Corporate Governance and the Prediction of the Impact of AIFRS Adoption
John Goodwin ([email protected]) is an Assistant Professor in the School of Accounting and Finance, The Hong Kong Polytechnic University, Kamran Ahmed is a Professor in the Department of Accounting, La Trobe University, and Richard Heaney is a Professor in the School of Economics, Finance and Marketing, RMIT University.
The authors acknowledge the helpful comments of seminar participants at the Australian National University, La Trobe University, RMIT University and The Hong Kong Polytechnic University, three anonymous reviewers and the editor. Any errors remain the authors’ responsibility.
Abstract
This study examines whether a firm's corporate governance system, particularly with respect to the characteristics of the board of directors and senior management, affects how accurately the impact of accounting changes is reported to shareholders. We concentrate on the relation between corporate governance measures and accounting forecast errors that arise with adoption of the International Financial Reporting Standards by listed Australian firms. Evidence reveals that corporate governance mechanisms are associated with the likelihood and magnitude of managerial forecast errors.
It is widely accepted that sound financial disclosure mitigates agency problems by bridging the information asymmetry gap existing between management and shareholders. In contrast, poor financial disclosure often misleads shareholders and has adverse effects on their wealth. In recent years, researchers have found that corporate ownership structure and board composition affect financial reporting quality, proxied by management forecast precision, earnings restatements and voluntary disclosure of informational items (e.g., Bamber and Cheon, 1998; Eng and Mak, 2003; Ajinkya et al., 2005; Karamanou and Vafeas, 2005; and Beekes and Brown, 2006). Karamanou and Vafeas (2005) find that attributes of corporate boards and audit committees are positively associated with voluntary disclosure of management earnings forecasts and that these forecasts are more precise when the board of directors and audit committees are effective. A recent study of Australian listed entities (Beekes and Brown, 2006) finds that a firm's corporate governance quality is positively associated with disclosure quality, as measured by its informativeness to the securities market.
This study extends this line of research on the association between corporate governance structure and management forecast accuracy following the recent adoption of Australian equivalents to International Financial Reporting Standards in Australia (hereafter AIFRS). Prior to full implementation of AIFRS, firms were required to disclose expected significant effects in their last AGAAP annual report and the actual effects on their earnings and equity after adopting AIFRS. In implementing the switch to AIFRS, the Australian Securities and Investments Commission (ASIC) warned entities to make their disclosures as accurate as possible by adequately informing users of the impact of the transition to the new standards. Otherwise, the ASIC would consider appropriate intervention in relation to significant non-compliance with the standards (ASIC, 2005). This study seeks to assess the precision of corporate managers in forecasting material effects on earnings and equity in implementing the various AIFRS. Our expectation is that the quality of disclosure, as measured by forecast errors, is related to the quality of a firm's governance structure.
Recent accounting scandals in Australia (e.g., HIH Insurance Ltd) and abroad (e.g., Enron Corporation) have resulted in increased regulation that it is argued will improve the quality of financial reporting with a focus on the responsibilities of the board of directors and senior management. For example, the HIH Royal Commission final report (Owen, 2003) contains no fewer than sixty-one regulatory recommendations, with about 34 per cent referring to governance, financial reporting and assurance issues. Part of the Australian response to the perceived failure of a firm's system of corporate governance, particularly with respect to the board of directors and the auditors, has been the issuance in 2003 of the ‘Principles of Good Corporate Governance and Best Practice Recommendations’ (hereafter ASX Principles), by the Australian Stock Exchange's (ASX's) Corporate Governance Council.1 The ASX Principles comprise ten core principles that the ASX expects to be consistent with international best practice. One of these recommends increased independence for the board. A definition of independence is provided. As might be expected, this definition is strict. A variable measured in accordance with this definition is included in the tests of the present paper.
Using a sample of 154 firms (27 firms with no AIFRS reconciliation, 102 firms with no errors and 25 firms with errors), with annual reporting dates of December through May, we find that firms with material errors have relatively smaller blockholder shareholdings than firms with no material errors from AIFRS adoption. Also revealed is that ‘error’ firms have managing directors with a longer tenure than ‘no error’ firms. These results are evident in both univariate and multivariate tests.
INSTITUTIONAL ENVIRONMENT AND MANAGERIAL AIFRS FORECASTS
AIFRS are effective for financial years beginning on or after 1 January 2005. However, comparatives need to be AIFRS compliant2 and so the changeover date to AIFRS is 1 January 2004 for December 31 reporting-date entities. The requirement to restate comparatives was introduced ‘to facilitate understanding of the current reporting period’ (Aus36.1, AASB 1). Reconciliations from AGAAP to AIFRS comparatives are required (para. 39 of AASB 1). In the last annual report, immediately prior to the use of AIFRS, firms are required to disclose the expected significant effects of implementing AIFRS under AASB 1047. These requirements give rise to a situation where a firm's forecast of significant effects can be compared with the actual effects arising from the introduction of AIFRS. This study documents the disclosures under AASB 1047 (pre AIFRS adoption disclosures) and then, with a view to detecting what are called ‘errors’, compares these disclosed items with the reconciliations required under AASB 1 (post AIFRS adoption disclosures).
The pre-AIFRS adoption disclosures are qualitative but quantification is strongly encouraged. Paragraph 4.1(b) of AASB 1047 requires ‘a narrative explanation of the key differences in accounting policies that are expected to arise from . . . IFRSs’. Paragraph 4.1.1 of AASB 1047 refers explicitly to disclosure of those policies that have ‘significant impacts on the reported financial position and financial performance’. Post AIFRS adoption disclosures include reconciliations of AGAAP net income (annual to 31 December 2004 and half yearly to 30 June 2004) to AIFRS net income (annual and half yearly), and AGAAP equity at 1 January 2004, 30 June 2004 and 31 December 2004 to AIFRS equity at those dates.3 These two note disclosures are examined for each of the 154 firms in searching for the existence of an error. We use annual net income and equity at the most recent balance date that AGAAP is used. Specifically, for a 31 December balance date our comparisons are made with net income for the year ended 31 December 2004 and equity at 31 December 2004. It is important to note that errors relating to the prior accounting period are also retrospectively adjusted to comparatives (see paras 7 and 31–4 of AASB 1). Under paragraph 41 of AASB 1 separate disclosure of these errors is required, which permits a clean comparison with only the AIFRS effects.
There are two main types of possible error with AIFRS adoption. The first concerns undisclosed accounting effects in the AASB 1047 required note that are subsequently disclosed in the AASB 1 required reconciliation note. The other is non-compliance with the standards, such as omitting a reconciliation. This study only examines the first type of error.4 Our empirical analyses focus on the error type related to the issue of management's knowledge of its firm's accounting systems, its preparedness for a major accounting change and its willingness to communicate the effects of that change.
With respect to the second error type, since companies failed to comply with accounting standards it is not possible to determine whether the errors are material on quantitative grounds, even though such omission may be the result of poor governance. However, four of the six firms with these types of error are classified in the ‘no error’ sample in this study, which may bias results toward failure to reject the null hypothesis.5
CORPORATE GOVERNANCE AND FORECAST ACCURACY
The literature on corporate governance has focused mainly on links between firm performance and board structure (in particular board composition and board size), governance choices, and investment and financing decisions (see Farhina, 2003, and Gillan, 2006, for a detailed review). In general, empirical evidence suggests that the market reacts positively to corporate decisions when firms have an effective governance structure. While most prior studies focused on board structure, recent studies have assessed the impact of a range of internal and external corporate governance attributes, measured by an index on firm performance (e.g., Black et al., 2002). However, in recent years some researchers have shifted attention to how corporate disclosure policy is affected by a firm's governance structure. The theoretical foundation for this relation is that good governance mitigates agency problems between managers and shareholders; voluntary disclosure is a way of reducing these problems (Jensen and Meckling, 1976; Botosan, 1997). For example, Eng and Mak (2003) find that ownership structure and board composition are associated with a firm's disclosure of voluntary information. Further, the literature on voluntary disclosure and corporate governance and institutional ownership has been extended by Ajinkya et al. (2005), Karamanou and Vafeas (2005) and Beekes and Brown (2006), who examine how corporate boards, audit committees and institutional ownership affect financial disclosure quality, as proxied by voluntary disclosure of management earnings forecasts and the informativeness of disclosures. Ajinkya et al. (2005) examine the association between outside directors and institutional ownership, and the likelihood of firms making earnings forecasts. They report a positive association between these factors and the frequency and accuracy of management earnings forecasts. Karamanou and Vafeas (2005) suggest that managers, guided by effective boards, audit committees and active shareholders, have greater pressure to provide better quality information. They find that firms with more effective board and audit committee structures, proxied by a higher percentage of outside directors, more frequent meetings, a lower percentage of insider ownership and a higher percentage of members with financial expertise, are more likely to make or update an earnings forecast, the forecast is likely to be more accurate and it is more likely to receive a favourable market response. Prior to that, Baginski and Hassell (1997) find that managerial earnings forecasts are more precise in firms that face external scrutiny such as by financial analysts. Tan et al. (2002) argue that the accuracy of management forecasts captures management competence and its credibility builds on the accuracy of forecasts made in previous years (Williams, 1996). Using earnings restatements to proxy for financial reporting quality, Agarwal and Chadha (2005) report that the probability of a restatement is lower in firms whose boards or audit committees have an independent director with financial expertise.
Beekes and Brown (2006) provide empirical evidence of a positive association between corporate governance quality and the informativeness of disclosures. Their corporate governance quality measure is based on the Horwath index of Australian firms and their measures of informativeness of disclosures include frequency of disclosure, timeliness and speed of price discovery, analysts’ following and forecast accuracy. Using a sample of 250 Australian firms rated in the 2002 Horwath Corporate Governance Report, they find that better governed firms disclose more timely and price sensitive information, have a larger analyst following and have less biased and more accurate earnings forecasts.
Here it is argued that firms with effective corporate governance will have implemented mechanisms that identify more precisely the significant items that affect earnings and equity on adoption of AIFRS. Further, this forecast precision will be reflected in fewer accounting errors when the accounts are prepared in accordance with AIFRS. Our research proposition is that there is a negative relation between the quality of a firm's corporate governance system and the likelihood and magnitude of AIFRS-related accounting errors.
VARIABLES, MODEL DEVELOPMENT AND DATA
Variables
Board Independence (BI) The percentage of independent directors on the board (BI) is measured using the firms’ application of the definition of independence in the ASX Principles. The composition of the board of directors has been examined in many recent corporate governance studies. Focusing on financial reporting issues, Dechow et al. (1996), Beasley (1996), Klein (2002) and Vafeas (2005) find that independent directors are more effective monitors of managerial actions. Beasley (1996) also finds that the likelihood of financial statement fraud is inversely related to the extent of independent directors on the board. Further, Peasnell et al. (2000) provide U.K. evidence of reduced income-increasing earnings management to achieve target earnings by firms whose boards comprise a higher proportion of outside directors. Karamanou and Vafeas (2005) find that the extent of independent directors on the board has a positive impact on managerial forecast precision. In all of these studies, the variable is measured as the number of non-executives on the board divided by board size.
The ASX Principles assume that independent directors are important in monitoring managerial actions. Independent directors rely on the firm's accounting system for basic information on the performance of the firm and on the behaviour of the firm's executives and so changes to the accounting system will be of particular interest to these directors. Prior research implies that the greater the control that independent directors have over the board the greater the care taken with the introduction of changes to accounting systems and on the reporting of those changes. Thus a negative association between board independence (BI) and the errors from AIFRS adoption might be expected.
However, the definition of independence in the ASX Principles differs from that used in most prior studies, and this difference permits a competing expectation. Several firms in our sample have directors who are appointed by substantial shareholders such as nominee companies, or who are executive officers of substantial shareholders, such as the parent entity. In some cases nominee companies have appointed 50 per cent of directors, who are also not shareholders. One expects these directors to improve reporting quality because they represent blockholders. Further, directors appointed by parent entities and who do not own shares in the firm, might be expected to impose a similar reporting standard to that used in the parent entity. Thus we have no expectation about the sign for the coefficient on this variable.
Managerial Ownership (MO) From an agency perspective, higher management percentage ownership implies higher firm value, since the goals of management and other shareholders are more closely aligned (Jensen and Meckling, 1976). However, empirical studies have failed to consistently support this theory.6 For example, Morck et al. (1988), Leech and Leahy (1991) and Griffith (1999) find that at higher managerial ownership levels the relation between managerial ownership and firm value is negative. These relations may be due to managers becoming more risk averse at higher ownership levels and less willing to invest in projects with potentially high payoffs (Wright et al., 1996). Further, in situations of higher insider ownership, management is less likely to invest in effective internal systems such as audit committees (Kalbers and Fogarty, 1998) and so the implementation of new accounting processes may not be as efficiently undertaken. Bergstresser and Philippon (2006) find that the more closely a CEO's compensation is tied to the value of stock and options the more likely that discretionary accruals will be used to manipulate profits. For a sample of Singaporean firms, Luoa et al. (2006) find that higher levels of managerial ownership moderate the usefulness of voluntary disclosures on the returns-future earnings relation. The above studies suggest that monitoring is weaker at higher managerial ownership levels, and therefore we expect a positive association between the managerial ownership variable and errors from AIFRS adoption.
Tenure of the Chief Executive Officer (CEOT) CEO tenure could have an impact on the likelihood and magnitude of AIFRS errors through its ability to proxy for CEO entrenchment. Studies generally argue that over time, the CEO becomes more powerful and has a stronger reputation (see, e.g., Adams et al., 2005, for evidence, and Hermalin and Weisbach, 1998, for theoretical discussion).
The literature on CEO tenure is richer in financial economics, where entrenchment has been found to have a positive relation with CEO compensation (see, e.g., Core et al., 1999, and Ryan and Wiggins, 2004). CEO tenure also partially explains corporate finance decisions including capital structure choice (Berger et al., 1997) and R&D investment decisions (Tosi et al., 1997) with the CEO behaving in a self-interested manner in both cases.
With respect to accounting, studies have examined the relation of CEO tenure and discretionary accruals (Davidson et al., 2007), timeliness of earnings (Bushman et al., 2004) and firm performance (Adams et al., 2005). Most studies find support for the entrenchment theory. Studies examining disclosure errors and CEO tenure are scarce,7 but those examining CEO tenure and accounting restatements (recognition errors) are nascent. The results of accounting restatements studies show mixed support for the entrenchment theory. For example, Burns and Kedia (2006) find no relation between CEO tenure and the likelihood of a firm restating its earnings, a result supported by Agrawal and Chadha (2005). Beasley (1996) finds no relation between CEO tenure and fraud. Conversely, Srinivasan (2005) reports that the consequences for directors of firms with accounting restatements are less severe for directors with greater tenure, possibly because they are more entrenched (p. 314), a finding supported by Desai et al. (2006). This suggests that CEOs with longer tenure might be less careful in their input into and their review of the accounting process, perhaps because the damage to their reputation and personal wealth will be relatively lower. In line with this discussion we predict that errors from AIFRS adoption will be positively associated with CEO tenure.
Audit Committee (AC) The audit committee primarily oversees the firm's financial reporting process. It meets regularly with the firm's outside auditors and internal financial managers to review the firm's financial statements, audit process and internal accounting controls (Leung et al., 2004). Prior research suggests that the audit committee's role as arbiter between the two parties is to weigh and broker divergent views of both parties to produce ultimately a balanced, more accurate report (Antle and Nalebuff, 1991). McMullen (1996) finds that the presence of an audit committee is associated with more reliable financial reporting; in particular, fewer errors, irregularities and illegal acts. Similarly, Bedard et al. (2004) find audit committee practices are positively related with the quality of financial reporting, as measured by the extent of income increasing or income decreasing accruals. The establishment of an audit committee is also associated with increases in earnings response coefficients (Wild, 1996). This result is consistent with investors expecting improvements in accounting quality. However, other studies such as Beasley (1996) and Farber (2005) find that the existence of an audit committee and its size do not affect the likelihood of financial statement fraud. On balance, we expect the existence of an audit committee will be positively associated with higher quality financial accounting information and therefore negatively associated with the errors from AIFRS adoption.
External Blockholders (EB) There is a growing body of literature on the effect of blockholders on corporate value (e.g., see Choi, 1991; Bethel et al., 1998). These studies find that if large blocks of equity are held by a few shareholders, then agency costs such as managerial monitoring expenditures may be reduced. Additionally, a more widely dispersed ownership structure could exacerbate agency incentive conflicts between contracting constituents in firms. External blockholders (BH) are expected to exert influence over the board and lead to stronger governance, especially if they have a representative on the board of directors. Further, being on the board's audit committee gives these investors the opportunity to monitor the firm's financial reporting process (Klein, 2002). Blockholders are expected to increase their monitoring activity as their holding in the firm increases. Thus, we expect a negative relation between the external blockholders variable and the errors from AIFRS adoption.
Big 4 Auditor (B4A) Big 4 audit firms8 are expected to have greater technical skills than non-Big 4 chartered accounting firms and to be associated with higher levels of accounting quality. Given their larger client base, Big 4 auditors have more to lose than non-big 4 auditors in terms of reputation capital. Thus Big 4 auditors should have more incentive to maintain brand reputation than non-Big 4 auditors (DeAngelo, 1981). Consistent with this argument, we expect that Big 4 auditors are more familiar with the requirements of AASB 1, AASB 1047 and the other AIFRS. We expect a negative association between the errors from AIFRS adoption and whether the firm is audited by a Big 4 firm (B4A).
Firm Size (FSIZE) Large firms are likely to have higher accounting quality than small firms because they have more resources available for the development of internal systems, are probably longer lived and are expected to have more established systems and more likely to have stronger internal controls. Further, Schipper (1991) argues that the annual reports of large firms are more likely to be scrutinized by financial analysts and government agencies than those of smaller firms. To control for these possible effects we include the natural logarithm of total assets as an additional independent variable (FSIZE). A negative relation is expected between the errors from AIFRS adoption and firm size.
Number of Reconciling Items (REC) Since AIFRS are likely to affect firms differently, error incidence is expected to be positively associated with the extent of the effects of AIFRS. To control for this potentially confounding effect, we measure the average number of reconciling items for net income and equity for each firm (REC), and include this variable in our analysis. A positive association between this variable and the errors from AIFRS adoption is expected.
MODEL DEVELOPMENT
In addition to univariate tests of mean and median differences, both logit and tobit analyses are used to test whether our variables explain the incidence and magnitude of accounting errors from AIFRS adoption. We use logit in analysis of the dummy variable, Errori, which identifies whether a firm has a material AIFRS error (Errori= 1) or not (Errori= 0).
()We use tobit in analysis of the absolute value of the material AIFRS errors, Abserrori. This dependent variable consists of either zero, where there is no material error, or the material error percentage. While not attempting to explain the sign of a material AIFRS error we are concerned with the magnitude of this error and so focus on the absolute value of the material AIFRS error. By construction this variable is censored at zero, where firms do not exhibit a material AIFRS error, and so we use Tobit to ensure that we properly estimate the standard errors for this model (Greene, 2003). The model takes the form:
()The variables are defined as follows:
Errori= variable that equals 1 if the firm made a material error from AIFRS adoption and zero otherwise,
Abserrori= absolute value of the error divided by total AGAAP assets at the end of the last year that AGAAP is used,9
BIi= number of non-executive directors who are not substantial shareholders, who have not worked for the firm for the past three years, have no material contractual relationships with the firm and are free from any interest which could materially interfere with the directors’ ability to act in the best interests of the firm, deflated by the board size at the date of the financial report,
MOi= number of the issued ordinary shares held directly or indirectly by the insiders on the board deflated by the total number of issued ordinary shares at the end of the last year that AGAAP is used,
CEOTi= number of years to the date of the financial report that the CEO has held the top ranking position,
ACi= dummy variable that equals 1 if the firm has an audit committee and zero otherwise,
EBi= sum of the percentage holdings of non-insider investors that hold at least 5 per cent of the issued ordinary shares at the end of the last year that AGAAP is used,
B4Ai= dummy variable that equals 1 if the firm is audited by a Big 4 chartered accounting firm and zero otherwise,
FSIZEi= natural logarithm of the book value of total AGAAP assets at the end of the last year that AGAAP is used, and
RECi= simple average of the number of reconciling items for annual net income and equity at the most recent date that AGAAP is used.
We report the Cox–Snell R2 prediction accuracy (R2) to indicate the explanatory power of the logit (tobit) model and p-values are reported for the coefficient for each explanatory variable.
Data
The FinAnalysis database from Aspect Huntley Pty Ltd provides accounting data and scanned images of financial reports for listed and delisted firms extending back to the early 1980s for some firms. All firms with a financial year-end of December through May were selected from the FinAnalysis database, giving a total of 229 firms. From these, 48 were deleted because they used non-Australian GAAP, 10 because they were suspended from trading and their accounts were not available and 9 because they did not provide interim accounts under AGAAP. The 2005 half-yearly accounts were sought from the FinAnalysis database for the remaining 162 firms. The sample was further reduced by 6 firms, for which the accounts could not be obtained, and by 2 firms that restated under AIFRS in foreign currency, giving 154 firms. Of these 154 firms there were 27 firms that had no change to report with the introduction of AIFRS and were dropped from the main analyses, leaving a final sample of 127 firms, 102 firms with no material error and 25 firms that had a material error.
For each firm we examine the notes and record information on the significant effects of AIFRS as required under AASB 1047. Specifically we identify the reason for the key difference and classify differences into one of twenty-four categories following Jubb (2005). Some firms provided quite detailed reconciliations while others were less forthcoming. We selected a classification scheme that was sufficiently broad so that it could be applied consistently across all firms in the sample. We obtain the first half-yearly accounts prepared under AIFRS and record the items in the reconciliation note for annual net income and equity into one of the twenty-four categories. If an item appears in the reconciliation that is not disclosed in the note about the expected effects, we flag it as a potential error.
Most potential errors were easily identified because the firm used the same descriptor heading in the pre AIFRS AASB 1047 disclosures as it did in the post AIFRS AASB 1 reconciliation note. We also read the notes to reduce the potential for misclassifying an error due to changes in presentation. Some of the potential errors were eliminated in this process. For example, in its AASB 1047 disclosures, a firm may provide a heading of ‘Impairment’ and state that its non-current assets may be written down under AIFRS. If, in that firm's AASB 1 reconciliation note goodwill impairment expense is recorded under the heading ‘Goodwill’, then we do not identify that expense as a potential error. Only the heading has changed in this case. Additionally, a small number of firms restated their comparatives due to a prior period error and these adjustments are not included in the error identification process.
To operationalize the ‘significant impacts’ requirement in paragraph 4.1.1 of AASB 1047, we use a threshold level of 5 per cent before classifying an item as an error. Therefore, in our empirical tests, an error occurs when an item in a reconciliation note has an absolute value greater than the absolute value of 5 per cent of A-GAAP net income (equity) and that item was not disclosed in the AASB 1047-required note. We gathered instances of other errors (untabulated) such as breaches of the standards, but these are not part of our empirical tests.
Table 1 presents descriptive statistics for the errors and Panel A shows the percentages and firm frequencies ranked by material frequencies. The error sample used consists of the material errors shown in the two right columns. Eleven of the 25 material errors (44 per cent) are accounted for by foreign currency adjustments (7 out of 25) and restoration provisions (4 out of 25), suggesting that firms either underestimated or were unaware of the effects of these when they disclosed the expected effects of AIFRS. Panel B shows the magnitudes of the errors by reconciliation. Most (23) errors occur with net income. The total frequencies for net income and equity do not total 25. The explanation is that six firms have an error which is material for net income and for equity.
|
Panel A: Percentages and frequencies of errors ranked by frequency |
||||
|---|---|---|---|---|
| All | Material | |||
| % | N | % | N | |
| Foreign currency | 13 | 18 | 28 | 7 |
| Restoration provisions | 9 | 13 | 16 | 4 |
| Other | 15 | 20 | 8 | 2 |
| Impairment | 4 | 5 | 8 | 2 |
| Income tax | 4 | 5 | 8 | 2 |
| Investments | 6 | 8 | 8 | 2 |
| Revenue recognition | 8 | 11 | 4 | 1 |
| Employee benefits | 1 | 2 | 4 | 1 |
| Exploration and evaluation costs | 1 | 1 | 4 | 1 |
| Intangibles | 3 | 4 | 4 | 1 |
| Property, plant and equipment | 6 | 8 | 4 | 1 |
| Provisions | 3 | 4 | 4 | 1 |
| Share based payment | 4 | 5 | 0 | 0 |
| Financial Instruments | 3 | 4 | 0 | 0 |
| Goodwill | 2 | 3 | 0 | 0 |
| Business combinations | 4 | 5 | 0 | 0 |
| Defined benefit superannuation | 6 | 8 | 0 | 0 |
| Leases | 4 | 5 | 0 | 0 |
| Debt–equity classification | 1 | 2 | 0 | 0 |
| Losses attributable to minority interests | 2 | 3 | 0 | 0 |
| Insurance contract definition change | 1 | 1 | 0 | 0 |
| Total | 135 | 25 | ||
|
Panel B: Error magnitudes |
|||||||
|---|---|---|---|---|---|---|---|
| N(firms) | Net income | Equity | |||||
| N | Mean | Median | N | Mean | Median | ||
| Foreign currency | 7 | 7 | 0.24 | 0.23 | 2 | 0.17 | 0.17 |
| Restoration provisions | 4 | 3 | 0.17 | 0.19 | 2 | 0.14 | 0.14 |
| Other | 2 | 2 | 0.09 | 0.09 | 0 | ||
| Impairment | 2 | 1 | 0.37 | 0.37 | 1 | 0.17 | 0.17 |
| Income tax | 2 | 2 | 0.26 | 0.26 | 1 | 0.19 | 0.19 |
| Investments | 2 | 2 | 0.10 | 0.10 | 1 | 0.11 | 0.11 |
| Revenue recognition | 1 | 1 | 0.07 | 0.07 | 0 | ||
| Employee benefits | 1 | 1 | 0.07 | 0.07 | 0 | ||
| Exploration and evaluation | 1 | 1 | 0.12 | 0.12 | 0 | ||
| Intangibles | 1 | 1 | 0.14 | 0.14 | 0 | ||
| Property, plant and equipment | 1 | 1 | 0.11 | 0.11 | 1 | 0.06 | 0.06 |
| Provisions | 1 | 1 | 0.07 | 0.07 | 0 | ||
| Total | 25 | 23 | 0.17 | 0.12 | 8 | 0.14 | 0.14 |
| The total sample of 154 observations is split into 27 firms that do not report an AIFRS effect, 102 firms that report an AIFRS effect with no error and 25 firms that report an AIFRS effect with error. In Panel B the error magnitude is the absolute value of the amount of the error divided by, alternatively, the absolute value of AGAAP net income for the last year that AGAAP is used or AGAAP equity at the most recent date that AGAAP is used. | |||||||
RESULTS
Univariate Analysis
Table 2 shows corporate governance and other characteristics of the test variables (Panel A) and other variables (Panel B). Data are shown for three groups, one for firms with no reconciliation, a second for firms with no errors and a third for firms with errors. Firms with no reconciliation are unaffected by AIFRS and we exclude them because the error and no-error classifications are redundant. Recall that errors in this study are those accounting adjustments that affect either annual net income or equity by at least 5 per cent, and that were not previously disclosed as expected effects. The two columns of p-values shown on the far right of the table are from univariate tests of mean and median or spread differences between the ‘no error’ and ‘error’ firm samples, these are the focus of our ensuing discussion. A 10 per cent threshold is used for tests of significance in this paper. When we have a directional prediction for the relation between two variables, results from one-tailed tests are reported. Otherwise results from two-tailed tests are reported.10
| No reconcilliation | No error | Error | p-values | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| N | Mean | Median | N | Mean | Median | N | Mean | Median | t-test | Mann-Whitney | |
| Panel A: Test variables | |||||||||||
| BI (Board independence) | 27 | 33.0 | 33.3 | 102 | 50.9 | 50.0 | 25 | 49.3 | 50.0 | 0.79 | 0.92 |
| MO (Managerial ownership) | 27 | 21.8 | 15.6 | 102 | 12.8 | 2.7 | 25 | 29.4 | 13.4 | 0.02 | 0.01 |
| CEOT (CEO tenure) | 24 | 6.9 | 4.7 | 102 | 5.7 | 3.7 | 25 | 10.2 | 7.5 | 0.01 | 0.01 |
| AC (Audit committee) | 27 | 0.4 | 0 | 102 | 0.75 | 1.0 | 25 | 0.72 | 1.0 | 0.36 | 0.36 |
| EB (External blockholders) | 27 | 31.9 | 19.1 | 102 | 37.4 | 33.9 | 25 | 20.4 | 21.4 | 0.01 | 0.01 |
| B4A (Big 4 auditor) | 27 | 40.7 | 0 | 102 | 80.4 | 1.0 | 25 | 72.0 | 1.0 | 0.20 | 0.18 |
| FSIZE (Firm size) | 27 | 6.8 | 6.8 | 102 | 8.2 | 8.1 | 25 | 7.7 | 7.7 | 0.01 | 0.06 |
| REC (Reconciling items) | 102 | 3.1 | 2.5 | 25 | 3.1 | 3.0 | 0.44 | 0.22 | |||
| Panel B: Other variables | |||||||||||
| Market capitalization ($ millions) | 27 | 10.4 | 7.5 | 102 | 1,183.2 | 142.9 | 25 | 353.9 | 75.8 | 0.01 | 0.09 |
| MB ratio | 27 | 4.4 | 0.9 | 102 | 2.3 | 1.6 | 25 | 1.8 | 2.1 | 0.69 | 0.64 |
| PE ratio | 8 | 64.7 | 6.0 | 64 | 23.0 | 13.9 | 17 | 80.6 | 15.2 | 0.17 | 0.65 |
| ROE | 27 | −0.07 | 0.01 | 102 | 0.04 | 0.02 | 25 | 0.01 | 0.02 | 0.27 | 0.92 |
| ROA | 27 | −0.10 | −0.02 | 102 | 0.01 | 0.01 | 25 | 0.01 | 0.01 | 0.44 | 0.40 |
| Leverage | 27 | 2.2 | 0.2 | 102 | 0.4 | 0.4 | 25 | 0.43 | 0.3 | 0.69 | 0.95 |
- As for Table 1, the total sample of 154 observations is split into 27 firms that do not report an AIFRS effect, 102 firms that report an AIFRS effect with no error and 25 firms that report an AIFRS effect with error.
- Panel A—Test variables: BI = percentage of independent board members, MO = the percentage of the issued ordinary shares held by the insiders on the board, CEOT = number of years to the date of the financial report that the CEO has held the top ranking position, AC = dummy variable that equals 1 if the firm has an audit committee and zero otherwise, EB = sum of the percentage holdings of those investors that hold at least 5 per cent of the issued ordinary shares and are not insiders. B4A = dummy variable that equals 1 if the firm is audited by a Big 4 firm and zero otherwise, FSIZE = natural logarithm of total assets measured at the end of the financial year, REC = simple average of the number of reconciling items for annual net income and equity at the most recent date that AGAAP is used.
- Panel B—Other variables: Market capitalization is expressed in millions of dollars and is measured at financial year end, MB ratio = market to book ratio at financial year end, PE ratio = market capitalization at financial year end divided by annual net income. The sample size for the PE ratio comparison is smaller because it is only measured when net income is positive. ROE = annual net income divided by average equity, ROA = annual net income divided by average total assets, Leverage = total liabilities divided by total assets at financial year end.
- p-values are from one-tailed tests of significance for all differences except BI, MB ratio, PE ratio, ROE, ROA and Leverage where p-values are from two-tailed tests of significance.
Looking first at Panel A, the differences between the means and medians or spreads of the MO, CEOT and EB variables are significant at the 5 per cent level. There are no instances of unexpected signs. The mean of the FSIZE variable is significantly larger for the ‘no error’ sample, and the non-parametric test shows significance at the 10 per cent level. This suggests there are some very large firms in the ‘no error’ sample. The differences between all other test variables are insignificant. It is evident from the other variables in Panel B that firms which provide no reconciliation are small (mean market cap =$10.4m), compared with the ‘no-error’ and ‘error’ samples (mean market caps =$1,183.2m and $353.9m respectively). Medians show similar relations. This result is consistent with prior research (Goodwin and Ahmed, 2006). Only the market capitalization differences are significant in Panel B.11
Multivariate Analysis
Table 3 shows Pearson correlation coefficients for the dependent and the independent variables. The dependent variable called Error is positively correlated with managerial ownership (MO) and CEO tenure (CEOT) and negatively correlated with external blockholders (EB) and firm size (FSIZE). The signs on these coefficients are as expected. For the AbsError dependent variable, only the correlations with CEOT, AC and FSIZE are significant and the coefficients have the expected signs. The FSIZE variable is significant only at the 10 per cent level (one-tailed test). The two control variables (FSIZE and REC) are correlated with five of the other six independent variables. CEO tenure (CEOT) is the exception.
| AbsError | BI | MO | CEOT | AC | EB | B4A | FSIZE | REC | |
|---|---|---|---|---|---|---|---|---|---|
| Dependent variables | |||||||||
| Error | .02 | −.03 | .26** | .26** | −.03 | −.26** | −.08 | −.16** | .01 |
| AbsError | −.01 | .11 | .22** | −.16** | −.04 | −.01 | −.14* | −.09 | |
| Independent variables | |||||||||
| BI (Board independence) | −.38** | .07 | .25** | .04 | .28** | .37** | .30** | ||
| MO (Managerial ownership) | .35** | −.20** | −.44** | −.20** | −.33** | −.20** | |||
| CEOT (CEO tenure) | −.19** | −.19** | −.01 | −.08 | −.09 | ||||
| AC (Audit committee) | .17* | .19** | .53** | .44** | |||||
| EB (External blockholders) | .27** | .33** | .19** | ||||||
| B4A (Big 4 auditor) | .37** | .31** | |||||||
| FSIZE (Firm size) | .66** | ||||||||
- ** significant at the 5% level,
- * significant at the 10% level.
- Results from one-tailed tests of significance are reported for the correlations between the dependent variables and all of the independent variables except for BI. Results from two-tailed tests of significance are reported for all other correlations.
- Error = 1 if the firm made a material error from AIFRS adoption and zero otherwise, AbsError = absolute value of the error divided by total AGAAP assets at the end of the last year that AGAAP is used, BI = percentage of independent board members, MO = the percentage of the issued ordinary shares held by the insiders on the board, CEOT = number of years to the date of the financial report that the CEO has held the top ranking position, AC = dummy variable that equals 1 if the firm has an audit committee and zero otherwise, EB = sum of the percentage holdings of those investors that hold at least 5 per cent of the issued ordinary shares and are not insiders. B4A = dummy variable that equals 1 if the firm is audited by a Big 4 firm and zero otherwise, FSIZE = natural logarithm of total assets measured at the end of the financial year, REC = simple average of the number of reconciling items for annual net income and equity at the most recent date that AGAAP is used.
Table 4 presents results from estimating multivariate logit and tobit models. For the logit model the dependent variable is one if the firm makes a material AIFRS error and zero otherwise. For the tobit model the dependent variable is the absolute value of the error divided by total AGAAP assets. Where a firm makes an error in both its earnings and equity we use the smaller of the two errors, though the results are qualitatively the same if we use the larger error instead. The second column of Table 4 shows the expected sign for each of the independent variable coefficients, based on the discussion above of corporate governance and forecast accuracy.
| Independent variable | Expected sign | Logistic regression(error) | Tobit regression(absolute value of error deflated by total assets) |
|---|---|---|---|
| Constant | 3.688 (0.19) | 0.138 (0.22) | |
| BI (Board independence) | ? | 0.096 (0.94) | 0.025 (0.61) |
| MO (Managerial ownership) | + | 0.369 (0.36) | −0.009 (0.42) |
| CEOT (CEO tenure) | + | 0.080 (0.02)** | 0.003 (0.02)** |
| AC (Audit committee) | − | 0.649 (0.17) | 0.020 (0.22) |
| EB (External blockholders) | − | −2.483 (0.03)** | −0.087 (0.04)** |
| B4A (Big 4 auditor) | − | −0.084 (0.44) | 0.004 (0.42) |
| FSIZE (Firm size) | − | −0.821 (0.02)** | −0.032 (0.03)** |
| REC (Reconciling items) | + | 0.321 (0.03)** | 0.008 (0.12) |
| R 2 | 0.15 | 0.08 | |
| Prediction accuracy | 81.90 | ||
| N | 127 | 127 |
- ** significant at 5% level,
- * significant at 10% level.
- The total sample of 154 observations is split into 27 firms that do not report an AIFRS effect, 102 firms that report an AIFRS effect with no error and 25 firms that report an AIFRS effect with error. Numbers in brackets are p-values associated with the test for significance of the independent variables. p-values are from one-tailed tests of significance for all coefficients except for Constant and BI where p-values are from two-tailed tests of significance. R2 for the logistic regression is the Cox–Snell R2. In the logistic regressions the dependent variable is equal to 1 if there is an error and zero otherwise. In the tobit regression the dependent variable is equal to the absolute value of the error deflated by total assets. BI = percentage of independent board members, MO = the percentage of the issued ordinary shares held by the insiders on the board, CEOT = number of years to the date of the financial report that the CEO has held the top ranking position, AC = dummy variable that equals 1 if the firm has an audit committee and zero otherwise, EB = sum of the percentage holdings of those investors that hold at least 5 per cent of the issued ordinary shares and are not insiders. B4A = dummy variable that equals 1 if the firm is audited by a Big 4 firm and zero otherwise, FSIZE = natural logarithm of total assets measured at the end of the financial year, REC = simple average of the number of reconciling items for annual net income and equity at the most recent date that AGAAP is used.
The results for the logistic regression show that CEOT is significantly associated with Error in the predicted direction (p= 0.02). This indicates that the longer the CEO tenure, the more likely it is that a disclosure error will be made from AIFRS adoption. This result supports the ‘entrenchment’ hypothesis that less care will be taken over accounting data quality, possibly because CEOs become less sensitive to events that could potentially damage their reputation. Wai Hui (2004) also finds that CEO tenure is inversely related to disclosure quality. The percentage of external blockholders (EB) is also significant (p= 0.03) in predicting the likelihood of an error. This suggests that where there is a large group of external blockholders the firm is less likely to exhibit errors with the adoption of AIFRS. Both control variable coefficients have the expected signs and are significant. The likelihood of errors is negatively associated with firm size (p= 0.02) and positively associated with the number of reconciling items (p= 0.03). The model is significant with an explanatory power of about 15 per cent. Its prediction accuracy is about 82 per cent. With respect to the tobit regression, only the REC coefficient is not consistent with the results from the estimated logit model (p= 0.12). Although the likelihood of making an error increases with the number of reconciling items, it seems there is no relation to the magnitude of error. The model explains about 8 per cent of the variation in the absolute value of the material AIFRS error values. Thus both the incidence and magnitude of the error are a function of CEO tenure, external blockholders and firm size.
In summary, we find the longer the CEO tenure the more likely that errors are made and the more material the errors. There is also evidence that both the percentage of external blockholders and firm size are negatively associated with AIFRS errors. Apparently, the greater the level of external blockholders the less likely it is that errors are made and the less material is the error. Also, the larger the firm the lower the probability of an error, and the less material is the error. The greater the extent of the effect of AIFRS, the greater the probability that an error will be made. However, there is no relation to error magnitude.
ADDITIONAL TESTS
Given the relatively small sample size in this analysis we use bootstrapping (Johnston and DiNardo, 1997; Lunneborg, 2000) to generate 95 per cent and 90 per cent confidence intervals as an additional check on the statistical tests. The results of these tests (untabulated) are generally in line with the logit and tobit tests, as reported in Table 4. Indeed, the reported statistical tests are perhaps a little conservative, given that the 95 per cent and 90 per cent bootstrap confidence intervals identify the significant coefficients in Table 4 to be more significant.12
We also estimated the models including a dummy variable for board size, since it has been argued that once a certain size is reached, the board functions worse and is more likely to be dominated by the CEO (Jensen, 1993, p. 865). In the first case, the variable was set equal to one if size is equal to or greater than the board size median of five, and zero otherwise.13 In the second case we use Jensen's (1993) suggestion of seven for the board size cut-off. In both logistic and tobit regressions, the coefficients are qualitatively similar to those reported in Table 4.
It is possible the influence exerted by external blockholders as a group is contingent upon their number, so we included the number of external blockholders as an additional explanatory variable. The coefficient on this variable is insignificant and inferences for the other variables are unchanged.
Since the effects of AIFRS are likely to vary by industry, we included industry dummies in the regression models. We used binary variables for the following sectors: mining, financial and non-financial, the ten ASX CIGS codes, and the twenty-four ASX CIGS codes. In both logistic and tobit regressions it is found that inclusion of industry effects makes no difference to the signs and has little impact on the significance of CEO tenure, external blockholders and number of reconciling items variables. However, the coefficient for firm size becomes insignificant for all three industry specifications, suggesting that industry effects subsume the information captured by this control variable.
As an alternative proxy for the extent of the effect of AIFRS we measured the absolute value of the reconciling items scaled by AGAAP total assets and replaced the REC variable with this variable. The coefficient for this variable was insignificant. Further, while there was little impact on the other coefficients, the coefficient for the size variable is generally statistically insignificant.14
Finally, since the reconciliations are also provided in the annual accounts, we carried out the same comparison and error identification process as we did with the reconciliations provided in the half-year accounts on the initial sample of 154 firms. Fifty-four firms (about 35 per cent) changed their reconciliations of earnings or equity in the annual accounts, resulting in twenty-six error firms and 103 no-error firms. Two firms provided reconciliations in their annual accounts and gave no reconciliations in their half-year accounts, hence the difference of two between this sample and the sample for our main tests. Repeating the logistic and the tobit regressions gave the same inferences as for the tests using the data provided in the half-yearly accounts.
SUMMARY AND CONCLUSIONS
This study examines the association between corporate governance mechanisms and financial reporting quality in the context of the AIFRS adoption, using a final sample of 127 listed Australian entities drawn from 154 available firms. Several prior studies have examined governance mechanisms associated with accounting errors (see, e.g., Dechow et al., 1996). Our study uses a unique accounting provision that allows comparison of a firm's expectation about the impact of AIFRS with its actual impact. Australian GAAP for AIFRS adoption provide an opportunity to examine the issue because firms are required to disclose the expected significant effects of AIFRS adoption and provide reconciliations to AGAAP after adoption. Focusing on errors in reporting reveals the most common error is foreign exchange adjustments and that most errors occur with the net income reconciliation.
There are several implications for researchers, account users and regulators. First, CEO tenure and the percentage held by external blockholders are found to be important governance predictors of accounting disclosure quality in addition to their more commonly documented associations with recognition quality. The results are consistent with ‘CEO entrenchment’, where the longer serving CEO devotes fewer resources to the external reporting function resulting in a greater probability, and magnitude, of AIFRS errors. Yet, there are also factors that limit the incidence of AIFRS errors. The monitoring function associated with external blockholders and the greater resources devoted to the external reporting function by larger firms appear to reduce the incidence and magnitude of AIFRS errors. Future studies could examine the relation of governance variables with disclosure in other contexts, such as voluntary disclosures. Users should be aware of the possible relation between CEO tenure and disclosure quality in general. Although we only examine disclosure quality in a specific setting, the literature on this relation is growing. For example, Wai Hui (2003) also reports a negative relation between CEO tenure and disclosure quality. On a more practical note, the ASX might reconsider its strict definition of independence, as some directors expected to improve accounting quality are classified as non-independent under the ASX guidelines. This situation seems counterintuitive. Our results are not consistent with the new measurement of independence as an indicator of disclosure quality. Future research could examine this issue.
Corporate governance has provided the basis for a rich and growing body of research that explains some of the cross-sectional variation that is observed in corporate reporting. In this study, we find that firms with accounting forecast errors have longer chief executive officer tenure, lower blockholdings and, more weakly, greater managerial ownership, though inside director ownership is not statistically significant in multivariate analysis. The CEO tenure and external blockholder variables explain both the incidence and magnitude of errors after controlling for other factors. In sum, our results illuminate a new setting in which managerial financial reporting choices are associated, albeit weakly, with the quality of a firm's corporate governance mechanisms.
