335 observations) and provide a similar percentage for each type of stakeholder priority.

While the purpose of auditing is to provide assurance of financial reporting quality for the users (e.g. shareholders, creditors) (ISA 200, p. 3), these statistics on auditors’ stakeholder priorities suggest that mostly auditors are not fully aware of whom they serve or they prioritise non-users. This may be because the training and education of auditors have not emphasised the perspective of financial statement users strongly enough. Hence, regulators, oversight bodies, audit firms, and/or education institutions should pay more attention to this aspect of auditor education.

Panel D of Table 1 presents a correlation matrix of the control variables. In general, there is no high correlation among these variables.

4. Results

4.1 Results for RQ1

Table 2 presents the regression results for RQ1. Panel A reports the results of regressing OAQ on AQTBr_All, AQTBr_Client and AQTBr_Others, respectively. All the coefficients are positive and significant at the 1% level, indicating that AQTBs are positively associated with assessments of overall audit quality (RQ1). Note that the coefficient on AQTBr_Client (1.243) is lower than that on AQTBr_All (1.692) and AQTBr_Others (1.529), suggesting that AQTBr_Client is relatively less related to OAQ compared to AQTBr_Others.³⁹ The adjusted R²s are around 41% for all regressions. This suggests that, although OAQ is significantly associated with AQTBs, the portion of OAQ that cannot be explained is still very large, highlighting that other factors weigh heavily in the auditors’ quality assessments.

Table 2: Main results for RQ1

Panel A: Regress OAQ on AQTBr_All, AQTBr_Client and AQTBr_Others and control variables using team fixed effects.

	(1)	(2)	(3)
	OAQ	OAQ	OAQ
AQTBr_All	1.692***
	(4.21)
AQTBr_Client		1.243***
		(3.68)
AQTBr_Others			1.529***
			(4.02)
Female	−0.654	−0.565	−0.653
	(−1.49)	(−1.30)	(−1.50)
Partner	1.582***	1.515***	1.611***
	(3.01)	(2.93)	(3.07)
Manager	0.947**	1.016**	0.857*
	(2.05)	(2.20)	(1.88)
HigherEducation	−0.505	−0.820	−0.361
	(−0.54)	(−0.87)	(−0.38)
EngExperience	0.071	0.080*	0.061
	(1.49)	(1.65)	(1.39)
N	251	251	251
Pseudo R^2	0.410	0.402	0.406

Panel B: Regress OAQ on AQTB1r-AQTB4r and control variables using team fixed effects.
	(1)	(2)	(3)	(4)
	OAQ	OAQ	OAQ	OAQ
AQTB1r	1.067***
	(3.93)
AQTB2r		0.368*
		(1.75)
AQTB3r			1.239***
			(4.01)
AQTB4r				0.711**
				(2.56)
Female	−0.451	−0.424	−0.603	−0.686
	(−1.05)	(−0.99)	(−1.40)	(−1.56)
Partner	1.788***	1.515***	1.730***	1.221**
	(3.37)	(2.96)	(3.32)	(2.40)
Manager	0.805*	0.903**	0.813*	0.759*
	(1.78)	(2.02)	(1.78)	(1.71)
HigherEducation	−0.735	−0.433	−0.815	−0.301
	(−0.77)	(−0.46)	(−0.85)	(−0.32)
EngExperience	0.063	0.068	0.057	0.072*
	(1.38)	(1.60)	(1.35)	(1.76)
N	251	251	251	251
Pseudo R^2	0.405	0.379	0.407	0.386

Panel C: Regress OAQ on AQTB5r-AQTB8r and control variables using team fixed effects.
	(1)	(2)	(3)	(4)
	OAQ	OAQ	OAQ	OAQ
AQTB5r	0.433*
	(1.77)
AQTB6r		1.118***
		(3.96)
AQTB7r			0.721***
			(2.90)
AQTB8r				0.515**
				(2.13)
Female	−0.483	−0.628	−0.403	−0.495
	(−1.12)	(−1.45)	(−0.94)	(−1.15)
Partner	1.572***	1.293**	1.445***	1.445***
	(3.06)	(2.48)	(2.85)	(2.87)
Manager	0.880**	0.951**	0.944**	0.782*
	(1.98)	(2.09)	(2.07)	(1.77)
HigherEducation	−1.077	−0.623	−0.255	−0.645
	(−1.12)	(−0.65)	(−0.27)	(−0.68)
EngExperience	0.066	0.106**	0.070	0.066*
	(1.52)	(2.00)	(1.64)	(1.66)
N	251	251	251	251
Pseudo R^2	0.379	0.407	0.390	0.382

This table presents the results of regressing overall audit quality (OAQ) on the average reversed AQTBs (AQTBr_All), client-related AQTBs (AQTBr_Client) and other types of AQTBs (AQTBr_Others), and each of the eight individual reversed AQTBs (AQTB1r-AQTB8r), respectively. All the variables are defined in Appendix 1. We use ordered logistic regression and use team fixed effects in all the analyses in this table. The z-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Panel B (Panel C) of Table 2 presents the results of regressing OAQ on AQTB1r-AQTB4r (AQTB5r-AQTB8r), respectively. The coefficients on all AQTBr are positive and significant, although there are variations in terms of both magnitude and significance level. The coefficient on AQTB3r (‘sign off an audit-program step without completing the work or noting the omission’) is the greatest (1.239), significant at the 0.01 level. To compare, Coram et al. (2008) suggest that auditors have the highest ethical concerns about performing false signoffs. Skipping to complete an audit step in the audit process can be viewed as a major departure from conducting a complete audit that involve the gathering of sufficient and appropriate audit evidence in accordance with the ISAs (e.g. ISA 520, p. 7).

Also significant at the 0.01 level are the coefficients on AQTB1r (‘reducing the amount of work performed below a reasonable level’), AQTB6r (‘accepting weak client explanations’) and AQTB7r (‘making superficial reviews on client documents’). The findings of Coram et al. (2008) suggest that auditors perceive that accepting weak client explanations and superficial review of client documents involve less moral intensity than performing false signoffs. Finally, AQTB8r (‘failing to investigate an accounting principle’) and AQTB4r (‘making an unauthorised reduction of sample size’) have coefficients 0.721 and 0.711, respectively, which are significant at the 0.05 level. Coram et al. (2008) report that not testing all items in a sample has the lowest levels of moral intensity.

The coefficients on AQTB2r and AQTB5r are marginally significant (at the 0.10 level). The results hence suggest that auditors consider ‘under-report audit time (AQTB2r)’ and ‘greater than appropriate reliance on client work (AQTB5r)’ as comparatively less relevant for OAQ. Prior studies imply that under-reporting audit time may deteriorate audit quality in the long run, since it helps perpetuate artificially tight budgets and unreliable time records (e.g. Otley and Pierce 1996b, Pierce and Sweeney 2004). However, it may be that the association between AQTB2r and the overall quality of the engagement is less direct from the perspective of auditors. The coefficient on AQTBr5 indicates that, on average, auditors do not very strongly perceive that relying on client work represents a threat to overall audit quality, even though the ISAs raise this issue in several sections (e.g. ISA 500, A.31; ISA 500, A.20), that is, the appropriateness of audit evidence is compromised. Prior studies have discussed how auditing is heavily dependent on the information provided by clients (Jeppesen 1998), and this is a common reason for challenges in auditor’s operational independence (Daoust and Malsch 2020).

Turning to the results of the variables Partner and Manager, we find that auditors’ rank has a very significant association with OAQ. Specifically, the coefficients on Partner are positive and significant at the 0.01 level for all the regressions in Table 2, suggesting that partners tend to assess OAQ significantly higher than associates. The coefficients on Manager are also positive and significant, although the magnitude of the coefficients are somewhat smaller than those of Partner, and the significance is at the 0.05 or 0.1 level. This indicates that managers tend to assess OAQ higher than associates, but lower than partners.⁴⁰

Regarding the results of other control variables, EngExperience has a significant and positive coefficient in a few regressions in Table 2. This suggests that auditors with more experience in the engagement may have assessed the overall engagement quality higher. Other control variables are insignificant.

4.2 Results for RQ2

Table 3 presents the regression results for RQ2 regarding auditors’ stakeholder priorities, and AQTBs and OAQ. In Panel A of Table 3, the dependent variables are AQTBr_All [column (1)], AQTBr_Client [column (2)], AQTBr_Others [column (3)] and OAQ [column (4)], respectively.

Table 3: Main results for RQ2

Panel A: Regress AQTBr_All, AQTBr_Client, AQTBr_Others, and OAQ on auditor priority -variables and control variables using team fixed effects. Benchmark is the group of auditors indicating ‘regulators’ as the highest priority.

	(1)	(2)	(3)	(4)
	AQTBr_All	AQTBr_Client	AQTBr_Others	OAQ
PriorityClient	−1.048	−2.481***	0.141	−0.950
	(−1.45)	(−3.22)	(0.20)	(−1.00)
PriorityEmployer	−2.585***	−3.159***	−1.724**	−0.404
	(−3.15)	(−3.64)	(−2.08)	(−0.37)
PriorityShareholder	−0.917	−1.981**	0.074	−0.756
	(−1.07)	(−2.14)	(0.08)	(−0.66)
PrioritySociety	0.106	−0.175	0.197	−1.541
	(0.10)	(−0.16)	(0.18)	(−1.16)
PrioritySelf	−0.798	−1.754*	−0.124	0.877
	(−0.92)	(−1.92)	(−0.14)	(0.77)
PriorityOthers	−1.287*	−2.322***	−0.298	0.105
	(−1.73)	(−2.91)	(−0.40)	(0.11)
Female	0.605*	0.622*	0.531	−0.398
	(1.72)	(1.71)	(1.50)	(−0.89)
Partner	−0.246	−0.175	−0.190	1.630***
	(−0.63)	(−0.43)	(−0.47)	(3.10)
Manager	−0.067	−0.187	0.048	1.052**
	(−0.18)	(−0.50)	(0.13)	(2.19)
HigherEducation	−0.263	0.325	−0.420	−1.198
	(−0.35)	(0.42)	(−0.58)	(−1.22)
EngExperience	0.005	−0.025	0.035	0.072*
	(0.13)	(−0.60)	(1.04)	(1.72)
N	251	251	251	251
Pseudo R^2	0.119	0.172	0.129	0.389

Panel B: Regress individual AQTBs (AQTB1r-AQTB4r) on auditor priority -variables and control variables using team fixed effects. Benchmark is the group of auditors indicating ‘regulators’ as the highest priority.
	(1)	(2)	(3)	(4)
	AQTB1r	AQTB2r	AQTB3r	AQTB4r
PriorityClient	−1.822**	0.956	−0.586	1.512
	(−1.98)	(1.15)	(−0.54)	(1.50)
PriorityEmployer	−3.783***	−0.768	−1.516	−1.336
	(−3.68)	(−0.80)	(−1.36)	(−1.24)
PriorityShareholder	0.665	−0.706	−1.977	1.204
	(0.61)	(−0.71)	(−1.58)	(0.98)
PrioritySociety	0.912	−0.155	−0.837	−1.484
	(0.69)	(−0.11)	(−0.60)	(−0.93)
PrioritySelf	−1.274	−0.102	−1.240	0.037
	(−1.16)	(−0.10)	(−1.03)	(0.03)
PriorityOthers	−1.428	0.110	−1.155	−0.714
	(−1.50)	(0.13)	(−1.06)	(−0.68)
Female	0.246	0.359	0.501	1.394***
	(0.60)	(0.93)	(1.14)	(3.06)
Partner	−1.104**	−0.602	−0.758	2.176***
	(−2.25)	(−1.32)	(−1.51)	(3.85)
Manager	−0.148	−0.435	0.509	0.445
	(−0.34)	(−1.00)	(1.11)	(0.93)
HigherEducation	−0.207	−0.788	0.741	−2.063**
	(−0.25)	(−0.96)	(0.77)	(−2.00)
EngExperience	0.078**	0.039	0.062	−0.006
	(2.00)	(1.14)	(1.43)	(−0.15)
N	251	251	251	251
Pseudo R^2	0.313	0.244	0.336	0.351

Panel C: Regress individual AQTBs (AQTB5r-AQTB8r) on auditor priority -variables and control variables using team fixed effects. Benchmark is the group of auditors indicating ‘regulators’ as highest priority.
	(1)	(2)	(3)	(4)
	AQTB5r	AQTB6r	AQTB7r	AQTB8r
PriorityClient	−2.016**	−2.965***	−1.660*	−0.058
	(−2.08)	(−3.29)	(−1.92)	(−0.06)
PriorityEmployer	−2.038**	−4.219***	−2.856***	−0.914
	(−1.98)	(−4.20)	(−2.97)	(−0.87)
PriorityShareholder	−0.899	−1.953*	−2.330**	0.564
	(−0.80)	(−1.82)	(−2.23)	(0.51)
PrioritySociety	2.277	−1.432	−1.789	−0.522
	(1.64)	(−1.00)	(−1.43)	(−0.39)
PrioritySelf	−1.253	−2.315**	−1.323	−0.461
	(−1.15)	(−2.20)	(−1.28)	(−0.41)
PriorityOthers	−1.350	−2.615***	−2.002**	−0.750
	(−1.35)	(−2.79)	(−2.22)	(−0.77)
Female	0.732*	0.619	−0.086	0.049
	(1.87)	(1.56)	(−0.22)	(0.13)
Partner	−1.422***	0.589	0.058	0.262
	(−3.11)	(1.25)	(0.13)	(0.56)
Manager	−0.211	−0.029	−0.423	0.418
	(−0.50)	(−0.07)	(−1.03)	(1.03)
HigherEducation	3.039***	−0.190	−1.672*	1.086
	(3.56)	(−0.22)	(−1.91)	(1.22)
EngExperience	0.025	−0.077	0.007	0.015
	(0.65)	(−1.64)	(0.19)	(0.41)
N	251	251	251	251
Pseudo R^2	0.270	0.279	0.264	0.242

This table presents the results of regressing AQTBs (AQTBr_All, AQTBr_Client, AQTBr_Others and eight individual AQTBs), and OAQ on auditor priority -variables and control variables. Benchmark is the group of auditors indicating ‘regulators’ as the highest priority. All the variables are defined in Appendix 1. We use ordered logistic regression and use team fixed effects in all the analyses in this table. The z-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

We find that the coefficient on PriorityEmployer is negative and significant at the 0.01 level when the dependent variable is AQTBr_All or AQTBr_Client, and negative and significant at the 0.05 level when the dependent variable is AQTBr_Others. Hence, the results indicate that, compared to auditors who prioritise the regulators, those who prioritise the employer report more AQTBs, i.e. a lower quality audit process. It may be that prioritising the employer is strongly linked to the commercial view on auditing, and hence these auditors are more affected by the time deadline pressures, for example.

Moreover, when the dependent variable is AQTBr_Client, auditors prioritising client (PriorityClient), shareholders (PriorityShareholder) and themselves (PrioritySelf) also report a lower quality audit process in terms of client-related AQTBs, compared to auditors prioritising regulators. These results suggest that professional view that emphasises the regulator is associated with least AQTBs, which could be expected, since the AQTBs capture various elements highlighted in the auditing standards.

However, when the dependent variable is OAQ, the results indicate that none of the priority variables is statistically significant. We suggest that OAQ in itself is difficult to interpret without a benchmark (such as AQTB), because of the various views on audit quality. As pointed out by Brivot et al. (2018), audit quality is a multifaceted and ambiguous social construct. Overall, our findings suggest that certain priorities can explain the differences in the reported AQTBs, which are based on more specific questions about how actual audit procedures and gathering of audit evidence were performed during the audit.

Panels B and C of Table 3 present the results of regressing individual AQTBs on priorities. These results indicate that, compared to the group of auditors prioritising the regulators, PriorityEmployer and PriorityClient are associated with significantly more AQTBs related to ‘reducing the amount of work performed below a reasonable level (AQTB1r)’, ‘greater than appropriate reliance on client work (AQTB5r)’, ‘accepting weak client explanations (AQTB6r)’ and ‘making superficial reviews on client documents (AQTB7r)’. AQTB6r and AQTB7r are more frequent also among auditors prioritising shareholders, compared to auditors prioritising regulators. Furthermore, auditors who consider themselves as the highest priorities are also associated with more AQTB6r.

To compare to the paper by Svanberg and Öhman (2015), they find that auditors’ client identification is associated with more AQTBs, but auditors’ professional identification is not associated with AQTBs. It hence seems that the commercial views on auditing are quite influential when it comes to behaviours in the audit process. Our study complements that study and provides new insights on this issue by offering evidence on the role of several relevant stakeholder priorities in the occurrence of different types of AQTBs.

The results of the control variables indicate that Female has marginally significant positive coefficients for some regressions, where the AQTBr is the dependent variable. This suggests that female auditors report fewer AQTBs compared to male auditors. This aligns with the audit literature that female auditors deliver higher quality (e.g. Cameran et al. 2018).

In column (4) of Table 3 Panel A, where the dependent variable is OAQ, Partner and Manager are again significant and positive, similar to the results reported in Table 2. Moreover, regarding the results of individual AQTBs in Panels B and C, the associations between auditor ranks and AQTBs vary. Compared to associates, partners report more frequent AQTB5 and AQTB1, but fewer AQTB4.

4.3 Results for RQ3

Table 4 presents the results of the association between auditor priorities and the match between OAQ and AQTB. Column (1) shows the results when the dependent variable is OAQ_AQTBr. Column (2) shows the results for a sub-sample of observations that range from zero (i.e. match) to positive values (OAQ_AQTBr≥0). Column (3) shows the results for a sub-sample of observations that range from zero (i.e. match) to negative values (OAQ_AQTBr≤0).

Table 4: Main results for RQ3

	(1) Total sample	(2) OAQ_AQTBr≥0	(3) OAQ_AQTBr≤0
	OAQ_AQTBr	OAQ_AQTBr	OAQ_AQTBr
PriorityClient	0.021	0.564***	−0.290
	(0.07)	(2.74)	(−1.37)
PriorityEmployer	0.441	0.650***	−0.003
	(1.39)	(3.10)	(−0.01)
PriorityShareholder	−0.017	0.489**	−0.506**
	(−0.05)	(2.04)	(−2.00)
PrioritySociety	−0.253	0.196	−0.237
	(−0.59)	(0.60)	(−0.84)
PrioritySelf	0.392	0.397	0.026
	(1.15)	(1.63)	(0.09)
PriorityOthers	0.297	0.468**	0.113
	(1.00)	(2.22)	(0.48)
Female	−0.202	−0.252***	−0.052
	(−1.59)	(−2.78)	(−0.57)
Partner	0.394***	0.028	0.419***
	(2.62)	(0.27)	(3.93)
Manager	0.256*	0.105	0.167*
	(1.87)	(1.06)	(1.75)
HigherEducation	−0.161	−0.379	−0.222
	(−0.58)	(−1.60)	(−1.00)
EngExperience	0.007	−0.002	0.004
	(0.62)	(−0.32)	(0.58)
Constant	0.135	0.966***	−0.909***
	(0.31)	(3.13)	(−2.90)
N	251	159	144
Adj. R^2	0.073	0.005	0.025

This table presents the results of regressing the difference between overall audit quality (OAQ) and the average reversed AQTBs (OAQ_AQTBr) on auditors’ stakeholder priorities and control variables. All the variables are defined in Appendix 1. We use OLS regression in column (1), and truncated OLS in columns (2) and (3), and include team fixed effects in all the analyses in this table. The t-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

In column (1), we do not find significant coefficients on auditors’ stakeholder priority. The results in column (2) demonstrate that PriorityEmployer and PriorityClient have positive coefficients, statistically significant at the 0.01 level. This suggests that auditors prioritising the employer or client tend to over-assess the engagement quality, whereas auditors prioritising the regulators are closer to the match between OAQ and AQTB. The results of PriorityShareholder indicate that auditors prioritising the shareholders tend to over-assess [column (2)], but also under-assess [column (3)] engagement quality.

Regarding the control variables, Female is negative and significant in column (2). This result suggests that female auditors’ OAQ assessments are more likely to be aligned with the AQTBs. Moreover, Partner and Manager are positive and significant in columns (1) and (3), suggesting that associates tend to under-assess engagement quality.

4.4 Additional tests on auditor rank

As auditors in different ranks have different skills and perform different audit tasks, they may have varying views on audit quality and on auditing behaviours (e.g. Cameran et al. 2018). For example, higher rank auditors have been found to have more commercial views on auditing (e.g. Suddaby et al. 2009). To examine whether the results are the same or how the results may vary for different ranks of auditors, we examine RQ1, RQ2 and RQ3 for the sub-samples of partners, managers and associates. Because we have only one auditor per rank in most of our teams, we cannot compare auditors in the same rank of one team, which means that we cannot include the team fixed effect. As different teams conduct audits for different clients and clients have different characteristics, we control for client characteristics in addition to auditor attributes for these subsamples.⁴¹ However, the analyses are likely less powerful without the team fixed effect, because with team fixed effects, all the team and clients fixed characteristics are controlled.

The descriptive statistics for the client characteristics are presented in Panel A of Table 5. The mean value of client size (LnTA) is 14. In the sample, 24% of the clients are publicly listed (Public). The average debt ratio (Leverage) is 0.57, and client firms’ sales growth (SalesGrowth) is 0.18 on average. Clients have an average return on assets (ROA) of 0.04, with a quite large variation (standard deviation of 0.17). The average number of years since the client was registered in the system of the audit firm (Tenure) is 12.6.

Table 5: Additional tests for RQ1 – different ranks

Panel A: Descriptive statistics for client characteristics

	N	Mean	SD	p5	p25	p50	p75	p95
LnTA	101	14	2	10.89	12.64	14.11	15.44	17.41
Public	101	0.24	0.43	0	0	0	0	1
Leverage	101	0.57	0.26	0.09	0.36	0.62	0.76	0.93
SalesGrowth	101	0.18	1.14	−0.26	0	0.02	0.12	0.47
ROA	101	0.04	0.17	−0.18	0	0.04	0.1	0.25
Tenure	101	12.6	5.06	4	9	13	17	19

Panel B: RQ1 for different ranks
	(1)	(2)	(3)	(4)
	Total sample OAQ	Sample of partners OAQ	Sample of managers OAQ	Sample of associates OAQ
AQTBr_All	1.225***	1.443***	1.054***	1.407***
	(5.48)	(3.09)	(2.98)	(3.45)
Female	0.154	0.166	0.147	0.296
	(0.67)	(0.35)	(0.40)	(0.69)
Partner	0.884***
	(2.92)
Manager	0.513*
	(1.91)
HigherEducation	0.269	1.275	−1.053	1.085
	(0.53)	(1.21)	(−1.34)	(1.25)
EngExperience	0.052**	0.069	0.033	0.077*
	(2.01)	(1.40)	(0.59)	(1.69)
LnTA	0.053	0.103	0.029	0.139
	(0.84)	(0.76)	(0.30)	(1.01)
Public	0.143	0.407	−0.062	−0.101
	(0.50)	(0.65)	(−0.14)	(−0.18)
Leverage	0.225	0.439	−0.394	0.767
	(0.59)	(0.48)	(−0.77)	(0.84)
SalesGrowth	−0.130	1.017	−0.069	−0.201
	(−1.14)	(1.48)	(−0.41)	(−1.29)
ROA	0.737	2.285	−0.294	1.496
	(1.21)	(1.54)	(−0.32)	(1.36)
Tenure	0.055**	0.015	0.063*	0.064
	(2.37)	(0.31)	(1.68)	(1.53)
N	335	98	130	106
Pseudo R^2	0.091	0.116	0.064	0.128

Panel A reports the descriptive statistics of client characteristics. Panel B presents the results of regressing overall audit quality (OAQ) on the average reversed AQTBs (AQTBr_All) and control variables for auditors in different ranks. We use the full sample of 335 observations. We first present the results for the full sample (column 1) before we report the results for the partner rank, manager rank, and associate rank, respectively, in the last three columns. All the variables are defined in Appendix 1. We use ordered logistic regression in all the analyses in this table. The z-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

As we are not comparing auditors in the same team (i.e. do not include the team fixed effect), we do not need to require at least two auditors in the same team. Therefore, we include the full sample (335 observations) in this subsection. To test the three RQs for the subsamples, we first report the results for the full sample (Column 1), and then for the partners (Column 2), managers (Column 3) and associates (Column 4), respectively.

4.4.1 Auditor rank and RQ1

Table 5 Panel B presents the results of regressing OAQ on AQTBr_All and auditor and client characteristics. The coefficients on AQTBr_All are positive and significant at the 0.01 level for all four columns. The differences between the coefficients for partners, managers and associates are statistically insignificant.⁴² These results suggest that auditors’ AQTBs are strongly associated with their assessment of the overall quality of their engagements regardless of if the auditors are partners, managers or associates.

4.4.2 Auditor rank and RQ2

The results for RQ2, using subsamples for different ranks, are reported in Table 6. In Panel A, the dependent variable is AQTBr_All. In the first column where we use the full sample, only the coefficient on PriorityEmployer is negative and significant at the 0.1 level. Comparing these results with those in Table 3, where we have used the team fixed effect, the coefficient on PriorityEmployer is negative and significant at the 0.01 level in Table 3. Hence, the results in Table 6 support our expectation that, without the team fixed effects, these tests provide weaker results compared with those in the main analyses.

Table 6: Additional tests for RQ2 – different ranks

Panel A: AQTBr_All is the dependent variable

	(1)	(2)	(3)	(4)
	Total sample AQTBr_All	Sample of partners AQTBr_All	Sample of managers AQTBr_All	Sample of associates AQTBr_All
PriorityClient	−0.143	−0.415	−0.033	−0.222
	(−0.97)	(−1.34)	(−0.12)	(−0.92)
PriorityEmployer	−0.327*	−0.793**	−0.276	−0.048
	(−1.95)	(−2.28)	(−0.94)	(−0.17)
PriorityShareholder	−0.078	−0.164	−0.061	−0.154
	(−0.46)	(−0.49)	(−0.20)	(−0.50)
PrioritySociety	−0.197	−0.588	0.464	−0.474
	(−0.91)	(−1.65)	(1.02)	(−1.00)
PrioritySelf	−0.128	−0.635*	0.388	−0.154
	(−0.74)	(−1.72)	(1.05)	(−0.60)
PriorityOthers	−0.158	−0.392	−0.066	−0.093
	(−1.08)	(−1.30)	(−0.24)	(−0.40)
Female	0.160**	0.066	0.164	0.329***
	(2.57)	(0.58)	(1.58)	(2.65)
Partner	0.012
	(0.15)
Manager	0.015
	(0.20)
HigherEducation	−0.037	−0.261	−0.129	0.325
	(−0.28)	(−1.15)	(−0.61)	(1.21)
EngExperience	0.005	−0.007	0.034**	0.006
	(0.86)	(−0.58)	(2.18)	(0.67)
LnTA	−0.030*	−0.057*	−0.012	−0.066*
	(−1.73)	(−1.81)	(−0.44)	(−1.70)
Public	0.098	0.168	0.112	0.044
	(1.25)	(1.14)	(0.90)	(0.26)
Leverage	0.078	0.144	0.088	0.087
	(0.77)	(0.66)	(0.63)	(0.31)
SalesGrowth	0.046	0.030	0.034	0.057
	(1.33)	(0.19)	(0.68)	(1.00)
ROA	0.092	−0.045	0.180	0.280
	(0.55)	(−0.13)	(0.66)	(0.87)
Tenure	0.001	−0.002	0.000	−0.001
	(0.19)	(−0.21)	(0.00)	(−0.10)
Constant	4.479***	5.463***	4.108***	4.569***
	(14.48)	(9.51)	(8.46)	(7.29)
N	335	98	130	106
Adj. R^2	0.008	0.018	−0.002	0.003

Panel B: OAQ is the dependent variable
	(1)	(2)	(3)	(4)
	Total sample OAQ	Sample of partners OAQ	Sample of managers OAQ	Sample of associates OAQ
PriorityClient	−0.129	1.400	−0.889	−0.065
	(−0.24)	(1.07)	(−0.93)	(−0.07)
PriorityEmployer	0.105	1.586	−1.297	0.619
	(0.17)	(1.09)	(−1.21)	(0.63)
PriorityShareholder	−0.155	2.162	−1.336	−0.985
	(−0.25)	(1.53)	(−1.23)	(−0.94)
PrioritySociety	0.179	0.702	0.390	1.538
	(0.23)	(0.47)	(0.24)	(0.94)
PrioritySelf	0.052	3.733**	−1.935	−0.335
	(0.08)	(2.20)	(−1.42)	(−0.37)
PriorityOthers	−0.089	0.928	−0.801	0.146
	(−0.17)	(0.73)	(−0.83)	(0.17)
Female	0.329	0.391	0.227	0.560
	(1.43)	(0.80)	(0.60)	(1.29)
Partner	0.797***
	(2.63)
Manager	0.497*
	(1.83)
HigherEducation	0.163	0.801	−0.988	1.489*
	(0.33)	(0.79)	(−1.25)	(1.65)
EngExperience	0.055**	0.077	0.070	0.089*
	(2.25)	(1.43)	(1.17)	(1.95)
LnTA	0.018	0.039	0.053	0.038
	(0.27)	(0.29)	(0.52)	(0.27)
Public	0.264	0.702	0.058	−0.256
	(0.92)	(1.14)	(0.13)	(−0.45)
Leverage	0.259	0.863	−0.469	0.827
	(0.67)	(0.92)	(−0.90)	(0.86)
SalesGrowth	−0.079	1.094	−0.010	−0.033
	(−0.69)	(1.48)	(−0.06)	(−0.20)
ROA	0.814	2.413	−0.706	1.562
	(1.34)	(1.64)	(−0.73)	(1.41)
Tenure	0.050**	0.021	0.069*	0.040
	(2.18)	(0.42)	(1.81)	(0.94)
N	335	98	130	106
Pseudo R^2	0.045	0.117	0.043	0.091

This table presents the results of regressing the average reversed AQTBs (AQTBr_All), and overall audit quality (OAQ), respectively, on auditors’ stakeholder priorities and control variables for auditors in different ranks. We use the full sample of 335 observations. We first present the results for the full sample (column 1) before we report the results for the partner rank, manager rank, and associate rank, respectively, in the last three columns. All the variables are defined in Appendix 1. We use OLS regressions for Panel A and ordered logistic regression for Panel B. The z-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

The second column provides the results for partners. Similar to column 1, the coefficient on PriorityEmployer is negative and significant, but at the 0.05 instead of 0.1 level. Moreover, as the coefficient on PrioritySelf is also negative and (marginally) significant, partners that consider themselves as most important also report more AQTBs compared to the benchmark. The next two columns examining managers and associates show no significance on any of the test variables. The results reported in this panel suggest that the association between AQTBs and auditors’ stakeholder priorities is mainly driven by partners, not by managers and associates. Interestingly, it seems that partners’ views on which stakeholder is the highest priority matters for the behaviours conducted in the audit process, while managers’ and associates’ views on stakeholder priority are not relevant for their reported AQTBs. These results provide further evidence on the influential role of partners, and on the varying views of audit quality across partners documented in Brivot et al. (2018).

Panel B presents the results when regressing OAQ on auditors’ stakeholder priorities and control variables. Only partners that consider themselves as the highest priorities have a positive and significant coefficient (t=2.20), suggesting that these partners have a higher perception of their engagement’s OAQ. There is no significance on other test variables.

4.4.3 Auditor rank and RQ3

Table 7 reports the results for RQ3 where we use subsamples based on auditor rank. The dependent variable is OAQ_AQTBr. The first column reports the results when we use all the observations, while the last three columns present the results for partners, managers and associates, respectively. There are significant coefficients for test variables, mostly for partners [column (2)], where all the test variables measuring auditors’ stakeholder priority have positive and significant coefficients, although the significance varies from the 0.10 to 0.01 level. Specifically, the coefficients on PriorityEmployer and PrioritySelf are positive and significant at the 0.01 level, suggesting that, compared to partners that consider regulators as the highest priority, partners viewing employer or themselves as most important have a significantly higher perception of their OAQ relative to their reported AQTBs. The coefficients on partners’ other stakeholders are also positive, though significant at the 0.1 or 0.05 level, indicating that partners considering clients, shareholders and wider society as highest priority also assess their OAQ higher relative to AQTB, compared to the benchmark. Regarding other results of the test variables, only the coefficient on PrioritySelf has a significant and negative coefficient in the subsample of managers [column (3)].

Table 7: Additional tests for RQ3 – different ranks

	(1)	(2)	(3)	(4)
	Total sample OAQ_AQTBr	Sample of partners OAQ_AQTBr	Sample of managers OAQ_AQTBr	Sample of associates OAQ_AQTBr
PriorityClient	0.082	0.763**	−0.255	0.121
	(0.42)	(2.03)	(−0.68)	(0.38)
PriorityEmployer	0.311	1.209***	−0.190	0.223
	(1.39)	(2.85)	(−0.46)	(0.60)
PriorityShareholder	0.009	0.721*	−0.339	−0.225
	(0.04)	(1.77)	(−0.82)	(−0.55)
PrioritySociety	0.280	0.840*	−0.319	0.871
	(0.97)	(1.94)	(−0.51)	(1.38)
PrioritySelf	0.119	1.527***	−0.994*	0.047
	(0.51)	(3.40)	(−1.95)	(0.14)
PriorityOthers	0.125	0.619*	−0.195	0.169
	(0.64)	(1.69)	(−0.52)	(0.54)
Female	−0.028	0.035	−0.071	−0.103
	(−0.34)	(0.25)	(−0.49)	(−0.63)
Partner	0.300***
	(2.79)
Manager	0.176*
	(1.82)
HigherEducation	0.085	0.449	−0.169	0.153
	(0.49)	(1.62)	(−0.58)	(0.43)
EngExperience	0.011	0.025	−0.010	0.016
	(1.40)	(1.61)	(−0.48)	(1.38)
LnTA	0.033	0.070*	0.029	0.054
	(1.44)	(1.82)	(0.77)	(1.05)
Public	−0.004	0.006	−0.108	−0.059
	(−0.03)	(0.03)	(−0.63)	(−0.26)
Leverage	0.007	0.059	−0.226	0.168
	(0.05)	(0.22)	(−1.18)	(0.45)
SalesGrowth	−0.070	0.215	−0.036	−0.081
	(−1.52)	(1.08)	(−0.52)	(−1.07)
ROA	0.222	0.615	−0.364	0.440
	(1.01)	(1.48)	(−0.97)	(1.03)
Tenure	0.017**	0.007	0.022	0.020
	(2.03)	(0.51)	(1.55)	(1.22)
Constant	−1.040**	−2.297***	0.033	−1.494*
	(−2.53)	(−3.29)	(0.05)	(−1.80)
N	335	98	130	106
Adj. R^2	0.050	0.165	−0.038	0.008

This table presents the results of regressing the difference between overall audit quality (OAQ) and the average reversed AQTBs (OAQ_AQTBr) on auditors’ stakeholder priorities and control variables for auditors in different ranks. We use the full sample of 335 observations. We first present the results for the full sample (column 1) before we report the results for the partner rank, manager rank, and associate rank, respectively, in the last three columns. All the variables are defined in Appendix 1. We use OLS regression in all the analyses in this table. The t-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

Overall, the findings suggest that stakeholder priorities play a role mainly in partners’ views on audit quality (here, captured by the match between OAQ and AQTB). This finding aligns with results reported for AQTBs that indicate the importance of partners’ views of priorities in audit work for quality assessments. The final quality assessments made by the partners are likely to impact at what point the team decides to stop gathering further audit evidence to support the audit opinion. In line with Bobek et al. (2015) who studied perceptions of the ethical environment, our findings imply that partners as firm leaders assess the quality of audit work more positively relative to reported AQTBs than non-firm leaders. A strong public interest orientation among these partners may assist in developing an ethical environment that foster professionalism and high audit quality standards (Bobek et al. 2015).

4.5 The moderating effect of second-highest stakeholder priority

Bauer (2015) finds that the harmful influence of auditors’ client identification is alleviated by the salience of their professional identity. Inspired by Bauer (2015), we conduct an additional test in which we examine whether the influence of client prioritisation and employer prioritisation is alleviated when the auditor has also ranked the society/regulators/shareholders high. Specifically, we define a new variable, SecondPriorityUser, which is equal to 1 if the second-highest priority is users of financial statements (i.e. shareholders, regulators or wider society). We then interact this variable with PriorityClient and PriorityEmployer, respectively, and define two interaction variables, PriorityClient_User and PriorityEmployer_User. We regress AQTBr_All, OAQ and OAQ_AQTBr, respectively, on all the types of stakeholder priorities (as included in the analyses for RQ2 and RQ3) and these two interaction terms. The results are reported in Table 8 and show that, when AQTBr_All is the dependent variable in column (1), the coefficient on PriorityClient_User is positive and significant, suggesting that auditors that consider clients as the highest priorities do report fewer AQTBs if their second-highest priority is shareholders, regulators or wider society. This finding aligns with Bauer’s (2015) view that auditors’ client identification is alleviated by their awareness of the importance of users. However, the interaction term PriorityEmployer_User is not significant. The coefficients on PriorityClient and PriorityEmployer are negative and significant, consistent with the results in the main tests. The coefficients on priorities in the second and third columns where the dependent variable is OAQ and OAQ_AQTB are all insignificant except the marginal significance for PriorityClient_User.

Table 8: Additional tests - the moderating effect of the second highest stakeholder priority

	(1)	(2)	(3)
	AQTBr_All	OAQ	OAQ_AQTBr
PriorityClient_User	1.969***	1.365*	−0.960*
	(3.26)	(1.72)	(−1.68)
PriorityEmployer_User	−0.011	−2.027	−0.446
	(−0.01)	(−1.29)	(−0.31)
PriorityClient	−2.076***	−1.559	0.366
	(−2.60)	(−1.47)	(0.44)
PriorityEmployer	−2.660***	0.099	1.300
	(−3.01)	(0.08)	(1.42)
PriorityShareholder	−0.977	−0.662	−0.530
	(−1.11)	(−0.57)	(−0.56)
PrioritySociety	0.235	−1.405	−1.292
	(0.22)	(−1.05)	(−1.03)
PrioritySelf	−0.863	1.008	1.197
	(−0.97)	(0.87)	(1.28)
PriorityOthers	−1.461*	0.194	0.829
	(−1.91)	(0.19)	(1.03)
Female	0.598*	−0.527	−0.607*
	(1.71)	(−1.17)	(−1.80)
Partner	−0.221	1.691***	1.438***
	(−0.56)	(3.21)	(3.68)
Manager	−0.074	1.006**	0.978***
	(−0.20)	(2.05)	(2.66)
HigherEducation	0.541	−0.787	−0.661
	(0.67)	(−0.78)	(−0.87)
EngExperience	−0.005	0.059	0.032
	(−0.13)	(1.37)	(1.08)
N	251	251	251
Pseudo R^2	0.127	0.398	0.121

This table presents the results of regressing AQTBr_All, OAQ, and the OAQ_AQTBr (difference between OAQ and AQTBr_All) on auditors’ stakeholder priorities and control variables. We interact PriorityClient and PriorityEmployer with SecondPriorityUser, and construct two interaction terms PriorityClient_User and PriorityEmployer_User. SecondPriorityUser is equal to 1 if the second highest priority is users of financial statements (i.e. shareholders, regulators or wider society). We use ordered logistic regression and include team fixed effects in all the analyses in this table. The z-statistics are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

5. Conclusion and discussion

This study is the first to investigate auditors’ self-assessment of their own engagement quality. Specifically, we study whether and how auditors’ overall engagement quality assessment is related to their assessment of process quality captured by AQTBs (RQ1), and whether and how the auditors’ assessments are associated with their priorities of stakeholders in audit work (RQ2 and RQ3). We argue that how auditors prioritise different stakeholders in their own work can reflect their view on audits.

There are several important findings in this paper. First, results for RQ1 indicate that, on average, self-reported behaviours in the audit process are reflected in the self-assessment of audit quality, which implies that the performance of procedures and tests in the gathering of audit evidence throughout the audit process impacts the overall quality assessment. This finding holds in each subsample based on the auditor’s rank: partners, managers and associates. Hence, it seems that, in general, the professional view on auditing is prevailing. When investigating each AQTB separately, we find that auditors view certain AQTBs (e.g. premature signing offs, and reduction of the amount of work performed on an audit step below a reasonable level) in the audit process as more relevant for audit quality than others (see also Coram et al. 2008 regarding auditors’ perceptions of the moral intensity of certain AQTBs). In particular, the behaviour ‘audit team members have a greater than appropriate reliance on client work’ appears to be perceived as less relevant, which is intriguing as the source of audit evidence is emphasised in the auditing standards in relation to reliability. It might be that, for auditors, the threshold to conclude that reliance on client work has affected audit quality is quite high, since auditing in general is heavily dependent on the information provided by clients (Jeppesen 1998).

Second, our findings for RQ2 indicate that, compared to auditors who prioritise the regulators, auditors who prioritise the employer report more AQTBs, i.e. lower audit process quality. Our analyses further show that this finding seems mainly driven by AQTBs that explicitly mention the client (i.e. ‘have a greater than appropriate reliance on client work’, ‘accept weak client explanations’ and ‘make superficial reviews of client documents). Reflecting these findings to the literature, it may be that the ‘social consensus’ (Jones 1991) of the commercial view on auditing (Malsch and Gendron 2013; Zeff 2003; Wyatt 2004) is stronger particularly among those auditors who prioritise their employer. Moreover, our findings reveal that the association between auditors prioritising the employer and AQTBs is significant only among the subsample of partners. Hence, the partner’s prioritisation has a significant relation to the behaviours that occur during the audit process. The result is important, since prior studies suggest that partner is the most influential on the team behaviour (e.g. Cameran et al. 2018). We also find that auditors prioritising the client, shareholders, or themselves report more AQTBs that explicitly mention the client, compared to auditors prioritising regulators. While this result can be expected from client prioritisation, the result for shareholder prioritisation further highlights that auditors prioritising the regulators are indeed very strict in following the audit conduct stipulated by the auditing standard. Overall, in addition to the commercial view, those auditors prioritising other stakeholders than regulators may be more affected by the challenge of balancing between maintaining professional view while keeping good relationship with the client for smooth conduct of audit (Guénin-Paracini et al. 2015).

When the dependent variable is OAQ, the results indicate that none of the priority variables is statistically significant – even though significant results could be found when the AQTB is a dependent variable. We assume that the OAQ captures various views of audit quality held by individual auditors that are not related to the priorities as prior studies have proposed that audit quality is a multifaceted construct (e.g. Knechel et al. 2013a; Christensen et al. 2016; Brivot et al. 2018).

Third, results from the test of RQ3 indicate that auditors who perceive the employer or the client as their highest priority assess OAQ higher relative to AQTBr. Also, those prioritising the shareholders both over-assess and under-assess OAQ. The findings for employer and client prioritisation imply that a more commercial view on auditing may lead to over-assessment of the engagement quality where AQTBs are not fully considered. These findings suggest that, as expected, auditors who prioritise the regulators have the closest match between the assessments of OAQ and AQTB. Our analyses of the subsamples based on auditors’ rank indicate significant associations between priorities and the match between OAQ and AQTB, but again only among partners.

Overall, the study findings demonstrate that the audit quality is assessed differently by audit team members working on the same engagement team and that priorities in audit work impact auditors’ quality assessments. These observations offer important insights by revealing certain patterns in auditors’ understanding of audit quality, which is useful for efforts to improve audit quality through training and oversight activities. In this regard, the recent findings by Herda et al. (2019) could be considered. They find that auditors who are coached by supervisors to appreciate the importance of their work to external financial statement users are more likely to be mindful in their work, leading to a reduced likelihood of auditor’s premature sign-off. Their findings and our findings highlight the importance of employer-driven practices that emphasise the professional view on audit.

Moreover, the results of this study indicate significant differences in quality assessment between auditor ranks. Specifically, partners are in general assessing audit quality higher than managers and associates on the same engagements. This is an interesting finding suggesting that there are different (institutional) pressures facing auditors of different ranks that affect, for example, views on commercialism and professionalism (Suddaby et al. 2009) and assessments of audit quality, and that we should not treat all auditors the same but instead consider the incentives and pressures that apply at the different rank levels.

We acknowledge that the study is subject to several limitations. First, this study, like other survey studies, suffers from the inability to lend clarity and nuance to responses. Therefore, we call for future studies to investigate how auditors understand audit quality with the use of a qualitative approach that allows for follow-up questions to gain deeper insights into how and why auditors make certain quality assessments. Second, our study is subject to the limitations inherent in the use of a survey questionnaire, such as the honesty of the responses. For example, auditors may be unwilling to admit to low performance quality, thereby biasing the responses. Third, we assume that AQTBs represent the assessment of audit process quality based on auditing standards. We acknowledge that AQTBs may not be an entirely comprehensive proxy for the audit process quality based on auditing standards, as they may not capture all the elements. Fourth, although we control for important individual auditor characteristics in our models, we do recognise that there may be other confounding factors affecting the quality assessments that we are unable to control for in our analyses.

Acknowledgement

The authors thank seminar participants at the 10th European Auditing Research Network (EARNet) Symposium in Parma, the Virtual Audit and Assurance Conference arranged by British Accounting and Finance Association (BAFA), the European Accounting Association (EAA) 2021 Virtual Congress, and at Umeå School of Business, Economics and Statistics for their comments. We appreciate the comments received from Peter Öhman and thank an anonymous Big 4 audit firm for the assistance in the distribution of the survey. The authors also thank the editor and two anonymous reviewers. Tobias Svanström gratefully acknowledges financial support from Jan Wallanders och Tom Hedelius Stiftelse (Grant P2014-0088 and P2018-0029).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Jan Wallanders och Tom Hedelius Stiftelse samt Tore Browaldhs Stiftelse: [Grant Number P2014-0088 and P2018-0029].

Notes

1 Exceptions are the recent papers by Brivot et al. (2018) and Bell et al. (2015). Brivot et al. (2018) interview audit partners to examine how they understand audit quality. Bell et al. (2015) use data from the assessments of the audit process completed by an audit firm’s internal reviewers.

2 There is much less research on the quality of auditors’ execution of the audit process than on audit quality based on audit outcomes, such as discretionary accruals, meeting or beating earnings benchmarks, or audit opinions (Bell et al. 2015; Gaynor et al. 2016). Although our study does not strive to capture the actual quality of the audit engagement, our investigation of the auditors’ self-assessments shifts the focus from outcome measures closer to the quality of the audit process.

3 This discussion is closely linked to the social identity theory. Prior research has presented that professionals – such as auditors – tend to identify both with their organisation and with their profession (Bamber and Iyer 2002; Hekman et al. 2009), or even with their clients (Bamber and Iyer 2007, Svanberg and Öhman 2015; Bauer 2015).

4 We examine eight AQTBs. The AQTBs specifically mentioning the client capture the extent audit team members ‘have a greater than appropriate reliance on client work’, ‘accept weak client explanations’, and ‘make superficial reviews of client documents’. The other audit procedure-related AQTBs capture the extent audit team members ‘reduce the amount of work performed on an audit step below reasonable level’, ‘sign off an audit-program step without completing the work or noting the omission’, ‘make an unauthorized reduction of sample size’, ‘under-report audit time’ and ‘fail to investigate an accounting principle’.

5 For example, Willett and Page (1996) and Lee (2002) in the UK; Otley and Pierce (1996a, 1996b) and Pierce and Sweeney (2004), Sweeney et al. (2013) and Sweeney and Pierce (2015) in Ireland; Coram et al. (2008) in Australia; Alderman and Deitrick (1982), Kelley and Margheim (1990), Malone and Roberts (1996), Donnelly et al. (2003), and Herda et al. (2019) in the US; Svanberg and Öhman (2013, 2015, 2019) in Sweden.

6 AQTBs could stem from time pressure, that is, the pressure to reduce the costs of the audit (e.g. Pierce and Sweeney 2004), but also from other factors such as lack of competence or training (Svanström 2016) and ethical culture at the audit firm (Svanberg and Öhman 2013).

7 Basically, all the AQTBs (except under-reporting audit time) should be avoided to comply with the International Standard of Auditing (ISAs). We hence assume that AQTBs largely capture the regulators’ view of audit quality.

8 The auditing standards stipulate that the reliability of audit evidence increases if the evidence is collected from a source outside the audited entity (ISA 500, A.31) and that the (skeptical) auditor should question contradictory audit evidence and the reliability of documents and responses to inquiries and other information obtained from management and those charged with governance (ISA 500, A.20).

9 Specifically, we add to the literature on auditor-specific determinants of AQTB (e.g., Peytcheva and Gillett 2012; Donnelly, Quirin, and O’Bryan 2003; Malone and Roberts 1996; Herda et al. 2019; Svanberg and Öhman 2015).

10 Knechel et al. (2013a) present the audit quality indicators in four categories: inputs, process, outputs and context. Common audit quality inputs include measures of competence and independence, such as auditor/audit firm tenure, experience and industry specialisation (Knechel et al. 2013a; DeFond and Zhang 2014). Christensen et al. (2016) find that investors emphasise the role of auditor characteristics as inputs for audit quality.

11 Furthermore, Aobdia (2019) document that most of the traditionally used audit quality proxies in research are not significantly associated with the two practitioner-related measures of audit process quality (i.e. audit firms’ internal inspections and PCAOB inspections).

12 For example, the findings of Carcello et al. (1992) indicate that audit partners considered a sceptical attitude (firm responsiveness to client needs) to be more (less) important than the preparers did. Moreover, partners considered compliance with general audit standards to be less important compared to both preparers and users. Another study by Gonthier-Besacier et al. (2016), however, argues that the differences in perception of audit quality go beyond the role-perception-gap. They report that the differences depend on professional characteristics such as expertise and shared values between the professionals.

13 The audit process represents the implementation of audit inputs, where the audit team applies the testing procedures (Francis 2011).

14 Martinov and Roebuck (1998) study how inherent risk and materiality assessments are made and how they are integrated in the planning of audit testing by using audit manuals, other decisions aids and interview data.

15 In some jurisdictions like Sweden, the group of interested users of audits are defined broadly and also include the wider society (Diamant 2004, p. 110). Auditing may potentially reduce tax errors and economic crime and warn the general public about forthcoming events and uncertainties (e.g. going-concern opinion).

16 For example, empirical evidence shows that receiving a disciplinary sanction (as an outcome of a regulatory inspection documenting insufficient audit work) leads to an unfavourable salary development for Big 4 auditors (Sundgren and Svanström 2017). At the same time, auditors are compensated for attracting new clients and keeping existing clients (Knechel et al. 2013b).

17 The clients also exercise direct and indirect pressure, often related to audit opinions that may include threats to dismiss and replace the auditor. Empirical studies have found evidence in support of client pressures paying off, at least in some settings, in terms of ‘successful shopping of audit opinions’ (Lennox 2000; Chen et al. 2016; Chung et al. 2019).

18 Brivot et al. (2018) interviewed partners of Big 4 firms and non-Big 4 firms and found two different perspectives on audit quality to dominate. Partners in Big 4 firms that audit public companies endorse a highly formalized process to accomplish a technically flawless and perfectly documented audit (i.e. ‘model’ audit quality convention), whereas partners working primarily on private company audits endorse instead the ‘‘value-added’’ audit quality convention where the audit is tailored to meet the client’s needs (Brivot et al. 2018).

19 Audit standards describe numerous procedures requiring that the auditor is not uncritically accepting clients’ standpoints. For example, ‘When using information provided by the audited entity, the auditor should assess whether the information is sufficient for the purpose of an audit’ (ISA 500, p.9), ‘reliability of audit evidence increases if the evidence is collected from a source outside the audited entity’ (ISA 500, A.31) and the auditor should question contradictory audit evidence and other information obtained from management and those charged with governance (ISA 500, A.20).

20 An underlying assumption in much of the earlier audit research has been that auditors, due to their common role, share similar views on auditing and audit quality (Litjens et al. 2015; Porter 1993). However, audit research document variation between individual auditors in both (actual) audit quality (Che et al. 2018; Cahan and Sun 2015) and perceptions about audit quality (Svanberg and Öhman 2019; Svanberg and Öhman 2015; Sweeney et al. 2013).

21 Moral reasoning reflects a person’s general tendencies to think about moral dilemmas in a particular way (Kohlberg and Kramer 1969). In audit and accounting research, moral reasoning has been positively associated with, for example, the ability to interpret ethical dilemmas (Arnold and Ponemon, 1991; Shaub and Lawrence, 1996; Uddin and Gillett, 2002).

22 ‘A moral issue is present where a person’s actions, when freely performed, may harm or benefit others’ (Jones 1991, p. 367). An auditor may, for example, have pressure to meet a time budget and therefore intentionally choose to reduce evidence gathering, which may ultimately be harmful for the users of financial statements.

23 Also, Malone and Roberts (1996) investigated whether organisational commitment or professional commitment are associated with AQTBs, but they did not find significant results.

24 Before the survey was sent to the participants, two audit partners signed an internal email sent to the invited auditors to encourage survey participation and to inform them about which specific audit engagement to consider when responding. The survey was delivered on December 8th, 2016 and closed by December 23rd, 2016, so the memory of the audit would not be too vague for the participants because the experience that they had to think about was very recent, 07/2015-06/2016.

25 Online surveys could reduce bias due to issues related to researcher-respondent contact (Jones III et al., 2010).

26 According to the regulation, no ethical review or permission was required because the survey questionnaire does not include information that is considered sensitive by regulators. The survey questionnaire was distributed from Umeå University and has been discussed at research seminars at the University.

27 Non-response was analysed with the use of ANOVA to compare differences between the first 50 and last 50 respondents’ results (Larson and Catton 1959), which gives a probable direction of any non-response bias. Results showed no significant difference between early and late respondents, which indicates low risk for non-response bias.

28 The audit teams in our final sample have responded based on experiences from engagements in either public or private clients.

29 There are 84 teams that have only one responding auditor per team.

30 There have been no national adjustments to the ISA standards in Sweden since 2011. Inability to meet the (quality) standards may results in disciplinary sanctions issues by the Swedish Inspectorate of Auditors (SIA). The most severe sanction is withdrawal of the audit certificate followed by warning and reprimand.

31 Research findings in Sundgren and Svanström (2017) indicate a rather limited impact of regulators in Sweden as they find that auditor reporting behavior and client losses do not change significantly after an auditor has received a disciplinary sanction.

32 For audit quality threatening behaviours (AQTBs), we use the following question: ‘Previous research that has studied auditors' behaviours has found that there are several different types of behaviours that occur during an audit. From 1 (Never) to 5 (Always), please indicate how often the audit team members of the specified audit engagement conducted the behaviours listed below: 1) Reduce the amount of work performed on an audit step below what you consider reasonable; 2) Under-report audit time; 3) Sign off an audit-program step without completing the work or noting the omission; 4) Make an unauthorised reduction of sample size; 5) Have a greater than appropriate reliance on client work; 6) Accept weak client explanations; 7) Make superficial reviews of client documents; and 8) Fail to investigate an accounting principle.’

33 The advantage of ordered logistic model is that it takes into account the different distance between each category. A common view of ordinal variables is that they are non-strict monotonic transformations of interval variables (e.g. O'Brien 1981). That is, one or more values of an interval-level variable are mapped into the same value of a transformed, ordinal variable. For example, a Likert scale may place individuals in one of a number of ranked categories, such as, "strongly agree," "somewhat agree," "neither agree nor disagree," "somewhat disagree," or "strongly disagree" with a statement. An underlying, continuous variable denoting individuals' degrees of agreement is mapped into categories that are ordered but are separated by unknown distances (Winship and Mare 1984).

34 The variable Female may also capture other auditors’ attributes, such as risk aversion. It is likely that female auditors are more risk averse than male auditors (e.g. Ittonen et al. 2013).

35 The rank ‘director’ is for auditors that have similar audit experience to partners, but do not have ownership of the firm. When we do not consider the director as a partner, the results hold.

36 Because partners (and managers) have more audit experience than associates, the variables for auditor rank are highly correlated with the auditor's professional experience. In additional tests, we replace auditor rank variables with audit experience, measured by the number of years of audit experience, and the inferences from the results remain similar to those from the main results.

37 Note that the values of AQTBs are reversed for convenience.

38 To understand whether certain types of auditors consider specific stakeholders as their highest priority, we regress each variable of stakeholder priority on auditor attributes. There are however no significant associations except for when auditors consider themselves as most important. We note that female auditors and auditors with higher ranks are less likely to consider themselves as the highest priority. These results indicate that the variation in priorities is not determined by the typical auditor characteristics. Instead, it could be that priorities are, for example, associated with different personality traits of auditors.

39 However, the differences between the coefficients for the three regressions in Panel A are not statistically significant.

40 The higher OAQs for the higher auditor ranks may be due to two reasons: (1) greater responsibility for the audit outcome (they need to proceed with the gathering of audit evidence until they themselves believe that they have sufficient audit evidence to conclude that there are no material misstatements in the financial statements); and (2) more information about the gathered audit evidence (a partner gets full access to all audit evidence gathered and is basing her assessment on full information – although they might not fully consider it all – while other auditors might have more limited knowledge about the audit evidence gathered).

41 We receive a limited number of client characteristics from the audit firm assisting us with the data. Anonymity was assured by using an identifier number.

42 The p-values for the differences between partners and managers, between partners and associates, and between managers and associates are 0.96, 0.53 and 0.50, respectively.

Appendix 1: Variable Definitions

Table

OAQ	=	Overall Audit Quality based on the question ‘Relative to your other engagements with similar client characteristics, how would you rate the overall audit quality of this engagement, from 1 (low) to 5 (high)?’
AQTB1	=	The extent audit team members reduce the amount of work performed on an audit step below a reasonable level, where 1 (5) indicates never (always).
AQTB2	=	The extent audit team members under-report audit time, where 1 (5) indicates never (always).
AQTB3	=	The extent audit team members sign off an audit-program step without completing the work or noting the omission, where 1 (5) indicates never (always).
AQTB4	=	The extent audit team members make an unauthorised reduction of sample size, where 1 (5) indicates never (always).
AQTB5	=	The extent audit team members have a greater than appropriate reliance on client work, where 1 (5) indicates never (always).
AQTB6	=	The extent audit team members accept weak client explanations, where 1 (5) indicates never (always).
AQTB7	=	The extent audit team members make superficial reviews of client documents, where 1 (5) indicates never (always).
AQTB8	=	The extent audit team members fail to investigate an accounting principle, where 1 (5) indicates never (always).
AQTB1r	=	Reversed AQTB1, where AQTB1r = 1, 2, 4, and 5, if AQTB1 = 5, 4, 2, and 1, respectively.
AQTB2r	=	Reversed AQTB2, where AQTB2r = 1, 2, 4, and 5, if AQTB2 = 5, 4, 2, and 1, respectively.
AQTB3r	=	Reversed AQTB3, where AQTB3r = 1, 2, 4, and 5, if AQTB3 = 5, 4, 2, and 1, respectively.
AQTB4r	=	Reversed AQTB4, where AQTB4r = 1, 2, 4, and 5, if AQTB4 = 5, 4, 2, and 1, respectively.
AQTB5r	=	Reversed AQTB5, where AQTB5r = 1, 2, 4, and 5, if AQTB5 = 5, 4, 2, and 1, respectively.
AQTB6r	=	Reversed AQTB6, where AQTB6r = 1, 2, 4, and 5, if AQTB6 = 5, 4, 2, and 1, respectively.
AQTB7r	=	Reversed AQTB7, where AQTB7r = 1, 2, 4, and 5, if AQTB7 = 5, 4, 2, and 1, respectively.
AQTB8r	=	Reversed AQTB8, where AQTB8r = 1, 2, 4, and 5, if AQTB8 = 5, 4, 2, and 1, respectively.
AQTBr_All	=	The mean value of the (reversed) responses to the eight questions about AQTBs
AQTBr_Client	=	The mean value of the (reversed) responses to the questions about AQTBs that are related to audit clients
AQTBr_Others	=	The mean value of the (reversed) responses to the questions about AQTBs that are not related to audit clients
OAQ_AQTBr	=	The difference between OAQ and reversed AQTB_All
PriorityClient	=	An indicator variable, which equals 1 if the auditor considers the audit client as the highest priority, and 0 otherwise.
PriorityEmployer	=	An indicator variable, which equals 1 if the auditor considers audit firm, audit office or audit team as the highest priority, and 0 otherwise.
PriorityShareholder	=	An indicator variable, which equals 1 if the auditor considers the shareholders as the highest priority, and 0 otherwise.
PrioritySociety	=	An indicator variable, which equals 1 if the auditor considers the wider society as the highest priority, and 0 otherwise.
PriorityRegulator	=	An indicator variable, which equals 1 if the auditor considers the regulators as the highest priority, and 0 otherwise.
PrioritySelf	=	An indicator variable, which equals 1 if the auditor considers herself as the highest priority, and 0 otherwise.
PriorityOthers	=	An indicator variable, which equals 1 if the auditor considers others as the highest priority or has a missing value, and 0 otherwise.
Female	=	An indicator variable, which equals 1 if the auditor is female and 0 otherwise.
Partner	=	An indicator variable, which equals 1 if the auditor is a partner or a director, and 0 otherwise.
Manager		An indicator variable, which equals 1 if the auditor is a senior manager, manager or associate manager, and 0 otherwise.
HigherEducation	=	An indicator variable, which equals 1 if the auditor has a master’s degree or above, and 0 if she has a bachelor’s degree.
EngExperience	=	The respondent’s experience of the audit engagement in years.