The market reaction when a firm changes
from auditor
Name: Patrick Bakkers
Student number: 10002056
Date: June 21st , 2014
University: Amsterdam Business School
Faculty: Faculty of Economics and Business, University of Amsterdam Master: Accountancy & Control, variant Accountancy
Supervisor: dr. B. Qin
2
Abstract
The aim of this study is to investigate the influence of an expected lower or higher audit fee and the corresponding market reaction at the time of an auditor change. The research question that has been developed is: “how does the market react when a company switches to an audit firm with a lower expected audit fee?”.
The market reaction is measured with an event study and the difference in audit fee is calculated with the audit fee model. With an event study the effect of an event is measured. The event is the auditor change and the moment of this event determined by the date the 8K-Form is filed. The auditor change must have occurred between 2005 and 2013. By using different time windows (3, 5 and 10 days) the Cumulative Abnormal Returns have been calculated. The results of the audit fee model have been used as a proxy to measure the different Cumulative Abnormal Returns.
The audit fee model finds a significant difference in audit fee between the Big 4, medium and small audit firms. The model shows that Big 4 audit firms charge a fee premium. According to the results of audit fee model the switch from a Big 4 to a non-Big 4 or from a medium to small audit firm should lead to a positive market return because of the decreased audit fee costs. Results from the event study show that there is a significant negative market reaction when there are switches from Big 4 to medium and medium to small audit firms. In both situations a positive market reaction was expected. A significant positive reaction is found for switches from small to medium. In this case a negative reaction was expected. For the switch from medium to Big 4 a negative market reaction was expected and confirmed by the results found. The cross-sectional analysis shows however that an expected lower or higher audit fee not significantly influences the reaction of the market.
The conclusion of this study is that there is a significant positive or negative market reaction when the expected audit fee is higher or lower but this reaction is not necessarily an effect of the changed audit fee.
3
Table of content
1. Introduction ... 4
2. Literature review ... 7
2.1. Audit Fees... 7
2.1.1. Determinants of audit fees ... 7
2.1.2. Differences between audit fees ... 9
2.1.3. Conclusion audit fees ... 10
2.2. Market reaction ... 10
2.2.1. Efficient market hypothesis ... 11
2.2.2. Market reaction to auditor switches... 11
2.3. Hypothesis ... 15
2.4. Summary ... 17
3. Methodology ... 18
3.1. Empirical approach and sample ... 18
3.2. Data sample ... 18
3.3. Event study ... 23
3.4. Audit fees ... 25
3.5. Cross-sectional analysis... 28
4. Data Descriptive ... 30
4.1. Descriptive of the audit fee model ... 30
4.2. Descriptives event study ... 33
4.3. Cross-sectional analysis... 35
5. Results... 37
5.1. Audit fee results ... 37
5.2. Market reaction results ... 38
5.3. Cross-sectional analysis... 41
5.4. Summary ... 45
6. Conclusion ... 46
7. References ... 48
8. Appendices ... 52
8.1. Appendix A: auditor specialization ... 52
8.2. Appendix B: audit fee models ... 61
4
1.
Introduction
There are many reasons to change from your incumbent auditor. A lower audit fee is one of the key considerations for an auditee (the firm that hires an auditor) to switch from auditor. Research states that auditees are trying to achieve a lower audit fee because they think they pay too much for the audit quality they get (Beattie & Fearnley, 1995). But there are more reasons to change from auditor.
Last year Aegon NV announced that they had changed from auditor for the annual report of 2014. In their announcement Aegon communicated that the auditor change was a consequence of their corporate governance principles (Aegon, 2013). In the corporate
governance principles of Aegon NV it is stated that the Audit Committee and the Supervisory Board conducts an assessment of the functioning of the external auditor at least every four years (Aegon, 2014). The result of this assessment was the reason for Aegon to change from auditor and they expect that the new auditor will fit better to their expectations. The auditor change they made was from a Big 4 to Big 4 audit firm (Aegon, 2013).
Switches from a Big 4 to another Big 4 auditor are not the only switches we have seen last years. Unitil made the switch from McGladrey to Deloitte (medium to Big 4). The main reason in this case was that the Audit Committee of Unitil Corp. concluded that Deloitte better suites the regulatory requirements and guidelines for external audits (Reuters, 2013). National Western Life Insurance Co. made a switch the other way around. From a Big 4 to a third-tier (small) audit firm. They stopped working with KPMG and contracted BKD. The preference of BKD over KPMG was the outcome of a competitive proposal procedure
(Reuters, 2014). This competitive proposal process can implicitly mean that National Western Life Insurance was in search of a lower audit fee.
These are just three examples of organizations that changed from auditor. Based on these examples the reason to change from auditor and to which type of auditor can be quite diverse.
From the research of Beattie & Fearnley (1995) it can be expected that organizations change from a Big 4 to a non-Big 4 audit firm because a non-Big 4 audit firm charges a lower audit fee. Nevertheless, the opposite has happened. The Big 4 dominate the audit market and it is known that they charge a fee premium (Francis, 1994; Hogan & Wilkins, 2008; Francis & Simon, 1987).
5 Although the audit fee level is a key consideration for an auditee to switch from
auditor the actual reason can be quite different. Some of these reasons have already been given in the examples earlier mentioned. Another reason could be that key stakeholders in the firm might want a higher quality and demand a Big 4 audit firm as auditor. Beside the audit fee premium that Big 4 audit firms charge, the audit quality is also higher (Boone et al., 2010). This theory can explain the current market condition that Big 4 audit firms dominate the audit market.
In the near future it is expected that the number of companies that will change from auditor will significantly increase in the European Union. This is a result of the
recommendations of Barnier. Michel Barnier is European Commissioner for the Internal Market and Services and publisher of the Green Paper: Lessons from the Crisis (Barnier, 2010).
One of the recommendations Barnier made was that it should become mandatory to change from auditor (Humphrey et al., 2011). As result of this recommendation the European Union is planning to introduce the mandatory auditor rotation. In the Netherlands it is already decided that the rotation will be mandatory as of the year 2016 (European Commission, 2014; Dutch Senate, 2012). In the United States of America the PCAOB (Public Company
Accounting Oversight Board) tried to introduce mandatory auditor rotation as well. They were not successful. The House of Representatives voted against the PCAOB proposal (Journal of Accountancy, 2013).
Although there are many reasons why organizations could change from auditor, the level of the audit fee (the lower, the higher the profit) and audit quality are key considerations for an auditor change. Investors, shareholders and other interested parties are mostly
interested in the value of - or in the profit on their shares. Therefore, how the market reacts on an auditor change is an interesting topic for them.
Studies done so far independently researched the audit fee level of auditors or how the market reacts on auditor changes. No research has been done on the correlation between these two topics. Since the interest of investors, shareholders and other interested parties in market reactions, this study aims to provide an understanding on how the market reacts on auditor switches when the expected audit fee level changes (lower or higher). Having this knowledge could result in an active auditor changing policy for both organizations, investors,
6 The following research question has been developed: how does the market react when a company switches to an audit firm with a lower expected audit fee?
The research question will be answered with an event study. The data is gathered with the use of Audit Analytics and Compustat. An event study is a common way to investigate the effect of an event. The event in this study is the auditor switch and the effect is measured with the market reaction. An expected lower audit fee is measured with the audit fee model. A cross-sectional regression analysis has been performed to validate the outcomes of the event study.
The results of this study show that the audit fees are different between audit firms. Big 4 audit firms have significantly higher audit fees than medium and small audit firms. This confirms the presence of the audit fee premium. The defined proxies for the difference in audit fees are used in the event study. The results of the event study show that the audit fee does not influence the market reaction as expected: a higher expected audit fee resulted in a positive market reaction (where a negative reaction was predicted). The cross-sectional regression analysis finds no relationship between an expected lower or higher audit fee and the market reaction.
The contribution of this paper is the increased understanding in differences of audit fees between audit firms and the associated market reaction. The results help to improve the understanding of the market reaction and where the market reaction depends on.
The second chapter starts with the literature review and contains background literature that is used to support and develop the hypotheses. The methodology is the main part of the third chapter of this study. In this chapter it is explained how the research is conducted. Chapter four and five are about the data, descriptive statistics and the results. After the results a conclusion is drawn with an answer on the research question.
7
2.
Literature review
Investors, shareholders and other interested parties are interested in how the market reacts on an economic event. In the first part of the literature review information on the differences in and the determinants of the audit fee will be covered. The following paragraph contains the components that influence the reaction of the market. Based on the theories explained in the literature review the hypothesis of this research is developed. This chapter will be concluded with a short summary.
2.1. Audit Fees
Since Auditees are continuously searching for a lower audit fee, the audit fee is one of the main reasons for an auditor switch (Beattie & Fearnley, 1995). To understand better on how the market responds on an auditor switch there needs to be at first a better insight in the audit fee. In this paragraph the audit fee will be explained. In the first section of this paragraph the determinants of the audit fee and how to measure them will be explained. Following the clarification of the audit fee determinants some differences between audit fees will be handled.
2.1.1. Determinants of audit fees
For most people it is very unclear how audit fees are realized. This section will give a clear guidance on the different audit fee components. With this knowledge a better understanding will be created on how the market reacts when an auditee switches to a new audit firm.
Total audit fees consist of audit and non-audit or service fees. There is a positive correlation between the audit fee and the non-audit fee (Whisenant et al., 2003A). Next to these two components of the total audit fee, researchers have tried to find determinants which explain the audit fees. The determinants found are auditee size, audit(ee) complexity and audit(ee) risk (Simunic, 1980; Francis, 1984) but also auditee profitability, ownership control, timing variables, auditor location and auditor size (Chan et al., 1993). The most important determinants and how to measure them will be handled in the next section.
Auditee size can be measured in different ways, for instance by total assets or by using turnover numbers (sales). Some researchers state that turnover is a better measure because assets can significantly differ between firms. This is a result of accounting policies or age profile. The use of total assets in the audit fee model as well as using total assets in other ratios or calculations can decrease the explanatory power of the audit fee model. On the other
8 hand, turnover isn’t flawless as well. The definition of turnover can differ very much between different industries. As a result of these shortcomings the square roots or (natural) log of total assets are used in the new models. This is done since the relationship between audit fees and auditee size is not linear (Chan et al., 1993).
Increased auditee complexity means, in general, a higher audit fee because the auditor needs more time for the audit. The complexity of the auditee can be explained by the nature of their business, the internal control system or the complexity of the transactions. The proxies that are used are the number of subsidiaries or balance sheet composition measures.
Subsidiaries is a decent proxy because the auditor has to make different financial statements, has to control intra-group transactions and the auditee needs additional monitoring if not all subsidiaries have the same auditor. These are just some reasons why the audit fee increases as a result of the complexity of the auditee. The second proxy for the auditee complexity is the balance sheet composition measures. Balance sheet composition measures are important since some balance sheet components, such as inventory or debtors, are more difficult to audit than cash or fixed assets. Balance sheet composition can be measured through ratios such as inventory to total assets or accounts receivable to total assets (Chan et al., 1993).
Auditee risk is together with auditee size and auditee complexity the most common used determinant for audit fees. Audit firms base their audit risk approach on the risk the auditee faces. Different studies confirmed the relationship between audit fees and auditee risk (Hogan & Wilkins, 2008). Most audit fee models explain auditees risk by using market based measures. The market based measures are liquidity ratio, current ratio, solvability ratio etc. (Chan et al., 1993). Explaining audit risk is a bit harder because this is perceived by the auditor.
Prior research shows a link between auditee profitability and audit fees. Companies that face a low profitability are more inclined to cut costs and as result more control is needed on the overhead costs. This reaction might result in increased audit work because the auditor needs to investigate more. To overcome this problem, shareholders’ equity or a loss indicator is used as a proxy for the auditees profitability (Chan et al., 1993).
The timing variables can be explained in different ways. The auditees may need to pay more if they want some non-audit services in the busy period. This period is between the first of December and 31st of March. This period is the busy season because some auditee years end at the 31st of December or at the 31st of March. The other explanation of the existence of
9 this determinant is the time between the auditees year end and the audit report date. A short lag can mean an inefficient audit which results in a higher fee. A longer lag can imply audit problems which will also result in a higher audit fee (Chan et al., 1993). The difficulty with this variable is how to measure the lag problem. Different studies use different busy periods. For example July instead of December and March (Carson et al., 2004) or only December (Hogan & Wilkins, 2008). Kamp (2002) finds that the most fiscal year-ends are in December. This suggests that the busy period is at least December and maybe even December up and until March. For computing the lag problem the literature has not found any solution. Because of this problem a year-end indicator is the most common proxy for the timing variable.
There is not a consistent answer on the importance of auditor size. Some papers suggest that Big N audit firms charge a fee premium to small auditees (Francis, 1984). Other papers did not find any evidence for a fee premium (Che-Ahmad & Houghton, 1996). Interviews with audit partners suggested the existence of Big N premiums but no significant results are found (Chan et al., 1993). The Big N indicator is used as proxy for the possible influence of the auditor size on the audit fee.
Ownership control is a variable which can influence the audit fee explained by the agency theory. A more complex structure of ownership control requires a better audit. Ownership control is difficult to measure directly, that’s why little research has been done so far (Chan et al., 1993).
Another variable which isn’t very much investigated is the auditor location. This explanatory variable emerged from interviews that in some regions, such as London, the audit fees are higher (Chan et al., 1993).
2.1.2. Differences between audit fees
It is logical that audit fees of large organizations are higher than the audit fee of small organizations. To compare these audit fees researchers use the natural logarithm of the total assets or in lesser extent the audit fee/total asset ratio (Simunic, 1980; Craswell & Francis, 1999; Carson et al. 2004; Whisenant et al., 2003A). The researches have all in common that the way the audit fees is determined is based upon some of the explanatory variables which were explained in the first section of this paragraph. The only explanatory variable that suggests that more research should be done is the auditor size in relation to the auditee size. This conclusion is based upon the finding that the studies are not consistent in their results. The theory suggests that the bigger the auditee, the more complex the financial statements
10 will be and the more time it will take to complete the audit (Simunic, 1980; Chan et al., 1993). This theory is supported by the research of Carson et al. (2004). They found evidence that suggests that Big N audit firms charge fee premiums to small organizations. Another reason why audit fees can vary is product differentiation. This product differentiation is a possible explanation for the difference in audit fee of Big N versus non-Big N audit firms (Francis, 1994; Hogan & Wilkins, 2008; Francis & Simon, 1987).
Audit firms invest to create industry expertise. Because of this investment you could expect a higher audit fee. However, these audit firms aim to get more clients which will result in economies of scope, thus lower audit fees. Crasswel et al. (1995) found that specialist Big N audit firms charge higher audit fees than non-Big N specialists. According to Ferguson et al. (2003) the findings of Crasswel et al. (1995) are no longer valid. They showed in their research that firm-wide expertise does not exist. It exists on a local office-wide or city-wide level. The top two audit firms in an industry only charge a fee premium when they are an industry expert at office-wide level (city leader). They also believe that you can expect the same results in the USA because the US audit market is even more dispersed than the Australian market.
2.1.3. Conclusion audit fees
Throughout the years the models to explain audit and non-audit fees have changed. It started with Simunic in 1980 with a model of 10 different variables and in 2003 there was already a model which consisted of 22 different variables. All these variables are derived from the theory that both audit fees and non-audit fees are influenced by factors such as auditee size, auditee complexity, auditee risk, auditee profitability, timing, auditor size and in a lesser degree ownership control and auditor location.
Prior literature does not show a clear picture on the height of audit fees. Some researchers suggest that the audit fee charged by Big N firms to small firms is relatively higher than when you compare audit fees charged by Big N audit firm to other big firms. Other researchers suggest that there is no evidence for this suggestion. Industry expertise does not result in a higher audit fee on national firm-wide level because industry expertise only exist on a local firm-wide level.
2.2. Market reaction
In this section a couple of different scientific papers will be covered to help explaining how the market reacts on an auditor switch. This paragraph starts with a section on the introduction
11 of the efficient market hypothesis. The efficient market hypothesis is used to explain how the market should react. Following this explanation the actual market reaction is handled. The similarity or difference in the reactions from the market forms the basis of the hypothesis that will be formulated.
2.2.1. Efficient market hypothesis
The easiest way to explain what the efficient market hypothesis is (Fama, 1991), is by stating that all information from the market is covered in the security price. Nowadays there are some strong and weak versions of this hypothesis. In the strong version of the information
hypothesis, the information and the trading costs are zero. In the weaker versions there should be a balance between the marginal benefits and the marginal costs.
The stock market cannot be perfectly efficient. Different studies suggest that the stock market reactions are sometimes predictable. This is based on price irregularities and
predictable patterns in stock returns (Malkiel, 2003). After an information announcement the market reacts very quickly and the stock price changes almost immediately (Fama, 1991). In an efficient market all publicly available news should be reflected immediately in the stock price. Despite what the theory expects, the market responds faster to ‘bad news’ than ‘good news’ (Basu, 1997).
When the markets work perfectly efficient there should not be any difference between the response time of good news and bad news. According to the efficient market hypothesis both good and bad news is information and all information should directly and equally influence the stock price. It can be concluded that the current security markets are not perfectly efficient.
2.2.2. Market reaction to auditor switches
In the USA a lot of research has been done on the market reaction to a company’s change of an audit firm. The results of the researches are not consistent with each other. One of the explanations for this inconsistency is the time window the researchers use. The use of the right time window is very important and affects the outcome of the research (Scott, 2011).
Using a 21-week time window resulted in a negative market reaction (Fried & Schiff, 1981). There was however no significant reaction when using an eight-week window (Nichols & Smith, 1983). Johnson & Lys (1990) also didn’t find any significant stock reaction when using a three-day time window. All these researches where conducted in the time period 1972
12 – 1982. Other researchers found negative stock reactions when using different time windows but they all investigated at negative events such as auditor resignations or weak internal controls (Chang et al., 2010).
Whisenant et al. (2003B) investigated the market reaction to disclosures of reportable events. One of the investigated events is the auditor change. Within their study the 8-K Form filing date was used as the event date. For calculating the Cumulative Abnormal Return they used a three and seven-day time window. They found a negative market return for both windows.
Chang et al. (2010) found in their study that more companies switched to small audit firms than to a Big 4 audit firm. They formulated small audit firms as all audit firms except Deloitte (Big 4), Ernst & Young (EY) (Big 4), KPMG (Big 4), PwC (Big 4), BDO Seidmand (medium) and Grant Thornton (medium). In their study two time periods were used: the period before and after August 23rd, 2004. They made this distinction because of three
reasons: (1) the introduction of SOX section 404 requirements, (2) PCAOB inspection reports and (3) the changed deadline of filing the 8-K Form. In their research they calculated the Cumulative Abnormal Returns over the event window and compared these results with the actual market return. The event window they used was five days. Four days before the filing of the 8-K Form up and until the date of filing in the 8-K Form. This is done because of the maximum of four business days for filing the 8-K Form after the auditor switch. They found in the second period a non-negative market response when companies changed from Big 4 to small, Big 4 to medium and Big 4 to Big 4. In all other cases they found a negative return. A non-negative market response can mean no response or a positive response. There is a positive market reaction when companies were looking for better audit services.
Researchers who investigated the different characteristics such as industry
specialization found that a switch to an industry specialized audit firm resulted in a positive market reaction and a switch from an industry specialized audit firm to a non-specialist audit firm resulted in a negative reaction (Knechel et al., 2007). This is consistent with the efficient market hypothesis because a switch to an industry specialist can be explained as good news while a switch to a non-specialized audit firm can be explained as bad news.
The actual market reaction and the expected market reaction can be the same. The expected market reaction is based on the efficient market hypothesis. For example, for the switch from non-specialized to specialized auditors the actual and the expected market
13 reaction are equal. The non-negative reaction which Chang et al. (2010) found cannot be explained with the efficient market hypothesis. According to the efficient market hypothesis and the audit fee literature a switch from Big 4 to medium or small should lead to a lower audit fee and thus a positive market response. This is because of the audit fee premium a Big 4 audit firm charges to their clients. The non-negative return that has been found can mean a positive return or no return. Thus the actual market return cannot completely be explained by the efficient market hypothesis theory.
Table 1 summarizes the different studies that have been examined. The author(s), year of publishing, the journal and important information such as the length of the test window and the findings is displayed.
14 Table 1: Summary market reaction studies
Authors Year Journal Test
window
Findings
Fried & Schiff 1981 The
Accounting Review
21 weeks Negative market reaction
Nichols & Smith 1983 Journal of Accounting Research
8 weeks No significant reaction
Johnson & Lys 1990 Journal of Accounting and Economics
Three-days No significant market reaction
Whisenant et al. 2003B A Journal of Practice & Theories
Three-days and seven-days
Negative market reaction to an auditor switch.
Knechel et al. 2007 A Journal of Practice & Theories
Three-days Switch from non-specialist to industry specialist results in a positive market
response and a switch from industry specialist to a non-industry specialist in a negative market reaction.
Chang et al. 2010 A Journal of
Practice & Theories
Five-days Non-negative market response when companies changed from Big 4 to any other audit firm. When companies were searching for better audit services than there was a positive market reaction.
15
2.3. Hypothesis
The hypotheses used in this research are based on different market reaction theories and the audit fee theory. One of the most important reasons for an auditee to consider an auditor change is the audit fee (Beattie & Fearnley, 1995). The ultimate objective of an auditee is to establish a lower audit fee combined with a higher audit quality or service. This is because an auditee wants to get the highest possible audit quality but also wants to realize lower costs (Beattie & Fearnley, 1995). According to the efficient market hypothesis theory, the market response after the switch is positive when the audit fee is lower. However an auditor switch from non-Big N to Big N (according to the theory this implies a higher audit fee) can result in a positive market reaction as well (Knechel et al., 2007). This reaction is unlike what the efficient market hypothesis suggests. The efficient market hypothesis predicts that a higher audit fee leads to a negative market response. The positive reaction can be explained with higher audit quality.
Audit quality can be split into an actual audit quality and a perceived audit quality. Boone et al. (2010) compared medium sized audit firms with Big N audit firms. They concluded that there is little difference in actual audit quality but that he difference in perceived audit quality is far higher. Medium sized audit firms have a lower perceived audit quality but a higher ex ante equity risk premium compared to Big N audit firms. Equity risk premium is the excess of the company-specific ex-ante cost of equity capital over the risk-free interest rate (Boone et al., 2008). In this case the ex-ante equity risk premium is a proxy for the perceived audit quality for investors. So the investors think that Big 4 audit firms deliver higher audit quality than medium size audit firms (Boone et al., 2010). A switch from medium to Big N results in a positive market reaction despite the higher audit fees and the lower ex ante equity risk premium. Boone et al. (2010) found that the actual audit quality of Big N and medium size audit firms is the same and because of the lower audit fees of medium sized audit firms, the switch from Big N to medium should lead to a positive reaction
Concluding the theories discussed the following hypotheses can be defined for this research: Hypothesis 1: The switch to an audit firm that charges a lower audit fee will lead to a positive market reaction.
This hypothesis can be explained by using the efficient market hypothesis and previous studies. Boone et al. (2010) explained that there is no difference between the actual audit quality of Big 4 and medium sized audit firms. With this assumption the first hypothesis
16 is formulated. If the audit quality stays the same lower audit fees are perceived as positive news and the market reacts accordingly positive. This hypothesis also suggests that a higher audit fee results in a negative market response. In contrast to the first hypothesis, there is also some evidence for the opposite reaction, a higher audit fee results in a positive market
response
Hypothesis 2: The switch to an audit firm with a higher audit fee leads to a positive market reaction.
The second hypothesis is the opposite of the first hypothesis and has a different supporting theory. A higher audit fee after an auditor switch may indicate a higher level of assurance or audit quality which can results in a positive market reaction.
17
2.4. Summary
Enclosed table provides an overview of the literature reviewed. The first row summarizes the most important findings on the audit fee. The second row contains the conclusions on how the market should react according the efficient information hypothesis and how the market did react. The last row provides the hypotheses formulated.
Table 2: Summary of the literature reviewed Audit Fees Different explanatory variables:
auditee size, auditee complexity, auditee risk, auditee
profitability, timing, auditor size, ownership control and auditor location.
Mixed results about higher audit fees when the auditor is a Big N firm.
Industry expertise does not result in a higher audit fee on national firm-wide level. Unless they are also industry experts on local-level.
Market Reaction
The market should react positive on a lower audit fee according to the efficient market hypothesis and negative to a higher audit fee. Unlike the efficient market hypothesis suggests, a higher audit fee combined with higher audit quality can be experienced as good news as well because of the higher audit quality. It is experienced as good news because of the positive reaction.
Switching to an industry specialist auditor results in a positive market reaction. This is interesting because industry specialization can result in a higher audit fee. Switching from Big N to non-Big N results in a negative market reaction.
Hypotheses 1. Switching to an audit firm which charges a lower audit fee, leads to a positive market reaction.
2. Switching to an audit firm with a higher audit fee, leads to a positive market reaction.
18
3.
Methodology
Within this paper an event study has been applied as a quantitative research methodology. The next paragraphs cover the empirical approach and the data sample that has been used. The empirical approach consists of the event study of the auditor change and the audit fee model.
3.1. Empirical approach and sample
This paper will apply a quantitative research and to be specific an event study. The event in this study is the auditor switch. The moment that is used to determine when an auditor switch has taken place is the date at which the 8K-Form is filed (SEC, 2012).
Because of some major adjustment to the SOX regulation in 2004 the sample will contain the available data from American firms from 2005 up and until 2013. 2005 is chosen as a starting point because of the combination of the new SOX 404 requirements on auditing of internal controls over financial reporting and the introduction of the Auditing Standard No. 2 from the PCAOB (Chang et al., 2010). In 2004 all firms were required to report on the effectiveness of their internal controls (Klamm & Watson, 2009). Because of the
incomparableness of the financial statements from the financial sector, these firms have been excluded from the sample (NAICS: 520.000-529.999 (NAICS, 2014)).
Important for an event study is the choice of the right time window. To find the causation between an economic event and the market reaction, a narrow time window is needed (Scott, 2011, pp. 160-162). To find the right causation some different time windows will be used.
The structure of the research consists of three parts. The first part is how to conduct an event study, the second part covers how to predict audit fees and the third part contains the defined hypotheses that will be tested. A combination of Audit Analytics and Compustat will be used to gather the data. In the next paragraph the sample selection is explained.
3.2. Data sample
The data on auditor changes is gathered through Audit Analytics. In the period 2005-2013 3,072 firms changed from auditor. These firms are traded at the stock exchange in the USA. After deleting missing data the sample is reduced to 1,855 auditor changes. The distribution of the changes is displayed in table 3.
19 Table 3: Auditor switches
To Big 4 Medium Small
From
Big 4 545 235 308
Medium 88 24 113
Small 99 116 329
To get a more detailed understanding of the movements table 4 shows the changes to the different audit firms. The Big 4 are losing more clients than they get new clients. The number of clients of medium and small audit firms is growing. For example PwC is losing 308 clients and gets 170 new clients. BDO USA loses 71 clients and gets 124 new clients. Table 4: Auditor switches between audit firms
To
From PwC Ernst & Young Deloitte & Touche KPMG Thornton Grant BDO USA McGladrey Small Total
PwC 0 50 45 47 39 14 17 96 308 Ernst & Young 49 0 39 58 32 22 6 69 275 Deloitte & Touche 68 55 0 30 31 15 5 66 270 KPMG 22 52 30 0 30 15 9 77 235 Grant Thornton 11 9 9 16 0 10 1 56 112 BDO USA LLP 5 8 6 7 5 0 2 38 71 McGladrey LLP 1 5 3 8 4 2 0 17 40 Small 14 36 20 29 29 46 41 329 544 Total 170 215 152 195 170 124 81 748 1,935
To define Big 4, medium and small firms the defined sample and existing literature will be combined. Industry specialists will be identified on only the sample. To define industry specialization the market share in each industry sector in each year will be used. To be recognized as a specialist you need at least 30% market share, be market leader in the region and you need to be one of the two market leaders in the USA (Knechel et al., 2007; Ferguson et al., 2003).
The Big 4 firms are according the literature: PWC, Deloitte, EY (Ernst&Young) and KPMG. To define the medium firms is more difficult. Based on the information of two USA organizations who investigated the net-revenue of accounting firms in the USA, the medium
20 accounting firms are Grant Thorton and McGladrey (IPA, 2013; Statistics, 2013). These firms are medium sized firms because their revenue is much smaller than the Big 4 firms but much higher than BDO, the 7th accounting firm in the USA. GAO (United States Government Accountability Office) stated McGladrey, Grant Thornton, BDO Seidman and Crowe Chizek & Company as midsize audit firms in 2008 (GAO, 2008). Other authors suggest that BDO Seidman and Grant Thornton are medium sized audit companies (Chang et al, 2010). Small audit firms are all the audit firms that not belong to the Big 4 or medium sized audit firms. Table 5: Ranking auditors based on audit fees 2005-2013
# Auditor name Audit fees % Market share based on audit fee
BIG 4, Medium, Small classification 1 PricewaterhouseCoopers $ 30,821,193,366 29.4%
BIG 4 2 Ernst & Young $ 24,952,576,049 23.8%
3 Deloitte & Touche $ 24,088,794,514 23.0%
4 KPMG $ 17,480,865,650 16.7% 5 Grant Thornton $ 1,937,028,444 1.8% Medium 6 BDO USA $ 1,384,021,296 1.3% 7 McGladrey $ 415,631,272 0.4% 8 Crowe Horwath $ 266,249,874 0.3% Small 9 Moss Adams $ 162,012,233 0.2% 10 Marcum $ 122,485,809 0.1%
By summarizing the total audit fees from our sample, as done in the table 5 shown above, the Big 4 accounting firms are also PwC, EY, Deloitte and KPMG. The medium sized firms are Grant Thornton, BDO and McGladrey. McGladrey is classified as a medium sized accounting firm because literature suggests it is one of the biggest accounting firms after the Big 4 firms. All other audit firms will be classified as small audit firms.
Auditor industry expertise can be measured by the market share of each audit firm in every industry. NAICS codes will be used to identify different industries (NAICS, 2014). Theory suggests that nation-wide level industry expertise is not the best way to identify industry specialism. City-wide level is the best way to identify industry specialism (Ferguson et al., 2003). In this study industry specialism will be identified by the market share in each region each year. Region-wide level is chosen over city-wide level because not enough information was available for city-wide level.
To be industry specialist the audit firm must be one of the top two audit firm nation-wide and have at least 30% market share in the specific industry. Industry specialization
21 nation-wide is displayed in table 6.Table 6 shows for each of the 20 industries the nation-wide industry specialists for each year. This is the first step in identifying auditor industry specialization. In the appendix the region-wide industry specialist are disclosed for each year. For example table 22 in appendix A contains for each industry and each region the industry specialist in 2005.
Table 6: Industry specialist nation-wide each year
2005 2006 2007 2008 2009 2010 2011 2012 2013
Nation-wide Nation-wide Nation-wide Nation-wide Nation-wide Nation-wide Nation-wide Nation-wide Nation-wide
1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
Agriculuture, Forestry,
Fishing and Hunting EY KPMG EY KPMG EY EY PwC EY PwC EY PwC EY PwC EY PwC EY McGladrey
Mining KPMG PwC KPMG PwC EY PwC PwC KPMG PwC EY EY PwC EY PwC PwC EY PwC EY
Utilities Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Construction EY PwC EY Deloitte Ey Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte
Manufacturing PwC EY Pwc EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY
Wholesale Trade EY KPMG EY PwC EY KPMG EY KPMG EY KPMG EY KPMG EY KPMG EY KPMG EY KPMG
Retail Trade Deloitte EY Deloitte EY Deloitte EY EY Deloitte Deloitte EY EY Deloitte EY Deloitte EY Deloitte EY Deloitte Transportation and
Warehousing EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY PwC PwC EY
Information EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC
Finance and Insurance PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte PwC Deloitte Deloitte PwC Real Estate Rental and
Leasing Pwc Deloitte PwC Deloitte PwC Deloitte PwC EY PwC EY PwC EY PwC Deloitte PwC EY PwC Deloitte
Professional, Scientific,
and Technical Services PwC KPMG PwC KPMG KPMG EY KPMG Deloitte PwC Deloitte Deloitte PwC PwC Deloitte PwC Deloitte Deloitte EY Management of
Companies and Enterprises
EY PwC PwC EY PwC EY EY Deloitte EY Deloitte EY Grant
Thornton EY KPMG EY Deloitte Grant
Thornton KPMG Administrative and
Support and Waste Management and Remediation Services
EY PwC EY Deloitte Deloitte EY EY Deloitte Deloitte EY Deloitte EY EY Deloitte EY Deloitte Deloitte KPMG
Educational Services EY PwC EY KPMG EY KPMG EY PwC EY PwC EY PwC EY PwC EY PwC EY Deloitte
Health Care and Social
Assistance PwC EY EY PwC EY PwC EY PwC EY Deloitte EY PwC EY Deloitte EY Deloitte EY PwC
Arts, Entertainment, and
Recreation Deloitte KPMG Deloitte EY Deloitte EY EY Deloitte EY Deloitte EY Deloitte EY Deloitte EY Deloitte Deloitte PwC Accommodation and
Food Services EY Deloitte EY Deloitte EY Deloitte Deloitte EY EY Deloitte EY Deloitte EY Deloitte EY Deloitte Deloitte KPMG Other services (except
Public Administration) PwC KPMG PwC KPMG PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY PwC EY
Public Administration EY Deloitte Grant
Thornton Deloitte Grant
Thornton Deloitte Deloitte Grant Thornton Grant Thornton PwC Grant Thornton Deloitte Grant Thornton PwC Grant Thornton PwC -- --
3.3. Event study
Event studies are used to see what the financial impact is on certain events. Different researchers and theories suggest that financial events are immediately reflected in the share price of firms. With this in mind the time period used in this paper to determine the impact of an auditor change on the share price is short. The impact of an auditor change is measured by computing the CAR (= Cumulative Abnormal Return). To compute this CAR, there is not one good structure for event studies but there is a general flow of analysis (MacKinlay, 1997). The following section explains how the CAR is computed during this research. The steps that need to be taken are:
1. Choose event window 2. Calculating actual return 3. Calculating normal return 4. Calculating abnormal return
5. Computing the Cumulative Abnormal Return (CAR)
The first step is to decide what the event window will be. In this study multiple event windows will be used to see (1) if there is any differences between time windows (because of premature information leakage) and (2) the day of filing the 8-K Form is not always the day of the announcement due to rules that suggest that you can file the 8-K Form four days after the announcement day.
The above figure shows that the time between T0 and T1 is the estimation window.
This is the period to calculate what the normal return should be. The time between T1 and T2
is the event time window. This is the period in which the economic event occurred (t). The last window is between T2 and T3 and this is the post-event window.
To calculate the abnormal return (AR) the following model is used: ( | )
24 Where i is the firm and t is the event date. is the abnormal returns for firm i on time t,
is the actual return and ( | ) is the normal return. is the normal market return. This can be measured with the constant mean return model or the market model. To compute it is common to take a time period prior to the event period. In this case a period of 120 days before the event window has been used (estimation window).
The second step in computing the CAR is finding the right actual return. This has been done with the following model:
In this formula is the actual return of firm i on time t. is the closing price of firm i on
time t. is the closing price of the previous trading day of firm t or it is the opening price of firm i on time t.
The third step is computing the normal return. The market model is used to predict the normal return related to the market portfolio. This model is a statistical model and is
perceived to be better than economical models like CAPM (Capital Asset Pricing Model) or APT (Arbitrage Pricing Theory). The market model is an improvement of the linear model because the market model reduces the variance of the abnormal returns compared with the linear model (MacKinley, 1997).
Market model:
In this model is the return of firm i in period t, is the market return in time period t
and is an error variable. is the intercept and is the slope of the market model. The first and last formula can be combined into one model:
̂ ̂ ̂
25 The fifth step is calculating the Cumulative Abnormal Return. To do this the observed abnormal returns of firm i on time t must be aggregated to conclude if the event has an effect on the stock price of firm i. The sum of the abnormal return is called the Cumulative
Abnormal Return and can be computed with the following model:
̂ ( ) ∑ ̂
With this formula the abnormal returns of firm i on time t is combined to get the Cumulative Abnormal Return of firm i when they decided to change from auditor.
The switch of an Auditee from audit firm A to audit firm B leads to a reaction in the market (measured with CAR). The market reaction is based on the characteristics of audit firm A and audit firm B. Multiple equal changes based on the characteristics of audit firm A and audit firm B indicate how the market reaction depends on the different characteristics of the audit firms. Figure 1 visualizes the event study schematically.
3.4. Audit fees
As stated before, one of the most common reasons for an auditee to change their auditor is the audit fee. Literature suggests however that a change to a Big 4 auditor results in a higher audit fee due to a perceived higher audit quality.
Auditee
Switch from firm A to firm B
Audit firm A
The switch leads to a marketAudit firm B
reaction (CAR)
That depends on
Characteristics of
audit firm A and B
26 In the second part of this research the audit fees will be examined. The answers to questions like ‘do Big 4 auditors indeed charge higher audit fees’ and ‘are there any differences between audit fees of Big 4 audit firms’ will be researched.
There are multiple ways to examine the differences between audit fees of Big 4 audit firms and difference between Big 4 firms and non-Big 4 firms. The first way is to look at the ratio assets to audit fees. With this method it is possible for every industry to get a ratio for every audit company. In addition you can calculate how high the expected audit fee after the audit switch should be. The second way to examine the mentioned differences is to use an audit fee model with some dummy-variables for the different audit firms. A third possible way is to use the real audit fees after the audit switch. The literature suggests however that neither of these methods contains the correct procedure. Audit fees increase after an auditor change and the audit fees decrease the longer the tenure. DeAngelo (1981) found that the incumbent auditor can earn quasi-rents. Quasi-rents are excess revenues above avoidable costs including opportunity costs. After a switch the audit fees are higher because of contract and specific client start-up costs.
To calculate the expected audit fees the model of Carson et al. (2004) is used. The explanatory variables for audit fees are auditee size, auditee complexity, auditee risk, auditee profitability, timing, auditor size, ownership control and auditor location. The Carson et al. (2004) model uses most of these variables in their model and this model will be used in this research. Due to the type of research the model will be modified. For instance the mining variable will not be used and an extra dummy variable will be introduced to make a
distinction between Big 4, medium and small audit firms. The study of Carson et al. (2004) is performed in Australia and this study uses listed US firms. In Australia there are a lot of mining firms in the small client sample. In USA this is not the case. That is why the mining variable is removed from the model. The new dummy variable is placed in the model to see whether there is a difference between audit firms and to calculate the expected audit fee better.
The explanatory variables and the chosen proxies are displayed in table 7 on the next page.
27 Table 7: Audit fee determinant and proxies
Explanatory variables Proxy variable(s)
Auditee size Natural logarithm of the total assets
Auditee complexity Inventory to assets ratio and receivable to assets ratio
Auditee risk CATA, quick ratio and leverage
Auditee profitability ROI, audit opinion and a recent loss indicator.
Auditor size Big 4, medium or small audit firm and specialism.
Timing Potential higher cost in December fiscal year-end audits.
Audit fee model 1:
( ) ( )
( )
Where:
LN(FEE) Natural logarithm of total audit fees (in thousands of dollars). LN(ASSETS) Natural logarithm of total assets (in millions of dollars). INVTA Ratio, inventory to total assets.
RECTA Ratio, accounts receivable to total assets. CATA Ratio, current assets to total assets.
QUICK Ratio, current assets less inventories to equity. LEVERAGE Ratio of long-term debt to equity.
ROI Ratio, earnings before interest and tax to total assets. OPINION 1 for going-concern opinion otherwise 0.
YE Indicator for fiscal year end in October, otherwise 0. LOSS 1 if company reported a loss in any of the last three years. LN(OTHER
SERVICE)
Natural logarithm of other services fee (in thousands of dollars). BIG 4 1 if audit firm is PwC, KPMG, Deloitte or EY.
MEDIUM 1 if audit firm is BDO, Grant Thornton or McGladrey otherwise it is a small firm. SPECIALIST 1 if audit firm is an industry specialist otherwise 0.
To see if there is any difference in the audit fee between Big4 audit firms and medium sized audit firms a second model is used. In this model all small sized audit firms are removed and in the audit fee model the variable MEDIUM is removed. In a third regression model the variable BIG4 is replaced with four dummy variables for each member of the Big 4. To test the third audit fee model all non-Big 4 items are removed from the sample. The next models
28 contain different variables to test for more differences in audit fees. For example the last audit fee model tests the influence of industry specialization on the audit fee.
The data that is necessary within this model is retrieved from two sources. Audit, non-audit fees and non-audit opinion is collected from Audit Analytics. Other data like assets, sales and returns are gathered from Compustat.
3.5.
Cross-sectional analysis
A cross-sectional analysis is used to explore whether the abnormal returns are a function of different control variables. The control variables are other variables than the audit fee that could influence the market reaction.
Where:
CAR Is the Cumulative Abnormal Returns of firm i. The different CARs are derived from the event study results.
UPAUDITFEE Is a dummy variable that is one when the expected audit fee is higher and zero if it is lower.
DOWNAUDIT FEE
Is a dummy variable that has the value one when the expected audit fee is lower and zero when the audit fee is higher. The expected audit fee is based on the audit fee model.
RESIGN Prior literature suggests that reasons for audit resignation are negative and can affect the market value (Dunn et al. 1999). The dummy variable RESIGN is one when the predecessor auditor resigns and otherwise zero. TIMING Is one when the switch is in the fourth quarter of the fiscal year. Efficient
planning and improve reporting timeliness are advantages of switches earlier in the fiscal year compared to switches in the last quarter of the fiscal year (Schwartz & Soo, 1996). A switch in the fourth quarter of the fiscal year negatively influences the market reaction.
SIZE Is the natural logarithm of the total assets in the year prior to the switch. Firm size is a proxy for the information availability (Knechel et al., 2007) DISAGREE Is a dummy variable with the value of one when the predecessor auditor
29 has an accounting disagreement with the company. Smith (1988) finds that there is a negative market return after the switch when there is a disagreement with the auditor.
DISTRESS Is measured with the Altmen z-score and is a proxy for firm’s financial distress. The Altmen z-score can be calculated with the following formula:
In general firms with a score higher than 3.0 are safe, between 2.7 and 2.99 firms are on alert, between 1.8 to 2.7 good chances of going bankrupt within 2 years and lower than 1.8 means a very high probability of going bankrupt (Altmen, 1968). DISTRESS is a control variable because
auditees with financial distress are more likely to switch from auditor. The auditee wants to pressurize the incumbent auditor for providing favourable treatment (Kluger & Shield, 1989).
TENURE Is a dummy variable that has the value of one when the tenure of the predecessor auditor is longer than five years. The longer the tenure of the audit period, the higher the audit quality (Ghosh & Moon, 2005).
AFILER Has the value of one when the firm is an accelerated filer. The market reacts more negative to a switch from an accelerated filer than to a switch from a non-accelerated filer firms (Cullinan et al., 2012).
The variables of interest are UPAUDITFEE and DOWNAUDITFEE. For these variables and the control variables three different CARs will be tested. A three-day, five-day and ten-day CAR is being tested.
30
4.
Data Descriptive
This chapter contains the descriptive statistics of the audit fee model and auditor changes. The audit fee model is based on data derived from both Compustat and Audit Analytics. The expected signs are based on different economic theories and previous studies. Descriptive statistics on auditor changes contain statistics about the different event-windows that are used and the expected signs. The expected signs are based on the audit fee model.
4.1. Descriptive of the audit fee model
The descriptive statistics of the audit fee model are not explained individually but as a group. The individual variables are linked to the determinants of the audit fee model. The
determinants are auditee size, auditee complexity, auditee risk, auditee profitability, auditor size and timing. The variables are displayed in table 8. To increase the power of the audit fee model some outliers are removed. The outliers are approximately the top and bottom 1% of the data sample. They are removed because outliers influence the power of the model (Zimmerman, 1994).
Three models are tested. The first model is to check the difference in audit fee when there is a switch from small to medium or to a Big 4 audit firm. The second model checks the difference in audit fees when a medium size audit firm is replaced for a Big 4 audit firm. In the third model the dummy variable BIG4 and MEDIUM are replaced for EY, DELOITTE and KPMG. This model shows the difference in audit fee between the Big 4 audit firms. The results of the models are shown in table 8.
Size: the proxy variable, LN(ASSETS), is as expected significant for all models. They
have all a positive relation at a 1% significance level. This is consistent with prior literature that suggests that the size of the auditee is one of the main components of the audit fee. The 0.523, 0.535 and 0.534 for the respective Model 1, Model 2 and Model 3 means that 1 percent increase in assets results in 0.523%, 0.535% and 0.534% increase in audit fee.
Complexity: the complexity is measured with the ratios of inventories to total assets
(INVTA) and receivable to total assets (RECTA). When these ratios are higher the
complexity of the balance sheet is also higher. This is why the expected signs are positive. For RECTA in all models the expected sign is also the realized sign. In model 2 the INVTA is negative but not significant. The INVTA is positive in the first and third model but not
31 significant. The relationship between the audit fee and the RECTA is significant. For all models RECTA is positive with a p-value of 0.000.
Table 8: Descriptive statistics audit fee model
Full Sample: N=25,523 N= 21,874 N=18.947
Model 1 Model 2 Model 3
Exp. Sign Mean
Std.
Dev. B t Sign. B T Sign. B T Sign.
(Constant) 9.436 429.786 0.000 * 9.736 396.334 0.000 * 9.989 350.732 0.000 * LN(FEE) 13.701 1.283 LN(ASSETS) + 6.139 2.126 0.523 199.307 0.000 * 0.535 193.897 0.000 * 0.534 183.525 0.000 * INVTA + 0.100 0.126 0.111 0.606 0.544 -0.152 0.775 0.438 0.024 0.118 0.906 RECTA + 0.137 0.114 0.953 26.932 0.000 * 1.095 27.491 0.000 * 1.229 28.351 0.000 * CATA + 0.500 0.255 0.198 1.085 0.278 0.137 0.707 0.480 0.302 1.499 0.134 QUICK - 0.398 0.244 0.401 2.201 0.028 ** 0.530 2.731 0.006 * 0.367 1.821 0.069 *** LEVERAGE + 0.426 4.694 -0.001 -0.688 0.492 0.000 -0.593 0.553 0.000 0.000 1.000 ROI - -0.003 3.096 -0.001 -1.107 0.268 0.005 3.625 0.000 * 0.006 4.066 0.000 * OPINION + 3.2% 0.407 18.156 0.000 * 0.313 10.44 0.000 * 0.310 9.400 0.000 * YE + 71.6% 0.029 3.377 0.001 * 0.133 5.791 0.000 * 0.147 6.102 0.000 * LOSS + 48% 0.204 24.854 0.000 * 0.205 23.628 0.000 * 0.203 21.719 0.000 * LN(OTHER SERVICE) +/- 2.887 4.451 0.009 10.749 0.000 * 0.11 12.076 0.000 * 0.010 10.295 0.000 * BIG4 + 74.3% 0.601 45.387 0.000 * 0.188 14.724 0.000 * PWC ? 18.6% EY ? 24.7% -0.094 -8.093 0.000 * DELOITTE ? 16.1% -0.061 -4.768 0.000 * KPMG ? 14.9% -0.117 -8.534 0.000 * MEDIUM + 11.5% 0.395 25.909 0.000 * SPECIALIST + 18% 0.026 2.537 0.011 ** 0.022 2.242 0.025 * * 0.007 0.306 0.760 R2 0.787 0.725 0.700 * 1% significance ** 5% significance *** 10% significance
Expected sign = is the expected relationship between the audit fee and the determinants. The expected signs are based on the literature.
Risk: auditors risk is captured with the variables CATA, QUICK and LEVERAGE.
The QUICK ratio is expected to be negative because this ratio is an indication in the capability of the firm to pay their short term debt. If this ratio is greater than one, the
organization is able to pay all their short term debt. If an organization is capable in paying all their short term debt the risk is lower and a lower risk results in a lower audit fee. The results of the three models shows however that QUICK has a positive relation with the audit fee and it is significant at the levels 5%, 1% and 10% respectively. CATA and LEVERAGE is
32 expected to positively influence the audit fee because higher debt to equity implies a higher risk. The CATA is positive but not significant in the models. LEVERAGE is in the first model negative but not significant. The beta of LEVERAGE in model 2 and 3 is 0.000 and not significant.
Profitability: the profitability of a firm is measured by the LOSS and ROI variables.
The ROI is slightly negative but not significant in the first model. Model 2 and 3 show that the ROI is positive with a p-value of 0.000. This means that when the ROI increases with 1% the audit fee also increases with 0.5% and 0.6%. When a company is making a loss than the company will try to lower their costs. As a result the auditor needs more time for the control what increases the audit fee. The OPINION variable is significant positive at a 1% level for all models. This is consistent with what the literature suggests.
Auditor size: the expected signs for BIG4 and MEDIUM are positive. This is because
a switch to a Big 4 audit firm leads to a higher audit fee because of the fee premium. Both the variable BIG4 and MEDIUM are positive and significant in the first model. The coefficient of BIG4 is higher than MEDIUM. This indicates that a switch from a small audit firm to a Big 4 or medium sized audit firm results in a higher audit fee. In the second model the small audit firms are removed. This results in a higher audit fee when there is a switch from medium to Big 4. The results are summarized in table 9.
Table 9: Expected audit fee after an auditor switch To
BIG4 MEDIUM SMALL PWC EY DELOITTE KPMG
From BIG4 ? - - MEDIUM + ? - SMALL + + ? PWC - - - EY + + - DELOITTE + - - KPMG + + +
+ Higher expected audit fee - Lower expected audit fee
? Unknown effect on expected audit
The third model shows that there is a difference in audit fee between the Big 4. PwC has the highest audit fee and KPMG the lowest. A switch from PwC to KPMG results in a lower audit fee of 11.7%. All the variables are significant at a 1% level.
33 In the first two models SPECIALIST is as expected positive at a significance level of 5%. In the third model where the BIG4 variable is replaced for the four members of the Big 4 the variable SPECIALIST is not significant. SPECIALIST still has a positive relation but is not significant anymore. Table 31 (model 6) in appendix B shows that a switch to a
SPECIALIST results in a significant higher audit fee. The p-value is 0.000 this means it is significant at a 1% level. The interpretation of the coefficient of 0.093 is that a switch to an industry specialist results in a 9.3% higher audit fee.
Timing: previous research indicates December as busy period. This means that YE
should be positive because a busy period could mean a higher audit fee. In the models this is the case. At a 1% significance level YE variable is positive significant in the models.
Other: LN(OTHER SERVICE) is the last variable that is not yet explained. The
expected sign is difficult to determine. When a company has high non-audit fee costs you could expect that the audit firm knows more about the auditee. If this knowledge is used within the audit process the audit fee could decrease. This is however not the case. When a company has higher non-audit fee costs, they probably need more advice on difficult topics. This could increase the complexity of the audit and the audit fee. The effect of the increased complexity is higher than the effect of knowledge about the client. The LN(OTHER
SERVICE) is significant positive for all the models. The coefficient is respectively 0.026, 0.022 and 0.007 for Model 1, Model 2 and Model 3. This indicates that an increase in the LN(OTHER SERVICE) of 1% leads to an increase in the audit fee of 2.6%, 2.2% and 0.7% respectively.
The R2 for the three models are 78.7%, 72.5% and 70%. This means that at least 70% of all the audit fees are explained with the models.
4.2. Descriptives event study
The descriptive statistics of the event study show a mixed result. The next table displays the results of using a 120-day estimation window and three different event-windows. CAR 3 means one day before the change, the day of the change and one day after the change. CAR 5 means three days before the auditor change, the day of auditor change and one day after the auditor change. CAR 10 is instead of three days before the event eight days before the event. The change date is not the actual day of the auditor change announcement or replacement. The change date is the date of filing the 8-K Form. The sample size is 1,855
34 instead of the previous 1,935 because some outliers and missing data are deleted. Outliers are the top and bottom 1% of the Cumulative Abnormal Returns.
After the t-test only the Cumulative Abnormal Return of a ten day event window is significant. Table 10 shows us that there is not one expected market return after an auditor switch. This is because the means for the different CARs are negative and positive but not significant. More tests are done to see which characteristics of the new auditor are significant. Table 10: CAR descriptives
CAR3 CAR5 CAR10
N 1,855 1,855 1,855 Maximum 1.195 2.800 3.857 Minimum -0.594 -0.975 -1.913 Mean 0.001 0.002 -0.002 Std. Deviation 0.079 0.121 0.196 Sign. 0.589 0.467 0.641
Based on the results of the audit fee model there are some expectations about the market reaction. Table 11 summarizes the expected market reactions which are used to examine if the first hypothesis is significant. The audit fee model shows a higher audit when there are switches from SMALL to BIG4 or MEDIUM. That is why the expected market reaction should be negative. The positive sign of SPECIALIST indicates that a switch from an industry specialist results in a positive market reaction because of the lower audit fee. A switch to SPECIALIST is the other way around. The higher audit fee results in a negative market reaction.
Table 11: Expected market reaction based on expected audit fee To
From BIG4 MEDIUM SMALL SPECIALIST
BIG4 ? + + -
MEDIUM - ? + -
SMALL - - ? -
SPECIALIST + + + x
+ Positive market reaction - Negative market reaction ? Unknown market reaction x Change is not possible
35
4.3. Cross-sectional analysis
The descriptive statistics of the cross-sectional analysis is shown in table 12. The full sample exists of 1,855 auditor switches. 25.4% (472) of all the switches are switches to auditors with an expected higher audit fee, 43.5% (806) are switches to an auditor with an expected lower audit fee. The other 31.1% (difference between FULL SAMPLE and combined
UPAUDITFEE and DOWNAUDITFEE), these are 577 switches which do not result in an expected higher or lower audit fee.
Table 12: Independent variable descriptive statistics
Type of Switch
FULL SAMPLE UPAUDITFEE DOWNAUDITFEE
N 1,855 472 806 % 100% 25.4% 43.5% RESIGN N 422 97 125 % 22.7% 20.6% 15.5% TIMING N 263 68 84 % 14.2% 14.4% 10.4% DISAGREE N 52 11 26 % 2.8% 2.3% 3.2% TENURE N 496 119 196 % 26.7% 25.2% 24.3% AFILER N 1,172 307 523 % 63.2% 65% 64.9% SIZE N 1,631 399 718 Mean 5.678 5.935 5.736 Median 5.695 5.879 5.692 Std. Dev. 2.149 1.989 2.112 DISTRESS N 1,645 407 717 Mean 4.016 5.940 1.175 Median 2.156 2.151 2.186 Std. Dev. 40.752 26.576 24.234 Variable Definitions:
UPAUDITFEE = Auditee that switch to an audit firm with an expected higher audit fee DOWNAUDITFEE = Auditee that switch to an audit firm with an expected lower audit fee
When auditors resign more firms change to an auditor with an expected higher than an expected lower audit fee. 22.7% of the changes are due to auditor resignations, 20.6% of the UPAUDITFEE firms report an auditor resignation as an auditor change and 15.5% of the DOWNAUDITFEE firms report this as well. Only 14.2% of the changes occur in the fourth quarter of the fiscal year. This is 14.4% for the UPAUDITFEE firm and 10.4% for the
36 in the full sample 2.8% and 2.3% and 3.2% respectively for UPAUDITFEE and
DOWNAUDITFEE firms. Approximately one fourth of the changes are changes where the previous audit tenure was longer than five years. For the full sample, UPAUDITFEE and DOWNAUDITFEE this is respectively 26.7%, 25.2% and 24.3%. 63.2% of the firms are accelerated filers. In the UPAUDITFEE sample this is 65% and in the DOWNAUDITFEE sample this is 64.9%.
SIZE is the natural logarithm of the total assets. The mean of this variable is 5.678 for the full sample and 5.935 and 5.736 for UPAUDITFEE and DOWNAUDITFEE. The mean for the bankruptcy z-score is 4.016 for the full sample. 5.940 for UPAUDITFEE and 1.175 for
