Page 1 of 68
AMSTERDAM BUSINESS SCHOOL
Auditor changes, auditor type and earnings
management in the U.S. context
Student Name: Oana-Corina Gheorghe Student Number: 10622721
Date: 23 June 2014
MSc Accountancy and Control, Track Accountancy
Faculty of Economics and Business, University of Amsterdam First Supervisor: Prof. Dr. Vincent O’Connell
Page 2 of 68 Abstract
The purpose of this study is to examine how earnings management impacts the probability of an audit firm to change, judging both from a general point of view and from specific cases. In order to do this, I examine how the manipulation of a company’s financial earnings, either directly or through indirect accounting methods, influences the probability of an auditor to be dismissed or to resign from an engagement. I also want to analyze the influence of the auditor’s type (Big 4 or Non-Big 4) on the choice of ending business relationships between the audit company and its client. I conduct my research in the U.S. context, in the period 2003 – 2008 because the American environment represents the ideal background in terms of litigation risk and legislative setting (the Sarbanes-Oxley Act). The results provide evidence contrary to my expectations: auditor change is not triggered by the accrual based earnings management; but in fact it is better explained by the abnormal levels of cash-flows from operations and production costs metrics of real earnings management. Moreover, discretionary accruals, production costs and discretionary expenses increase the probability of auditors’ resignations.
Key words: accrual-based earnings management, real earnings management, auditor change, auditor dismissals, auditors’ resignations, auditor type, litigation risk, financial condition, Sarbanes-Oxley Act.
Page 3 of 68 Table of contents
1.Introduction ... 4
2. Literature review, hypothesis development and conceptual background ... 8
2.1. Agency theory ... 8
2.2. Audit service quality ... 9
2.3. Earnings management ... 11
2.4. The effect of Sarbanes – Oxley Act ... 13
2.5. Litigation risk of auditor, deep pockets theory, auditor and client characteristics . 144 2.6. Auditor change: auditor resignations and auditor dismissals ... 15
2.6.1. Auditor resignations ... 16
2.6.2. Auditor dismissals ... 18
3. Research design, sample selection and empirical models... 19
3.1 Sample selection... 19
3.2. Research models and variables ... 21
3.3. Empirical model ... 21
4.Empirical results ... 29
4.1 Descriptive statistics ... 29
4.2 Results of the multivariate tests ... 36
4.2.1. Hypothesis 1 ... 36
4.2.2. Hypothesis 2 ... 40
5. Conclusions ... 51
References ... 55
Page 4 of 68
1. Introduction
Accounting rules can be applied in a very flexible manner and this treatment leads to accounting techniques that can be used to characterize a positive image of the company and its financial position. Managers prefer higher rather than lower earnings and referring to agency theory, they have incentives to manipulate earnings in order to obtain private gain. These are some of the reasons why there is a need for the external involvement of auditors. Financial auditors are hired to verify if the financial statements are properly prepared. The final product of this process is the audit opinion which provides reasonable assurance that the financial statements are “presented fairly, in all material aspects, and/or give a true and fair view in accordance with the financial reporting framework”. If this opinion is not correct and if it leads to audit failures, the probability of a lawsuit is high and it may lead to a possible litigation risk. Beyond its obvious threat, litigation risk is a potential factor that could impair a firm’s reputation and, in addition to this, Palmrose (1988) documented that it is a “negative signal about the quality of the firm’s audit services”.
One measure taken after the bankruptcy of Arthur Andersen was the issuance of the Sarbanes-Oxley Act in 2002 which, further on, lead to the creation of Public Company Accounting Oversight Board (PCAOB). This organization oversees the audits of public companies, having as its main purposes the protection of investors and that of the public interest; so the importance of this organization is, without doubt, extreme in the certification of audits. According to an article in “The New York Times” published on August 23rd 2012, by Philip Scott Andrews, PCAOB reported that in recent months, none of the audits of twenty three brokerage firms was acceptable. The main target of this article were small audit companies, but there was also a specific negative reference towards big auditing firms: “Among the Big Four - Deloitte & Touche, PricewaterhouseCoopers, KPMG and Ernst & Young - the board found something wrong in nearly one in six audits it reviewed that year” and a year later, the number doubled. This is not an encouraging sign and I make use of this information in order to sustain my point of view that audit companies face a loss of reputation risk and might be put in the situation of being sued.
Since the demise of Arthur Andersen in 2006, at least 6,543 companies switched their audit company and this action continued but in a more limited way until 2006, when 1,322 U.S.
Page 5 of 68
companies changed their independent auditor. However, both practitioners and academics consider that in the audit market there exists a transparency problem which rises from the fact that many companies do not disclose in their 8-K forms (a form broadly used in the United States to inform shareholders about material events) the reasons for the auditor change. This fact negatively impacts investors, the ultimate beneficiaries of the audit report. There are clear signs from practice for this concern, for example, one of the largest accounting firms in the world, Grant Thornton LLP has demanded SEC to require companies to provide a reason for all the auditor changes. In the majority of cases, companies decide to dismiss their auditors. This can happen for many reasons: when there is a policy of changing auditors after a period of time, to change to an audit company that provides higher quality services or when the audit company disagrees with management on key accounting policies, companies can opt to move to an auditor which agrees with their viewpoints. On the other hand, auditors might resign for multiple reasons among which: when they decide that a client is not worth the risk or when they simply do not trust management.
There has been conducted extensive research about earnings management which has demonstrated that there are incentives to positively inflate accounting numbers. According to Matsumoto (2002), in addition to using discretion over reported earnings to meet expectations, managers can influence analysts’ opinions. When looking at earnings management, we can divide between accrual-based earnings management and real earnings management.
Accrual-based earnings management is achieved by modifying estimates or accounting methods when presenting an economic transaction in the financial statements. As opposed to the first method, real earnings manipulation is a deliberate activity to alter reported earnings in a precise direction, which is accomplished by modifying the timing or structuring of an operation, investment or financing transaction and which has substandard business consequences. There hasn’t been much emphasis put on real earnings management, even though this kind of action increased after the implementation of SOX. One reason for this might be the fact that it is harder to detect as opposed to accrual-based earnings management and is more costly for shareholders. Graham et al. (2005) document that most earnings management is achieved via real actions and not through accounting techniques. Their paper brought to light the fact that management casually admits engaging in real economic actions, such as: delaying maintenance or advertising
Page 6 of 68
expenditure. Moreover, they also would give up positive NPV projects just to meet short-term earnings benchmarks.
The construct of earnings management is outlined by the evidence presented above which shows that managers have incentives to report enhanced earnings. They can do this by using their power over the financial statements in order to create a prosperous image of the company.
If there are unsolved disagreements regarding the accounting policies between the audit company and its client, then there is a high possibility that the aforementioned relationship will end. Thus, the first research question is: Is auditor change, either in the form of dismissal or resignation, connected to earnings management?
Moreover, in order to complete my work, my research tries to demonstrate to which type of earnings management this auditor change is more connected: accrual-based or real earnings management. In the light of the above arguments, the second research question is: To which form of earnings management is auditor change more correlated?.
From an academic point of view, my research could bring valuable knowledge because I distinguish between the two types of earnings management. One of the strengths of my thesis is the fact that it also addresses real earnings management and from my readings, there is little attention given to this topic. In my paper, I would like to separate accrual-based earnings management from real earnings management and correlate this to the situation in which exists a change in audit firm. My research could improve current knowledge about earnings management and explain how these attempts lead to auditor dismissal or auditor resignation and improve both academics’ and practitioners’ understandings.
Another aspect that is innovative about my topic is that it extends knowledge about auditor change. From my readings, there is very little research regarding the grounds of this situation; few authors examined the division between auditor resignations and auditor dismissals. So far, current research has focused on linking the change of auditor with discretionary accruals and not with real earnings management and the litigation risk of auditor. Based on these facts, I believe that my paper will add to prior literature and enhance current understandings.
From a societal point of view, my work could be very useful for investors (shareholders). First of all, they will understand that auditors can sometimes provide inappropriate audit reports. Second, they might acknowledge that auditors may face situations that could impede them to serve the scope for which they were hired, that of ensuring fair financial reporting.
Page 7 of 68
The remainder of this paper is as follows. The next section concentrates on the literature review and theoretical background in relation to auditor changes, audit quality, earnings management, litigation risk and culminates with the development of the hypotheses. The data sources, the sample used, the variables employed and the empirical models are presented in section 3. Section 4 contains the results of my regressions together with the analysis of the results and the most relevant tables. Finally, section 5 discusses the most important limitations of my paper and contains the overall conclusions regarding the association between earnings management and auditor changes.
Page 8 of 68
2. Literature review, hypothesis development and conceptual background
This section provides an overview of the most relevant papers regarding earnings management, audit quality, litigation risk and their relation with auditor change. I will begin with a broad description of auditor change. Afterwards, I will document wider two possible scenarios that might occur when the relationship between the audit firm and its client does not end with an unqualified audit opinion. These prospects refer to the situations when an auditor decides to resign from an engagement or when the client dismisses its auditor. The literature review will be used as a basis for the hypotheses statement and further on, to support my findings.
2.1. Agency theory
Agency theory represents one of the most prominent theoretical issues in accounting during the last years and its importance relies upon how it incorporates conflicts of interest and incentive problems. My motivation for including this theory in my paper is based on the fact that accounting and further on, auditing are strongly linked to the problem of incentives and how incentives are used by managers of companies. One of the most relevant and important papers which focused on the agency problem is the research of Jensen and Meckling (1976, pg. 4). They define the agency relationship as a “contract under which, one person (the principal) engages another person (the agent) to perform some service on their behalf which involves delegation of some decision making authority to the agent”. Problems in the principal-agent relationship tend to occur if both parties are utility maximizers. In this case, it is very probable that the agent will not act in the best interest of the principal. Divergences arise when the parties have different interests and asymmetric information, thus the principal cannot monitor the agent because it is too costly for him to do so. Furthermore, there are incentives to write contracts in such a way that the marginal cost equals the marginal gains from reducing the residual loss, as documented by Jensen and Smith Jr. (1985).
Agency theory plays a significant role in emphasizing auditor switches. Moreover, auditors play a major part in a company’s activities since they can mitigate information asymmetry and reduce agency costs by signaling the credibility of the financial statements.
Page 9 of 68 2.2. Audit service quality
Auditors are hired in order to enhance the degree of confidence of users of financial information in the financial statements. Maintaining confidence in businesses, contributing to the quality of financial reporting and working in the public interest are some of the auditors’ core responsibilities. Thus, in order to provide high quality services, prior literature has demonstrated that Big 4 audit firms: Deloitte Touche Tohmatsu, Ernst & Young, KPMG and PricewaterhouseCoopers deliver more professional services because they have more to lose if they fail to report weaknesses in clients’ financial records (DeAngelo (1981)).
However, audit quality can be viewed as well from the perspective of the audited company. Larger clients are more likely to hire high quality auditors because they have better training, are more specialized on particular industries and deal better with organizational complexity. Also, as documented by Liu and Lai (2012) they can reduce information asymmetry costs, thereby raising firm value. Moreover, larger auditors gain the benefit of holding more financial resources and structure better their auditors’ training, while smaller auditors might not enjoy such advantages.
A different perspective on audit quality is the one which takes into consideration how it differs across individual auditors. One significant factor that separates high quality auditors from other auditors that share the market for audit services is industry specialization. Auditors included in this category provide higher quality audits because they are more likely to detect irregularities. Furthermore, Gul et al. (2013) provide evidence for this statement and argue that this is possible even in the absence of client-specific knowledge. The quality of an audit can be determined as a joint probability that the auditor will discover a breach in the client’s accounting system and will report that breach. In order to discover that breach, an auditor depends on their technological capabilities and the procedures employed. In addition to this, higher quality auditors, considered in this paper to be the Big 4 accounting firms, have incentives to protect their reputation, especially considering the loss of faith in the audit profession after the Enron scandal. Given the past circumstances, auditors are more focused on protecting their brand name. Francis et al. (2013, pg. 1629) reveal that the purpose of audit standards is to require accounting firms “to establish quality controls over their operations”. They also state that Big 4 firms have national training programs for all of their offices, standardized audit programs, and are thus better prepared as compared with other audit firms. They consider this a form of audit quality.
Page 10 of 68
The Big 4 companies have received a lot of criticism regarding audit quality in recent years. These negative comments refer to several aspects of the audit process, such as: failures to properly audit the internal controls over financial reporting and the tendency to become consultancy organizations by the amounts obtained through their consulting activities. For example, Deloitte increased its revenues from consulting by 13.5% compared to just 6.1% for audit. These facts lead one to question the independence of such companies and the audit quality which is associated with the Big 4. Moreover, in 2012, the Public Company Accounting Oversight Board revealed criticism for two of the Big 4 group: KPMG and PwC, while Deloitte is the only Big 4 firm to present a significant development from 2011 to 2012. For KPMG, inspectors noted instances where the audit firm identified fraud situations but failed to complete the audit procedures that addressed them. As concerning PwC, inspectors identified nineteen cases where the firm failed to properly audit internal control and nine cases where it failed to properly audit accounting estimates. These situations tend to show the other side of Big 4 companies, demonstrating that there exist instances where there are limitations in Big 4 audit quality as well.
The question of Big 4 superiority is important since many studies rely on the Big 4 versus Non-Big 4 separation as a proxy for audit quality. There is a vast literature studying the relationship between discretionary accruals and auditor type. For example, Becker et al. (1998) discover that Big 4 clients disclose lower absolute discretionary accruals as compared to the clients of Non-Big 4 while Francis et al. (1999) reveal that Big 4 auditors constrain opportunistic reporting because their clients have higher total accruals but lower discretionary accruals. In addition to the papers mentioned above, Lawrence and Zhang (2011) try to provide an answer to the steep separation between Big 4 and Non-Big 4 auditors but their results demonstrate that the treatments of Big 4 auditors are insignificantly different from those of Non-Big 4 auditors in what concerns discretionary accruals.
After a proper review of the literature concerning the separation of auditor types, I decide to select the segregation between Big 4 and Non-Big 4 as a proxy for audit quality as mentioned in the arguments above.
Page 11 of 68 2.3. Earnings management
Earnings management represents one of the main topics of my paper because the purpose of my research is to investigate how audit companies decide to end a business relationship with their clients when they discover earnings manipulation attempts. My goal is to distinguish between the two types of earnings management and thus, a very important aspect is to provide different fields of research for each of them in particular.
Prospect theory plays an important role in the section referring to earnings management. This theory was first presented in the paper of Kahneman and Tversky (1979). The core of this doctrine is explained by the tendency of people to underweight outcomes that are hardly possible as compared with outcomes that are achieved with certainty. This tendency leads to risk aversion in choices concerning sure gains and to risk seeking in choices involving sure losses.
Earnings management has been a well documented area so far. Different authors have focused on presenting the theory that firm managers have incentives to manipulate reported earnings in order to avoid reporting losses, mainly because of the need for “consistent profitability” and also because of the effect of this disclosure on the stock market through an aggressive reduction of the share price. Prior research, Carslaw (1988), Thomas (1989) but most of all, Hayn (1995) report the existence of a “point of discontinuity around zero”. This means that many companies concentrate their efforts in obtaining small earnings, close to zero while there are few cases when companies report small losses. This idea is rounded off by the compelling evidence found in the paper of Burgstahler and Dichev (1997) which stress the fact that 8% to 12% of the firms with small pre-managed earnings use discretion in order to disclose increases in earnings and, furthermore, 30% to 44% of the companies with slightly negative pre-managed earnings exercise discretion to report positive returns. There are two possible explanations for this behavior. The first one refers to the desire of managers to avoid reporting losses in order to prevent high costs associated with transactions with stakeholders. The second refers to the aversion managers encounter to the absolute and relative losses introduced by the prospect theory.
Another aspect that motivates earnings management is managers’ compensation. DeGeorge, Patel and Zeckhuser (1999) show the relationship between earnings and CEO manipulation is influenced by the fact that managers’ salary and more importantly, their bonuses are strictly tied to the reported earnings during their mandate. Moreover, option-based
Page 12 of 68
compensation exhibits incentives with respect to current stock prices as documented by Yermack (1997) and Aboody and Kasznik (2000) because management has an incentive in lowering the stock price near the dates of stock option award.
Moving from the motivations to engage in earnings management to more practical aspects, one needs to differentiate between two types of earnings management: accrual-based earnings management which has been the center of attention for many years and real earnings management, an area which still requires consideration.
Accrual-based earnings management focuses on the management’s use of discretionary accruals. Since it is an area of such importance, many authors have focused on identifying the method that really captures the essence of this practice. In this respect, one of the papers that distinguishes itself from others is the paper of Dechow et. al (1995). They analyze alternative models in order to discover which is best for estimating discretionary accruals and, after conducting different tests conclude that the method which guarantees the best outcome is the Modified Jones Model.
Many authors have focused on explaining discretionary accruals through the Modified Jones Model. I rely on prior literature when I proxy for discretionary accruals, for example the work of Cohen et. al (2008) which compared the two types of earnings management in different time periods and concluded that firms which achieved important earnings benchmarks used less accruals and more real earnings management.
In order to determine which form of earnings management is better connected to auditor change, I make use of the paper of Cohen and Zarowin (2010). They analyze accrual and real earnings management activities around seasoned equity offerings. It is desirable to separate earnings management into its components because they have different effects on the operating activities of firms. It has been demonstrated that accrual-based earnings management does not have any effect on cash flows as opposed to real earnings management. Thus, from this point of view, real earnings management activities are more harmful in the long-term for the company since it focuses on the manipulation of real operating activities. Moreover, the damage that is made on the company as explained by Roychowdhury (2008) include overproduction in order to reduce the cost of goods sold, offering price discounts in order to increase the level of sales and reducing drastically discretionary expenses in order to report lower expenses and thus higher profits.
Page 13 of 68
Due to the attention accrual-based earnings management draws, companies face higher risks of being exposed to the capital market. Consequently, the company’s tendency is to move towards real earnings management. I build my argument on the paper of Graham et. al (2005) which documents that managers turn to manipulation of real economic actions instead of accounting manipulations and even admit that they will give up positive NPV projects in order to meet short-term benchmarks. Moreover, Cohen, Dey and Lys (2008) demonstrate that accrual-based earnings management increased until the passage of SOX, while real earnings management declined prior to SOX and increased after it was passed. In addition, Zang (2012) shows that companies substitute real earnings management for accrual-based earnings management and vice versa. The decision to engage in real earnings management exists when accrual-based earnings management is constrained due to a higher level of scrutiny of accounting practice and also when there is limited accounting flexibility due to accrual manipulation in prior years. Moreover, companies use more accrual-based earnings management when real earnings management is more costly for them.
2.4. The effect of Sarbanes – Oxley Act
The Sarbanes – Oxley Act of 2002 was enacted as a response to major corporate collapses such as: Enron, Tyco International and WorldCom. The main purpose of this law was to protect the investors which lost billions of dollars when the aforementioned companies went bankrupt. It also tries to restore public confidence in the audit profession which suffered a high impact along with these scandals. The aftermath of SOX shows that the impact of the law on financial reporting was significant in that auditors were not allowed to provide some non audit services (such as: bookkeeping, investment-advisory work) in order to protect their independence. In their paper, Bartov and Cohen (2009) demonstrate that firms which meet or beat analysts’ expectations, in the post-SOX period show a decrease in accrual-based earnings management and an increase in real earnings management. This finding is also supported by the findings of Cohen et al. (2008) who show that firms which achieved significant earnings benchmarks used less accruals and engaged in more real earnings management activities after SOX when compared to the pre-SOX period.
Page 14 of 68
2.5. Litigation risk of auditor, deep pockets theory, auditor and client characteristics
Managers can manipulate accruals in order to portray a strong financial position; however, evidence has shown that the signs of manipulation can be seen at the level of abnormal accruals. In order to obtain a guarantee that the financial statements present a true and fair view of the company, auditors are hired by shareholders and are expected to ensure fair financial reporting. They have an important role in identifying and reducing management attempts of earnings manipulation. If the manager has inflated reported financial results and the auditor does not attenuate these manipulations, then the audited financial reports may not provide adequate warning of impending financial problems. If they fail in pursuing this mission, the probability of being sued increases, thus increasing the litigation risk.
Litigation against auditors represents a complex process, starting with the initial discovery of potential false financial statements, to filling of such lawsuits and concluding with their resolution. Palmrose (1988, pg. 56) documents the relevant relationship between litigation of auditors and the quality of their services, stating that “users can view auditors with relatively low (high) litigation activity as higher (lower) quality suppliers”.
A potential litigation has expensive consequences for an audit firm since it has unlimited liability in connection with cases against them, not only considering the material aspects of a lawsuit but considering the indirect costs as well. According to DeAngelo (1981), high quality auditors have more to lose than other auditors because, for them, their brand name is a very important asset.
Litigation risk is closely related to deep pockets theory. Deep pockets theory plays an important role in my paper by explaining the risk of litigation faced by auditors. Many clients take legal actions against their auditor when an audit failure occurs, but according to Palmrose (1988), legal action is taken by some when there is no audit failure. This occurs because companies see auditors as “deep pockets”, and their “pockets” vary in a directly proportional manner with their services’ level of quality.
Litigation risk of auditors also takes into consideration other factors such as some characteristics of both auditors and clients. First of all, Heninger (2001) showed that auditors could face an increased litigation risk by not succeeding in attenuating managers’ tendency to engage in earnings management. Moreover, litigation risk is higher for auditors whose clients report increasing abnormal accruals, who find themselves in weaker financial condition and are
Page 15 of 68
larger than other clients. Secondly, the above mentioned evidence can be complemented by the paper of Lys and Watts (1994); litigation risk is higher for clients with poor stock price performance and those who receive qualified audit reports. The probability of a lawsuit increases if the client represents a large portion of the auditor’s revenues. It is recommended that auditors should not depend financially on certain clients so that their independence remains intact. Furthermore, Stice (1991) hypothesizes and reveals that auditors’ ability to detect errors in financial statements, which leads to a higher probability of audit failure, is increased also due to the asset structure and the sales growth of the client. The paper of Stice (1991) provides useful evidence regarding auditor litigation risk by showing that the greater the market value and the greater variability of return a client has, the greater chances are that the auditor will be sued.
Previous literature is rich in documenting factors that expose auditors to different levels of litigation risk and audit companies can make use of extensive information to protect them against this risk. Some of the measures they can take at the beginning, as proven by Krishnan and Krishnan (1997) include being more selective when choosing their clients. Moreover, after a proper evaluation, auditors can decide to withdraw from the existing engagements which are high risk. Further on, they can increase their audit fees by including a premium associated with this risk or can increase the number of modified opinions they issue in order to signal to the market the company’s riskiness.
2.6. Auditor change: auditor resignations and auditor dismissals
Auditor changes represent the central theme of my research paper and from this point of view it requires a considerable amount of attention.
First, the scandal of Enron and other important companies made the remaining Big 4 companies re-evaluate their client portfolios and terminate contracts with several clients that exposed them to significant risk. Even more, they became more selective towards potential clients. Landsman et al. (2009) demonstrate that neither resignations nor dismissals represent an increase in the sensitivity to client risk in the period following the Enron scandal.
There are many studies that see client changes from the point of view of clients, but not so many investigate auditors’ perspective. However, a paper that studies auditors’ resignation is the paper of Shu (2000) which explains the resignation of auditor in two ways. First, in line with the litigation risk literature presented earlier, auditors decide to resign from client engagements
Page 16 of 68
because some companies expose them to significant risk which can have negative consequences on the long-term. Secondly, auditors have become specialized on certain industries; for example Deloitte presents on its website the list of industries it serves, among others: consumer business, energy and resources, financial services, life sciences and healthcare, manufacturing, public sector, real estate and technology. This implies that audit firms choose their clients depending on their industry, thus they adjust their client mix in such a way that they move towards clients who are better matched with the services they provide.
Auditor change is a process that emerges from different perspectives. Beattie and Fearnley (1995) correctly emphasize in their paper that auditor change can emerge either from client characteristics, auditor’s characteristics but also from the audit environment; any change in one of these factors can trigger a change in auditors. For example, Kohler and Ratzinger – Sakel (2012) identify three main factors that characterize the audit market in Germany and can be used as an example for this research: endogeneity effects, effects of the audit market segments and institutional setting effects. For example client characteristics such as: firm size, the level of debt and management share ownership, influence a company’s decision regarding its auditor. Moreover, Beattie and Fearnley (1995) provide some examples of auditor characteristics that can influence this decision, such as: the audit team and the association between the auditor and client structures. The paper of Beattie and Fearnley (1995) is of particular importance to my study because it brings useful evidence about auditor change in UK. The results of their survey which are of most interest to me are represented by the client’s dissatisfaction with the audit quality (the ability of the audit company to detect errors in the financial statements, ranked second) and disagreements over the accounting policies used (ranked fifteenth).
2.6.1. Auditor resignations
The overall audit environment suffered a lot of change in the post 2000s and audit companies have become more conservative regarding their clients due to deep pockets theory. This conservative approach and, in addition, the presence of SOX has lead to an increase in the number of resignations, especially for Big 4 auditors. The number of resignations increased substantially after 2003, Rama and Read (2006, pg. 2) state: “the number of SEC audit client resignations by the Big 4 increased from 103 in 2001 to 140 in 2003”. The turbulent period the audit market faced, together with the overwhelming media attention aimed at auditors has forced
Page 17 of 68
them to give up a significant number of clients. However, positive outcomes resulted from this, mainly benefitting smaller audit firms. After resignation, many companies opted for smaller auditors; the papers of Rama and Read (2006) and Shu (2000) support this by revealing that companies have headed towards this direction.
Auditors tend to drop clients more often when they identify high levels of real earnings management in order to protect themselves from increased levels of litigation risk. Kim and Park (2014) find that auditor resignations occur when they discover real earnings management attempts. Real earnings management attempts can be separated into different actions: acceleration of the timing of sales through increased price discounts or more lenient credit terms, reporting of lower costs of goods sold through increased production and decreases in discretionary expenses (advertising, research and development, selling, general and administrative expenses). In addition, Kim and Park (2014) find that auditor resignations occur when they discover real earnings management attempts through decreases in discretionary expenses and acceleration of the timing of sales.
Prior literature has documented the ability of auditors to restrict accrual-based earnings management. However, due to the fact that companies have the tendency to report higher earnings, they will attempt to replace incumbent auditors with audit companies that allow them use income-increasing abnormal accruals. Based on this argument and on the aforementioned statements, my first hypothesis is:
H1: There is an increased probability that the auditor, in particular a Big 4 auditor, will change
as the detected accrual-based earnings management increase.
Later, I want to provide more thorough evidence on auditor changes by separating them into auditor dismissals and auditor resignations; thus the following two sub-hypotheses are tested:
H1a: There is an increased probability that the auditor, in particular a Big 4 auditor, will resign
as the detected accrual-based earnings management increase.
H1b: There is an increased probability that the auditor, in particular a Big 4 auditor, will be
Page 18 of 68 2.6.2. Auditor dismissals
According to SEC (1974) disagreements are defined as “disputes that occur at the decision-making level and involve accounting principles or practices, financial statement disclosures or auditing scope or procedures”.
Clients who are in a poor financial condition have incentives to portray a positive image of the company using different methods. This can occur due to the fact that when companies have deteriorating financial conditions they incur higher costs than other companies; an example of such a cost is the cost of borrowing capital. Therefore, in order to minimize these costs, companies tend to prefer an auditor who permits them to inflate earnings. However, the interests of these two parties are not always aligned. On one hand, one desires to reduce its costs or meet certain benchmarks while, on the other hand, the other party concentrates on reducing the litigation risk to which it is exposed. Dwaliwal, Schatzberg and Trombley (1993) demonstrate that disagreements between auditors and clients, preceding auditor change are negatively influenced by the client’s financial condition.
In addition to this, Ettredge et al. (2007) reveal that dismissals are associated with companies that have going-concern reports and which lack strong internal controls.
Thus, my second hypothesis is as follows:
H2: There is an increased probability that the auditor, in particular a Big 4 auditor, will change
as the detected real earnings management increase.
As in the case of the first hypothesis, I divide the sample in two particular cases, auditor dismissals and auditor resignations.
H2a: There is an increased probability that the auditor, in particular a Big 4 auditor, will resign
as the detected real earnings management increase.
H2b: There is an increased probability that the auditor, in particular a Big 4 auditor, will be
dismissed as the detected real earnings management increase.
I will now discuss the methodology employed to test the aforementioned hypotheses. Section 4 presents the descriptive statistics, correlation matrixes between the variables used and additional analyses in order to obtain a better understanding.
Page 19 of 68 3. Research design, sample selection and empirical models
The structure of this chapter is as follows. First of all, I will begin with a concise description of the sample choice, the databases used in order to collect my sample and an indication of the data excluded from the initial sample. The next step is to explain the research model and the variables used to test the research questions.
3.1 Sample selection
The research is intended to be a quantitative research situated in the area of auditing and financial accounting research and is designed by combining theoretical constructs from both areas.
First and foremost, I start by collecting the data I am going to use in my research from Compustat and AuditAnalytics about U.S. companies over the period January 1st 2003 – December 31st 2008. I chose to begin with the year of 2003 because this is the first year after the Sarbanes-Oxley Act implementation. This can be a valuable piece of information in order to obtain information regarding a change in practices as an understanding of the lessons from the Arthur Andersen case. I would like my study to provide information about auditor changes that refer to earnings management attempts until 2008 because I do not want to capture in my research the effects produced on the economic context by the financial crisis. Further, a series of observations from both databases were removed and the two datasets were merged into one. Thereafter, the variables have been imported in Stata 12 statistical package. The last step was to run the regressions and to interpret the results.
The initial sample of companies obtained from Compustat gathered 66,596 firm-year observations while the AuditAnalytics sample contained 10,759 companies who reported changes in their audit firm in the analyzed period. The next step made was to remove the companies which were not headquartered in the U.S. since they are not part of the purpose of my study. This lead to a reduction of the Compustat sample to 50,148 firm-year observations.
Regarding the AuditAnalytics sample, I removed data for which I couldn’t find information about the previous auditor (10), data referring to companies who dismissed their auditor prior to 2003 (272) and data referring to companies for which I was not able to find information about the current auditor (265). Until this point, the final sample consists of 10,212
Page 20 of 68
companies, from which companies who were audited by Big Four companies 3,500 and companies which were audited by Non-Big Four companies (including second-tier and small audit firms) 6,712.
Last but not least, the Compustat sample (which contained financial information about the companies such as: total assets, accounts receivable, book value of equity) and the AuditAnalytics sample (which contained information about the previous auditor, the engagement date, current auditor) were matched into the final version. The matching criterion included a fit by company name, year and also SIC Code as has been done in previous research. I considered necessary a match by SIC Codes since I found companies with similar names but with different economic activities which could create an incorrect image of the companies included in this study. Therefore, from the matching of the companies, the final sample used to test the hypotheses consists of 1,654 companies and 8,685 firm-year observations.
Table 1 reports a full description of the data that I excluded from the sample selection accompanied by a short explanation.
Table 1
Sample Selection
Sample Construction Number of observations
Companies' financial information for period 2003-2008 in Compustat
66,596
Data referring to Non-US companies 16,488
Observations from AuditAnalytics that are not present in the
Compustat sample 33,814
Total number of observations present in both samples 16,294
Observations without fiscal year 38
Observations with different SIC Codes 7,571
Total number of observations 8,685
In addition, the industry to which a company belongs is of great importance to my study, because it has been demonstrated that the probability of earnings management attempts depends on this aspect. For this reason, one of my control variables includes a dummy variable for the company’s industry as found in Cohen & Zarowin (2010).
Page 21 of 68 Table 2
Observations per industry
SIC Code Description Number of Companies
2833 - 2836
Drugs: medicinal chemicals and botanical preparations; pharmaceutical
preparations; etc
111
8731 - 8734 Research and Development and Testing
Services 6
7371 - 7379 Computer Programming, Data
Processing and Other Computer Related 108
3570 - 3577 Computer and Office Equipment 26
3600 - 3674
Electronic and Other Electrical Equipment and Components except
Computer
116
Total 367
3.2. Research models and variables
In order to test the hypotheses mentioned earlier, I use the Probit regression model because in all cases, my dependent variable (Auditor Change and further, Auditor Resigned/Auditor Dismissed) are dummy variables.
My research is based on the separation of earnings management in accrual-based earnings management and real earnings management. The purpose of my paper is to identify if there was any auditor change related to accrual-based earnings management and then sub-divide at the dismissal and resignation levels. After I do this, the next step is to analyze the change in audit firms regarding real earnings management activities and also in this case sub-divide at the dismissal and resignation levels.
3.3. Empirical model
Hypothesis 1
Based on the above discussion, I estimate the following empirical model for hypothesis 1: Auditor_Ch= + Abn_Accruals + Auditor_Type + FD + ROA + +
Page 22 of 68 Table 3
Explained variables
Variables Detailed explanation Data Source
Dependent Variable
Auditor Change A dummy variable which takes value 1 if the auditor
has changed and 0 otherwise AuditAnalytics Independent/ Explanatory
Variables
Abnormal Accruals Abnormal accruals determined based on the
Modified Jones Model Compustat
Auditor Type A dummy variable which takes value 1 if the
current auditor is Big 4 and 0 otherwise AuditAnalytics
Control Variables
Financial Distress
Financial distress, computed using Altman's Z-score; Z-score = 0.012*X1 + 0.014*X2 + 0.033*X3 +
0.006*X4 + 0.999*X5 Compustat
ROA Return on Assets computed as a ratio of net income
and total assets Compustat
Litigation
A dummy variable which takes value 1 if the SIC Codes are: 2833-2836; 8731-8734; 7371-7379; 3570-3577; 3600-3674 and 0 otherwise
Compustat
Market Capitalization
Market value of equity/ market capitalization, computed as a product of number of shares outstanding and price close annual fiscal
Compustat Book to Market Equity Book to Market, computed as a ratio of book value
of total equity and market value of equity Compustat
Leverage Debt to equity ratio Compustat
Sales Growth (Net sales 2003 - Net Sales 2002)/Net sales 2002 Compustat
Size Log of total assets Compustat
My final sample comprises 8,685 observations for the change in auditors. These observations refer to the period 2003-2008. However, in most cases, the auditor has not changed every year, thus in the years in which there was recorded a resignation or a dismissal, the variable “Auditor Change” received value 1 and if in a specified year, the was no movement regarding the audit company, “Auditor Change” receives value 0. Therefore, from the total sample, only 2,010 observations refer to a dismissal or a resignation of the auditor, more specifically, there are 513 resignation cases and 1,477 cases of dismissal in my final sample.
Page 23 of 68
In order to test hypotheses 1a and 1b, the empirical model remains the same with only one difference. In these cases, the dependent variable “Auditor Change” will be replaced by “Auditor Resigned/ Auditor Dismissed” when the previous auditor resigned from the engagement or was dismissed. A representation can be found in Table 4.
Table 4
Explained variables
Variables Detailed explanation Data Source
Dependent Variable
Auditor_Resigned/ Auditor_Dismissed
A dummy variable which takes value 1 if the auditor has resigned and 0 if the auditor was dismissed from the engagement by the client
AuditAnalytics
The primary independent variable in my empirical model is “Abnormal Accruals” and it is determined using the Modified Jones Model. The reasoning behind this is based primarily on the study of Dechow et al. (1995) which evaluated the ability of different models existing in practice to detect earnings management attempts. There were various models tested, such as: the Healy Model, the DeAngelo Model, the Jones Model, the Modified Jones Model and the Industry Model. The results of their analysis showed that all the models taken into consideration appeared to produce good results for a random sample of event years but the model which demonstrated the most powerful tests concerning earnings management was the Modified Jones Model.
I make use of this model by firstly computing total accruals using the approach found in Cohen et al. (2008). Total accruals are of two types: normal and discretionary accruals but the type that is of interest to my paper are the discretionary accruals. Secondly, I calculate the normal accruals using the same approach and thus, the difference between total accruals and normal accruals represent my discretionary accruals.
Based on Dechow (2005), the total accruals are estimated as follows: = + + + (1)
Where for a year t and a company i:
- ; EXBI is the earnings before extraordinary items and discontinued operations and CFO represents the operating cash flows (from continuing operations), computed
Page 24 of 68
as a difference between “Operating Activities – Net Cash Flow” (Compustat symbol: OANCF) and “Extraordinary Items and Discontinued Operations” (Compustat symbol: XIDOC)
Total assets;
= Change in revenues from the previous year; = Gross value of property, plant and equipment
Based on the coefficients from Equation (1), I can determine the normal accruals, as shown by the following regression:
= + +
Where is the change in accounts receivable from the previous year. Following the methodology already used in prior literature, for computing discretionary accruals I adjust the revenues reported by firms for the change in accounts receivable in order to capture any accounting discretion arising from credit sales. Therefore, the measure of discretionary accruals is the difference between total accruals (as mentioned above: / ) and .
The second independent variable I am going to include to research the auditor change is “Auditor Type” because many studies have introduced the problem of audit quality being related to earnings management. For example, in the study of Becker et al. (1998), the relation between audit quality and earnings management is tested. Their results reflect my expectations, that higher quality auditors (Big Four auditors) are more likely to capture earnings management attempts and also, are more likely to object to any potential fraudulent accounting choices of management in order to increase earnings to satisfy their own personal objectives. Hence, I expect that companies who are not audited by a Big Four audit firm to have higher discretionary accruals as compared to those audited by Big Four companies.
Control variables
The first control variable used in my empirical model is “Financial Distress”. In order to determine its value, I use Altman’s Z-score to compute companies’ financial distress score. The reason I chose this measure lies in its components.
Working Capital/ Total Assets (X1) refers to the liquidity aspect of a company. If a company has financial difficulties, it will reduce its current assets in relation to the total assets. Moreover, from all the liquidity ratios evaluated, this one is considered to be the most important.
Page 25 of 68
Retained Earnings/ Total Assets (X2) focuses primarily on the size of the company because it stresses the idea that younger firms need time to consolidate their profits and thus, the probability of going bankrupt increases with size.
Earnings before Interest and Taxes/ Total Assets (X3) emphasizes the importance of the earnings power of a company’s assets to continue its operations over a long period of time.
Market Value of Equity/ Book Value of Total Debt (X4) is considered to be a more effective indicator of bankruptcy than others, since this measure indicates how much a company’s assets can decline in value before the liabilities outrun the assets and the company becomes insolvent.
Sales/ Total Assets (X5) illustrates a company’s sales capacity to generate assets and its importance is based on the ability of management to deal with other competitive forces.
Therefore, the model for financial distress used in my research, as emphasized in the paper of Altman (1968) is:
Z-score = 0.012*X1 + 0.014*X2 + 0.033*X3 + 0.006*X4 + 0.999*X5
In order to determine in which category a company is located, a series of intervals must be discussed. A company is included in the “Safe” zone if the score is higher than 2.9, if the Z-score is between 1.22 and 2.9, a company is situated in the “Grey” zone and last, if the Z-Z-score is less than 1.22, a company is in financial distress.
The control variables included in my test can be divided in multiple categories.
A company’s external stakeholders expect the independent auditor to limit the earnings management trials that might occur and rely heavily on the auditor report they issue. However, if the company goes bankrupt, there is a high probability that the audit company will be sued, due to their “deep pockets”. Therefore, an aspect that is relatively important to my study is litigation risk because auditors might decide to give up a client or a potential client because of the litigation risk it involves. In order to control for this aspect, I took into consideration a wide range of factors that could influence litigation risk, such as: tenure, the size of the client, the industry in which the client operates and its growth.
Previous studies such as Stice (1991) and Carcello and Palmrose (1994) have shown that the tenure plays an important role in a company’s litigation risk. Stice (1991) showed that the risk an auditor faces is higher in the early years of the engagement due to the unfamiliarity of the
Page 26 of 68
auditor with the clients’ operations and that it takes time for an auditor to understand the client’s business.
A clients’ high growth is a relevant aspect that must be understood properly. Firms that fall in this category might incur difficulties in establishing and maintaining internal controls as compared to low-growth companies. As the study of Watts (1994) shows, clients experiencing high growth pose higher litigation risk to the auditor. In order to control for this, I use a variable named “Sales Growth” computed as the annual percentage sales growth in total sales.
Another factor that could influence the litigation risk faced by the auditor is represented by the industry in which the client operates. As demonstrated by Watts (2003) and implemented by Cohen and Zarowin (2010), I control for the industry using “Litigation” which is a dummy variable which takes value 1 for industries with SIC Codes: 2833 – 2836, 8731 – 8734, 7371 – 7379, 3570 – 3577, 3600 – 3674 and 0 otherwise.
Hypothesis 2
The second hypothesis takes into consideration the real earnings management explanation of the potential auditor change.
As opposed to accrual-based earnings management, real earnings management has not received so much attention even though it is a relevant part of accounting manipulations. The main reason for this is because it is harder to detect and has negative consequences on the subsequent operating performance by influencing future cash flows. Graham et al. (2005) conducted a study in which they analyzed the managers’ opinion regarding earnings management and the results proved that they prefer real earnings management over accrual-based earnings management. Moreover, there is evidence showing that managers perform real economic actions in order to safeguard accounting appearances.
In order to construct my proxy for real earnings management, I will make use of prior literature and focus on three metrics of particular importance regarding this concept, which are:
1. The abnormal levels of cash flow from operations – acceleration of the timing of sales
through increased price discounts or more lenient credit terms. These conditions:
acceleration of sales and price discounts have the effect of increasing sales volume on the short-term but will disappear once the company reverts to the initial prices. However,
Page 27 of 68
even though sales might increase over a short period of time, the cash-flows associated will decrease in the same period.
2. Discretionary expenses – such expenses include those in research and development (R&D), advertising and selling, general and administrative expenses (SG&A). Management could decide to reduce such expenses, which will increase current period earnings but could jeopardize the company’s value creation over time.
3. Production costs – another way in which a company can increase its earnings in a financial period is by increasing the production more than necessary. When this happens, managers can spread the fixed overhead costs over a larger number of units produced and in this way the fixed cost per unit will be reduced.
In order to test these three metrics, three separate regressions will be run based on the model developed by Dechow et al. (1998) and implemented in the research of Roychowdhury (2006).
Firstly, for the abnormal levels of cash flow from operations, the following cross-sectional regression will be run for each year individually:
= k1 + k2 + k3 + , (2) where:
= Cash flow from operations from the current year; = The value of total assets from the previous year; = Sales from the current year;
= The change in sales: current year sales – previous year sales
However, from the real earnings management perspective, I am interested in the abnormal CFOs. This will be computed by difference between the CFO and the normal level of CFO using the coefficients (k1, k2, k3) estimated in Equation number 2.
Secondly, I will explain the regression used for computing the production costs. Production costs are represented by the sum of costs of goods sold (COGS) and changes in inventory during the year. Therefore, the model for COGS is:
= k1 + k2
+ (3)
Next, the model for the change in inventory is presented: = k1 + k2 + k3 + (4)
Page 28 of 68
From Equations (3) and (4), I will express the model for production costs as presented below: = k1 + k2 + k3 + k4 + (5)
Thirdly, the model for discretionary expenses can be described as a linear function of sales:
= k1 + k2
+ (6)
The purpose of these regressions is to construct the real earnings proxy.
As mentioned before, in the second hypothesis I expect auditor change to decrease in relation with real earnings management.
The empirical model used for the second hypothesis is the following:
Auditor_Ch= + REM + Auditor_Type + FD + ROA + + + + + + + ε
Table 5
Explained variables
Variables Detailed explanation Data Source
Dependent Variable
Auditor Change A dummy variable which takes value 1 if the
auditor has changed and 0 otherwise AuditAnalytics Independent/ Explanatory
Variables
REM The three cross-sectional regressions for real
earnings management Compustat
Auditor Type A dummy variable which takes value 1 if the
current auditor is Big 4 and 0 otherwise AuditAnalytics
Control Variables
Financial Distress Financial distress, computed using Altman's
Z-score Compustat
ROA Return on Assets computed as a ratio of net
income and total assets Compustat
Litigation A dummy variable which takes value 1 if the SIC Codes are: 2833-2836; 8731-8734; 7371-7379; 3570-3577; 3600-3674 and 0 otherwise
Page 29 of 68 Market Capitalization
Market value of equity/ market capitalization, computed as a product of number of shares outstanding and price close annual fiscal
Compustat Book to Market Equity Book to Market, computed as a ratio of book
value of total equity and market value of equity Compustat
Leverage Debt to equity ratio Compustat
Sales Growth (Net sales 2003 - Net Sales 2002)/Net sales 2002 Compustat
Size Log of total assets Compustat
Again, as in the first hypothesis, I will separate the study of auditor change in two sub-hypotheses, referring to the possibility that the audit company will resign or it will be dismissed from the engagement.
Table 6
Explained variables
Variables Detailed explanation Data Source
Dependent Variable
Auditor Resigned/ Auditor Dismissed
A dummy variable which takes value 1 if the auditor has resigned and 0 if the auditor was dismissed from the engagement by the client
AuditAnalytics
4.Empirical results 4.1 Descriptive statistics
The purpose of this section is to provide relevant information and summaries about the sample used and the observations that have been made in order to test the main hypotheses. To start with, this section forms the basis for the initial description of the data in order to gain useful insights about the sample and is part of a more extensive statistical analysis.
Regarding the discretionary accruals, I am interested in the mean, standard deviation and the minimum and the maximum, separated in quartiles taking into consideration the variable “Auditor Change”. Table 7, presented below makes a short description of the two cases: the one in which there is an auditor change and the other one which reflects no auditor change. One can observe that the mean of the discretionary accruals is higher when the auditor has been dismissed from the engagement as opposed to the case that occurs when the auditor resigns. This arises because of the higher number of cases in which the auditor is dismissed.
Page 30 of 68 Table 7
Summary measure Auditor Change = 0
--- Quantiles ---
Variable n Mean S.D. Min 0.25 Mdn 0.75 Max
Discretionary Accruals 3991 69.28 733.43 -1.70E+04 -4.79 1.49 27.87 13639.88 Auditor Change = 1 --- Quantiles ---
Variable n Mean S.D. Min 0.25 Mdn 0.75 Max
Discretionary
Accruals 1194 45.83 739.09
-1.30E+04 -6.42 0.25 16.03 15549.86
Descriptive statistics for the change of auditor considering accrual-based earnings management are presented in Table 8. This table contains the mean, standard deviation, the minimum and the maximum of both the dependent and independent variables and also, of the control variables.
Table 8 - Descriptive Statistics – Accrual based earnings management
Variable Obs. Mean Std. Dev. Min Max
Auditor Change 5185 0.23 0.42 - 1.00 Discretionary Accruals 5185 63.88 734.73 (16,998.07) 15,549.86 Big 4/ Non-Big 4 5185 0.06 0.24 - 1.00 Financial Distress 5185 3.49 41.79 (47.73) 2,317.11 Return on Assets 5185 (0.51) 8.25 (342.17) 27.02 Litigation 5185 0.21 0.41 - 1.00 Market Capitalization 5185 1,815.55 8,855.47 0.00 191,641.00 Book To Market Equity 5185 (10.95) 1,259.95 (76,503.87)
39,187.76 Sales Growth 5185 211.10 6,798.36 (101.36) 475,819.80 Size 5185 26.41 45.29 0.05 697.00
Page 31 of 68
According to the data, almost 23 percent of the sample is represented by auditor changes, while the difference indicates that, in the chosen period, there was no auditor change. This is observable because the mean of discretionary accruals is positive. From the total sample, 21 percent of observations refer to companies that operate in high litigation industry, those with SIC Codes: 2833 – 2836, 8731 – 8734, 7371 – 7379, 3570 – 3577 and 3600 – 3674 which is not a negative outcome since the percentage is lower than a quarter of the total observations. It would have been a significant outcome if this percentage was higher. In that case, it would have indicated the activity of companies that pose an increased litigation risk to the auditing company. This is an interesting outcome since the aim of my paper is to focus on companies that operate in normal conditions and are not perceived as being a risky client by the audit company. In addition, one can observe that the mean of Financial Distress is 3.49, higher than 2.9 which indicates that generally, the companies included in the sample do not have significant economic difficulties, being situated in the “Safe” zone of financial condition based on Altman’s Z-score. Regarding the audit quality expressed by the type of auditor (Big 4/ Non – Big 4), one can notice that the mean of Auditor Type is 6 percent. This indicates that the type of the auditor inclines in favor of Non-Big 4 auditors. This aspect is in line with the paper of Ettredge et al. (2007) which documents that companies switch to Non-Big 4 companies. One of the reasons for this reaction is to obtain substantial cost savings by eliminating the higher audit fees demanded by Big 4 companies.
In order to test the strength of the relationships between all the variables employed in each of the models, the correlation matrix was used. As Table 9 shows, there is a negative and not significant relationship (-0.0134) between discretionary accruals and auditor change which does not support my first hypothesis. According to the first hypothesis, the probability that the auditors will change increases as accrual-based earnings management increases. However, the relation between auditor change and the type of the auditor is positive and highly significant (0.461) indicating that it is more probable that Big 4 auditors will change, as I expected. Raghunandan and Rama (1999) demonstrated that Big 6 auditors are less likely to serve as auditors for clients if the previous auditor resigned, thus they will drop clients which meet this criteria. In addition to this, Rama and Read (2006) documented that after the Enron scandal and the introduction of the Sarbanes – Oxley Act in 2002, auditors became more conservative regarding client retention and acceptance decisions. Thus, in the period following these stressful
Page 32 of 68
events, it is expected that Big N auditors will resign more often than in the preceding period. This evidence is, therefore, in line with the strong association between the change in audit firms and the type of the auditor provided by my research.
Among the factors that have a negative influence on the probability that the auditor will change are: size (almost 4 percent) which is significant at the 1% significance level, market capitalization (2.16 percent), return on assets (1.22 percent) and book to market equity (0.2 percent).
However, there are also factors that have a positive influence over the change in audit fees. The most significant one is represented by sales growth (2.2 percent), significant at the 1% significance level followed by leverage (0.3 percent) and litigation (0.26 percent).
An important aspect that needs to be pointed out regarding the correlation matrix is the high, positive influence market capitalization has on discretionary accruals (63.64 percent). Linck et al. (2013) reveal that companies with high-accruals experience a greater rise in stock price at the moment of their earnings announcements. Moreover, these companies issue more equity and raise more debt than the opposite category: low-accrual firms.
