Reliance on the internal auditors’ work: experiences of Swedish external auditors

All the EAs mentioned ISA when discussing the evaluation of the IAF and its work. EA4 and EA5, working in two different audit firms, mentioned that their firm has interpreted the standard and drafted its own checklist based on it. This checklist has to be followed in the evaluation before deciding whether to use the IA’s work. However, even if there is a standard concerning the cooperation between EAs and IAs, it is the respondents’ opinion that the professional judgment of the EAs is of great importance when evaluating IAs’ work. This finding supports the argument by Knechel et al. (2013), Francis (2011) and Grey (1998) that professionals possess specific knowledge that should be used to perform their work. EA7 argued that if using the work of the IAF:

One should not downplay the fact that it makes tremendous demands. You must see that they [the IAs] are objective, that they are competent. If one is to rely on an IA’s work, then you need to do all that and ensure that they have done the task right.

This indicates that there are high requirements on the EA when doing the evaluation (Chen et al. 2017; Desai et al. 2017; Pizzini et al. 2015). However, one of the respondent’s statements about the time required to evaluate the work of the IAF is quite interesting:

I just have to take one hour to say I trust the internal audit, its work, and have assessed its competence and quality. I can get an idea of what they write in their reports. And then I just sign off the internal audit’s reports, in the section that concerns my work, and say that the internal audit considers the internal control as satisfactory and I consider it to be also.

This quote shows that the work style can also influence how EAs decide whether to rely on and use the work of IAs (Brody 2012). The respondents claimed that the evaluation of the IA’s independence was based on where the IAF was located in the organization, whom the IA reported to, and whether there were any incentive systems, corroborating the arguments of Chen et al. (2017). EA4 admitted that it was difficult to evaluate the IA’s objectivity and said that in the evaluation of the IAF he controlled how the work was performed and what conclusion the IA had drawn, and he checked whether he would have come to the same conclusion. As for the evaluation of IAs’ competence (i.e. professionalism), the EAs stated that this evaluation was based on the IA’s education and experience. EA1 added that the competence of the IA would be apparent once the EA had worked with him/her. Thus that EA seemed to follow his/her feelings, which is in line with Jonassen (2012), who states that emotions and experiences could have an impact on the decision making.

When asked if the evaluation resulted in more or less work for the EA in the audit process, five respondents clearly stated that the total time would decrease if the IA’s work was assessed to be useful. Such positive evaluation leads to efficiency improvements for the EAs. However, the situation differs for the EAs working with financial institutions. EA3 explained that when working with financial institutions there was no difference in time, because the evaluation of the IAF and its work is part of the engagement: ‘There must be a communication with financial companies, because it is part of my mission to check what the IAF does and how its function is designed’. Since this obligation exists, it might be tempting for the EA to make use of the IA’s work when the evaluation has to be performed anyway, which would suggest that the EA is a rational decision maker (see Edwards 1954; Jonassen 2012). Overall, EAs execute the evaluation in accordance with ISA 610, but this guidance can be interpreted differently and is largely a matter of the EAs’ professional judgment. It also seems that IAs’ independence is harder to evaluate than their professionalism (i.e. competence). Finally, respondents confirmed previous findings from the literature (see Chen et al. 2017; Glover et al. 2008) as outsourced IAFs were deemed more objective than in-house IAFs.

In general, the EAs had a positive attitude towards the establishment of IAFs within their clients. Some highlighted that IAFs are a good corporate governance and control mechanism, therefore agreeing with Mihret and Admassu (2011) and Soh and Martinov-Bennie (2011).

Other respondents emphasized that competent IAFs were good for the EA’s work. EA1 stated:

Many times they [IAs] are at the forefront of what is about to happen and what has happened and what is going on and they see risks. So, it is an incredible support to the general audit of the business to have a competent internal audit, absolutely.

The fact that the auditee has good internal control results in the EA saving time on the audit because it indicates orderliness in the auditee. Otherwise, ‘the controls on the company may become worse and then we would have to do more’ (EA7). This quote shows how EAs appreciate the efficiency gains stemming from good cooperation with IAs, something that is already recognized in the literature (Brody 2012; Pike et al. 2016).

However, the respondents also stated that the IA was not as controlled as the EA and did not have the same requirements, which means that the quality of the internal audit did not reach the same level as the external audit. This opinion is in line with previous findings by Roussy and Brivot (2016) and Allegrini et al. (2006). One reason for lower quality is related to the argument that the IA is not as independent as the EA (Arnold et al. 2013; Al-Twaijry et al. 2004; Glover et al. 2008; Stewart and Subramaniam 2010). Four respondents distinctly pointed out that the IAs were not independent. Specifically, EA5 argued that the IA was working for only one stakeholder, while EA1 pointed out that the IAF was an extension of the Board, and therefore IAs could not be considered independent. Still, the existence of an audit committee may, as highlighted by EA6 and EA7, spread the perception that IAs are more independent, confirming the arguments by Roussy and Brivot (2016).

According to some respondents, the independence of the IAs depended also on the size of the IAF. If a company does not have sufficient resources for its IAF, independence and competence of IAs can be problematic and, in turn, can affect the likelihood that the EA will decide to use their work, corroborating the findings by Al-Twaijry et al.’s (2004). Contradicting this, however, EA4 declared that the size of the IAF was not important, in the sense that the competence and work of the IAs is of greater significance.

Building on these findings, it appears that respondents showed an awareness of the constraints in using the IAF’s work. IAs are seen as not sufficiently independent and competent, and focused on different things than the EAs (Pike et al. 2016; Roussy and Brivot 2016). That is, IAs work for the Board and have other tasks that the EA may perceive as too narrow and specific (see Al-Twaijry et al. 2004; Arnold et al. 2013). EA6 and EA7 claimed that the IAs’ work was not reliable because they were either doing the wrong things or doing the right things wrong; therefore, one could not always use the IA’s work, for example, just to avoid duplication of work. One respondent stated that sometimes EAs had to check the work that had already been done by IAs. In such cases, the external audit process does not make efficiency gains but, instead, its overall effectiveness decreases (Brody 2012). In such circumstances, the EA had to ‘explain that if we are to rely on what you are doing, you must test according to our rules […] And it is not always the case […]. You are usually much freer being an IA’ (EA7).

Two respondents pointed at the advantages of outsourcing the IAFs (i.e. increased competence and legitimacy) that are also recognized in literature (see Glover et al. 2008; Roussy and Brivot 2016). Regarding IAs working in-house (i.e. within the auditee), EA1 said that ‘being in one place and sitting at the coffee table and talking, and being at the same staff parties with those you later audit and evaluate—I think that almost seems impossible’.

In summary, the existence of an IAF in an organization can be positive for both the auditee and the EA as it usually leads to fewer inaccuracies in the organization. At the same time, the IAs are not perceived as independent and professional by the EAs, meaning, as Al-Twaijry et al. (2004) and Morrill and Morrill (2003) argue, that the work of IAs cannot be used to an unlimited extent by EAs since this could have a negative impact on audit quality.

The theoretical framework suggests that the way cooperation evolves over time is linked to feelings of trust (Bachmann and Inkpen 2011; Tomkins 2001; Vangen and Huxham 2003). Half of the EAs thought that their communication with the IAs affected their trust in them. As an example, EA3 expressed it this way:

It depends of course on what the communication contains […] but you can put it like this, if in my external audit I intend to rely in any way on what the IAF does, it is more likely that I will trust someone that I have a higher degree of cooperation and exchange with than someone I meet more rarely or communicate with more rarely.

This quote corroborates the arguments by Brody (2012), who claims that good communication may engender greater trust. Half of the respondents stated that even the length and the type of communication influenced the EAs’ usage of the IAs’ work. The longer there had been cooperation and the EAs knew that the IAs were competent, the more they felt that they could rely on the IAs’ work. This finding supports Bachmann and Inkpen (2011) in that trust is based, among other things, on the counterpart’s competence. Other respondents were of the opposite opinion and claimed that the length and the type of communication do not affect the cooperation: ‘It is not as if you have been working together for three years, then suddenly you start to work more together’ (EA3).

In addition, emotions, experience and personal identity can be considered when making decisions (Jonassen 2012). Some EAs thought that personal chemistry was important in the cooperation with IAs while others did not. Most of them seemed to think that it made the cooperation work more smoothly but was not a crucial aspect in the decision to cooperate with IAs and to use their work.

In summary, the extent and type of communication seems to influence some of the EAs’ trust in the IAs. A tentative interpretation is that a long-lasting and well-functioning cooperation sometimes leads to greater trust, as Vangen and Huxham (2003) argue, or in other words, to a greater use of IAs’ work by EAs. At the same time, the EAs stated that they made the evaluation yearly, which would imply that the EAs tried to maintain their professional judgment so that the audit would not be negatively affected by a long cooperation. While the EAs did not seem to cooperate with IAs merely because of a favourable evaluation in terms of social matters, there was still some room for emotions and experience.

A number of our respondents agreed that the purpose of their cooperation with an IA was to gain different types of benefits, which is also in line with findings by Etzioni (2014), Delbufalo (2012) and Vangen and Huxham (2003). One of our respondents stated: ‘I prefer to take advantage of the IAs’ work. If they do good work, I can benefit from it in some way’ (EA4).

Our findings are in line with those of Brody (2012), Morrill and Morrill (2003) and Suwaidan and Qasim (2010) as the responding EAs indicated that cooperation with IAs leads to efficiency improvements. These gains, which appear to be interrelated, are manifested in reduction of work due to avoidance of duplication, time savings and/or cost savings.

Some respondents also claimed that the auditees were cost-conscious and wanted cooperation to occur in order to reduce the cost of the external audit (see Brody 2012). The cooperation could therefore result in a win–win situation in which the audit firm could gain competitive advantages by responding positively to the client’s expectations (Broberg et al. 2013).

Focusing on cost reduction and related efficiency gains might risk compromising the audit quality. Respondent EA4 mentioned that time saving often leads to more client orientation and a stronger focus on the client’s needs. This may indicate a shift of focus from audit quality to client satisfaction/orientation, as was also found by Duska (2005) and Broberg et al. (2013), who report that gaining competitive advantage by reducing audit fees, plus greater customer and market orientation, have become a common discourse in the audit industry.

Reducing the time spent on an audit could also result from pressure exerted by the audit firm (Broberg et al. 2013). Audit firms allocate a certain amount of time for audit engagements, which may result in pressure that influences how EAs perform their tasks and, in turn, could compromise audit quality (Broberg et al. 2017; Gundry and Liyanarachchi 2007; Svanberg and Öhman 2013). However, our respondents did not explicitly mention time pressure as a factor pushing them towards cooperation. According to EA4:

Assessment that I have too many clients or that I do not have enough time really has nothing to do with IAF, so to speak. But if I feel I do not have time and have too many engagements, then I cannot say, ‘Oh, what a relief that there is an IAF, then I do not need to work as much.’ [It does not work like that, and if it does then something is wrong.]’

Corroborating previous findings (Abbott et al. 2012; Pizzini et al. 2015; Lin et al. 2011), EA1 asserted that there might be a possible audit quality gain due to cooperation with the IA, because of the IA bringing ‘insider’ information to the table. Additionally, some respondents pointed out that EAs and IAs specializing in different areas can supplement each other in the audit work. One respondent went as far as to claim that he is more comfortable working with more experienced IAs, showing that experience matters in the cooperation (see Brody 2012).

When discussing the negative aspects of cooperation, our respondents evoked the issue of putting too much focus on the auditee and its needs at the expense of their independence, which is in line with Jeppesen (1998), who claimed that EAs’ independence could be compromised if they act in the interest of the auditees. One of our respondents (EA8) seemed to be aware of this risk, commenting that:

[The auditee has expectations] that one should use the internal audit, but when they are not that competent and you cannot really use it, it can become a bit tricky. […] It will not always be effective, but still one is going to use the work that fundamentally cannot be properly used.

This quote is in line with the findings by Desai et al. (2017) arguing that EAs may use the work of IAs even if doing so is unwarranted. Most EAs perceived their clients (auditees) with IAFs to have some expectations about cooperation. One respondent mentioned that focusing on the auditee view and overusing IAFs’ work could threaten EAs’ independence. The EA might disregard audit quality and prioritize client wishes/satisfaction. If an EA uses an IA’s work, even if doing so will have several negative effects, then audit quality can be compromised.

All these findings indicate that self-interest could be embedded in the decision to cooperate, which can be related to the agency problem. EAs may seek to maximize their own benefit and fail to consider the good of the wider public and society (see Jensen and Meckling 1976; Ross 1973), which in turn could adversely affect the quality of the external audit.

Our findings indicate that all EAs feel invested with great responsibility and that the maintenance of their legitimacy relies heavily on their ability to perform an independent audit, which is in line with the findings of previous studies (Carrington 2010; Duska 2005; Jeppesen 1998). In line with Baker and Hayes (1995), the responding EAs claimed that they have a responsibility not only towards the auditee but also towards the users and all stakeholders. In addition, our respondents agreed with Knechel et al. (2013), Francis (2011) and Grey (1998), claiming that professionals possess specific knowledge they require to perform their work.

Still, respondents EA4 and EA5 discussed the risk that the EA could come to a wrong conclusion and rely on the work of an IAF that is of poor quality. Further, EA8 mentioned that there is the risk that an EA could consciously choose to use work of poor quality. Thus, the quality of the external audit could be compromised. One EA argued that ISA 610 is clear, and as long as it is followed, the EA’s independence and professionalism would not be affected. However, we maintain that there might be a tension between EAs’ interest in cooperating and their responsibility towards the public. Taking into account the narratives of the respondents and the work tasks and the cooperation aspects they told us about, our own opinion is that a threat to their independence, while not acute, nevertheless exists.

In addition, one EA stated that the audit firm he was working in had opinions on the cooperation between IAs and EAs, indicating that such cooperation is perceived as sensitive and is being problematized at the firm level. At the same time, some EAs argued that the users of the external audit might not know that there could be cooperation between EAs and IAs because such information is not disclosed in any external reports. If the public is not aware of the cooperation, then it is not in a position to question it. An EA could therefore continue to act in self-interest even though the external audit could be perceived by users as having been prepared in an independent and professional way.

In sum, the findings of the qualitative part of this study resonate well with the elements of the theoretical framework presented in Fig. 1 and are also aligned with the results obtained in the quantitative part as explained in the following sub-sections.

Concerning Block 1, Table 2 shows that 24% of our respondents were women, the average professional tenure was approximately 23 years, while the average tenure as authorized auditors was approximately 17 years. Our average respondent was approximately 49 years old, 60% of the respondents were partners, and 50% of our sample consisted of auditors employed in Big 4 audit firms. Although our sample is small, the characteristics of our respondents do not significantly differ from the total population of Swedish auditors as reported in Broberg et al. (2018), apart from over-representation of the partners in our sample.

In relation to Block 2, one can observe that IA’s professionalism and IA’s competence have the highest mean values and are deemed by the responding EAs to be the most important aspects in the decision to use the work of the IAs. These findings resonate well with the findings of the qualitative part of the study, where respondents were in agreement that professionalism and competence are important starting points in the initiation of cooperation as well as its satisfactory continuation (cf. Brody 2012; Pike et al. 2016; Desai et al. 2017). IA’s staff size, tenure at the client firm, and work experience with IAs at another firm together with the inaccuracy reduction were deemed least important in the cooperation decision. The findings related to the IA and the firm are aligned with the findings of the qualitative study where however these findings were not as evident, which was also the case when it comes to the work experience with IAs. The latter could be explained by potential rarity of such a situation and thus EAs’ lack of exposure to it. The lower ranking of inaccuracy reduction could potentially be related to EAs’ confidence as well as overconfidence in performing audit assignments (cf. Hardies et al. 2011) as well as potential bias in answering this question due to the desire to maintain the status of the profession. The remaining aspects that had mean values between 3.19 and 3.63 were related to efficiency aspects (Time saving and Cost reduction), IA characteristics (IA attitude and IA availability), firm characteristics (Client firm’s size and Client firm industry) as well as regulatory (ISA regulations) and audit-quality-related aspects (Improvement in audit quality). These aspects were quite evident in the qualitative part of the study and resonate well with the findings in previous literature (cf. Suwaidan and Qasim 2010; Brody 2012; Al-Twaijry et al. 2004).

Concerning Block 3, EA’s efficiency gain stood out as the most important perceived outcome of the cooperation between EA and IA, which was also highlighted in the qualitative part of this study and has generated a lot of attention in past research (cf. Brody 2012). Relatively low means of the other three aspects were reflected in the interview-based findings, in that the EAs de-emphasized IAs’ cooperation influence on their independence, professional judgment and gatekeeping role. This could potentially represent both the assertion and gestalt of the professional discretion but also the EAs’ unwillingness to recognize potential risk associated with cooperating with IAs.

Given the limited number of observations and the number of variables that are not normally distributed, we have explored our data by means of Spearman Correlation (Table 3). From Table 3 we can observe that all the aspects in Block 2 are significantly positively correlated with EA’s efficiency. Inaccuracy reduction appears to highly positively correlate with Improvement in audit quality as well as Tenure at client’s firm, and has a weakly negative significant correlation with IA’s professionalism. EA’s professional judgment has a significant positive relationship with Tenure at client’s firm and has weakly significant negative correlation with Cost reduction, Client’s firm industry and Work experience with IAs at other firms. Finally, EA’s independence appears to have a weakly significant positive relation with Cost reduction, Tenure at client’s firm and ISA regulations, and weakly significant negative relation with IA’s competence and Work experience with IAs at other firms. The bivariate relations here provide just some indications of the relationships between the reason for working with IAs and EAs’ perceived outcomes of this cooperation. In order to explore what might be driving the EAs’ perceptions, we perform the multiple linear regression analysis. Yet, due to the limited sample, we explore reasons for working with the IAs by the means of a Principle Component Analysis (PCA).

We have performed the PCA on all the questions in Block 2. We checked the correlations between the variables (which only in 3 cases did not exceed .3), sampling adequacy (MSA, which was high) and Kaiser–Meyer–Olkin (KMO, above 0.7). The eigenvalues and screen plot indicated that three components should be used (all communality values were preferred 0.4). From Table 4: Varimax rotated component matrix we can observe that a number of factors have a crossloading, that is, Time saving, IA’s availability, Work experience with IAs at other firms, Improvement in audit quality and Cost reduction. Based on our interpretation of the components within each factor we have called Factor 1: IA’s Characteristics and efficiency; Factor 2: Firm characteristics and Audit Quality (AQ) and Factor 3: Efficiency and external audit quality (AQ). That is how the respondents in our sample appear to perceive and relate different reasons for working with IAs.

To explore our data further, we perform a multiple linear regression analysis. Given the limitation of our sample, we only include the three factor scores as well as two aspects from Block 1, namely the Big 4 and the Professional tenure of the responding auditor, given that both have been shown to relate to the choice of using IAs in previous research (e.g. Soh and Martinov-Bennie 2011; Mohamed et al. 2012). We perform four regression tests, one for each outcome variable in Block 3. Our testing indicated that only one out of four regression models was significant, namely, the model with EAs’ efficiency as a dependent variable. This regression is reported in Table 5, showing that it is IAs’ characteristics and efficiency that are primarily associated with it, as well as Efficiency and AQ considerations that are significantly correlated with EAs’ perceived efficiency associated with the use of IAs’ work.

The findings reported in this regression model resonate well with those of the qualitative part of this study showing that EA’s efficiency gain is an important outcome of cooperation with an IA. These findings also reflect EAs’ interview-deduced reflections as to what might be triggering this view. Similar to the qualitative part, the quantitative findings suggest that it is IAs’ characteristics, audit quality improvements and ideas related to efficiency that are important parts of the decision to use IAs’ work and are related to the perceived efficiency gains on the part of EA. Yet what the quantitative study shows is a more nuanced picture as to how the triggers of that perception are aligned with each other and their relation to the EAs’ efficiency perception. It appears that decisions related to efficiency are highly intertwined with decisions related to IA characteristics and improvements in audit quality. Thus it might indicate that when a decision to cooperate with IAs is made, the EAs consider the IA’s characteristics and efficiency gains simultaneously and consider them to be closely associated with each other, which in turn influences EAs’ perceived efficiency (cf. Endaya and Hanefah 2016). The same applies to the interrelation between efficiency and improved audit quality and their subsequent relation to EAs’ perceived efficiency, which has also been highlighted in the previous literature (Abbott et al. 2012).

Regression models with EA’s independence, EA’s professional judgment and EA’s gatekeeping role as dependent variables were found to be not significant, thus indicating that different EAs do not consider the IAs’ related factors to be associated with these specific outcomes. There could be several explanations to these non-findings. First, EAs perceive these three outcome aspects in a very homogeneous way, and the strength of embeddedness in the profession makes the respondents uniform in their answers, thus showing a low degree of variation (cf. Emby 1994; Broberg 2013). This explanation finds its support in other quantitative studies on auditors and difficulties with getting variability in answers given the homogeneity of professional identity and values in this profession (cf. Umans et al. 2016). Another possible explanation may lie in the single-item measures of the complex concepts that do not allow the respondents to assess the full meaning of the terms (cf. Fuchs and Diamantopoulos 2009). Here the qualitative part of the study provides a better-articulated insight into how EAs reason about their independence, professional judgment and gatekeeping role and how these relate to their decision to enter into cooperation with an IA. Finally, non-findings might signal EAs’ caution in answering questions related to them as individuals rather than as members of the profession, something that the qualitative part of the study has captured by allowing the respondents to make a choice of answering for themselves or for the role/profession.

Overall, the findings of the quantitative investigation highlight the complexity of the EAs’ decision-making concerning reliance on the IAs’ work, yet they provide a more nuanced understanding of the importance and interrelation of the triggers of these decisions.