2.1 Management's expert

The Australian auditing standard ASA 500 (identical to the equivalent International Standard on Auditing ISA 500) defines an ME as ‘an individual or organisation possessing expertise in a field other than accounting or auditing, whose work in that field is used by an entity to assist the entity in preparing the financial report’ (ASA 500, para 5e). Examples of expertise that may be classified as being provided by MEs include the valuation of complex financial instruments, land and buildings, plant and machinery, jewellery, art, intangible assets, assets and liabilities acquired in business combinations, information used for impairment assessments, estimation of resources such as oil and gas reserves, valuation of environmental liabilities, actuarial calculations, provision of engineering data required in the valuation of roads and the interpretation of contracts and laws (Agrawal, Tarca, & Woodliff, 2020; AUASB, 2020; Chartered Professional Accountants Canada, 2019; South African Institute of Chartered Accountants, 2020).

If an auditor intends to use information provided by an ME as audit evidence, then auditing standards require the auditor to evaluate the capabilities of the ME to undertake the assigned task along with the appropriateness and reasonableness of the work undertaken by the ME (ASA 500, para 8). In spite of the guidance provided by the auditing standards, the ASIC continues to find instances where the audit files did not contain sufficient evidence to support auditors' reliance on MEs [ASIC's Audit Inspection Reports issued in 2011 (para 72,74), 2012 (para 47), 2014 (para 76), 2015 (para 86), 2017 (para 90), 2019a (para 102), 2019b (para 108)]. This suggests that auditors tend to over-rely on MEs.

To address concerns raised by audit practitioners about the lack of authoritative guidance on the use of experts and their work (Bratten, Gaynor, McDaniel, Montague, & Sierra, 2013; Griffith, 2013; Griffith et al., 2015) and the regulator about auditors' over-reliance on MEs, the AUASB issued additional guidance by revising its Guidance Statement GS 005 in March 2015 and then again in March 2020.³ The latest revisions to GS 005 provide additional guidance on procedural matters in relation to determining when an individual is considered as an ME, identifying when and how management uses MEs, assessing an ME's credibility, and on issues auditors have to consider when evaluating the work of MEs (AUASB, 2020). The updates aim at enhancing audit quality in the area of the auditor's use of MEs, improving consistency among audit firms of all sizes, reflecting requirements of current and revised auditing standards [ASA 620, 2010, ASA 540, 2018b (revised), and ASA 500] and aligning the guidelines to the Public Company Accounting Oversight Board amendments to the auditing standard for using the work of specialists (AS 1105, 2020, Appendix A).

2.2 Early-stage conversation

One of the recommendations made by GS 005 is for auditors to have an early-stage conversation with MEs and confirm whether the ME's intended approach and methodology is appropriate for financial reporting purposes (GS 005, para 22).⁴ The aim is to ensure that all involved parties agree upon the use of the ME's work and the expert's approach towards the task.

Based on interview findings, Agrawal (2018) reported that although an early-stage conversation is recommended by GS 005, auditors do not always consider it necessary to engage with the ME in the early stages. Auditors are more likely to initiate an early-stage conversation with an ME when the task is a one-off, unusual or complex. The author also reports that although the primary aim of an early-stage conversation may be to understand and to agree upon the experts' intended methodology and approach towards the task, auditors often use this opportunity to ensure that the ME understands the purpose and intended use of their work. In addition, auditors clarify their expectations regarding the extent of documentation and the depth of information required. In return, they gain insights into the ME's mindset and views regarding the assigned task. Therefore, an early-stage conversation with an ME appears to allow auditors to get a better understanding of what to expect and to raise concerns, if any, in a timely manner. Given the cost involved in engaging MEs, sorting out the issues prior to the commencement of the task is presumably beneficial to all the parties involved.

3.1 Impact of early-stage conversation on perceived client pressure

DeZoort and Lord (1997, p. 47) defined client pressure as ‘the pressure to yield, or the anticipation of the pressure to yield, to a client's wishes or influence whether appropriate or not’. Similarly, Brown and Johnstone (2009, p. 67) defined client pressure as the ‘extent of persuasion believed to be necessary to make the client move towards a more conservative alternative’. Prior literature identifies various economic considerations and environmental conditions that may exert pressure on auditors to acquiesce to clients' interests. For example, research finds that auditors' economic interests (Felix, Gramling, & Maletta, 2005; Hatfield et al., 2011; Trompeter, 1994); client pressure to reduce fees (Ettredge, Fuerherm, & Li, 2014; Gramling, 1999); fear of losing a client (Arnold, Bernardi, & Neidermeyer, 1999); lack of authoritative guidance (Salterio & Koonce, 1997); subjectivity of the accounting issue (Hackenbrack & Nelson, 1996); audit tenure and prior knowledge of a client's position (Brown & Johnstone, 2009; Ng & Tan, 2007; Salterio & Koonce, 1997); and perceived client flexibility (Gibbins et al., 2010; Hatfield, Houston, Stefaniak, & Usrey, 2010) can result in bias surviving the audit process. Because ‘audit firms' commercial success requires profitable audit engagements’ (Messier & Schmidt, 2018, p. 338), motivations to meet clients' expectations and accommodate their interests continue to persist.

Agrawal et al. (2020) raise concerns that MEs' perceived cognitive authority status increases the risk of insufficient scepticism exercised by auditors in seeking and evaluating audit evidence. This is because auditors tend to perceive MEs as being specialists in their field, so consequently ‘know what they are talking about’ and ‘know something that auditors do not know’. This perceived credibility of MEs enhances the reliability of the information provided by MEs (Griffith, 2015) and, thus, makes it difficult for auditors to challenge them (Cannon & Bedard, 2017). Following on from the prior literature, we argue that the inherent difficulty in persuading the client to make changes to the work of someone who is considered as an expert further increases when auditors fail to uncover the issue(s) during the early-stage conversation.

The primary purpose of an early-stage conversation is to agree on the intended methodology and basis of valuation prior to the ME undertaking the task (GS 005). Hence, it would not be unreasonable to assume that a client might expect auditors to discuss and resolve all potential issues with the ME during the early-stage conversation and not raise them once the work is completed. The anticipation that the client may not favourably look upon an auditor raising issues after the ME has completed their work will create a pressure to yield to the client's wishes of a no-surprise audit. The adverse impact of an early-stage conversation on perceived client flexibility to proposed audit adjustments presents yet another condition of client pressure that an auditor needs to withstand.

Hence, by measuring client pressure as the auditors' perception of client flexibility to an adjustment proposed to the work of the ME, we propose that

3.2 Impact of an early-stage conversation on auditors' acquiescence to a client's interests

In her seminal work in social psychology, Kunda (1990) suggested that goals or motives can affect reasoning. Kunda (1990) proposed that the motivated reasoning phenomena are driven by two types of goals: accuracy and directional. When people are motivated to be accurate, they are likely to put more cognitive effort into issue-related reasoning, attend to information more carefully and process it more deeply (Kunda, 1990). When people are motivated to arrive at a desired conclusion, they attempt to construct a reasonable justification for their preference. In other words, their belief construction, evidence collection and evaluation become biased by directional goals (Kunda, 1990). Kunda (1990) argued that motivated reasoners make self-interested goal-consistent judgements but they are subject to the ‘reasonableness constraint’. That is, they cannot display unrestrained self-interest but attempt to be rational and construct a justification for their desired conclusion that would persuade a ‘dispassionate observer’.

Both the accuracy and directional goals are present in an audit context. Auditors are responsible for ensuring that a client's financial statements are free from material misstatement to warrant an unqualified audit opinion. Failure to provide an appropriate (‘accurate’) audit opinion may result in disciplinary actions against the firm (Lennox & Li, 2020) and cause reputational damage to both the audit firm and its clients (Chaney & Philipich, 2002; Davis & Simon, 1992; Dechow, Ge, & Schrand, 2010; Dechow, Sloan, & Sweeney, 1996; Wilson & Grimlund, 1990). Hence, auditors' actions and decisions will be driven by accuracy goals. However, research also indicates that in an auditor–client relationship, auditors are the relationship managers whose job is also to see that the client remains ‘happy’ (McCracken, Salterio, & Gibbins, 2008). The damage that a confrontation would cause to the auditor–client relationship and client retention exposes auditors to directional goals and may result in introducing a subconscious bias into auditors' judgements, decisions and actions. The ‘reasonableness’ criteria would however require auditors to be able to justify their decision as being unbiased and reasonable.

Consistent with the theory of motivated reasoning, Blay (2005) found that auditors' interpretation of information related to issues of going concern aligns with management when they perceive a threat of losing the engagement. Similarly, Hackenbrack and Nelson (1996) observed that auditors interpret accounting standards to justify a client's aggressive financial reporting. Ng and Tan (2007) found that auditors' propensity to insist on an adjustment is reduced in the presence of expressed client concerns, particularly when the audit difference is subjective and the auditors' decision to acquiesce to the client can be reasonably justified. Various archival studies also provide evidence that auditors are more likely to accept client-preferred outcomes (i.e. driven by directional goals) when the audit difference is subjective (Braun, 2001; Libby & Kinney, 2000; Nelson, Elliott, & Tarpley, 2002; Ng, 2007; Wright & Wright, 1997). The subjectivity involved not only provides an opportunity to rationalise the client's preference but also makes it difficult for auditors to define what is reasonable and appropriate and, consequently, justify and insist on an adjustment.

As noted earlier, an early-stage conversation allows auditors to work with the ME and to confirm whether the ME's intended approach and methodology is appropriate for financial reporting purposes. A satisfactory early-stage conversation will consequently assist the auditor in planning the audit. It is also reasonable to assume that a client might expect auditors to discuss and resolve all potential issues with the ME during an early-stage conversation and not raise them once the work is completed. Taken together, a subjective audit difference, perceived high capabilities of an ME and perceived pressure to yield to the client's expectations of a no-surprise audit (because of an early-stage conversation) may weaken an auditor's relative bargaining power. Other things being equal, this may reduce the likelihood of auditors succeeding in a negotiation. As depicted in Figure 2, this may trigger the adoption of self-interest directional goals (e.g. preventing confrontation and keeping the client happy) that facilitate a greater likelihood of auditors accommodating the client's preferences, provided that support for the decision can be reasonably provided.

Accordingly, we propose that

3.3 Auditor–client relationship

Auditor–client negotiations are not a one-time occurrence (Beattie, Fearnley, & Brandt, 2000; Gibbins, Salterio, & Webb, 2001). Past experiences with a client will allow auditors to predict how a particular client is likely to react to a proposed adjustment and be better prepared for the negotiation (Brown & Wright, 2008; Brown-Liburd & Wright, 2011; Gibbins & Salterio, 2000; Trotman, Wright, & Wright, 2005). When an auditor faces the prospect of negotiating with a difficult and contentious client, the auditor is likely to anticipate that the client will challenge the auditor's position and strongly object to any adjustment proposed. However, as a relationship manager, it is the role of the auditor to come up with strategies to negotiate (Hatfield, Agoglia, & Sanchez, 2008), determine appropriate persuasion tactics (Perreault & Kida, 2011), develop potential solutions that will satisfy the client, ensure at least minimum compliance to the GAAP and maintain the ongoing relationship with the client (McCracken et al., 2008). Accordingly, many researchers have investigated the impact of the auditor–client relationship on the negotiation process. For example, a past auditor–client relationship can influence auditors' pre-negotiation planning (Brown-Liburd & Wright, 2011), auditors' negotiation strategies (Brown-Liburd & Wright, 2011; Fu, Tan, & Zhang, 2011; Gibbins et al., 2010; Hatfield et al., 2008) and the eventual outcome of negotiations (Fu et al., 2011; Gibbins et al., 2001; Ng & Tan, 2003).

Evidence from prior literature suggests that auditors generally adopt a reciprocity approach when dealing with clients. Accordingly, studies examining auditor–client negotiations find that auditors take a contending approach with difficult clients by making higher initial offers and making fewer concessions during the negotiation. A contentious approach adopted by a client during past negotiations results in auditors forming a negative perception about the client (Rennie, Kopp, & Lemon, 2010), thereby resulting in auditors taking a tougher stand (Pruitt & Carnevale, 1993). Further, a tougher stand during negotiation legitimises the professional role of auditors in the eyes of the client (Bame-Aldred & Kida, 2007; Beattie, Fearnley, & Brandt, 2004; Beattie, Fearnley, & Hines, 2015), sets the tone for future negotiation (Emby & Favere-Marchesi, 2010; Gibbins et al., 2001; Kleinman & Palmon, 2000) and prevents future exploitation by the counterparty (Roering, Slusher, & Schooler, 1975). However, when auditors perceive their bargaining power to be low due to a weak audit committee, ambiguous accounting standards or estimation uncertainties (Cannon & Bedard, 2017; DeZoort et al., 2003), auditors are found to contend less by adopting a more accommodating approach when negotiating with contending clients (Brown-Liburd & Wright, 2011; Ng & Tan, 2003).

Following on from the prior research, we expect that a client's expectation of a no-surprise audit adds additional pressure on auditors, which may lower their perceived bargaining power. Consequently, to protect their own self-identity that they have professional expertise and did not err in the past by not resolving all potential issues during the early-stage conversation, it is likely that auditors' acquiescence to the client's interests will be more pronounced when the past auditor–client relationship is contentious. In contrast, if the past experiences with a client suggest that the client is collaborative and receptive, it is likely that the client will continue to work with the auditors openly on contentious issues, irrespective of the early-stage conversation. Due to the amenable nature of the client, we expect an early-stage conversation to have a smaller impact on perceived client flexibility towards audit adjustments and, consequently, on auditors' accommodation of the client's interest.

Consequently, we propose that

4.1 Sample and administration

Participants were drawn from eight accounting firms, including both Big 4 and non-Big 4 firms located in Australia.⁵ An invitation letter with a link to the experiment was sent to audit partners for distribution. The invitation letter explicitly requested participation only from auditors who had experience in evaluating the work of experts (preferably manager level and above). Participation was voluntary, and participants were offered $50 upon completion of the experiment.⁶ Qualtrics, the online data collection software, was used to collect data. Qualtrics allocated a participant randomly into one of the four treatments.

Data were collected over a period of 7 months from September 2016 to March 2017. Out of the 95 responses recorded by Qualtrics, 26 were removed from the sample as they were either incomplete or did not demonstrate sufficient attention to detail. Accordingly, the results reported henceforth are based on the final 69 observations.

Demographic data are reported in Table 1. Among the 69 participants, 57.97% were from the Big 4 accounting firms, and 62.32% were male. Participants had an average of 9.55 years of experience in the field of audit and assurance (standard deviation = 5.57), and 72.46% of them were less than 35 years of age. The majority of participants were manager level or above and were regularly involved in evaluating the work of experts.⁷

4.2 Experimental task

The experimental instrument required participants to evaluate the work done by an ME engaged by an audit client and make decisions pertaining to the acceptability of the work of the ME as audit evidence. The instrument used in this study was developed in consultation with two audit directors from a Big 4 accounting firm, who identified ‘impairment’ as a key audit issue that is worth examining.⁸ The discussions with the audit directors also informed the setting that could provide an opportunity for auditors to acquiesce to a client's interests without compromising their own self-interest. Feedback on the draft instrument was sought from five renowned academics in the field, one audit practitioner and two technical advisers with a Big 4 accounting firm. Consequently, the terminology used in the experiment, the content and the wording of the manipulations, the placement of the information and the overall layout of the experiment were modified. The revised instrument was later pilot tested with academics within the discipline of accounting, and minor changes were made. The above process ensured that the task developed was clear, realistic and well designed.

The case began by providing participants with a background description of a hypothetical medium sized listed company in Australia. The case informed participants that the company had undertaken an impairment test with the help of a third-party expert. The audit evidence obtained by the audit team indicated that the third-party expert was a credible information source and all assumptions, except the discount rate used in the discounted cash flow model, were considered reasonable.

The third-party expert estimated the discount rate at between 10% and 12% and had used 10.8% in the computation of the recoverable amount of the client's operations. Other evidence such as the recommended range for discount rates by the audit firm's internal valuation division (11%–13%) and the average discount rate used by comparable companies (11.8%) implied a potential under-estimation of risk by the client's expert. Sensitivity analysis indicated that a discount rate of 11.2% would trigger an impairment charge, which if recognised would materially impact the client's financial statements. Overall, the audit evidence suggested that a small adjustment to the discount rate (i.e. increase of discount rate from 10.8% to either 11% or 11.1%), if insisted by the auditor would ensure that the client met with at least the lowest recommended discount rate without any materially adverse impact on their financial statements. However, any upward movement in the discount rate (i.e. from 11.2% onwards) would result in recognition of an impairment charge and therefore require a material adjustment to the financial statements. Importantly, the case provided an audit setting that provides some room for ambiguity of interpretation (i.e. an opportunity for auditors to accept the client's position without compromising their self-interests). Consistent to the motivated reasoning theory, by proposing a discount rate of 11% or 11.1%, auditors will not only be able to arrive at a desired conclusion (i.e. safeguard their client's interests by not requiring an impairment which will adversely affect the client's financial interests) but will also be able to reasonably justify their decision (i.e. the proposed adjustment ensures that the discount rate is within the range recommended by the audit firm's internal valuation division). Other risk factors were controlled, and the narrative deliberately kept generic to prevent potential noise that additional factors could bring to the setting.

After reading the above information, participants (1) assessed client flexibility to a proposed adjustment to the work of the ME (i.e. perceived client pressure) and (2) made decisions regarding the acceptability of the ME's work (i.e. acquiescence to the client's interests). Participants then went through manipulation checks, provided demographic information and lastly reflected upon their own personal experience with MEs at the end of the instrument.

4.3 Dependent variables

4.4 Independent variables

Two independent variables (early-stage conversation and auditor–client relationship) are manipulated between participants resulting in a 2 × 2 analysis of variance (ANOVA) design. The manipulation of an early-stage conversation (ES) with an ME refers to the presence or absence of the early-stage conversation and is consistent with its stated purpose by the auditing standards. According to GS 005, communication with an ME as early as practicable during the engagement would allow an opportunity to the auditor to confirm whether the ME's intended approach and methodology is appropriate for financial reporting purposes (GS 005, para 22). The manipulation of auditor–client relationship (RELATION) differentiates a collaborative and cordial client from a difficult and contentious one.⁹ Consistent with the elements used by Hatfield et al. (2008) and Brown-Liburd and Wright (2011) to describe auditor–client relationships, this study manipulates the relationship by focusing on two characteristics: (1) auditor–client interactions in the past and (2) flexibility of the client towards proposed audit adjustments in the past. See Figure 3 for a description of the wording of the manipulated variables.

5.1 Manipulation and other checks

Two manipulation check questions were employed.¹⁰ For the auditor–client relationship condition, participants were asked, ‘What do you think best describes the past relationship your audit firm has had with the CFO when dealing with proposed audit adjustments?’ Responses were recorded on an 11-point Likert scale (0 = extremely difficult; 10 = extremely cooperative). The average rating of participants in the contentious-client condition is significantly lower than those in the collaborative-client condition (means = 1.90 and 8.34 respectively, p < 0.001), suggesting this manipulation was perceived as desired.¹¹

To check if participants recognised the presence or the absence of an early-stage conversation, participants were asked to indicate if the audit manager within the case material had an early-stage conversation with the client's expert. Three options were provided: Yes, No and Unsure. Five participants answered incorrectly, and three were unsure of the condition. Possibly, the participants considered the information obtained through the early-stage conversation as immaterial, hence, the lack of attention to detail.¹²

Participants were also asked to indicate the extent to which they agreed that the case material realistically represented a situation they might come across during the course of an audit. The participants answered on an 11-point Likert scale (0 = completely disagree, 5 = neutral and 10 = completely agree). The mean response (standard deviation) was 8.04 (1.499). This indicates that overall, participants considered the case to be realistic.

5.2 Test of hypotheses

5.2.1 Test of H1 and H3 on perceived client pressure (inferred from perceived client flexibility)

A two-way ANOVA is performed to test if early-stage conversation influences auditors' perception of client flexibility to an adjustment proposed after the completion of the ME's work.¹³ As reported in panel A of Table 2, perceived client flexibility (FLEXIB) is lower under ES conditions (mean = 3.457) as compared with non-ES conditions (mean = 5.011). This result supports H1 that having an early-stage conversation with an ME reduces perceived client flexibility towards audit adjustments proposed in later stages. Our results also indicate that collaborative clients are perceived to be relatively more flexible (mean = 5.121) towards proposed adjustments than contentious clients (mean = 3.346). A two-way ANOVA indicates significant main effects for both early-stage conversation (F = 8.842, p = 0.002, one-tailed) and auditor–client relationship (F = 12.188, p = 0.001, one-tailed) confirming that perceived client flexibility is significantly lower when clients are considered contentious and when auditors have an early-stage conversation with MEs (refer panel B of Table 2).

TABLE 2. Auditors' perceived client flexibility towards proposed adjustment to the work of the ME

Panel A: Means (standard error) for perceived client flexibility towards a proposed adjustment to the work of the ME by treatment conditions
	ES	Non-ES	Overall
Collaborative	4.743 (0.458) N = 21 Cell 1	5.499 (0.509) N = 17 Cell 3	5.121 (0.341) N = 38
Contentious	2.170 (0.542) N = 15 Cell 2	4.522 (0.524) N = 16 Cell 4	3.346 (0.376) N = 31
Overall	3.457 (0.355) N = 36	5.011 (0.366) N = 33	4.234 (0.254) N = 69

Panel B: Test of hypothesis by treatment conditions
Source	Type III sum of squares	df	Mean square	F	p-value (one-tailed)
Corrected model	94.541	3	31.514	7.252	0.000
Intercept	1216.885	1	1216.885	280.046	0.000
ES	38.42	1	38.42	8.842	0.002
RELATION	52.963	1	52.963	12.188	0.001
ES*RELATION	10.748	1	10.748	2.473	0.061
Error	282.445	65	4.345
Corrected total	376.986	68
R2 = 0.251 (adjusted R2 = 0.216)

Panel C: Ordinal interaction contrast coding test
Planned contrast	t-statistic	p-value (one-tailed)
(+3, −1, +2, −4)	1.833	0.036

• Note: Dependent variable, FLEXIB, is perceived client resistance towards an adjustment proposed post early-stage conversation. Participants were asked, ‘If you were to propose an adjustment to the discount rate, how likely is it that the client would be willing to accept your recommendation and recognise an impairment of its assets?’ Responses were measured on an 11-point Likert scale ranging from 0 = being extremely unlikely, 5 = being neutral and 10 = being extremely likely. Independent variables = ES and RELATION. Early-stage conversation (ES) is manipulated at two levels: presence or absence. Past auditor–client relationship (RELATION) is manipulated at two levels: collaborative and contentious.
• Abbreviation: MEs, management's experts.

To test if the adverse impact of early-stage conversation on perceived client flexibility was more pronounced under contentious auditor–client relationship (H3), we conduct a planned contrast test. Buckless and Ravenscroft (1990) suggested that contrast coding provides greater statistical power than the conventional ANOVA when examining ordinal interactions. We test if the mean difference in perceived client flexibility towards an audit adjustment between the ES-Contentious condition (cell 2) and non-ES-Contentious condition (cell 4) is greater than the mean difference between ES-Collaborative condition (cell 1) and non-ES-Collaborative condition (cell 3). As a result, the following contrast weights are used to examine the ordinal interaction: ES-Collaborative condition (cell 1 = +3), ES-Contentious condition (cell 2 = −1), non-ES-Collaborative condition (cell 3 = +2) and non-ES-Contentious condition (cell 4 = −4).¹⁴ The results from the planned contrast test (refer to panel C of Table 2) and the downward slopes shown in Figure 4 support our predictions in H3, that is, an early-stage conversation lowers perceived client flexibility and that the size of the effect is larger for contentious clients. See Figure 4 for a graphical presentation of the impact of the independent variables on perceived client flexibility.¹⁵

5.2.2 Test of H2 and H4 on acquiescence to client's interests

Accept the work of an ME (ACCEPT)

Auditors' acquiescence to client's interests is primarily measured by the likelihood of auditors' accepting the work of the ME without any changes (ACCEPT). We conduct a two-way analysis of covariance (ANCOVA) to examine the effects of an ES and RELATION on ACCEPT. Participants' total audit experience (EXP) is included as a covariate.¹⁶ Descriptive statistics across treatment conditions, and the results for the ANCOVA are reported in panels A and B of Table 3, respectively.

TABLE 3. Auditors' acceptance of the work of the ME

Panel A: Means (standard error) for the likelihood of auditors accepting the work of the ME by treatment conditions
	ES	Non-ES	Overall
Collaborative	3.530 (0.396) N = 21 Cell 1	3.003 (0.441) N = 17 Cell 3	3.267 (0.296) N = 38
Contentious	2.935 (0.469) N = 15 Cell 2	2.486 (0.454) N = 16 Cell 4	2.711 (0.326) N = 31
Overall	3.233 (0.469) N = 36	2.745 (0.317) N = 33	2.989 (0.220) N = 69

Panel B: Test of hypothesis by treatment conditions (two-way ANCOVA)
Source	Type III sum of squares	df	Mean square	F	p-value (one-tailed)
Corrected model	21.901	4	5.475	1.668	0.084
Intercept	226.309	1	226.309	68.957	0.000
ES	3.963	1	3.963	1.208	0.138
RELATION	5.242	1	5.242	1.597	0.106
ES*RELATION	0.026	1	0.026	0.008	0.465
EXP	9.981	1	9.981	3.041	0.043
Error	210.041	64	3.282
Corrected total	231.942	68
R2 = 0.094 (adjusted R2 = 0.038)

Panel C: Ordinal interaction contrast coding test
Planned contrast	t-statistic	p-value (one-tailed)
(+3, −1, +2, −4)	1.589	0.059

• Note: Dependent variable, ACCEPT, is the likelihood of auditors accepting the work of the client's expert with no changes. Participant responses were measured on an 11-point Likert scale where 0 = extremely unlikely, 5 = neutral and 10 = extremely likely. A lower assessment suggests that auditors are less likely to accept the work of an ME in the presence of inconsistent audit evidence. That is, auditors are more likely to propose an adjustment. Independent variables = ES and RELATION. Early-stage conversation (ES) was manipulated at two levels: Presence or Absence. Past auditor–client relationship (RELATION) was manipulated at two levels: Collaborative and Contentious. Covariate (EXP) is participants' experience in the field of audit and assurance. Experience was measured by the number of years.
• Abbreviations: ANCOVA, analysis of covariance; MEs, management's experts.

As shown in panel A of Table 3, the mean acceptance of the client's position is less than ‘5’ across all of the four treatments. This indicates that overall, auditors are unlikely to accept the work of the ME without an adjustment. Consistent with our expectations, the pattern of cell means reveals that the presence of an early-stage conversation has a positive impact on auditors' acceptance decision (mean for ES condition = 3.233 and for non-ES condition = 2.745) under both collaborative and contentious auditor–client conditions. However, the insignificant main effect of early-stage conversation on auditors' acceptance decisions indicates that the impact of early-stage conversation is not statistically significant (refer to panel B: F = 1.208, p = 0.138, one-tailed). Similarly, the relative likelihood of auditors accepting the work of an ME appears to be higher when auditors have a collaborative relationship with their clients (mean acceptance likelihood for collaborative clients = 3.267 and for contentious clients = 2.711). However, the insignificant main effect of RELATION on auditors' acceptance decisions also indicates that the impact of the auditor–client relationship is not statistically significant (refer to panel B: F = 1.597, p = 0.106, one-tailed). These findings imply that in the presence of inconsistent audit evidence, auditors are unlikely to accept the work of an ME without any changes.¹⁷ The presence of an early-stage conversation will not influence their decision to propose an adjustment. The findings do not support H2.¹⁸

We conduct a planned contrast to test the impact of an early-stage conversation on the auditors' acceptance decisions for both collaborative and contentious clients (H4). As shown in panel C of Table 3, the presence of an early-stage conversation does not have a significantly larger effect on auditors' decisions to accept the work of the ME when the auditor–client relationship is contentious (t = 1.589, p = 0.059). The interaction plot shown in Figure 5 confirms that the size of impact of early-stage conversation on auditors' acceptance of the work of MEs does not differ due to the auditor–client relationship, suggesting that H4 is not supported.

Insist on an impairment (IMPAIR)

The second measure of auditors' acquiescence to clients' interests is auditors' willingness to insist on an impairment—IMPAIR. Table 4 reports the frequency statistics for the variable MINIMUM. As shown in panels A and B of Table 4, the raw scores for MINIMUM (i.e. the minimum acceptable discount rate indicated by the participants) across the four treatments are not significantly different from each other (F = 0.886, p = 0.453). However, a frequency table based on auditors' indications of the minimum acceptable discount rate for an unqualified audit opinion (refer to panel C of Table 4) shows that only 40.6% of the participants actually insisted on an impairment. Participants' willingness to accommodate the client's financial interests by not insisting on impairment is suggestive of the auditors' acquiescence to clients' interests.

TABLE 4. Auditors' indication of the minimum acceptable discount rate

Panel A: Means (standard error) for minimum discount rates required by auditors by auditors by treatment conditions
	ES	Non-ES	Overall
Collaborative	11.148 (0.090) N = 21 Cell 1	11.141 (0.100) N = 17 Cell 3	11.144 (0.067) N = 38
Contentious	11.34 (0.106) N = 15 Cell 2	11.256 (0.103) N = 16 Cell 4	11.298 (0.074) N = 31
Overall	11.244 (0.070) N = 36	11.199 (0.072) N = 33	11.213 (0.049) N = 69

Panel B: Test of differences between cell means (one-way ANOVA)
	Sum of squares	df	Mean square	F	p-value
Between groups	0.449	3	0.15	0.886	0.453
Within groups	10.989	65	0.169
Total	11.438	68

Panel C: Distribution of participants by minimum discount rate
Discount rate (%)	Frequency	Per cent	Cumulative Per cent
10.00	1	1.40	1.40
10.80	5	7.20	8.70
11.00	35	50.70	59.40
11.20	4	5.80	65.20
11.30	2	2.90	68.10
11.40	1	1.40	69.60
11.50	11	15.90	85.50
11.80	8	11.60	97.10
12.00	1	1.40	98.60
13.00	1	1.40	100.00
Total	69	100.00

• Note: Panel A reports the cell means of the minimum discount rates indicated by the participants across the four experimental conditions. Panel B reports the results of one-way ANOVA with minimum discount rate as the dependent variable and the four treatment cells as the independent variables. Panel C reports the frequency table for participants' selection of the minimum discount rate.
• Abbreviation: ANOVA, analysis of variance.

Based on the raw scores of MINIMUM, we create a binary variable IMPAIR to measure whether the auditor is willing to insist on an impairment. By design, if a participant indicates the minimum acceptable discount rate to be equal to or greater than 11.2%, then the dependent variable, IMPAIR, takes the value of ‘1’, otherwise ‘0’. Table 5 provides the results of the binary logistic model.¹⁹

TABLE 5. Likelihood of auditors to insist on impairment

Panel A: Proportion (standard error) of auditors likely to insist on an impairment by treatment conditions
	ES	Non-ES	Overall
Collaborative	33.3% (0.108) N = 21 Cell 1	29.4% (0.120) N = 17 Cell 3	31.6% (0.079) N = 38
Contentious	46.7% (0.128) N = 15 Cell 2	56.3% (0.124) N = 16 Cell 4	51.6% (0.088) N = 31
Overall	38.9% (0.083) N = 36	42.4% (0.087) N = 33	40.6% (0.060) N = 69

Panel B: Test of hypothesis by treatment conditions
Generalised linear model (function-logit, distribution-binomial)
Source	Wald χ2	df	p-value (one-tailed)
Intercept	2.065	1	0.076
ES	0.040	1	0.421
RELATION	2.790	1	0.048
ES*RELATION	0.316	1	0.287

Panel C: Ordinal interaction contrast coding test
Planned contrast	t-statistic	p-value (one-tailed)
(−3, +1, −2, +4)	1.736	0.044

• Note: Dependent variable, IMPAIR, is a binary variable that takes the value of 1 if a participant has indicated the minimum acceptable discount rate to be 11.2% or greater, 0 otherwise. A value of 1 indicates that the participant will require the client to recognise an impairment in order to receive an unqualified audit opinion. Independent variables = ES and RELATION. Early-stage conversation (ES) is manipulated at two levels: Presence or Absence. Past auditor–client relationship (RELATION) is manipulated at two levels: Collaborative and Contentious.

Panel A of Table 5 reports the cell means of the proportion of participants that are likely to insist on an impairment across the four treatments. The proportion is higher in the contentious client treatment (51.6%) as compared with the collaborative client treatment (31.6%). The significant main effect of RELATION (χ² = 2.790, p = 0.048), as reported in panel B of Table 5, suggests that past auditor–client relationships influence the auditors' motivated reasoning. That is, auditors appear to adopt a more compromising approach with collaborative clients by setting a higher tolerable limit, thereby being less likely to insist on an impairment. On the other hand, auditors adopt a more contending approach with difficult clients by setting a lower tolerable limit and are thereby more likely to insist on an impairment.

The effect of an early-stage conversation on auditors' decisions is also as expected. The proportion of auditors that will insist on an impairment are relatively lower in the ES condition (38.9%) compared with the non-ES condition (42.4%). However, the difference between the means of the two conditions is not statistically significant (χ² = 0.040, p = 0.421). This suggests that having an early-stage conversation with an ME does not significantly influence the auditors' decision to insist on an impairment. Overall, the findings do not support H2.

To examine the interaction effect as predicted in H4, we conduct a planned contrast. As predicated in H4, the mean difference between auditors' decisions under the ES-Contentious condition and the non-ES-Contentious condition is found to be greater than the mean difference between the ES-Collaborative condition and non-ES-Collaborative condition. We employ the same contrast coding as used in Table 3 but reverse the signs as rejection of the ME's work is examined instead of acceptance. The results shown in panel C of Table 5 highlight a significant ordinal interactive effect between ES and RELATION on the auditors' decisions to insist on an impairment (t = 1.736, p = 0.044). We then plot the interaction between ES and RELATION on auditors' decisions to insist on an impairment in Figure 6. As shown in Figure 6, the downward slope of the line for contentious clients indicates that having an early-stage conversation weakens the contending approach auditors adopt with difficult clients. Therefore, our prediction in H4 is supported, suggesting that the presence of an early-stage conversation has a larger negative impact on auditors' decisions to require an impairment when the auditor–client relationship is contentious.

Overall, our results indicate that having an early-stage conversation with an ME adversely affects perceived client flexibility towards audit adjustments, thereby indicative of pressure on the auditor to yield a no-surprise audit. This perceived pressure is significantly greater when the client is considered contentious. However, in spite of the perceived pressure, auditors are unwilling to accept the work of the ME without any changes in the presence of inconsistent audit evidence. Neither having an early-stage conversation nor the past auditor–client relationship influences auditors' decisions. Our findings do provide some support for motivated reasoning phenomena with auditors more likely to propose an adjustment that meets a client's interests (i.e. achieves the desired conclusion) without compromising the auditor's position (i.e. within reasonable constraints).²⁰ Although, having an early-stage conversation does not accentuate auditors' motivated reasoning behaviour, the unintended negative impact of early-stage conversation appears to be more pronounced under contentious auditor–client relationship conditions.

5.3 Additional tests

We conduct three additional tests to better understand our main findings. First, to further analyse auditors' perceptions of the impact of an early-stage conversation, participants answered three reflective questions at the end of the experiment. These questions serve to understand auditors' perceptions about the outcome of an early-stage conversation. Based on their personal experience, participants indicated the extent to which they agreed to the following questions: (1) Are MEs more approachable after an early-stage conversation? (2) Do clients expect auditors to resolve all potential issues during the early-stage conversation with MEs? and (3) Does an early-stage conversation with an ME ensure a no-surprise audit? Participants' responses were measured on an 11-point Likert scale where 0 = completely disagree, 5 = neutral and 10 = completely agree.

As reported in Table 6, an early-stage conversation opens doors for further communication with the ME. A majority of the participants (63.8%) agreed that MEs are more approachable after an early-stage conversation (mean = 6.48, median = 7). Approximately half of the participants (52.2%) further agreed that their clients expect them to resolve potential issues during an early-stage conversation (mean = 5.93, median = 6). Lastly, more than half of the participants (56.5%) agreed that an early-stage conversation, in fact, ensures a no-surprise audit (mean = 5.77, median = 6). The above outcomes observed in actual practice further confirm that an early-stage conversation can increase client expectations of a no-surprise audit and, thereby, potentially have an adverse impact on client flexibility towards audit adjustments.

Secondly, prior research indicates that perceived source credibility can influence auditors' decisions (Bernardi, 1994; Caster & Pincus, 1996; Danos & Imhoff, 1982; Goodwin, 1999; Griffith, 2015; Margheim, 1986; Rebele, Heintz, & Briden, 1988). To examine the possible influence of the perceived ME's credibility on auditors' decisions to use the ME's work, the participants were asked to indicate their confidence in the capability of the client's expert to undertake the task on an 11-point Likert scale (0 = not at all confident and 10 = very confident). Other than the early-stage conversation, the case material controlled for other credibility indicators such as the ME's experience, reputation, qualifications and objectivity. The mean assessment of the ME's capabilities is not significantly different across the treatments. We conduct ANCOVAs by incorporating the perceived ME's credibility as a control variable, and our main results still hold. The perceived ME's credibility is not significantly associated with either of the dependent variables, suggesting that it is not a concern in our experimental design.

Lastly, Griffin (2014) found that auditors view supplemental disclosures by clients regarding imprecision in fair value estimation as a substitute for adjustment and hence show a lower propensity to propose an audit adjustment. Therefore, as an additional dependent variable (ACCEPT-DISC), participants were asked to indicate the extent to which more disclosure of sensitivity analysis by the client would affect the likelihood of them accepting the work of the ME without any changes. Participants' responses were recorded on an 11-point Likert scale where 0 = significantly decrease, 5 = no change and 10 = significantly increase. The results from the main tests still hold. We conduct a two-way ANCOVA by using ACCEPT-DISC as a dependent variable, ES and RELATION as independent variables and participants total audit experience (EXP) as a covariate. We find that although additional disclosures increase the likelihood of auditors' accepting the work of MEs across all treatments, the likelihood of acceptance is significantly higher under collaborative client conditions. The findings once again indicate a reciprocity approach where auditors appear to be more considerate with collaborative clients and take a more rigid stand with contentious clients.