• No results found

The effect of conservatism on audit fees and audit quality

N/A
N/A
Protected

Academic year: 2021

Share "The effect of conservatism on audit fees and audit quality"

Copied!
44
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The Effect of Conservatism on Audit Fees and Audit Quality

Name: Xueyan Jia

Student number: 11638354

Thesis supervisor: Dr. Alexandros Sikalidis

Date: June 24, 2018 Word count: 12,148

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Xueyan Jia who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Accounting conservatism is one of the main characteristics of accounting information. Prior researches mainly focus on the effect of conditional conservatism or unconditional conservatism on earning, financing and governance, and yet few studies investigates whether there is an association between accounting conservatism and audit fees. This paper mainly investigates the respective effect of conditional conservatism and unconditional conservatism on audit fees and whether they can affect audit fees as well. Although there is a related literature stating that conditional conservatism has a negative association with audit fees, this paper complements and extends the study through considering the effect of unconditional conservatism on audit fees and taking effects of conditional conservatism and unconditional conservatism on audit quality into account. The empirical research proves that both conditional conservatism and unconditional conservatism are negatively associated with audit fee. However, the associations between conditional conservatism and audit quality and between unconditional conservatism and audit quality are converse to each other. The findings of this paper can attract the attention from researchers, regulators, audited companies and auditors.

(4)

Contents 1 Introduction ... 6 2 Literature Review ... 9 2.1 Accounting Conservatism ... 9 2.2 Conditional Conservatism ... 10 2.3 Unconditional Conservatism ... 10 2.4 Prior Literature ... 11 2.5 Theoretical Development ... 13 3 Methodology ... 15

3.1 Measurement for conditional conservatism ... 16

3.2 Measurement for unconditional conservatism ... 18

3.3 Measurement for audit quality ... 19

3.4 Empirical Model ... 19

3.5 Sample Selection ... 21

4 Results ... 23

4.1 Descriptive Statistics ... 23

4.2 Estimation results for conditional conservatism... 27

4.3 Estimation results for audit quality ... 28

4.4 The relation between conditional conservatism and unconditional conservatism and audit fee ... 29

4.5 The relation between conditional conservatism and unconditional conservatism and audit quality ... 31

5 Conclusion ... 33

References ... 36

(5)
(6)

1 Introduction

Accounting conservatism, as one of the qualitative characteristics of accounting information, becomes more and more important to conform to financial report's users' needs in terms of transparency and timeliness of financial reporting. The rewards of implementing accounting conservatism for accounting firms and organizations are still being researched. Prior researches mainly focus on the effect of conditional conservatism or unconditional conservatism on earning, financing and governance, and yet, few studies investigates whether there is an association between accounting conservatism, both conditional conservatism and unconditional conservatism, and audit fees. According to Lee et al. (2015), implementing conditional conservatism reduces the litigation risk for auditors, which leads to a decrease in audit fees. This paper is grounded in this idea and extends the theory with some specific points. My paper primarily inquiry the effect of conservatism, both conditional conservatism and unconditional conservatism, on the audit fees.

In addition, this paper also aims to verify whether conditional and unconditional conservatisms have positive effect on audit quality. As accounting firms continually expand their scope of business to cater to client's needs in many different ways like management report, the relationship between accounting firm and client becomes complicated. Users of audit report raise the question whether auditors can be independent and retain satisfying audit quality, when they are bothered by this kind of complicated beneficial relationship. Also, many researchers claim that this kind of relationship has the potential to determine auditor's independence, which can indirectly influence the perceived audit quality and actual audit quality.

How to deal with the relationship between auditors and clients to assure the audit quality has always been a hot issue. Many relevant paper come up with different determinants that may have an impact on audit quality, like time budget pressure, audit committee expertise, auditor relationship, etc. Although those related literatures dig into the associations between those determinants and audit quality, few of them considers either conditional conservatism or unconditional conservatism as one key determinant to affect audit quality.

As reported by Venkataraman et al. (2008), the remission of litigation risk can set off the declines of both audit fee and audit quality, because auditors are not obliged to put too much efforts into retaining litigation risk within an acceptable level. Wherefore, the mitigation of

(7)

litigation risk caused by realizing either conditional conservatism or unconditional conservatism possibly engenders a shrinkage for audit quality.

Furthermore, although many relevant paper present distinct points about the relationship between audit fees and audit quality, most literatures share the same point that audit fees do affect audit quality in an unfixed pricing domain. Therefore, since the conditional conservatism and unconditional conservatism may have negative effects on the audit fees, the diminution of audit fee may brings about an escalation of audit quality.

Therefore, this paper also aims to testify whether the audit fees affected by either conditional conservatism or unconditional conservatism can impact the audit quality. If it can, both of these accounting conservatisms can be productive to handle the audit quality issue in a complex relationship between auditor and client.

To sum up, this paper mainly investigates the effect of conservatisms (both conditional and unconditional conservatisms) on audit fees and audit quality.

RQ: Does the conservatism (both conditional conservatism and unconditional conservatism) affect audit fees and audit quality?

All data necessary for empirical research is collected from WRDS (Compustat and AuditAnalytics). After filtering out, I have 1074 samples for approximately 200 companies. The results of regression for audit fee comfirm that both conditional conservatism and unconditional conservatism have the similar moderating effect on audit fee, suggesting that an organization with higher degree of either conditional conservatism or unconditional conservatism is charged less by auditors. This result is partly consistent with the findings from Lee et al. (2015). However, according to the results of regression for audit quality (abnormal accruals), the associations between conditional conservatism and audit quality and between unconditional conservatism and audit quality are reverse to each other. Conditional conservatism has a negative association with audit quality, while conversely, unconditional conservatism can enhance audit quality. Though some possible explanations about this converse results are presented, further investigations are needed.

This paper contributes to both academic research and practical application in three ways. First, although there are many relative literatures researching either conditional conservatism or unconditional conservatism, most of the prior researches have not involved with the kind of study related with audit fees or audit quality. For example, Lara et al. (2016) elucidates

(8)

that accounting conservatism improves investment efficiency by means of reducing overinvestment. Penman and Zhang (2002) certifies that accounting conservatism affects the quality of earnings reported on the income state. This paper complements and extends the academic study through considering the effect of conditional conservatism and unconditional conservatism on audit fees and taking effects of conditional conservatism and unconditional conservatism on audit quality into account. Although academic research on accounting conservatism has increased rapidly, this paper is a fresh start to analyze accounting conservatism from the perspective of auditing. Simultaneously, through comparing the effects of conditional conservatism and unconditional conservatism on audit fees, I can figure out which one has more effects on audit fees, and ultimately on audit quality, which can help researchers have a much deeper understanding about conditional conservatism and unconditional conservatism.

Secondly, the findings of this research question are valuable for audited companies, accounting firms and policy regulators. Companies that have been audited frequently may take a keen interest in this kind of topic, because the audit fees can be rationally priced in a reasonable way and they can have a deeper understanding about the potential attributes of the audit quality. Uncontentiously, accounting conservatism has a far-reaching impact on the auditees' auditing activity. And auditees may have the motivation to ponder about whether they should undertake either conditional conservatism or unconditional conservatism to amend its audit quality. In addition, when companies aim to adjust their corporate governance mode into accounting conservatism, the comparison between the effects of conditional conservatism and of unconditional conservatism on audit fees and audit quality can provide more relevant information about these two conservatisms to companies, which assist in choosing the more suitable one for organizations.

This paper also aids accounting firms to reasonably charge audit fees, since accounting conservatisms (both conditional and unconditional conservatisms) reduce litigation risk that accounting firms may take. This kind of reduction in audit fees is unequivocally favorable in the present situation that accounting firms have to face the ever-growing competition from peers. Meanwhile, a noticeable increase in audit quality offers accounting firms more preponderances as well. Therefore, the reasonable pricing and increasing audit quality boosts competiveness of accounting firms.

Although several literatures that are concerned with the accounting conservatism encourage organizations to carry out either conditional conservatism or unconditional conservatism,

(9)

there is no compulsory regulation for execution of accounting conservatism. Policy makers can take the findings of this paper into account when they aim to improve the legitimacy of audit through enacting more regulations about accounting conservatism. And because of the adverse impact of conditional conservatism and unconditional conservatism on audit quality, regulators can provide more regulatory guidance to users of audit report and organizations. The reminder of this paper is structured as follows: The next section explains the definition of accounting conservatism, the detailed information about conditional conservatism and unconditional conservatism and the theorytical development. The third section describes the methodology of this paper. To specify, the proxies, empirical model and data selection are stated in this section. The forth section sums up the empirical results. And the last section presents the conclusion of the whole paper.

2 Literature Review

2.1 Accounting Conservatism

Accounting conservatism is one of the qualitative characteristics of accounting information. The definition of accounting conservatism used to change a lot. According to Neag and Maşca (2015), the first argument about accounting conservatism appearing in Gliman and MacNeal (1939). That paper mainly discusses the asset valuation bases for financial report. After that, there are many similar academic researches doubting the different parts of accounting conservatism. Financial Accounting Statements Board (FASB) announced Statement of Financial Accounting Concepts 2 in 1980 and defined conservatism as "a prudent reaction to uncertainty to try to ensure that uncertainties and risks inherent in business situations are adequately considered". Nowadays, accounting conservatism is just like what FASB defined in essence.

Mounting literatures probe into the strength and weakness of implmenting accounting conservatism and discuss the necessity of implementing accounting conservatism. Watts (2003) claims that accounting conservatism is effective to address agency problems resulted by lenders having asymmetric information and asymmetric payoffs. García Lara et al. (2016) also mentions the negative implications of accounting conservatism to overinvestment. Yet, no explicit consensus about this kind of necessity has been reached. All in all, whether it is wide to perform accounting conservatism is contentious, since this whole topic is premature. Accounting conservatism is categorized into conditional conservatism and unconditional

(10)

conservatism. Although accounting conservatism has been prevalent as a topic in academic research, compared with unconditional conservatism, conditional conservatism attract more attention from scholars. There are no specific reason or any relevant research investigating why this kind of difference happens, but the distinction between conditional conservatism's definition and unconditional conservatism's one may be one clue to explain it.

2.2 Conditional Conservatism

Conditional conservatism is defined as a higher verification requirement for recognizing good news (gains) than bad news (loss) (Basu 1997). Because of the asymmetric timeliness between different kind of news, many paper imply that the implementation of conditional conservatism is beneficial both for investors and organizations. Kim and Pevzner (2010) proves that higher current conditional conservatism is associated with lower probability of future bad news and that the stock market reacts more sensitively to those organizations that carry out more conditional conservatism than to others when it comes to a good (bad) news. Many academic literatures such as Liu and Elayan (2015) and Lafond and Watts (2008) also agree that the implementation of conditional conservatism can mitigate the information asymmetry between managers and investors. Liu and Elayan (2015) points out that as it is sincerely acknowledged that the implementation of conditional conservatism can partly settle the information asymmetry problem, a higher level of litigation risk may motivates organizations to consider the effect of conditional conservatism on information asymmetry more often. Lafond and Watts (2008) refers to that conservatism does not only play an information role for the organization but also demotivate management to manipulate accounting information and financial statement.

Likewise, the reasonable doubt to conditional conservatism arises often as well. Wibawa and Wardhani (2018) states that "as the recognition of economic losses becomes more timely, the sensitivity of firm investment to cash flow decreases". Hence, although more and more paper are studying the advantages and disadvantages of conditional conservatism, it is hard to draw a conclusion about whether the organizations should put conditional conservatism in effect or not and if they do, its unknown impact.

2.3 Unconditional Conservatism

Unconditional conservatism, as another form of accounting conservatism, is not prevalent in various literatures. Unconditional conservatism means that the recognized book values of net

(11)

asset are below their expected market values during their lives at inception. There is expected unrecorded goodwill yielded from the inception of assets and liabilities (Beaver and Ryan 2005). Compared with conditional conservatism, unconditional conservatism is more news-independent. As conditional conservatism reacts asymmetrically to bad news and to good news, unconditional conservatism has no connection with any kind of news in stark contrast. While conditional conservatism is one kind of ex post action, unconditional conservatism is the kind of ex ante action. Most academic researchers pay more attention on studying conditional conservatism not on studying unconditional conservatism. Although there are many paper researching conditional conservatism and unconditional conservatism simultaneously (Kohansal et al 2017, Beaver and Ryan 2005, Gassen et al. 2006 and Iatridis 2011), unconditional conservatism is seldom solely studied.

There are no specific reason or any relevant research investigating why this kind of difference happens, so it is a great topic to investigate further in the future.

2.4 Prior Literature

Seetharaman et al. (2002) states that audit fees is positively related with the litigation risk. And it also mentions that there are two possible cases to explain the association between litigation risk and audit fees. The first one is that fairly high litigation risk activates auditors to make some efforts to reduce the possibility of being prosecuted. To deal with the high litigation risk and reach the expected audit risk, auditors devote many more resources like human resources or time to the auditing process, directly causing an increasing audit cost. On the other hand, auditors take possible future litigation loss into consideration when they set the audit pricing. To cover the possible litigation loss in the future, auditors may take this kind of loss into account in advance when they try to submit an audit offer to auditees. The cases can either solely or jointly explain the impact of litigation risk on audit fees, since both of them are reasonable and can happen practically.

There are many relevant literatures that express the similar point about the relationship between audit fees and litigation risk. For instance, Abbott et al.(2017) states that there is an impact of litigation risk on auditor pricing behavior and proves the point with the evidence from reverse mergers. Venkataraman et al. (2008) utilizes IPO setting to ascertain the relationship between litigation risk, audit fees and audit quality and provides strong evidences to justify that since the increasing litigation exposure sways auditors incentives, a higher-litigation regime can lead to increases in both audit fees and audit quality. Similarly,

(12)

Seetharaman et al. (2002) discusses the association between litigation risk and audit fees by focusing on UK firms cross-listed on US markets and certifies that the incremental litigation risk from US market can be one of the reasons to explain why those kind of UK cross-listed on US market are charged more than other UK companies that are not listed in US market. Most of these relevant literatures approve that along with reduction of litigation risk, audit fee is getting lower to some extent.

And Lee et al.(2015) notes that the conditional conservatism has an negative effect on the audit fees, since it may attenuates the litigation risk for auditors. To specify, conditional conservatism limits the management's opportunistic behavior and reports bad news in a timely manner, which can lower the litigation risk. Since many academic researchers insist that litigation risk is positively associated with audit fees, the reduction of litigation risk definitely results in changes of audit pricing. Nevertheless, the efficacy of conditional conservatism is debilitated when the governance of organizations is high-leveled. The paper puts forth that an excellent corporate governance can lower the hazard litigiousness as much as possible, then the benefit of carrying out conditional conservatism to mitigate litigation risk is not so notable any more.

Moreover, although the paper has quoted some words from Watts (2003a, 2003b) to assume that since unconditional conservatism can limit the possibility of overstating the firms' assets and understating the firms' liability, unconditional conservatism may lowers the litigation risk and then has a negative effect on audit fees just like conditional conservatism does, this kind of research is not discussed in the paper. And other relevant research about this topic has not been found yet. Therefore, whether unconditional conservatism has the similar impacts on audit fees still needs to be digged.

By comparing the financial information from pre-IPO period and from post-IPO period, Venkataraman et al. (2008) explores the relationship between litigation risk and audit fee and audit quality affirms that since auditors must make more efforts to handle the litigation risk, litigation risk is positively associated with both audit fee and audit quality.

However, without strong testimony, it is rough to conclude straightforwardly that audit quality declines as soon as litigation risk decreases

In addition, there is a widely shared view in some related literatures that abnormal audit fee has an impact on audit quality. And one divergence about this topic exists. Most of the literatures like Choi et al. (2010), are supportive of the view that the relationship between

(13)

audit fees and audit quality is asymmetric. To specify, the positive abnormal audit fees can lower the audit quality, but the negative abnormal audit fees have no significant association with audit quality. However some paper like Asthana and Boone (2012) do not agree with this idea. They believe that below-normal audit fees can negatively impact audit quality as well. Asthana and Boone (2012) enunciates that as the absolute value of negative abnormal audit fee increases, audit quality declines with it. Asthana and Boone (2012) deems that the bargaining power of audited organizations in the process of negotiation may exacerbate audit quality. Although the specific form of the relationship between audit fees and audit quality is not universally acknowledged, what all relevant literatures mentioned above consent is that audit fee does affect audit quality in an unfixed pricing domain.

However, apart from these two views mentioned above, a latest literature, Dabor and Benjamine (2018), proposes an unusual argument that abnormal audit fee is not significantly related with audit quality.

Hence, whether abnormal audit fee is associated with audit quality is still arguable. And if the association exists, the specific form of this association should be figured out.

2.5 Theoretical Development

This paper is grounded in the paper Lee et al.(2015) and extends the theory with some specific points. To reprove the negative association between conditional conservatism and audit fee, more timely data is utilized. Under the circumstance that litigation risk for auditors can be decreased by the implementation of conditional conservatism, audit cost for auditors is correspondingly decreased. Similarly, unconditional conservatism, another form of accounting conservatism should has achieved the same result. Since Lee et al. (2015) just concentrates on the effect of conditional conservatism on audit fees and mentions the possibility of unconditional conservatism to lower the litigation risk without proving it, I think it is necessary to complement the theory with the effect of unconditional effect on audit fees.

Although unconditional conservatism, as another form of accounting conservatism except conditional conservatism, is not as prevailing as conditional conservatism among companies, many literatures articulate the positive effects of unconditional conservatism on the governance and operation of firms, which proves the research values of unconditional conservatism and may attracts more attention from researchers. Due to the fact that few paper

(14)

looks into the real value of unconditional conservatism solely, this extension also helps us to have a better understanding about unconditional conservatism.

While conditional conservatism and unconditional conservatism are essentially two forms of accounting conservatism, their differences in practice may make them differ in the extent of impact on the audit fees. The assumption that unconditional conservatism may works the same way like conditional conservatism in audit fees is not reliable enough, so it is necessary to separate the hypothesis into the conditional conservatism one and unconditional conservatism one. And through making a comparison between these extents, I can figure out which conservatism affects audit fees and audit quality more. Therefore, the hypothesis should be listed separately.

H1a: Unconditional conservatism is negatively associated with audit fees. H1b: Conditional conservatism is negatively associated with audit fees.

Except the relationship between accounting conservatism (both conditional one and unconditional one) and audit fees, the relationships between conditional conservatism and audit quality and between unconditional conservatism and audit quality is another object of this study. How to sustain audit quality in an acceptable level is a popular research topic, and many determinants for audit quality are mentioned in a variety of literatures. For example, Broberg et al. (2017) claims that there is a negative relationship between audit quality and time budget pressure. The findings suggest that the quality of audits in smaller listed firms that benefit from greater expertise on their audit committees (Ghafran and O'Sullivan, 2017). Ball et al. (2015) find a positive relation between audit firm engagement by the client (firm-to-firm) and audit quality. However, researchers scarcely express concerns about whether audit quality can be influenced by conditional conservatism and unconditional conservatism. About the relation between conditional conservatism and unconditional conservatism and audit quality, there are two contrary assumptions.

One assumption is that both conditional conservatism and unconditional conservatism are negatively related with audit quality, since the mitigation of litigation risk motives less auditing work. Inspired by Venkataraman et al. (2008), I doubt whether the audit fee lowered by conditional and unconditional conservatism is negatively associated with audit quality, since the cutback of litigation risk caused by the performances of conditional conservatism and unconditional conservatism may elicits the situation that auditors endeavors less in auditing. Under the premise that both conditional conservatism and unconditional

(15)

conservatism can lessen litigation risk, the likelihood of earning management's occurrence is likely diminished, which motivates auditors to do less work in auditing.

However, it is not reasonable to infer directly that both conditional conservatism and unconditional conservatism negatively affect audit quality after taking what is mentioned in Choi et al. (2010) and Asthana and Boone (2012) into consideration. Based on these two paper , audit fee does affect audit quality in an unfixed pricing domain, since abnormally high audit fee can be cut back after the conservatism mitigates the litigation risk. Then whether the abatement of audit fee caused by either conditional conservatism or unconditional conservatism affect audit quality positively is questionable. Also, I cannot deny the fact that both of conditional conservatism and unconditional conservatism possibly impact on audit quality directly. Hence, another assumption is that both conditional conservatism and unconditional conservatism are positively related with audit quality, since audit fee is negatively related with audit quality.

It is impossible to draw a conclusion about which one assumption is more reasonable or true without any empirical tests. And if both of these assumptions are true, the extents of their impact on the association between accounting conservatism and audit quality must be disparate.

No matter which kind of way either conditional conservatism or unconditional conservatism impacts on audit fees, directly or indirectly, this paper assumes initially that audit quality is negatively related with both conditional conservatism and unconditional conservatism.

H2a: Unconditional conservatism is negatively associated with audit quality. H2b: Conditional conservatism is negatively associated with audit quality.

Therefore, this paper also aims to investigate whether conditional and unconditional conservatisms can positively affect audit quality.

By certifying these hypothesis, we can comprehend the link between accounting conservatism (both conditional one and unconditional one) and audit fee and the link between accounting conservatism and audit quality better.

3 Methodology

In this section, I first introduce the proxies for conditional conservatism, unconditional conservatism and audit quality, since they are not directly measurable. Next, the empirical

(16)

models for H1 and H2 are elucidated. In the end, the standard of data collection process and the reasons why I select those data are clarified.

3.1 Measurement for conditional conservatism

To measure the conditional conservatism, inspired by Khan and Watts (2009) and Basu (1997), I resort to the C_SCORE (Khan and Watts ; 2009) as the proxy for conditional conservatism. And Ahmed and Duellman (2013) works as a reference to adjust the model a little bit.

The Basu (1997) cross-sectional regression is specified as

NIt = β1+ β2Dt+ β3RETt+ β4DtRETt+ ε

where NI is net income before extraordinary items and after noncontrolling interest scaled by market value of equity; RET is the stock return for the 12-month period of fiscal year ; D is a dummy variable equal to 1 when Return<0 and equal to 0 otherwise, and ε is the residual. The good news timeliness measure is β3. And β4 is the measure of incremental timeliness for

bad news over good news, or conservatism. The total bad news timeliness is β3+β4.

Based on the regression from Basu (1997), Khan and Watts (2009) elaborates two new notions (G_SCORE and C_SCORE) to estimate the timeliness of good news and bad news separately. And the paper states that the timeliness of good news each year (G_SCORE) and the incremental timeliness of bad news (C_SCORE) are two linear functions of firm-specific characteristics each year:

G_SCORE = β3 = μ1+ μ2SIZEt+ μ3M/Bt+ μ4LEVt

C_SCORE = β4 = λ1+ λ2SIZEt+ λ3M/Bt+ λ4LEVt

where SIZE is the natural log of the total assets; M/B is the market-to-book ratio (market value of equity divided by the book value of equity; and LEV is total liability divided by total asset.

In Khan and Watts (2009)'s derived regression, the variables used to estimate C_SCORE contain firm size, market-to-book ratio and firm leverage.

Even though firm size is one crucial element in the process of estimating conditional conservatism, no absolute judgment about whether the association between firm size and conditional conservatism is positive or negative. On the one hand, it is sincerely acknowledged that larger firms have much more mature information environment, which

(17)

promotes the reduction of information asymmetries. This kind of inner situation propels large firms not to adopt conditional conservatism to evade information asymmetries. On the other hand, large firms like Big 4 are destined to face more challenges or threats from outside than other small organizations. They have to deal with much more litigation risk than others, since outsiders can acquire much more economic benefits. Hence, a large firm may follows conditional conservatism to allay litigation risk.

As stated in Khan and Watts (2009), when organizations have a high market-to-book ratio, they own more growth options, which signifies a higher agency cost. And conditional conservatism, as one of the efficient ways to settle agency cost, are applied more often in those organizations.

For those firms with a high leverage, their agency conflicts between lenders and shareholders. And conditional conservatism is also a solution to ease the conflict.

Thus and thus, these three variables can be combined together into C_SCORE regression so as to estimate conditional conservatism appropriately

Then Khan and Watts (2009) substitutes G_SCORE and C_SCORE into Basu's regression: NIt= β1+ β2Dt+ RETt1+ μ2SIZEt+ μ3M/Bt+ μ4LEVt)

+ DtRETt(λ1+ λ2SIZEt+ λ3M/Bt+ λ4LEVt)

+ (δ1SIZEt+δ2M

Bt + δ3LEVt+ δ4DtSIZEt+ δ5DtM/Bt+ δ6DtLEVt) + ε where D is a dummy variable equal to 1 when Return<0 and equal to 0 otherwise; RET is the stock return for the 12-month period of fiscal year; SIZE is the natural log of the total assets; M/B is the market-to-book ratio (market value of equity divided by the book value of equity; and LEV is total liability divided by total asset.

After regressing the above model of NI, I obtain the coefficients (λ1, λ2, λ3, λ4) for DtRETt,

DtRETtSIZEt, DtRETtM/Bt and DtRETtLEVt . Then, I substitute the estimated coefficients

( λ1, λ2, λ3, λ4) into Basu (1997) 's C_SCORE model to calculate every sample's C_SCORE

which is the proxy for conditional conservatism.

The regressions from Khan and Watts (2009) and Basu (1997) have been trendy and played a prominent role to measure conditional conservatism in many pertinent literatures, such as Gassen et al. (2016), Goh et al.(2017), Banker et al. (2017), etc. After these certification and

(18)

practice from these paper, the regressions are mature and reliable to measure conditional conservatism accurately and comparably.

3.2 Measurement for unconditional conservatism

Just like what I mentioned in the theory part, researchers seldom explore fields related with unconditional conservatism. In spite of the fact that some literature are involved with conditional conservatism and unconditional conservatism at the same time, there is no consensus about the most applicable measurement for unconditional conservatism. The amount of useful proxy specified in relevant essay is pretty small, by reason of the small number of paper related with unconditional conservatism. Even though there are a few ways to estimate unconditional conservatism, there is no guarantee that the selected proxy can be truly behalf of the degree of unconditional conservatism for sampled organizations, which may restricts the accuracy of final results.

Ahmed and Duellman (2012) resorts to 2 proxies to estimate unconditional conservatism. One is, Con-ACC and another one is skewness. According to Givoly and Hayn (2000) and Ahmed et al. (2002), Con-ACC stands for the accumulation of negative accruals, while skewness is the difference between cash flow skewness and earning skewness. Both of them are not practically investigated about whether they can truly reflect the extent of unconditional conservatism. Before making a decision about which one should be designated for the measurement of unconditional conservatism in this paper, I pre-calculate the value of data under those two measurements. Since the skewness's results contain plenty of extreme values, inspired by Givoly and Hayn (2000) and Ahmed et al. (2002), I decide to utilize Con-ACC to measure unconditional conservatism. And some basic introduction and pre-calculation process about skewness are presented in Appendix B

Uncon = Con– ACC =−(NI0− CFO + DE) AVA

where NI0 is the net income before extraordinary items and after noncontrolling interest; CFO

stands for the net cash flow of operation; DE is the depreciation expense; AVA is the average asset of that year.

Since the detailed information about depreciation expense is not provided by WRDS, I make use of OI (operating net income before depreciation) to replace NI0 + DE. And Con-ACC is

(19)

Uncon = Con– ACC = −(OI − CFO) AVA

3.3 Measurement for audit quality

Because audit quality cannot be directly measured, I utilize the earning quality in lieu of audit quality like what prior literatures do (Asthana and Boone 2012; Choi et al. 2010; Fung et al. 2017). And abnormal accruals is a proxy for earning quality, as a result of the fact that this one is prevailing in this issue.

The model for abnormal accruals is developed by Dechow and Dichev (2002) and modified by Ball and Shivakumar (2006):

ACCt = α0+ α1CFt+ α2CFt−1+ α3CFt+1+ α4DCFt+ α5DCFtCFt+ et

where ACC stands for total accruals (earnings before extraordinary items) scaled by average total assets; CF is the operating cash flows scaled by average total assets; CFt-1 is the

operating cash flows at year t-1 scaled by average total assets at year t; CFt+1 is the operating

cash flows at year t+1 scaled by average total assets at year t; DCFt equals one when the

change in cash flows is negative (CFt - CFt-1 < 0), and zero otherwise; et is the error term.

Even if unoberservability of audit quality generates the situation that no peculiar and precise measurement for audit quality exists, earning quality can replace audit quality in empirical model since it has been tested in different literatures frequently.

3.4 Empirical Model

To testify H1a and H1b, I make a regression to test the relationship between audit fees and conditional and unconditional conservatisms.

LAUDITt = β0+ β1CONt+ β2UNCONt+ β3BIG4+ β4SIZEt+ β5LEVt+ β6ROAt + β7FYENDt+ ε

where LAUDIT is the natural log of total audit fees; CON and UNCON are the proxies for conditional conservatism and unconditional conservatism; BIG4 is a dummy variable equal to 1 when the accounting firm who is responsible for auditing is one of the Big 4 and equal to 0 otherwise; SIZE is the natural log of the total assets; LEV is total liability divided by total asset; ROA is return on assets calculated as income before extraordinary items divided by beginning of the year total assets; FYEND is a dummy variable equal to 1 when the fiscal

(20)

year of audited organization ends on December 31th and equal to 0 otherwise, and ε is the residual.

Next, some explanations about why these variables are in this model is specified. First, when the accounting firm who is in charge of auditing is one of Big 4, it denotes that auditee demands to pay more for auditing. There are some possible reasons to explain why Big 4 charge much more than others. Sirois et al. (2012) supports that in order to enhance the audit efficiency and audit quality, Big 4 engage in investing much more energy and money on audit technology. Also, higher payment for auditors is probably another reason.

In addition, if most organizations ' fiscal year end on December 31th, this kind of assemblage can bring on a busy season for auditors and a higher charge for auditing services during that period. However, because of the fierce competition in U.S. audit market (Sirois et al., 2012), I assume that some accounting firms may take advantage of low-ball audit fees to attract more clients during the busy seasons. Therefore, it is difficult to tell whether FYEND impacts on the audit fee. Moreover, if the audited firm has complex businesses and intricate operations, it will trouble the whole auditing process and require to cost more resources from auditors, resulting in an increase in audit fee. That's why I pick up some other variables to stand for the complexities of audited firm.

To testify H2a and H2b, the association between conditional and unconditional conservatisms and audit quality, the regression is based on the regression from Choi, Kim and Zang (2010) and modified a little bit.

AA = μ0+ μ1CONt+ μ2UNCONt+ μ3LNTAt+ μ4BIG4t+ μ5BTMt+ μ6LOSSt+ μ7LEVt + ε

where AA is the proxy for abnormal accruals (et ); CON and UNCON are the proxies for

conditional conservatism and unconditional conservatism; LNTA is the natural log of total assets; BIG4 equals one if the auditing company is one of the Big 4, otherwise zero; BTM is book-to-market ratio; LOSS equals one if the company reports a loss in that year, otherwise zero.

There are two assumptions about BIG4 and LNTA in this regression of abnormal accruals. First of all, normally, the size of organizations determines the maturity of their whole governance system, the investment in supervising management and other matters to some extent, which may positively influences the audit quality. Next, since some literatures like Sirois et al. (2012) supports that Big 4 engage in investing much more energy and money on

(21)

audit technology and talents, the audit quality is assumed to be higher for those organizations audited by Big 4.

3.5 Sample Selection

All data that I employ are collected from WRDS. Some of them like CFO are available in Compustat of WRDS and some of them like audit fees can be found in AuditAnalytics. I reorganize and splice the files from Compustat and AuditAnalytics into one file. Moreover, most of the required data can be directly extracted from database, but some data like book-to-market ratio and natural log of total assets have to be computed after gathering data from WRDS.

The samples of data are from North American database. To ensure the accuracy of the data, I choose to regard the fiscal year as the time limitation.

Moreover, the Sarbanes–Oxley Act of 2002 enacted on July30, 2002 issue many new or strict requirements for all US public company, which must enormously influence the governance structure, accounting practices, management decision of organizations in different industries. Also, in order to gain a footing in this sea change of policy and claim a share of the pie in the reconstructed auditing market, accounting firms have to alter their audit procedure and audit efficiency correspondingly. Asthana and Boone (2012) claims that the promulgation of SOX drives auditors to be more independent. And since the Sarbanes-Oxley Act of 2002 has influences on many fields, industries and accounting conservatism, the data range is planned from 2003 to 2017, so that no drastic fluctuation in data can make a mess of the final testing. Since I have to utilize 3-year data to calculate the proxy for audit quality, some data like CFO may be gathered from 2002. In a conclusion, to avoid effect of SOX on final data verification, the sample period for this paper ( other than the data of CFO and ROA) is restricted to the period from 2005 to 2016.

However, due to the situation that AuditAnalytics has not disclosured every organizations' 2017 audit fee, I finally determine that data (except CFO and ROA) are primarily from 2005 to 2015.

In Compustat database, the initial sample consists of 196,834 observations for 28400 companies from fiscal year 2003 to 2017. Since the database from Compustat lack a little of history data of some organizations' financial situation and marketing prices, the database is cut back as appropriate in order to sustain the completeness and stability of data from

(22)

selected organizations. I delete observations missing data of dividends, net cash flow of operation, market value of equity, net income before extraordinary items and after noncontrolling interest and the close price and the number of shares outstanding so as to ascertain that all data needed for estimating conditional conservatism, audit quality and other control variables are well prepared. After this action, I have 1737 observations left. Then I delete some observations that is either from the first two years of every organizations in this database or the their last year in data base, because I have to utilize 3-year data to calculate Table 1

Data distribution by year

Year Number Percentage

2005 82 9.99% 2006 78 9.50% 2007 77 9.38% 2008 55 6.70% 2009 86 10.48% 2010 85 10.35% 2011 79 9.62% 2012 79 9.62% 2013 75 9.14% 2014 69 8.40% 2015 56 6.82% Total 821 100%

the proxy for audit quality, yielding 1256 observations left. Also, considering that there are some extreme values in some variables, I delete some data that are not suitable for the whole data testing. Finally, I get 821 observations for validation of H1 and 1074 observations for H2.

In AuditAnalytics database, the initial sample consists of 177,052 observations for 28400 companies from fiscal year 2004 to 2015. I just collect the data about audit fee and the name of accounting firms that are in charge of auditing for respective audited firms, so there is not any omission in this database. Later, I combine these two database into one. In order to avert possible impacts of extreme values on the whole data analysis, I winsorize all variables at the top and bottom 1 percent of their distributions

(23)

Table 1 outlines the specific data distribution by year. From the table, apparently, the percentage that each year covers in the whole sample varies from 6.70% to 10.48%, which illustrates that data is evenly distributed.

4 Results

4.1 Descriptive Statistics

Table 2 states the descriptive statistics of the variables for Hypothesis 1 and Hypothesis 2 and introduces the general information of the whole sample.

The medians of conditional conservatism variable (CON) and unconditional variable (UNCON) are 0.055 and 0.021 respectively. Since both of these numbers are positive, it implies that more than 50% selected companies decided to put conditional conservatism and unconditional conservatism into its accounting action for some reasons in the period from 2004 to 2015, proving that accounting conservatism (both of them) became well-accepted . The median of net cash flow of operation (CF) is 0.069, suggesting that most organizations in the database experience a positive net cash flow of operation. The median leverage rate (LEV) is 69.4%, so it seems that most of the organizations' liabilities was kind of beyond control, which called for management's attention. The min value, max value and median of M/B are 0.331, 5.357 and 1.623 respectively. The latter two figures specifies that more than half of the selected companies were underestimated in its book value or overstated in its market value. Plus, this kind of estimation or overestimation can be intense for some organizations, because their market values were approximately 5.4 times than their book values. Considering the narrow difference between minimum and maximum of client size (SIZE), the sampled organizations were on a similar scale. There are abounding relative literatures asserting that audit fee is greatly associated with client size. Carson and Fargher (2007) presents that the additional audit services demanded for larger clients explains the association between audit fee and client size. Accordingly, the narrow difference between the minimum and maximize of LAUDIT (the variable standing for audit fee) can be explained by the similar scales of sampled organizations on some level. Approximately 89.7 percent of the whole sample were audited by Big 4 (BIG4), which likely had a positive impact on the audit quality. Moreover, 38.6 percent of the whole samples experienced a negative net cash flow in that period (DCF). The median of FYEND is 1, hinting that most sampled organizations regard December 30

(24)

Table 2

Descriptive statistics

Variables n Mean Std. Dev. Min Max Median

RET 1074 2.041 9.032 -32.410 27.530 2.51 D 1074 0.30 0.46 0 1 0 ACC 1074 0.026 0.032 -0.190 0.100 0.028 CF 1074 0.069 0.035 -0.066 0.191 0.069 CFt-1 1074 0.063 0.032 -0.074 0.168 0.063 CFt+1 1074 0.077 0.038 -0.058 0.262 0.074 LEV 1074 0.674 0.109 0.190 0.894 0.694 M/B 1074 1.745 0.732 0.331 5.357 1.623 SIZE 1074 22.291 1.725 16.471 25.087 22.467 BIG4 1074 0.897 0.304 0 1 1 BTM 1074 0.658 0.278 0.140 1.937 0.615 DCF 1074 0.386 0.487 0 1 0 LNTA 1074 22.291 1.725 16.471 25.087 22.467 LOSS 1074 0.062 0.241 0 1 0 UNCON 1074 0.021 0.019 -0.089 0.090 0.021 CON 1074 0.066 0.071 -0.054 0.433 0.055 AA 1074 -0.009 0.035 -0.158 0.083 -0.007 NI 821 0.053 0.031 -0.131 0.110 0.057 LAUDIT 821 14.422 1.057 11.261 16.713 14.447 BYEND 821 0.898 0.303 0 1 1 ROA 821 0.028 0.020 -0.150 0.119 0.029

(25)

Table 3

Correlation among variables for hypothesis 1 and 2

NI BIG4 ACC BTM CF CFt-1 CFt+1 RET D DCF LAUDIT

NI 1.0000 BIG4 0.0149 1.0000 ACC 0.0361 0.1522* 1.0000 BTM -0.4322* -0.0011 0.0390 1.0000 CF 0.0154 0.1454* 0.7139* 0.0595 1.0000 CFt-1 0.0078 0.1255* 0.5489* 0.0593 0.6339* 1.0000 CFt+1 0.0139 0.1317* 0.3671* 0.0625 0.2037* 0.2137* 1.0000 RET 0.0765 0.0281 0.0031 -0.1719* 0.0124 0.0091 0.0100 1.0000 D -0.0764 -0.0109 0.0046 0.1544* -0.0175 -0.0600 -0.0230 -0.6496* 1.0000 DCF -0.0782 -0.0149 -0.0035 0.0549 -0.0380 0.0374 -0.0172 -0.0499 0.0727 1.0000 LAUDIT 0.0127 0.6159* 0.1151* 0.0731 0.1452* 0.1304* 0.1293* 0.0354 -0.0353 -0.0179 1.0000 LEV -0.0105 -0.1630* -0.7714* -0.0835 -0.8664* -0.8708* -0.4638* -0.0104 0.0404 -0.0027 -0.1643* LNTA 0.0568 0.6261* 0.3139* 0.0645 0.3822* 0.3370* 0.3379* 0.0357 -0.0522 -0.0315 0.8704* LOSS -0.3942* -0.0737 -0.2209* 0.2554* -0.1910* -0.1618* -0.1749* -0.1124* 0.1350* 0.0528 -0.0327 M/B -0.0032 0.0186 0.5494* -0.0409 -0.0272 -0.0579 -0.0812 -0.0196 0.0459 0.0350 -0.0466 ROA 0.0348 0.1527* 0.9670* 0.0365 0.6169* 0.4114* 0.5275* 0.0025 0.0096 -0.0068 0.1122* SIZE 0.0568 0.6261* 0.3139* 0.0645 0.3822* 0.3370* 0.3379* 0.0357 -0.0522 -0.0315 0.8704* AVA 0.0312 0.2528* 0.0253 0.0201 0.0372 0.0335 0.0316 0.0352 -0.0018 -0.0444 0.7157* OI 0.0654 0.2559* 0.0280 -0.0280 0.0414 0.0370 0.0370 0.0280 0.0025 -0.0524 0.7052* CON -0.0036 0.0122 0.5215* -0.0440 -0.0590 -0.0899 -0.0983 -0.0200 0.0474 0.0349 -0.0533 UNCON 0.0171 0.1795* 0.7984* 0.0809 0.7707* 0.6610* 0.7088* 0.0081 -0.0321 0.0012 0.1805* FYEND 0.0501 0.0343 -0.0122 0.0346 -0.0187 -0.0165 -0.0160 -0.0007 0.0511 -0.0136 0.1980* AA 0.3142* -0.0543 0.0895 -0.0594 -0.0193 -0.0144 -0.0923 0.0589 -0.0815 -0.3140* 0.0185

(26)

Table 3 (continued)

Correlation among variables for hypothesis 1 and 2

LEV LNTA LOSS M/B ROA SIZE AVA OI CON UNCON FYEND AA

NI BIG4 BIG4 ACC BTM CF CFt-1 CFt+1 RET D DCF LAUDI T LEV 1.0000 LNTA -0.4301* 1.0000 LOSS 0.2161* -0.1914* 1.0000 M/B -0.0358 -0.1101* -0.0193 1.0000 ROA -0.7035* 0.3065* -0.2266* 0.5895* 1.0000 SIZE -0.4301* 1.0000* -0.1914* -0.1101* 0.3065* 1.0000 AVA -0.0414 0.6765* -0.0537 -0.0211 0.0238 0.6765* 1.0000 OI -0.0420 0.6715* -0.0937 -0.0205 0.0271 0.6715* 0.9648* 1.0000 CON 0.0008 -0.1266* -0.0114 0.9993* 0.5641* * -0.1266* -0.0232 -0.0225 1.0000 UNCON -0.9052* 0.4575* -0.2579* 0.0380 0.8034* 0.4575* 0.0436 0.0516 0.0048 1.0000 FYEND 0.0190 0.1974* 0.0119 0.0095 -0.0118 0.1974* 0.1937* 0.1863* 0.0100 -0.0268 1.0000 AA 0.0126 0.0164 -0.3613* 0.0794 0.0708 0.0164 0.0670 0.0274 0.0799 0.0405 0.0445 1.0000

(27)

as the end of their fiscal year, which may promotes a higher audit fees. The medians of NI and ROA are positive.

Table 3 describes the correlation between the variables for testifying H1 and H2. By cause of the different numbers of sample for H1 and H2, only the 821 samples that are subject to test H2 are picked out to inquire into the correlation between those variables. From this panel, there are some findings that can justify the viewpoints of other pertinent academic literature. Initially, audited firms have to pay more audit fees for auditing services, when the accounting firm is one of the big 4, which is consistent with my prediction and other relevant literatures like Carson and Fargher (2007), Carsen et al. (2004), and Rusmanto and Waworuntu (2015). Further, the greater book value of asset is, the more likely the client's accounting firm is one of big 4. These two finding about big 4 make sense, in that normally big 4 charge higher audit fees for auditing services. Thirdly, it seems that leverage rate is highly negatively correlated with total accruals (ACC), which reminds of much more consideration on leverage rate in the organization's operation.

4.2 Estimation results for conditional conservatism

Table 4 reveals the estimated coefficients through making a regression in Khan and Watts' model so as to estimate C_SCORE ( variable for conditional conservatism). Although all these coefficients regressed from the model have be recorded in this panel, only the coefficients for D*RET, D*RET*SIZE, D*RET*M/B and D*RET*LEV are exploited to measure C_SCORE.

Table 4 shows that NI is negatively associated with D and M/B, which can explain why D*M/B has a positively association with NI. Unlike the result from Khan and Watts (2009), RET is negatively related with NI, which is pretty strange. Because the sign for the coefficient of RET is unexpected, accordingly, D*RET, D*SIZE, RET*M/B, RET*LEV, D*RET, D*RET*SIZE, D*RET*M/B and D*RET*LEV are influenced to some extent. Compared the result of my regression for C_SCORE and the result from Khan and Watts (2009), there are some differences between them. The significant one is the opposite signs for RET. After investigating the originality of this distinction, it turns out that the extent of variation of equity's market value is greater than the extent of variation of net income in my samples. While NI in this regression model is the net income before extraordinary items and after noncontrolling interest scaled by market value of equity, NI is shrank. Hence, when a

(28)

Table 4

Coefficients from estimation regressions for C_SCORE

Variable Coefficient t-stat.

D -0.2239 -4.46 D*SIZE 0.0033 1.25 D*M/B 0.0416 3.79 D*LEV 0.1166 2.29 RET -0.0085 -1.85 RET*SIZE 0.0002 1 RET*M/B 0.0005 0.75 RET*LEV 0.0041 1.08 D*RET -0.0165 -1.06 D*RET*SIZE -0.0004 -0.51 D*RET*M/B 0.0092 2.71 D*RET*LEV 0.0147 1.16 SIZE 0.0001 0.07 M/B -0.0087 -1.84 LEV 0.0044 0.16

positive change of net income before extraordinary items and after noncontrolling interest occurs in two years, on the contrary, NI decreases. Thus, under the premise that the time periods of data between this paper and Khan and Watts (2009) is disparate, this opposite signs for RET can be explicable.

4.3 Estimation results for audit quality

The estimation results of audit quality (abnormal accruals) are outlined in table 5 and table 6. Since the estimated result of residuals contains two extreme values, I winsorize the residuals at the top and bottom 1 percent of their distributions without compromising the integrity of whole samples for H2.

From the table 6, the median of et (RESIDUAL) is negative (-0.0064), suggesting that

abnormal accruals is commonly negative.

However, as a result of the situation that audit quality cannot be measured directly, the estimation of audit quality (abnormal accruals) may be not accurate or reliable enough to

(29)

Table 5

Coefficients from estimation regressions for audit quality

Variable Coefficient t-stat.

CF 0.8964 3.13 CFt-1 -0.0341 -0.16 CFt+1 0.4508 2.96 DCF 0.0590 1.24 DCF*CF -0.3137 -0.86 Table 6

Descriptive statistics for residuals

Variables n Mean Std. Dev. Min Max Median

et (RESIDUAL) 1074 -0.0088 0.0353 -0.1579 0.0826 -0.0064

represent the real situation of audit quality, which is regarded as one of the limitations of this paper. The concrete explanation and discussion are mentioned in the conclusion part.

4.4 The relation between conditional conservatism and unconditional conservatism and audit fee

Table 7 presents the result of regression analyses of audit fee (LAUDIT) and yields the verification about the relationship between conditional conservatism and unconditional conservatism (H1). CON for conditional conservatism is calculated by using the coefficients of some variables from NI regression. UNCON for unconditional conservatism is estimated by the use of CON-ACC model. As there is an alternative model for the measurement of unconditional conservatism, only the CON-ACC model is kept in order that data stability is not disrupted.

From the regression of LAUDIT, the coefficient of BIG4 is positive and significant (p < 0.01) like I presumed, proving that if the accounting firm is one of the big4, companies have to pay much more for auditing services. And from the descriptive statistics, we can find that most of the sampled organizations would prefer to hire big4 than other accounting firms, for whatever reasons. Like predicted, the coefficient of SIZE is positive and significant (p < 0.01) as well, indicating that the size of auditees is one momentous factor that accounting firms consider

(30)

Table 7

Conditional conservatism and unconditional conservatism and audit fee

Predicted Sign Coefficient t-statistic

BIG4 + 0.2802 4.71 (0.000)*** LEV ? -0.5648 -3.09 (0.002) *** SIZE + 0.5715 46.29 (0.000)*** ROA ? -2.3773 -3.00 (0.000)*** FYEND ? -0.0605 -1.30 (0.194) CON - -1.0286 -2.92 (0.004)*** UNCON - -2.3153 -2.94 (0.003)*** R2 0.8593 N 821

while fixing a price for auditing services. Further, from the regression of audit fee, I find that LEVERAGE and ROA are significantly negatively related with audit fees, which is consistent with some results of Khan and Watts (2009). Also, FYEND is insignificantly (p > 0.1) negatively associated with audit fee, implying that FYEND is not associated with audit fee.

(31)

As predicted, the coefficient of CON (conditional conservatism) is negative and significant ( p < 0.01), suggesting that those companies whose conditional conservatism is enforced more thoroughly can decrease more expenditures on auditing. This outcome is consistent with the findings from Lee et al.(2015). As well, the coefficient of UNCON (unconditional conservatism ) is negative and significant ( p < 0.01). And the absolute value of UNCON is greater than the one of CON, proving that unconditional conservatism does not only share the same function with conditional conservatism in pricing procedures for auditing services, but also act the function better. However, since the model for measuring unconditional conservatism is not tested by many literatures, the statement that unconditional conservatism functions better than in constricting audit fee still calls for further investigation. In summary, the results of regression support the viewpoint that both conditional conservatism and unconditional conservatism are negatively related with audit fee.

4.5 The relation between conditional conservatism and unconditional conservatism and audit quality

Table 8 presents the result of regression analyses of audit quality (AA) and testify the relationships between conditional conservatism and unconditional conservatism and audit quality. The absolute value of AA stands for audit quality in the regression of abnormal accruals. The greater absolute value of AA is, the poorer audit quality an organization has. Hence, ideally, when the coefficient of AA approaches to zero, it confirms that audit quality is quite high.

From the regression of abnormal accruals, the coefficient of BIG4 is positive and insignificant (p > 0.1), which signifies that it is not possible to enhance audit quality even if organizations hire Big 4 that have well-developed audit technology and more talents in different industries for auditing services. Analogously, the coefficient of LNTA is negative and insignificant (p > 0.1). This insignificant association between LNTA and AA denies the assumption mentioned above that large-scaled companies owning more mature governance system and better supervision system may indirectly impact on audit quality. The coefficients of BTM is positive and significant (p < 0.01), stating that BTM is negatively related with audit quality. Similarly, the estimated coefficient of LOSS is positive and significant as well (p < 0.01). This proves that those organizations that report a loss in the fiscal year probably experience an increase in their absolute value of abnormal accruals and then a reduction in the audit quality. On the contrary, the coefficient of LEV is negative and significant, denoting

(32)

Table 8

Conditional conservatism and unconditional conservatism and audit quality

Predicted Sign Coefficient t-statistic

BIG4 - 0.0036 1.28 (0.201) LEV ? -0.0317 -4.45 (0.000)*** LNTA - -0.0001 -0.3 (0.765) BTM ? 0.0124 4.05 (0.000)*** LOSS ? 0.424 14.16 (0.000)*** CON - 0.1219 8.7 (0.000)*** UNCON - -0.1188 -3.37 (0.001)*** R2 0.2761 N 1074

that LEV is positively associated with audit quality. One possible explanation about this kind of association is that accounting firms probably endeavor to investigate every detail about financial information for those organizations that have high leverage rate so as to keep a lid on audit risk.

Unlike I predicted, the associations between conditional conservatism and audit quality and between unconditional conservatism are asymmetry. The coefficient of CON is positive and significant ( P < 0.01) in the regression of abnormal accruals, while the coefficient of UNCON is negative and significant ( P < 0.01). Hence, conditional conservatism is

(33)

negatively related with audit quality, while unconditional conservatism is positively related with audit quality. In the previous section, there are two contrary assumptions about the relationships between conditional conservatism and audit quality and between unconditional conservatism and audit quality. Interestingly, the regression results for CON and UNCON are reverse. Hence, the inferences about the final results for those relationships are presented respectively.

For the association between conditional conservatism and audit quality, there are two suppositions. The first one is that this negative association is solely resulted in by the reduction of litigation risk. As the implementation of conditional conservatism can mitigate litigation risk, the reduction of litigation risk may relaxes auditors' vigilance in the process of auditing, thereby negatively affecting audit quality. Or, this negative association is jointly caused by the reduction of litigation risk and cutback of audit fee, but cutback of audit fee has less impact on audit quality than the other one does.

Oppositely, the first assumption to explain the positive relationship between unconditional conservatism and audit quality is that the decrease of audit fee may affect the audit quality solely. Based on Choi et al. (2010) and Asthana and Boone (2012), audit fee does affect audit quality in an unfixed pricing domain, so it is reasonable that the unconditional conservatism can indirectly affect audit quality through cutting down the audit fee. Secondly, under the situation that both the reduction of litigation risk and cutback of audit fee have an influence on audit quality, conversely, the cutback of audit fee is the more effective one.

In conclusion, these adverse results about the relationship between conditional conservatism and audit quality and unconditional conservatism and audit quality cannot be explained easily. Therefore, more researchers should inquiry this topic further.

5 Conclusion

In this paper, I further explore the effects of accounting conservatism, both conditional conservatism and unconditional conservatism, on some issues of auditing. Inspired by the findings from Lee et al. (2015), I initially reprove whether audit fee can be curtailed by the use of conditional conservatism in the new database, and then probe into whether unconditional conservatism shares the same efficacy with conditional conservatism. I hypothesis that as organizations are in pursuit of either conditional conservatism or unconditional conservatism for accounting practices, the audit fee is reduced correspondingly. In other words, both conditional conservatism and unconditional conservatism are negatively

(34)

related with audit fee. Besides, audit quality, as one of key subject about auditing, is also examined as well. This paper present two assumptions about the relationship between conditional conservatism and unconditional conservatism and audit quality. Although these two assumptions analyze this relationship in different perspectives and produce divergent results about the relation, I initially hypothesis that both of these conservatism are negatively associated with audit quality.

My results of empirical research prove that both conditional conservatism and unconditional conservatism have moderating effect on audit fee, perhaps because the application of accounting conservatism alleviates litigation risk that the accounting firms undergo, directly or indirectly cutting the spending for accounting firms. Hence, the finding is consistent with H1 and the findings from Lee et al. (2015). Except that, I also find that audit fee is higher for those companies accompanied with low leverage rate, large scale and low ROA or audited by Big 4.

However, the findings of this paper demonstrates that unlike what I expected in H2, the relationships between conditional conservatism and audit quality and between unconditional conservatism and audit quality are contrary to each other. The implementation of conditional conservatism suppresses audit quality, while unconditional conservatism plays a key role in boosting audit quality. Several conjectures about why this counter results exist are mentioned, but more investigations about this topic are required to determine which kind of conjecture can truly elucidate this inconsistency.

Since this paper expands on the findings from Lee el al. and certifies the effectiveness of conditional conservatism and unconditional conservatism to some auditing issues, the findings of this paper is advantageous for academic researchers, regulators, audited companies and auditors. Nevertheless, there are some limitations in this paper.

First of all, the narrow scale of data and exclusive territorial restrictions may sway the credibility and stability of final findings in this paper. Because of confined availability of partial data, I have to filter out abundant fragmentary samples without compromising the integrality of the database. Thus, I just have 821 samples for H1 and 1074 samples for H2. Only approximately 200 companies' data are involved in this data, whereas I expected that the number of sampled organizations is as great as possible. I cannot deny that the instability of final findings would be recovered a little bit if the data were scaled up. Moreover, the exclusive territorial restrictions and time limitations also block the contribution of this paper

(35)

in some ways. For a further research, many literatures can put more other areas' samples and data before SOX into consideration to broader the findings of this paper.

Secondly, the proxies for unconditional conservatism are not enough convincing. Under the present circumstance that most literatures pay less attention on studying unconditional conservatism, viable model to estimate unconditional conservatism is not ample. Even though some proxies for unconditional conservatism like Con-ACC or skewness are acknowledged by a few literatures, more evidence-based practices are desired to validate the accuracy of unconditional conservatism's measurement whilst using these proxies. Therefore, unless proxies for unconditional conservatism will be practically tested and developed by mounting literature, it is going to be tough to lessen this kind of constraints.

Thirdly, the reason why conditional conservatism and unconditional conservatism function adversely on audit quality is unknown. Though this paper discusses several possible explanations, the root causes of these opposite results cannot be specified without an intensive research.

Referenties

GERELATEERDE DOCUMENTEN

Wordt wel een vroege voorjaarsbemesting gegeven dan kunnen juist trage witte klaverrassen met een klein aandeel in het bestand, een hogere eerste snede geven, omdat het gras meer

For instance, in its recommendation on the establishment of the Youth Guarantee, the Council of the European Union has emphasized the need for member states to build a

Since no distinct seasonal cycles could be identified for the measured aromatic hydrocarbons (Fig. 2), the possible influ- ence of air masses passing over different source regions

Deze Big Data Revolutie wordt ook uitmuntend beschreven in het boek ‘De Big Data Revolutie’, waarin big data wordt beschreven als bron van economische waarde en

To hide the search pattern, we make use of techniques used in oblivious RAM [14], [21], [22] (ORAM) and private information retrieval [3], [9] (PIR), which solve this problem

affordable, reliable, clean, high-quality, safe and benign energy services to support economic and human

Het verschil in effect tussen live en uitgesteld kijken op merkbekendheid en merkreputatie van de sponsor kan mogelijk worden gemedieerd door de mate van betrokkenheid bij

This chapter described the running-in of rolling-sliding contacts on macroscopic and microscopic level. 1) On macro-scale, the geometrical change of the contacting