Governance regulatory changes, International Financial Reporting Standards adoption, and New Zealand audit and non-audit fees: empirical evidence

1.1. Background

Compared to its US and Australian counterparts, the initial response in New Zealand to strengthen corporate governance and audit quality was limited at best. Marshall (2002) claims that New Zealand would experience fewer problems than in the USA and elsewhere, because of New Zealand's putative less complex environment, greater board independence, lower fixation on meeting analysts’ expectations, and the use of principle-based rules. In October 2003 (amended 1 May 2004), the NZX imposed changes in its listing rules to improve the governance and audit quality of New Zealand public companies, and required compliance within a year of the company's 2003 annual meeting. The new rules required the establishment of an audit committee with majority of independent director membership (listing rule 3.6), imposed a minimum quota for independent directors (listing rule 3.3), and introduced a non-mandatory Corporate Governance Code of Best Practice (listing rules, appendix 16). In addition, the NZSC in 2004 promoted a non-mandatory set of nine corporate governance principles supporting the general thrust of the NZX rules (NZSC, 2004).

The New Zealand Accounting Standards Review Board announced on 19 December 2002 that all New Zealand listed companies should adopt NZ IFRS for periods commencing on or after 1 January 2007, with early adoption from 1 January 2005. This move was influenced by Australia's decision to adopt IFRS for periods commencing on or after 1 January 2005, which was itself influenced by the IFRS adoption in European Union countries (Bradbury and van Zijl, 2005).

From an auditor's perspective, while some NZ IFRS standards equate broadly to the existing standards, they require more detail and disclosure and, therefore, impose greater audit effort and audit risk. This would have been especially so in the years of transition and adoption (NZICA, 2007). The adoption of NZ IFRS also created more awareness of governance practices generally, and the costs and benefits of the auditing process (Ministry of Economic Development, 2004). However, although Pickens (2005) and Muriwai (2005) generally argued in favour of the status quo, they also acknowledged the need to review auditor independence and oversight issues. Diplock (2005), the NZSC, and the Big Four (NZSC, 2007) also called for increased oversight of the auditing profession.⁶

Despite the global initiatives for better governance and audit quality, few New Zealand listed companies were directly affected by SOX or similar overseas reforms; for example, few New Zealand companies dual-list on a US exchange,⁷ and New Zealand has no equivalent of section 404 of SOX (on internal control reporting). As such, initial calls for such companies’ boards and auditors to increase the quality of auditing assurance as per SOX or similar solutions should have had only a limited impact on New Zealand audit effort and fees. While company boards and auditors are free to set an appropriate level of assurance for their investors, our review of the New Zealand environment suggests a far greater preoccupation by such groups with the adoption and implementation of NZ IFRS than with SOX or CLERP 9. We would expect audit and non-audit fees over our study period of 2002 to 2007 to reflect this preoccupation.

1.2. Prior literature

Our study builds upon the literature of how regulation might affect audit and non-audit fees. Although such fee issues have been considered in contexts outside of New Zealand (e.g. Griffin and Lont (2007) and Salman and Carson (2008) document the impact of SOX on US and Australian company audit fees, respectively), much less is known about the New Zealand audit market and its possible response to SOX. Hay and Knechel (2006) and Baskerville and Hay (2006) suggest that audit fees stabilized prior to 2002 (before SOX) but do not consider more timely fee changes as we do here. We are the first of which we are aware to examine New Zealand audit fees in the post-SOX era.

Our study also relates to the literature on how auditor independence rules might change audit firms’ fees from consulting services. For instance, if the provision of non-audit services threatens auditor independence and audit quality, then non-audit fees should decline following the adoption of stricter rules. Prior empirical evidence yields mixed results in this regard. For instance, some studies suggest that the provision of non-audit services increases the economic bond between the auditor and client, which leads to either actual or perceived impairment of independence (DeAngelo, 1981; Frankel et al., 2002; Kinney et al., 2004). Others conclude that the provision of audit and non-audit services to the same client enhances audit effectiveness and efficiency (Antle et al., 1997; Mishra et al., 2005; Joe and Vandervelde, 2007). We extend this idea by considering the provision of non-audit services by auditors in New Zealand. While New Zealand auditors are not specifically restricted in the type of non-audit work they can undertake and, therefore, the impact on this type of work might be less than in the USA, we expect New Zealand auditors’ non-audit fees to decline in the post-SOX period following the stricter local independence rules, such as those in the NZX listing rules and the NZICA Code of Ethics: Independence in Assurance Engagements, effective from 1 January 2004.

However, the transition to and adoption of IFRS could also have an impact on audit and non-audit fees, and we need to incorporate this in our research design. Schadewitz and Vieru (2008) find that small and medium size companies in Finland paid higher auditors’ fees, particularly non-audit fees, in their first year of IFRS adoption. A report by the Institute of Chartered Accountants in England and Wales discloses that European Union companies ranked increased auditing costs as one of the largest IFRS-related costs (2007). That report also listed several types of support and services auditors may offer, suggesting a positive relation between permitted non-audit fees and IFRS adoption.

1.3. Research issues

Based on the foregoing, we examine the following issues. Each builds upon the theoretical notion that to the extent that mandated accounting and auditing requirements change the optimal level of audit assurance, they should also change the optimal level of fees. We first predict that audit fees should increase as a reflection of the additional audit effort and risk, and that non-audit fees should decrease as the increased assurance is, in part, achieved by stricter independence rules. Also, clients or auditors, particularly some Big Four auditors, may have voluntarily imposed restrictions on themselves regarding the supply of non-audit services. Our first research issue is therefore:

One way the impact of international regulatory change could affect New Zealand audit fees is through listing requirements. While Telecom Corporation of New Zealand is the only company in our sample directly affected by US requirements, several companies in our sample are dual listed in Australia. Therefore, fees could be impacted if boards and shareholders of dual-listed New Zealand companies impose in response to the regulations a higher level of audit effort and risk and auditors increase audit fees to cover these factors, although, ultimately, this is an empirical question. We control for listing status in our research design.

We also contend that such audit and non-audit fee changes should reflect a stronger association with local governance regulations such as the adoption of NZ IFRS and NZX rules, and a weaker association with international regulations such as SOX. SOX is not a requirement in New Zealand (other than for a few companies with a dual-listing); although such reforms could still affect local fees if New Zealand boards and auditors increase audit quality commensurate with the provisions of the overseas rules. On the other hand, regulators and many in the financial press expressed the view that the transition to and adoption of NZ IFRS should be a costly endeavour for New Zealand companies (e.g. Fisher, 2007). The move to NZ IFRS creates extra work and risk for auditors in their preparation of an audit and in the audit itself following the new standards and, therefore, could increase audit fees. While we expect, as per our first issue, non-audit fees to decline generally because of stricter auditor independence rules, such decline could also be offset by a demand by clients for auditors’ consulting services to assist in IFRS transition and adoption. Therefore, our second issue is:

The paper continues as follows. Section 2 outlines the methods and data. Section 3 provides descriptive analyses of the level and the change of audit and non-audit fees, and discusses the results from regression analysis. Section 4 examines the robustness of the results to alternative procedures, and Section 5 states the major findings and conclusions.

2.1. Data

We analyse annual audit fee, non-audit fee and financial data for companies in the OSIRIS database with fiscal year ends from 2002 to 2007. Initially, we obtain a total of 724 company-year observations. To test better how the different regulations might affect audit and non-audit fee changes over our study period, we limit our analysis to companies with at least 5 years of data. Where necessary, we manually collect missing data from companies’ annual reports. This yields a final sample of 653 company-year observations.

2.2. Method

Our basic approach uses an audit fee model with fee determinants as per the prior literature and a series of temporal indicator variables to examine if the higher or lower audit fees correspond with key governance events that may have changed the audit firm's environment.⁸ To determine the potential effect of NZ IFRS on audit fees, we include indicator variables for the year immediately prior to IFRS adoption, the year of adoption, and subsequent years. We specify a parsimonious statistical model because of the small number of New Zealand public companies.

The prior literature guides our selection of the control variables. In a review of 147 published studies, Hay et al. (2006a) report client size as the most important determinant of audit fees. That study also suggests that client complexity and audit risk explain audit fees. For example, Simunic (1980) suggests that client leverage, liquidity, and profitability are risk variables that drive audit fee increases, as highly levered, less liquid, and low return companies are more likely to incur losses. Prior studies also show that Big N firms command a fee premium for reputation and quality (DeAngelo, 1981; Simon and Taylor, 1997; Hay et al., 2006a). In addition, companies in regulated industries such as finance are easier to audit (Turpen, 1990) and we, therefore, expect lower audit fees for these companies relative to those in other industries. Finally, audit report lag – the period from the balance date to the issuance date of the audit report – is often interpreted as an indicator for audit efficiency, as a longer delay can indicate problems during the course of the audit and difficulties in resolving sensitive audit issues (Knechel and Payne, 2001). Therefore, we expect a positive association between audit report lag and audit fees. Finally, we expect a positive relation between dual-listing and audit fees, as these companies often demand increased audit services and require the auditor to increase the scope and complexity of audit procedures (Carson and Fargher, 2007).

We use log of total assets to proxy for client size (LTA), the ratio of inventory and receivables to assets to measure client complexity (INVAR), the ratio of long-term debt to assets (DA), the ratio of earnings before interest and tax to assets (ROA) and current ratio (CURRENT) to proxy for client risk, log of the number of days between the balance date and auditor signature date (LAG) to proxy for audit efficiency, a zero–one indicator variable for auditor quality (BIG4), a zero–one indicator variable for companies in the finance and investment services industry (FINANCE), and a zero–one indicator variable for dual-listed companies (DUAL).⁹ We also include natural log of non-audit fees (LNAF) as prior research shows non-audit fees may be a determinant of audit fees.

We contend that the auditors of NZ IFRS adopters will perform additional work to comply with the new standards. The audit fees should be higher in the first year of adoption, when companies need to restate their opening balance sheet and provide prior year comparative figures, and also on an ongoing basis, reflecting the more onerous accounting, measurement, and disclosure requirements of IFRS. Audit fees should also be higher in the transition year (the year prior to NZ IFRS adoption) because of preparatory work related to the move to IFRS. Therefore, we include three IFRS-related indicator variables, specifically, PREIFRS, IFRSY1 and IFRSY2&3, which equal one for companies in the year prior to IFRS adoption, the year of adoption, and the second and third year of adoption, respectively, and zero otherwise. We also control for the effect of an auditor change (AUDCHG), as previous studies document audit fee differences by new auditors (Simon and Francis, 1988; Turpen, 1990; Pong and Whittington, 1994).

Finally, the model includes year indicator variables to adjust for audit fee changes in the 2002–2007 study period not controlled for by the other determinants. To accomplish this, we set YR2003, YR2004, YR2005, YR2006 and YR2007 equal to one for data from 2003 to 2007, respectively, and zero otherwise. The 2002 effect is reflected in the intercept. These fixed effect variables capture overseas and local governance regulatory changes in general and, also, possible macro trends in IFRS. Therefore, the IFRS variable coefficients are incremental to these general effects and, hence, more conservative given that we de-trend the regression residuals. Overall, we expect positive coefficients for LTA, INVAR, DA, BIG4, FINANCE, DUAL, LAG, LNAF, the three IFRS variables, and all year indicator variables, and negative coefficients for CURRENT and ROA. Since we are unable to distinguish between dismissals and resignations, which could have a different impact on audit fees, we make no expectation about the coefficient sign for AUDCHG.

We motivate the variables in our non-audit fee model as follows. We include LTA, INVAR, DA, ROA, BIG4, FINANCE, DUAL, LAF, three IFRS variables and a series of year indicators. Houghton and Ikin (2001) argue that companies are more likely to engage an auditor viewed as a non-audit services specialist or industry leader to perform non-audit services. Therefore, we include a non-audit services industry leader variable (INDLEAD) in the non-audit fee model, and define INDLEAD equal to one if an audit firm is the largest supplier of non-audit services (based on total dollar fees) in the industry in which a company operates, and zero otherwise.¹⁰ We expect the same signs for the common variables in the non-audit fee model, a positive coefficient for INDLEAD, and negative coefficients for the year indicators.

2.3. Models

We specify the following pooled cross-sectional regression models. For convenience, the models omit the time and company subscripts. Our audit fee model is:

()

Our non-audit fee model is:

()

where

Since the pooling of observations over time for the same companies can lower the power of the statistical tests because of lack of independence, we also examine alternatives to this approach and report the results in Section 4 on robustness tests.

3.1. Descriptive statistics

Panel A of Table 1 reports descriptive statistics for audit and non-audit fees by year. Mean audit fees increase from $NZ142 373 in 2002 to $NZ234 402 in 2007, at a constant annual compound growth rate of 7.3 per cent.¹¹ Median audit fees also increase. Mean and median non-audit fees, in contrast, show an overall decrease. For example, mean non-audit fees experience a compound negative annual growth rate of 2.5 per cent over 2002–2007. Panel B reports similar statistics by audit firm. Focusing on the most recent 2007 data, the Big Four, collectively, perform more than 82 per cent of the audits and receive around 95 per cent of combined audit fees.¹² PricewaterhouseCoopers captures more than one-third of the market followed by KPMG, which comprises 27.5 per cent (30/109). However, the market share of all Big Four firms remains relatively stable over time, with a large increase for KPMG and decreases for the other firms, including the non-Big Four.

Panel B of Table 1 also shows that audit fees per company, on average, increase each year from 2002, with the exceptions that Deloitte and Ernst & Young experience a decline from 2002 and 2003. Mean KPMG audit fees almost tripled from 2002 to 2007, in particular, the large increase in 2002–2003 is mainly because the firm acquired several large clients. The trend for non-audit fees is less uniform. While KPMG achieves a significant increase in its non-audit fee income to become the largest audit firm provider of non-audit services,¹³ the non-audit fees for Ernst & Young and Deloitte decline significantly. Non-Big Four audit firms experience a slight increase in non-audit fees, but from a smaller base. The mean ratio of non-audit fees to total fees decreases for all Big Four audit firms, reflecting mostly the decline in non-audit fees, including a decline for KPMG, whose non-audit fees increase.

Because large changes in the mean Big Four audit fees and non-audit fees shown in Panel B could be due to auditor changes,¹⁴ Panel C restricts the sample to those companies audited by the same Big Four auditor over the 6 year period. These results confirm the temporal increase in audit fees for all Big Four auditors. For non-audit fees, a more tempered increase is observed for KPMG, while Ernst & Young grew its consultancy fees, but from a low base, with the loss of key clients explaining the decline reported in Panel B. Deloitte is the only Big Four firm that shows a clear downward trend in non-audit fees, and it may be that they have voluntarily restricted the non-audit services they provide.

Panel D of Table 1 presents descriptive statistics for the dependent and control variables. We take the log of audit and non-audit fees, company size (total assets), and audit lag (the number of days between balance date and auditor signature date) to reduce the effects of outliers in the regression analysis. Companies with a Big Four auditor represent 79.2 per cent of the sample, companies in the finance and investment services industry represent 9.2 per cent, 6.3 per cent of the companies experience an auditor change, and non-audit service specialists perform 34.7 per cent of the audits.

Panel E of Table 1 reports the sample distribution of the adoption of NZ IFRS. Among the 110 companies, 47 companies (42.7 per cent) adopted NZ IFRS between 2005 and 2007, with 5, 24, and 18 companies in each year, respectively.¹⁵ Therefore, there are 62 companies that should report under NZ IFRS for their 2008 fiscal year; these companies fall into the PREIFRS category in 2007.

3.2. Univariate results

Table 2 documents how companies’ audit fees differ on the basis of selected factors. Consistent with our expectations, the results show that audit fees are higher for clients that are larger (LTA), more complex (INVAR) and more risky (DA). Audit fees are also higher when clients buy more non-audit services (LNAF ), when clients adopt IFRS early (IFRS ), when a company chooses a Big Four firm as its audit provider (BIG4 ), and when a company lists on an overseas exchange (DUAL). We also observe lower audit fees for companies in the finance and investment services industry (FINANCE) and companies with auditor changes (AUDCHG), although the differences are not significant.

3.3. Regression results

Panel A of Table 3 documents the results for the audit fee model (model 1).¹⁶ Regressions 1 and 2 outline results for the combined sample and those audited by a Big Four auditor, respectively. We also summarize the audit fee regressions for larger and smaller companies (partitioned on median total revenue) in regressions 3 and 4, respectively.

The combined sample estimation (regression 1 in Panel A), explains 74.1 per cent of the variation in audit fees. As expected, the coefficients for LTA, INVAR and LAG are significantly positive; that is, audit fees increase as companies become bigger and more complex. The significant and negative coefficient for ROA indicates that audit fees are also higher for riskier clients. We observe that the coefficient for BIG4 is insignificant and, therefore, our results do not provide evidence of a Big Four premium in New Zealand, consistent with Hay et al. (2006b).¹⁷ Regression 1 also shows that clients with an auditor change (AUDCHG) and dual listing (DUAL) experience an increase in audit fees, and there is a positive and significant relation between LAF and LNAF.¹⁸ The variables CURRENT, DA and FINANCE are generally not significant in our audit fee model.¹⁹

We also find significant audit fee increases that coincide with the transition to and adoption of NZ IFRS. Our audit fee model reflects these as positive and significant coefficients for PREIFRS, IFRSY1 and IFRSY2&3. For example, the coefficient for PREIFRS (0.307) indicates that the mean LAF in the year of transition is significantly higher than the mean LAF in the years prior to the transition year (the base group, denoted NONIFRS in the following discussion). Likewise, the coefficient estimates for IFRSY1 and IFRSY2&3 indicate that the mean LAF for the first year of IFRS adoption and the second/third year of adoption are 0.314 and 0.706 higher, respectively, than the mean LAF for the NONIFRS years.

The estimated marginal means reported at the bottom of Panel A show that there is no significant fee increase from the year of transition to the first year of adoption (the means of LAF for PREIFRS and IFRSY1 are 11.396 and 11.403, respectively). However, the mean LAF is significantly higher for the second and third years of IFRS adoption (mean LAF for IFRSY2&3 is 11.795) compared to the LAF in the first year of adoption (IFRSY1). Overall, these results support our expectation that companies began their transition to NZ IFRS in the year prior to the adoption, as reflected in a higher audit fee in the transitional period. Audit fees are also higher in the later years of IFRS adoption than the first year of adoption, consistent with an ongoing cost of IFRS. However, such higher audit costs may moderate in the future to the extent that they reflect one time start-up or learning costs.

Our results in regression 1 of Panel A also indicate that audit fees increased generally between 2004 and 2006, given the positive and significant coefficients for YR2004, YR2005 and YR2006 and a negative (but insignificant) coefficient for YR2007. The insignificant coefficient for YR2003 suggests that audit fees did not increase reliably through 2003. In other words, the early effects of SOX do not appear to associate with a similar impact on New Zealand companies’ audit fees. However, the significant fee increase in 2004 could reflect a lagged association with SOX and related overseas governance changes in 2003 and 2004 or an anticipatory effect of the New Zealand governance reforms.

The results for companies with Big Four auditors (regression 2) and smaller New Zealand companies (regression 4) are qualitatively equivalent to those for the combined sample. None of the coefficients for the year indicators and the IFRS variables is significant for the non-Big Four sample (not reported). Our results also imply that the audit fee increases in the later years for larger firms (regression 3) associate mostly with the transition to and implementation of NZ IFRS rather than more general factors, since none of the coefficients for YR2005, YR2006 and YR2007 is significant after we control for IFRS.

Panel B of Table 3 summarizes the results of estimating the non-audit fee model (model 2). We use 513 company-year observations for this analysis, excluding companies with zero non-audit fees. The adjusted R²s are approximately 64 per cent for the combined and Big Four samples. The adjusted R²s are smaller for the non-Big Four (not reported) and the larger and smaller company samples. Our results for regression 1 show that the model's explanatory power is driven mostly by LTA, INVAR, BIG4, INDLEAD, LAF, and the time trend in non-audit fees, particularly from 2005 to 2007.

We apply the same method as used in the audit fee model to analyse the association between NZ IFRS adoption and non-audit fees. First, the insignificant coefficients for PREIFRS, IFRSY1 and IFRSY2&3 in regression 1 indicate that non-audit fees did not change reliably in relation to IFRS transition or adoption, although the coefficients are positive, consistent with the IFRS results for audit fees. To the contrary, non-audit fees decreased generally over the study period. Indeed, all the year indicator coefficients are negative after we control for the other determinants and the IFRS variables; and significantly so for YR2005, YR2006 and YR2007. We conjecture that these non-audit fee decreases associate with the stricter auditor independence requirements of NZX, NZSC guidance, and the NZICA code of ethics. The results for Big Four and larger companies (regressions 2 and 3, respectively) show similar findings. On the other hand, besides the significantly negative coefficients for YR2005 and YR2006, we observe insignificant coefficients for the smaller company sample.²⁰

In short, we find no evidence of significant non-audit fee increases in the year prior to the IFRS adoption, year of the adoption, and subsequent years. While this is inconsistent with our expectation that auditors may have provided consulting services to audit clients regarding the move to IFRS, it is, on the other hand, broadly supportive of the contention that the stricter rules for auditor independence, as reflected in changes in NZX rules, NZSC guidance, and the NZICA code of ethics, may have impacted non-audit fees. Some New Zealand Big Four auditors, also, may have been required to follow the rules and guidance from their international affiliate or partner and, hence, provided fewer non-audit services to their audit clients due to those factors as well.²¹ The results for the year indicators show that New Zealand auditors (especially the Big Four) reduced their provision of non-audit services to their audit clients, as the YR2005, YR2006 and YR2007 coefficients in Panel B are significantly negative. We also find insignificant coefficients for the YR2003 and YR2004 variables, which suggest that SOX and similar legislation had little impact on New Zealand non-audit fees.²²

Finally, we use the coefficients on the year indicator and IFRS variables to approximate an upper bound on the amount of audit fee increase over our study period in general, and the fee increase associated with IFRS adoption in particular. Based on the coefficients in regression 1 of Panel A of Table 3, we estimate that audit fees increased 21.8 per cent from 2002 to 2006 (about $NZ31 000 given an average audit fee of $NZ142 373 in 2002). After controlling for other determinants of audit fees, including year effects generally, we calculate that companies that adopted NZ IFRS early paid approximately 36 per cent higher audit fees in the pre-IFRS period (the transition year) and the first year of IFRS adoption, and an additional 48 per cent higher audit fees in their second or third year of IFRS adoption. Similarly, we also estimate that New Zealand companies’ non-audit services dropped by 54.8 per cent ($NZ80 200 per company) between 2002 and 2007, in which a 44 per cent decrease from 2005 to 2007.

4.1. Event year analysis of IFRS

Rather than estimate the effects of NZ IFRS in calendar time, as we report in the previous section, an alternative approach estimates the impact of NZ IFRS in event time and, therefore, exploits the fact that the initial year of adoption for the companies in our sample could be 2005, 2006 or 2007. Such an approach does not rely on calendar estimates and may reduce calendar year dependence by combining the observations of different years. We define event year 0 as the adoption year. Event year –1 is therefore the year prior to adoption, or the year of transition.

We first estimate the residuals from the audit fee model (model 1) each year without PREIFRS, IFRSY1, IFRSY2&3 and YR200X. The companies are divided according to which event year they belong to relative to IFRS adoption date. We then estimate the mean and median audit fee residual in adoption event years –2 to 2 and compare these data to the mean and median audit fee residual for companies that have yet to adopt IFRS, although by the mechanics of regression estimation such mean of the residuals for non-adopting companies should be near zero. We perform the individual year t-tests relative to a mean of zero. The residuals for ‘other years’ (all observations other than the IFRS observations) are also tested relative to a mean of zero.

Table 4 presents the results. The combined sample shows positive mean residual audit fees (significant in adoption years –1, 1, and 2), and the trend of mean residual fees generally increases monotonically from event years –2 to 2 (although the results for event year 2 derive from five observations only). Therefore, consistent with our results in Panel A of Table 3, audit fees in the second year of NZ IFRS adoption appear to be higher than those in the first year. We show similar results for larger companies and companies with a Big Four auditor. The mean residuals for Big Four companies are significantly greater than zero for each of event years –1, 0, and 1. In other words, clients of the Big Four appear to pay higher audit fees in the year of transition and the first 2 years following the adoption of NZ IFRS. These results further hold for median residual fees, so the results are unlikely caused by a few extreme observations. Unreported results also show that the standard deviations of the yearly residuals are essentially constant. In summary, these results are consistent with our earlier evidence of an increase in New Zealand company audit fees coincident with the adoption of NZ IFRS.

4.2. Individual year analysis

The analysis in Table 3 uses a pooled cross-sectional sample to test the impact of NZ IFRS on audit fees. As there is a potential problem of lack of independence of observations for a constant sample, we examine the robustness of our results undertaking individual year audit fee estimations from 2004 to 2007. We omit the results for the individual year analysis in 2002 and 2003 since no company adopted IFRS until 2005. Our results, presented in Table 5, show that PREIFRS, IFRSY1 and IFRSY2&3 are uniformly positive and significant for all 4 years, with exceptions of the coefficients for PREIFRS in 2006 and IFRSY1 in 2007.

Specifically, companies that adopted NZ IFRS in 2005 show statistically significant audit fee increases in 2004 and 2005 and marginal audit fee increases in 2006 and 2007. This is reflected in the coefficients for PREIFRS in 2004, IFRSY1 in 2005, and IFRSY2&3 in 2006. Similarly, companies that adopted IFRS in 2006 experience higher audit fees for the year prior to IFRS adoption, the year of adoption, and subsequent years. In particular, the fee increases are more significant in the year prior to the adoption and the subsequent years than the year of adoption. Finally, we observe that companies adopting in 2007 appear to have significantly higher audit fees in the year prior to adoption or upon adoption. We suspect this is because auditors reduce audit costs as they become more familiar with the IFRS implementation. Overall, the results for the individual year analysis are qualitatively equivalent to those for the pooled regressions.