Disclosure of Critical Audit Matters, Audit Risk and Audit Opinion

Xiaojuan YANG; Shixiang DAI

doi:10.21078/JSSI-2022-354-34

PDF(215 KB)

Journal of Systems Science and Information ›› 2022, Vol. 10 ›› Issue (4) : 354-387. DOI: 10.21078/JSSI-2022-354-34

Disclosure of Critical Audit Matters, Audit Risk and Audit Opinion

Author information +

History +

Abstract

We explore the association between the number of critical audit matters and auditing opinion based on the Chinese capital market data. In addition, we investigate the effect of audit risk on the relationship between the number of critical audit matters and auditing opinion. Using 6, 662 firm-year observations listed in the Chinese capital market from 2017—2019, we find that the number of critical audit matters has significantly positive association with clear audit opinion. Specially, the greater the number of critical audit matters, the more likely getting clear audit opinion. We also find that the audit risk restricts the effect of the critical audit matters on the audit opinion. Lastly, we find that Big4 accounting firms are less likely to be influenced by the number of critical audit matters; the audit complexity restricts the effect of the critical audit matters on the audit opinion; the more board members, the less effect of the number of critical audit matters on the audit opinion. We also use several robustness tests to strengthen the conclusion. Our research may contribute to the understanding of the new audit standard.

Key words

critical audit matters / audit opinion / audit risk / confirmation bias theory

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

Xiaojuan YANG , Shixiang DAI. Disclosure of Critical Audit Matters, Audit Risk and Audit Opinion. Journal of Systems Science and Information, 2022, 10(4): 354-387 https://doi.org/10.21078/JSSI-2022-354-34

1 Introduction

There is a significant reform of audit report between 2015 and 2017. As early as 2015, the IAASB made significant changes to the traditional audit reporting, including the requirement of disclosing key audit matters (KAMs) in audit report. After that, in 2017, the PCAOB promulgate AS3101, The Auditor's Report on an Audit of Financial Statements When the Auditor Expresses an Unqualified Opinion, which makes it clear that auditor should report critical audit matters (CAMs) in financial audit report^[1]. In 2016, the Ministry of Finance of the People's Republic of China adopted a new auditing standard that contains the same requirement of the IAASB and the PCAOB. According to the definition of PCAOB and Ministry of Finance of the People's Republic of China, CAMs refer to the material matters that should be chosen from matters which auditor communicated with the audit committee and are the most significant in the audit process. More specifically, since there are many professional judgments in the audit process, CAMs are intended to communicate the most challenging, subjective, or complex auditor judgments to the investors. The Big4 firms and investors generally support that the new implementation of critical audit matters can be a meaningful method to improve the usefulness and information value of the audit report^[2]. However, there is little research express the concern that CAMs may have significant impact on auditor's judgments. In this paper, we investigate the association between the number of critical audit matters and audit opinion. Specifically, we examine whether the number of critical audit matters influences audit opinion.

The reform of audit report is a significant milestone in auditing history. Before 2015, there is no extra information in the audit report but only pass/fail audit opinion. After 2008 economic crisis, audit report has been criticized by providing little information. Providing critical audit matters in the audit report may improve the information environment^{[1, 3]}. Current studies demonstrate that the additional disclosure in audit report is expected to fortify the information content^[2].

Numerous research claim that the disclosure of critical audit matters not only increases the transparency but also makes the audit report more informative^[4-6]. At the same time, abundant research find that the inclusion of CAMs in audit report can reduce auditor liability perceived by jurors^{[7, 8]}. Even though the new audit standard reform leads to a lot of research, there are few studies about the association between the critical audit matters and audit opinion.

We extend existing research by investigating whether the number of CAMs in the audit report has significant impact on the auditor's audit opinion. We provide empirical evidence to this question. According to the theory of Confirmation Bias^[9-11], people tend to pursue evidence that supports their beliefs and decisions. Auditors generally hold the belief that their responsibility has been met and there is little audit liability should be attributed to auditor themselves if misstatements are found. Audit opinion with CAMs provide an opportunity to solidify this kind of "will not be blamed'' belief. This is because audit reporting with CAMs can send more information to financial statement users than the traditional pass/fail audit opinion and the more CAMs included, the more information can be provided. In addition, since CAMs are intended to communicate the most challenging, subjective, or complex auditor judgments to the investors, exposing important subjective judgments to investors can improve auditor's belief that they can use audit report to defense themselves if misstatements have been found. In other words, the perceived audit risk is lower when audit report with CAMs is released than when audit report with pass/fail opinion is released. Further, the more CAMs are released in the audit report, the lower audit risk will be perceived. Audit risk model guides auditors to express a clear opinion when auditors decrease audit risk to a tolerance interval^{[12, 13]}. Therefore, we make a hypothesis that the more critical audit matters, the more likely a clear audit opinion and we provide empirical evidence to our hypothesis.

In our paper, we also examine the effect of audit risk on the association of the number of critical audit matters on audit opinion. Psychology research suggests that people's control ability of situation lead them to be more cautious^[11]. Since high audit risk may lead to less controllable situation, auditor will be more willing to monitor their behavior when audit risk is high. In this case, real audit risk will be more influential than perceived audit risk. Therefore, we assume that the real audit risk may moderate the correlation between the number of critical audit matters and audit opinion. In this paper, we confirm this prediction by providing empirical evidence using Chinese market data.

We provide several significant contributions to the existing research. First, our research contribute to the research method in the filed of audit critical matters. Because of limited research data, most of the studies about critical audit matters use experimental method. Even though the experimental method is more detailed, most of the experiments are limited by external validity. Archival research can provide remedy to this disadvantage, providing more general conclusion by using comprehensive market data. Therefore, our conclusion can be used in broader scope.

Second, we extend the current research by confirming that critical audit matters have significant impact on auditor judgment. Audit opinion is an important information source for investors. Even though financial statements provide a lot of information to investors, audit opinion determines whether investors can trust the information in the financial statements. Thus, improving the understanding of whether and how audit opinion has changed under new audit standard is valuable. However, current research has not made an absolutely clear conclusion on the impact of critical audit matters on auditor behavior. Brasel, Doxey, Grenier, and Reffett (2016) have used experimental method to prove that critical audit matters have significant influence on auditor liability^[7]. However, they only test one possibility of the influence of critical audit matters on auditors. Our findings go further and extend this area of the researches by linking critical audit matters and audit opinion.

Third, our conclusion has significant implications for regulators and auditors. While numerous research prove that disclosure of critical audit matters is profitable^[2], our conclusion suggests that auditors themselves may unconsciously influenced by critical audit matters and issue a more clear opinion than the company deserves. In this case, auditors should be more cautious on audit opinion when they must disclose critical audit matters in the audit report. Our research inform an unintended consequences associated with the reform of audit reporting. According to our findings, regulators not only need to inform users the unintended influence of CAMs, they also need to carefully monitor auditor's decision-making process. We arrange the remainder of our paper as follows. Section 2 provides background information and development of our hypotheses. Sections 3 and 4 describe our empirical design and results. Section 5 states additional analysis. Section 6 provides robustness check and Section 7 offers research conclusions.

2 Literature Review and Hypothesis Development

2.1 Regulatory Background and Literature Review

In ISA701, the IAASB adopted the requirement to report key audit matters in audit report. Later, in 2017, the PCAOB announced a new auditing standard and admitted the disclosure of critical audit matters in the audit report. This means several years of discussion of the reform of audit report finally comes to results. In 2016, the Ministry of Finance of the People's Republic of China also issued a final audit standard to claim that all Chinese companies' audit reports should include critical audit matters since 2017. According to the new proposed auditing standard, auditors not only need to discuss critical audit matters in the audit report, but also should include the reasons why the matters were appropriate to be classified as critical audit matters^{[14, 15]}.

The development of audit report is the result of the investor's and regulators' concern that the traditional audit report, which only provide a pass or fail conclusion, cannot inform financial statements users the sufficient information and therefor is useless for users^[16]. Even though the initial purpose of critical audit matters is to solve the concern of limited information in audit report and to make audit report more informative, abundant research have voiced that CAM not only increases the transparency and makes the audit report more informative^[4-6], but also influences the behavior of auditor, audit committee, investor and juror^[7].

Since critical audit matters are aimed to disclose companies' risk information during the auditing, Christensen, et al. found that the inclusion of CAMs in the audit report decreases investor's likeliness to invest^[2]. This result indicates that users treat critical audit matters as important information and pay much attention to them^[17]. In fact, critical audit matters not only make auditors unconsciously exert more effort in auditing, but also uncover potential managerial misreporting^[18]. This kind of risk information may alert investors to rethink before investing in real business. Even though the inclusion of risk information in the audit report is harmful for the company to attract investors, some studies find that it is actually a benefit thing for auditors. Brasel, et al. found that the inclusion of CAMs in audit report can reduce auditor liability perceived by jurors^[7]. In addition, Wright and Wright found that the disclosure of judgment process information can mitigate auditor attribution^[8]. However, there is also some research illustrates that key audit matters increase investors' perceived auditors' negligence^[19] and responsibility^[20]. In practical, critical audit matters could be replaced in the next year or reported continually for several years. Vinson, et al. found that jurors assess higher auditor liability when a critical audit matter is removed than when a critical audit matter is disclosed for more than one year^[19]. Since critical audit matters are chosen from matters that have been communicated to the audit committee, some research also focus CAM problem on audit committee's judgment and decision making. Kang claimed that driving by perceived greater oversight duty, audit committee members' are more intend to ask challenging questions when critical audit matters are required^[21].

Most studies use experimental method to discover the effect of audit matters. This is because of limited data. In American, the earliest data for critical audit matters is contained in financial report of 2019. In 2017, some companies voluntarily disclose critical audit matters in financial report in China. In 2018, critical audit matters are compulsive demand in Chinese capital market. Therefore it is appropriate for us to investigate the problem of critical audit matters by using Chinese capital market data.

2.2 Hypothesis Development

Confirmation bias theory indicates that people tend to interpret or seek out evidence in ways that are consistent with their existing beliefs, expectation or hypothesizes and to ignore or overlook information that is supposed inconsistent^{[9, 10]}. In fact, this preference is considered as a kind of biased and unconscious thinking. Confirmation bias widely exists in the judgment and decision making process. A simple example of confirmation bias is that when examining whether a certain person is honest or not, people will interpret evidence as pointing to the answer they think is most likely, no matter what the truth is.

As far as the auditing is concerned, auditor's existing belief is that they should not be blamed. According to confirmation bias theory, this belief will be reinforced when appropriate fact or evidence appears. In our paper, the additional fact is the disclosure of critical audit matters. There are two reasons why critical audit matters can reinforce auditor's existing belief.

Firstly, providing more information to users can increase auditor's belief that they have provided enough risk information to users and therefore, there is little ground that they should be attributed. In 2008 recession, a furor at auditors arose when financial statements users were informed that great financial flops such as AIG, Citigroup, Washington Mutual and Lehman Brothers had received clean opinions from their auditors while these companies were actually being on the brink of collapse. The core of the questioning and condemnation of auditors is the failure to provide adequate information and 2008 recession directly led to the rethink of traditional audit report. With the new development of audit report, this situation improved. Christensen, Glover, and Wolfe found that audit opinion with critical audit matters is more informative^[2]. Providing more internal information, CAMs can reduce the information gap between managements and investors. As CAMs are eventually enacted, investors can get information they have long been asking auditors to provide and get the ability to look inside the companies. In this case, auditor has reasonable grounds for thinking that the attribution of irrational decisions should shifts from auditors to investors themselves and the more CAMs provided, the stronger the belief will be. Therefore, when more CAMs are detailed in the audit report, auditor's perceived audit risk will be lower.

In addition, CAMs provide significant textual evidence that auditors can use to defense themselves if misstatements have been detected. Auditors must keep the companies' internal information confidential. Therefore, when audit reports only had two options (pass or fail), the information provided in audit report was generally boilerplate and there was little evidence can auditor use to defense themselves if they have been accused of missing their duty. However, CAMs provide a promising opportunity for auditor to clear themselves. Since CAMs refer to the most subjective and challenging matters in the auditing process, CAMs actually inform users the matters keeping auditors awake all day. In this way, users can see inside the mind of auditors and in that case, auditors can use CAMs to defense themselves that they have done what they can and already have undiscovered the uncertainty and risk information to investors. Since the source of risk perceived by auditors has been decreased, perceived audit risk will be lower if CAMs are discovered in the audit report. Further, the greater number of critical audit matters, the more defense power and the lower perceived audit risk.

Numerous studies have made it clear that there is a causal relationship between audit opinion and audit risk. As a predictive tool, the audit risk model guides auditors to express a clear opinion when auditors decrease audit risk to a tolerance interval^{[12, 13]}. Practically, the auditor may issue an unqualified opinion when their perceived audit risk is low rather than there are little material errors or misrepresentations^[22].

Accordingly, we expect that the more CAMs auditors have provided, the lower perceived audit risk, furthermore, the more likely auditor will disclose a clear audit opinion. We state our first hypothesis as following:

Hypothesis 1 Auditor is more likely to issue unqualified audit opinion when more CAMs are disclosed in the audit report than less CAMs are included.

Psychology research suggests that people's control ability of situation lead them to be more cautious. Bar-Tal, et al. found that only individuals who doubt their ability to control the situation will actively monitor their own behavior^[11]. Based on the logical chain we have claimed earlier, we assume that the number of CAMs decrease auditors' perceived audit risk and then influence the audit opinion. However, since high audit risk may lead to less control of situation and therefor actively influence auditors' behavior, we assume that the real audit risk may moderate the correlation between the number of critical audit matters and audit opinion. To be more precisely: When the real audit risk is high, auditors ponder whether they are in control of the situation and ponder their own ability to avoid responsibility. Based on the psychology theory, auditors more tend to be cautious and actively monitor their own behavior when audit risk is high. Therefore, in respect of auditors in high audit risk, cautiousness of their own behavior may result in accurate self-awareness of biases. If so, the positive correlation between the number of critical audit matters and the audit opinion should be restricted by high audit risk. In sum, we propose our second hypothesis as following:

Hypothesis 2 The impact of CAMs on the audit opinion is less significant when companies' audit risk is higher than when companies' audit risk is lower.

3 Research Design

3.1 Sample Selection

While CAMs data are manually collected from the audit reports from fiscal 2016 year to 2018 year, we obtain all other data from China Stock Market and Accounting Research (CSMAR) database. Our sample begins in 2016. This is because 2016 is the first year when data on critical audit matters were available. Even though in China, annual report of 2017 is required to include critical audit matters, many firms voluntarily implement the new auditing standard in annual report of 2016. We include all firms listed on the Shenzhen and Shanghai Stock Exchanges from 2016

\sim

2018. Our final sample consists of 6, 662 firm-year observations. We exclude firms belongs to the financial industry not only because the characteristics of financial industry are different from others, but also there is no possibility for them to develop some real earnings management variable. In addition, we exclude firms whose financial data are missing. All the continuous variables are winsorized at 1% and 99% to avoid the effect of outliers.

3.2 Measurement of All Variables

Our main independent variable of interest is denoted CAM_num

_{i, t}

and is measured as the number of critical audit matters disclosed in the auditing report of firm

i

in time period

t

. Our independent variable is audit opinion. Audit opinion is denoted AuditOpin

_{i, t}

and defined as one if audit opinion is a clear opinion and otherwise AuditOpin

_{i, t}

is measured as zero.

Another variable of interest is denoted AuditRisk

_{i, t}

. We use different ways to measure audit risk. Firstly, based on the current research^{[23, 24]}, we define AuditRisk_accrual

_{i, t}

as 1 if accrual earnings management is higher than average accrual earnings management for company

i

in time period

t

and otherwise it will be measured as 0; secondly, based on the research of Roychowdhury^[25], we define AuditRisk_real

_{i, t}

as 1 if real earnings management is higher than average real earnings management for company

i

in time period

t

and otherwise AuditRisk_real

_{i, t}

will be measured as 0; thirdly, we use the willingness of avoiding loss to measure audit risk. According to Dechow, Sloan, Sweeney^[23], companies with low positive earnings might be the result of earnings management. Therefore, we define Avoidloss

_{i, t}

as 1 if the earnings of firm

i

in time period

t

is higher than zero and less than 0.5^{[26, 27]}.

To be more specially, following Jones^[24], accrual earnings management is measured as the residual of Equation (1). To estimate the model, we run a cross-sectional regression for every industry-year.

\frac{{T A}_{i, t}}{{Assets}_{i, t - 1}} = α_{1} \frac{1}{{Assets}_{i, t - 1}} + α_{2} \frac{{P P E}_{i, t}}{{Assets}_{i, t - 1}} + α_{3} \frac{Δ {Sales}_{i, t}}{{Assets}_{i, t - 1}} + ε_{i, t},

(1)

where

{TA}_{i, t}

= total accruals for company

i

at year

t

; this variable is measured as the difference between income before extraordinary items and cash flows of operation.

{PPE}_{i, t}

= gross property, plant, and equipment for company

i

at year

t

Δ {Sales}_{i, t}

= sales revenue for company

i

at year

t

minus sales revenue for company

i

at year

t - 1

{Assets}_{i, t}

= total assets for company

i

at year

t

When it comes to real earnings management (REM), we cite Roychowdhury. Roychowdhury claims that real earnings management has three aspects: Abnormal cash flow from operation, abnormal discretionary expenses and abnormal production costs^[25].

We use the residual of Equation (2) to measure the real earnings management comes from cash flow from operation and we multiply the residuals of Equation (2) by negative one (

- 1

) to make the proxy an increasing function of REM. To estimate the model, we run a cross-sectional regression for every industry-year.

Abnormal cash flows from operation:

\frac{{CFO}_{i, t}}{{Assets}_{i, t - 1}} = α_{1} \frac{1}{{Assets}_{i, t - 1}} + α_{2} \frac{{Sales}_{i, t}}{{Assets}_{i, t - 1}} + α_{3} \frac{Δ {Sales}_{i, t}}{{Assets}_{i, t - 1}} + ε_{i, t},

(2)

where,

{CFO}_{i, t} =

cash flows from operation for company

i

in year

t

;

{Assets}_{i, t - 1} =

total assets for company

i

at the end of year

t - 1

;

{Sales}_{i, t} =

sales revenue for company

i

in year

t

;

Δ {Sales}_{i, t}

= sales revenue for company

i

in year

t

minus sales revenue for company

i

in year

t - 1

Other variables are the same as in Equation (1).

Managers can decrease several expenditures, such as research and developments (R & D), advertising, and selling, general and administrative expenses (SG & A), to increase the reported earnings in the current year. We use the residuals of Equation (3) as a proxy of abnormal discretionary expenditures and we multiply the residuals of Equation (3) by negative one (

- 1

) to make the proxy an increasing function of REM.

Abnormal discretionary expenses:

\frac{{DISE}_{i, t}}{{Assets}_{i, t - 1}} = α_{1} \frac{1}{{Assets}_{i, t - 1}} + α_{2} \frac{{Sales}_{i, t - 1}}{{Assets}_{i, t - 1}} + ε_{i, t},

(3)

where,

{DISE}_{i, t}

= discretionary expenses for company

i

in year

t

; discretionary expenses is measured as the sum of advertising, R & D and SG & A expenses;

{Sales}_{i, t - 1}

= sales revenue for company

i

in year

t - 1

;

Other variables are the same as in Equation (1).

Firms can produce more goods than necessary. Since fixed costs cannot be changed, increased production lowers fixed costs per unit. This leads to the decrease of cost and the increase of the earnings. Therefore, abnormal production cost is also a significant part of real earnings management. We use the residuals of Equation (4) to measure abnormal production costs.

Abnormal production costs:

\frac{{P R O D}_{i, t}}{{T A}_{i, t - 1}} = α_{1} \frac{1}{{T A}_{i, t - 1}} + α_{2} \frac{{Sales}_{i, t}}{{T A}_{i, t - 1}} + α_{3} \frac{Δ {Sales}_{i, t}}{{T A}_{i, t - 1}} + α_{4} \frac{Δ {Sales}_{i, t - 1}}{{T A}_{i, t - 1}} + ε_{i, t},

(4)

where,

{PROD}_{i, t}

= the sum of cost of goods sold and change in inventory for company

i

in year

t

Other variables are the same as in Equation (1) and Equation (2).

Abnormal levels of cash flows from operation (REM_CFO), discretionary expenses (REM_DISE) and production costs (REM_PROD) can be measured as the residual from Equations (2)

\sim

(4). We use Equation (5) to measure real earnings management:

Real earnings management:

{R E M}_{i, t} = | {R E M}_{-} {P R O D}_{i, t} - {R E M}_{-} {C F O}_{i, t} - {R E M}_{-} {D I S E}_{i, t} |,

(5)

where,

{REM_PROD}_{i, t}

= abnormal production costs; measured by the residual of Equation (2);

{REM_CFO}_{i, t}

= abnormal cash flows from operations; measured by the residual of Equation (3);

{REM_DISE}_{i, t}

= abnormal discretionary expenses; measured by the residual of Equation (4);

3.3 Research Model

The main objective of this study is to analyze the effect of the number of critical audit matters on audit opinion. Further, we investigate the effect of audit risk on the association of the number of critical audit matters and audit opinion. We use Equation (6) to test our Hypothesis 1:

{AuditOpin}_{i, t} = α_{0} + α_{1} {CAM_num}_{i, t} + β Controls + ε_{i, t},

(6)

where,

{AuditOpin}_{i, t}

= audit opinion; this is an indicator variable. When there is a clear opinion of firm

i

in time period

t

, AuditOpin is 1, otherwise equals 0.

{CAM_num}_{i, t}

= the number of critical audit matters,

Control variables are listed in Appendix A.

We test our Hypothesis 1 by the significance and direction of

α_{1}

in Equation (6). Since

{AuditOpin}_{i, t}

is an indicator variable and when there is a clear opinion of firm

i

in time period

t

, it equals 1, otherwise equals 0, significantly positive

α_{1}

suggests that the greater number of critical audit matters, the more likely a clear opinion. We use Equation (7) to test our Hypothesis 2. The controls variables in Equation (6) and Equation (7) are the same.

\begin{aligned} {AuditOpin}_{i, t} = & α_{0} + α_{1} {CAM_num}_{i, t} + α_{2} {AuditRisk}_{i, t} \\ + α_{3} {CAM_num}_{i, t} \times {AuditRisk}_{i, t} + β Controls + ε_{i, t}, \end{aligned}

(7)

where,

{AuditRisk}_{i, t}

= audit risk; the variable is measured by AuditRisk_accrual, AuditRisk_real and AuditRisk_avoidloss, respectively.

AuditRisk_accrual, AuditRisk_real, AuditRisk_avoidloss and control variables are listed in Appendix A.

We test our Hypothesis 2 by the significance and direction of

α_{3}

in Equation (7). If

α_{3}

is significantly negative, audit risk restricts the effect of the number of critical audit matters on the audit opinion; otherwise audit risk strengthen or has no effect on the association between the number of critical audit matters and the audit opinion.

In order to strengthen our conclusion, we include fixed effects, individual clustering and random effects in our models, respectively. In addition, we control for year and industry effect in all regressions to mitigate the effect of industries and year on results.

4 Empirical Results

4.1 Descriptive Statistics of Audit Opinion

Table 1 and Table 2 describe the characteristics of critical audit matters from 2016 to 2018. We count the number of companies with or without critical audit matters and present the distribution by year in Table 1. In Table 2, Panel A and Panel B state the audit opinion of companies with critical audit matters and present the distribution by industry and year, respectively.

Table 1 The number of company with or without critical audit matters

The number of company with or without critical audit matters (distributed by year)
Year	The number of companies without critical audit matters	The number of companies with critical audit matters	Total
2016	1, 922	42	1, 964
2017	6	2, 174	2, 180
2018	12	2, 506	2, 518
Total	1, 940	4, 722	6, 662

Table 2 Audit opinion of companies with critical audit matters

Panel A
Audit opinion fof companies with critical audit matters (presents the distribution by industry)
Industry	Qualified opinion	Unqualified opinion		Total
Agriculture, Forestry, Animal Husbandry and Fishery Industry	6	42		48
Extractive Industry	6	75		81
Manufacturing Industry	70	3, 141		3, 211
Production and Supply of Electricity, Gas and Water	2	101		103
Construction Industry	0	118		118
Wholesale and Retail Trade	6	202		208
Transportation, Warehousing and Postal Services	1	94		95
Hotel and Catering Sectors	0	14		14
Information Transmission, Software and Information Technology Services	15	377		392
Real Estate	3	146		149
Leasing and Business Service	4	59		64
Scientific Research and Technical Services	0	51		51
Water, Environment and Utilities Management	2	66		68
Residential Services, Repairs and Other Services	0	1		1
Education	0	13		13
Health and Social Work	0	20		20
Culture, Sport and Entertainment	5	60		65
Comprehensive Industry	0	21		21
Total	121	4, 601		4, 722
Panel B
Audit opinion of companies with critical audit matters (presents the distribution by year)
Audit opinion type	2016	2017	2018	Total
Qualified opinion	1	47	73	121
Unqualified opinion	41	2, 127	2, 433	4, 601
Total	42	2, 174	2, 506	4, 722

Table 1 indicates that in 2016, only 42 companies included CAMs in their auditing report. However, in 2017 and 2018, only 6 and 12 companies did not include CAMs in their auditing report, respectively. This is because companies voluntarily disclose critical audit matters in 2016 but after 2017, critical audit matters are required to be disclosed in the financial statement.

In Table 2, Panel A indicates that 4, 601 samples received clear opinion while 121 samples did not. Panel B shows that in 2016, only forty-one companies received a clear opinion; in 2017, 2, 127 companies received a clear opinion and in 2018 this number increase to 2, 433. Panel B also states that in 2016, only one company did not receive a clear opinion; in 2017, 47 companies did not receive a clear opinion and in 2018 this number increase to 73. From Table 2, we can see that in the samples with critical audit matters, audit opinion type vary widely. The differences between audit opinions provide us the opportunity to examine the association between the audit opinion and the critical audit matters.

Table 3 presents statistics results of our main variables. Table 3 indicates that the average (median) number of critical audit matters in their auditing report is 0.7853 (1.0986) with a maximum of 1.9459 in a single firm and 97.48% companies received a clear audit opinion in our whole sample. In our sample, 4.35% of the firms employed a Big4 audit firm. Definitions of these variables are listed in Appendix I.

Table 3 Descriptive statistics of main variables

Variables	$N$	Mean	Sd	Med	Min	Max
AuditOpin	6, 662	0.9748	0.1568	1	0	1
CAM_num	6, 662	0.7853	0.5347	1.0986	0	1.9459
Size	6, 662	22.2465	1.2390	22.0952	19.9607	26.1503
Lev	6, 662	0.4067	0.1952	0.3967	0.0603	0.8869
Growth	6, 662	0.2338	0.4619	0.1489	-0.5635	3.39
ROA	6, 662	0.0395	0.0591	0.0385	-0.2545	0.1872
AuditCox_rec	6, 662	0.1355	0.1063	0.1173	0.0005	0.4887
AuditCox_inv	6, 662	0.1356	0.1185	0.1084	0.0003	0.6534
Avoidloss	6, 662	0.0940	0.2918	0	0	1
Equity	6, 662	1.9783	1.0694	1.6553	1.061	8.2399
Turnover	6, 662	0.5802	0.3765	0.4929	0.0672	2.3589
Quick	6, 662	1.9147	1.8847	1.3077	0.2024	12.0328
Top1	6, 662	32.3503	13.4583	30.3800	9.5600	73.7000
Dual	6, 662	0.3134	0.4639	0	0	1
Dirpct	6, 662	0.3766	0.0532	0.3636	0.3333	0.5714
Dirsize	6, 662	16.0032	3.4432	17.2988	7.0901	20.6119
Salary	6, 662	14.4823	0.6526	14.4440	12.9820	16.3514
Committee	6, 662	3.9437	0.3944	4	2	5
Big4	6, 662	0.0435	0.2041	0	0	1
Seperation	6, 662	5.1171	7.5557	0	0	29.4550
AuditRisk_accrual	6, 662	0.0555	0.0576	0	0	1
AuditRisk_real	6, 662	0.1568	0.1680	0	0	1

4.2 Main Results

Table 4 reports the main results from the equation 6 which is used to test the association between the number of critical audit matters and audit opinion. Specially, column 1 reports the regression results using least square method (OLS); column 2 indicates the regression results using panel model (Panel); column 3 shows the regression results using random-effect panel model (Panel_RE); column 4 suggests the regression results using fixed-effect panel model (Panel_FE); column 5 manifests the regression results using individual-clustering and fixed-effects panel model (Panel_cluster).

Table 4 The regression between audit opinion and critical audit matters

Variables	(1)	(2)	(3)	(4)	(5)
Variables	OLS	Panel	Panel_RE	Panel_FE	Panel_cluster
CAM_num	0.0441***	0.0395***	0.0395***	0.0376***	0.0376**
	(2.9253)	(4.4163)	(4.4160)	(3.2825)	(2.1538)
Size	0.0100***	0.0101***	0.0101***	0.0096	0.0096
	(3.6068)	(4.0318)	(4.0318)	(0.7611)	(0.5839)
Lev	-0.0198	-0.0014	-0.0014	0.1081**	0.1080
	(-0.4670)	(-0.0520)	(-0.0520)	(2.0514)	(1.2059)
Growth	0.0056	0.0111***	0.0111***	0.0270***	0.0270***
	(1.0856)	(2.7683)	(2.7683)	(5.2788)	(3.3628)
ROA	0.6231***	0.6081***	0.6081***	0.5467***	0.5467***
	(7.6169)	(16.2430)	(16.2430)	(10.5371)	(5.6854)
AuditCox_rec	0.0387	0.0315	0.0315	0.0243	0.0243
	(1.5664)	(1.3814)	(1.3814)	(0.3505)	(0.2296)
AuditCox_inv	0.0860***	0.0965***	0.0965***	0.2806***	0.2806**
	(3.4785)	(4.1701)	(4.1701)	(4.3887)	(2.4581)
Avoidloss	0.0274***	0.0223***	0.0223***	0.0079	0.0079
	(3.4459)	(3.4495)	(3.4495)	(0.9736)	(0.8698)
Equity	-0.0191**	-0.0230***	-0.0230***	-0.0531***	-0.0531***
	(-2.3803)	(-6.0544)	(-6.0544)	(-7.4829)	(-2.8917)
Turnover	-0.0079	-0.0135**	-0.0135**	-0.1012***	-0.1012***
	(-1.0775)	(-2.0758)	(-2.0758)	(-5.3598)	(-2.6515)
Quick	-0.0008	-0.0000	-0.0000	0.0038	0.0038
	(-0.5240)	(-0.0543)	(-0.0543)	(1.2558)	(1.3637)
Top1	0.0002*	0.0003*	0.0003*	0.0008	0.0008
	(1.6753)	(1.8035)	(1.8035)	(1.0478)	(1.0018)
Dual	0.0030	0.0033	0.0033	0.0077	0.0077
	(0.7326)	(0.7245)	(0.7245)	(0.8433)	(0.5820)
Dirpct	-0.0436	-0.0417	-0.0417	-0.0147	-0.0147
	(-1.2443)	(-1.0725)	(-1.0725)	(-0.1819)	(-0.1448)
Dirsize	-0.0001	-0.0002	-0.0002	-0.0037	-0.0037
	(-0.1611)	(-0.3265)	(-0.3265)	(-1.6009)	(-1.3578)
Salary	-0.0024	-0.0013	-0.0013	0.0100	0.0100
	(-0.6440)	(-0.3317)	(-0.3317)	(1.1288)	(0.9730)
Committee	0.0030	0.0039	0.0039	-0.0003	-0.0003
	(0.6190)	(0.7546)	(0.7546)	(-0.0281)	(-0.0411)
Big4	0.0088	0.0081	0.0081	0.0058	0.0058
	(1.6157)	(0.7335)	(0.7335)	(0.1705)	(0.5497)
Seperation	0.000130	0.0000	0.0000	0.0003	0.0003
	(0.5827)	(0.3055)	(0.3055)	(0.4202)	(0.4264)
Cons	0.7151***	0.6983***	0.6983***	0.7057**	0.7056*
	(8.9884)	(10.1712)	(10.1712)	(2.4392)	(1.9218)
Random-effect	No	No	Yes	No	No
Fixed-effect	No	No	No	Yes	Yes
Individual-cluster	No	No	No	No	Yes
Industry	yes	yes	yes	yes	yes
Year	yes	yes	yes	yes	yes
Obs	6, 662	6, 662	6, 662	6, 662	6, 662
Adj $R^{2}$ /Within $R^{2}$	0.0930	0.0798	0.0534	0.0533	0.0833

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Hypothesis 1 is focus on the relationship between audit opinion and the critical audit matters. A review of Table 4 columns 1

\sim

5 indicates that the coefficient on CAM_num is positive and significant. The

t

-value of coefficient on CAM_num are respectively 2.9253 in column 1, 4.4163 in column 2, 4.4160 in column 3, 3.2825 in column 4, 2.1538 in column 5. This result shows that there is a significantly positive relationship between the number of critical audit matters and clear audit opinion. This result is consistent with the theory and Hypothesis 1.

4.3 Moderating Effect of Audit Risk

To analysis the moderating effect of audit risk, we show the regressing results of Equation (7) using different proxy variable of audit risk. Specially, Table 5 reports the regression results using the likelihood of loss aversion as audit risk (audit risk refers to variables Avoidloss); Table 6 shows the regression results using real earnings management (audit risk refers to variable AuditRisk_real); Table 7 indicates the regression results using accrual earnings management (audit risk refers to variable AuditRisk_accrual). In each table we also show the regression results that distinguish high and low audit risk.

Table 5 The moderating effect of audit risk (audit risk refers to the degree of loss aversion

	(1)	(2)	(3)
		High audit risk	Low audit risk
Variables		Avoidloss equals one	Avoidloss equals zero
CAM_num	0.4997***	0.1027	0.0275**
	(2.8319)	(1.5541)	(2.2697)
CAM_num $\times$ Avoidloss	-0.1604***
	(-4.1066)
Avoidloss	0.1220
	(0.9963)
Size	0.1482***	-0.1130	0.0103
	(3.1654)	(-1.2900)	(0.7571)
Lev	-1.2706***	0.2190	0.1138**
	(-2.8371)	(0.5229)	(2.0700)
Growth	0.1238	0.0542**	0.0240***
	(1.3313)	(2.0413)	(4.3050)
ROA	3.8060***	7.2940	0.5748***
	(6.7750)	(1.4907)	(10.6703)
AuditCox_rec	0.0941	-0.9424*	-0.0032
	(0.2342)	(-1.9202)	(-0.0433)
AuditCox_inv	1.2376***	-1.0912**	0.3232***
	(2.9426)	(-2.4040)	(4.7146)
Equity	-0.0539	-0.0315	-0.0522***
	(-1.1672)	(-0.7373)	(-6.7887)
Turnover	0.0119	-0.3500***	-0.0681***
	(0.1069)	(-4.2078)	(-3.2202)
Quick	0.0280	-0.0009	0.0051
	(0.6531)	(-0.0416)	(1.5954)
Top1	0.0077**	0.0030	0.0008
	(2.2915)	(0.5148)	(0.9586)
Dual	0.0373	0.0145	0.0115
	(0.4111)	(0.2577)	(1.1948)
Dirpct	-0.4955	0.1100	0.0052
	(-0.6740)	(0.2709)	(0.0602)
Dirsize	0.0053	-0.0050	-0.0022
	(0.4267)	(-0.4296)	(-0.8656)
Salary	0.0554	0.0075	0.0141
	(0.7874)	(0.1334)	(1.5007)
Committee	-0.0056	0.0233	0.0046
	(-0.0547)	(0.4870)	(0.3895)
Big4	-0.5492	-0.0352	0.0015
	(-1.4772)	(-0.1641)	(0.0417)
Seperation	0.0018	0.0014	0.0006
	(0.3066)	(0.3253)	(0.7267)
Cons	-2.4638**	3.5945*	0.5474*
	(-2.0045)	(1.6726)	(1.7589)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	626	6, 036
Adj $R^{2}$ /Within $R^{2}$	0.0795	0.0474	0.0612

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Table 6 The moderating effect of audit risk (audit risk refers to variable AuditRisk_real)

Variables	(1)	(2)	(3)
		High audit risk	Low audit risk
		AuditRisk_real is higher than the average	AuditRisk_real is Lower than the average
CAM_num	0.0513***	0.0137	0.0409**
	(4.9405)	(0.5818)	(2.2060)
CAM_num $\times$ AuditRisk_real	-0.1372***
	(-3.4669)
AuditRisk_real	-0.0524***
	(-3.6410)
Size	0.0113***	0.0358	-0.0150
	(3.8051)	(1.4356)	(-0.5762)
Lev	-0.0126	0.0272	0.3357***
	(-0.4147)	(0.2613)	(3.2950)
Growth	0.0159***	0.0280***	0.0189
	(3.4017)	(3.0866)	(1.4259)
ROA	0.5705***	0.7130***	0.5608***
	(13.6627)	(7.4312)	(6.2089)
AuditCox_rec	0.0081	-0.1950	0.0954
	(0.2975)	(-1.4943)	(0.6625)
AuditCox_inv	0.1121***	0.2630**	0.2400
	(4.1823)	(2.3252)	(1.5862)
Avoidloss	0.0217***	0.0172	-0.0084
	(3.0778)	(1.1239)	(-0.5950)
Equity	-0.0231***	-0.0300**	-0.0978***
	(-5.7293)	(-2.5672)	(-6.7868)
Turnover	-0.0117	-0.0604*	-0.0342
	(-1.5642)	(-1.8081)	(-0.7868)
Quick	0.0006	0.0006	0.0059
	(0.3051)	(0.0780)	(0.9957)
Top1	0.0003*	-0.0010	0.0043***
	(1.7438)	(-0.7088)	(2.5944)
Dual	0.0023	0.0183	-0.0138
	(0.4322)	(1.0654)	(-0.8662)
Dirpct	-0.0354	-0.1750	-0.0502
	(-0.7778)	(-1.1179)	(-0.3290)
Dirsize	-0.0003	-0.0024	-0.0048
	(-0.3842)	(-0.4989)	(-1.2737)
Salary	-0.0002	-0.0272*	0.0226
	(-0.0559)	(-1.6697)	(1.3857)
Committee	0.0027	0.0073	-0.0173
	(0.4571)	(0.3235)	(-0.9056)
Big4	0.0074	-0.0401	0.0262
	(0.5887)	(-0.6084)	(0.4355)
Seperation	0.0001	0.0022	-0.0008
	(0.2563)	(1.4369)	(-0.5842)
Cons	0.6696***	0.7000	1.0218*
	(8.3811)	(1.2328)	(1.6491)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	3, 800	2, 862
Adj $R^{2}$ /Within $R^{2}$	0.0612	0.0230	0.0710

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Table 7 The moderating effect of audit risk (audit risk refers to variable AuditRisk_accrual)

Variables	(1)	(2)	(3)
		High audit risk	Low audit risk
		AuditRisk_accrual is higher than the average	AuditRisk_accrual is lower than the average
CAM_num	0.0466***	0.0255*	0.0856**
	(3.6573)	(1.7130)	(2.2717)
CAM_num $\times$ AuditRisk_accrual	-0.0156*
	(-1.6596)
AuditRisk_accrual	0.0019
	(0.1870)
Size	0.0128	-0.0106	0.0176
	(0.8801)	(-0.5663)	(0.4245)
Lev	0.0742	0.1541*	0.0480
	(1.2645)	(1.9497)	(0.3358)
Growth	0.0248***	0.0228***	0.0017
	(4.2426)	(2.8606)	(0.1407)
ROA	0.5385***	0.5433***	1.0123***
	(9.7300)	(8.0217)	(4.1088)
AuditCox_rec	0.1001	0.3264***	-0.2150
	(1.2755)	(2.8930)	(-1.2575)
AuditCox_inv	0.2643***	0.3512***	0.4714***
	(3.6707)	(3.6067)	(3.1580)
Avoidloss	0.0062	-0.0070	0.0023
	(0.7063)	(-0.6482)	(0.0833)
Equity	-0.0516***	-0.0586***	-0.0275
	(-7.1350)	(-6.0522)	(-1.3548)
Turnover	-0.1133***	-0.1844***	-0.0736
	(-5.5054)	(-7.0867)	(-1.3981)
Quick	0.0031	0.0030	-0.0016
	(0.8199)	(0.6286)	(-0.1986)
Top1	0.0006	0.0007	-0.0023
	(0.8293)	(0.7410)	(-0.8898)
Dual	0.0064	0.0123	0.0025
	(0.6419)	(0.9915)	(0.0869)
Dirpct	-0.0316	0.0043	0.3240
	(-0.3390)	(0.0371)	(1.3341)
Dirsize	-0.0048*	-0.0068**	-0.0083
	(-1.8883)	(-2.1957)	(-1.1767)
Salary	0.0078	0.0228*	-0.0058
	(0.8017)	(1.9168)	(-0.2169)
Committee	-0.0038	0.0040	-0.0133
	(-0.3094)	(0.2591)	(-0.3803)
Big4	-0.0010	0.0016	-0.0235
	(-0.0278)	(0.0371)	(-0.1468)
Seperation	0.0002	0.0008	-0.0027
	(0.2525)	(0.7497)	(-1.4104)
Cons	0.7208**	0.9944**	0.8210
	(2.1624)	(2.3737)	(0.8106)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	4, 297	2, 365
Adj $R^{2}$ /Within $R^{2}$	0.0514	0.0650	0.0318

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

In Table 5, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's loss aversion of earnings is higher (when variable Avoidless equals one); column 3 indicates the regression results when firm's financial earnings loss aversion is lower (when variable Avoidless equals zero).

Table 5 column 1 indicates that the coefficient of Cam_num

\times

Avoidloss is negative and significant (

t

-statistics

= - 4.1066

). This result is consistent with the hypothesis that when audit risk is higher, the effect of number of critical audit number on auditing opinion decreases. Column 2 shows that when firm's loss aversion of earnings is higher, the coefficient of CAM_num is insignificant (

t

-statistics = 1.5541). This result suggests that when audit risk is higher, there is an insignificant relationship between the number of critical audit matters and auditing opinion; on the contrary, column 3 indicates that when firm's loss aversion of earnings is lower, the coefficient of CAM_num is significant (

t

-statistics = 2.2697). This result suggests that when audit risk is lower, there is a significant relationship between the number of critical audit matters and auditing opinion. According to Table 5, when audit risk is higher, auditor is less likely to be affected by the disclosure of critical audit matters. Therefore, we find evidence in favor of Hypothesis 2.

In Table 6, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's real earnings management is higher than the average real earnings management (when variable AuditRisk_real is higher than the average); column 3 indicates the regression results when firm's real earnings management is lower than the average real earnings management (when variable AuditRisk_real is lower than the average).

Table 6 column 1 indicates that the coefficient of Cam_num

\times

AuditRisk_real is negative and significant (

t

-statistics =

- 3.4669

). This result is consistent with the hypothesis that when audit risk is higher, the effect of the number of critical audit number on auditing opinion decreases. Column 2 shows that when firm's real earnings management is higher than the average real earnings management, the coefficient of CAM_num is insignificant (

t

-statistics = 0.5818). This result suggests that when audit risk is higher, there is an insignificant relationship between the number of critical audit matters and auditing opinion; on the contrary, column 3 indicates that when firm's loss aversion of earnings is lower, the coefficient of CAM_num is significant (

t

-statistics = 2.2060). This result suggests that when audit risk is lower, there is a significant relationship between the number of critical audit matters and auditing opinion. This result is consistent with the theory that when audit risk is higher, auditor is less affected by the disclosure of critical audit matters. We find evidence in favor of Hypothesis 2 in Table 6.

In Table 7, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's accrual earnings management is higher than the average accrual earnings management (when variable AuditRisk_accrual is higher than the average); column 3 indicates the regression results when firm's real earnings management is lower than the average real earnings management (when variable AuditRisk_accrual is lower than the average).

Table 7 Column 2 shows that when firm's accrual earnings management is higher than the average accrual earnings management, the coefficient of CAM_num is significant (

t

-statistics = 1.7130). Column 3 indicates that when firm's accrual earnings management is lower than the average accrual earnings management, the coefficient of CAM_num is significant (

t

-statistics = 2.2717). These results suggest there is a significant relationship between the number of critical audit matters and auditing opinion in both high and low audit risk. However, column 1 indicates that the coefficient of CAM_num

\times

AuditRisk_accrual is negative and significant (

t

-statistics =

-

1.6596). Since we expect that audit risk restricts the relationship between the number of critical audit matters and the audit opinion rather than expecting audit risk eliminates the relationship between these two variables, the negative coefficient of CAM_num

\times

AuditRisk_accrual is consistent with the hypothesis that when audit risk is higher, the effect of the number of critical audit number on auditing opinion decreases. We still find evidence in favor of Hypothesis 2 in Table 7.

5 Additional Analyses

5.1 Moderating Effect of Audit Complexity

High audit complexity has a negative effect on the auditors' evaluation process. To be more specifically, since companies with high audit complexity have more complicated control systems, more subsidiaries and more complicated business model, the evaluation process performed by auditors become more difficult and audit risk increases. Based on the Hypothesis 2 that the audit risk moderates the relationship between the number of critical audit matters and the audit opinion, we expect that audit complexity can also moderate the relationship between the number of critical audit matters and the audit opinion. If this is true, we provide more evidence for Hypothesis 2.

In order to test whether audit complexity restricts the association between the number of critical audit matters and the audit opinion, we use equation 8 to test the relationship between audit complexity, the number of critical audit matters and the audit opinion:

\begin{aligned} {AuditOpin}_{i, t} = & α_{0} + α_{1} {CAM_num}_{i, t} + α_{2} {AuditCox}_{i, t} \\ + α_{3} {CAM_num}_{i, t} \times {AuditCox}_{i, t} + β Controls + ε_{i, t} . \end{aligned}

(8)

According to Simunic^[12], audit complexity denoted by AuditCox and measured by two variables (AuditCox_rec and AuditCox_rec). Variable AuditCox_rec refers to the ratio of accounts receivable to total assets and variable AuditCox_rec refers to the ratio of inventory to total assets. Other variables are listed in Appendix A.

In Table 8, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's auditing is more complex than others (when variable AuditCox_rec is higher than the average); column 3 indicates the regression results when firm's auditing is less complex than others (when variable AuditCox_rec is lower than the average).

Table 8 The moderating effect of audit complexity (audit complexity refers to variable AuditCox_rec)

Variables	(1)	(2)	(3)
		High audit complexity	Low audit complexity
		AuditCox_rec is higher than the average	AuditCox_rec is lower than the average
CAM_num	0.0542***	0.0223	0.0409***
	(5.2335)	(0.6508)	(2.7874)
CAM_num $\times$ AuditCox_rec	-0.1604***
	(-4.1066)
AuditCox_rec	0.0212
	(0.7845)
Size	0.0119***	-0.0026	0.0256
	(3.9738)	(-0.0777)	(1.3707)
Lev	-0.0128	0.0898	0.1565**
	(-0.4225)	(0.5155)	(2.1584)
Growth	0.0136***	0.0328**	0.0244***
	(2.9281)	(2.3821)	(3.5225)
ROA	0.5467***	0.5036***	0.6459***
	(13.2393)	(3.1133)	(9.0391)
AuditCox_inv	0.1062***	0.3361	0.2234***
	(3.9585)	(1.5031)	(2.7629)
Avoidloss	0.0211***	-0.0026	0.0194*
	(2.9851)	(-0.1973)	(1.8250)
Equity	-0.0234***	-0.0514	-0.0597***
	(-5.8111)	(-1.6199)	(-6.4711)
Turnover	-0.0154**	-0.1210**	-0.0487*
	(-2.0610)	(-2.0000)	(-1.7673)
Quick	0.0005	0.0077	0.0030
	(0.2453)	(1.3786)	(0.7901)
Top1	0.0003*	0.0035**	-0.0004
	(1.7427)	(2.0314)	(-0.4874)
Dual	0.0019	0.0171	-0.0056
	(0.3604)	(1.0919)	(-0.4567)
Dirpct	-0.0384	-0.0059	-0.0917
	(-0.8416)	(-0.0323)	(-0.8345)
Dirsize	-0.0004	-0.0006	-0.0082***
	(-0.4828)	(-0.1446)	(-2.7910)
Salary	-0.0010	0.0346	-0.0116
	(-0.2300)	(1.6039)	(-1.0285)
Committee	0.0029	-0.0090	-0.0002
	(0.4809)	(-0.9095)	(-0.0105)
Big4	0.0054	-0.0412	0.0142
	(0.4304)	(-0.8330)	(0.3703)
Seperation	0.0000	-0.0017	0.0009
	(0.1688)	(-1.4349)	(0.9168)
Cons	0.6668***	0.5315	0.7656*
	(8.3196)	(0.7627)	(1.8452)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	2, 877	3, 785
Adj $R^{2}$ / Within $R^{2}$	0.0756	0.0875	0.0622

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Table 8 column 1 indicates that the coefficient of CAM_num

\times

AuditCox_rec is negative and significant (

t

-statistics =

-

4.1066). This result provides evidence that when audit complexity is higher, the effect of the number of critical audit number on auditing opinion decreases. Column 2 shows that when firm's auditing is more complex, the coefficient of CAM_num is insignificant (

t

-statistics = 0.6508). This result suggests that when auditing complexity is higher, there is an insignificant relationship between the number of critical audit matters and auditing opinion. Column 3 indicates that when firm's auditing is less complex, the coefficient of CAM_num is significant (

t

-statistics = 2.7874). This result suggests that when auditing complexity is lower, there is a significant relationship between the number of critical audit matters and auditing opinion. Table 8 consistents with the result that audit complexity moderates the relationship between the number of critical audit matters and the audit opinion.

In Table 9, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's auditing is more complex than others (when variable AuditCox_inv is higher than the average); column 3 indicates the regression results when firm's auditing is less complex than others (when variable AuditCox_inv is lower than the average).

Table 9 The moderate effect of audit complexity (audit complexity refers to variable AuditCox_inv)

Variables	(1)	(2)	(3)
		High audit complexity	Low audit complexity
		AuditCox_inv is higher than the average	AuditCox_inv is lower than the average
CAM_num	0.0506***	0.0199	0.0571***
	(4.9830)	(0.7382)	(3.8659)
CAM_num $\times$ AuditCox_inv	-0.0605***
	(-4.1615)
AuditCox_inv	0.1088***
	(4.0483)
Size	0.0121***	0.0106	0.0101
	(4.0439)	(0.3927)	-0.5861
Lev	-0.0176	0.1912	0.1112
	(-0.5787)	(1.4358)	(1.6440)
Growth	0.0139***	0.0167*	0.0151**
	(2.9970)	(1.7215)	(2.1249)
ROA	0.6073***	0.4505***	0.4825***
	(15.1709)	(2.7117)	(7.4016)
AuditCox_rec	0.0098	-0.0593	0.1835*
	(0.3620)	(-0.5538)	(1.9170)
Avoidloss	0.0234***	0.0169	-0.0018
	(3.3231)	(1.1361)	(-0.1692)
Equity	-0.0230***	-0.0857***	-0.0590***
	(-5.7059)	(-2.9456)	(-6.0762)
Turnover	-0.0136*	-0.0965	-0.0747***
	(-1.8213)	(-1.5219)	(-2.7230)
Quick	0.0002	(0.0032)	0.0019
	(0.0872)	(-0.6680)	(0.5538)
Top1	0.0003*	0.0016	-0.0003
	(1.6991)	(1.3416)	(-0.2977)
Dual	0.0024	0.0008	-0.0026
	(0.4417)	(0.0423)	(-0.2064)
Dirpct	-0.0389	0.1166	-0.04546
	(-0.8513)	(0.8150)	(-0.4135)
Dirsize	-0.0004	-0.0025	-0.0076**
	(-0.4988)	(-0.8322)	(-2.4926)
Salary	-0.0007	-0.0386**	0.0248**
	(-0.1591)	(-2.3603)	(2.2154)
Committee	0.0027	0.0036	-0.0075
	(0.4575)	(0.4214)	(-0.4647)
Big4	0.0089	0.0326**	0.0017
	(0.6991)	(2.2813)	(0.0355)
Seperation	0.0000	0.0008	0.0005
	(0.2078)	(0.6365)	(0.4329)
Cons	0.6568***	1.3757**	0.6352
	(8.1848)	(2.1954)	(1.5879)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	2, 468	4, 194
Adj $R^{2}$ / Within $R^{2}$	0.0875	0.1091	0.0653

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Table 9 column 1 indicates that the coefficient of CAM_num

\times

AuditCox_inv is negative and significant (

t

-statistics =

-

4.1615). This result provides evidence that when audit complexity is higher, the effect of the number of critical audit number on auditing opinion decreases. Column 2 shows that when firm's auditing is more complex, the coefficient of CAM_num is insignificant (

t

-statistics = 0.7382). This result suggests that when auditing complexity is higher, there is an insignificant relationship between the number of critical audit matters and auditing opinion. Column 3 indicates that when firm's auditing is less complex, the coefficient of CAM_num is significant (

t

-statistics = 3.8659). This result suggests that when auditing complexity is lower, there is a significant relationship between the number of critical audit matters and auditing opinion. Audit complexity moderates the relationship between the number of critical audit matters and the audit opinion.

5.2 Moderating Effect of Big4

We further test whether there is a difference between Big4 account firm and Non-big4 account firm. This is necessary because studies show that auditors in Big4 account firm are more profession, more cautious and have more expertise than others^{[28, 29]}. Expertise benefits audit practice and prudence is good for reducing biased behavior. Since the critical audit matters lead to biased behavior of auditors, this phenomenon is less likely to happen in Big4. Therefore, we expect that the effect of critical audit matters is less likely to appear in Big4.

In order to test whether Big4 auditors are less influenced by critical audit matters, we use equation 9 to test the relationship between Big4, the number of critical audit matters and the audit opinion:

\begin{aligned} {AuditOpin}_{i, t} = & α_{0} + α_{1} {C A M}_{{num}_{i, t}} + α_{2} Big 4_{i, t} + α_{3} {CAM_num}_{i, t} \times Big 4_{i, t} \\ + β Controls + ε_{i, t} . \end{aligned}

(9)

Big4 refers to an indicator variable that equals one if a firm's external auditor belongs to a Big4 auditor, i.e., Deloitte & Touche (DT), PricewaterhouseCoopers (PWC), Ernst & Young (EY) and KPMG, and zero otherwise. Other variables are listed in Appendix A.

In Table 10, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's auditors came from Big4 (when variable Big4 equals one); column 3 indicates the regression results when firm's auditors came from non-big4 (when variable Big4 equals zero).

Table 10 The moderating effect of audit firm

Variables	(1)	(2)	(3)
Variables		Big4	Non-Big4
CAM_num	1.0369**	0.0001	0.0454***
	(2.0571)	(0.0081)	(3.5116)
CAM_num $\times$ Big4	-1.4234***
	(-3.3897)
Big4	-2.9055***
	(-2.6468)
Size	0.3071***	-0.1110***	0.0122
	(2.7040)	(-3.5277)	(0.9375)
Lev	-3.0947***	0.0971	0.1087**
	(-2.7684)	(0.6906)	(2.0036)
Growth	0.2659	0.0299**	0.0264***
	(0.8332)	(2.2868)	(4.9504)
ROA	8.1942***	0.3391**	0.5423***
	(7.5506)	(2.2473)	(10.1416)
AuditCox_rec	0.2273	-0.0488	0.0113
	(0.2273)	(-0.2784)	(0.1589)
AuditCox_inv	2.4181**	0.2430***	0.2838***
	(2.2455)	(2.9561)	(4.2275)
Avoidloss	0.2276	-0.0216*	0.0089
	(0.8163)	(-1.8283)	(1.0630)
Equity	-0.0579	-0.0157	-0.0541***
	(-0.6057)	(-1.1342)	(-7.3431)
Turnover	-0.0084	-0.3337***	-0.0851***
	(-0.0327)	(-15.9805)	(-4.2546)
Quick	0.0680	0.0077	0.0038
	(0.5725)	(0.5476)	(1.1933)
Top1	0.0165**	0.0007	0.0008
	(2.0737)	(0.6914)	(1.0507)
Dual	0.0292	-0.0107	0.0072
	(0.1461)	(-0.8891)	(0.7477)
Dirpct	-1.2911	0.1380	-0.0045
	(-0.9530)	(1.5859)	(-0.0526)
Dirsize	0.0159	0.0032	-0.0038
	(0.5176)	(0.6607)	(-1.5716)
Salary	0.1079	0.0206*	0.0110
	(0.6154)	(1.6883)	(1.1627)
Committee	0.0844	-0.0091	0.0001
	(0.3917)	(-0.9533)	(0.0095)
Seperation	0.0061	0.0006	0.0004
	(0.4735)	(0.4860)	(0.5146)
Cons	-5.5252**	3.4064***	0.6220**
	(-1.9675)	(4.9439)	(2.0617)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	290	6, 372
Adj $R^{2}$ / Within $R^{2}$	0.0307	0.0109	0.0530

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

Table 10 column 1 indicates that the coefficient of CAM_num

\times

Big4 is negative and significant (

t

-statistics =

-

3.3897). This result provides evidence that compared to non-big4, auditors came from big4 are less likely to be affected by the disclosure of critical audit matters during auditing process. Column 2 shows that when firm's auditors came from Big4 audit firm, the coefficient of CAM_num is insignificant (

t

-statistics = 0.0081). This result suggests that there is an insignificant relationship between the number of critical audit matters and auditing opinion in Big4. Column 3 indicates that in non-big4, the coefficient of CAM_num is significant (

t

-statistics = 3.5116). This result suggests that there is a significant relationship between the number of critical audit matters and auditing opinion in non-big4. Audit firm type moderates the relationship between the number of critical audit matters and the audit opinion.

5.3 Moderating Effect of Audit Committee

Current studies indicate that audit committee plays a critical role in monitoring management, helping auditors communicating with senior management and protecting minority shareholders' rights^[30]. According to Fama (1980), the board of directors and audit committee is a significant structure of the corporate governance. It is a useful design to supervise and control the behavior of the senior management. In addition, its core task is to coordinate the behavior of auditors and senior management. When it comes to our research question, audit committee can become a supervision structure on auditors, helping correcting auditors' biased behavior. Therefore, we expect that the effect of critical audit matters on audit opinion will diminish when the audit committee size is greater than the average number of audit committee members.

In order to test whether audit committee restricts the association between the number of critical audit matters and the audit opinion, we use equation 10 to test the relationship between audit committee size, the number of critical audit matters and the audit opinion:

\begin{aligned} {AuditOpin}_{i, t} = & α_{0} + α_{1} CAM_{num}_{i, t} + α_{2} {Committee}_{i, t} \\ + α_{3} {CAM_num}_{i, t} \times {Committee}_{i, t} + β Controls + ε_{i, t} . \end{aligned}

(10)

Variable Committee refers to the number of audit committee members. Other variables are listed in Appendix A.

In Table 11, column 1 shows the regression results of whole samples; column 2 reports the regression results when firm's audit committee size is greater than average (when variable committee is greater than average); column 3 indicates the regression results when firm's audit committee size is smaller than average (when variable committee is smaller than average).

Table 11 The moderating effect of audit committee

Variables	(1)	(2)	(3)
Variables		Committee is greater than average	Committee is smaller than average
CAM_num	0.0625***	-0.0227	0.0398***
	(3.6675)	(-0.6199)	(3.2826)
CAM_num $\times$ Committee	-0.2104***
	(-3.2229)
Committee	0.0019
	(0.3725)
Size	0.0112***	0.1355***	0.0026
	(3.5345)	(2.6737)	(0.1977)
Lev	-0.0302	-0.0006	0.0823
	(-0.6603)	(-0.0033)	(1.4772)
Growth	0.0095	0.1105***	0.0210***
	(1.6232)	-(5.5253)	(3.9483)
ROA	0.5638***	-0.0681	0.5516***
	(6.7995)	(-0.3326)	(10.1783)
AuditCox_rec	0.0307	0.2515	0.0269
	(1.0962)	(1.2082)	(0.3691)
AuditCox_inv	0.0967***	0.9181***	0.1692**
	(3.5107)	(5.2945)	(2.4420)
Avoidloss	0.0273***	0.0022	0.0088
	(3.2991)	(0.0798)	(1.0508)
Equity	-0.0191**	-0.0753***	-0.0510***
	(-2.3838)	(-3.1800)	(-6.8192)
Turnover	-0.0067	-0.2143***	-0.1040***
	(-0.8311)	(-3.0266)	(-5.2532)
Quick	-0.0002	0.0041	0.0038
	(-0.0996)	(0.2813)	(1.2189)
Top1	0.0003	0.0006	0.0007
	(1.5985)	(0.1709)	(0.9297)
Dual	0.0021	0.0735*	0.0007
	(0.4342)	(1.9615)	(0.0720)
Dirpct	-0.0380	-0.6241**	0.0160
	(-0.9724)	(-2.0353)	-0.1885
Dirsize	-0.0002	-0.0021	-0.0034
	(-0.2376)	(-0.2411)	(-1.4159)
Salary	-0.0022	0.0034	0.0081
	(-0.5254)	(0.0963)	(0.8733)
Big4	0.0058	-0.3811**	0.0064
	(0.9596)	(-2.0147)	(0.1845)
Seperation	0.0001	-0.0076*	0.0004
	(0.4729)	(-1.8574)	(0.5503)
Cons	0.6893***	-1.6922	0.8953***
	(7.8996)	(-1.3818)	(2.9968)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	6, 662	480	6, 182
Adj $R^{2}$ / Within $R^{2}$	0.0737	0.0305	0.0559

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

In Table 11, column 1 indicates that the coefficient of CAM_num

\times

Committee is negative and significant (

t

-statistics =

-

3.2229). This result provides evidence that compared to small audit committee, companies with bigger audit committee are less likely to be affected by the disclosure of critical audit matters during auditing process. Column 2 indicates that when audit committee size is greater than the average, the coefficient of CAM_num is insignificant (

t

-statistics =

-

0.6199). This result suggests that there is an insignificant relationship between the number of critical audit matters and auditing opinion in companies with greater audit committee; however, column 3 shows that when audit committee size is smaller than the average, the coefficient of CAM_num is significant (

t

-statistics = 3.2826). This result suggests that there is a significant relationship between the number of critical audit matters and auditing opinion in companies with greater audit committee. Audit committee moderates the relationship between the number of critical audit matters and the audit opinion.

6 Robustness Checks

We further use different models to test whether there is a positive and significant effect between the number of critical audit matters and audit opinion. Table 12 shows the results of different models. Column 1 and column 3 show the results without control variables. Column 1 and column 2 show the result using Logit model while column 3 and column 4 show the result using Poisson model.

Table 12 Robustness check (using different models)

Variables	(1)	(2)	(3)	(4)
Variables	Logit	Logit	Poisson	Poisson
CAM_num	1.2028**	1.0015**	0.0328**	0.0468***
	(2.1919)	(2.0243)	(1.9929)	(2.9466)
Size		0.2347**		0.0107***
		(2.1780)		(3.6699)
Lev		-2.8477***		-0.0109
		(-2.6241)		(-0.2308)
Growth		0.2190		0.0057
		(0.7098)		(1.0907)
ROA		9.7796***		0.6797***
		(8.6970)		(7.2462)
AuditCox_rec		0.4952		0.0418
		(0.5003)		(1.6234)
AuditCox_inv		2.4073**		0.0912***
		(2.3075)		(3.5361)
Avoidloss		0.1287		0.0306***
		(0.4685)		(3.6204)
Equity		-0.0625		-0.0223**
		(-0.6214)		(-2.3644)
Turnover		0.0284		-0.0092
		(0.1078)		(-1.1906)
Quick		0.0675		-0.00077
		(0.6281)		(-0.4902)
Top1		0.0187**		0.0002
		(2.3899)		(1.6117)
Dual		0.0686		0.0033
		(0.3492)		(0.7819)
Dirpct		-1.4839		-0.0459
		(-1.1203)		(-1.2614)
Dirsize		0.0268		-0.0002
		(0.8734)		(-0.2498)
Salary		0.1232		-0.0003
		(0.7066)		(-0.7580)
Committee		0.1470		0.0031
		(0.6878)		(0.6102)
Big4		1.5862		0.0086
		(1.5363)		(1.5339)
Seperation		0.0089		0.0001
		(0.6862)		(0.5569)
Cons	2.3419***	-4.6978*	-0.1208***	-0.3007***
	(6.3882)	(-1.6687)	(-2.8808)	(-3.5351)
Fixed-effect	yes	yes	yes	yes
Industry	yes	yes	yes	yes
Year	yes	yes	yes	yes
Obs	6, 662	6, 662	6, 662	6, 662
Adj $R^{2}$ / Within $R^{2}$	0.0600	0.0476	0.0544	0.0297

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

The results in Table 12 show that the coefficient of CAM_num is significantly positive and the

t

-statistics of column 1 to column 4 are 2.1919, 2.0243, 1.9929, 2.9466, respectively. Table 12 provides robust evidence to our hypothesis that the number of critical audit matters has significant positive effect on audit opinion.

We use 6, 662 firm-year observations for the period 2016–2018 to test our main hypothesis. However, the disclosure of critical audit matters is not compulsory in 2016. Since there is significant difference between voluntary disclosure and compulsory, there is a potential our results will change if we use different time period. Therefore, we regression the number of critical audit matters and audit opinion using 4698 firm-year observations for the period of 2017–2018. Table 13 shows that results. Table 13 column 1 shows the result using panel data; column 2 indicates the result using logistic model; column 3 suggests the result using Poisson model. According to Table 13, the coefficients of CAM_num are all positive and significant in three models. Specially, the

t

-statistics of panel model, logit model and poisson model are 2.7148, 2.6157 and 3.4572, respectively. This result consistents with our hypothesis that the more critical audit matters, the more likely a clear opinion.

Table 13 Robustness check (Sample period: 2017-2018)

Variables	(1)	(2)	(3)
	Panel	Logit	Poisson
	AuditOpin	AuditOpin	AuditOpin
CAM_num	0.0498***	1.2546***	0.0650***
	(2.7148)	(2.6157)	(3.4572)
Size	0.0280	0.1102	0.0084**
	(1.2265)	(0.8798)	(2.3907)
Lev	0.1009	-2.8480**	-0.0009
	(1.1864)	(-2.2279)	(-0.0146)
ROA	0.4598***	10.3250***	0.7631***
	(6.3235)	(7.8438)	(6.8496)
AuditCox_rec	0.3158***	1.4403	0.0648**
	(2.9529)	(1.2171)	(2.0647)
AuditCox_inv	0.3039***	2.4449*	0.1018***
	(2.9658)	(1.9349)	(3.0769)
Avoidloss	-0.0034	0.0820	0.0324***
	(-0.2922)	(0.2521)	(3.0359)
Equity	-0.0688***	-0.1000	-0.0272**
	(-6.5170)	(-0.8972)	(-2.2551)
Turnover	-0.1222***	0.0086	-0.0115
	(-3.7693)	(0.0307)	(-1.2165)
Quick	0.0039	0.0036	-0.0015
	(0.7866)	(0.0292)	(-0.7033)
Top1	0.0005	0.0131	0.0001
	(0.3695)	(1.4492)	(0.7924)
Dual	-0.0059	-0.0139	0.0016
	(-0.4115)	(-0.0616)	(0.2971)
Dirpct	-0.0475	-2.0278	-0.0635
	(-0.3862)	(-1.4100)	(-1.3677)
Dirsize	-0.0089**	-0.0219	-0.0016
	(-2.5198)	(-0.5765)	(-1.6189)
Salary	0.0277**	0.1540	-0.0035
	(2.0253)	(0.7914)	(-0.7173)
Committee	-0.0043	0.1023	0.0027
	(-0.2788)	(0.4411)	(0.4379)
Big4	0.0147	1.4261	0.0159**
	(0.2559)	(1.3798)	(2.1843)
Seperation	-0.0015	0.0059	0.0000
	(-1.3474)	(0.4139)	(0.0476)
Cons	0.1238	-2.6248	-0.3124***
	(0.2356)	(-0.8682)	(-3.0834)
Fixed-effect	yes	yes	yes
Industry	yes	yes	yes
Year	yes	yes	yes
Obs	4, 698	4, 698	4, 698
Adj $R^{2}$ / Within $R^{2}$	0.0475	0.0356	0.0490

Note: *

P

< 0.1, **

P

< 0.05, ***

P

< 0.01, two tails.

7 Conclusion

Using a sample of 6, 662 firm-year observations for the period 2016–2018, we analyze the association between the number of critical audit matters and audit opinion. Specifically, we investigate whether the number of critical audit matters has an effect on audit opinion and whether audit risk plays a moderating effect on their relationship.

Our results indicate that the number of critical audit matters has a significant and positive effect on audit opinion. This is consistent with the conclusion that the more the critical audit matters, the more possibility of clear opinion. However, when audit risk is higher, the positive association between the number of critical audit matters and audit opinion will be less significant. In the further analysis, we find that: 1) When the audit firm is Big4 audit firm, the influence of the number of critical audit matters on audit opinion is less significant; 2) When the auditing is more complex, the influence of the number of critical audit matters on audit opinion is less significant; 3) When the audit committee is greater, the influence of the number of critical audit matters on audit opinion is less significant.

Our results contribute to the research of critical audit matters and support calls for keeping cautious during the audit process. With the adaption of critical audit matters, audit report become more informative, however the disclosure of critical audit matters may also influence the judgment and decision making of auditors. In our paper, we find evidence that critical audit matters have a significant positive effect on the audit opinion. This result warns that both regulators and auditors should pay more attention to the potential effect of critical audit matters.

Variable Definitions

variable	definition
AuditOpin	An indicator variable, equals one if a firm gets an unqualified audit opinion other than a modified audit opinion, and zero otherwise.
CAM_num	The number of critical audit matters.
Size	The natural logarithm of the total assets at the end of the year
Lev	The ratio of total liabilities divided by total assets at the end of the year
Growth	Growth rate of sales between year $t$ and year $t - 1$
ROA	The ratio of net income divided by total assets at the end the year.
AuditCox_rec	the ratio of accounts receivable divided by total assets
AuditCox_inv	the ratio of inventory divided by total assets
Avoidloss	An indicator variable, equals one if the earnings of firm is higher than zero and less than 0.5, and zero otherwise.
Equity	The natural logarithm of the equity at the end of the year
Turnover	The ratio of sales revenue divided by the average of assets
Quick	The ratio of quick assets divided by the current liabilities at the end of the year
Top1	The share percentage of the largest shareholder at the end of the year.
Dual	An indicator variable, equals one if the CEO and the chair of the board are the same person, and zero otherwise
Dirpct	The proportion of independent directors in the total number of board memebers
BoardSize	the number of directors on the board
Salary	Natural log of the salary of senior management
Committee	the number of audit committee members on the board
Big4	An indicator variable, equals one if the company is audited by Big4, and zero otherwise.
Separation	The degree of separation of ownership and control
AuditRisk_accrual	An indicator variable, equals one if accrual earnings management is higher than average accrual earnings management, otherwise it will be measured as 0
AuditRisk_real	An indicator variable, equals one if real earnings management is higher than average real earnings management, otherwise it will be measured as 0

References

Publishing order | Descend order by publishing year | Descend order by cited within

Public Company Accounting Oversight Board (PCAOB). The auditor's report on an audit of financial statements when the auditor expresses an unqualified opinion and related amendments to PCAOB standards. Release No. 2017–001, Rulemaking Docket Matter No. 034, 2017. https://pcaobus.org/rulemaking/docket034/2017-001-auditors-report-final-rule.pdf.

Cited in this article [2]

2	Christensen B E, Glover S M, Wolfe C J. Do critical audit matter paragraphs in the audit report change nonprofessional investors' decision to invest?. Auditing: A Journal of Practice & Theory, 2014, 33(4): 71- 93. https://doi.org/10.3969/j.issn.1004-8278.2014.04.020 Cited in this article [5]

Public Company Accounting Oversight Board (PCAOB). Proposed auditing standard on the auditor's report on an audit of financial statements when the auditor expresses an unqualified opinion. PCAOB Release No. 2013–005. (August 13). Washington, DC., 2013. http://pcaobus.org/rules/rulemaking/docket034/release_2013-005_arm.pdf.

Cited in this article [1]

4	Prasad P, Chand P. The changing face of the auditor's report: Implications for suppliers and users of financial statements. Australian Accounting Review, 2017, 27(2): 348- 367. Cited in this article [2]

5	Simnett R, Huggins A. Enhancing the auditor's report: To what extent is tere support for the IAASB's proposed changes?. Accounting Horizons, 2014, 28(4): 719- 747. https://doi.org/10.2308/acch-50791

6	Sneller L, Bode R, Klerx A. Do IT matters matter? IT related key audit matters in Dutch annual reports. International Journal of Disclosure and Governance, 2017, 14(2): 139- 151. https://doi.org/10.1057/s41310-016-0017-0 Cited in this article [2]

7	Brasel K, Doxey M, Grenier J, et al. Risk disclosure preceding negative outcomes: The effects of reporting critical audit matters on judgments of auditor liability. The Accounting Review, 2016, 91(5): 1345- 1362. https://doi.org/10.2308/accr-51380 Cited in this article [4]

8	Wright A M, Wright S. Modification of the audit report: Mitigating investor attribution by disclosing the auditor's judgment process. Behavioral Research in Accounting, 2014, 26(2): 35- 50. https://doi.org/10.2308/bria-50662 Cited in this article [2]

9	Wu X L, Xu F M, Wang W. The confirmation bias in judgment and decision making. Advances in Psychological Science, 2013, 20(7): 1080- 1088. https://doi.org/10.3724/SP.J.1042.2012.01080 Cited in this article [2]

10	Nickerson R S. Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 1998, 2(2): 175- 220. https://doi.org/10.1037/1089-2680.2.2.175 Cited in this article [1]

11	Bar-Tal Y, Brycz H, Dolinska B, et al. When saying that you are biased means that you are accurate? The moderating effect of cognitive structuring on the relation between metacognitive self and confirmation bias use. Current Psychology, 2017, 38(2): 1- 7. Cited in this article [3]

12	Simunic D A. The pricing of audit services: Theory and evidence. Journal of Accounting Research, 1980, 18(1): 161- 190. https://doi.org/10.2307/2490397 Cited in this article [3]

13	Simunic D A. The impact of litigation risk on audit pricing: A review of the economics and the evidence. Auditing: A Journal of Practice and Theory, 1996, 15(1): 119- 134. Cited in this article [2]

14	Deloitte. Comment Letter: PCAOB Rulemaking Docket No. 034, 2013. Http://Pcaobus.Org/Rules/Rule- making/Docket034/132b_Deloitte.Pdf. Cited in this article [1]

15	Gaetano C. PCAOB's expanded audit report proposal raises litigation concerns. The Trusted Professional: Newspaper of the New York State Society of Certified Public Accountants, 2014, 17(1): 1- 2. Cited in this article [1]

16	Carcello J. What do investors want from the standard audit report?. The CPA Journal, 2012, 82(12): 22- 28. Cited in this article [1]

17	Louis-Philippe S, Jean B, Palash B. The informational value of key audit matters in the auditor's report: Evidence from an eye-tracking study. Accounting Horizons, 2018, 32(2): 141- 162. https://doi.org/10.2308/acch-52047 Cited in this article [1]

18	Qi C, Xu J, Zhang Y. The effects of audit quality disclosure on audit effort and investment efficiency. The Accounting Review, 2019, 94(4): 189- 214. https://doi.org/10.2308/accr-52286 Cited in this article [1]

19	Vinson J M, Robertson J C, Cockrell R C. The effects of critical audit matter removal and duration on jurors' assessments of auditor negligence. Auditing: A Journal of Practice & Theory, 2019, 38(1): 183- 202. Cited in this article [2]

20	Gimbar C, Hansen B, Ozlanski M E. The effects of critical audit matter paragraphs and accounting standard precision on auditor liability. Accounting Review, 2016, 91(6): 1629- 1646. https://doi.org/10.2308/accr-51382 Cited in this article [1]

21	Kang Y J. Are audit committees more challenging given a specific investor base? Does the answer change in the presence of prospective critical audit matter disclosures?. Accounting, Organizations and Society, 2019, 77(3): 1- 14. Cited in this article [1]

22	Snyder H, Jongward S. Watching the watchmen: Applying audit risk models to auditor's opinions. Journal of Theoretical Accounting Research, 2007, 2(7): 1- 33. Cited in this article [1]

23	Dechow P M, Sloan R G, Sweeney A P. Detecting earnings management. Accounting Review, 1995, 70(2): 193- 225. Cited in this article [2]

24	Jones J J. Earnings management during import relief investigations. Journal of Accounting Research, 1991, 29(2): 193- 228. Cited in this article [2]

25	Roychowdhury S. Earnings management through real activities manipulation. Journal of Accounting and Economics, 2006, 42(3): 335- 370. Cited in this article [2]

26	Bhattacharya U, Daouk H, Welker M. The world price of earnings opacity. The Accounting Rreview, 2003, 78(3): 641- 678. Cited in this article [1]

27	Biddle G, Hilary G. Accounting quality and firm-level capital investment. The Accounting Review, 2006, 81(5): 963- 982. Cited in this article [1]

28	Francis J R, Wang D. The joint effect of investor protection and Big4 audits on earnings quality around the world. Contemporary Accounting Research, 2008, 25(1): 157- 191. Cited in this article [1]

29	Francis J R, Krishnan J. Accounting accruals and auditor reporting conservatism. Contemporary Accounting Research, 1999, 16(1): 135- 165. Cited in this article [1]

30	Fama E F. Agency problems and the theory of the firm. Journal of Political Economy, 1980, 88(2): 288- 307. Cited in this article [1]

PDF(215 KB)

975

Accesses

Citation

Detail

Sections

Recommended

Abstract
Key words
Cite this article
1 Introduction
2 Literature Review and Hypothesis Development
2.1 Regulatory Background and Literature Review
2.2 Hypothesis Development
3 Research Design
3.1 Sample Selection
3.2 Measurement of All Variables
3.3 Research Model
4 Empirical Results
4.1 Descriptive Statistics of Audit Opinion
Table 1 The number of company with or without critical audit matters
Table 2 Audit opinion of companies with critical audit matters
Table 3 Descriptive statistics of main variables
4.2 Main Results
Table 4 The regression between audit opinion and critical audit matters
4.3 Moderating Effect of Audit Risk
Table 5 The moderating effect of audit risk (audit risk refers to the degree of loss aversion
Table 6 The moderating effect of audit risk (audit risk refers to variable AuditRisk_real)
Table 7 The moderating effect of audit risk (audit risk refers to variable AuditRisk_accrual)
5 Additional Analyses
5.1 Moderating Effect of Audit Complexity
Table 8 The moderating effect of audit complexity (audit complexity refers to variable AuditCox_rec)
Table 9 The moderate effect of audit complexity (audit complexity refers to variable AuditCox_inv)
5.2 Moderating Effect of Big4
Table 10 The moderating effect of audit firm
5.3 Moderating Effect of Audit Committee
Table 11 The moderating effect of audit committee
6 Robustness Checks
Table 12 Robustness check (using different models)
Table 13 Robustness check (Sample period: 2017-2018)
7 Conclusion
Variable Definitions
References

Received	Accepted	Published
2020-04-22	2021-06-01	2022-08-25
Issue Date
2022-08-15

模态框（Modal）标题

Please choose a citation manager

Content to export

Abstract

Key words

Cite this article

1 Introduction

2 Literature Review and Hypothesis Development

2.1 Regulatory Background and Literature Review

2.2 Hypothesis Development

3 Research Design

3.1 Sample Selection

3.2 Measurement of All Variables

3.3 Research Model

4 Empirical Results

4.1 Descriptive Statistics of Audit Opinion

Table 1 The number of company with or without critical audit matters

Table 2 Audit opinion of companies with critical audit matters

Table 3 Descriptive statistics of main variables

4.2 Main Results

Table 4 The regression between audit opinion and critical audit matters

4.3 Moderating Effect of Audit Risk

Table 5 The moderating effect of audit risk (audit risk refers to the degree of loss aversion

Table 6 The moderating effect of audit risk (audit risk refers to variable AuditRisk_real)

Table 7 The moderating effect of audit risk (audit risk refers to variable AuditRisk_accrual)

5 Additional Analyses

5.1 Moderating Effect of Audit Complexity

Table 8 The moderating effect of audit complexity (audit complexity refers to variable AuditCox_rec)

Table 9 The moderate effect of audit complexity (audit complexity refers to variable AuditCox_inv)

5.2 Moderating Effect of Big4

Table 10 The moderating effect of audit firm

5.3 Moderating Effect of Audit Committee

Table 11 The moderating effect of audit committee

6 Robustness Checks

Table 12 Robustness check (using different models)

Table 13 Robustness check (Sample period: 2017-2018)

7 Conclusion

Variable Definitions

{{custom_sec.title}}

{{custom_sec.title}}

References

{{custom_fnGroup.title_en}}

Footnotes