Auditors on auditing – an empirical study
Peter Ohman*, Einar Hackner**, Anna-Maria Jansson*** & Finn Tschudi****
* Mid Sweden University Department of Social Sciences 851 70, Sundsvall, Sweden Email: [email protected] ** Lulea University of Technology Department of Business Administration and Social Sciences 971 87 Lulea, Sweden Email: [email protected] *** Mid Sweden University Department of Social Sciences 831 25, Ostersund, Sweden Email: [email protected] **** University of Oslo Department of Psychology 0317 Oslo, Norway Email: [email protected]
Auditors on auditing – an empirical study
Based on agency theory, an accountability model has been developed to describe and analyse the views of Swedish auditors regarding their reviewing of information provided by listed companies. 82 auditors were interviewed using the repertory grid technique and open interview questions. A sample of respondents was chosen to represent differences in experience, level of seniority, gender, firm and location. To check the stability in the thought patterns of the respondents, six retests were made and, to validate the results, two expert panels consisting of auditors and other representatives of the accounting and auditing profession were consulted. Principal component analysis of the mean grid for all the respondents showed a number of distinct patterns. The first component was related to judgement and the time-dimension and the second to auditing in practice. Auditors devote a relatively long time and considerable effort to issues that can be satisfactorily verified, but that are not of primary importance to the investors or to share prices. This auditing does not in any way violate the law or regulations, but the fact that they spend so much time reviewing auditing issues that they themselves do not consider to be of primary importance to stakeholders can be seen as a problem. Moreover, the auditors were very reluctant to make statements about any information except for that elicited according to current praxis. In addition to this traditional attitude, the auditors appear to be more concerned about their own situation than that of the accountees they are supposed to be protecting. There are no significant differences in the way in which the different groups of auditors perceive their reviewing on accounting information, with one exception. Auditors of listed companies perceive ???soft??? elements, such as CEO comments in the annual report and business strategies, as being of greater importance to investors and other interested parties than other auditors do.
Key words: Accounting, agency theory, auditing, repertory grid, stewardship, thought patterns.
Corporate failures have given rise to questions and doubts amongst the public about the quality of accounting information and financial audit failure (Power, 1997). The main focus has been on public limited liability companies and their reports and disclosures. Revelations of ???shady??? annual reports accompanied by vague audit reports have been in focus (Chandler & Edwards, 1996), together with severe losses for investors and other interested parties. Recent examples such as the Enron, WorldCom and Skandia accounting disasters have further highlighted the dilemma of information asymmetry. How and why was it possible for large listed companies to misinform the markets, investors and analysts so badly without alarm signals from the auditors
Generally, the provision of accounting information and auditing are regulated in law. The accounting and auditing professions interpret the laws and create ???rules??? and recommendations for accounting and auditing practice (Cf. Jonsson, 1988, 1994). Widespread lack of confidence and doubts about the quality of accounting and auditing among investors and other interested parties, as a result of the recent accounting disasters, have given rise to strong demands for changes in the rules of the game, i.e. the laws, regulations and recommendations. The conditions necessary to change and the effects of such changes are now being studied and evaluated in many countries.
Because of the great influence of the accounting and auditing profession in Sweden (Edenhammar & Hagg, 1997), it is essential to examine how professional auditors perceive their area of responsibility in auditing accounting information and any possible changes in
their duties. This paper presents the results of an empirical study of the thought patterns of Swedish auditors on these matters.
The study has four main purposes:
to develop an accountability model of the relationship between accountees, accountors and auditors, based on agency theory and the stewardship approach.
to describe and analyse the views of Swedish auditors on their reviewing of information provided by listed companies, according to Swedish legislation and praxis.
to examine the opinions of Swedish auditors about making statements about the information provided by listed companies, if auditors were permitted to exceed present Swedish legislation and praxis.
to compare and draw conclusions about the difference in thought patterns between different categories of auditors, with reference to their experience, seniority level, gender, firm and location.
An accountability model
The overall purpose of company accounting is to provide useful information. More specifically, the information should provide investors, creditors, employees, public authorities and other interested parties with a ???true and fair??? view of the firm and its operations. The 2
focus should be on the requirements of the users of the information for the purpose of decision-making (Hendriksen & van Breda, 1992).
Accounting involves a principal-agent relationship (Gjesdal, 1978; Mellemvik, Monsen & Olson, 1988). Agency theory analyses the relationship between two parties: investors and managers. The agent (the manager) undertakes to perform certain duties for the principal (the investor) and the principal undertakes to reward the agent (Jensen & Meckling, 1976). It is argued that in a modern corporation, in which share ownership is widely spread, managerial behaviour does not always maximise the returns of the shareholders (Pratt & Zeckhuser, 1985; Donaldson & Davis, 1991). The degree of uncertainty about whether the agent will pursue self-interests rather than comply with the requirements of the contract represents an agency risk for an investor (Cf. Fiet, 1995). Given that principals will always be interested in the outcomes generated by their agents, agency theory demonstrates that accounting has an important task in providing information after an event has taken place. This task is often associated with stewardship, in which an agent reports to the principal on the events of the past period (Ijiri, 1975).
In agency theory, incomplete or distorted information often constitutes a problem. Situations in which not all factors are known to both parties and certain consequences are not considered by both are known as information asymmetries (Hendriksen & van Breda, 1992). The agent may be able to take actions that are impossible to verify and therefore cannot be included in a contract. Such a state of affairs is referred to as moral hazard. Moral hazard usually implies an adverse selection, i.e. situations in which one party, typically the agent, has private information that is relevant for performing according to the contract (Besanko, Dranove & Shanley, 1996). Some information is then withheld and remains unknown to the principal. 3
One possible way in which to address the problem of adverse selection is to regulate the agents actions regarding the content and form of the accounting information to be reported. Auditors can be engaged by the principals to monitor the activities of the agent (Eisenhardt, 1989) and to ensure the quality of the information provided (Wolnizer, 1987). Without correct information, the principal will not be able to check up on the agent (Mellemvik et al, 1988).
Using agency theory and the stewardship approach, we can identify three significant parties to include in an accountability model. Based on the theories of Ijiri (1975) we have included the accountee, accountor and auditor in the model, and focus on their relationship concerning the fact that auditors are unable to review all the information provided by the accountor (Lee, 1993).
Figure 1 about here
The model can be linked to entity theory (Paton, 1962) as well as enterprise theory (Soujanen, 1954). According to entity theory, the operations of the company are seen as de-coupled from the owners of the company. At the same time the company has social responsibilities, as emphasised by enterprise theory. The information provided affects several interested parties, not only investors.
There are two types of information provided by the accountors. One type of information (a) goes directly to the accountees without being quality assured by the auditors before it is distributed. Accountors can choose what information to distribute and when this should take place. Such a situation may often involve moral hazard and obvious risks of adverse selection. 4
Another type of information is first presented to the auditors (b) for quality assurance before it is distributed to the accountees (c). Some of the information presented to the auditors (b) by the accountors is controlled by laws and regulations. It is left to the discretion of the accountors to decide about the content and form of the remainder of this information. In addition, the accountors must provide the auditors with further information on which they can make an assessment of the quality of the information presented to the accountees (c). Thus, not all the information delivered to the auditors is forwarded to the accountees.
Whether the information distributed to the accountees is controlled by laws and regulations or not, the information distributed is perceived as important (Burchell, Clubb, Hopwood, Hughes & Nahapiet, 1980). Accounting and information in general about a certain area of business directs the reader??™s attention to that area, sometimes thereby diverting attention away from other areas (Mellemvik et al, 1988). This has been referred to as ???the magic of accounting??? (Gambling, 1977; Grojer & Stark, 1978). This mechanism creates opportunities for accountors, primarily via the left arrow (a) in the model, to influence the overall picture presented to the accountees.
As can be seen in the model, the auditor is a link between the accountor and the accountee. The main duty of the auditors is to protect the principal and the public interest (Flint, 1988). However, auditors have been criticised repeatedly for their inability to protect investors and other interested parties (Lee, 1995). There is an obvious risk of the auditors neglecting their main duty and instead becoming the advocates of the agents (Bazerman, Morgan, & Loewenstein, 1997; Haynes, Jenkins & Nutt, 1998). Auditing firms are dependent on their clients to increase their share of the market (de Ruyter & Wetzels, 1999), and revenues from non-audit services to audit clients have become an increasingly important way for auditing 5
firms to improve their profitability (Arnold, Bernardi & Neidermeyer, 1999). The auditors may give more attention to arrow (b) than arrow (c) or disregard arrow (a) in the model. If that occurs, the independence of the auditors can be questioned.
The independence of auditors has been called into question and a growing interest has developed in examining what auditing is supposed to deliver compared to what it really ???delivers???. This ???expectation gap??? emerges when auditors and the public hold different beliefs about the duties of auditors and the responsibilities and the message conveyed by audit reports (Monroe & Woodliff, 1993; Koh & Woo, 1998). For example, according to Epstein & Geiger (1994) an expectation gap exists between auditors and investors regarding the level of assurance an audit provides. The role of auditors in fraud detection has been given as one particularly critical component of the expectation gap (Humphrey, Moizer & Turley, 1993). The public expects auditors to play a role in fraud detection, while auditors claim that their task is to deliver an opinion on the financial statements that can be described as ???true and fair??? (Power, 1997).
Similarities and differences between different categories of auditors
One of the core activities of auditing is to obtain and evaluate documentary and other types of evidence. Even so there are no definitive guidelines for this evaluation process. Auditing is a regulated function, but it is also based on the professional judgement of the auditors (Eklov, 2001). Differences in the application of auditing routines have been described in terms of ???structure vs. judgement??? (Dirsmith & McAllister, 1982) indicating the difference between a formal approach and allowing for individual judgement. Structure represents the 6
preformatted, systematic work that is dependent on instructions, manuals, computer support etc. In contrast, judgement emphasises the human capacity. Auditors are expected to rely on their own experience, to bring many aspects into consideration and to make assessments accordingly. There are differences in opinion regarding the merits of increased structure in auditing (Cf. Messier, 1995). Power (2003) claims that the trend towards greater structure is about legitimacy and control, which is not necessarily consistent with better or more efficient auditing.
Within strongly regulated social spheres, such as accounting and auditing, the actors tend to employ elaborated thought patterns and frames of reference. These thought patterns form the basis of selective action and behaviour (Patterson, 1996). The issues addressed in this study take place in the accounting and auditing context and deal with auditors??™ core activities and, to some extent, their professional identities, i.e. issues that can be assumed to be of great psychological significance to them. Well-developed thought patterns tend not only to be more complex and more balanced, but also more conservative and more difficult to change than other thought patterns (Hellgren & Lowstedt, 1997).
On the one hand it seems reasonable to focus on general thought patterns and to assume a high degree of homogeneity in the thought patterns of auditors, as professional auditors have a similar education and often face similar situations and tasks in their work. Auditors seek support from their colleagues and are involved in on-going interaction with other members of their profession (Pentland, 1993). Furthermore, Fischer (1996) claims that auditors have a tendency to rely on objectified knowledge, and that they appear to be highly constrained by the conventions of their profession regarding how to conduct an audit. Thus, auditors seem to show a high degree of socialisation. 7
On the other hand, auditors also interact with members of their firm and their audit team (Pentland, 1993; Grey, 1998). Auditors tend to ???identify??? themselves with their firm and those belonging to the same firm tend to adopt a similar approach to how to conduct an audit. Fischer (1996) reports that the conventions of how to conduct an audit appear to differ not only between firms, but also between audit teams within the same firm. Furthermore, auditors??™ knowledge and level of understanding develop gradually over time. New experiences are interpreted on the basis of prior knowledge (Cf. Berger & Luckmann, 1967; Patterson, 1996). It is therefore reasonable to assume that auditors interpret their experiences in the light of such prior knowledge, but also in the light of the common stock of knowledge within the auditing firm. Auditors mediate meaning through their behaviour, and they build their frames of reference through action, in interaction with other people in the same firm. Are there differences in the thought patterns of auditors with short experience and those with long experience, between auditors working in one auditor firm and those in other auditor firms etc Comparisons between different ???categories??? of auditors with respect to the corresponding ???external variables??? (experience, seniority level, gender, firm, location, etc.) are of interest. Do differences between different categories of auditors influence the general picture with respect to all auditors
Given that frames of reference are gradually and continuously developed (Cf. Bannister & Fransella, 1986) and that different audit tasks are normally performed at different stages of an auditor??™s career (Bonner & Pennington, 1991) it may be possible that auditors take a different view of the relative importance of different kinds of auditing issues depending on their level of experience. In this respect, it seems reasonable to make a distinction between ???hard??? and ???soft??? elements of auditing (Hackner, 1988). ???Hard elements??? refers to that which can be measured relatively easily, such as the purchase of shares, stock price drops, bad receivables 8
and capitalised organisation expenses. ???Soft elements??? refers to phenomena that are difficult to measure and often difficult to verify, such as comments from the CEO in the annual report and company strategies. Experienced auditors may devote more time to making judgements about whether the information provides a true and fair view than less experienced auditors do. These judgements demand a holistic view, which includes the soft elements. It is reasonable to assume that more experienced auditors may see such soft information as more relevant to investors??™ decision-making than less experienced auditors do. The relationship between accountees, accountors and auditors is particularly significant and problematic in the case of listed companies. Guided by information provided by the accountor, the accountee expects to be able to make well-founded decisions. The auditors of listed companies should be well aware of what information is relevant to investors for decision-making. The soft elements carry especial significance in these cases. Hence, it is important to investigate whether auditors with longer experience perceive soft elements as more important to investors than auditors with shorter experience do, and if the auditors of listed companies perceive soft elements as more important to investors than auditors of non-listed companies do.
Our approach is neither synonymous with objectivism nor with pure subjectivism. Even if we reject the view that it is possible to determine a given reality ???out there???, our scientific position can rather be described as a double perspective, in line with Campbell (1988) and Bryman (2001). Analytical and interpretative approaches are combined in the study. Data is 9
collected in a structured way, but the data is comprised of the thought patterns and interpretations of actors. We combine statistical analyses of data with interpretations of these analyses and actor responses to open questions in interviews.
We assume that thought patterns on the actor level can be represented and analysed using cognitive maps (Fransella, Bell & Bannister, 2003) and also that it is possible to analyse cognitive maps on a group level (Tschudi, 2000). A basic question is to what extent frames of reference are comparable on a group level as well as on the individual actor level. The auditor makes sense of his or her world, interprets the situation at hand and acts accordingly. The subjective meaning of the individual auditor can be inter-subjective among auditors as a group, and thus be perceived as objective by the actors (Cf. Berger & Luckmann, 1967).
The repertory grid
Data was collected through structured personal interviews with a sufficient number of Swedish auditors to enable a certain degree of generalisation to be drawn from the findings. To chart the thought patterns of the auditors to reviewing and making statements about information provided by listed companies, we used a special type of cognitive map: the repertory grid. The psychologist George Kelly (1955) introduced the repertory grid technique as a method for exploring personal construct systems (Stewart & Stewart, 1981). It is a set of related (interview) methods that elicit responses in a structured manner, but the content and specific structure of responses varies between the respondents. It is a well-tested method that provides a reliable representation of the thought patterns of the respondents (Reger, 1990; Fransella et al, 2003). 10
When collecting empirical data using the repertory grid technique there are three specific steps to be carried out: 1) selecting elements, 2) eliciting constructs and 3) eliciting evaluations of the elements in terms of the constructs.
1) The elements can be people or objects, situations, actions, etc. (Tschudi, 1998, 2000). In general, either the researcher or the respondent can select the elements. A third possibility is to combine the two procedures (Reger, 1990). In the latter case previous studies can serve as a basis for the choice of elements and the respondents can then also choose their own elements.
2) To make sense of the elements, constructs need to be elicited. The classical way to elicit constructs is to use Kellys triadic method (Kelly, 1955). He emphasises that constructs are bipolar, so the respondent is faced with three elements and asked to describe in what way two of them are similar and how the third one is different. The characterisation of the similar pair is called the likeness pole; the characterisation of the third one is called the contrast pole. The procedure continues until a sufficient number of constructs have been generated. A variety of other methods of elicitation have also been suggested (Stewart & Stewart, 1981, Reger, 1990? Fransella et al, 2003).
3) Having established a relevant set of constructs and elements the respondent is asked to rate each of the elements in terms of each of the constructs, i.e. on the dimension formed by the two poles of the construct. There is a simplified example of a grid for an auditor in figure 2. In the example the auditor finds bad receivables to be the easiest (1) and productivity the most difficult (7) element to review. The other elements are given scores in between these two.
Figure 2 about here
It was considered important that the auditors in the sample represented different categories fairly well with respect to experience, seniority level, gender, firm and location. Large as well as small auditing firms were included. The whole population of auditors in the greater area of a medium-sized Swedish city was selected to be studied. According to statistics from the Supervisory Board of Public Accountants there were a total of 43 active auditors in the area. Two of them were excluded as they had prior knowledge of the study. There were four nonresponses due to lack of time, leaving 37 to be interviewed. All 37 took part and produced complete data. To include auditors from more than one city, auditors from two larger cities and a small town were also included, making the total number of respondents 82. Access was the most decisive criterion when additional auditors were selected, but ensuring that different categories of auditors were represented was also considered important.
Pilot study and interview forms
A pilot study was made in order to test the techniques for data collection and to receive qualified feedback regarding our prior knowledge and assumptions. The pilot study took place over a period of two months. Two senior authorised auditors from two big auditing firms in different cities co-operated as ???co-researchers??? (Cf. Hackner, 2001). With their help, the grid, 12
the complementary interview questions and the manuals for the data collection process were continuously revised and improved.
A preliminary list of elements (auditing issues) was elicited based on previous research results, the accountability model, pilot studies using the repertory grid and suggestions from the two co-researchers. The auditing issues represented three primary types of audit: financial statement audits, operational audits and compliance audits (Arens & Loebbecke, 1997: 4-6) in addition to both hard and soft elements of auditing (Hackner, 1988). To elicit constructs we used Kellys triadic method, but supplemented this with a less constrained conversational approach and informal discussions (Cf. Stewart & Stewart, 1981). The constructs were treated as 7-point scales for the evaluation of the elements, representing a wide range of aspects relevant to the inherent problems in the accountability model. By repeated grid studies and discussions with the co-researchers, the final design of the repertory grid was determined. In the instructions for the interview sessions it was emphasised that the respondents should use their experience as auditors when answering the questions. In addition to the instructions for the grid interviews, a brief description of a fictive information technology company was provided as an illustration.
The grids generated in the pilot study were analysed statistically using Principal Component Analysis (PCA). An examination of the results and discussions with the co-researchers resulted in adjustments being made to the grid to be used in the main study. It was considered that an auditor was likely to only be prepared to devote limited time to such a non-profitgenerating activity, so it was decided that interview sessions should not exceed 90 minutes. Consequently the final grid was limited to 14 elements and 12 constructs, see appendix A.
In addition to the grid interviews, background questions and questions of an open character were included. The background questions were used to categorise the auditors. The open questions focused on whether the respondents considered it possible to review an issue less extensively without this having a negative effect on the information needed by investors and other interested parties, and if they considered it possible to review an issue more extensively in order to provide more significant information. The final open question addressed the auditors opinions about the possible effects and consequences of expanding the auditing area. They were asked to comment on their feelings about making statements regarding information not normally commented on in current praxis.
Data collection in the main study
The fixed elements and constructs elicited in the pilot study were used in the grid interviews in the main study. Each interview session took place at the respondent??™s office and was conducted by one of the participants in the research team. The researchers took care to avoid introducing into the interviews any preconceived notions based on the frames of reference of the research team (Cf. Hackner, 2001). All of the respondents produced their data without communicating with anyone, except for the interviewer. Most of the interviews were carried out with only one respondent, but in some cases two or more auditors from the same auditing firm were interviewed simultaneously. The interviews were held within a period of seven weeks from November 2002 to January 2003. The respondents were encouraged to rely on their intuitive feelings, rather than to try to ???analyse??? the scores of the elements on the construct scales. Most respondents took 60 – 90 minutes to read the instructions, fill in the
grid form and give responses to the other interview questions. Generally the respondents were motivated and had no problem filling in the grids.
As only Swedish-speaking auditors took part in the study all the documentation was written in Swedish. In the report, all the elements and construct have been translated into English. To make sure not to miss any of the conceptual content of the concepts a ???back translation??? procedure was carried out. The process did not reveal any translation problems of importance.
The repertory grid technique yields geometric, quantitative and qualitative representations of the respondents??™ construct systems. Different statistical methods can be used, depending on the purpose of the research (Stewart & Stewart, 1981, Reger, 1990). To analyse the individual grids we used the Flexigrid program (Tschudi, 1998) to perform the PCA. The PCA analyses the internal structure of a set of variables in order to identify the underlying components. This technique makes it possible to reveal basic dimensions in the thought patterns that are essential to the respondent, but which may be difficult to see or articulate during the interview. The program generates the components and they can be interpreted by examining the constructs with high factor loadings and the elements with high factor scores. When distinctive variables load high on a component or form tight clusters, a pattern may emerge (Walsh, 1990), mapping the thought patterns of the respondent with respect to area in focus.
One commonly used measure of the degree of complexity in a grid is the variance described by the first component. If a high percentage of the variance is described, the thought patterns 15
have a one-dimensional character, which indicates low complexity. An alternative way of determining the degree of complexity, is to find out how many of the components can be interpreted in a meaningful way when the data structure is compared with theoretical models, prior knowledge of the field studied and data collected using other methods (Tschudi, 1998).
For comparing individual grids and analysing homogeneity within as well as between different categories of auditors we used the Multigrid program (Tschudi, 2000). This program allows for a wide range of analyses, including computing the overall mean grid and computing mean grids for specified categories. Each individual grid can be treated as a string (vector) of 14 * 12 = 168 numbers (nr. of elements * nr. of constructs) and these vectors can then be correlated and the principal components describing the respondents can be computed. Furthermore, each person can be described by a vector, in which the values from an analysis of the grid of this person are either element factor scores for the first principal component or construct loadings for the first principal component. Regardless of which of these three options is used, high correlation between respondents or between mean values from categories of respondents represents homogeneity in the thought patterns of these groups.
To allocate the auditors to different, possibly homogenous, categories, 15 external variables (including one control variable) were used, see appendix B. Guided by the respondents??™ answers to the background and open questions, these external variables were chosen for the comparison of different categories of auditors.
The general view of the auditors
The general picture for the 82 auditors shows moderate homogeneity. As can be seen in table 1, the homogeneity in the full grids is relatively low (.360), according to the mean intercorrelation between the variables. The same conclusion applies when looking at the results of a PCA in which the auditors are treated as ???constructs??? and their vectors of factor scores regarding the first component as ???elements???. For the constructs the homogeneity is higher, but still rather moderate. One possible interpretation of the finding that the mean intercorrelation is higher for constructs than for elements would be that the auditors have a more common ???language??? related to their use of constructs than similarity when they use that language to characterise the elements.
Table 1 about here
The second column in the table clearly shows that when ???noise??? is filtered away by using the vectors of factor scores or factor loadings instead of the vectors of all the combinations of elements and constructs, the percentage of the variance that is explainable increases. This observation supports the expectation that the first component purifies inherent patterns in the data.
For each of the three approaches above we can think of the person as a ???construct???. A twodimensional PCA analysis of these ???constructs??? reveals possible clusters. Distinct clusters (of respondents) represent a specific point of view. Had there been many such clusters a common 17
frame of reference for all the auditors could not be posited. We did not, however, find any such clusters, but the majority of points clustered around the central axis, so the mean grid can be said to give a fair representation of the respondents??™ thought patterns.
One measure of complexity used was the percentage of the variance in the mean grid explained by the components. The first component explains 52 % and the two first components together explain 76 % of the variance. Together these findings indicate a moderate complexity according to the mean grid. As a relatively high explanation value was represented by two of the components, while the third component did not allow an easy interpretation, the mean grid was further analysed in two dimensions. Figure 3 is a plot presenting a varimax rotated mean grid for all 82 auditors.
Figure 3 about here
The PCA of this mean grid showed some distinct patterns. The relations between the constructs that form three clusters give rise to the following interpretations:
In the first cluster five constructs are included. Furthest to the right in the plot we find ???very prediction dependent??? (the contrast pole for construct number 3), ???statements significant for share price??? (8), ???statements very valuable for investors??? (2) and even ???industry considerations required??? (9) and ???internal control not important??? (4). These can be interpreted as judgements of information related to the future. The other extreme of the five constructs (furthest to the left in the plot) represents the past.
In the second cluster three constructs are represented. To the lower right in the plot we find ???very oral information dependent??? (12), ???statements significant for company image??? (11) and ???difficult to review??? (1). These have to do with judgements of soft information that is difficult to verify. The other extreme of the three constructs (to the upper left in the plot) focuses on hard information that is possible to verify.
In the third cluster four constructs are represented. At the bottom of the plot we find ???seldom reviewed??? (5), ???limited effort (time)??? (7), ???low review precision??? (10) and ???competence not sufficient??? (6). These can be interpreted as judgements showing a creative and comprehensive view. The other extreme of the four constructs (at the top of the plot) represents ???number crunching??? of selected parts.
The two axes in the mean grid represent two essential dimensions of the respondents thought patterns. The location of the elements in the plot supports the interpretations. The first dimension has to do with the time perspective. Auditing issues such as attest routines and CEO comments are two opposites on the time axis. Attest routines represent information from the past while CEO comments represent information about the future. The second dimension has to do with auditing practice. The auditors devote a relatively large amount of time to issues such as balance sheet items. They are well qualified to review these issues. On the other hand, issues such as productivity, strategies and environmental crime are seldom reviewed and the auditors are not qualified to review such matters with high precision.
The interpretations of the mean grid also indicate a gap between what auditors actually review and what is perceived as important for investors and other interested parties. The auditors
devote a relatively large amount of time and effort to issues that are not of major importance to shareholders, and little time to issues that exert a major influence on share prices.
Comparisons between categories of auditors
To what extent can the overall mean grid be said to characterise the different categories of auditors according to each external variable One approach to this question is to correlate each external variable with the mean grids for the categories of that variable. For variables with just two categories, for example gender, there will be just one such intercorrelation. If there are more categories, there will be an intercorrelation matrix for each of the categories of a variable. Computing a total mean intercorrelation for all the variables gives a mean correlation of .877.
Even though the mean grids for different categories of auditors are quite similar to each other, the configurations might still show some differences. We therefore made a separate analysis of each category, making a total of 45 analyses for the 15 external variables. For 24 configurations the clusters identified above ??“ both for constructs and elements ??“ could be recognised as described above. Looking closely at the clusters we see that construct 12 is quite close to the cluster (3, 8, 2, 4, 9), and it is thus not surprising that in 11 configurations construct 12 seemed to belong more clearly to this cluster than to form a cluster with (1, 11). In the remaining 10 configurations the overall pattern could also be recognised, although in 2 or 3 cases it was fairly weak.
To compare different categories of auditors according to their experience, seniority level, gender, firm and location we examined the content and homo-/heterogeneity of their thought patterns. If the correlation between members within a category is higher than the correlation between members of different categories, this indicates that the first category of auditors has developed a unique homogeneity in thought patterns not shared by those other categories. Generally, there were rather limited differences between the categories identified. No typical thought patterns could be found in any firm or city. Neither could differences between men or women, between authorised public accountants and approved accountants or between auditors with great or little experience be found. All correlation values (both inter- and intracorrelation) were between .30 and .45. The relatively low correlation values indicate a moderate homogeneity within as well as between categories of respondents.
It was not possible to see any clear differences between auditors with different experience and their view of the importance of soft elements (CEO comments and business strategies) to investors. However, one finding linked the degree of experience of auditors from listed companies with the soft elements of auditing issues. Auditors in listed companies perceive soft elements as more important to investors than auditors who do not work for listed companies.
To detect a conceivable interviewer bias we also examined possible differences between respondents interviewed by different interviewers. The results showed no sign of differences in the thought patterns of the categories of auditor checked.
Open interview questions
The answers of the respondents to the open interview questions revealed a traditional attitude. Most of the respondents did not want to see any or only small changes in the way in which auditing was carried out. 30 % of the 82 respondents were of the opinion that they could not carry out the reviewing in a less extensive way without this having a negative effect on the information needed by investors and other interested parties for decision-making. Almost 20 % of the respondents felt that more extensive reviewing would not result in more useful information.
Among the respondents that were not entirely negative to changes, there was considerable consensus concerning their willingness to reduce the emphasis on one specific type of auditing issue. They considered that too much of their work involved tax supervision. Of those respondents who were not negative to changes every second argued that supervising taxes and other charges, such as employers??™ contributions to social security, is the duty of the tax authorities. Some of the respondents also criticised the fact that they are obliged to report economic crimes. They commented on the fact that criminal law is not included in the basic auditor training.
In general, the respondents were very negative to the idea of providing statements about any information except for that elicited according to current praxis. Only 7 % of the respondents were in favour of making statements that go beyond the present legislation and praxis. Despite the general reluctance to provide additional statements, none of the respondents considered such statements insignificant to investors and other interested parties.
Although we failed to detect any clear differences between the grids for different categories of auditors, there could still have been individual differences between respondents. One way to test this, and to check the stability in the thought patterns of the respondents and the reliability of the data collection, was to make some retests. Therefore, six of the respondents were interviewed a second time, five months after the main interviews. They were presented with the same elements and constructs, but this time placed in a different order.
If no genuine individual differences existed, the similarity between two grids from the same person would be of the same magnitude as the correlation between any two grids from different respondents. For the six auditors that participated in the follow-up tests the mean intercorrelation between the first test and retest was .573. Although this is not an impressive degree of reliability it is clearly larger than the mean intercorrelation between different individuals, which was .360.
A related way of checking for consistent individual differences is to ask whether two grids from the same person have more in common than can be accounted for by the general factor (represented by the overall mean grid). When the general factor is partialled out will the correlation then drop to 0 All the partial correlations were positive, and the mean was .30, again testifying to the existence of ???real??? individual differences.
When asked about possible explanations to the differences between the first test and the retest, the six respondents argued in a rather similar way. Their frames of reference were partly 23
influenced by experiences in their daily work and general events relevant to their work or their business. Some of the respondents testified that specific events reported in the press had influenced his or her opinions and, consequently, his or her answers in the grid form. The auditors who participated in the retest were also asked about possible errors or omissions concerning the choice of elements and constructs. Generally the respondents were quite satisfied with these choices, thus substantiating that a thorough pilot study had been made and increasing the confidence in the validity of the study. Their confirmation indicates that the elements are representative of the domain.
Discussion and conclusions
The findings of this study indicate that there are two different types of tasks carried out by auditors. One type is the traditional tasks that focus on hard facts and number crunching, and involve checking certain parts of the company??™s business. These are the traditional tasks noted by Power (1997). Matters that are generally possible to verify are checked by the auditors. These are either historical events or matters that are regulated by reliable auditing procedures. These are the sorts of tasks that Swedish auditors approve of. Hence, the information that is quality assured by auditors ??“ arrow (b) and (c) in the accountability model ??“ can be described as primarily fragmentary, hard and historical. When reviewing this kind of information, for example balance sheet items, the auditors feel relatively secure (Cf. Pentland, 1993). The items can be reviewed using reliable auditing procedures and according to given laws and recommendations. When auditors have a clear structure and objectified knowledge to work 24
with, the responsibility is partly shifted from the individual level to the auditing profession (Cf. Power, 2003). The auditors feel that this shift in responsibility reduces the risk of their being held liable to pay damages.
The second type of task carried out by auditors focuses on information seen from the perspective of the investors. This kind of comprehensive, soft and future-oriented information is more difficult to review. The lack of reliable auditing procedures makes the auditors feel rather insecure and the fear of being held liable for damages is seen as a problem. This kind of information is rarely reviewed and commented on by auditors. Swedish auditors tend to leave this area of information to the companies, financial analysts and journalists. As a consequence this information ??“ arrow (a) in the model ??“ is normally presented to the accountees without being quality assured.
Figure 4 about here
Swedish auditors do not seem to be particularly eager to incorporate a more forward-looking quality control as part of their auditing domain. A possible explanation is that they feel secure on their own territory, together with accounting consultants and bookkeeping firms. ???Sticking to the knitting??? also produces higher revenues than jeopardising one??™s reputation in new and perhaps dangerous areas. According to Arnold et al (1999) revenue from non-audit services has become increasingly important as auditing firms seek to increase their profitability. The assistance and consulting offered by Swedish auditors is closely related to traditional auditing.
In June 2003 two expert panels consisting of notable representatives of the Swedish accounting and auditing profession and regulative institutions were invited to comment on the 25
results of the study. One of these panel debates was organised as a public seminar. These sessions confirmed the reluctance to see a more forward-looking quality control as part of the auditing domain. The representatives of the accounting and auditing profession used expressions such as ???back to basics???, ???cleaner review??? and ???not playing financial analysts??? when they discussed the results of this study and the auditing domain. The common feeling was that auditors should review and comment on some of the information, primarily hard and historical, provided by the accountors and delivered to the accountees. The importance for accountees??™ decision-making of other information, primarily soft and future-oriented, should be left to financial analysts and other actors to judge.
The findings show that Swedish auditors appear to be more focused on the content of the information rather than the usefulness of the information to investors and other interested parties. It seems more important that accountors are seen to have done the ???right things??? ??“ that some of the information is checked and verified ??“ than that the information forms a useful basis for decision-making. Bazerman et al (1997) and Haynes et al, 1998) emphasise that auditors might be abdicating their duty of protecting investors. The results of this study point in the same direction. The respondents consider themselves to be less responsible to the accountees than to themselves or to the accountors. Two examples indicate that this is the case. First, auditors devote little time and effort to issues of major importance to investors and share prices. Instead they devote a relatively large amount of time to issues that are not of vital importance to investors, but that can be satisfactorily verified and reviewed by the auditor. Second, of those respondents who thought that reviewing information about the 26
future was significant to investors and other interested parties, only a few were also prepared to make statements about such information. Accordingly the auditors are rather uninterested in arrow (c) in the model.
The gap between what auditors do and what they perceive to be important to investors and other interested parties can be related to the ???expectation gap??? (Cf. Koh & Woo, 1998). The gap described here is maybe even more interesting, because the auditors spend their time reviewing auditing issues that they themselves do not consider to be of primary importance to investors and other interested parties. The auditors??™ work does not in any way violate laws and regulations, but they appear to be aware that they do not review the issues of most importance to investors and other interested parties. To solve this problem, they claim that they share the duty to protect investors and other interested parties with financial analysts. In fact, the auditors seem to be most concerned with their own interests. For example, two thirds of the respondents referred to the problems it would cause them as an argument for not reviewing and providing statements about information that they do not produce according to current praxis.
The expert panels confirmed the finding that Swedish auditors feel less responsible to the accountees than to themselves. The representative of the shareholders expressed the need for more independent auditors, i.e. auditors not afraid of questioning managers (accountors). There is a need for auditors with integrity who are willing to open a dialogue with investors and other interested parties (accountees). The representatives of the accounting and auditing profession were more interested in discussing the fact that laws and recommendations had become more detailed and that there is a risk that in the future auditors will be expected to act more like lawyers when reviewing accounting information. 27
The findings also point to the traditional orientation of Swedish auditors. A majority of the respondents were rather unwilling to make more than minor changes in their work over and above that required by adjustments of laws and recommendations. The most common proposal for change was to reduce the aspect of tax supervision included in the work. In general the respondents were very negative to the idea of providing statements about any information except for that elicited according to current praxis.
The expert panels confirmed this traditional attitude. External pressures have only succeeded in bringing about two significant changes in the work of auditors in Sweden in the last 25 years. They are both related to compliance audits. These changes have taken place partly because investors and other interested parties would like to see fraud detection included in the auditors??™ tasks (Cf. Humphrey et al, 1993). One of the changes was that auditors should check up on taxes and other charges. In this study more than four of ten respondents explicitly argued that it is the tax authorities that should review taxes and other charges, such as employers??™ contributions to social security. The other change was that auditors are now obliged to report economic crimes, another issue criticised by auditors in this study. The unwillingness of auditors to handle certain issues or to accept a role in the detection of fraud or to provide statements about information outside current praxis, are facts that indicate that Swedish auditors are not only traditional, but also rather uninterested in reducing the ???expectation gap??? between auditors and investors and other interested parties.
This traditionalism is connected to the characteristics of the auditing profession. Auditors work to great extent according to legislation and praxis without really considering if things can be done in any other way. Fischer (1996) refers to several examples of traditionalism. Auditors rely on past experiences, both their own and those of others, as documented in guides for audit planning and the execution of current audits. Working methods tend to be taken for granted and are not questioned. Auditors have a tendency to rely strongly on objectified knowledge such as ???what was done last year??? when planning and executing their work. The widespread traditional attitude among Swedish auditors may be a problem. It makes it rather difficult to extend the information flow in arrow (b) and (c) in the accountability model and thus reduce the information flow in arrow (a).
Comparisons made in this study point to rather limited differences between different categories of auditors. Irrespective of experience, seniority level, gender, firm or location, the thought patterns of the auditors do not diverge significantly from those found in the mean grid. In fact, the correlation values within and between different categories of respondents are in the same range as the correlation values in the mean grid. Nor could any decisive differences be detected between experienced and less experienced auditors in their view of soft elements and what is considered important for investors. The lack of differences may indicate a relatively rapid process of socialisation into the auditor profession, and also the importance of structure and traditionalism.
However, some respondents recognised the importance of judgements about soft and futureoriented information to investors and other interested parties for decision-making. Auditors in listed companies perceive soft elements as more important to investors and other interested
parties. A possible interpretation is that auditors in listed companies focus more on arrow (c) in the model than other auditors.
Possible future reforms and further research
Regardless of whether the information is controlled by laws and regulations or not, it would be helpful to both investors and other interested parties if the information for their decisionmaking was quality assured. Therefore it would be desirable for the information flow from accountor to accountee following arrow (b) and (c) to increase and the flow following arrow (a) in the accountability model to decrease.
If the auditors are to handle this task and quality assure the comprehensive, soft and future oriented information for accountee decision-making, some changes need to be made. It is necessary to develop improved auditing procedures for auditing and assessing this kind of information. Partly these procedures can build upon the familiar auditing procedures for fragmentary, hard and historical issues that the auditors spend time and effort on today. Serious judgements about future-orientated information can hardly be made without considering what has happened earlier. But this is not sufficient. For example, in order to check and judge CEO comments the auditors must be capable of making trend-analyses and industry-related judgements. One possibility is to use some kind of benchmarking. Comments from one manager could be compared with previous comments from the same manager and with current comments from other managers, active in the same industry or in other industries. An extended auditor domain also requires a change in the professional training of
auditors. A deeper understanding of creative judgement and a more comprehensive approach must be learnt, in addition to the more traditional elements of the training.
We do not claim that every auditor should be able to check and judge all sorts of events. Instead auditors could specialise on different tasks, and they could co-operate to produce more comprehensive auditing than at present. A retrospective auditor could be a teammember that focuses on information that can be relatively easily verified. A future-orientated auditor could ??“ with assistance from internal and external experts ??“ concentrate on information of great importance to investors and other interested parties. The question of payment need not necessarily constitute a major problem. It can be solved according to existing praxis in Sweden. Retrospective auditors could be paid by the accountors whereas future-oriented auditors, like the financial analysts of today, could be paid by the accountees.
Considering the traditional attitude of Swedish auditors, other parties involved must take action if any changes are to be made within the scope of a market-oriented solution. Without pressure from accountees it seems difficult to reduce the information flow in arrow (a) and extend the information flow in arrow (b) and (c) in the model. For example, if it became easier to attract capital to companies that allow auditors to make their internal memorandum public, it could be a driving force towards increased openness and a larger proportion of quality assured information.
One suggestion for further research is to link the accountability model to field-studies. An increasing number of field-oriented studies of organisational life in its context have been made in accounting research. Nevertheless, there are few studies into the practice of auditing in its real context and little research that can be called ???field-work??? has been conducted in the 31
sphere of auditing (Gendron & Bedard, 2001; Power, 2003). We therefore suggest that there is a need for more field-studies in auditing research, i.e. in-depth research and longitudinal studies of what auditors really do, what issues they work with, how they interact in working groups, and with clients and investors. As it is also relevant to discover how other actors perceive auditing and auditors, it is important to focus attention not only on the auditors.
The accountability model can also be used to compare the thought patterns of accountors (management representatives) and accountees (investors and other interested parties) regarding their view of the way in which auditors review and make statements about information provided by different kinds of public companies. It can also be of interest to compare the complexity and homogeneity of the thought patterns of different groups of actors in the model.
Figures and tables
Quality assured information for decision-making by accountees (c) Auditor***
Information for decision-making by accountees (a) (b) Accountor*
Information to be quality assured
* Accountor: Management representatives, agents in charge of operations with stewardship obligations and who are obliged to provide true and fair information. ** Accountee: Investors and other interested parties entitled to information for their decisionmaking process. *** Auditor: Independent professional, responsible for assuring quality of information, i.e. that it gives a true and fair view and is reasonably complete.
Figure 1: The accountability model
Easy to review (1) – Difficult to review (7) Not prediction dependent (1) – Very prediction dependent (7) Limited effort (time) (1) – Great effort (time) (7) Low review precision (1) – High review precision (7) Statements insignificant (1) – St. sign. for company image (7)
Bad receivables Environmental crime Stock price drop CEO comments Productivity
3 4 7 4 3
1 1 6 7 3
5 7 3 3 6
5 2 1 1 7
7 4 3 5 1
Figure 2: A grid example
1 Easy to review 11 Statements insignificant 12 Not oral info. dependent
5 10 6 7
Often reviewed High review precision Competence completely suf. Great effort (time)
COMPONENT 2 ???Number crunching??? 3 Very prediction dependent 8 St. sign. for share price 2 St very val for investors 4 Internal control not imp. 9 Industry cons. required
5 10 6 7 1 11
(Purchase of shares)
(Stock price drop)
9 4 2 8
???Past??? COMPONENT 1 ???Future???
8 2 4
9 I. considerations not req. 4 Int. ctrl. very important 2 Statements not valuable 8 Statements insignificant 3 Not prediction dependent (Productivity)
12 11 1
K 7 6
???Creative comprehension??? 7 6 10 5 Limited effort (time) Competence not sufficient Low review precision Seldom reviewed 12 Very oral info. dependent 11 St sign for company image 1 Difficult to review
Figure 3: Rotated mean grid for 82 auditors
Quality assured fragmentary, hard and historical information for decision-making by accountees (c) Auditor Comprehensive, soft and future-oriented information for decision-making by accountees (a) (b) Accountor
Fragmentary, hard and historical information to be quality assured
Figure 4: Information flow and auditor focus in the accountability model
Mean inter correlation Full grids: the auditors and their vectors .360 with scores for all elements on all constructs as elements .330 Elements: grids with the auditors and their vectors of factor scores on the first component Constructs: grids with the auditors and .469 their vectors of factor loadings on the first component
Variance described by two components (%) 42
Table 1: PCA of mean grid for all respondents. Mean intercorrelation and percentage of the variance explained by two components.
References Arens, A. and Loebbecke, J. (1997) Auditing – an integrated approach (7th ed.). Upper Saddle River, New Jersey: Prentice Hall. Arnold, D., Bernardi, R. and Neidermeyer, P. (1999) The effect of independence on decisions concerning additional audit work: a European perspective. Auditing: A journal of practice and theory, Supplement 18: 45-67. Bannister, D. and Fransella, F. (1986) Inquiring man (3rd ed.). London: Routledge. Bazerman, M., Morgan, P. and Loewenstein, B. (1997) The impossibility of auditor independence. Sloan Management Review, (Summer): 89-94. Berger, P. L. and Luckmann, T. (1967) The social construction of reality. New York: Doubleday Anchor. Besanko, D., Dranove, D. and Shanley, M. (1996) Economics of strategy. New York: John Wiley & Sons. Bonner, S. E. and Pennington, N. (1991) Cognitive processes and knowledge as determinants of auditor expertise. Journal of Accounting Literature, 10: 1-50. Bryman, A. (2001) Social research methods. Oxford: Oxford University Press. Burchell, S., Clubb, C., Hopwood, A., Hughes, J. and Nahapiet, J. (1980) The roles of accounting in organizations and society. Accounting, Organizations and Society, 5(1): 5-27. Campbell, D. T. (1988) Methodology and epistemology for the social sciences. Chicago: The University of Chicago Press. Chandler, R. A. and Edwards, J. R. (1996) Recurring issues in auditing: back to the future Accountability, Auditing and Accountability Journal, 9(2): 4-29. Dirsmith, M. W. and McAllister, J. P. (1982) The organic versus the mechanistic audit. Journal of Accounting, Auditing and Finance, (spring): 214-28. de Ruyter, K. and Wetzels, M. (1999) Commitment in auditor-client relationships: antecedents and consequences. Accounting, Organizations and Society, 24(1): 57-75. Donaldson, L. and Davis, J. H. (1991) Stewardship theory or agency theory: CEO governance and shareholder returns. Australian Journal of Management, 16 (1): 49-64. Edenhammar, H. and Hagg, I. (1997) Makten over redovisningen [The power over accounting]. Stockholm: Studieforbund Naringsliv & Samhalle. Eisenhardt, K. M. (1989) Agency theory: An assessment and review. Academy of Management Review, 14(1): 57-74.
Eklov, G. (2001) Auditability as interface: negotiation and signification of intangibles. School of Business Research Reports no. 2001:09. Stockholm: School of Business, Stockholm University. Epstein, M. J. and Geiger, M. A. (1994) Investor views of audit assurance: recent evidence of the expectation gap. Journal of Accountancy, 17 (January): 60-6. Fiet, J. (1995) Reliance upon informants in the venture capital industry. Journal of Business Venturing, 10: 195-223. Fischer, M. J. (1996) ???Real-izing??? the benefits of new technologies as a source of audit evidence: an interpretive field study. Accounting, Organizations and Society, 21(2/3): 21942. Flint, D. (1988) Philosophy and principles of auditing ??“ an introduction. London: Macmillan. Fransella, F., Bell, R. and Bannister, D. (2003) A manual for repertory grid technique (2nd e