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

The University of Melbourne, Victoria 3010, Australia.

Email: cajubb@unimelb.edu.au

Fax: +61 3 9349 2397

December 2001

The governance literature has only just begun to consider the role of the audit as a component governance device (Anderson et al [1993]; Matolcsy et al [1999]). This study seeks to provide evidence suggestive of an association between internal governance practices and the external audit quality. The empirical literature presents substitution theory, the insurance hypothesis, and signalling as theories that would predict an association exists between the two.

The results of the analyses indicate evidence of a relationship between various internal governance mechanisms and the level of external audit quality demanded. Some support for each of substitution, and insurance and signalling is found. Governance mechanisms that are associated with additional monitoring via the external audit quality, rather than less monitoring, are observed to share a common element (theoretically, at least)- independence from management. Additionally, supplementary analyses indicate that the association between the various forms of internal governance and audit quality is contingent on circumstances of company size, financial risk profile, and overall governance quality in a manner beyond that which can be controlled for using traditional variables to capture these constructs. As such, applying a model to predict auditor quality on randomised observations is not particularly helpful and helps explain the weak and conflicting results in prior literature.

Key Words: Corporate governance, audit quality

Governance is the form by which stakeholders monitor their company of interest. This monitoring process consists of both internal and external governance practices; internal forms including the board of directors and remuneration packages of company executives, and key external corporate governance mechanisms being the external audit and the take-over market (Jensen and Meckling [1976]). This study examines the independent audit as an external governance device.

The primary objective of the study is to examine whether an association exists between the level of audit quality demanded by a company and its managers, and the internal governance factors of the company. The relationship is analysed by way of multinomial logistic regressions, which indicate whether variables representing governance are capable of predicting audit quality choice. A secondary objective is to investigate whether associations (if any) between governance and choice of audit quality level are of the form of the substitution hypothesis proposed by Williamson [1983]. The study examines whether external audit quality is sacrificed when agency conflicts are reduced by way of internal monitoring. It proposes alternative explanations- the insurance hypothesis, and signalling - as to why the external audit quality may not be a substitutable governance tool. The current interest in governance (eg Matolcsy et al [1999], Dalton et al [1999], Stapledon [1998]) suggests that the findings of this study will be of interest to researchers, regulators, and the auditing profession alike.

The remainder of the paper is organized as follows. Section 2 presents the motivation for the study. Section 3 discusses the prior literature foundation for the study and supports development of the hypotheses. Section 4 describes the research design. Empirical results and their analysis appear in Section 5. The final section discusses conclusions of the paper, limitations and avenues for future research.

This study is motivated by the gap in the auditor choice literature in terms of governance, the gap in the governance literature in terms of external audit as a governance tool, and the implications this study potentially suggests for future academic research, practitioners, and regulators.

The Auditor Choice Literature

That auditor choice is influenced by various client risk factors has been well established in different areas of the auditing literature (eg Francis et al. [1999], DeFond [1992], Johnson and Lys [1990], Francis and Wilson [1988], Schwartz and Menon [1985]). However, less established is the association between internal governance and external audit quality.

Internal governance controls are often used in the auditor choice literature as a proxy for the extent of agency conflicts (eg Francis and Wilson [1988], DeFond [1992]). The studies include one or two governance tools, in addition to measures such as leverage, to proxy for the level of agency conflict. These studies find that, generally, as agency conflicts increase, monitoring increases via a higher standard of audit quality demanded.

This study extends the auditor choice literature by providing evidence with respect to the relationship between audit quality and a comprehensive range of internal governance devices. Many of the variables included have been before included in such analyses, but should theoretically also affect the extent of agency conflicts and the demand for additional monitoring via a higher degree of audit quality.

The UK Cadbury Report states that: "The annual audit is one of the cornerstones of corporate governance" (Cadbury Report [1992] para 5.1), yet the governance literature, excluding research on audit committees, has generally only recently begun to consider the association between the external audit quality as a governance device, and internal corporate governance mechanisms (eg Anderson et al [1993], Matolcsy et al [1999], Jubb [2000]). This study extends the literature by examining whether a relationship does exist between the two when a greater range of internal governance devices than those examined previously is taken into account. Additionally, the study is motivated to provide empirical findings to facilitate evaluation of the question of substitution of governance mechanisms with respect to the demand for audit quality and to extend the prior research of Anderson et al [1993] and Matolcsy[1999]. Anderson et al [19931 argue that the corporate governance measures of the internal audit, external audit quality (or quantity), and the board of directors, are substitutable dependent on the company characteristics of greater assets-in-place versus growth. They find that companies with greater stability employ more monitoring via audits than through directorships, and that more is spent on internal auditing than on the external audit. Their findings are supported and extended in Matolcsy et al [1999], which finds that governance from directors is greater compared to governance from the external audit, in companies with high growth options.

The prior literature surrounding audit committees in association with auditor choice indirectly tests for the substitution hypothesis, however, generally these studies show no consistent result with regard to the issue of the substitutability of audit quality (eg Menon and Williams [1994b], Pincus et al [1989], Bradbury [1990]). Bradbury [1990] indicates support for audit quality demanded being substitutable, while Menon and Williams [1994b] and Pincus et al [1989] suggest it more likely that audit quality is non-substitutable. However, Pincus et al. [1989] finds only weak significance for audit quality being a non-substitutable aspect of governance.

Implications for Future Governance Research

The governance literature has generally failed to recognise audit quality as being encapsulated within the corporate governance umbrella. Also, the literature, excluding studies of audit committees, has largely assumed the audit to be a homogenous product within the sample data. Yet, the auditing literature has long shown that there are distinguishable levels of audit quality (DeAngelo [1981], Craswell and Taylor [1991], Craswell et al. [1995]). As such, in some past studies there is potentially an omitted variable relating to the value-relevance of specific governance mechanisms to corporate performance (eg Boyd [1995], Pearce and Zahra [1992], Wild [1998]), as audit quality both directly and indirectly affects many performance-related units of measurement used in these studies.

Audit quality directly affects the share price, since, as shown by the literature, the market reaction to the release of the earnings figure is greater the higher the audit quality (eg Teoh and Wong [1993], Krishnan and Yang [1999]). Indirectly, audit quality also affects many of the measures used by past governance researchers as proxies for firm performance (eg ROA, ROE, EPS. The literature (eg DeFond and Jiambalvo [1993], Francis et al [1999]) shows that higher quality auditors are more likely to ensure that the accounting numbers are a better representation of commercial reality than those examined by lower quality auditors.

Nevertheless, before it can be suggested that governance studies need to control for audit quality, an association between the various governance devices and audit quality needs to be established. If governance, as is suggested theoretically, can explain in part the choice of auditor, and the choice of auditor affects performance measures then audit quality needs to be controlled for in future studies of the value-relevance of various governance devices. It may also provide an explanation for conflicting findings in the past empirical literature with regard to the form various governance mechanisms should take to be indicative of good governance, and, by implication, have a positive impact on performance.

Agency theory postulates that the interests of principals and agents may not be aligned and that monitoring of managers is a method of reducing agency costs (Jensen and Meckling [1976]). Monitoring is conducted via corporate governance information systems (Eisenhardt [1989]), the external audit being a monitoring device (Watts and Zimmerman [1983]). Both managers and stakeholders have incentives to encourage such monitoring (Fama and Jensen [1983a,b]).

The governance mechanisms employed are by and large internal methods of control. The audit is differentiated in that it is an external form of monitoring and, theoretically at least, independent of management influence. Also, unlike that other external form of governance, the take-over market, audit is not a governance mechanism of "last resort".

It is argued that different audit firms specialise in different levels of audit quality (DeAngelo [1981], Simunic and Stein [1987]) exhibited through their size or industry specialist designation (eg Francis and Simon [1987]; Craswell and Taylor [1991]; Craswell et al Taylor [1995]). The earnings management literature shows higher quality auditors constrain discretionary accruals (Becker et al [1998]; DeFond and Subramanyan [1998]; Francis et al [1999]) and report more auditor-client disagreements over accounting policy choice (DeFond and Jiambalvo [1993]). Studies of IPOs report less underpricing in the presence of a Big 6 auditor (Jang and Lin [1993], Beatty [1989]). The auditor change literature observes that the market reacts unfavourably to a switch from a Big 6 to a non-Big 6 auditor (Eichenseher et al [1989]; Persons [1995]), and favourably when the switch is in the opposite direction.

That differing audit quality is recognised and rewarded by the market leads us to query whether Anderson et al's [1993] findings of a substitutable relationship between audit quality and internal governance are valid. Additionally, other literature seems to indicate that audit quality could very well be a non-substitutional governance mechanism because of insurance and signalling dimensions. These dimensions are explored in the next section.

Prior literature establishes a negative association between the external audit quality demanded and specific internal governance mechanisms (Anderson et al [1993]). This study aims to observe whether such a relationship exists given a wider range of internal monitoring devices that may encourage increased external monitoring rather than moderate its demand.

The substitution effect suggests corporate governance measures are interchangeable amongst each other. This has been shown to hold for the takeover market (eg John and Senbet [1998], Walsh and Seward [1990], Morck et al [1989], Brickley and James [1987]). However, while the substitutability of internal governance devices with the takeover market has been considered (eg Williamson [1983]), the literature is just beginning to explore the substitutability of the external audit quality. Anderson et al [1993] and Matolcsy [1999]) are the first governance studies to investigate the audit as a monitoring device and its substitutability in certain types of companies.

Generally, tests indicate that when internal governance practices are poor, the external governance mechanism of the takeover market dominates. Morck et al [1989] finds that as management ownership increases, there is less likelihood of a hostile turnover and more of a 'friendly acquisition'. Brickley and James [1987]) results suggest that where there is a market for takeovers, there is less need for internal control, proxied by outsiders on the board of directors (BOD) and ownership concentration. In instances where takeovers are not as efficient a market tool, the internal governance mechanisms played a larger role in monitoring. Brickley and James [1987] and Anderson et al [1993] also find internal governance devices to be substitutable amongst one another. The former finds that either a greater number of outsider directors or less diffusion of ownership reduced managerial opportunism, while Anderson et al's [1993] results indicate that the relative expenditure on directors and internal auditors changes given different corporate circumstances.

The body of literature indicating that governance mechanisms are substitutable suggests that the substitution hypothesis may be a significant predictor of audit quality choice, whereby monitoring via high external audit quality is replaced with internal governance devices so that an audit of a lesser quality becomes acceptable.

The insurance hypothesis proposes that the auditor functions as a potential indemnifier against investment losses for investors, creditors (Menon and Williams [1994a], Schwartz and Menon [1985]) and regulators (Wallace [1987]). Literature finds support for this proposition (eg Menon and Williams [1994a]; Franz et al [1998]). The need for insurance will drive companies to demand large auditors (DeAngelo [1981], Francis and Wilson [1988]), as they are seen to be more capable of paying damages awarded or settling the case for a large sum (Schwartz and Menon [1985]). Schwartz and Menon [1985] further suggest that larger auditors have a comparative advantage in the provision of insurance, as they are able to spread the risk of litigation over a larger number of clients.

The role of the insurance hypothesis is above and beyond the monitoring function suggested by agency theory. It suggests that there may be a positive association between some corporate characteristics that are concurrently governance mechanisms relating to owners of the company (eg blockholders) or to directors, and audit quality choice, rather than the negative relationship proposed by the substitution hypothesis.

Signalling via auditor choice is based on agency theory, and is a manner by which managers and/or directors may impart to the market additional information about their company and their own behaviour. As the types of financial statements produced have become standardised, potential information differentiation that a company can use to send a signal to the market through its financial statements is reduced. Companies are thus provided an incentive to signal, other than through transparency in their notes to the accounts and other voluntary disclosures, through their choice of auditor. Moreover, even voluntary disclosures that may be used as signals achieve enhanced credibility in the presence of a quality auditor.

A high quality audit sends a signal to the market that the financial statements are more credible than those audited by lower quality auditors. The market perceives Big Six and specialist auditors to be of a higher quality than others and rewards (punishes) companies with larger improvements or falls in share prices accordingly (eg Teoh and Wong [1993], Krishnan and Yang [1999], Menon and Williams [1994a]).

Signalling theory does not actually require higher audit quality, it merely needs the market to believe that Top Tier firms are associated with higher audit quality because of the fee premiums they are able to command (Moizer [1997]). It has been shown that the market's perception of the quality of the company’s auditor influences that company's share price. As such, directors and management may want to signal to stakeholders that their interest is being well monitored.

Therefore, signalling should, theoretically, affect the demand for audit quality over and beyond the monitoring function alone. The positive signal of transparency and credibility it sends to the market and the assurance it provides to stakeholders about the quality of directors and management performance suggests a positive association between governance and audit quality choice. This needs to be compared to the negative relationship that would hold if substitution amongst the internal and external governance mechanisms operates.

Theory (substitution, insurance needs, signalling) suggesting that there is an association between audit quality choice and various governance mechanisms was presented in the preceding section. The study formulates its hypothesis based on these explanations; however, as the predictive direction of alternative explanations is inconsistent, the hypothesis is non-directional.

The predictive direction of audit quality choice made by companies under the three explanations when internal governance measures are present or absent is indicated below. It highlights that there is a negative association between the level of audit quality demanded and internal governance mechanisms under substitution theory. Insurance and signalling suggest a positive relationship between the two.

The hypothesis is tested by way of a multinomial logistic regression. The model is of the following form:

Where the variables are as follows:

Audit quality is the dependent variable for the study. It is represented via the trichotomous variable AQUAL depending on whether the auditor is a non-Big 6 firm, a Big 6/non specialist or a Big 6/specialist firm. The use of Big 6/ Non Big 6 (DeAngelo [1981]) and auditor industry specialisation (Craswell et al. [1995]) as a proxy for quality is well established in the literature.

The relationship of most of the governance variables included in this study with audit quality has not yet been tested, and especially not in the governance literature. The variables are representative of those found in that literature and are coded in a manner consistent with posited good governance controls. Table 1 documents variable definitions, coding and expected direction in association with audit quality.

Board Size BCAT is represented by a dichotomous variable indicating whether the board size of the company is in a direction suggestive of good governance. Separation of the Chair and CEO functions CHAIR as suggested by The Cadbury Code of Best Practice (1992) is also represented by a dichotomous variable. Board composition is represented by the variable OUTSIDE and is measured as a percentage of the number of outsiders over the total board size, and a larger value is interpreted as better governance. Insiders’ shareholdings INSSTK is a variable representing the percentage of shareholdings held by all insider directors. Having a company financier representative on the board is represented by a dummy variable BANKER and given a value 1 if one of the outside directors is a representative of one of the company’s bankers and 0 otherwise. The presence of an audit committee AUDCOM is measured as a dummy variable taking on the value 1 if there is an audit committee present and 0 otherwise. To capture whether the audit committee is independent from management, ACINDEP takes on a value of 1 if there are no insiders on the committee and 0 otherwise. Since substantial shareholders are more likely to monitor, blockholders BLOCK, a categorical variable, is represented by shareholders holding at least 5% of the ordinary voting shares. Also, if the top shareholder is an institutional shareholder, this suggests increased monitoring and is captured by the dummy variable TOPINST taking the value 1 if the company top shareholder is also an institutional shareholder and 0 otherwise. The percentage of holdings of the top 20 investors is captured by TOP20 and serves as an indication of the diffusion of shareholdings.

Prior literature has indicated that the riskiness of a company affects its demand for monitoring (Francis and Wilson [1988], DeFond [1992]). It is necessary therefore to control for client risk factors that may affect the association of internal monitoring devices with audit quality as the importance of monitoring changes. Distress has been commonly used as an indicator of the state of the company and has been shown to have a relationship with audit quality choice (Schwartz and Menon [1985]). This study uses to proxy for distress the five Altman (1968) ratios consisting of the ratios of working capital to total assets WCTA, retained earnings to total assets RETA, earnings before interest and tax to total assets EBITTA, market value to debt MKTDEBT and sales to total assets SALTA. Additionally debt has been shown to have a negative association with audit quality (DeFond [1992]). In this study, the inverse of a commonly used leverage measure, debt to equity, EQDEBT is used to proxy for the increased risk presented by debt.

Corporate size LNTA is included as a control variable as large companies are more likely to hire a higher quality auditor. Because larger companies are more likely to implement governance strategies but have these strategies benefit at a decreasing rate with size, it is necessary to log total assets. It is necessary to control also for the number of years listed YRLIST, as this influences governance policies and also the quality of external reporting (McMullen [1996]). There is no predicted sign as to its relationship with audit quality as both just listed and long-time listed companies may demand similar levels of audit quality. The relative importance of the client in its industry POR4 is measured by a ratio of the total assets of a company to the total assets of the largest four companies in that industry. This variable captures political cost influences and has been shown to influence choice of an industry specialist auditor (Kwon [1996]). Finally, companies may choose a higher quality (large) auditor because of NAS services it is capable of providing. This possibility is captured by NASAUD, which is the amount of total NAS purchases made by a company from any of its auditors relative to total audit fees.

INSERT TABLE 1 ABOUT HERE

For this study, 1995 is chosen as the year of interest; the year being suited to the purposes of this study as it allows time for companies to implement their corporate governance practices after the publicity of the early 1990s emphasising their importance. Data was collected from ASX Data disc, the Annual Report Collection (Connect4), Craswell [1996] Who Audits Australia, and The Australian Financial Review: Shareholder. In order to ensure a variety of governance and risk levels, the sample was split such that the data was collected from each of the 164 largest and smallest listed companies based on 1995 total assets after deleting those companies which failed the selection criteria. The final sample numbered 328 companies. The selection criterion excluded all companies in the banking and finance industry (Industry 16) as well as any company where at least one piece of data was unavailable from one of the four sources. Also, companies reporting in foreign currencies were deleted from the sample in order to reduce confusion surrounding exchange rates.

Table 2 shows the descriptive statistics for the sample selection process and the frequencies of industries of companies included in the sample. The most frequent industry included in the sample is Industry 1 (Gold), which accounts for 25.3% of the sample companies. Breakdown into large and small companies indicates that this is predominantly driven by the small company sample.

INSERT TABLE 2 ABOUT HERE

Table 3 reports the descriptive statistics for all the variables included in the study. It can be seen from Table 3 that on average 71% of the sample companies chooses a Big 6 auditor, and that 28% also choose a specialist auditor- that is, a little more than one third of the Big 6 auditors demanded are also specialist auditors (39.4%). Of the sampled companies, 73% have board sizes that are less than the median for their subset (large or small), and 86% exhibit separation of the CEO and Chairman functions, indicative of posited effective monitoring. Also, on average, insiders held 7% of a company. In terms of board composition, some companies have boards of either extreme: no sitting inside directors, or a board that consists of purely outside directors. On average, companies elect to have about half of their board being represented by outside directors. Less than half the sample companies had an audit committee, and only 16% of those were independent from management.

The control variables exhibit indications that some companies in the sample may be in situations of distress, as the mean values for EBITTA, WCTA, and RETA are negative. These descriptives affirm the decision to include these controls for distress in the analyses. Also, on average, companies have 30.14 times more equity capital than debt, and have been listed for almost 15 years.

A correlation matrix of the independent variables for the study is reported in Table 4. The table indicates high collinearity between LNTA and AUDCOM (0.77), however this is to be expected. Table 4 highlights also a strong relationship between LNTA and POR4 (0.62). That a correlation is found between these two measures is not surprising, given that they are both indicative of the size of a company and wholly measured by total assets. There is correlation of 0.50 between EQDEBT and MKTDEBT; again some measure of correlation is expected as they share the same denominator, and their numerator is but a different measurement of a similar construct. The correlation matrix indicates that the rest of the variables share correlations below 0.50.

INSERT TABLE 4 ABOUT HERE

Table 4 indicates that on a univariate basis, there seems not to be any association between the individual governance measures and two proxies of audit quality (BIG6 and SP15). The highest observable correlation between the study’s proxies for governance and audit quality is 0.33, between AUDCOM and BIG6.

The results of the multinomial logistic regressions are reported in Table 5. The regression is run twice, altering the base category each time in order to observe comparisons between all combinations of the audit quality categories. The chi-square of the model is significant at the 0.00 level, indicating that the overall model can be interpreted. The Pseudo R² of the regression is 19.93%, suggestive of an adequate model fit.

Table 5 reports the three-way comparison between companies choosing Big 6 auditors, or Specialists, and those who choose Non-Big 6 auditors. Table 5 reveals that when the choice is between a Big 6 and a Non Big 6 auditor, TOP20 is the only significant governance variable (p<0.03). However, this finding is not carried across to when the choice is between a Big 6/Specialist and a Non Big 6/ Non Specialist. The significantly positive finding indicates that companies with less diffusion of ownership are more likely to choose a Big 6 auditor as compared to one of a lower quality. The positive association between the TOP20 and audit quality is in a manner inconsistent with the substitution hypothesis, but consistent with that of insurance and signalling theory. It is likely that insurance needs is the dominant theory, as less diffusion suggests greater shareholdings by any one top shareholder.

Table 5 also indicates that, when the choice of auditor involves a Specialist, the importance of various other governance measures increases. As predicted by substitution theory, less equity holdings (INSSTK, p<0.05) by inside directors result in the company demanding more monitoring in the form of a Specialist auditor rather than choosing an auditor of lower quality– this finding is consistent with that of DeFond [1992]. AUDCOM is also weakly significant (p<0.09), and has a negative relationship with the choice of a Specialist auditor compared to a Non Big 6. This finding is contrary to that of Menon and Williams [1994b] and Pincus et al [1989] but consistent with that of Bradbury [1990]. The negative relationship further indicates support for the substitution hypothesis.

CHAIR is shown in Table 5 as being positively significant (p<0.06), indicating that companies without a CEO–Chairman are more likely to demand external monitoring in the form of a Specialist auditor rather than be satisfied with auditors of a lower quality. This is contrary to what substitution would predict. However, like TOP20, it is consistent with the rationale of signalling and the insurance hypothesis.

INSERT TABLE 5 ABOUT HERE

In addition to the governance finding, Table 5 reveals significant findings in the control variables, indicating the ability of those variables to explain audit quality choice. The same variables are significant whether the choice is between a Specialist or Big 6 firm. Table 5 indicates a negative association between EQDEBT (p<0.01) and the choice of auditor quality, as previously demonstrated (DeFond [1992]). As leverage decreases, financial risk is reduced, and consequently less monitoring via external devices is demanded. MKTDEBT is also positively associated with higher audit quality choice (p<0.01), indicating that companies with higher market-to-debt ratios are more likely to choose a high quality auditor. The relationship is not in the expected direction, and is in contrast to EQDEBT.

As predicted, the size of a company is positively related (p<0.01) to the selection of a higher quality auditor. This shows consistency with the political cost hypothesis. Also, larger companies tend to be more complex, requiring the services of auditors with technological advantages (DeAngelo [1981]). NASAUD is also found to be associated with audit quality (p<0.02), with companies with higher proportions of NAS to audit fees selecting a higher quality auditor. This finding affirms that companies choose auditors who are able to simultaneously supply them with NAS services. YRLIST is significantly negative (p<0.06), indicating that younger companies are more likely to choose higher quality over a lower quality auditor. As the study controls for distress, this finding is consistent with companies with relatively recent IPOs choosing higher quality auditors to minimise underpricing (Jang and Lin [1993], Beatty [1989]) and so potentially contributes to the IPO literature.

Table 5 highlights that there is no significant difference in the governance or financial riskiness of companies that choose to hire a Big 6 or a Specialist auditor. The only significant indicator is POR4 (p<0.04), indicating that companies relatively more important in the industry are more likely to select a Big 6/ Specialist auditor. This finding suggests that a political cost argument is important in the decision of auditor quality choice.

The model correctly explains companies choosing Non Big 6 auditors 76% of the time; and those choosing Big 6 auditors 66% of the time. However, the model is unable to correctly predict the choice of a Specialist auditor 78% of the time. This finding is not unexpected, given that the multinomial logit showed little significance for any variable (except POR4) when discriminating between Big 6 non Specialist and Specialist auditors.

These findings suggest an association exists between various internal governance devices and the level of external audit quality demanded, and therefore suggest some support for Hypothesis 1. The associations indicate support for both substitution theory and insurance and signalling as valid predictors of the level of audit quality demanded. The control variables (except MKTDEBT) demonstrate associations consistent with prior literature and/or the predicted sign suggested earlier.

It is possible that there exist associations between other of the governance variables and the audit quality demanded which conducting the multinomial logit on the overall sample may mask. The sample for the study was deliberately chosen from the extremes of size of listed companies in order to capture a full range of governance combinations. It may be however, that small companies demonstrate a different relationship between governance and audit quality as compared to large companies, since other factors such as the political cost hypothesis are likely to be in operation for large, but not small companies. Further tests designed to explore the association between governance and audit quality choice when the sample is partitioned on the basis of size, financial riskiness, and overall governance quality respectively, follow. The findings indicate that the model when applied to the full sample does indeed mask the effects of underlying relationships that exist and are strongly significant depending on the nature of the corporate characteristics.

To partition the sample on the basis of size, median total assets is used. To partition on the basis of risk and overall governance quality, factor analysis is conducted in order to obtain a score for the risk variables, and a score for the governance variables. Variables entered into the factor analysis process were the Altman’s distress score, EQDEBT, and all the variables proxying for governance. It is necessary to factor analyse the governance variables together with the risk indicators as they do not have mutually exclusive properties.

The initial supplementary test conducted is to divide the sample on the basis of size (total assets) at the median. The descriptive statistics for both size subsamples are shown in Table 6. T-tests reveal many significant differences in governance structure and control variables between the two. Larger companies are more likely than smaller companies to have a board size above the median, less instances of duality (Heidrick and Struggles [1987]), and less insider shareholdings. They are also more likely to have a banker, an audit committee, and a top institutional shareholder, and choose a Big 6 or Specialist auditor. Larger companies are also on average listed longer and are less financial risky.

The multinomial logit was performed once for each subsample of 164 companies, changing the comparison each time. The regression output for both subsamples is reported in Table 6. Total assets (logged) remains in the analysis because there is large variation in size even within each subsample, demonstrated by significant t-tests (not reported).

Table 6 reports a Pseudo R² of 18.01% and 24.01% for the small and large companies subsamples respectively. The model fit is much better for the large companies subsample, and the table reveals that many of the significant findings of the full sample are driven by the governance practices of the large companies, and that many relationships are lost when the full sample is used. This suggests that different optimal models may exist depending on the sample characteristics.

INSERT TABLE 6 ABOUT HERE

The results reported in Table 6 Panel A indicate that the reasons small companies choose a Big 6 auditor are related more to the age of the listed company (p<0.05), its demand for NAS (p<0.08), and the financial riskiness of the company (p<0.01 for MKTDEBT), rather than any lack (or presence) of governance mechanisms. These findings affirm the earlier conclusions of a link between recent IPOs and the choice of a quality auditor, as IPOs are by definition younger companies, and riskier in their early years.

Table 6 Panel B indicates however, that governance is more important to small companies when the choice is between a Specialist and a Non Big 6 firm. Table 6 Panel B reveals that small companies choosing Specialist auditors have relatively less insider equity holdings (p=0.014) and tend to have a top shareholder who is also an institutional shareholder (p=0.045). The presence of institutional shareholders shows evidence supportive of insurance and signalling, rather than substitution. This indicates that even though institutional shareholders may have management’s ear (Monks and Minow [1995]), they still seem to value additional assurance through an independent monitor. Also, Table 6 Panel B highlights that these small companies making a Specialist choice are comparatively larger in size (p=0.089), more recently listed (p=0.042) and have higher market-to-book ratios (p<0.01) than those choosing non-Big 6 firms. Small companies choosing Specialists are also relatively more leveraged than their counterparts (Table 6 Panel B, p=0.057). Panel C compares companies that choose Specialist over just Big 6 firms; small companies that choose Specialists have less insider shareholdings (p=0.027) and are relatively larger (p=0.031).

In contrast to small companies, Table 6 Panels A & B reveals that governance structure affects the demand Big 6 or Specialist auditors in large companies. Specifically, large companies that choose high quality auditors are less likely to have duality of the CEO-Chairman roles (p<0.03) and banker representation (p<0.08), and have more blockholders (p<.04). These companies tend also to buy more NAS (p<0.058) relative to audit services. However, as shown in Table 6 Panel A, company size (positively, p=0.062) and its age (negatively, p=0.035) are also explanators of auditor choice when choosing between a Big 6 or Non Big 6 auditor.

Table 6 Panel C reveals that the only structural difference between large companies demanding a Specialist as opposed to a Big 6 firm is their relative importance in the industry (p=0.058), and the presence or absence of a audit committee (p<0.05). Companies with audit committees appear to be more willing to sacrifice some form of quality, perhaps because the audit committee’s role includes monitoring the quality of the reporting process.

The classification accuracy of the model indicates the model is able to correctly predict small companies that choose a low audit quality 84% of the time compared to only 60% for large companies. In addition, prediction accuracy of the model for small companies is higher than that of the large (42% and 32% respectively) when predicting choice of a Specialist auditor. However, the model is better able to predict large companies that select Big 6 auditors (83%) than small companies (22%) doing the same.

Overall, the supplementary test by size partitioned subsamples reveals that the full sample model masks associations between iternal governance devices and the external governance device of the audit. One implication is that future governance or auditor choice studies selecting samples from both large or small companies may not be generalisable to the population.

This study finds evidence consistent with an association between internal governance and external audit quality. It finds support for more than one theory of the choice. These findings suggest theories other than substitution alone are at work between audit quality and internal governance mechanisms indicating that further research into the substitutability of internal governance devices is needed. It may be that internal governance mechanisms are substitutable only within similar entity groupings, and future research in this area will extend the literature surrounding optimal governance guidelines. For example, future studies could investigate whether the contribution of some absent governance tool is adequately compensated by other governance measures in place in a similar grouping, such that the absence is not associated with poorer corporate performance, ceteris paribus.

Future research could also consider investigating the substitutability of other monitoring devices for which data is not readily available in Australia (eg. employee pension plans, internal auditing). It has been indicated, for instance, that institutional shareholders in Australia and overseas behave differently (Monks and Minows [1995], pp. 315-316). The importance of external monitoring by way of the external audit may also change.

This study contributes to the auditor choice literature by finding evidence consistent with the notion that governance devices, can, partially at least, explain auditor choice. The relationship between governance and audit quality is such that good internal governance will never eliminate the demand for quality auditors. It may also suggest greater gains to accounting firms if production cost functions change, as good governance is suggestive of greater auditor reliance on internal controls, with cost savings not necessarily passed on to the client. This has potential implications for lowballing and auditor independence.

The economic consequences to audit firms as a result of improving client governance structures have not yet received much research attention. Additionally, the audit literature, when considering various governance devices, has often regarded them within an agency framework setting. As such, the relationship proposed is often of a substitution form. This study shows that while such a setting may be appropriate for some governance indicators, namely those found to be substitutable, this predicted relationship is not necessarily appropriate when the governance proxies used are shareholder, or outsider, related devices.

The study also suggests implications for research into the capital market effects of auditor-related events (eg auditor change). Previous studies have tended to consider the capital market impact of audit quality in isolation of governance mechanisms (eg, Teoh and Wong [1993]); this study however, suggests future studies need to either control for governance, contingent on the circumstance of the sample, or examine capital market impact in different governance circumstances.

While the findings of this study suggest avenues for future research, the limitations need to be recognised. First, the sample is taken from the extremities of listed companies by size; as such, it may be not representative. Supplementary analyses indicates governance affects differentially audit quality choice across the extremes of company size. It may also be that different governance devices are important in middle-sized firms. A random sample would not have facilitated such further tests, at least not on such a large subsample, and may have masked the findings of an association between governance and audit quality choice.

A limitation of the study is that the measures representing governance are ambiguous. It needs to be recognised that good governance is firm-specific and not all governance devices hailed as being ‘sound’ may work necessarily for individual companies, or may not work for reasons not controlled for here (eg relationships between outside directors and management). Findings of a relationship between governance and audit quality are therefore only on average, and despite analysis to indicate if there exist differences due to company size, financial risk, or overall governance quality, the findings may not be extendable to the individual company to explain auditor choice decisions.

Additionally, the variables the study captures to proxy for the quality of governance devices but indicate the existence of specific governance tools; however, as the prior literature illustrates, existence does not necessarily mean effectiveness. For instance audit committee presence might be more ‘image management’ than serving any real monitoring purpose (Menon and Williams [1994a], Bradbury [1990]). It would also have been useful to the present study to control for more of the prior variables found to be significant in predicting audit quality choice, for example, the level of accruals (Francis et al [1999], DeFond [1992]), growth (DeFond [1992]), or interlocking directorates (Jubb [2000]).

Lastly, the study demonstrates potential limitations in its research design. Factor analysis is utilised to divide the sample into high/low risk and and poor/good governance subsamples. While less arbitrary than standardized scores, measurement error and some misclassifications in partitioning on the basis of median score generated from the factor analysis will exist. Also, the study is conducted on a cross-sectional basis; it has made the assumption that governance devices work immediately upon implementation. However, it may be that the effects of the governance tool are not evident until a period of time has passed– this presents a negative bias against the findings, suggesting that possibly other governance devices are important in the different circumstances analysed. It will be useful to replicate this study on a longitudinal basis.

To summarise, the findings of this study, although constrained by limitations of the research design and the difficulty in capturing the quality of governance, are indicative of there being an association between internal governance devices and the external audit quality. The study demonstrates that different internal governance tools become differentially important given different company characteristics of size, financial risk, and overall governance quality.

This study finds a positive association for internal governance devices demonstrating an element of independence from management, thus suggestive of insurance and signalling needs. Other governance mechanisms demonstrate substitutability with the external audit quality. These conclusions suggest a role for external audit beyond that of governance alone.

The association between internal governance and audit quality suggest avenues for future research in a number of areas. It also indicates that future research needs to consider the role of governance in audit quality choice decisions, and that the governance literature needs to pay greater attention to the role of audit quality, especially in studies considering internal governance mechanisms substitutable by external audit quality. As illustrated by this study, topics for research in governance are by no means yet exhausted.