Correspondence to: Dr C.A. Jubb, Department of Accounting,

The University of Melbourne, Victoria 3010, Australia.

Email: cajubb@unimelb.edu.au

Fax: +61 3 9349 2397

Auditors and directors may develop personal attachments over time based on trust and familiarity, and these personal ties seem important for the maintenance of long-term auditor-client relationships. This study examines the tenure of the audit engagement in the presence of these links, which is expected to be longer than auditor-client relationships not so linked. Results indicate director-auditor links are positively associated with auditor tenure, and the retention of auditors beyond the critical four-year period identified by Levinthal and Fichman (1988).

Key Words: Auditor Tenure, Interlocking Directorates, Auditor Rotation, Audit Quality

It has been argued that auditing is a service that is difficult to evaluate without being experienced, since its quality is not easily discernible (Pennings, Lee and van Witteloostuijn, 1998; Craswell and Francis, 1999). In such a circumstance relationships between individuals are likely to influence the decision to select, or continue, relationships with service providers (Koreto and Harding, 1996).

The impact of personal connections in exchange relationships has been well established (eg Pfeffer, 1994) and these ties have been examined in the context of auditing service provision. One of these studies, Davison, Stening and Wai (1984), investigates the impact of personal attachments (captured by interlocking directorates) on choice of auditor. Davison et al. (1984) report a positive and significant association between the number of director interlocks attributable to a company and the probability that the interlocked companies are audited by the same public accounting firm. Jubb (2000), employing a more robust and detailed empirical analysis, finds results consistent with those of Davison et al. (1984).

The primary purpose of this study is to investigate the association of director-auditor links with auditor tenure. That is, where companies with interlocking directors are audited by the same public accounting firm, it is expected that the tenure of that auditor is significantly longer than for that in respect of companies not so linked.

A secondary purpose of this study is to investigate whether the personal attachments measured by director-auditor links are associated with mitigation of the pressure for an auditor switch during the critical initial four year period found by Levinthal and Fichman (1988).

For some time the incidence of interlocking directorates has been acknowledged and studied (for example Dooley, 1969; Allen, 1974; Alexander, Murray & Houghton, 1994) and explanations have been offered for this phenomenon (Mizruchi, 1996; Allen, 1974). One of the main reasons suggested for the existence of interlocking directorates is that these associations reflect corporate strategies to reduce or control important sources of uncertainty in companies’ environments (Allen, 1974; Schoorman, Bazerman & Atkin, 1981). In addition, interlocking is seen as a means of exchange of information and expertise between companies. It is therefore surprising that more research has not been focussed on the relationship between interlocking directorates and the selection and retention of auditors, especially since auditing is a relatively complex service performed in an uncertain environment (Crosby, Evans & Cowles, 1990).

As noted earlier, the primary paper addressing this issue thus far is by Davison et al. (1984) who provide evidence that the presence of interlocking directorates is important to the choice of auditor, creating significant links between directors and auditors. Another study, Seabright, Levinthal & Fichman (1992), investigated the effect of attachment of individuals primarily responsible for the auditor-client exchange on the likelihood of auditor switching. The study showed that while changes in resource requirements (of the client) and resource provisions (of the auditor) increased the likelihood of an auditor switch, the development of attachments between boundary spanners attenuated this effect on auditor change. Seabright et al. (1992) suggest that the auditor-client relationship relies largely on personal knowledge and trust and that these act as disincentives for clients to change auditors.

This current study provides a more direct and arguably more meaningful measure of the impact of personal attachments on exchange relationships than that employed by other researchers, especially Seabright et al. (1992) who operationalised their individual or personal attachment variables through tenure of company officers (that is CEO, CFO etc). Those authors noted that there might be other consequential attachments within a larger network of relationships in which an auditor and client are involved (Seabright et al., 1992: 155). The current study responds to this call by using director-auditor links as a measure of personal attachment in testing an hypothesised association between director-auditor links and auditor tenure.

It can be argued that factors increasing the likelihood of a company changing its auditor can be seen also as factors reducing the likelihood of continued tenure. While this is true, such an ‘auditor change’ approach risks ignoring variables impacting specifically on the length of auditor tenure rather than on auditor change. Further, examination of the determinants of tenure length may be as important, if not more important, than the determinants of auditor change to accounting firms. As much as it may be useful to explain why auditor-client relationships end, an indication as to what might make them last is arguably of greater practical significance to firms engaged in these relationships.

Additionally, this study potentially informs the debate concerning mandatory auditor rotation. Concerns have been raised about the impact on audit quality and auditor independence when auditor tenure is for particularly short or long periods (eg Latham, Jacobs & Roush, 1998; Raghunathan, Lewis & Evans III, 1994; Aldhizer III & Lampe, 1997) because of the impact on familiarity with the client. While audit quality is not specifically tested in the context of this study, any association between director-auditor links and longer auditor tenure may accentuate concerns over auditor independence, an issue of major concern to the accounting and auditing professions and those who regulate them (eg Levitt, 1998). In fact, auditing Standards and ethics statements, regardless of the jurisdiction in which they are based, generally include commentary on tenure length.

Where an interlocking director comes into contact with the same auditor across other companies on whose boards (s)he sits and auditor tenure is relatively long, the potential impact on auditor independence is unclear. Longer tenure in its own right has been criticised as potentially reducing independence (Raghunathan et al., 1994; Aldhizer and Lampe, 1997). Interlocking directorates (regardless of their links with auditors) have also attracted criticism. It is therefore likely that any significant positive association between director-auditor links and auditor tenure will heighten concerns with respect to auditor independence. The aim of this paper is to investigate whether such a positive association exists.

The remainder of this paper is presented as follows. The next section discusses the prior literature and develops the hypotheses to be tested. The variables selected for the testing of the hypotheses and their measurement are then discussed. An outline of the research design and sample data is then provided. Finally, the results are presented and discussed, and limitations and opportunities for future research are outlined.

An interlocking directorate arises when a director sits on two or more company Boards. Many explanations have been offered for the existence of interlocking directorates covering a range of theoretical prescriptions. These perspectives have included transaction costs (Williamson, 1991), agency theory (Eisenhardt, 1989) and class theories (Koenig and Gogel, 1981). However, the most relevant explanation for their existence, in terms of the context relevant to this study, is that they serve to reduce or control uncertainty in business environments (Allen, 1974; Schoorman et al. 1981; Mizruchi, 1996). Allen (1974: 395) specifies three main ways in which interlocking directorates attempt to reduce environmental uncertainty. These are (1) by the exchange of information and expertise between companies; (2) by providing a stable means of communication and liaison between companies; and (3) by advising management concerning the relationship of the company to its external environment. However, when interlocking directors are systematically associated with a common auditor across their various board holdings, it is not clear that benefits exist for all stakeholders.

Unlike other products or services, the quality of an audit is not readily discernible. It cannot be judged from the outside and must be experienced to be evaluated (Pennings, Lee and van Witteloostuijn, 1998; Craswell and Francis, 1999). Interlocking directors holding multiple board positions are in one of the best positions to judge the relative quality of audits due to their experience with various service providers. Their experience gives them the ability to advise on and perhaps contribute to selection of the most appropriate auditor for companies on whose boards they sit. Sharing this knowledge with boards of other companies on which they sit reduces the costs of evaluating the strengths and weaknesses of potential auditors.

Zajac (1988) indicates that multiple directorships allow directors to view a panorama of their companies’ environments within which to monitor and control uncertainties. Sharing this outlook with auditors, who may be knowledgable about their clients’ business environments, creates synergies that potentially enable difficulties to be overcome more smoothly.

In the accounting literature, little attention has been paid to the relationship between interlocking directors and auditors. From the research that does exist, it appears that there is a tendency for interlocking directors to employ the same auditor across the group of companies through which they are interlocked (Davison et al., 1984; Jubb 2000).

Many factors have been found to influence the length of auditor tenure. However, the focus of this section is the expected positive association of the hypothesis variable, director-auditor links, with auditor tenure.

As noted, director-auditor links develop, in part, due to the building of attachments and personal ties between directors (in their roles as boundary spanners) and auditors. Literature in the field of management and marketing would suggest that the development of personal ties is important to the development and continuance of corporate relationships. De Ruyter and Wetzels (1999) found that trust and pleasant business partnerships increase the commitment of clients to the relationship and their intentions to continue it. Similarly, numerous researchers have concluded that the choice of continuing business relationships depends on trust that emerges between organisations due to repeated personal attachments and ties (eg. Cook, 1977; Levinthal and Fichman, 1988; Gulati, 1995).

With respect to the impact of personal ties on auditor tenure specifically, Seabright et al. (1992) examined the effect of personal attachments on the dissolution of the auditor-client relationship. They found that attachments between client and auditor organisations occur mainly at the individual level and their findings suggest that while other factors may act as pressures for auditor change, it is personal attachments that attenuate the impact of these influences and are critical to the maintenance of the relationship.

Relationships generated in the presence of director-auditor links are argued to allow the development of mutual dependence due to the greater stability of the alliance. This dependence relies on trust as an integral ingredient in the relationship. Therefore, it is posited that the trust and dependence manifested in the auditor-client relationship will be influential in client decisions to retain the auditor, and that the development of personal ties or attachments over time resulting from director-auditor links will be positively associated with auditor tenure.

The results of Levinthal and Fichman (1988) provide the basis for the second hypothesis. Those authors found that the likelihood of auditor change increased up to the fourth year of tenure, before reducing. They posited that since the client receives feedback about the desirability of the auditor-client attachment only on an annual basis, it might take a number of years (up to four years) for initial favourable beliefs to change sufficiently for the attachment to be ended. Given this result, Hypothesis 2 examines the impact of director-auditor links on the realisation of auditor tenure surrounding what appears to be the critical four-year period. That is, it is hypothesised that personal attachments between directors and auditors will ameliorate the pressure for an auditor switch within the first four years of auditor tenure.

Auditor tenure is the dependent variable for the testing of Hypotheses 1 and 2 However, a different measure of tenure is employed in each as appropriate. For Hypothesis 1 a continuous (but capped) measure of tenure is used (AUDTEN). Auditor tenure can, and often does, last for many years. Given the findings of Levinthal and Fichman (1988) that the majority of auditor switches occur in the first four years of the auditor-client relationship, the limit on tenure should be greater than four years but short enough to maximise company survival over the examined period. In this study, auditor tenure is measured continuously and is censored at seven years. While accepting the survivorship bias that any censoring of tenure may induce, a trade-off must be made between the logistics of data collection and adequately capturing the length of auditor engagement of the companies in the sample.

For Hypothesis 2, which involves examination of auditor tenure surrounding the critical four-year period in the auditor-client relationship, a dichotomous variable is used to represent periods either side of the four-year period. As discussed earlier, Levinthal and Fichman (1988) found that the likelihood of auditor switching increases up to the fourth year of auditor tenure, where it peaks, before declining. Therefore, the dependent variable for Hypothesis 2, a categorical measure of auditor tenure (CATTEN) takes the value 1 if the tenure of the incumbent auditor is greater than four years and 0 otherwise.

Director-Auditor Links (ALOCKYRS)

The preceding discussion indicates the rationale behind the presence of director-auditor links as the hypothesis variable. This variable represents the cumulative total of director-auditor links per company observation over the measurement period. That is, ALOCKYRS measures the total number of years that an observed company’s interlocking directors have had a personal attachment link with a particular audit firm. This is appropriate given that the hypothesis is concerned with personal attachment links over the observed length of auditor tenure. Furthermore, it is likely to be the endurance of these director-auditor links that will most affect auditor tenure, not the occurrence of a link at any particular point in time.

Director-auditor links are measured over their existence and capped at seven years. ALOCKYRS is measured as the total number of director-auditor links for each focal or observed company (regardless of the number of directors interlocking with another company) for each given year, summed over the seven-year period.

The following example of the calculation of ALOCKYRS is based on a hypothetical sample of three companies over three years. ALYRS represents the number of total director-auditor links for any given year which, when summed across the potential seven-year period, gives ALOCKYRS. The measurement of this variable over a potential seven-year period is consistent with the measurement of auditor tenure, the dependent variable.

For a director-auditor link (ALYRS) to be included in the calculation of ALOCKYRS, two conditions must be met. Firstly, a given director must sit on at least one board amongst the sample companies other than that of the observed or focal company and secondly, the observed and ‘other’ company(ies) must engage the same auditor. Note that failure to meet the second condition reveals why, in the above example, there are no director-auditor links (ALYRS) for Company 2 in Year 3. Although some of Company 2 directors sit on other sample company boards, Company 2 does not share the same auditor in year 3 as either Company 1 or Company 3.

The control variables used in the testing of Hypotheses 1 and 2 are based primarily on those used in prior research that has examined auditor tenure or auditor switching. Table 2 provides a summary of the relevant auditor tenure and change studies that have influenced the models used in the current study. The studies by Seabright et al. (1992) and Ritson, Jubb and Houghton (1997) are the source of the majority of variables. The variables used in a number of auditor change models (eg Haskins and Williams, 1990; Levinthal and Fichman, 1988) are relevant also because of the aforementioned link between auditor tenure and change.

Complexity (COMPL)

Simunic’s (1980) work indicates that providing assurance on financial statements is more demanding where an audit is more complex. Further, auditees are more likely to choose an auditor most capable of dealing with such complexities. The inclusion of complexity recognises the positive significance of this variable found by Walker, Casterella & Moet (1998).

Consistent with Simunic (1980) and Walker et al. (1998), complexity is measured as the proportion of inventory and receivables to total assets, given that both studies find these areas to add to the complexity of an audit.

Distress (DISTRESS)

Previous studies have identified the importance of financial distress for the likelihood that a company will change auditor (for example Schwartz and Menon [1985]; Haskins and Williams [1990]). In addition, distressed companies are more likely to be associated with damage to the reputation of auditors in the event of litigation (Krishnan and Krishnan, 1997), and shareholders of distressed companies are likely to seek compensation for losses in the event of failure (Menon and Williams 1994). For these reasons, auditors may be less likely to retain such clients (DeFond, Ettredge and Smith, 1997; Krishnan and Krishnan, 1997).

Distress is measured using re-estimated parameters of Altman’s (1968) model (the original model has been used in prior literature eg. Seabright et al., 1992; Schwartz and Menon, 1985). This revised model was re-estimated by Constable and Woodliff (1994) based on Australian company data and was found to improve the original model’s predictive ability.

Risk (RISK)

Similar to distress, risk (in the context important to this study) is a measure of the attractiveness of a client to an auditor. Highly leveraged companies have a greater chance of failure than do those with less debt. Since auditors are perceived to have ‘deep pockets’ (Wallace, 1987), the stakeholders of failed companies (such as shareholders or creditors) may seek compensation from the auditor for losses incurred. This acts as a disincentive to auditors for continuance of the auditor-client relationship and is predicted to result in lower auditor tenure.

Risk is measured as the proportion of debt to total assets. While other factors may influence the risk of a company, leverage has the advantage of explanatory power and parsimony.

Qualified Opinion (PRIORQUAL)

Chow and Rice (1982) and Schwartz and Menon (1985) found that clients in receipt of a qualified audit opinion in the prior year have a higher tendency to switch auditor. This may be because the directors seek to engage an auditor whose views are more in line with those of management. These new auditors are therefore assumed less likely to qualify.

A dichotomous (dummy) variable is used to operationalise a qualified opinion, coded 1 if the audit report of a company was qualified in the prior year, 0 otherwise.

Auditee Age (AGE)

Auditor tenure may depend on the longevity of companies in the sample. If an audit client has been in existence only for a limited period, it would follow that the tenure of its auditor cannot exceed the client’s age (although it may be less than this). Further, older companies have had the time, and therefore the opportunity, to build personal attachments of the nature tested by the hypothesis variable in this study. Thus, the age of the client is predicted to positively influence auditor tenure and needs to be controlled for.

Age is measured continuously as the number of years a company has been listed on the Australian Stock Exchange (ASX).

Audit Fee (AUDFEE)

Audit fees that are perceived to be excessively high have been found to influence auditor change (Haskins and Williams, 1990; Eichenseher and Shields, 1983; Shockley and Holt, 1983). Thus, the higher the audit fee, ceteris paribus, the lower the expected tenure.

Audit fees are measured as the dollar amount paid to the principal auditor only for the auditing of a client’s financial statements as disclosed in the company’s annual report.

Auditee Growth (GROWTH)

The resource requirements of a company change throughout its existence. Thus, growth may influence the decision to change auditor due to a difference between current resource requirements of clients and the ability of an audit firm to provide these resources (Seabright et al., 1992). Indeed, Haskins and Williams (1990) found growth to be a significant determinant of auditor change. Consequently, auditor tenure is expected to be lower for companies experiencing growth.

Consistent with previous studies (for instance Haskins and Williams, 1990) growth is operationalised as the percentage change in revenue from the prior year.

Director Tenure (DIRTEN)

This study proposes that directors may gain a familiarity with, and attachment to, an auditor over time. Furthermore, Seabright et al. (1992) found the tenure of the CFO and audit committee members are negatively associated with auditor changes. Thus, auditor tenure is predicted to increase with the average tenure of directors on a company’s board given that longer director tenure provides a greater time frame for an attachment to develop.

Director tenure is measured as the average number of years’ tenure (capped at seven years) for a focal company’s board members.

Non Audit Services Purchased (NAS)

Beck, Frecka and Solomon (1988) report that companies that purchase high levels of recurring NAS from their auditor tend to have longer auditor tenure. In an Australian context, Butterworth and Houghton (1995) report findings consistent with this result.

NAS is measured as the remuneration paid to a company’s principal auditor for non-audit services as disclosed in the company’s annual report.

Auditee Size (SIZE)

Prior research by De Angelo (1981) found that as the size of clients increases, they are more likely to select larger auditors. Thus, it is expected that as the auditee size increases companies will tend to select Big 6 (now Big 5) auditors. Additionally, Levinthal and Fichman (1988) found that auditor size is highly significant in explaining the expected duration of the auditor-client relationship.

Consistent with prior research (eg Francis and Wilson, 1988; Levinthal and Fichman, 1988), client size is measured as total assets.

Big 6 (BIG6)

The size of audit firms has been shown to have a systematic effect on the duration of the auditor-client relationship. For instance, prior research [Levinthal and Fichman (1988)] has shown that client relations with Big 8 (Big 6 in this sample) firms are likely to last longer than those with non-Big 8 (non-Big 6) auditors.

BIG6 is captured by a dichotomous variable taking the value 1 if a company’s auditor is a member of the Big 6, and 0 otherwise.

Industry Specialist (SPECAUD)

The existence of an industry premium specialist has been shown to result in fee premia attributed to the expertise and quality that such an auditor exhibits (Craswell et al., 1995, DeFond, Francis and Wong 2000). Further, research by Haskins and Williams (1990) indicates that auditor switches can be explained partly by clients preferring to choose a specialist auditor. Thus, auditor tenure for industry specialists is likely to be longer than that for non-specialists.

An industry specialist auditor is deemed to exist if at least one audit firm within in an industry receives at least 15% of the total industry audit fees. Consistent with Craswell and Taylor (1991) and Craswell, Francis and Taylor (1995), at least 30 companies must exist in an industry for a specialist to be deemed to exist. Auditor specialisation is measured in the prior year, because the decision to retain or change auditor may depend, in part, on a company’s perception of which audit firm(s) is a specialist auditor in its market. If an auditor meets the aforementioned criteria in the prior year, the dichotomous variable is coded 1, and 0 if it does not.

Therefore, the model to be tested takes the following form:

a + b ₁ALOCKYRS + b ₂COMPL - b ₃DISTRESS -

b ₄RISK - b ₅PRIORQUAL + b ₆AGE - b ₇AUDFEE - b ₈GROWTH + b ₉DIRTEN + b ₁₀NAS + b ₁₁SIZE + b 1₂BIG6 + b ₁₃SPECAUD + e

The variable definitions and their expected direction are summarised in Table 3.

TABLE 3 ABOUT HERE

Hypothesis 1 examines the association of director-auditor links with auditor tenure (measured continuously). Rather than using an ordinary least squares (OLS) regression, multivariate Tobit analysis is used due to the censoring of auditor tenure.

Hypothesis 2 examines the potential impact of director-auditor links in influencing the extension of auditor tenure past the four-year barrier argued to be critical by Levinthal and Fichman (1988). Given that the dependent variable for the testing of this hypothesis is dichotomous, logistic regression is used.

The same sample companies are used for the testing of both hypotheses. The sampling frame consists of the top 242 companies by total assets listed on the ASX in the year 1995, and meeting the required data considerations. To be included, companies had to be audited by a single (3 deletions), private sector auditor (2 deletions) and have been listed for at least 2 years (16 deletions). In addition, only companies with financial statements denominated in Australian dollars (19 deletions) and all data (2 deletions) were included. The final sample size was 200 companies (see Table 4 for the sample criteria).

It was necessary to delete companies with more than one auditor (2) because of the inherent difficulty this would cause in the calculation of auditor tenure and ALOCKYRS. Similarly, the inclusion of companies listed for at least two years enabled the collection of prior year data for the growth, auditor industry specialist and qualification variables. The ‘other’ category consists of companies for which not all data was currently available.

The data was hand collected from a variety of sources including Who Audits Australia? (Craswell, 1996), The Australian Financial Review Shareholder handbook, Jobson’s Year Book of Australian Listed Companies (1989-1996), Jobson’s Year Book of Mining Companies (1989-1996), Australian Stock Exchange Datadisc, Australian Graduate School of Management (AGSM) Annual Report Microfiche File and Connect4.

Hypothesis 1 examines the association of director-auditor links with auditor tenure measured continuously over a maximum seven year period, left censored at two years and right censored at seven years. The descriptives reported in Table 5 show that the mean tenure for all companies in the sample is 5.4 years. This indicates that the average company observation has an audit relationship that has lasted for more than five years.

The majority of companies (59.5%) exhibit auditor tenure of at least seven years. This is interesting given the findings of Levinthal and Fichman (1988) that auditor switches are most likely to occur in the first four years of the auditor-client relationship. These findings suggest two possible alternatives. Firstly, average auditor tenure [since the study by Levinthal and Fichman (1988)], has tended to increase. Alternatively, given that Big 6 auditors have, generally, longer tenure than non-Big 6 auditors (Fichman and Levinthal, 1991), the incidence of auditor switching among the largest 200 companies (which tend to be audited by Big 6) may be so low as to have no significant impact on average auditor tenure. The latter explanation appears most pertinent given that auditor switching in Australia is relatively rare in any given year.

The descriptive statistics for the overall sample (Table 5) show that the mean number of director-auditor links (ALOCKYRS) is 9.19, with a maximum of 68. This indicates that on average, over the seven year measurement period, there are 9.19 links created where the directors of a sample company encounter the same auditor at other companies where they have board membership. The mean number of years the sample companies have been listed on the ASX (AGE) is 21, indicating that on average the companies are well established. This is to be expected since the sample is comprised of large, and generally stable, companies. In addition, the mean tenure of directors on the sample companies’ boards (DIRTEN) indicates that on average directors in the sample were on those boards for almost four years.

In terms of financial characteristics, the mean level of leverage (RISK), 0.478, suggests that on average the companies in this sample do not have overly high debt to equity ratios. The mean distress score is 0.036, indicating that the sample companies have relatively sound financial health. The dichotomous variables show that on average, only 3.5% of companies were issued a qualified opinion in the prior year (which is expected given that the sample includes large companies), 89% of companies are audited by a Big 6 auditor and 18% of companies engage a specialist Big 6 auditor.

The Pearson’s correlation matrix is reported in Table 6. Only two instances exist where independent variables have correlations over 0.5 (LOGFEES has a 0.555 correlation with LOGNAS and a 0.729 correlation with LOGSIZE).

INSERT TABLE 6 ABOUT HERE

The multivariate results for testing Hypothesis 1 are shown in Table 7. The model is significant with an adjusted R² of 43%. Robust analysis that applies the Huber-White technique is used since testing reveals the presence of hetereoscedasticity. These results show that six of the thirteen independent variables are significant with respect to auditor tenure. The coefficient of the hypothesis variable, ALOCKYRS, is positive and highly significant (p-value = 0.005 [one-tailed]) indicating that director-auditor links are significantly associated with the length of auditor tenure. This finding supports the results of Seabright et al. (1992) who found that personal attachments between those integral for the auditor-client relationship decreased the likelihood of auditor switching.

A number of control variables were also significantly associated with auditor tenure (AUDTEN). Director tenure (DIRTEN) (p-value = 0.000 [one-tailed]) is highly significant in the direction hypothesised suggesting that the average tenure of directors on the Board is positively associated with auditor tenure. The number of years since listing (AGE) is also significant (p-value = 0.039 [one-tailed]) and positive as hypothesised, as is BIG6 (p-value = 0.020 [one-tailed]), while LOGGROWTH is significantly negatively associated with auditor tenure (p-value = 0.019 [one-tailed]). The significance of LOGGROWTH further supports the hypothesis and findings of Seabright et al. (1992) that changes in resource requirements of clients increase the likelihood of auditor switches.

The other control variable significant in the model is LOGFEE (p-value = 0.023 [one-tailed]), although in the opposite direction to that hypothesised. The results indicate that audit fees are positively associated with auditor tenure suggesting that fee expensiveness is not associated with auditor change. This is consistent with Simon and Francis (1988) who found that any lowballing is fully recouped within four years. In addition, given that larger audit firms predominate in the sample and are generally considered to provide a higher quality audit (De Angelo, 1981), presumably companies are willing to pay a premium (Craswell et al., 1995; De Fond et al. 2000). The positive direction of this variable therefore seems to reflect the hypothesised positive, significant relationship between BIG6 and auditor tenure.

Interestingly, RISK, DISTRESS and complexity (COMPL) are all insignificant suggesting that the financial health of clients, and difficulty of the audit task, do not significantly influence auditor tenure. However, given that the sample consists of large companies tending to be well established, there is likely to be little variation in these variables. Qualification in the prior year (PRIORQUAL) is insignificant in this study. Levinthal and Fichman (1988) suggest that where relationships have survived for several years, there is likely to be less conflict between client and auditor and, hence, a low frequency of qualified opinions. Given that the companies in this sample predominantly have incumbent auditors with at least seven years’ tenure and are healthy financially, a low incidence of qualified opinions may be expected.

The significance of the hypothesis variable, director-auditor links (ALOCKYRS), provides strong support for Hypothesis 1.

The explanatory variables used in the estimation of Hypothesis 1 are included also for the testing of Hypothesis 2. However, a categorical measure of auditor tenure is used to test Hypothesis 2, which proposes a positive association between director-auditor links and the retention of the auditor for more than the critical four year period noted by Levinthal and Fichman (1988). Based on the findings of Levinthal and Fichman (1988), the dependent variable takes the value 1 if auditor tenure is greater than four years and 0 if tenure is less than, or equal to, four years. Of the observations, 67% exhibit auditor tenure longer than four years.

The descriptive statistics in Table 5 show that the mean level of director-auditor links is greater for companies where the incumbent auditor has tenure more than four years as opposed to less than, or equal to, four years. This trend is evident for most of the explanatory variables in Table 5. Companies with auditors exhibiting more than four years’ tenure tend to have longer director tenure, are more complex and riskier, have higher fees, and are more likely to be audited by a specialist auditor than those companies where the auditor does not have tenure greater than four years. The only exception is company growth, which suggests companies whose auditors have less than, or equal to, four years’ tenure (61%) grow more quickly than those companies whose auditors have greater than four years tenure (33%).

The univariate results are presented in Table 5. The independent samples t-tests for the continuous variables (Table 5 Panel A) indicate that most of the means in the two groups (tenure greater than four years and tenure less than, or equal to, four years) are significantly different from each other (two-tailed test). The only exceptions are complexity (COMPL), DISTRESS and RISK where the means for the two groups are insignificantly different from each other. The significant result for ALOCKYRS (p=0.000 [two-tailed]) indicates that the mean incidence of director-auditor links is greater for those company observations with auditor tenure longer than four years on a univariate basis.

Univariate Chi-square tests conducted on the categorical variables (Table 5 Panel B) suggest that there is no significant difference in the incidence of prior year audit qualifications (p=0.284), BIG6 (p=0.108) or use of a specialist auditor (SPECAUD) (p =0.160) between the two tenure groups.

Although not hypothesised, initially it is necessary to replicate the findings of Davison et al. (1984) in order to establish whether a relationship exists between ASSETS, ALOCKS and NLOCKS. Path analysis (not reported) revealed results consistent with, but stronger than, those found by Davison et al. (1984). The direct path between ASSETS and ALOCKS was not statistically significant (p = 0.519), and the paths between ASSETS and NLOCKS and NLOCKS and ALOCKS were highly significant (p = 0.000). This is consistent with the findings and conclusions of Davison et al. (1984) and hence progression to the two hypotheses takes place from a solid foundation.

Table 8 presents the results of the logistic regression. Table 8 Panel B documents the results when auditor tenure is coded as hypothesised at greater than or less than four years. Table 8 Panels A and C document the results of sensitivity analysis coding as greater than or less than three and five years respectively. Panel B shows that overall, the model is significant with a Chi-square statistic of 109.45 (p =0.000), and has a Pseudo R² of 0.18%. The Hosmer and Lemeshow goodness of fit test calculates a Chi-square value of 6.859 (p = 0.552) indicating that the model provides a good fit. Furthermore the classification accuracy, another indicator of the goodness of fit, comparing predictions to the observed outcomes shows that the model overall correctly predicts 83% of the observed cases. No other studies of auditor tenure known to the author have published this information. Consequently, these results are difficult to compare. However, the findings indicate that approximately 35% of the time, the model predicts auditor tenure greater than four years when it is less than, or equal to, four years (Type II errors). Furthermore, approximately 9% of the time, the model predicts that tenure is less than, or equal to, four years when it is actually greater than four years (Type I errors).

Regulators and Professional Bodies are likely to be more concerned with longer auditor tenure in the presence of links between directors and auditors. Hence, regulators are likely to be interested in a model that more accurately predicts tenure greater than four years than tenure less than, or equal to, four years. While overall predictive accuracy is important, this implies that a model that minimises Type I errors would be of more value than a model minimising Type II errors.

As Table 8 Panels B and C demonstrate, the hypothesis variable, ALOCKYRS, is positive and significant as hypothesised when the categorical dependent variable is conditioned on tenure greater than four years (p=0.025 [one-tailed]) or five years (p=.009 [one-tailed]), but not as Table 8 Panel A shows, when conditioned on tenure greater than 3 years (p=.078 [one-tailed]). This finding supports Hypothesis 2, which predicts that director-auditor links are positively associated with auditor tenure greater than four years. As distinct from the interpretation of the findings for Hypothesis 1, the results of Hypothesis 2 suggest that director-auditor links may alleviate the pressure for an auditor switch within the critical first four years of auditor tenure (when an auditor switch is most likely [Levinthal and Fichman, 1988]).

Two control variables are also significantly positively associated with auditor tenure over four years in duration. Director tenure (DIRTEN) is highly significant (p=0.000 [one-tailed]), as is BIG6 (p-value = 0.038 [one-tailed]). Panel C reveals that the age of the company and its growth become significant once longer tenure of five or more years becomes the partitioning variable. No other independent variables are found significant in this logistic regression. Many control variables important to a continuous measure of tenure are not associated with this dichotomous measure that partitions the sample according to whether or not auditor tenure lasts beyond the critical four-year stage.

The purpose of this paper is to investigate the impact of director-auditor links (a personal attachment where interlocking directors engage the same audit firm across their company directorships) on auditor tenure. Motivation is provided not only by the paucity of empirical analysis in the auditing literature on the relationship between interlocking directorates and auditors, but also by potential policy implications of any findings for the debate surrounding mandatory auditor rotation.

The results provide support for the findings of Seabright et al. (1992) that personal attachments between directors and auditors diminish the pressure for auditor switches considering the significant, positive relationship between director-auditor links (ALOCKYRS) and auditor tenure. The results also suggest that director-auditor links facilitate continuance of the relationship beyond the first four years of tenure. This four-year period was demonstrated by Levinthal and Fichman (1988) to be important to the likelihood of auditor switching.

The findings may also inform the debate over mandatory auditor rotation. For example, the existence of director-auditor links in an environment of longer auditor-tenure could arguably appear to be an example of the nurturing of "close personal or professional relationships with clients" by auditors proscribed under the Auditing Standards and Codes of Professional Conduct in many jurisdictions.

The results show that over half of the largest 200 Australian listed companies in the sample engaged the same auditor for at least the last seven years prior to and including 1995. However, this study does not investigate whether the same engagement partner was present in each of the years of incumbency. Nevertheless, the pressure for mandatory auditor rotation, on the grounds of ensuring actual or perceived independence, may gain momentum if auditor tenure is accompanied by director-auditor links. As noted earlier, longer auditor tenure has been criticised for impairing independence regardless of any attachments between directors and audit firms. Such attachments may heighten these independence concerns.

This study is further motivated by the implications it may have for public accounting firms. As mentioned earlier, many studies have focussed on the determinants of auditor change as opposed to auditor tenure. This study could provide guidance to accounting firms by indicating some of the factors that are associated with longer auditor tenure. Nevertheless, this guidance should be acknowledged with regard to the possible independence concerns noted above.

While the findings of this research do have some interesting policy implications, they must be considered in the light of the limitations of the study. Given that the study includes data on only the largest 200 Australian listed companies, it is questionable whether the results generalise to the population of Australian companies (including smaller and private companies). In addition, the strict criteria for the inclusion of companies in the sample may detract from the study’s generalisability. Nevertheless, the sample used here is comparable to that used by Davison et al. (1984) and Jubb (2000). The capping of auditor tenure also creates a limitation in this paper. However, a measure of tenure over longer periods is accompanied by non-trivial survivorship bias issues.

Even capping tenure at seven years, the study may suffer from survivorship bias because some of the companies included in the sample may not have existed for the seven-year measurement period. However, the inclusion of the variable AGE attempts to control for this potential limitation.

This study also assumes that ALOCKYRS captures accurately the explanatory power of personal attachments. However, other types of personal attachment may exist that are not captured by director-auditor links (for instance managerial links with auditors). In addition, the effect of partner turnover, and its potential impact on auditor tenure due to the loss of personal attachment, is not considered. This study also assumes that where an audit firm merger occurs, auditor tenure is continuous if a client of either of the two firms continues to engage the merged audit firm. As noted earlier, companies choosing to measure an end to the auditor-client relationship due to a merger between audit firms may confound the study’s results.

Nevertheless, the findings also provide interesting avenues for future research. This study raises the potential concern over auditor independence when director-auditor links and long auditor tenure occur concurrently. Future studies could investigate the impact of auditor-director links, in conjunction with longer auditor tenure, on audit quality as measured by, for instance, litigation against auditor. The impact on tenure of director-auditor links and other types of relationships (eg at the audit partner level) could be investigated over longer periods. Researching the impact of changes in the number of links or participants to the links in their association with auditor changes might be of benefit to the auditor change literature. Finally, examining whether the purchase of non-audit services from the incumbent auditor is contingent on the relationship between director-auditor links and tenure might add insight to the independence debate in the context of joint provision of services.

Auditors, audit firms, regulators, professional accounting bodies and purchasers of assurance services are likely to find the results of this study useful in informing the debate over both auditor independence, when audit firms are associated repetitively with directors, and rotation of auditors.

Aldhizer III, G.R. and Lampe, J.C. (1997). "Competitive Bidding, Auditor Tenure and the Extent of Single Audit Findings." The Government Accountants Journal 46(4): 45-49.

Alexander, M., Murray, G., and Houghton, J. (1994). "Business Power in Australia: The Concentration of Company Directorship Holding Among the Top 250 Corporates". Australian Journal of Political Science 29: 40-61

Allen, M.P. (1974). "The Structure of Interorganizational Elite Cooptation: Interlocking Corporate Directorates." American Sociological Review. 39(June):393-406.

Beck, P.J., Frecka, T.J. and Solomon, I. (1988). "An Empirical Analysis of the Relationship between MAS Involvement and Auditor Tenure: Implications for Auditor Independence." Journal of Accounting Literature 7:65-84.

Butterworth, S. and Houghton, K.A. (1995). "Auditor Switching and the Pricing of Audit Services." Journal of Business Finance and Accounting (April): 323-344.

Carroll, R., Stening, B. and Stening, K. (1990). "Interlocking Directorships and the Law in Australia". Company & Securities Law Journal (8): 290-302.

Chow, C.W. and S.J. Rice (1982). "Qualified Audit Opinions and Auditor Switching." The Accounting Review LVII(2) (April): 326-335.

Constable, J and D. Woodliff (1994). "Predicting Corporate Failure Using Publicly Available Information." Australian Accounting Review, 4(1) May:13-27.

Cook, K. (1977). "Exchange and Power in Networks of Interorganizational Relations." Sociological Quarterly 18: 62-82.

Craswell, A.T. (1985). "Studies of the Information Content of Qualified Audit Reports." Journal of Business Finance and Accounting, Spring: 93-115.

Craswell, A.T. and J.R. Francis (1999). "Pricing Initial Audit Engagements: A Test of Competing Theories", The Accounting Review, 74(2): 201-216.

Craswell, A.T., J.R. Francis and S.L. Taylor (1995). "Auditor Brand Name Reputations and Industry Specializations." Journal of Accounting and Economics. 20: 297-322.

Craswell, A.T. and S.L. Taylor (1991). "The Market Structure of Auditing in Australia: The Role of Industry Specialization." Research in Accounting Regulation. 5: 55-77.

Crosby, L.A., Evans, K.R., and Cowles, D. (1990). "Relationship Quality in Services Selling: An Interpersonal Influence Perspective". Journal of Marketing 54: 68-81.

Davison, A.G., B.W. Stening and W.T. Wai. (1984). "Auditor Concentration and the Impact of Interlocking Directorates." Journal of Accounting Research 22(1): 313-317.

De Angelo, L.E. (1981)."Auditor Size and Audit Quality." Journal of Accounting and Economics 3 (December): 183-199.

DeFond, M., M. Ettredge and D.B. Smith. (1997). "An Investigation of Auditor Resignations". Research in Accounting Regulation. 11: 25-45.

DeFond, M.L., J.R. Francis and T.J. Wong. "Auditor Industry Specialization and Market Segmentation: Evidence from Hong Kong". Auditing: A Journal of Practice & Theory. 19(1) (2000): 49-66.

De Ruyter, K., and Wetzels, M. (1999). "Commitment in Auditor-Client Relationships: Antecedents and Consequences". Accounting, Organizations and Society 24:57-75.

Dooley, P.C. (1969). "The Interlocking Directorate." American Economic Review 59:314-323.

Eichenseher, J.W. and D. Shields (1983). "The Correlates of CPA-Firm Change for Publicly-Held Corporations." Auditing: A Journal of Practice and Theory Spring: 23-37.

Eisenhardt, K.M. (1989). "Agency Theory: An Assessment and Review". Academy of Management Review. 14(1): 57-74.

Fichman, M., & Levinthal, D.A. (1991). "History dependence in professional relationships: The ties that bind." In S.B. Bacharach, S. Barley & P.S. Tolbert (Eds.), "Research in the Sociology of Organizations": 119-153.

Francis, J.R. and E.R. Wilson (1988). "Auditor Changes: A Joint Test of Theories Relating to Agency Costs and Auditor Differentiation". The Accounting Review LXIII(4) October: 663-682.

Gulati, R. (1995). "Does Familiarity Breed Trust? The Implications of Repeated Ties for Contractual Choice in Alliances". Academy of Management Journal 38(1):85-112.

Haskins, M.E. and Williams, D.D. (1990). "A Contingent Model of Intra-Big Eight Auditor Changes". Auditing: A Journal of Practice and Theory 9(3) (Fall): 55-74.

Jubb, C.A. (2000). Choosing an Auditor: Corporate Governance, Interpersonal Associations and Investor Confidence, Unpublished PhD, The University of Melbourne.

Koenig, T. And R. Gogel. (1981). "Interlocking Corporate Directorships as a Social Network." American Journal of Economics and Sociology 40: 37-50.

Koreto, R.J. and F. Harding. (1996). "How to Build a Network". Journal of Accountancy. 181(5) (May): 79-82.

Krishnan, J. and Krishnan, J. (1997). "Litigation Risk and Auditor Resignations". The Accounting Review 72(4): 539-560.

Latham, C.K., Jacobs, F.A. and Roush, P.B. (1998). "Does Auditor Tenure Matter?" Research in Accounting Regulation 23: 165-177

Levinthal, D.A. and Fichman, M. (1988). "Dynamics of Interorganizational Attachments: Auditor-Client Relations." Administrative Science Quarterly 33(3): 345-369.

Menon, K. and Williams, D.D. (1994). "The Insurance Hypothesis and Market Prices". The Accounting Review. (April):327-342.

Mizruchi, M.S. (1996). "What Do Interlocks Do? An Analysis, Critique, and Assessment of Research on Interlocking Directorates." Annual Review of Sociology 22: 271-298.

Pennings, J.M., K. Lee and A. van Witteloostuijn. (1998). "Human Capital, Social Capital, and Firm Dissolution". Academy of Management Journal 41(4): 425-440.

Pfeffer, J. (1994). Competitive Advantage Through People Boston: Harvard Business School Press.

Raghunathan, B., Lewis, B.L., and Evans III, J.H. (1994). "An Empirical Investigation of Problem Audits." Research in Accounting Regulation (8): 33-58

Ritson, B., C.A. Jubb and K.A. Houghton (1997). "Specialisation in the Market for Audit Services: Its Role in Explaining Auditor/Client Realignment." Working Paper 1, The University of Melbourne.

Schoorman, F.D., M.H. Bazerman and R.S. Atkin. (1981). "Interlocking Directorates: A Strategy for Reducing Environmental Uncertainty." Academy of Management Review 6(2): 243-251.

Schwartz, K.B. and Menon, K. (1985). "Auditor Switches by Failing Firms." The Accounting Review LX(2) (April): 248-261.

Seabright, M.A., D.A. Levinthal and M. Fichman (1992). "Role of Individual Attachments in the Dissolution of Interorganizational Relationships." Academy of Management Journal. 35(1) (March): 122-160.

Shockley, R.A. and R.N. Holt (1983). "A Behavioral Investigation of Supplier Differentiation in the Market for Audit Services." Journal of Accounting Research 21(2) Spring: 545-564.

Simon, D.T. and J.R. Francis (1988). "The Effects of Auditor Change on Audit Fees: Tests of Price Cutting and Price Recovery." The Accounting Review, April: 225-269.

Simunic, D.A. (1980). "The Pricing of Audit Services: Theory and Evidence." Journal of Accounting Research 18: 161-190.

Walker, P.L., Casterella, J.R., and Moet, L.K. (1998). "An Investigation of Audit Failures in New Audit Engagements." Research in Accounting Regulation 12: 61-75.

Wallace, W.A. (1987). "The Economic Role of the Audit in Free and Regulated Markets: A Review." Research in Accounting Regulation 1: 7-34.

Williams, D.D. (1988). "The Potential Determinants of Auditor Change." Journal of Business Finance & Accounting. 15(2) Summer: 243-261.

Williamson, O.E. (1991). "Comparative Economic Organization: The Analysis of Discrete Structural Alternatives." Administrative Science Quarterly. 36: 3-37.

Zajac, E.J. (1988). "Interlocking Directorates as an Interorganizational Strategy: A Test of Critical Assumptions." Academy of Management Journal. 31(2):428-438.

**Correlation is significant at the 0.01 level (2-tailed). Refer Table 3 for variable definitions

Refer Table 3 for variable definitions