The effects of audit firm tenure, audit firm size, and auditor industry expertise on audit quality as perceived by investors

(1)

‘The effects of audit firm tenure, audit firm size, and auditor

industry expertise on audit quality as perceived by investors’

Name: Laura Knuppe Student number: 10596887

Date of submission final version Master’s thesis: 22 June 2015 Word count: 12,234

Master Accountancy & Control, Accountancy specialisation Amsterdam Business School

Faculty of Business and Economics, University of Amsterdam Supervisor: dhr. prof. dr. V.S. (Victor) Maas

(2)

2

Statement of originality

This document is written by student Laura Knuppe who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

This study examines whether audit firm tenure, audit firm size, and audit firm industry specialisation affects audit quality as perceived by investors. Audit quality is proxied by the Earnings Response Coefficient (ERC). I use data for U.S. firms from Compustat, Audit Analytics, and CRSP for firm years 2002 – 2012. For the full sample period I find that investors perceive audit quality to increase with audit firm tenure. Also, investors’ perceptions of audit quality increase when a firm is audited by a Big 4 audit firm, as compared to firms audited by a non-Big 4 audit firm. However, no significant results are found regarding the effects of audit firm industry expertise on audit quality.

The accounting scandals in 2001 and 2002, and the global financial crisis commencing in 2007 might have influenced the results. Therefore, as a sensitivity test, the sample was partitioned into three subsamples: 2001 – 2002, 2003 – 2006, and 2007 – 2012. The sensitivity analysis implies that after the events in 2001 and 2002, investors perceive audit firm tenure to negatively affect audit quality. In addition, they perceive Big 4 audit firms to decrease audit quality in the 2001 – 2002 subsample. In the 2003 – 2006 subsample, investors perceive Big 4 audit firms to increase audit quality again, but they remain to perceive longer auditor tenure to decrease audit quality. This could be an excess effect from the 2001 and 2002 events. In the 2007 – 2012 subsample, investors again perceive longer audit firm tenure to positively affect audit quality.

This research contributes to the accounting literature by being one of the first studies to jointly examine the effects of tenure, firm size, and industry expertise on perceived audit quality, as proxied by the ERC. To the best of my knowledge, there is no such research yet in the recent U.S. setting. The majority of the studies in this field either examine the effects of one of the current proxies on actual audit quality, or examine the effects of one of the proxies on perceived audit quality, but not as captured by the ERC.

(4)

4

1 Introduction

The case of restricting audit firm tenure, also known as ‘mandatory audit firm rotation’, and its effects on the quality of audits has been a widely discussed topic in the past years. It has been a focus of discussions for, amongst others, the U.S. Congress, the U.S. Securities and Exchange Commission (SEC), the International Accounting Standards Board (IASB), the Financial Accounting Standards Board (FASB), the Public Company Accounting Oversight Board (PCAOB) and the users of financial statements. In the light of events such as the 2007 - 2008 global financial crisis, and the numerous accounting scandals arising in the same period, audit quality has gained more attention of the general public, and is again researched extensively.

Following the major Enron and WorldCom corporate accounting scandals in the United States in the 2000s, the Sarbanes-Oxley Act of 2002 (further referred to as SOX) was enacted. Amongst others, the federal legislative bill made sure that the mandatory rotation period of audit partners was shortened from seven to five years. In addition, the cooling-off period between auditing the same firm for partners was lengthened from two to five years (Litt, Sharma, Simpson, & Tanyi, 2014). The House of Representatives (2002) reported that the goal of mandatory rotation was to increase the quality of financial statement information reported to investors. Another element that was raised in SOX was the mandated research by the U.S. Comptroller General into the possible effects of mandating audit firm rotation, instead of audit partner rotation. The research was done by the U.S. Government Accountability Office (GAO), which reported that mandatory audit firm rotation would not be the most efficient way to increase audit quality. However, it did leave open the option to introduce mandatory audit firm rotation if it was considered necessary (Ruiz-Barbadillo, Gómez-Aguilar, & Carrera, 2009). Nonetheless, on August 16, 2011, 10-year mandatory audit firm rotation was introduced in a concept by the PCAOB. However, it was subject to much criticism; the PCAOB received 628 comment letters, mostly opposing mandatory audit firm rotation. Since then, mandatory audit firm rotation has been a topic of much debate, which eventually led to the recent ruling by Congress not to support the PCAOB’s recommendation to rotate audit firms (Litt et al., 2014).

The fact that mandatory audit firm rotation is a global issue can be seen from the legislation across the world. For example, in April 2014, the European Parliament accepted

(6)

6

rules to mandate European listed firms to rotate their audit firms every 10 years. This means that all 28 members of the European Council need to comply with these rules. Michel Barnier, who is the European Internal Market and Services Commissioner, argued that “these new measures will reduce risks of excessive familiarity between statutory auditors and their clients, encourage fresh thinking, and limit conflicts of interest” (Chasan, 2014). In addition, mandatory audit firm rotation was adopted in Brazil in 1996 for banks, and also for other listed companies in 1999. Once again, the mandatory rotation was provoked by fraudulent activities and bankruptcy of major banks. In South Korea, companies listed on the Korean Stock Exchange (KSE), or companies registered with Korea Securities Dealers Automated Quotations (KSDAQ) were obliged to rotate their auditors since 2003. The rotation period in Korea is six years (Cameran, Di Vincenzo, & Merlotti, 2005).

Advocates of mandatory audit firm rotation claim that restricting the tenure of audit firms results in higher independence of auditors, thereby enhancing financial statement quality. For instance, Dopuch, King, & Schwartz (2001) document that auditors in a system with short audit firm tenure are less likely to issue a biased report, than auditors in a system without mandatory audit firm rotation, and thus longer audit firm tenure. This bias is one of the main arguments in favour of restricting audit firm tenure; mandated rotation restricts auditors from being too aligned with the audited company (Jackson, Moldrich, & Roebuck, 2008). In addition, Ruiz-Barbadillo et al. (2009) state that auditors are more likely to be influenced in their reporting behaviour if there is no audit firm rotation, since they are economically dependent on the client and possibly its managers. Compulsory audit firm rotation may also help prevent corporate collapses such as WorldCom, and Enron/Arthur Andersen. Consequently, it is argued that preventing the costs of these collapses outweighs the high initial costs of mandatory audit firm rotation (Jackson et al., 2008). The competition amongst audit firms when mandatory audit firm rotation is imposed is also likely to increase an audit firm’s level of service, and thus audit quality (Hoyle, 1978).

On the contrary, opponents of restricting audit firm tenure argue that mandated rotation is likely to increase audit failures (Myers, Myers, & Omer, 2003). This increased likelihood of audit failures is due to a lack of client-specific knowledge in the initial years of the audit period (Chi, Huang, Liao, & Xie, 2009). It is thus argued that auditors gain important client-specific knowledge as the length of the audit progresses. Azizkhani, Monroe, & Shailer (2004) argue that audit quality is thus lower in the first years of the audit, and increases with

(7)

7

longer tenure. This is due to the reduction of information asymmetry between the client firm and the auditor that happens over the years. It is also argued that mandatory audit firm rotation is deemed unnecessary, since auditors are provided with enough market-based incentives to be independent. For example,DeAngelo (1981) argues that firm reputation is a major incentive for an auditor to be independent, and thus increases audit quality. Also, Reynolds and Francis (2001) agree that reputation is a major determinant of auditor decision-making.

Prior auditing literature has indicated that audit quality can be divided into two components; actual audit quality and perceived audit quality. Actual quality is described as “the degree to which the risk of reporting a material error in the financial accounts is reduced” (Jackson et al., 2008, p. 422). On the other hand, perceived quality is outlined as “how effective users of financial statements believe the auditor is at reducing material misstatements” (p. 422). Since much of the prior research has focused on actual audit quality, this research will focus on the effects of audit firm tenure, audit firm size and auditor industry specialisation on perceived audit quality. The users addressed in this study are investors, since they are one of the principal users of financial statements (Ghosh & Moon, 2005). Thus, in this study, the following research question stands central:

“Do audit firm tenure, audit firm size and auditor industry specialisation affect audit quality as perceived by investors?”

The primary motivation for this study stems from calls for incorporating both audit firm size and auditor specialisation in determining the effects of audit firm tenure on perceived audit quality. In the current situation, the majority of the studies either examine the effects of one of these proxies on actual audit quality, or examine the effects of one of the proxies on perceived audit quality, but not as captured by the ERC. As such, Jackson et al. (2008)stress that their research only focuses on actual audit quality, but leaves out the equally important perceived audit quality. As indicated by Fernando, Abdel-Meguid, and Elder (2010), audit firm size, auditor industry specialisation and audit firm tenure are important attributes of audit quality as perceived by capital market participants. They proxy perceived audit quality as the ex-ante cost of equity capital and show that all these three characteristics are negatively associated with perceived audit quality. The question remains whether these

(8)

8

results will persist when the Earnings Response Coefficient (ERC) is used as a proxy for investors’ perceptions of audit quality.

Teoh and Wong (1993) were one of the first researchers to document the relation between audit firm size and market perceptions of the credibility of financial statements. The credibility of earnings reports is measured by the Earnings Response Coefficient. As a key result, Teoh and Wong (1993) indicate that clients audited by Big 8 firms have significantly higher ERCs, indicating that investors perceive the presence of a Big 8 auditor as indicating a higher credibility of financial statements, and thus a higher quality of the audit.

Chi et al. (2009), are one of the few known studies to incorporate both actual and perceived audit quality in their study of audit partner rotation, in which perceived audit quality is measured by the Earnings Response Coefficient. In addition, their research takes place in a Taiwan setting, where 5-year audit firm rotation is mandatory since 2004. Chi et al. (2009) find no support that restricting audit partner tenure enhances investors’ perceptions of audit quality. Also, this study does take auditor size (Big 4 or Big 5 audit firms) into consideration, but equally fails to take auditor specialisation into account.

In the U.S. setting, Ghosh and Moon (2005) study the effects of auditor tenure on investors’ and information intermediaries perceptions of audit quality, proxied by the Earnings Response Coefficient from returns-earnings regressions. They find that longer audit firm tenure is perceived by capital market participants as positively influencing audit quality. However, like Chi et al. (2009), this study fails to look at the effects of auditor specialisation on the perception of investors on audit quality.

Another U.S. based research is that of Mansi, Maxwell, and Miller (2004), which focuses on the relation between audit firm tenure and audit firm size on the perception of capital market participants (in this case, bondholders). The perception of bond holders is proxied by the cost of debt financing, an alternative measure of audit quality. Mansi et al. (2004) find that auditor quality, proxied by Big 4 audit firms, and audit firm tenure, are negatively associated with the cost of debt financing. This implies that both audit firm tenure and audit firm size matter to capital market participants. Once again, this research does not incorporate an important attribute of perceived audit quality; auditor industry specialisation. Taking the above facts into consideration, the aim of this study is to determine whether there is a relation between audit firm tenure, audit firm size and audit firm specialisation and audit quality as perceived by investors. To the best of my knowledge,

(9)

9

there is no research in the U.S. setting examining the effect of these attributes of perceived audit quality as proxied by the Earnings Response Coefficient. In order to reach this aim, it is hypothesised that audit quality increases with increased audit firm tenure, having a Big 4 firm as auditor as compared to a non-Big 4 audit firm, and being audited by an industry specialist.

This research contributes to the current stream of research regarding attributes of audit quality and their interactions in the following way. Primarily, the academic contribution of this study is to jointly examine the effects of audit firm tenure, audit firm size (Big4/Non-Big4), and auditor industry specialisation on audit quality as perceived by investors in a recent U.S. setting. As mentioned before, the majority of the research in this field has rather focused on the effects of tenure, size and industry expertise on actual audit quality. Perceived audit quality is studied to a much lesser extent. Moreover, while other studies do use the Earnings Response Coefficient as a measure of audit quality as perceived by capital market participants, they do not take into consideration the relative importance of audit firm tenure, audit firm size, and auditor industry specialisation.

The key results of this study indicate that investors perceptions’ of audit quality as proxied by audit firm tenure increase as the tenure increases. This could be due to the preservation and achievement of client-specific knowledge as the audit progresses. It is also found that investors perceive audit quality to increase if the firm is audited by a Big 4 audit firm. Nevertheless, this study finds no significant relation between audit firm industry expertise and perceived audit quality. This could be a result of investors being unable to adequately register whether an audit firm is an industry specialist or not.

The remainder of this paper is organised as follows. The next chapter provides a literature review of the concepts of audit quality, audit firm tenure, audit firm size and auditor specialisation. In addition, this chapter states the hypotheses. Chapter three describes the sample, research methodology employed, and descriptive statistics of the sample. The fourth chapter contains the results. The last section comprises of a discussion and conclusion of the results, and poses limitations to the research and suggestions for further research.

(10)

10

2 Literature review and hypotheses development

This study’s objective is to examine whether audit firm tenure, audit firm size and audit firm industry specialism change capital market participants’ perceptions of audit quality. The capital market participants addressed in this study are investors. This chapter provides further background on the key concepts of this study. Firstly, the concept of audit quality is explained. Furthermore, the relationships and expected associations between audit firm tenure, audit firm size and audit firm specialisation and perceived audit quality will be addressed. This section will conclude with developing the hypotheses that will be used to test the main research question.

2.1 Audit quality

Firms prepare financial statements for their financial statement users, such as shareholders, managers, suppliers, investors, and employees. However, the difference between internal and external users of financial statements creates information asymmetry (Chi et al., 2009). In order to reduce this information asymmetry, the financial statements of companies are audited by audit firms. The worth of this audit depends on the quality that the audit firm delivers. Audit quality encompasses a continuum ranging from very low to very high audit quality, whereby audit quality is inherently connected to the amount of audit failures (Francis, 2004). As previously stated, it is shown in academic research that audit quality comprises both actual audit quality and perceived audit quality. Actual audit quality has been the focus of many academic researchers and legislators, indicating that actual quality is “the degree to which the risk of reporting a material error in the financial accounts is reduced” (Jackson et al., 2008, p. 422). This relates to the definition of Francis (2004), focusing on the amount of audit failures or material errors in financial statements as an indicator of audit quality.

However, audit quality also consists of the equally important perceived audit quality. Jackson et al. (2008) define audit quality as perceived as “how effective users of financial statements believe the auditor is at reducing material misstatements” (p. 422). This dimension of audit quality will be the focus of this research. As can be derived from the above definition, perceived audit quality encompasses financial statement users beliefs of audit quality. The financial statement users addressed in this research are investors, since they are one of the most important users of financial statement.

(11)

11

In the overall auditing literature, an important development is the assumption that differences in audit quality exist, and that these can be discovered by comparing different groups of auditors (Francis, 2004). In his study, Francis (2004) identifies that an underlying assumption here is that all auditors meet the minimum legal requirements of an audit, and differences are therefore distinguished beyond the legal framework. Specifically, differential attributes to audit quality are the ‘big firm - small firm dichotomy” (p. 352) and industry expertise. In addition, Fernando et al. (2010) identify that an important attribute of audit quality as perceived by investors is audit firm tenure. These three differential attributes to perceived audit quality will be the focus of this resarch.

2.2 Audit firm tenure and audit quality

The association between audit firm tenure and audit quality has been a subject of discussion for many years. According to Francis (2004), research focusing on audit firm tenure examines whether the length of the client-auditor relationship influences audit quality. The main motivation for this stream of research stems from calls for mandatory audit firm rotation. As described in the first section of this study, there are two main streams of opinion regarding mandatory audit firm rotation.

On the one hand, it is argued by various researchers (e.g. Dopuch et al., 2001; Jackson et al., 2008; Hoyle, 1978) that a long audit firm tenure has a negative influence on audit quality. For instance, it is said that restricting audit firm tenure results in a higher independence of auditors, thus, in a higher quality of the audits performed. This expected higher independence of companies with a short client-auditor relationship is documented by Dopuch et al. (2001), who state that an audit firm with a long tenure is more likely to issue a biased audit report. Another main argument in favour of mandatory audit firm rotation is that restricting audit firm tenure keeps auditors from being too aligned with management of the auditee, which again, results in a higher independence. Ghosh and Moon (2005) state that as the client-auditor relationship lengthens, auditors are more likely to agree with the auditee’s management. This might be a dangerous development if clients are e.g. pursuing nonconservative accounting. If the auditor is more likely to agree with these practices, independence is impaired, and consequently, financial statement quality is impaired as well.

(12)

12

On the other hand, researchers argue that restricting audit firm tenure comes with high initial costs for the auditor. A further important argument against mandatory audit firm rotation is that the mandated rotation off firms results in a loss of client-specific knowledge, which is needed to perform high quality audits (Chi et al., 2009). Ghosh and Moon (2005) also state that problem audits happen more often for new audit clients. If the clients are rotating their audit firms more often, these problems in the audit will happen more often. In addition, in the begin years of the audit, the auditor has to rely on managerial estimates and management information more often. Also, it is argued that auditors are incentivised by the market to enhance their independence, and that mandatory audit firm rotation will not contribute to this. Ghosh and Moon (2005) indeed find that there is a positive association between audit firm tenure and investor perceptions of audit quality. In the field of actual audit quality, Jackson et al. (2008) also find that audit quality increases when there is long audit firm tenure.

To conclude, the opinions and results of prior research regarding the relationship between audit firm tenure and audit quality are mixed. However, most outcomes of conducted research regarding perceived audit quality indicate a positive relationship between audit firm tenure and audit quality. The underlying assumption regarding audit firm tenure and its effects on audit quality thus is that audit quality as perceived by investors increases with an increase in audit firm tenure. Nonetheless, there has not been a distinct answer to the association, and overall results indicate both positive and negative associations.

2.3 Audit firm size and audit quality

Audit firm size, or as Francis (2004) states the “big firm - small firm dichotomy”, was one of the first differential factors to audit quality to be researched. For example, DeAngelo (1981) argues that audit firm size is a proxy for auditor independence, and thereby audit quality, since large audit firms are not dependent on one client. In addition, large audit firms have a reputation to protect if they misreport. It is therefore argued that Big 4 international audit firms have built a large clientele and brand name, and subsequently are determined to deliver high quality audits to protect this brand name and international reputation (Francis, 2004). In contrast, small audit firms are more likely to be economically dependent on one client, and thus have more to lose if they are being tough on their client and risk being let go

(13)

13

as that firm’s auditor. However, it should be noted that audit failures can and do still happen within Big 4 audit firms (e.g. known auditing scandals such as Enron and WorldCom by Arthur Anderson, and Lehman Brothers under the supervision of Ernst & Young). Still, despite these high impact cases there is support for the statement that, generally seen, Big 4 audit firms will deliver audits of higher qualities than non-Big 4 audit firms. This is also evidenced in the study done by Francis (2004). He provides prove that Big 4 audit firms generally charge their clients with higher audit fees. Later on in his study, he shows that, even though some audit firms have pricing power over their clients, high audit fees indeed are a proxy for high audit quality. In addition, Francis, Maydew, and Sparks (1999) shows that financial statements are of higher quality when they are audited by a Big 4 company. He uses abnormal accruals as a proxy for earnings quality, and finds that firms audited by Big 4 audit firms have lower abnormal accruals. This is a sign of higher quality earnings, and thus higher quality audits. Teoh and Wong (1993) are one of the first to study perceived audit quality by means of the ERC. They examine whether audit quality differs between Big 8 audit firms and non-Big 8 audit firms. Using abnormal stock returns and earnings surprises, Teoh and Wong (1993) find that firms that are audited by a Big 8 audit firm have higher Earnings Response Coefficients. This indicates that larger audit firms are better able to precisely report earnings in comparison to non-Big 8 audit firms. Since most of the research clearly indicates a tendency towards concluding that Big 4 firms provide higher quality audits, the underlying assumption of this study will be that investors will perceive the audits performed by a Big 4 audit firm to be of higher quality than those performed by a non-Big 4 audit firm.

2.4 Auditor specialisation and audit quality

Francis (2004) indicates audit firm industry expertise as one of the contributes to audit quality that make a difference between audit firms. Since audit firm specialisation is hard to observe, prior research used several proxies for audit firm industry specialisation. For example, Balsam, Krishnan, and Yang (2003) use six proxies for industry expertise. Three of them are based on the number of clients. Furthermore, they also identify an industry specialist as the industry’s first, second, and third largest supplier, as long as there are significant differences in market share between them. In addition, they use a proxy of auditor dominance, and identify an industry specialist if the auditor has the largest market share in audit fees, and if that market share is 10 % larger than the second largest auditor in

(14)

14

that industry. This is the definition of audit firm industry specialisation that is also used in this study. Balsam et al. (2003) find that for all six proxies, firms that are audited by an industry specialist, have higher Earnings Response Coefficients, than firms that are audited by a non-industry specialist. In addition, they find that for five out of six proxies, firms audited by an industry specialist have lower discretionary accruals. This implies that audit firm industry specialisation is positively associated with actual audit quality, and investors’ perceptions of audit quality.

In addition, Francis (2004) states that, empirically seen, there is evidence that industry expertise is not evenly distributed among the Big 4 accounting firms. It is argued that, logically, industry experts have a deeper knowledge than non-experts as a result of more experience in the respective industries. If the Big 4 firm has more clients or audit fees in an industry, their employees have better chances to gain industry-specific knowledge. These opportunities lead to industry expertise, and thus higher audit quality. The evidence on the effects of auditor specialisation on audit quality equals the research on audit firm size and audit quality. Audit fees are higher for industry experts, implying higher audit quality. Consequently, there are observations that financial statements are of higher quality when the auditor is an industry expert (Francis, 2004).

In the field of perceived audit quality, next to Balsam et al. (2004), Krishnan (2003) finds that earnings surprises are priced better by investors, which also is in line with higher audit quality for industry specialists. Interestingly, Jenkins, Kane and Velury (2006) research whether, during a period of exceptional stock market activity, audit quality was impaired. They also look at whether the presence of an industry expert as the audit firm affected the ability of that firm to decrease the impairment of audit quality. Their first finding suggests that there was a significant drop in Earnings Response Coefficients during that period. This indicates a drop in audit quality. Furthermore, Krishnan (2003) finds that industry specialists were not able to stop or lessen this decrease in audit quality. It might thus be implied that, during a period of economic distress, industry specialists are not able to deliver higher quality audits than non-industry experts. A more recent study done by Gul, Fung, and Jaggi (2009) also provides evidence that there is an association between audit firm industry specialisation and audit quality. The study shows that audit firm industry specialisation has a moderating effect on the relation between audit firm tenure and audit quality. If the client-audit firm relationship has a short tenure, normally, this has a negative effect on client-audit

(15)

15

quality. However, this relationship is weaker for clients that are audited by an audit firm that is identified as an industry specialist. Taking all the above arguments in consideration, a clear tendency towards a positive effect of audit firm industry specialisation on audit quality is present. Therefore, it is expected in this study that investors will perceive that audits are of higher quality when they are performed by an industry specialist.

2.5 Hypotheses

Following prior research (e.g. Teoh & Wong, 1993; Ghosh & Moon, 2005; Chi et al., 2009), the Earnings Response Coefficient is used to measure perceived quality. To make conclusions about investors’ perceptions on earnings quality, researchers often use stock-market-based metrics such as the ERC from regressions of returns on earnings (Ghosh & Moon, 2005). Teoh and Wong (1993) document that investors pay a larger premium for high-quality earnings, since they view high-quality earnings as more sustainable. This implies that examining the effect of audit firm tenure on the pricing of earnings via the ERC is prone to provide valuable insights into investors’ perceptions of this effect. In order to research perceived audit quality, the following hypotheses are tested:

Hypothesis 1: The perceived quality of the audit of firms audited by audit firms with a long tenure is higher than the audit quality of companies audited by audit firms with a short audit firm tenure

Hypothesis 2: The perceived audit quality of companies audited by Big 4 audit firms is higher than the audit quality of companies audited by non-Big 4 audit firms

Hypothesis 3: The perceived audit quality of companies audited by an industry specialist is higher than the audit quality of companies audited by a non-industry specialist

(16)

16

3 Methodology

3.1 Sample selection

The initial sample is collected from Compustat 2014 files, for firm years 2002 – 2012. The sample contains publicly traded firms from North America. The analysis begins with 2002 and ends with 2012 to be able to distinguish between short and long audit firm tenure and the effects over time on the dependent variable. Data on stock returns and firm age is then collected from CRSP 2014 files. In order to construct the industry specialist variable, data on audit fees is collected from Audit Analytics. The initial sample from Compustat contains 123,626 observations from 19,469 unique companies. The following restrictions are imposed upon the initial sample:

- Firms in highly regulated industries (SIC 40 – 49 and SIC 60 – 67) are removed; - Firms with missing financial data are removed;

- The top 1 and bottom 99 percent of observations for Cumulative Market-Adjusted Returns (CAR), Earnings (E), ΔE, GROWTH, BETA, SIZE and LEVERAGE are winsorized.

Firms in highly regulated industries are removed since managers in those sectors have less room for opportunistic behaviour. Also, the balance sheet structure of those firms are substantially different from other firms. This could significantly affect the results of this research. This approach is very common in other academic accounting literature. After merging the data from Compustat, CRSP and Audit Analytics and applying the restrictions described above, the final sample consists of 23,992 observations. Panel A and B of Table 1 provide an overview of the construction of the final sample and industry composition, respectively.

Although the final sample is constructed carefully, I follow Ghosh & Moon (2005) in constructing a second, restricted sample. It is indicated in previous accounting research that firms engage in opinion-shopping amongst different audit firms, in order to avoid negative audit opinions. Lennox (2000) finds evidence that firms in the UK indeed engage in opinion-shopping to avoid negative audit opinions. In addition, it is argued that firms might switch to an audit firm that supports accounting decisions that are profitable for management. DeFond and Subramanyam (1998) find evidence for this, in that discretionary accruals are

(17)

17

income decreasing in the year before an auditor switch. This implies that the previous auditor supported conservative accounting choices, and management diverged from the previous auditor in order to behave opportunistically with the new, less conservative, audit firm. Alternatively, audit firms that pursue a high quality audit might prematurely end the engagement with a firm that pursues a low quality of their financial reporting. In order to control for these biases, the restricted sample consists of firms with a client-auditor relationship of five years or longer. The restricted sample comprises 20,278 observations.

TABLE 1

Sample and industry composition Panel A: Sample derivation

Description N

Initial sample from Compustat 123,626

Removal of firms in highly regulated industries (40,960) Removal of firms with missing financial data (58,674)

Final sample 23,992

Panel B: Industry composition

Description N (%) 01 – 09 Agriculture 0 (0%) 10 – 14 Mining 211 (6,1%) 15 – 17 Construction 45 (1,3%) 20 – 39 Manufacturing 1951 (56%) 50 – 51 Wholesale Trade 130 (3,7%) 52 – 59 Retail Trade 286 (8,2%) 70 – 89 Services 839 (24,1%) 91 – 99 Public Administration 22 (0,6%)

(18)

18

3.2 Research model

Following prior research (Ghosh & Moon, 2005; Heydari, 2015; Jenkins, Kane, & Velury, 2006; Chi et al., 2009), I use the following regression model to analyse whether investors perceive earnings quality as being affected by auditor tenure, audit firm size and audit firm industry specialisation:

CAR = α + ß1E + ß2ΔE + ß3E*TENURE + ß4ΔE*TENURE + ß5TENURE+ ß6+2(j-1)E*Variablej

+ ß7+2(j-1)ΔE*Variablej + ß19+j Variablej + ε.

In this equation, E and ΔE represent reported earnings and changes in reported earnings on a year-to-year basis, respectively. TENURE is the length in years of the client-auditor relationship, which equals 1 in the first year of this research (2002). The subscript j describes the number of variables in the equation (1 to 9). As can be seen from the equation, every variable is included in the regression as a separate variable, and both interacted with E and ΔE. The sum of ß1 and ß2 is the ERC, and the proxy for perceived audit quality. Since I research the effects of auditor tenure, audit firm size and audit firm industry specialisation, my variables of interest are the sum of E*TENURE and ΔE*TENURE (ß3 + ß4), the sum of E*BIG4 and ΔE*BIG4 (ß5 + ß6), and the sum of E*INDSPEC and ΔE*INDSPEC (ß7 + ß8). The rest of the variables serve as control variables.

3.2.1 Variable specification

The dependent variable used in the regression model is cumulative market-adjusted returns (CAR). These returns are cumulated for 12 months. Market-adjusted returns are the difference between raw returns and CRSP value-weighted market returns.

E represents Income Before Extraordinary Items and, logically, ΔE represents the change in Income Before Extraordinary Items between the current year and the previous year. Both E and ΔE are deflated by market value of equity. TENURE is the length in years of the client-auditor relationship, which equals 1 in the first year of this research (2002). The largest tenure in this research is 12.

(19)

19

BIG4 is an indicator variable, which equals 1 if the firm is audited by a Big 4 audit firm, and 0 otherwise. A Big 4 audit firm is identified as Auditor 4, 5, 6 or 7 in Compustat, which are Ernst & Young, Deloitte & Touche, KPMG, and PricewaterhouseCoopers, respectively.

INDSPEC is an indicator variable, which corresponds to 1 if the auditor in that year is identified as an industry specialist, and 0 otherwise. Industry specialism is measured following Reichelt and Wang (2010). They distinguish between two types of industry specialism; national and city-level industry specialism. In this research, I use national industry specialism. An audit firm is identified as a national industry specialist if it has the largest market share of audit fees in a 2-digit SIC industry in a year. Furthermore, the definition if national level industry specialism focuses on audit firm dominance. Thus, the auditor is only an industry specialist if it (i) has the largest market share in a 2-digit SIC industry, and (ii) its market share is 10% points larger than the second largest auditor in a 2-digit SIC industry.

I construct several control variables, to control for other effects that may influence the ERC, and may possibly be associated with tenure, audit firm size and industry specialism (Ghosh & Moon, 2005; Mansi et al., 2004; Chi et al., 2009). FIRMAGE represents the number of years that a firm is listed on the stock exchange. This is computed using begin dates and end dates of stock data in the CRSP 2014 files. GROWTH equals the sum of market value of equity and the book value of debt, divided by the book value of total assets. Market value of equity is computed as Common Shares Outstanding * Price Fiscal Year Close. BETA corresponds to the systematic risk using past 60 monthly stock returns in CRSP. Stock returns are compared to expected S&P 500 returns for that stock to calculate systematic risk. SIZE is a variable that is the logarithmic transformation of market value of equity of the previous year, calculated at fiscal year-end. LEVERAGE is the final control variable. It is the ratio of total debt to total assets. Table 2 provides an overview of the calculation of variables in the regression model.

FIRMAGE is included as a control variable because it is expected that older firms are more stable, and thus have less information asymmetry. This would result in a higher ERC.

GROWTH and BETA are included to control for differences in firm quality and firm riskiness (Ghosh & Moon, 2005). It is expected that high growth firms with high earnings volatility are more risky. BETA controls for this financial risk.

(20)

20

SIZE is considered an appropriate control variable because, according to Ghosh and Moon (2005), “managers of large, politically sensitive firms are more likely to exploit the latitude in accounting to reduce political costs, which affects earnings quality” (p. 591).

Lastly, LEVERAGE is included to control for contract violations. It is argued that firms with high leverage are more likely to use gaps in accounting regulation to avoid debt covenant violations.

TABLE 2

Variable definitions

Variable name Variable definition (source)

Dependent variable

CAR Cumulative 12-month raw returns minus

cumulative 12-month value-weighted market returns (CRSP)

Test variables

E Income before extraordinary items, deflated

by market value of equity at the beginning of the year (Compustat)

ΔE Difference between income before

extraordinary items between the current and the past year, deflated by market value of equity at the beginning of the year (Compustat)

TENURE Client-auditor relationship duration in years

(Compustat)

BIG4 Dummy variable that equals 1 if the auditor

is a Big 4 audit firm, 0 otherwise (Compustat)

INDSPEC Dummy variable that equals 1 if the auditor

is an industry specialist, 0 otherwise (Audit Analytics)

FIRMAGE Number of years that a firm is listed on a

stock exchange (CRSP)

GROWTH Market value of equity plus book value of

debt, divided by book value of total assets (Compustat)

BETA Systematic risk using past 60 monthly stock

returns (CRSP)

SIZE Logarithmic transformation of market value

of equity of the prior year (Compustat)

LEVERAGE Ratio of total debt to total assets

(21)

21

3.3 Descriptive statistics

Table 3, Panel A and B present descriptive statistics of the full sample and the restricted sample, respectively. The variables CAR, E, ΔE, TENURE, FIRMAGE, GROWTH, BETA, SIZE and LEVERAGE are reported for both samples. Consistent with Ghosh and Moon (2005), the TENURE variable mean is larger (longer in years) for the restricted sample, as compared to the full sample. In the full sample, the mean of TENURE is 4.4515 years, while this is 4.8947 years in the restricted sample. The amount of years for the FIRMAGE variable also differs between the full sample and the restricted sample. In the full sample, the mean of FIRMAGE is 26.0645 years, whereas in the restricted sample, the mean of FIRMAGE equals 27.6876. The variable ranges between 6 and 89 for both samples.

TABLE 3

Descriptive statistics Panel A: Full sample

Variables Mean Min 25% 75% Max

CAR .1139 -3.0797 -.1709 .3312 10.5486 E -.0997 -78.9485 -.0337 .0627 2.4054 ΔE -.0199 -74.0439 -.0228 .0303 41.3242 TENURE 4.4515 1 2 6 12 FIRMAGE 26.0645 6 16 32 89 GROWTH 1.7329 .0102 .8493 1.9673 131.3021 BETA 1.2805 -3.4345 .6912 1.7117 8.5664 SIZE 6.0992 2.1272 4.5745 7.4885 13.1308 LEVERAGE .1889 0 .0050 .2939 4.9099

Panel B: Restricted sample

Variables Mean Min 25% 75% Max

CAR .1106 -3.0797 -.1603 .3201 10.5486 E -.0739 -78.9485 -.0189 .0641 1.9131 ΔE -.0247 -74.0439 -.0206 .0289 41.3242 TENURE 4.8947 1 2 7 12 FIRMAGE 27.6876 6 17 33 89 GROWTH 1.7364 .0342 .8653 1.9738 131.3021 BETA 1.2779 -3.4345 .7018 1.6976 8.5664 SIZE 6.2709 -.1068 4.7891 7.6504 13.1308 LEVERAGE .1880 0 .0071 .2900 4.9099

(22)

22

The descriptive statistics for CAR, E, and ΔE show few differences between the two samples. The mean of CAR in the full sample is .1139, as compared to .1106 in the restricted sample. The mean of E in the full sample equals -.0997, whereas in the restricted sample this is -.0739. Finally, the mean of ΔE in the full sample is -.0199, as compared to -.0247 in the restricted sample.

Similarly, there is few change in the variables GROWTH, BETA, SIZE and LEVERAGE. The mean of GROWTH in the full sample equals 1.7329, and 1.7346 in the restricted sample. The variable BETA has a mean of 1.2805 in the full sample, and a mean of 1.2779 in the restricted sample. The mean of SIZE equals 6.0992 in the full sample, whereas the mean in the restricted sample equals 6.2709. Lastly, the mean of LEVERAGE in the full sample is .1889, whereas the mean of LEVERAGE in the restricted sample is .1880. Overall, these results are consistent with prior research.

Table 4 presents the Pearson correlation matrix of the variables CAR, E, ΔE, TENURE, FIRMAGE, GROWTH, BETA, SIZE, and LEVERAGE. This correlation matrix is solely for the full sample. The bracketed numbers represent the significance levels of the correlation. The highest, statistically significant correlation coefficient (Pearson’s r) in the correlation matrix is that of E and ΔE (0.7303). This high, positive correlation coefficient indicates that there are significant amounts of overlapping data between the two variables. This can be explained by the fact that the variable ΔE is created with information from E. It can also be seen from the correlation matrix that all the control variables are significantly correlated with TENURE and CAR. GROWTH is negatively correlated with TENURE, whereas all the other control variables are positively correlated with TENURE. As for the correlation with CAR, FIRMAGE, SIZE, and LEVERAGE are negative and significantly correlated. GROWTH and BETA are positively correlated with CAR, and significant at the 1 % level. The significant correlation of the control variables demonstrates the importance of including them in the regression model.

In addition, other strong positive relationships are seen between FIRMAGE and SIZE, and FIRMAGE and BETA. The correlation coefficient of FIRMAGE and SIZE is 0.3281, and is statistically significant with a p-value of 0.000. This indicates that if SIZE increases, FIRMAGE increases as well, and vice versa. The correlation between FIRMAGE and BETA has a coefficient of -0.1545, and a p-value of 0.000. This implies that if BETA decreases, FIRMAGE increases, and if BETA increases, FIRMAGE decreases. Since this multicollinearity happens between two control variables, it is not such a big problem. BETA, SIZE and FIRMAGE do not

(23)

23

influence the coefficients of the variables of interest, and their function of control variables is not affected by their collinearity.

(24)

24

TABLE 4

Pearson correlation matrix

Variables CAR E ΔE TENURE FIRMAGE GROWTH BETA SIZE LEVERAGE

CAR 1.0000 E 0.1481 (0.0000) 1.0000 ΔE 0.1363 (0.0000) 0.7303 (0.0000) 1.0000 TENURE -0.0721 (0.0000) 0.0276 (0.0469) -0.0105 (0.1209) 1.0000 FIRMAGE -0.0341 (0.0000) 0.0522 (0.0000) 0.0080 (0.2373) 0.1654 (0.0000) 1.0000 GROWTH 0.1685 (0.0000) 0.0318 (0.0000) 0.0224 (0.0009) -0.0318 (0.0000) -0.0942 (0.0000) 1.0000 BETA 0.0563 (0.0000) -0.0618 (0.0000) -0.0119 (0.0796) 0.0636 (0.0000) -0.1545 (0.0000) -0.0344 (0.0000) 1.0000 SIZE -0.1628 (0.0000) 0.0855 (0.0000) -0.0120 (0.759) 0.2568 (0.0000) 0.3281 (0.0000) 0.0260 (0.0001) -0.0384 (0.0000) 1.0000 LEVERAGE -0.0365 (0.0000) -0.1222 (0.0000) -0.0338 (0.0000) 0.0143 (0.0264) 0.0491 (0.0000) -0.0283 (0.0000) -0.0145 (0.0249) 0.0939 (0.0000) 1.0000

(25)

4 Results

This chapter presents the results of the regression. Firstly, the results regarding the first hypothesis are discussed. Secondly, the results regarding the second hypothesis are discussed. The chapter will end with a discussion of the results regarding the third hypothesis. The regression carried out to come to the results is an Ordinary Least Squares (OLS) regression.

4.1 Perceptions of audit quality and audit firm tenure

The results of the full sample are presented in Table 5. The results of the restricted sample can be found in Table 6. To recap, the first hypothesis poses that investors perceive audit quality to increase with longer tenure. This is mainly due to the fact that client-specific knowledge is kept within the audit firm, and thus enables the firm to better carry out the audit. If this is the case, then the variable of interest ß3 + ß4 will be positively associated with CAR. I am also interested in the ERC, which is represented by ß1 + ß2. The combination of the ERC and the variable of interest will help to determine the effect of longer audit firm tenure on the willingness of investors to pay for earnings. In addition, since the restricted sample controls for short tenure, it is expected that the magnitude of how investors perceive audit quality to increase is greater than for the full sample.

I find that E is positively and significantly associated with CAR. However, in contrast with prior research, I find that ΔE is negatively, but not significantly, associated with CAR. The sum of the beta coefficients of E and ΔE, ß1 + ß2, is positive (.1082), and significant at the 5 % level.

Moving to the more important part of the results. The main variable of interest for this hypothesis, ß3 + ß4, equals .0092, and is statistically significant at the 10% level. The regression indicates that, when the variable of interest is divided by the ERC (ß1 + ß2), investors will pay 8,5 % more (= .0092 / .1082) for a company’s earnings if tenure increases with one year. Thus, it can be concluded that, in accordance with H1, investors perceive audit quality to increase with longer tenure.

In the restricted sample the variable of interest equals .0096, and is statistically significant at the 10 % level. The ERC in the variable of interest is .0689. This means that investors are willing to pay 13,9 % more for a company’s earnings if tenure increases with

(26)

26 one year, when short tenure is controlled for. This is consistent with the assumption that investors perceive firms that change audit firms before the client-auditor relationship reaches five years, to be shopping for opinions in order to avoid negative audit opinions or exploit accounting regulation breaches.

4.2 Perceptions of audit quality and audit firm size

The second hypothesis predicts that audit quality as perceived by investors is larger for firms that are audited by one of the Big 4 audit firms, as opposed to firms that are audited by a non-Big 4 audit firm. The variable of interest in the regression for this hypothesis is ß6 + ß7. Again, this measure will be combined with the ERC to test the impact of Big 4 audits on earnings.

The results indicate that E * BIG4 is negatively associated with the dependent variable. At -.0629, it is significant at the 1 % level. In addition, ΔE * BIG4 is positively associated with CAR. The results report a value of .1499, which is significant at the 1 % level as well. The variable of interest for this hypothesis equals .0871. It is significant at the 1 % level. This result is in line with the second hypothesis. In terms of percentages, this implies that investors will pay 80 percent more (= .0871 / .1082) for a company’s earnings if the company is audited by a Big 4 audit firm. To conclude, the results indicate that investors perceive audit quality to be enhanced when a firm is audited by a Big 4 audit firm, as opposed to when it is audited by a non-Big 4 audit firm.

The results of the restricted sample in Table 6 indicate that the variable of interest ß6 + ß7 is positive and significant at 5 %. The value of the variable equals .0764. E * BIG4 in the restricted sample takes a value of -.0803, and ΔE * BIG4 has a positive value of .1567. Both variables are significant at 1 %. The BIG4 variable itself is also positive (.1297) and significant at 1 %.

4.3 Perceptions of audit quality and industry specialism

It is argued that when a company is audited by an audit firm that is considered an industry specialist, the quality of the audit is higher. This is, logically, due to the fact that audit firms that audit more of the same type of companies, have better knowledge of the industry and the market that the firms operate in, and are thus better able to carry out an audit. The third

(27)

27 hypothesis supports this view, and poses that investors perceive audit quality to be higher for firms that are audited by an industry specialist.

From the results it can be observed that there is a positive association between INDSPEC (ß11) and CAR. Also, there is a positive association between ΔE * INDSPEC. The variable of interest for the third hypothesis, ß9 + ß10, indicates a negative association (-.0375) with the dependent variable, CAR. However, only the E * INDSPEC interaction is statistically significant, at the 1 % level. These results provide some evidence in favour of the third hypothesis. However, since the variable of interest is not statistically significant, no consistent support for the third hypothesis is found. Thus, based on this research, it cannot be confirmed that investors perceive audit quality to increase for firms that are audited by an industry specialist.

Table 6 shows that E * INDSPEC is again the only significant variable regarding the third hypothesis. With a value of -.0745 it is significant at 1 %. The variable of interest, ß9 + ß10, is negative but insignificant. All these results are in line with the results of the full sample.

4.4 Control variables

In this paragraph, the coefficients of the control variables will be discussed. As can be seen from Table 5, the coefficient of FIRMAGE (ß22) is positive (.0019) and significant at the 1 % level. In addition, both the interactions of FIRMAGE with E and FIRMAGE with ΔE are significant at 1 %. The sum of ß12 and ß13 is shows a positive sign but is insignificant. This implies that as FIRMAGE increases, CAR increases as well. In the restricted sample, the coefficient of FIRMAGE is positive and significant at the 1 % as well. The only difference with the full sample is the significance level of E * FIRMAGE and ß12 + ß13. The interaction of E with FIRMAGE’s significance fell from 1 % to 5 %. However, the significance of ß12 + ß13 changed from insignificant to significant at 10 %. This implies that FIRMAGE has a larger effect on the quality of an audit when the audit firm tenure is longer.

The interactions of E * GROWTH and ΔE * GROWTH are both positive and significant at 1 %. Also, the association of GROWTH with the ERC is positive and significant at the 1 % level. In addition, the sum of ß14 and ß15 is positive and significant at the 1 % level. It can thus be concluded that the ERC does vary with GROWTH. This finding is consistent with the findings in the restricted sample.

(28)

28 TABLE 5

Regression results ERCs and perceptions of audit quality

Variables (Coefficients) Full sample

Intercept (α) .2457(0.000)*** E (ß1) .1450 (0.000)*** ΔE (ß2) -.0368 (0.193) (ß1 + ß2) .1082 (0.018)** E * TENURE (ß3) -.0014 (0.626) ΔE * TENURE (ß4) .0106 (0.000)*** (ß3 + ß4) .0092 (0.054)* TENURE (ß5) -.0061 (0.000)*** E * BIG4 (ß6) -.0629 (0.000)*** ΔE * BIG4 (ß7) .1499 (0.000)*** (ß6 + ß7) .0871 (0.004)*** BIG4 (ß8) .1471 (0.000)*** E * INDSPEC (ß9) -.0684 (0.000)*** ΔE * INDSPEC (ß10) .0309 (0.130) (ß9 + ß10) -.0375 (0.157) INDSPEC (ß11) .0059 (0.485) Control variables E * FIRMAGE (ß12) -.0024 (0.002)*** ΔE * FIRMAGE (ß13) .0036 (0.000)*** (ß12 + ß13) .0012 (0.226) E * GROWTH (ß14) .3109 (0.000)*** ΔE * GROWTH (ß15) .0572 (0.000)*** (ß14 + ß15) .3682 (0.000)*** E * BETA (ß16) -.0087 (0.105) ΔE * BETA (ß17) .0301 (0.000)*** (ß16 + ß17) .0214 (0.033)** E * SIZE (ß18) .0287 (0.000)*** ΔE * SIZE (ß19) -.0348 (0.000)*** (ß18 + ß19) -.0062 (0.246) E * LEVERAGE (ß20) -.3743 (0.000)*** ΔE * LEVERAGE (ß21) -.2521 (0.000)*** (ß20 + ß21) -.6265 (0.000)*** FIRMAGE (ß22) .0019 (0.000)*** GROWTH (ß23) .0596 (0.000)*** BETA (ß24) .0699 (0.000)*** SIZE (ß25) -.0754 (0.000)*** LEVERAGE (ß26) .0502 (0.002)*** Adjusted R2 .1720

* notes variables that are statistically significant at 10 %. ** notes variables that are statistically significant at 5 %, and *** notes variables that are statistically significant at the 1 % level.

(29)

29 TABLE 6

Regression results ERCs and perceptions of audit quality

Variables (Coefficients) Restricted sample

Intercept (α) .2525(0.000)*** E (ß1) .0644 (0.038)** ΔE (ß2) .0045 (0.913) (ß1 + ß2) .0689 (0.169) E * TENURE (ß3) .0099 (0.005)*** ΔE * TENURE (ß4) -.0003 (0.937) (ß3 + ß4) .0096 (0.093)* TENURE (ß5) -.0064 (0.000)*** E * BIG4 (ß6) -.0803 (0.000)*** ΔE * BIG4 (ß7) .1567 (0.000)*** (ß6 + ß7) .0764 (0.037)** BIG4 (ß8) .1297 (0.000)*** E * INDSPEC (ß9) -.0745 (0.001)*** ΔE * INDSPEC (ß10) .0382 (0.164) (ß9 + ß10) -.0363 (0.236) INDSPEC (ß11) .0088 (0.315) Control variables E * FIRMAGE (ß12) -.0018 (0.033)** ΔE * FIRMAGE (ß13) .0046 (0.000)*** (ß12 + ß13) .0028 (0.058)* E * GROWTH (ß14) .3033 (0.000)*** ΔE * GROWTH (ß15) .1429 (0.000)*** (ß14 + ß15) .4462 (0.000)*** E * BETA (ß16) -.0052 (0.451) ΔE * BETA (ß17) .0281 (0.001)*** (ß16 + ß17) .0229 (0.069)* E * SIZE (ß18) .0264 (0.000)*** ΔE * SIZE (ß19) -.0397 (0.000)*** (ß18 + ß19) -.0132 (0.196) E * LEVERAGE (ß20) -.3455 (0.000)*** ΔE * LEVERAGE (ß21) -.2831 (0.000)*** (ß20 + ß21) -.6286 (0.000)*** FIRMAGE (ß22) .0017 (0.000)*** GROWTH (ß23) .0521 (0.000)*** BETA (ß24) .0617 (0.000)*** SIZE (ß25) -.0694 (0.000)*** LEVERAGE (ß26) .0421 (0.019)** Adjusted R2 .1603

(30)

30 ß16 and ß17, which represent E * BETA and ΔE * BETA, are negatively and positively associated with the ERC. ΔE * BETA is significant at the 1 % level. In addition, BETA (ß24) is also positively and significantly associated with ERC. This is consistent with the outcomes in Table 6. The sum of ß16 + ß17 is positive and significant at the 5 % level for the full sample. In case of the restricted sample, the sum of ß16 + ß17 is still positive, but significant at the 10 % level. This implies that if BETA increases, CAR increases as well.

The interactions of E * SIZE and ΔE * SIZE are both significant at the 1 % level. The interaction of E * SIZE is positive, while the interaction of ΔE * SIZE is negative. The sum of ß18 + ß19 is negative but insignificant. The variable SIZE (ß25) is negative and significant at the 1 % level. These results are similar for the restricted sample.

Lastly, the interactions of E * LEVERAGE and ΔE * LEVERAGE are both negatively associated with the ERC. The variable LEVERAGE is positively associated with the ERC (at 1 %). The sum of E * LEVERAGE and ΔE * LEVERAGE is negative and significant at the 1 % level. For the restricted sample, the variable LEVERAGE itself is significant at the 5 % level. The rest of the results are comparable to the full sample.

4.5 Sensitivity tests

Several sensitivity tests are conducted in order to test the reliability of the results. The sample contains two very important economic factors that might influence the outcomes. Firstly, as mentioned before, various large accounting scandals occurred in the years 2001 and 2002. In addition, the Sarbanes-Oxley Act was enacted in 2002 as well. This caused the notion of financial statement quality to be recognised by the general public.

Another incident that cannot be ignored is the global financial crisis that commenced in 2007. Therefore, I partition the sample in three separate subsamples to control for these events. The first subsample contain the years 2001 and 2002. The second subsample contains the years 2003 – 2006. Consequently, the final subsample contains the observations in years 2007 – 2012. The results for the sensitivity analysis can be found in Table 7. Panel A shows the results for the first subsample. Panel B reflects the results for the 2003 – 2006 subsample, and Panel C shows the results for the final subsample.

(31)

31 Table 7 shows that the results differ for the analysis in Table 5. The ERC in Table 5 Is positive and significant at the 5 % level. In Panel A, Table 7, this coefficient is not significant anymore. Also, the variable of interest for testing the effects of audit firm tenure has changed. It was positive and significant, but shows a negative and insignificant sign in the sensitivity analysis. This may indicate that the events that occurred in 2001 and 2002 have negatively influenced the perceptions of investors on audit firm tenure and audit quality. However, since this number is not significant, possibly due to the small subsample size of 3,369 observations, no concrete inferences can be made from these results. Interestingly, the variable of interest for Big 4 audit firms changes from positive and significant to negative and significant. This implies that after the accounting scandals and the enactment of SOX, Big 4 audit firms are no longer perceived to improve audit quality. The variable that indicates audit firm industry specialism remains insignificant for the 2001 – 2002 subsample.

Panel B of Table 7 also shows that the ERC is not significant as compared to the ERC in Table 5. In contrast to Table 5, the variable of interest for audit firm tenure has a negative sign, and is significant at the 1 % level. Since this negative sign is also the case for the first subsample, this can be the excess result of the events that happened in 2001 and 2002. However, in the 2003 – 2006 subsample, the variable of interest for Big 4 audit firms does change to positive again. This finding is consistent with the results in Table 5. Unfortunately, this value shows no significance yet. Again, there is no change in significance for ß9 + ß10.

Panel C, Table 7 shows the results for the subsample from 2007 – 2012. The sum of coefficients ß3 + ß4 is positive and significant at 5 %. This is in line with the results tabulated in Table 5. It seems that investors’ perceptions of audit firm tenure have changed to positive after the occurrence of the global financial crisis. Also, ß6 + ß7 is still positive and significant. From the sensitivity analysis it also is implied that investors perceive that Big 4 audit firms are better able to ensure audit quality after the financial crisis of 2007 – 2008. The variable of interest for industry specialism remains insignificant.

(32)

32 TABLE 7

Sensitivity analysis: Subsamples Panel A: Subsample 2001 – 2002

Variables (Coefficients) Subsample 1

E (ß1) -.0821 (0.660) ΔE (ß2) .4681 (0.003)*** (ß1 + ß2) .3860 (0.132) E * TENURE (ß3) .1521 (0.021)** ΔE * TENURE (ß4) -.1565 (0.025)** (ß3 + ß4) -.0044 (0.487) TENURE (ß5) -.0086 (0.683) E * BIG4 (ß6) -.2602 (0.003)*** ΔE * BIG4 (ß7) .0605 (0.373) (ß6 + ß7) -.1997 (0.099)* BIG4 (ß8) .0774 (0.009)*** E * INDSPEC (ß9) .0284 (0.557) ΔE * INDSPEC (ß10) .0866 (0.109) (ß9 + ß10) .1114 (0.124) INDSPEC (ß11) .0232 (0.283) Adjusted R2 0.3816

Panel B: Subsample 2003-2006 Subsample 2

E (ß1) .0891 (0.228) ΔE (ß2) .0887 (0.289) (ß1 + ß2) .1778 (0.129) E * TENURE (ß3) -.0135 (0.114) ΔE * TENURE (ß4) -.0373 (0.001)*** (ß3 + ß4) -.0508 (0.006)*** TENURE (ß5) -.0175 (0.000)*** E * BIG4 (ß6) .0013 (0.974) ΔE * BIG4 (ß7) .1009 (0.034)** (ß6 + ß7) .1023 (0.126) BIG4 (ß8) .0909 (0.000)*** E * INDSPEC (ß9) -.0351 (0.432) ΔE * INDSPEC (ß10) .0547 (0.244) (ß9 + ß10) .2148 (0.415) INDSPEC (ß11) -.0075 (0.483) Adjusted R2 0.1917

(33)

33

Panel C: Subsample 2007 – 2012 Subsample 3

E (ß1) .1597 (0.000)*** ΔE (ß2) -.1419 (0.001)*** (ß1 + ß2) .0178 (0.409) E * TENURE (ß3) .0138 (0.001)*** ΔE * TENURE (ß4) .0025 (0.550) (ß3 + ß4) .0163 (0.022)** TENURE (ß5) -.0038 (0.091)* E * BIG4 (ß6) -.1387 (0.000)*** ΔE * BIG4 (ß7) .2143 (0.000)*** (ß6 + ß7) .0757 (0.065)* BIG4 (ß8) .1400 (0.000)*** E * INDSPEC (ß9) -.0869 (0.002)*** ΔE * INDSPEC (ß10) .0569 (0.101) (ß9 + ß10) -.0300 (0.317) INDSPEC (ß11) .0081 (0.577) Adjusted R2 0.1751

(34)

34

5 Discussion

In this study, I examined whether audit firm tenure, audit firm size, and audit firm industry expertise have an effect on audit quality as perceived by investors. The data that was used in order to conduct this research is empirical archival data from the United States, over a time span of 2002 to 2012. Firstly, the effects of audit firm tenure on perceived audit quality were examined. Next, it was researched whether clients audited by a Big 4 audit firm, versus clients audited by a non-Big 4 audit firm, have a different relation to audit quality as perceived by investors. Lastly, the third hypothesis tested whether perceived audit quality is affected by industry specialism of the audit firm.

Using a sample of 23,992 observations from 3,484 unique forms, a positive and significant relation between audit firm tenure and the earnings response coefficient is found. It is indicated that investors, on average, are willing to pay 8.5 % more for a company’s earnings when the firm-auditor relationship increases with one year. Furthermore, when short tenure is controlled for in the restricted sample, this percentage increases to 13.9 %. The results of the first section are therefore in accordance with the first hypothesis, which poses that audit quality as perceived by investors increases with audit firm tenure. As is indicated from the literature review, there is mixed evidence regarding the relation between audit firm tenure and audit quality. On the one hand, it is argued that longer tenure exhibits a positive influence on audit quality, since client-specific knowledge is retained during a longer client-auditor relationship. On the other hand, it is debated that longer tenure impairs audit quality by means of the auditor being dependent on the client, and the client being able to influence the auditor. The results show a clear tendency towards the first argument in this specific context.

The second hypothesis poses that audit quality increases if a firm is audited by a Big 4 audit firm, as opposed to firms audited by non-Big 4 audit firms. The results indicate that investors are willing to pay a premium of 80 % for earnings of firms that are audited by a Big 4 client. Thus, from this study it is confirmed that investors perceive audit quality to increase if firms are audited by a Big 4 auditor, in comparison to non-Big 4 audit firms. These results are robust for the restricted sample. In prior literature, it is indicated that Big 4 audit firms deliver higher quality financial statements as compared to non-Big 4 audit firms. The main reason behind this is that Big 4 audit firms are less dependent on their clients, and are

The effects of audit firm tenure, audit firm size, and auditor industry expertise on audit quality as perceived by investors