Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
AUDIT OPINION MODIFICATION AND ITS REGULATION: EMPIRICAL EVIDENCE 1 Shuang Li and Xi Wu 2 ABSTRACT This study observes the game between the regulator and the regulated surrounding audit opinion modification. Our first finding is that the earnings management tendency of the listed companies has significant negative impact upon the auditor’s reporting behavior. This confirms the necessity for the securities regulators to pay close attention to the audit opinion modifications and to implement relevant policy (Document No. 14 promulgated by the CSRC by the end of 2001). Our second finding is that Document No. 14 does not have any significant containment effect upon the selection by auditors between the unqualified opinion with explanatory paragraphs and the qualified opinion. In the meantime, the top 10 accounting firms in terms of the number of listed clients have shown a stronger tendency to modify audit opinion in the year 2001. We argue that the unfavorable environment has a significant negative impact upon the auditor’s reporting, which has prevented the regulatory policy from giving play to its expected utility. The quality of accounting information needs to be improved through overall measures. Keywords: audit opinion modification, avoidability, earnings management, regulation
I. LITERATURE REVIEW AND RESEARCH MOTIVATION In regulating the accounting and auditing behaviors of the securities market, the regulators tend to focus on the audit opinions in the annual reports of the listed companies. Previous literature (e.g. Chen, Chen and Su, 2001; Zhang and Liu, 2002) studied the relation between the earnings management of the listed companies and the auditor’s issuing non-standard reports, and found a significant positive correlation between the two. These studies believed that Chinese auditors could play a role in the assessment of accounting information as they can report the earnings 1
The authors would like to thank Dr. Gongmeng Chen, the executive editor of China Accounting and Finance Review and two anonymous referees for their valuable comments and suggestions for revisions, and Prof. Xijia Su and Prof. Charles Jieping Chen of City University of Hong Kong, and Dr. Shuhua Li of the Chief accountant’s office under the China Securities Regulatory Commission (CSRC) for providing us with valuable materials. We are, however, responsible for all the errors in this article. The opinions expressed in this article are those of the authors and do not represent any view or policy of the organization we are working for. 2 Shuang Li, Professor of Accounting and Ph.D supervisor at the Central University of Finance and Economy, and advisor for the Chinese Institute of the Certified Public Accountants (CICPA), Beijing (100081); Xi Wu, Ph.D. candidate of accounting at the Central University of Finance and Economy, the Professional Standard Division of the CICPA, Beijing (100081). 1
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
management behavior by the listed companies. Chen, Chen and Su (2001) and Zhang and Liu (2002) emphasized the positive effect of the audit opinions and provided empirical evidence that auditors disapprove the earnings management by the clients. Starting from different perspectives, the academia and the regulatory agencies are more concerned about the audit opinion modification: whether the difference between different auditors in the audit reports (i.e. in the type and expression of the audit opinions) after the discovery of earnings management is well grounded, as it directly influences the quality of accounting information disclosed to the public. Sun and Wang (2000a) made a descriptive analysis over the explanatory paragraphs of the audit opinions, suggesting that Chinese auditors tended to modify the nature of audit opinions in recent years by using the explanatory paragraphs, in particular to change the qualified opinion to unqualified opinion with explanatory paragraphs. They argued that such modification partly resulted from the pressure and demand of the listed companies, and partly from some auditors’ belief that the explanatory paragraphs can provide them an important means to change the nature of audit opinion without any legal obligation. Another study (Wang and Chen, 2001) revealed that auditors respond differently, after Chinese public accounting profession’s structural reform, towards companies with minor profits and companies on the threshold for refinancing in the securities market. Their statistical results showed that, for those companies that use earnings management to realize minor profits in order to avoid PT (particular transfer) or ST (special treatment), the non-standard audit opinions prior to and after the structural reform show a significant positive correlation with the occurrence of minor profits, which represents that companies with minor profits (with ROE falling between [0, 1%)) are more likely to be presented with non-standard opinions. For companies using earnings management to be qualified for refinancing (with ROE falling between [10%, 11%)), the non-standard audit opinions do not show a significant statistical correlation with the occurrence of an ROE marginally above the threshold prior to the structural reform. However, a significant negative correlation is observed after the structural reform, which represents that auditors are less likely to issue non-standard opinions. The CSRC has already noticed the discrepancies between different auditors in their determination of the nature and degree 3 of the quality of the annual reports by the listed companies. On December 22, 2001, it issued Document No. 14: “Rule on Information Disclosure for Listed Companies No. 14 -- Treatment of Non-standard Audit Opinions and Relevant Issues”. Article 5 of the Document stipulates that auditors shall not replace qualified opinion with 3 Such as Document No. 40 [1998] by the CSRC - "On Circulating the Speech by Comrade Fan Fuchun at the Conference of Directors of Accounting Firms Certified to Practice in Securities and Futures Businesses" (August 14, 1998) with the attachment of “Speech by Comrade Fan Fuchun - Recognize the Situation, Speed up the Reform and Further Improve the Professional Standards of auditors”.
2
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
unqualified opinion with explanatory paragraphs, or adverse opinion with qualified opinion. This shows that the regulators are paying close attention to and want to contain the modification of audit opinions. Article 7 specifies the economic consequences of non-standard opinion, which says: “If a listed company refuses to adjust items that obviously violate the accounting principles and systems or relevant information disclosure requirements, or auditors insist that items after adjustments still obviously violate the accounting principles and systems or relevant information disclosure requirements and thus present non-standard audit opinions, the stock exchange must suspend the trading of that listed company after its disclosure of regular report and requires it to correct in due course.” The following Article 8 and 10 further specifies the consequences of the potential regulated companies. Article 8 says: “If the trading of a listed company is suspended due to causes specified in Article 7, the CSRC shall, during the suspension period, investigate and give proper treatment according to law, while the listed company shall continue to carry out its legal obligations in information disclosure.” Article 10 stipulates that “If the qualified opinion or the adverse opinion involves items with an influence on the profit of a listed company and that influence has been estimated by auditors, the listed company shall eliminate the influence estimated in the audit opinion from its profit distribution scheme. If auditors present a report with a disclaimer of opinion, the listed company shall not carry out any profit distribution scheme in that year.” The concern and regulation over the audit opinion modification by the academia and the regulatory agencies draw us to the fact that the previous studies (e.g. Chen, Chen and Su, 2001; Zhang and Liu, 2002) all adopted the binary classification between auditor’s standard report and non-standard report, while in reality, among the different types of non-standard reports, the unqualified opinion with explanatory paragraphs is still an unqualified opinion in nature, which is fundamentally different from qualified opinion, adverse opinion or disclaimer of opinion. Ignoring the different types of unqualified opinions may exaggerate the positive effect of the audit opinion while neglecting its negative influence. To provide the regulators with necessary empirical evidence and reasonable policy suggestions, we distinguish between the different types of unqualified opinions and study how auditors select among these types and what the determinants are. In order to prove the negative effect of some audit opinions, this paper tries to provide empirical evidence that some auditors take an ambiguous attitude towards the earnings management behavior of their clients, and to perform a preliminary test on the necessity and efficacy of the related regulatory policy. Part II to Part IV of the article seek to use the empirical evidence from 1998 to 2000 to confirm the existence of audit opinion modification and to support the necessity of the related regulatory policy. Part II clarifies the research scope. Part III describes the sample selection, identifies some
3
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
influential factors and defines variables. Part IV presents and analyzes the statistical results. By adding the empirical data of the audit reports of 2001, part V makes a preliminary assessment of the related regulatory policy. Part VI presents the conclusion, a further discussion, and the limitations of this study.
II. RESEARCH SCOPE From the perspective of whether the disagreement between the listed companies and auditors can be avoidable, the audit opinions can be divided between the avoidable opinion and the unavoidable 4 . Melumad and Ziv (1997) believed that the discrepancy on avoidability results from the fact that the management always has the opportunity to react to the non-standard opinion and adjust in such a way that no unfavorable opinion will be disclosed. Specifically speaking, when the management disagrees with auditors, it can choose to adjust or refuse to adjust. Our research extends the avoidability hypothesis of Melumad and Ziv (1997), as the management is more likely to negotiate with auditors, rather than merely to adjust or refuse to adjust, and reach collusion so that more favorable one will replace the unfavorable opinion, or the explicit expression will be replaced by ambiguous wording. This is what we mean by “audit opinion modification”. The avoidability hypothesis, however, does provide us with a fundamental theoretical framework, and guide us, in examining the audit reports, to first analyze the avoidability of the audit opinion and then to investigate the different behaviors of the auditor towards companies with different types of avoidability of audit opinions. For the unavoidable opinion mostly on going concern uncertainties, the regulatory agencies have noticed the problem of “imprecise measurement” (please refer to Chief Accountant’s Office, CSRC, 2001, p 41-43, 421-427). Li and Wu (2002b) observed the audit opinions towards going concern uncertainties and provided empirical evidence to confirm that such unavoidable opinions have been modified greatly in the expression. Namely, with no substantial difference in financial risks and related information disclosure of the listed companies, quite a lot of audit opinions express the going concern uncertainties only in the explanatory paragraphs or in a less unfavorable way. As Li and Wu (2002b) have already made empirical tests on the modification of unavoidable opinions, this article will confine its study to avoidable audit opinions, i.e. audit opinions over 4
Melumad N.D. and Amir Ziv, who divided the non-standard opinion between the ex post avoidable and the ex post unavoidable, put the avoidability hypothesis forward. Avoidable opinion results because the audited client disagrees with the auditor on certain items, which can be avoided when the listed company adjusts its report according to the opinion of the auditor. Such opinion is often given to items like asset disposal, timing of revenue recognition, adequacy of loss provisions and other violations of generally accepted accounting principles. Unavoidable opinion refers to the situation when the audited client cannot prevent the auditor from issuing non-standard report no matter whether to make adjustments. Such opinion is usually given to items like going concern uncertainties, debt default or other uncertainties. For a detailed analysis, see Melumad and Ziv (1997) or Li and Wu (2001). 4
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
clients with a strong motivation for earnings management (e.g. to avoid losses or maintain the qualification of rights issue). Such an audit opinion is often alleviated through “negotiation” between the audited clients and the auditors, and consequently, the utility of the audited accounting information is undermined. In recent years, items in violation of national accounting standards and systems in the securities market were mainly revealed in qualified opinions, rather than in adverse opinions. 5 This, in a way, confirms the concern of the CSRC in Document No. 14 over the behavior that auditors tend to substitute qualified opinion for adverse opinion, because this may well mean that auditors with securities practicing license would sacrifice their professional standards when they faced severe auditing problems. On the other hand, however, due to the small number of observations available for adverse opinions, we are hardly able to make an empirical test on the behavior of substituting the qualified opinion for adverse opinion (at least at the present stage). In the meantime, we notice that every year, among the non-standard opinions, quite many are unqualified opinions with explanatory paragraphs on the recognition and measurement of profits and losses. In theory, such audit opinion shall transmit to the information user the following message: the financial statements of the listed companies do not violate the accounting standards and systems in all significant aspects and the audit opinion is unqualified in nature. In practice, however, what the information users see is possibly the modified form of potential qualified opinion or even adverse opinion. As such, the article confines the test to the situation when qualified opinion is modified to unqualified opinion with explanatory paragraphs, and ignores the situation when adverse opinion is modified to qualified opinion or unqualified opinion with explanatory paragraphs.6
III. RESEARCH METHODOLOGY 1. Sample Selection All the samples are selected over consecutive three fiscal years from 1998 to 2000. The reasons behind such a choice are: first, a similar study by Sun and Wang (2000a) selected samples from 1995 to 1997; and second, the structural reform in Chinese auditing market was first completed in 1998 in accounting firms with securities practicing license, which has a significant and positive influence upon the audit quality in Chinese securities market (e.g. Wang and Chen, 2001; Li and Wu, 2002b). Therefore, our choice can be seen as a continuation of the previous studies and a further study of the practices of auditors in a new, stable environment.
5
The first adverse opinion in Chinese securities market appeared in 1997, and in 1998, 1999 and 2000 only one adverse audit opinion appeared each year. 6 From another perspective, the modification from qualified opinion (or even adverse opinion) to unqualified opinion with explanatory notes significantly changes the nature of the audit opinion, while the modification from adverse opinion to qualified opinion can be regarded as a change of degree within the same nature. Therefore, we believe that the former modification is more significant than the latter in research. 5
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
As our focus is on the modification of qualified opinion to unqualified opinion with explanatory paragraphs, we merely select non-standard audit reports with qualified opinions or unqualified opinions with explanatory paragraphs. The criterion for the selection is whether the corresponding listed companies have a positive book profit in the reporting period. Our study targets the typical or relatively “successful” earnings management behavior (i.e. with a net profit in the reporting period). In terms of the audit reports, we try to divide between grave modifications with the audited net profit more than zero and given an unqualified opinions (though with explanatory paragraphs) and no (or light) modifications with the audited net profit more than zero but given a qualified opinion. As a result, we select the audit opinions of companies with a positive book profit in the reporting period. Based upon the above criterion and procedure, we have examined the data by the Chief Accountant’s Office of the CSRC (2001a, b, c) and obtained non-standard audit reports with qualified opinions or unqualified opinions with explanatory paragraphs from 1998 to 2000, selecting those observations with a net profit as samples. The selection process and results are described in Table 1, which shows a total of 337 samples (company/ year) from 1998 to 2000, 251 of which are unqualified opinions with explanatory paragraphs and 86 are qualified opinions. This part is to test the factors that influence the selection between unqualified opinions with explanatory paragraphs and qualified opinions. The presentation of unqualified opinions with explanatory paragraphs is seen as a modification of audit opinions, while the presentation of qualified opinions is seen as no modification.
Table 1 Process and Results of Sample Selection 1998
1999
2000
Total
94
114
106
314
Panel A：Unqualified opinions with explanatory paragraphs Total Less：Number of audited clients with a net loss The resulting number of samples
20
21
22
63
74
93
84
251
38
58
57
153
15
25
27
67
Panel B：qualified opinions Total Less：Number of audited clients with a net loss The resulting number of samples Total number of samples
23
33
30
86
97
126
114
337
Note: The data of the audit opinions are from the Chief Accountant’s Office of the CSRC (2001a, b, c), and the figures of the net profits in the current period are manually collected from the annual reports of the listed companies.
2. Influential Factors and Definition of Variables Auditors are under the influence of many factors when they express their audit opinions. These factors can be roughly divided into two types: the first type is related to the characteristics of the 6
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
audited clients, and the second type is related to the characteristics of auditors. When it comes to the audit opinion modification by auditors, we will focus on the motives for earnings management and corresponding financial features of the audited clients in the first type, and the selection of auditors in the second type, such as the size and reputation of accounting firms, or recent change of auditors or audit tenure. (1) Characteristics Related to the Listed Companies (i)
Non-major business profit as a percentage of the total pre-tax profit
This indicator reflects the degree of profit manipulation of a listed company by means of such non-recurrent items as other business profit 7 , investment returns, and non-operating gains. The more the indicator is, the more obvious the earnings management by a listed company. However, the correlation between the variable and the audit opinion modification is uncertain. On one hand, the more obvious the earnings management is, the stronger the motives for the earnings management. Consequently, the company puts more pressure on auditors who are more likely to modify the audit opinions. On the other hand, when other things are held constant, the more obvious the earnings management is, the more likely a company is noticed or punished by the regulators, and therefore the more serious consequences auditors will face for their modification of the audit opinions. This has de facto contained the modification behaviors. Definition of variable: IRRGLPRF= (pretax profit of the current period – operating profit + other business profits)/ pretax profit of the current period 8 . (ii) Return on equity (ROE) 7
Other business profits do not refer to under-the-line items in a common sense. During the data collection process, however, we did find that a large number of sample companies recognized significant other business profits (as a percentage of the total profits). We also examined the notes to the accounting statements and discovered that many cases were resulted from related party transaction, such as trustee fees, revenues from occupation of funds and subsidies from related parties. According to Q&A on Information Disclosure by Listed Companies: Document 1--Non-Recurrent Profits and Losses issued by the CSRC on April 25, 2001, these profit items should be classified as non-recurrent profit items. Therefore, we take into account other business profits. Despite the possibility that the other business profits of some sample companies do not belong to non-recurrent items, it is difficult to identify these companies. In particular, when some listed companies disclose ambiguous or even wrong information, elimination of other business profits from the non-operating profits may result in errors. According to statistics (with 337 samples), if other business profits are not considered, the value of IRRGLPRF will drop about 59 percent on average (the T statistics of mean variance before and after the elimination of other business profits is 5.856, and the significance level is 0.000). In addition, if the percentage of other business profits is not significant, its inclusion will save the cost of data processing while without producing a significant impact upon the related variables. If we replace the original variable IRRGLPRF with a new variable IRRGLPRF without taking into account other business profits and make the regression analyses, we cannot find any significant impact upon the overall statistical effects of the research model or upon the correlation with other variables. However, the correlation between variable IRRGLPRF and dependent variables is statistically significant when the other business profits are excluded. The significance level goes up from 0.061 to 0.040. Such a result shows that we must be aware of the possibility that some listed companies might use other business profits to manipulate their earnings. However, this should not apply to all listed companies. Therefore, even if we totally ignore the other business profits, the statistical results and the conclusions are still robust. Here, we would like to express our gratitude to one anonymous referee who suggested that we should make the above analysis over the potential influence of other business profits. 8 Among the 337 samples we have selected, 5 have a net profit but a negative pretax profit. As these companies also have strong motivation for earnings management, they are not excluded from our samples. To avoid any extraordinary value, however, we use net profits rather than pretax profits in the calculation of the variable of these
7
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
When a listed company tries to qualify for refinancing by earnings management, its ROE is often maintained at “6 percent” or “10 percent” (Jiang and Wei, 1998; Sun and Wang, 2000b; Yan, Geng and Liu, 2001). This represents on the one hand that ROE can be an ideal substitute variable for the earnings management behavior of the listed companies, and on the other hand, when this variable is included, a single continuous variable may not bring its role into full play. In other words, the correlation between the audit opinion modification and the ROE should be observed in intervals, and the audited clients with “6 percent” or “10 percent” characteristics should be distinguished from other types of clients. The variable can be set up in two ways: either to set up dummy variables 9 , according to the critical point of ROE (i.e. 6 percent or 10 percent) related to refinancing qualification, or to set up a continuous variable and perform a test on sub-samples in different intervals of ROE. It should be noted that the ROE is the fully diluted index obtained by net profits divided by the number of common shares at the end of the period, rather than the weighted average value. Basic Definition of variable: ROE6=1, when the ROE is more than or equals to 6 percent but smaller than 10 percent in the current period, or =0 when it is not. ROE10=1, when the ROE is more than or equals to 10 percent but smaller than 50 percent in the current period, or =0 when it is not. BIZAROE=1, when the ROE is more than or equals to 50 percent or smaller than zero in the current period, or =0 when it is not 10 . Alternative Definition of variable: For sub-samples in different intervals of ROE, ROE= returns on net assets in the current period. (iii) Type of audit opinion of the previous year. The audit opinions are highly consistent in china’s securities market. When auditors present this year’s audit opinions, they will usually track the auditing emphasis of the previous year and judge in light of the adjustments or improvements made by the audited clients. The inclusion of the previous year’s variable can also help control the previous discrepancies of the listed companies. Definition of variable: PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is either qualified opinion, disclaimer of opinion, or adverse opinion, or =0 when it is not 11 . 5 samples. 9 Such a setup can test the significant impact of different intervals of ROE upon the audit opinion modification. 10 Among the 337 samples, 1 sample has a negative ROE in the current period, and 7 samples have a ROE more than 50 percent (including an extreme case of 6391.25 percent), because these companies have very low or even negative earnings per share (EPS). To avoid distortions by these extreme cases, we group them in the same category and treat them with a dummy variable BIZAROE. 11 In Li and Wu (2002b), the “audit opinions of the previous year” are divided into four types: standard unqualified opinion, unqualified opinion with explanatory notes, qualified opinion, disclaimer of opinion and adverse opinion. The latter 3 types correspond to the three variables (MOD, QUAL, DISCLADV). In our study, as there are only 2 samples with disclaimer of opinion or adverse opinion, we include them in PREQUAL variable rather than set up DISCLADV variable. 8
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
(2) Characteristics Related to auditors A listed company can increase the probability of modifying the audit opinions through selection of auditors. The variables related to auditors include the size of auditors, auditor changes, and the audit tenure. These are all factors that might affect the auditor independence. (i)
The size of auditors
We wish to observe the differences between auditors of different sizes. We focus on three types of accounting firms according to the number of listed clients: the top 10, the next top 10 and the rest. The valuation criterion is based on the data from the Chief Accountant’s Office of the CSRC (2001a, b) on the auditing market of various years (ranking by the number of clients). Definition of variable: TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10, or =0 when it is not. (ii) Auditor Changes If there is a change of the auditor in the reporting period, the succeeding auditor might lower the audit quality to attract clients by means of modifying the audit opinion. To this end, we observe the impact of change of auditors over the modification of audit opinions. Definition of variable: AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. (iii) Audit tenure A lengthy tenure may undermine the auditing independence and quality. For this reason, we include this variable in our model and observe its impact upon the audit opinion modification. Definition of variable: TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company 12 .
3. Research Model As we observe how auditors select between the two types of the audit opinions, we use a binary method in setting up dependent variables. Consequently, the test method of this study adopts the logistic regression, with the following model:
P( AOM ) = γ 0 + γ 1 IRRGLPRF + γ 2 ROE10 + γ 3 ROE6 + γ 4 BIZAROE + γ 5 PREMOD 1 − P( AOM ) + γ 6 PREQUAL + γ 7 TOP10 + γ 8 SECTIER + γ 9 AUDCHG + γ 10TENURE + ε Ln
(1)
AOM=1, when the audit opinion is an unqualified opinion with explanatory paragraphs, or =0
12
The valuation of the variable will meet with such a problem: for a sample company in the IPO year, we only consider the audit tenure only last one year (the listing year) without giving a thought to the tenure of the auditor prior to IPO, and make TENURE=1. 9
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
when the audit opinion is a qualified opinion 13 . IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE10=1, when the ROE of the current period is more than or equals to 10 percent but smaller than 50 percent, or =0 when it is not. ROE6=1, when the ROE of the current period is more than or equals to 6 percent but smaller than 10 percent, or =0 when it is not. BIZAROE=1, when the ROE of the current period is more than or equals to 50 percent or smaller than 0 percent, or =0 when it is not. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is either qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company. The data involved in the above variables are collected from the publicly disclosed annual reports and the statistics of the Chief Accountant’s Office of the CSRC (2001a, b). We use SPSS10.0 to perform related statistical tests. The regression results based on an alternative definition of the independent variable ROE will be discussed in a separate section.
IV. STATISTICAL RESULTS 1.
Descriptive Statistics
Table 2 compares between the samples with unqualified opinions and the samples with qualified opinions in some variables. The non-operating profitability is significantly higher in the former than in the latter, but no significant difference can be observed between two groups in ROE. This represents that the motivation for earnings management is not any weaker in the former than in the latter. Consistence is obvious in terms of the type of audit opinion, and when it comes to the characteristics related to auditors, the first group tends to use large accounting firms, but no significant difference can be observed in the change of auditors and the audit tenure.
13
AOM represents “audit opinion modification”. 10
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
Table 2 A Comparison between Samples with Unqualified Opinions with Explanatory Paragraphs
and the Samples with Qualified Opinions AOM=1 Number of samples
AOM=0 Mean
Number of samples
237.22% 0.35 0.088 0.37 0.19 0.092 4.71
86 86 86 86 86 86 86
T-test T-statistics
Significance level
143.31% 0.16 0.35 0.22 0.21 0.15 4.27
2.031* 3.670*** -4.775*** 2.750** -0.363 -1.387 1.461
0.043 0.000 0.000 0.007 0.717 0.168 0.145
Mean
Panel A：Total samples（n=337） IRRGLPRF PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE
251 251 251 251 251 251 251
Panel B：Comparison of ROE 0%≤ROE<6%
87
2.06%
49
1.84%
0.761
0.448
6%≤ROE<10%
73
7.63%
20
7.49%
0.496
0.621
10%≤ROE<50%
88
14.70%
12
16.77%
-0.994
0.322
Note: ***, **, and * refer to statistical significance at 0.0005, 0.01 and 0.05 levels respectively. AOM=1, when the audit opinion is an unqualified opinion with explanatory paragraphs, or =0, when the audit opinion is a qualified opinion. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in the reporting period, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company. ROE= returns on net assets in the current period.
2.
Logistic Regression of the Total Samples
As there are many control variables while the number of total samples is not large, we use the combination of ENTER method and conditional parameter estimation method 14 in the logistic regression. For the second method, we choose “Forward: Conditional” to describe the test results.
14 According to the principle of conditional parameter estimation, new independent variables are added according to the principle of maximum Chi-Square increment (i.e. most significant in Chi-Square test). This method can help us to select significant independent variables automatically. Although the results may differ with different selection methods, we can still find stable correlation between variables.
11
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
Table 3 Logistic Regression of the Total Samples during 1998-2000 (n=337) Enter Regression coefficient IRRGLPRF ROE10 ROE6 BIZAROE PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE Pseudo-R2 Chi-Square Correct classification of the model
0.091 1.687 0.873 -1.102 0.704 -1.357 0.440 0.130 -0.362 0.057
Wald statistics 3.516 18.728*** 6.445* 1.640 3.621 14.256*** 1.682 0.127 0.504 0.575 0.263 66.181***
Forward: Conditional Significance 0.061 0.000 0.011 0.200 0.057 0.000 0.195 0.722 0.478 0.448
Regression coefficient 0.097 1.733 0.992 0.677 -1.405 -
78.0%
Wald statistics
significance
4.065* 20.783*** 8.818** 3.706 16.164*** 0.239 63.790***
0.044 0.000 0.003 0.054 0.000 -
77.4%
Note: Pseudo-R2 adopts Nagelkerke R2 index, and Chi-Square is the index of the overall effectiveness of the model. ***, **, and * represent statistical significance at 0.0005, 0.01 and 0.05 levels respectively. Under “Forward: Conditional” method, new independent variables are added according to the principle of maximum Chi-Square increment (i.e. most significant in Chi-Square test). The variables which enter the model follow the sequence of: PREQUAL, ROE10, ROE6, PREMOD and IRRGLPRF. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE10=1, when the ROE of the current period is more than or equals to 10 percent but less than 50 percent, or =0 when it is not. ROE6=1, when the ROE of the current period is more than or equals to 6 percent but less than 10 percent, or =0 when it is not. BIZAROE=1, when the ROE of the current period is more than or equals to 50 percent or less than 0 percent, or =0 when it is not. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified Opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in reporting period, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company.
12
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
On the whole, the effect of logistic regression is acceptable. When the above variables are controlled, we can see from Table 3: (i) The two critical points of ROE (i.e. 6 percent and 10 percent) have a significant impact upon the auditor’s reporting behavior. When the ROE of a listed company falls between [6%, 10%) or [10%, 50%), given other conditions being constant, the audit report is more likely to have the unqualified opinion with explanatory paragraphs. When it falls between [0%, 6%) or is extraordinary, the audit report is more likely to have qualified opinion. (ii) The audit opinions of the previous period were significantly consistent. In particular, when a listed company was presented with qualified opinion in the previous year, it is more likely to be presented with the same opinion again. (iii) There is a positive correlation (although the significance might be different under different regression methods) between the non-operating profitability of a listed company and the presence of an unqualified opinion with explanatory paragraphs. In other words, the more the non-operating profit as a percentage of the total profits, the more likely for a company to be presented with unqualified opinion with explanatory paragraphs. (iv) Characteristics related to auditors, such as the auditor size, auditor changes and audit tenure, do not show significant influence upon auditor’s reporting behavior in the regression of total samples.
3.
Logistic Regression of Sub-samples in Different Intervals of ROE
We perform logistic regressions on the sub-samples in different intervals of ROE, and use the real value as the value of ROE variable. Given the limited number of samples with extraordinary ROE and lack of economic significance, the regression is limited to sub-samples with their ROE falling between [10%, 50%), [6%, 10%), and [0%, 6%). Likewise, we use the combination of ENTER method and conditional parameter estimation method to describe the statistical results. (1) Regression Results of Sub-samples with Their ROE Falling between [10%, 50%)
Table 4 describes the regression results of sub-samples with their ROE falling between [10%, 50%). The result shows that PREQUAL variable continues to be significant and that the non-operating profitability is somewhat significant. The real value of ROE is not significantly correlated to auditor’s selection of opinion types.
Table 4 Regression Results of Sub-samples with Their ROE Falling between [10%, 50%) (n=100) 13
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57 Enter Regression coefficient IRRGLPRF ROE PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE Pseudo-R2 Chi-Square Correct classification the model
2.073 -0.046 7.099 -4.587 0.050 -1.601 2.676 0.209
of
Wald statistics 3.905* 0.923 0.058 10.005** 0.002 2.435 2.453 0.958 0.479 28.640*** 90.0%
Forward: Conditional Significance 0.048 0.337 0.810 0.002 0.960 0.119 0.117 0.328
Regression coefficient 1.414 -3.239 -
Wald statistics
Significance
3.368 13.434*** 0.339 19.406***
0.066 0.000 -
91.0%
Note：Pseudo-R2 adopts Nagelkerke R2 index, and Chi-Square is the index of the overall effectiveness of the model. ***, ** and * represent statistical significance at 0.0005, 0.01 and 0.05 levels respectively. Under “Forward: Conditional” method, new independent variables are added according to the principle of maximum Chi-Square increment (i.e. most significant in Chi-Square test). The variables which enter the model follow the sequence of: PREQUAL and IRRGLPRF. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE= returns on net assets in the current period. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in the reporting period, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company.
(2) Regression Results of Sub-samples with Their ROE falling between [6%, 10%)
Table 5 describes the regression results of sub-samples with their ROE falling between [6%, 10%). The regression result under ENTER method is not significant on the whole, which means the dependent variable cannot be properly explained. However, TOP10 variable is significant at 10 percent level. Under “Forward: Conditional “method, only TOP10 variable enters the model. The real value of ROE is not significantly correlated with the type of audit opinions. Such a result tell
14
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
us that when the ROE of a listed companies falls beneath [6%, 10%), the top 10 accounting firms are more inclined to issue unqualified opinions with explanatory paragraphs. This is a unique statistical finding, which will be discussed in other parts of this paper.
Table 5 Regression Results of Sub-samples with Their ROE Falling between [6%, 10%) (n=93) Enter Regression coefficient IRRGLPRF ROE PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE Pseudo-R2 Chi-Square Correct classification of model
-0.191 0.145 1.323 -0.571 1.576 0.229 -0.617 -0.081
Wald statistics 0.416 0.378 2.451 0.547 4.305* 0.120 0.304 0.310 0.204 13.143
Forward: Conditional Significance 0.519 0.539 0.117 0.459 0.038 0.729 0.581 0.578
Regression coefficient 1.597 -
Wald statistics
Significance
5.706* 0.116 7.251**
78.5%
0.017 -
78.5%
Note：Pseudo-R2 adopts Nagelkerke R2 index, and Chi-Square is the index for the overall effectiveness of the model. ***, ** and * represent statistical significance at 0.0005, 0.01 and 0.05 levels respectively. Under the “Forward: Conditional” method, new independent variables are added according to the principle of maximum Chi-Square increment (i.e. most significant in Chi-square test). The only independent variable that enters the model is TOP10. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE= returns on net assets in the current period. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in the reporting period, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company.
(3) Regression Results of Sub-samples with Their ROE Falling between [0%, 6%)
15
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
Table 6 lists the regression results of sub-samples with their ROE falling between [0%, 6%), which show that PREQUAL variable remains significant and the non-operating profitability is significant at 5 percent level under both methods. This represents that, for listed companies with their ROE falling between [0%, 6%), the higher the non-operating profitability, the more likely for auditors to present unqualified opinions with explanatory paragraphs. The real value of ROE is not significantly correlated with the type of audit opinions chosen by auditors. Under “Forward: Conditional” method, AUDCHG variable is significant at 5 percent level and is negatively correlated with the modification of audit opinions.
Table 6 Regression Results of Sub-samples with Their ROE Falling between [0%, 6% ) (n=136) Enter Regression coefficient IRRGLPRF ROE PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE Pseudo-R2 Chi-Square Correct classification of model
0.135 0.240 0.169 -1.528 0.209 0.442 -0.719 0.126
Wald statistics
Forward: Conditional Significance
5.222* 2.777 0.119 8.139** 0.185 0.673 0.821 1.082 0.255 27.968***
0.022 0.096 0.730 0.004 0.667 0.412 0.365 0.298
Regression coefficient 0.112 -1.591 -1.346 -
Wald statistics
Significance
4.862* 10.917** 4.774* 0.216 23.287***
71.3%
0.027 0.001 0.029 -
72.8%
Note：Pseudo-R2 adopts Nagelkerke R2 index, and Chi-Square is the index for the overall effectiveness of the model. ***, ** and * represent statistical significance at 0.0005, 0.01 and 0.05 levels respectively. Under the “Forward: Conditional” method, new independent variables are added according to the principle of maximum Chi-Square increment (i.e. most significant in Chi-square test). The variables which enter the model follow the sequence of：PREQUAL, IRRGLPRF and AUDCHG. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE= returns on net assets in the current period. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in the reporting period, or =0 when it is not. : 16
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57 AUDCHG=1, when there is a change of the auditor in the reporting period, or =0: when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company.
(4) Summary
The regressions of sub-samples in different intervals of ROE show that PREQUAL variable has the same significant correlation in the sub-samples with their ROE falling between [0%, 6%), or [10%, 50%) as in the total samples. The non-operating profitability in these two groups is stably significant. In addition, despite the use of the real value, ROE does not show a significant correlation with auditor’s selection of audit opinions. This in part supports our setup of the dummy variable by interval in the total samples. In other words, the setup of the dummy variable based on intervals of ROE will not incur great loss of information, and such a method is necessary for our study.
4.
Characteristics Related to the Auditor
(1) Auditor Size
In most case, the above tests do not find any significant difference between accounting firms with a large number of clients and those with a small number of clients in the selection of the type of audit opinions. In some particular cases, the large accounting firms are more likely to give unqualified opinions with explanatory paragraphs (as shown in Table 5). Even if we set the value of the “Big Six” 15 as top 10, i.e. TOP10=1, the statistical significance of the above results does not change. Therefore, for listed companies that have a net profit and are given a non-standard opinion, there is no significant difference between large and small accounting firms in their audit reporting behavior. For this reason, we cannot jump to the conclusion that large accounting firms do not modify their audit opinions. Although Li and Wu (2002b) found that large accounting firms are more unlikely to modify the audit opinions, their research is based on unavoidable opinions. Considering that our research focuses on avoidable audit opinions, our study does not conflict with previous studies and may even draw a conclusion that auditor’s reporting behavior is related to the avoidability of the audit opinions. Though previous studies (e.g. Li, 2002; Zhang and Liu, 2002) reported a significant difference between large firms and small firms in audit quality, we think that these studies are not comparable to ours, as they are different in the sampling period, number of samples, study objects and setup of variables. In addition, these studies did not distinguish different types of non-standard opinions, and had mixed qualified opinions with 15 Our sample period is from 1998 to 2001. During this period, Price Waterhouse merged with Coopers and Big Six became Big Five. However, there were still six Sino-foreign cooperative accounting firms in China during this period (Price Waterhouse Coopers Zhongtian, Arther Andersen Huaqiang, KPMG Huazhen, Ernst & Young Huaming, Deloitte & Touche Hujiang, and Coopers Zhongxin). Influenced by the Enron Case, Big Five became Big Four in the year 2002 and six Chinese Sino-foreign cooperative accounting firms turned into five ones.
17
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
unqualified opinions with explanatory paragraphs. All these will lead to discrepancies in results. (2) Auditor Changes
Li and Wu (2002a) found that after a listed company changes its auditor, the audit opinion tends to be more stringent (while it is opposite for listed companies who were presented with non-standard opinions in the previous period). Table 6 of our study also reveals that, for listed companies with their ROE falling between [0%, 6%) and given a non-standard opinion, auditors who serve them for the first time are more likely to express qualified opinions. There is some sense in such a result. First, companies with their ROE falling between [0%, 6%) are facing more imminent financial difficulties. Second, the management of these companies with a strong motivation for loss-aversion often engages in earnings management involving the critical point of either making profits or suffering losses, and therefore its nature is more severe. These two factors expose auditors to a higher audit risk. From this perspective, the audit opinion for such companies has stronger inavoidability. In general, the new auditor is under greater pressure from the regulator (many companies in our samples changed their auditors because the former auditors were punished by the regulator), which will have a natural influence upon the audit reporting behavior of the successor 16 . (3) Audit Tenure
Li and Wu (2002b) found that, the longer the audit tenure, the more likely for auditors to make audit opinion modifications over the going concern uncertainties. However, our study does not find any significant influence of the audit tenure upon the auditor’s choice of non-standard opinion for companies with a net profit. One prevailing idea is to implement a compulsory rotation system to prevent the potential impairment of audit quality because of overlong tenure, but our study does not provide any empirical evidence to support such a proposal.
V. POLICY EFFECT OF DOCUMENT NO. 14 OF THE CSRC: PRELIMINARY TEST ON THE FIRST YEAR 1. Expected Effect of the Regulatory Policy Article 5 of Document No. 14 has expressed clearly the idea of the regulator on the containment of the modification of audit opinions. It requires auditors to observe the principle of prudence, and to maintain independence in face of pressure or persuasion from the listed companies. Articles 7, 8 and 10 stress the legal responsibility and economic consequence when the auditor expresses an
However, this change occurred after the end of the 2001 fiscal year auditing and had no impact on this study. 16 As the number of samples with compulsory replacement of auditors (mainly because the former auditor was punished by the regulator and had to be replaced) is small, we cannot perform an independent test on the influence of voluntary and compulsory replacement upon the audit reporting behavior of auditors . 18
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
audit opinion. From the perspective of the regulatory intention, Document No. 14 is not only to contain the potential audit opinion modifications, but also increases punishments over any violations of the accounting standards and systems. On the other hand, however, the legal system and corporate governance is still immature in China, and the motivation for fraudulent financial reporting remains strong. Therefore, the release of Document No. 14 may aggravate the inclination to avoid the non-standard opinion and to express audit opinions that “do not obviously violate the accounting standards and systems as well as information disclosure requirements”. Our research interest lies in: How will auditors change their discretionary audit reporting behaviors under the dual pressure from the audited clients and the regulators? In other words, can Document No. 14 play its expected role in the first year after it comes into effect?
2. Audit of Financial Statements of Listed Companies for the Year 2001 and Sample Selection In year 2001, among a total of 1173 listed companies (excluding 2 companies that did not disclose their annual reports on time), 157 listed companies were presented with non-standard opinions, in which 19 were presented with disclaimer of opinion, 44 were presented with qualified opinion and 94 were presented with unqualified opinion with explanatory paragraphs (including 2 companies which only issue B-shares). Based on the same selection criteria and procedures in Part II and Part III, we select 77 A-share companies as samples for 2001. As shown in Table 7, all of them had a net profit and were presented with an unqualified opinion with explanatory paragraphs. Together with the 337 samples between 1998-2000, we test on a total of 414 samples from 1998 to 2001. It is our belief that, with the inclusion of the samples in 2001, we can perform a preliminary test on the policy effect of Document No. 14. Table 7 Sample Selection Procedures and Results for the Year 2001 Year 2001 Panel A：Unqualified opinion with explanatory paragraphs Total number Less：Number of audited clients with a net loss The resulting number of samples
92 37 55
Panel B：Qualified opinion Total number Less： Number of audited clients with a net loss
44 22
The resulting number of samples
22
Total of samples in the current year #
77
Note: Data on audit opinions are manually collected through the reading of the 2001 annul reports of the listed companies, and checked against the statistical data of the Chief Accountant’s Office of the CSRC 17 . Data on the
17
We would like to express our gratitude to Dr. Shuhua Li from the Chief Accountant’s Office for providing the 19
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57 net profit of the current period are manually collected from the 2001 annual reports of the listed companies. #: In the 77 samples of 2001, one company is in a special case. It first disclosed a net profit in the 2001 annual report that was given an unqualified opinion with explanatory paragraphs, but later it adjusted its financial reports as required by the regulatory agencies, and restated its financial statements with a net loss in the same year. We believe that this sample alone has revealed the strong opportunism of the management of the sample company and the auditor in their external reporting behavior. Therefore, despite a net loss in the end, we still include this company in our samples and use its first disclosed data for our tests.
3. Expansion of the Original Research Model The research model in Part III mainly considers two types of factors that may influence the audit reporting behavior of auditors: characteristics related to audited clients and characteristics related to auditors. For the reason discussed in this part, we now need to consider the third factor external regulation. Given the increasing rigid environment in regulation in 2001, we expect that Document No. 14 will play a direct and active role in containing the behavior of auditors to present unqualified opinion with explanatory paragraphs. Similarly, auditors are expected to be more conservative and the tendency of audit opinion moderation will decrease. A dummy variable Y2001 is set up to observe the impact of this variable upon the auditor’s reporting behavior. We expect a negative correlation between Y2001 and the dependent variable. Here is the expanded model: P( AOM ) = γ 0 + γ 1IRRGLPRF + γ 2 ROE10 + γ 3 ROE6 + γ 4 BIZAROE + γ 5 PREMOD 1 − P( AOM ) + γ 6 PREQUAL+ γ 7TOP10 + γ 8SECTIER + γ 9 AUDCHG + γ 10TENURE + γ 11Y 2001+ ε Ln
（2）
All the data involved in the above variables for 2001 are manually collected from the publicized 2001 annual reports.
4. Logistic Regression of the Total Samples from 1998 to 2001 The sample size of year 2001 is not large, and the number of samples divided according to the different intervals of ROE is even smaller. Consequently, they do not have a substantial impact upon the ROE distribution features of the original sampling period (1998-2000). In addition, we have noticed that on March 15, 2001, the CSRC released the new qualifications 18 for issuance of new shares and abolished the 1999 qualifications for refinancing (the average net ROE for recent three years shall not be less than 10 percent, and the net ROE for each year shall not be less than 6
statistical data. 18 According to the "Circular On Improving the Work Surrounding New Share Issuance by Listed Companies" (CSRC [2001] 43) promulgated on March 15, 2001 by the CSRC, any company that applies for IPO must meet the following requirements besides the qualifications stipulated in the "Methods on Managing New Share Issuance by Listed Companies": audit by auditors, and at least 6 percent of a weighted average net ROE for the recent 3 years. The weighted average should be the lower between the net ROE after non-recurrent profits and losses and the net profits before the deduction. 20
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
percent). For this reason, the “6 percent” or “10 percent” phenomenon is significantly reduced in the 2001 annual reports. Our test for 2001 will be limited to the total samples (altogether 414 observations) and no division will be made as in Part III and Part IV. The regression model (2) still reserves the variable of ROE intervals, but our explanations for the statistical results will take into account the economic meanings as a result of policy changes. Table 8 describes the regression results of the total samples from 1998 to 2001.
Table 8 Regression Results of the Total Samples from 1998 to 2001 (n = 414) Enter Regression coefficient IRRGLPRF ROE10 ROE6 BIZAROE PREMOD PREQUAL TOP10 SECTIER AUDCHG TENURE Y2001 Pseudo-R2 Chi-Square Correct classification of the model
0.079 1.655 0.853 -1.067 0.542 -1.336 0.612 0.247 -0.450 0.018 0.336
Wald statistics 4.482* 21.053*** 7.689** 2.133 2.818 17.769*** 4.255* 0.561 1.114 0.070 1.018 0.249 76.986***
Forward: Conditional Significance 0.034 0.000 0.006 0.144 0.093 0.000 0.039 0.454 0.291 0.792 0.313
Regression coefficient 0.072 1.560 0.827 -1.556 0.642 -
78.0%
Wald statistics 4.246* 19.935*** 7.615** 28.202*** 5.410* 0.224 68.731***
Significance 0.039 0.000 0.006 0.000 0.020 -
77.8%
Note：Pseudo-R2 adopts Nagelkerke R2 index, and Chi-Square is the index for the overall effectiveness of the model. ***, ** and * represent statistical significance at 0.0005, 0.01 and 0.05 levels. Under the “Forward: Conditional” method, new independent variables are added according to the principle of maximum Chi-Square increment (or most significant in Chi-square test). The variables which enter the model follow the sequence of：PREQUAL, ROE10, TOP10, ROE6 and IRRGLPRF. IRRGLPRF= (pretax profit of the current period – operating profit +other business profits)/ pretax profit of the current period. ROE10=1, when the ROE of the current period is more than or equals to 10 percent but less than 50 percent, or =0 when it is not. ROE6=1, when the ROE of the current period is more than or equals to 6 percent but less than 10 percent, or =0 when it is not. BIZAROE=1, when the ROE of the current period is more than or equals to 50 percent or less than 0 percent, or =0 when it is not. PREMOD=1, when the audit opinion of the previous year is unqualified opinion with explanatory paragraphs, or 21
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57 =0 when it is not. PREQUAL=1, when the audit opinion of the previous year is qualified opinion, disclaimer of opinion or adverse opinion, or =0 when it is not. TOP10=1, when the auditor is ranked among the top 10 in the reporting period, or =0 when it is not. SECTIER=1, when the auditor is ranked among the next top 10 in the reporting period, or =0 when it is not. AUDCHG=1, when there is a change of the auditor in the reporting period, or =0 when there is no change. TENURE=the number of consecutive years that the existing auditor has been the auditor of a sample company. Y2001=1, when the sample is in year 2001, or =0 when it is in any year of 1998, 1999 and 2000.
On the whole, the logistic regression results of Table 8 are basically the same with those in Table 3. A comparison between them reveals that after including samples of year 2001, there is almost no change in the statistical results of the original model as to most correlations of variables and their significance level, and the dummy variable Y2001 has no significant correlation with the dependent variable as well. However, there are two interesting changes: (1) The variable IRRGLPRF demonstrates a more significant correlation with the dependent variable from the original level of 10% to 5%. (2) The variable TOP10 becomes significantly and positively correlated with the dependent variable at five percent level. 19
5. Analysis of Statistical Results (1) Role of the Dummy Variable Y2001
The first important issue that deserves our discussion is the role of the variable of year 2001. The statistical results show that, though the correlation between Y2001 and the dependent variable is not statistically significant, the coefficient is positive. Further, the variable IRRGLPRF demonstrates a more significant correlation with the dependent variable from the original level of 10% to 5%, representing that Document No. 14 at least did not produce any effect upon audit reports that is significantly different from the previous years (i.e. the choice of non-standard opinions). On the contrary, the auditor’s inclination to express the unqualified opinion with explanatory paragraphs increases to some extent. As mentioned before, the unfavorable practicing environment in China has a significant negative impact upon the audit reporting behavior of auditors (which is also the conclusion supported by the evidences from Part II to Part IV). Document No. 14 is expected to deal with this problem, but the test results show that its containment is quite limited. 19 To increase the robustness of our test results, we use "Backward: Conditional" method to perform a logistic regression. The unreported regression results show that the variable of TOP10 remains in the model, and the regression results of significance level, variable coefficient and variable symbol are basically the same with those obtained under both the ”Enter” and "Forward: Conditional" method.
22
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
(2) Statistical Results of TOP 10
Another important issue is the change of audit reporting behaviors in the top 10 accounting firms 20 . The statistical tests on samples between 1998-2000 do not show any significant difference between accounting firms with a large number of clients and those with a small number of clients in their choice of audit opinions (but as shown in Table 5, large accounting firms are more likely to present unqualified opinion with explanatory paragraphs in some cases). However, when samples of 2001 are included, the top 10 accounting firms are found to be more inclined to present unqualified opinion with explanatory paragraphs 21 . Therefore, we can conclude that the top 10 accounting firms are more likely to modify the audit opinions. To explain the above statistical results, we further observe 77 samples of year 2001 and find that 33 samples are employing the top 10 accounting firms. The composition of the top 10 does not change much between 2001 and 2000, as 8 accounting firms are listed as top 10 in both years 22 . To understand the modification tendency of different accounting firms, we perform tests on accounting firms with observations more than or equal to three 23 . The method is as follows: for the first tested accounting firm, the original coding of TOP10 as 1 is replaced by zero, and model (2) is run under the “Forward: Conditional” method to observe the resulting differences in the statistical results (i.e. entry into the model, variable coefficient, variable symbol and significance level). When the second accounting firm is tested, we resume TOP10=1 for the first tested accounting firm, and then repeat the above process. These tests reveal that, when TOP10 is set at 0 for two of the top 5 accounting firms, the variable TOP10 does not enter the model, representing that the rest of the top 10 do not significantly differ from other accounting firms in the modification tendency. The two accounting firms in question are the main actors contributing to the statistical features in Table 8. When the other three in the top 10 are tested, TOP10 still enters the model, and the test results of two firms do not show any significant change, while the test results of one firm show that the significance level of TOP10 is increased from 5 to 1 percent. From these tests, we can conclude that the statistical features of TOP10 are mainly brought by a few large firms. Such a behavioral inclination might be closely related to their business size, practicing style and negotiation power with the regulator. 20
According to our statistics as of April 30, 2002, the top 10 accounting firms should be: Shanghai Lixin Changjiang (51), Da Hua (46), Shenzhen Pengcheng(44), Zhejiang Tianjian(40), Hubei Daxin (36), Beijing Jingdu (33), Wuzhou United (31), Shanghai Shangkuai (30), Sichuan Junhe (29), Arthur Andersen Huaqiang (29) and Wulian United (29). Except Arthur Andersen Huaqiang, the other 10 accounting firms are all taken as top 10 in model (2). In the sensitivity test, Arthur Andersen Huaqiang is also included in TOP10. 21 Likewise, even when we include the joint ventures of the "Big Six" in the variable of TOP10, the statistical results do not change significantly. 22 This means that the change of statistical features of the variable TOP10 was not a result of the change of market structure in China’s audit market in 2001 (seven accounting firms with securities practicing license were deprived of their qualifications because of practicing quality, which triggered off a major replacement of auditors and a major change in market structure). 23 Altogether 26 observation values involving 5 accounting firms, with 7 for Shanghai Lixin Changjiang, 5 for Shenzhen Pengcheng, 7 for Zhejiang Tianjian, 4 for Hubei Daxin and 3 for Beijing Jingdu. 23
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
VI. CONCLUSIONS AND DISCUSSIONS 1. Major Research Conclusions The securities regulator is highly concerned over the audit opinion modifications as these behaviors often conceal the motivation and schemes of the listed companies to manipulate profits and reduce the usefulness of accounting information in the securities market. This paper attempts to systematically discuss the listed companies that had a net profit during 1998-2001 and were given a non-standard opinion. Some factors are examined to observe the potential differences between different accounting firms in their audit reporting behaviors (i.e. selection between unqualified opinion with explanatory paragraphs and qualified opinion). Our major findings are: First, when the ROE of a listed company falls between the sensitive intervals (including [6%, 10%) and [10%, 50%)), auditors are more inclined to give unqualified opinion with explanatory paragraphs 24 . We believe that such a phenomenon is closely related to the motivation of earnings management. In other words, when a company has a strong motivation to maintain its qualification for refinancing, it tends to impose more pressures to or provide more “baits” for auditors who in turn are more likely to modify the audit opinions. This finding is consistent with Wang and Chen (2001) who discovered that auditors react totally differently to companies who use earnings management to realize minor profits and those who use earnings management to maintain their qualification for rights issue. They explained, “ The former was risk based, while the latter was the result of competition”. We believe that this also shows the differences between these two types of companies in avoidability of audit opinions. Second, the larger the non-operating profits as a percentage of pretax profits, the more likely for auditors to give unqualified opinion with explanatory paragraphs. This reflects the negative impact of a company’s motivation for earnings management upon the auditor’s reporting behavior. Third, the audit reporting behavior in year 2001 does not differ significantly from that of previous years. On the contrary, the auditor’s inclination to express the unqualified opinion with explanatory paragraphs increases to some extent. For companies who had a net profit but were presented with a non-standard opinion in 2001, Document No. 14 does not have a significant effect to contain the selection of non-standard opinion by auditors. Finally, the characteristics of auditors do not have a significant impact upon the choice of different audit opinions. Big accounting firms do not differ significantly from small accounting firms when they give audit opinions to companies with potential motivation for earnings management. In 2001, the top 10 accounting firms were more inclined to modify their audit opinions, and a further sensitivity test reveals that the statistical significance of TOP10 variable 24
The unreported tests show that when the ROE falls between [10%, 50%], auditors are more inclined to present unqualified opinions with explanatory paragraphs than when the ROE falls between [6%，10%). 24
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
are mainly caused by a few big accounting firms.
2. Policy Implications and Suggestions As auditor’s reporting behavior is significantly influenced by the earnings management of the listed companies which will in turn incur severe economic consequences for information users (i.e. deterioration of accounting information quality and motivation for rights issue through earnings management), this paper provides empirical evidence to support the attention by the CSRC to the modification of audit opinion and accordingly the necessity to release Document No. 14. However, tests on the data of 2001 do not find the expected effects of the related regulatory policy. Between the clients and the regulatory policy, auditors do not obviously favor the latter. According to our observation, the stock exchanges did not suspend, after the promulgation of Document No. 14, the trading of any listed company presented with non-standard opinion. Many accounting firms have attempted to avoid such expressions as “obviously violating the accounting standards, systems and information disclosure requirements”. It is possible that auditors are taking the advantage of the nature of unqualified opinion with explanatory paragraphs for modification, and likely to replace the due responsibility of “professional judgment” with “mere description of dubious transactions or events”. One reason contributing to this phenomenon is that auditors are a weak group and their actions are under great pressure from external interest groups, including the listed companies, the local governments and other related interest groups. When the regulator attempts to impose stricter punishments for violations, the management of a listed company will transfer such pressure to auditors. As long as the pressures from the external interest groups are great enough, the inclination to modify the audit opinions will not be decreased. On the contrary, the possibility for auditors to express standard unqualified opinions may well be increased 25 . On the other hand, the fact that not a single listed company that has been suspended from trading has in fact encouraged some listed companies and their auditors to take an opportunistic view towards financial frauds. This has certainly undermined the expected effects of the regulatory policy. Therefore, regulation is an important means to improve the audit quality, but its role is also limited by the external environment. Auditors alone cannot increase their independence. They need to be relieved of the external pressures, which relies very much on the improvement of the ownership structure and corporate governance. If the external pressures continue to dominate, the independence of audit cannot be realized merely through regulation or profession self-discipline. We believe that the improvement of accounting information in the securities market is a process of taking overall measures. A balance of regulation toward different players in the market must be achieved. 25
Out of the consideration of regulatory costs and performance, the regulator is more likely to track those companies whose problems have already been identified by auditors, and this may encourage auditors to shift from “modifying the audit opinion” to presenting standard unqualified opinion directly. This of course increases the 25
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
We also observe that the top 10 accounting firms have a stronger inclination to modify the audit opinions, and such an inclination is even more obvious in 2001. This means that the regulatory policy does not have an obvious deterring effect upon large accounting firms. In addition, the behaviors of large accounting firms may encourage other accounting firms to follow suit, and thus jeopardize the regulation and healthy development of the audit market. When large accounting firms have a stronger bargaining (lobbying) power or a close relationship with the regulatory agencies, the containment role of the regulatory policy may be greatly reduced. To this end, the regulators should be aware of and try to inhibit such modification behaviors by large accounting firms. Our research also supports the CICPA to put the work of revising audit report standard on their 2002 agenda. It is suggested to take drastic measures to effectively inhibit the modification behavior (the presentation of unqualified opinion with explanatory paragraphs in particular). In some newly released exposure drafts of auditing standards in September 2002, such as “Specific Independent Auditing Standard No. 7 - Audit Report” (revised), “Specific Independent Auditing Standard No. 17 - Going Concerns” (revised), and “Guideline for Practices of Chinese CPAs No. 5 - Audit Report”, the CICPA has shown its self-disciplinary intention and policy orientation towards audit report modifications. In particular, these standards have provided some specific rules on unqualified opinion with explanatory note (Yang, 2002). Users of accounting information are advised to pay close attention to the audit reports when they are reading the annual reports by listed companies whose ROE falls between the sensitive interval. Our research reveals that, under the existing environment, information users should “undervalue” or even “ignore” the type of audit opinions presented by auditors, but give special attention to the unqualified opinion with explanatory paragraphs and analyze the implications of the explanatory paragraphs.
3. Research Limitations and Necessary Explanations As the modification of audit opinions is not observable, our research chooses an alternative way to observe the differences in characteristics between unqualified opinion with explanatory paragraphs and qualified opinion. Our samples are limited to companies that have a net profit and are given either of the above two audit opinions. This may inevitably include some samples who do not have any motivation for earnings management. In addition, the differences in financial characteristics are not under full control when we set up the variables, and as a result, any manipulation of operating profits is not captured by any variable in our regression models. Also, the effects of a regulatory policy can be evaluated from different perspectives, and a knowledge of the background and the implementation of the regulatory policy is crucially
difficulty and costs of regulation. 26
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
important for any researcher. Our study is only based on one perspective, and for this reason, we might have lost the sight of other perspectives that can reflect the policy effects of Document No. 14. In addition, the variable of Year 2001 is expected to reflect the policy effect of Document No. 14, while in reality it can mirror the overall effects of all events that could have influenced the audit reporting behaviors of auditors in 2001. It is really difficult for us to segregate one event to observe its effect 26 . We have considered one more criterion of “expressing in explicit terms the problems concerning recognition and measurement of profits and losses in the audit opinion”. However, this is not mentioned in many audit opinions, as it is the more covert modification of audit opinions (Zhang and Liu (2002) had a similar idea in their sample selection process). Finally, we choose companies with non-standard opinions and a net profit as our samples, and hypothesize implicitly that unqualified opinion with explanatory paragraphs is a modification of audit opinion. We do not exclude audit opinions that are truly reasonable, and in fact we can hardly quantify the complexities in audit practices. We only attempt to identify statistically significant factors without any evaluation of or judgment over the audit quality of any particular listed company or accounting firm. In addition, we have restricted our study to qualified opinion and unqualified opinion with explanatory paragraphs, but we do find the same obvious motivation for earnings management in many listed companies who are presented with standard unqualified opinion. These will be our future interests of research.
26
Of course, this limitation does not affect our conclusions. In the year of 2001, as the total regulatory environment of the securities audit market in China was more stringent on the whole, regulatory policies other than Document 14 of CSRC may have reinforced the regulatory policy effects. Thus, Year 2001 variable should in theory reveal stronger regulatory policy effects on the professional behavior of the auditors and our tests have actually observed the change of professional behavior of auditors under the combined effects of these policies. As our test results do not observe any significant regulatory effect, the regulatory effect of Document 14 is of course weaker as against the combination of many policies. 27
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
REFERENCE Chen, C., S. Chen, and X. Su (2001), ‘Profitability Regulation, Earnings Management and Modified Audit Opinions: Evidence from China’, Auditing: A Journal of Practice & Theory 20 (Fall). Chief Accountant’s Office of the CSRC (2001a), Who Audits China’s Securities Market: Analysis of the Audit Market (1997-1999), China Finance and Economics Press. Chief Accountant’s Office of the CSRC (2001b), Who Audits China’s Securities Market: Analysis of the Audit Market (2000), China Finance and Economics Press. Chief Accountant’s Office of the CSRC (2001c), auditors say ‘no’--- Analysis of the Audit Opinions of the Listed Companies in China (Exposure Draft), China Finance and Economics Press. Jiang, Y. H. and G., Wei (1998), ‘ROE and Qualifications for Refinancing’, Empirical Studies on Accounting Issues in the Securities Market, etd. by Jiang, Y. H. and S.H., Li, Shanghai Finance and Economics University Press.223-235. Li, S. and X., Wu (2001), ‘Market Reaction to Audit Opinion: Model Analysis and Implications under Avoidability Hypothesis’ (translation and compilation), Accounting Research 6:58-63 Li, S. and X., Wu (2002a), Research on Auditor Changes: Preliminary Evidence from China’s Securities Market, China Finance and Economics Press. Li, S. and X., Wu (2002b), ‘System factors and Audit Quality: Empirical Evidence from Audit Opinions on Going Concern Uncertainties’, Chinese Certified Public Accountants 6: 29-34. Li, S. H. (2000), Increasing Independence of Audit and Deviation of the Audit Market, Shanghai Sanlian Publishing House. Melumad N.D., and A. Ziv (1997), ‘A Theoretical Examination of the Market Reaction to Auditors’ Qualifications’, Journal of Accounting Research 35: 239-256. Sun, Z. and Y. T., Wang (2000a), ‘An Empirical Study of the Explanatory Paragraphs and Modification of Audit Opinions”, China’s Securities Market: A Financial and Accounting Perspective, etd. by Sun, Z. and Z. Q., Li, Shanghai Finance and Economics University Press. 33-43 Sun, Z. and Y. T., Wang (2000b), ‘An Empirical Study on Resource Allocation and Earnings Manipulation”, China’s Securities Market: A Financial and Accounting Perspective, etd. by Sun, Z. and Z. Q., Li, Shanghai Finance and Economics University Press. 119-130. Wang, Y. T. and S. M., Chen (2001), ‘An Empirical Study on the Impact of the Ownership Reform and Separation from Affiliated Entities upon the Independence of Audit’, Auditing Research 3: 2-9. Yan, D. W., J. X., Geng and W. P., Liu (2001), ‘An Empirical Study on the Financing Activities of 28
Li Shuang and Xi Wu. Audit Opinion Modification and Its Regulation: Empirical Evidence. China Accounting and Finance Review. Vol.4 No.4 (December 2002): 29-57
the Listed Companies through Rights Issues’, Accounting Research 9: 21-27 Yang, Z. G. (2002), ‘Some Issues on Audit Report for the Listed Companies’, Chinese Certified Public Accountants 10: 27-31. Zhang Y. K., and F., Liu (2002), ‘An Empirical Study on the Correlation between Earnings Management and Audit Opinions’, China Accounting and Finance Review 4: 1-14.
29
