Does office and individual partner industry expertise enhance audit quality?

Does office and individual partner industry

expertise enhance audit quality?

Student Douwe Wijma

Student number 2236648

Track MSc Accountancy

Supervisor Carel Huijgen Co-assessor Vlad Andrei Porumb

Date 20-06-2017

Abstract

In the past there have been numerous situations where the audited financial statements did not represent the real financial situation of a company. The gap between the reported financial situation and actual financial situation at a company determines the audit quality. We address the incomplete knowledge about the determinants of audit quality by comparing industry and non-industry experts on the office, individual, and both office and individual level and its associated audited quality of public German corporations. Archival data containing the office location, individual identity, and financial statement information between 1999 and 2009 has been used for the analysis of the potential association between industry expertise and audit quality. The results indicate that the relationship between industry expertise and audit quality is non-existent. This study contradicts previous results that find a positive association between industry expertise and audit quality. Conclusion is that financial statement users will not be able to allocate resources between investment opportunities more efficient when considering the industry expertise on the office, individual or both office and individual level in our German setting.

1 Introduction

This study tries to expand the knowledge on the transferability of industry expertise on the office and individual partner and its related audit quality. Auditing is considered an important task in creating confidence for financial statement users in the earnings reported by corporations. The quality of the audited financial statements is important when investors wish to allocate resources among different investment opportunities. Reporting restatements during the collapse of Enron in 2002 was one of the major examples where financial statement users could not rely on the

audited financial statements (Browning and Weil, 2002). The need to improve and understand audit quality became a more prominent subject of debate among regulators, investors and academics. The solutions offered after the numerous scandals were increased regulation and increasing transparency in order to increase audit quality (Zerni, 2012). Recent investigations of the AFM in the Netherlands still revealed quality problems among the audited financial

statements of big 4 accountants (AFM, 2014). The PCAOB (2013) stated that competence levels between engagements partners differ.

Audit quality is a concept that has multiple determinants that are influencing the quality of audited financial statements (Balsam et al, 2003). Audit quality is defined as “the market-assessed joint probability that a given auditor will both (a) discover a breach in the client’s accounting system and (b) report the breach” (DeAngelo, 1981). One of the determinants of audit quality is industry expertise. By using industry specialization as a determinant of audit quality we would like to measure where industry expertise is created in an audit firm. Industry specialization is defined as a differentiation strategy where investments are being made in knowledge and facilities to gain superior knowledge of an industry (Hay and Jeter, 2011). Industry experts with superior knowledge are better able to constrain earnings management. Earnings management in this thesis is defined as “the process of taking deliberate steps within the constraints of generally accepted accounting principles to bring about a desired level of reported earnings” (Davidson et al. 1987).

The relationship between industry expertise and audit quality has been studied in past by several scholars. On the firm level the industry expertise is related to allow a lower amount of accruals in

comparison of non-industry experts (Balsam et al., 2003; Krishnan, 2003). The firm level perspective assumes that the transformation of industry expertise through the firm leads to homogenous audit quality (Chi and Chin, 2011). The office and individual partner level industry expertise relaxes the assumption that audit quality is the same throughout the firm. The argument for this relaxation is that industry expertise is related to the knowledge of individual engagement partners that work at a local office (Chi and Chin, 2011). Reichelt and Wang (2010) studied the relationship between industry expertise and audit quality by comparing the effect of firm and office expertise on discretionary accruals. The combined industry expertise on both firm and office levels produced the most significant results (Reichelt and Wang, 2010). DeFond and Francis (2005) suggest that an important area for expanding the knowledge is testing the association between industry expertise and audit quality on the individual partner level. The industry expertise, on the firm and individual level, and its related audit quality have been studied recently (Chi and Chin, 2011; Gul et al., 2013). Chi and Chin (2011) report significant

relationships on the firm and partner level, however industry expertise on both levels is most significantly related towards audit quality. Gul et al. (2013) provide evidence that audit quality is explained better by industry expertise on the individual level. The link between industry expertise and audit fees is also studied on an individual level (Knechel et al. 2013; Goodwin and Wu, 2014). Knechel et al. (2013) find an association between industry expertise on the individual level and compensation. Goodwin and Wu (2014) provide research where the office and

individual partner level of industry expertise are related towards audit fees. The results reveal that industry expertise on the office level and joint firm-office is non-existent if controlled for

industry expertise on the individual level (Goodwin and Wu, 2014). Recent studies suggest that individual or both firm/office and individual industry expertise is related towards audit quality. However, there is not enough evidence from different techniques and research settings to be conclusive. A gap in existing knowledge still exists because industry expertise on the office, individual, and both office and individual level and its related earnings management are not studied so far.

This study is important for several reasons. First, we will analyze the transferability of industry expertise on the office and individual level. On the office level the industry expertise is based upon the industry expertise of engagement partners that have superior knowledge of clients that can be distributed among other engagement partners in the same audit office (Reichelt and Wang,

2010). Industry expertise on the individual level assumes that industry expertise of individual industry experts cannot be distributed among other engagement partners of a local audit office and between offices. Second, by investigating audit quality among office and individual industry specialists and non-industry specialists in Germany we can give investors additional information in order to make an informed decision in allocating resources. The identity of the engagement partner is not a mandatory requirement of the audited financial statements in most countries. The consequence is that investors are not able to relate industry expertise on the individual level towards the quality of the audited financial statements. Kilgore et al. (2012) report fund managers and analysts find individual engagement partner characteristics more important than audit firm characteristics. Third, this study may provide justification for increased transparency in other countries by presenting results that indicate the usefulness of the identity of the engagement partner in the annual report of public companies. Mandatory disclosure of the partner identity will implicitly acknowledge that the work performed, when auditing a public company, involves the work of a highly skilled engagement partner (Zerni, 2012).

We want to understand audit quality better by testing its relationship with industry expertise on the office and individual level at the same time. By testing this relationship, we want to answer the following research question:

How does industry specialization on the office and partner level in Germany relate to the audit quality of the audited financial statements in public German firms?

The paper is organized as follow. The second section contains the literature background, which reviews relevant literature and hypotheses. The third section describes the research design, sample selection, variables, and regression model used for analysis. The fourth section describes the results. The fifth section concludes this thesis.

2 Literature Background

2.1 Industry expertise and audit quality

The big four audit firms are international organizations that have administrative organizations in order to coordinate and support all the partnerships at the national level (Ferguson, 2003). The international and national country structures of the professional services firms do not directly influence the decision making of the individual auditor. The activities of the audit firms are performed in decentralized offices that are semi-autonomous (Narayanan 1995; Wallman 1996; Francis et al. 1999). National offices provide the administration between the local office and international administrative organization (Ferguson, 2003). Engagement partners on the office level provide audit services to clients and sign the audit report. However, in most cases only the name of the office can be identified and not the identity of the engagement partner. Most audit organizations have a decentralized structure and place most authority for decision making on the engagement partner at the local office level. It is the individual engagement partner that has the most influence on the activities related towards the audit.

This study will examine how industry specialization on the office and individual auditor level can be related towards audit measured by discretionary accruals. The transferability of industry knowledge on the three levels of industry expertise is the main issue of interest. The three waves of research provide three main explanation of how industry expertise is related to audit quality. The first wave of audit research literature focused on industry expertise as a firm wide

characteristic that can explain audit quality differences between the expert and non-expert firms (Balsam et al. 2003). The transferability of audit firm industry expertise can be achieved when the firm is able to spread all expert knowledge from individuals, offices, and the firm towards every level of the audit firm to achieve homogeneous industry expertise. Homogeneous audit quality is achieved when proxy variables like audit quality cannot be predicted by industry expertise on the firm, office or individual level. The assumption is that industry expertise is homogeneous between engagement partners who are employed in the same firm. The theory behind this assumption is that standard procedures and knowledge sharing makes the quality of audits uniform regardless of the different characteristics of offices and engagement partners (Zerni, 2012).

The second wave of research relaxes the assumption that industry expertise is homogeneous on a firm wide perspective.When industry expertise is an office phenomenon the transferability of specialist knowledge of industries is only efficient at the office level of the audit firm. The assumption being made is that the deep knowledge between engagement partners of the same office is being transferred in an efficient manner to achieve homogeneous audit quality on the office level, while the efficient processing of expertise knowledge towards the firm is not

possible. The engagement partners of the local offices are authorized to make the most important decisions, thus the local office industry expertise can be an important determinant of audit quality (Wallman, 1996; Francis et al, 1999). Francis et al. (2005) provided a rationale for the identified differences in quality between offices by stating that the process of auditing is related to decisions based on the client’s characteristics that occurred primarily at the office level. The decentralized structure of the local audit office makes important decisions possible that are related to the contracting with clients, performing the audit of financial statements, and issuing the audit report with a modified or unmodified audit opinion (Francis and Yu, 2009; Reichelt and Wang, 2010). Three reasons are provided which indicate that homogeneous audit quality on the firm level is not to be expected. First, the decentralized structure of audit office leads to incentive problems between the different engagement partners of different audit offices (Fama and Jensen, 1983). Second, industry expertise is not transferred throughout the firm because the engagement partners tend to solve their problems within the audit office itself (Danos et al. 1989). Third, industry expertise is gained by the investment in human capital of the skilled professional that is often client and office specific (Ferguson et al. 2003). Solomon et al. (1999) claim that industry expertise is gained by a combination of training and experiences related to auditing of financial statements on the office level. Office and client specific expertise is therefore office specific and is not distributed towards the whole audit firm (Ferguson et al. 2003).

The third wave examines the relationship between industry expertise of the engagement partner and the related audit quality. Specialization on the individual level occurs when the transferability of specialist knowledge is not efficient enough to spread all knowledge from the engagement partner towards other engagement partners in the local audit office or towards the audit firm. Audit services are used to decrease agency costs related to information asymmetry between the agent and principal (Watts and Zimmerman, 1986). Ferguson et al. (2003) describes that

individual engagement partners at the office level bear the authority and responsibility for the audit-client relationship, determining contractual engagements, execute the audit work, determine which audit work should be done, interpret the audit information that has been collected, and issue an audit report with the appropriate opinion. Engagement partners need general knowledge that is standardized among the whole firm and specific knowledge related to specific clients like the client’s internal control systems, accounting system, specialized accounting rules,

organizational activities, contracts, ownership structure, and agency conflicts (Goodwin and Wu, 2014). Specific knowledge of an engagement partner can solve agency problems because of the ability to identify problems related to accounting and contractual arrangements (Craswell et al. 1995). This kind of specific client knowledge of the engagement partner is not transferrable without practical experience within the same industries (Bonner and Walker, 1994). Decision making is decentralized in audit firms, thus differences in relevant industry knowledge and capabilities of engagement partners explain differences in audit quality. Vera-Munoz et al. (2006) provide four reasons why sharing information with audit partners is complicated. First, the

documentation and transportation of information from individual towards the office and firm level is complicated. Second, interpretation differences of the transferred information between individual engagement partners exists, which will lead towards differentiated audit quality. Third, knowledge sharing is not always possible by using IT-systems are not embraced by all audit engagement partners. Fourth, personal interest of the engagement partners that are in conflict with the principles of knowledge sharing will lead to differences in audit quality. It is expected that individual auditors have differentiated audit quality based on the difficulties related to the sharing of deep knowledge.

The three views that represent the firm, office and individual industry expertise can also relate simultaneously to the earnings quality. In this thesis we will examine the relationship between industry expertise and audit quality on the office, individual and both office and individual level. Reichelt and Wang (2010) presented significant results when investigating industry expertise and audit quality on the office level. This leads towards the following hypothesis;

Earlier studies indicate that the industry expertise of the engagement partner constrains earnings management as well (Chi and Chin, 2010; Gul et al. 2013; Knechel et al. 2013; Goodwin and Wu, 2014). This leads to the following hypothesis:

H2 There is a positive relationship between the individual industry specialization and audit quality

Previous literature indicates earnings management is constrained the most by specialists on multiple levels of analysis (Reichelt and Wang, 2010; Chi and Chin, 2011; Goodwin and Wu, 2014). This leads to the following hypothesis:

H3 There is a positive relationship between industry specialization and audit quality on both the office and individual level.

3 Research design

3.1 Sample selection

Quantitative data will be used from listed companies in Germany from between 1998 and 2010 from a database provided by the supervisor. The database provides panel data of 5930 firm year observations which includes the identity of the engagement partner, the audit office name, the audit firm name, the name of the audited public company, the industry of the public company, and financial statement data. The sample does not include companies from financial services sector that can be identified with the two digit SIC code between 6000 and 6999 (Frankel et al. 2002; Francis et al. 2005). German listed companies are audited in 56.9% of the cases by a Big N auditor and 43.1% of the cases by a Non-Big N auditor between 1999 and 2010. Because of the relative large market share of the Non-Big N auditors, we will not remove firm year observation of Non-Big N auditors.

Table 1: Sample selection for the office and individual level

Total sample 5930

Fiscal years 1998 and 2010 (917)

Industry contains less than 10 observations during an industry-year combination (2028)

Final sample size 2985

Table 1 outlines the procedures performed to select the appropriate samples on the office and the individual engagement partner level. First, we will remove all firm year observations without the required data to identify the engagement partner, audit office name, and audit firm name. This will remove all firm year observations from the fiscal years 1998 and 2010. After this procedure, we have 917 firm year observations removed. After the first selection procedure, we have a general sample size of 5013 firm year observations. We will use this sample and provide additional procedures to provide in order to calculate the portfolio shares, market shares and discretionary accruals. The minimum requirement for sample requires 10 firm-year observations per industry year combination to estimate discretionary accruals (Peasnell, 2000; Chi and Chin,

2011). Required is a minimum of 10 firm year observations over a period of 11 year. This will remove 2028 observations from the sample and we will have a final sample of 2985 firm-year observations remaining in 9 two-digit SIC industries.

3.2 Calculation of discretionary accruals

Audit quality will be measured by calculating discretionary accruals based on the cross-sectional modified Jones model. The cross-sectional modified Jones model is based on the model

developed by Jones (1991) that has been modified by Dechow et al. (1995), and Bartov et al. (2000) using cross-sectional data instead of time-series data. The difference between the Jones and the Modified Jones model arises when the change in revenues are calculated. The modified Jones model calculates the change in revenues while considering the change in receivables while the Jones model does not consider the change in receivables when calculating the change in revenues (Krishnan, 2003). Dechow et al. (1995) made this modification to the Jones model, because the discretionary accruals tend to be calculated with error for companies where managers have discretion over revenue recognition between fiscal years. The modified Jones model has been the most accurate in detecting earnings management when comparing several models which contain the Jones and modified Jones model (Dechow et al. 1995). The difference between the modified Jones model and the cross-sectional modified Jones model is that the calculation of the parameter estimates a1, a2, a3 is based on the use of cross-sectional data instead of time-series

data (Bartov et al. 2000). By using a cross-sectional version of the modified Jones model, we are able to include newly listed companies for analysis, which reduces survivorship bias (Bartov et al. (2000). Bartov et al. (2000) provided another comparison which contains the five models used for comparison by Dechow et al. (1995), and two newly developed cross-sectional Jones and modified Jones models. Results of Bartov et al. (2000) and Becker et al. (1998) indicate that the cross-sectional versions of the Jones and modified Jones models provide the best estimation methods for earnings management.

The first procedure towards the calculation of discretionary accruals based on the cross-sectional modified Jones model is the calculation of total accruals (TA). Healy (1985) and Jones (1991) calculate TA by using the balance sheet approach. The calculation of TA is as follow:

𝑇𝐴_𝑡 = ∆𝐶𝐴_𝑡− ∆𝐶𝑎𝑠ℎ_𝑡− ∆𝐶𝐿_𝑡+ ∆𝐷𝐶𝐿_𝑡− 𝐷𝐸𝑃_𝑡 (1)

Where:

TAt = Total accruals in year t

∆CAt = Change in current assets in year t

∆Casht = Change in cash and cash equivalents in year t

∆CLt = Change in current liabilities in year t

∆DCLt = Change in debt included in current liabilities in year t

DEPt = Depreciation and amortization expense in year t

After the calculation of total accruals, the regression analysis will produce the numbers for firm-specific parameter estimates a1,a2, a3 by using cross-sectional data (DeFond and Jiambalvo,

1994). The parameter estimates are industry (two-digit SIC industry) and year (fiscal year) specific instead of firm specific (Bartov et al. 2000). The equation is as follow:

𝑇𝐴𝑡 𝐴𝑡−1

= 𝑎

(

) + 𝑎

(

) + 𝑎

(

) + 𝜀

At-1 = Total Assets at the end of year t-1

∆REVt = Revenues in year t less revenues in year t-1

PPEt = Gross property plant and equipment at the end of year t

a1, a2, a3 = Firm-specific parameters

After calculating the TA and the firm-specific parameters for every fiscal year industry combination the non-discretionary accruals (NDA) can be calculated as follow:

𝑁𝐷𝐴𝑡 𝐴𝑡−1

= 𝑎

(

) + 𝑎

(

) + 𝑎

(

)

Now, we can calculate discretionary accruals(DA) as follow:

𝐷𝐴_𝑡 = 𝑇𝐴_𝑡− 𝑁𝐷𝐴_𝑡 (4)

3.3 Auditor specialization on the office level

Office specialists will be identified by using a market share approach. This approach has been used among the most scholars when identifying industry experts (Gramling and Stone, 2001; Balsam et al. 2003; Krishnan, 2003; Chi and Chin, 2011). Office industry expertise is an

independent variable used in our model and is measured indirectly as a proxy variable. This paper will use the logarithm of total assets audited by the local office as the proxy variable, which is also used by other studies within this area (Jiang et al. 2012). Industries will be categorized based upon two-digit SIC industries. The market share at the office level will be calculated by dividing the sum of logarithm of total assets audited by an office in an industry-year combination by the sum of logarithm of total assets audited within an industry-year combination. The office with the largest market share in an industry-year combination will be identified as an industry expert. The market share will be calculated as follow:

𝑀𝑆_𝑜𝑖𝑡 = ∑ 𝐿𝑁𝑎𝑠𝑠𝑒𝑡𝑠_𝑜𝑖𝑡 / ∑ 𝐿𝑁𝑎𝑠𝑠𝑒𝑡𝑠_𝑖𝑡 (5)

Where:

MSoit = Market share of audit office o within an industry i and year t combination

LNassetsoit = The sum of the natural logarithm of total assets audited by office o in

industry i and year t combination

LNassetsit = The sum of the natural logarithm of total assets audited within an industry i

and year tcombination

o = office code

i = industry code

The following example will demonstrate how the market share method is calculated. Office 289529 performs the audit of 3 clients within industry 20 and fiscal year 1999 combination, which contain 13.91 LNassets. The total industry 20 and year 1999 combination contains 31 public clients. The total LNassets within industry 20 and fiscal year 1999 combination is 148.36 LNassets. The calculated market share is then 13.91 / 148.36 is 0.0938. Office 289529 holds the largest market share within the identified industry-year combination. Thus, all 3 companies audited by office 289529 in the industry 20 and year 1999 combination will be identified as audited by an industry expert. Market shares of audit offices will be calculated for all industry-year combinations, because industry experts differ between industries and time.

3.4 Auditor specialization on individual level

Individual auditor specialists can be measured indirectly by using a proxy variable. The approach used to identify industry experts on the individual level is again the market share approach. Industry expertise will be based on the logarithm of total assets audited by the engagement partner. Individual auditor specialization is calculated by dividing the sum of natural logarithm of total assets in an individual-industry-year combination audited by an engagement partner by the sum of natural logarithm of total assets in an industry-year combination. The individual

engagement partner with the largest market share within an industry-year combination will be identified as an industry expert. The market share approach will be used to calculate industry expertise as follow:

𝑀𝑆_𝑝𝑖𝑡 = ∑ 𝐿𝑁𝑎𝑠𝑠𝑒𝑡𝑠_𝑝𝑖𝑡 / ∑ 𝐿𝑁𝑎𝑠𝑠𝑒𝑡𝑠_𝑖𝑡 (6)

Where:

MSpit = Market share of engagement partner pin industry i and year y

LNassetspit = Sum of the natural logarithm of total assets audited by engagement partner

p in industry i and Year t

3.5 Regression model

The OLS regression model in equation (7) tests if earnings management is constrained by industry expertise. The regression also contains control variables that have been selected based upon previous studies (Frankel et al. 2002; Ferguson et al, 2003; Choi et al. 2008; Gul et al. 2009; Reichelt and Wang, 2010).

𝐴𝐵𝑆𝐷𝐴𝑡 = 𝑏1+ 𝑏1𝐵𝑖𝑔5𝑡+ 𝑏2𝐿𝑜𝑠𝑠𝑡+ 𝑏3𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑡+ 𝑏4𝑆𝑖𝑧𝑒𝑡+ 𝑏5𝐵𝑀𝑡+ 𝑏6𝐶𝐴𝑇𝐴𝑡+

𝑏7𝐶𝑅𝑡+ 𝑏8𝑂𝐸(1)𝑖𝑡+ 𝑏9𝐼𝐸(1)𝑖𝑡+ 𝑏10𝐵𝐸(1)𝑖𝑡+ 𝜀 (7)

Where:

ABSDAt = Absolute discretionary accruals scaled by total assets in year t

Big 5t = Dummy variable equal to 1 if the client is audited by a Big 5 audit

firm and 0 otherwise in year t

Losst = Dummy variable equal to 1 if the client has a net loss and 0 otherwise in

year t

Leveraget = Fraction of assets financed by debt in year t

Sizet = Natural logarithm of total assets in year t

BMt = book value of equity scaled by the market value of equity in year t

CATAt = Current assets scaled by total assets in year t

CRt = Current assets scaled by current liabilities in year t

OE(1)it = Dummy variable equal to 1 if an office holds the largest market share

within an industry-year combination it and 0 otherwise

IE(1)it = Dummy variable equal to 1 if an individual engagement partner holds the

largest market share within an industry-year combination it and 0 otherwise

partner hold the largest market share within an industry-year combinationit and 0 otherwise

𝜀 = Error term

Regarding the control variables, Reichelt and Wang (2010) indicate that Big 5 clients have lower amounts of discretionary accruals. Loss is included to control for differentiation in earnings management between profitable and unprofitable public companies (Choi et al. 2010). Choi et al. (2008) claims that loss is an indicator of client specific risk, thus we expect higher discretionary accruals for companies with losses. Leverage is a control variable that is related to income

decreasing discretionary accruals (DeFond and Jiambalvo 1994; Becker et al. 1998; Frankel et al. 2002; Balsam et al. 2003). Reason for this negative association is that financial distressed firms can engage in contractual renegotiations that provides incentives for income decreasing

discretionary accruals (DeAngelo et al. 1994). Reichelt and Wang (2010) suggest that the size of the company relates towards a lower amount of discretionary accruals, which is supported by their results. Chen et al. (2011) use BM as a control variable. The results indicate that BM is related to lower discretionary accruals. CATA is a control variable that indicates the degree of audit risk and is associated with higher audit fees (Ferguson et al. 2003). The expectation is that CATA, representing audit risk, is positively associated with earnings management. CR is a control variable that represents firm complexity and is positively associated with income increasing discretionary accruals (Gul et al. 2009).

4 Results

4.1 Descriptive statistics

Descriptive statistics of all dependent, independent, and control variables are presented in table 2. The descriptive statistics contain the number of observations, means, standard deviations,

minimum values, first quartile, median, third quartile, and the maximum values of each variable.

Table 2: Descriptive statistics for the specialization variables

Variable N Mean SD Min Q1 Mdn Q3 Max

ABSDA 2985 0.108 0.129 0.000 0.031 0.073 0.137 1 TA 2985 -0.062 0.232 -7.035 -0.120 -0.056 0.000 3.572 Big 5 2985 0.610 0.489 0 0 1 1 1 Loss 2985 0.310 0.462 0 0 0 1 1 Leverage 2985 0.188 0.216 0.000 0.022 0.144 0.304 5.653 Size 2985 5.080 2.144 -0.462 3.633 4.837 6.169 12.264 BM 2985 0.739 0.936 -17.322 0.346 0.603 0.943 10.315 CATA 2985 -0.572 0.196 0.0215 0.443 0.582 0.714 0.999 CR 2985 2.955 17.749 0.0239 1.210 1.706 2.572 765.092 OE(1) 2985 0.090 0.291 0 0 0 0 1 IE(1) 2985 0.050 0.225 0 0 0 0 1 BE(1) 2985 0.020 0.156 0 0 0 0 1

The maximum value of ABSDA is 1, because measurements with discretionary accruals that surpass the amount of total assets are windsorized. The dependent variable ABSDA has a mean of 0.108 and a median of 0.073 of total assets. The total accruals are negative on average, being -0.062 of total assets. 0.09 of all financial statements is audited by an office that has been

identified as an industry expert. 0.05 of all financial statements is audited by an engagement partner that has been classified as an industry expert. And 0.02 of all clients is audited by both an

industry expert on the office and engagement partner level. 31% of all companies reports a loss. Leverage is on average 0.188 of total assets and the maximum leverage identified is 5.653. Book-to-market value is on average 0.739 and the minimum and maximum are -17.322 and 10.315. The current ratio has a mean of 2.955. However, the minimum and maximum values indicate extreme observations within our sample.

4.2 Pearson correlation matrix

Table 3 present the Pearson correlation matrix of the dependent, independent and control

variables. Based on the correlations we can identify potential significant correlations and possible multicollinearity issues between the independent variables. IE(1) and BE(1) are highly correlated. The correlation of 0.672 is not a problem. However, because we do not include both variables in our model. Surprising is the fact that ABSDA is not significantly correlated with OE(1), IE(1), and BE(1). Most studies show a significant correlation between ABSDA and the different levels of industry expertise. Unexpected is the insignificant positive association between OE(1) and ABSDA. The correlation between ABSDA and IE(1), and BE(1) are all negative, but

insignificant. The correlations between Size and OE(1), IE(1), and BE(1) are relatively high because of the use of a market share approach in determining industry expertise. The control variables Big 5, Leverage, BM, and CATA are not related with ABSDA, where a significant relationship was expected. The control variables Loss, Size and CR show their predicted associations with ABSDA.

4.3 Multivariate analysis

The results of the three models in table 4 are obtained by using equation 7 in an OLS regression model. The estimation coefficients in the first column represent the relationship of the control and independent variables with the dependent variable, when all other variables are held constant. The p-value represents the significance of the estimation coefficients of all control and

independent variables. The F-value for all three models are significant, which means the model has an appropriate fit. The adjusted R2 for model 1 is 0.025, and for model 2 and 3 it is 0.024. This indicates that a maximum of 2.5% of the variation in ABSDA is explained by our models. This means that our three models have low predictive ability.

Table 4 shows the results of the multivariate analysis of the three models from equation (7). For all three models the control variables Loss, Size, and CATA are significantly correlated with ABSDA and have the predicted direction. Big 5, Leverage, BM, and CR are not significantly correlated with ABSDA in our OLS regression models.

The first two columns in table 4 present model 1 and test the relationship between industry expertise on the office level and the discretionary accruals. Office expertise is assumed when the local audit office has the largest market share within a certain industry-year combination. The estimated coefficient is 0.009, while the associated p-value is 0.290. The results indicate that OE(1) and ABSDA are positively related to each other, but this relationship is insignificant on a 5% level. Therefore, we reject hypothesis 1 that industry expertise on the office level is positively related to audit quality. This is inconsistent compared to the results of Reichelt and Wang (2010) that report improved audit quality on the office level alone.

Table 4: Multivariate analysis of discretionary accruals and industry specialization Dependent variable is the discretionary accruals scaled by total assets

Model 1: Office level audit Model 2: individual level Model 3: Both office

specialist. audit specialist and individual level

audit specialist

Estimate p-value Estimate p-value Estimate p-value Constant 0.139 <0.001 0.138 <0.001 0.138 <0.001 Big 5 0.002 0.755 0.003 0.618 0.002 0.741 Loss 0.025 <0.001 0.025 <0.001 0.025 <0.001 Leverage -0.005 0.633 -0.006 0.597 -0.006 0.629 Size -0.005 <0.001 -0.005 <0.001 -0.005 <0.001 BM -0,005 0.070 -0.005 0.065 -0.005 0.071 CATA -0.032 0.013 -0.032 0.011 -0.032 0.013 CR 0.000 0.063 0.000 0.058 0.000 0.062 OE(1) 0.009 0.290 IE(1) -0.003 0.752 BE(1) 0.006 0.698 OOE(1) 0.009 0.340 OIE(1) -0.009 0.527 F-value 5.181 <0.001 5.122 <0.001 4.682 <0.001 Adj R2 0.025 0.024 0.024 N 2985 2985 2985

Model 2 shows the relationship between industry expertise of the individual engagement partner and discretionary accruals. The results are presented in column three and four of table 4.

Individual engagement partner expertise is assumed when the market share of the engagement partner is ranked first in an industry-year combination. Model 2 tests hypothesis 2, which expects a positive relationship between industry expertise on the individual level and audit quality. The estimation coefficient is -0.003, but the associated p-value is only 0.752. The results indicated that IE(1) and ABSDA are negatively related to each other. However, the association is

insignificant on a p<0.05 level. Therefore, we will reject hypothesis 2 that industry expertise of the engagement partner is related to industry expertise. This is inconsistent with Chi and Chin (2011), who find a positive relationship between industry expertise on the individual level and audit quality.

Model 3 shows the relationship between industry expertise on both the office and individual level and discretionary accruals as presented in the fifth and sixth column of table 4. By combining the market share approaches on both the office and individual level we will test the association between both industry expertise on the office and individual level and earnings management. Model 3 tests hypothesis 3. Model 3 also includes industry expertise only on the office level (OOE(1)) and industry expertise only on the individual level (OIE(1)). The association of OOE(1) and ABSDA is insignificant with a p-value of 0.340. The relationship between OIE(1) and ABSDA is also insignificant, because the p-value of 0.527 is beyond the 0.05 level. The association of between industry expertise on both levels (BE(1)) and ABSDA produces an estimated coefficient that has a positive value of 0.006 and its associated p-value is 0.698. That means that the relationship between BE(1) and ABSDA is insignificant. This means that a combination of both office and individual industry expertise does not produce the positive

synergies that leads towards lower abnormal accruals. The results are inconsistent with the results of other studies that combine different levels of industry expertise (Reichelt and Wang, 2010; Chi and Chin, 2011; Goodwin and Wu, 2014). This means that we reject hypothesis 3 that predicts a positive relationship between industry expertise on both levels and audit quality.

4.4 Robustness test 1

The first robustness test will analyze the stability of the model described in equation (7) by using a different market share approach for the identification of office, individual, and both office and individual industry experts. On both the office and individual level we will calculate the market share by dividing the sum of logarithm of total assets audited by an office/individual engagement partner in and industry-year combination by the sum of logarithm of total assets audited within an industry-year combination. Then we will identify the two offices and individual engagement partners with the largest market share within an industry-year combination as industry experts. The condensed results are presented in table 5 and compared with the results from table 4. Model 1 analyses the association between office industry expertise (OE(1.2)) and discretionary accruals. The results presented in table 5 show the same positive insignificant relationship between

industry expertise on the office level and discretionary accruals. Model 2 tests the relationship between the industry expertise of the individual (IE(1.2)) and discretionary accruals. The insignificant negative association between individual industry expertise and discretionary accruals is the same as the results presented in table 4. Model 3 tests the association between discretionary accruals and industry expertise on both the office and individual level (BO(1.2)). Results from table 5 indicate that OOE(1.2) and OIE(1.2) have the same insignificant positive and negative relationship with discretionary accruals. The coefficient estimates indicate that the only difference exists when the client is audited by an auditor identified as both an industry specialist on the office and individual level. Table 4 presents a positive insignificant relationship, while table 5 presents a negative but insignificant relationship between discretionary accruals and industry expertise on both the office and individual level. The relationship between industry expertise and discretionary accruals seems to be non-existent and robust in Germany based on the results from tables 4 and 5.

Table 5: Multivariate analysis of discretionary accruals and industry specialization Dependent variable is discretionary accruals

Model 1: Office level audit Model 2: individual level Model 3: Both office

specialist. audit specialist and individual level

audit specialist

Estimate p-value Estimate p-value Estimate p-value Constant 0.139 <0.001 0.138 <0.001 0.099 <0.001 OE (1.2) 0.002 0.760 IE (1.2) -0.006 0.867 BE (1.2) -0.004 0.714 OOE(1.2) 0.004 0.604 OIE (1.2) -0.006 0.647 F-value 5.122 <0.001 5.141 <0.001 4.638 <0.001 Adj: R2 0.024 0.024 0.024 N 2985 2985 2985

Control variables Big 5, Loss, Leverage, Size, BM, CATA, CR are included Year dummies from 2000 till 2009 are part of the regression model

4.5 Robustness test 2

The second robustness check tests whether the relationship between industry expertise and audit quality is stable, when using another accrual model. The model we use is developed by DeFond and Park (2001) that measures discretionary working capital accruals (DWCA). By using DWCA as a dependent variable we can compare the results from the two accrual models and determined the robustness of the results presented in table 4. The model measures realized working capital as compared to the working capital needed to support current revenue levels (DeFond and Park, 2001). The difference between realized working capital and working capital needed to support current revenues is the discretionary working capital accrual (DeFond and Park, 2001). DeFond and Park (2001) assume that discretionary working capital accruals are unsustainable and tend to align with earnings over time. The model is described as follow:

𝐷𝑊𝐶𝐴_𝑡 𝐴_𝑡−1

=

−

DWCAt = Discretionary working capital accruals in year t

WCt = Non-cash working capital in the current year computed as (current assets –

cash and short-term investments) – (current liabilities – short term debt) in year t

WCt-1 = Working capital in year t

Revt = Revenues in year t

REVt-1 = Revenues in year t-1

The calculation of DWCA contains three additional procedures. First, we will remove 6

observations from clients where revenues are zero. Second, we will calculate the absolute value of the discretionary working capital accruals. The last procedure is to windsorize all observations where DWCA is larger than lagged assets.

The condensed results are presented in table 6 and will be compared with the results in table 4. Model 1 displays the same positive insignificant relationship between industry expertise on the office level and earnings management. Model 2, which measures industry expertise on the partner level, presents the same negative insignificant association between industry expertise and

earnings management. Model 3 presents negative insignificant results for the relationship between OOE(1) and OIE(1) and discretionary accruals, which are similar in both tables. Model 3 also examines the association between industry expertise and discretionary accruals on both the office and individual level (BIE(1)) and finds the same insignificant positive relationship as presented in table 4. The two models presented in table 4 and table 6 have similar results,

therefore we can assume that our insignificant results are robust. This means that the association between industry expertise and audit quality does not exist in our sample of German companies.

Table 6: Multivariate analysis of discretionary working capital accruals and industry specialization

Dependent variable is abnormal working capital

Model 1: Office level audit Model 2: individual level Model 3: Both office

specialist. audit specialist and individual level

audit specialist

Estimate p-value Estimate p-value Estimate p-value Constant 0.106 <0.001 0.105 <0.001 0.105 <0.001 OE(1) 0.009 0.299 IE(1) -0.007 0.500 BE(1) 0.008 0.594 OOE(1) 0.007 0.444 OIE(1) -0.019 0.193 F-value 5.220 <0.001 5.184 <0.001 4.782 <0.001 Adj: R2 0.025 0.025 0.025 N 2979 2979 2979

Control variables Big 5, Loss, Leverage, Size, BM, CATA, CR are included Year dummies from 2000 till 2009 are part of the regression model

5 Discussion and Conclusion

Previous research results showed that the positive association between industry expertise and audit quality is the strongest when using a model where industry expertise on multiple levels are tested simultaneously (Reichelt and Wang, 2010; Chi and Chin, 2011). The models show a stronger negative relationship with discretionary accruals when combining the multiple levels of industry expertise together. This thesis adds towards this stream of research by conducting tests of industry expertise on multiple levels that have not tested earlier by analyzing the relationship between industry expertise on the office and individual level and audit quality.

For both the dependent and independent variables we use proxies, because industry expertise and earnings quality cannot be observed directly. The analysis is based on the financial information from German companies over a period from 1999 to 2009. We have used a market share approach for the identification of industry expertise on the office and individual level.

Discretionary accruals are calculated by using a cross-sectional version of the modified Jones model. The tests undertaken in this thesis do not show an association between industry expertise and audit quality. When performing two robustness tests we again do not find an association between industry expertise and audit quality.

The results can be used to provide evidence that the association between industry expertise and audit quality may be non-existent. Results from the linear regression indicate that no relationship exists between industry expertise and earnings quality of the audited clients in Germany. Adding the identity of the auditor in the audited financial statements does not increase the decision usefulness of the annual reports of German corporations even when the engagement partner is identified as an industry expert. Trying to create industry expertise by investing in facilities related towards increasing the efficiency of the transformation of knowledge might not an appropriate strategy. The results from this study indicate that these investments that stimulate knowledge sharing do not result in industry expertise.

Two limitations are related to our study. The first limitation in our study is the use of proxy variables in order to determine audit quality and industry expertise. Discretionary accruals are used as a proxy for audit quality and a market share approach is used to determine industry

experts on the office and individual level. There are multiple methods that can be used as a proxy for industry expertise and earnings quality, however which proxy has the best fit with industry expertise is still unknown. The second limitation is that industry experts specialized in smaller industries will not be classified as industry experts when we use the market share approach as a proxy variable.

Future research should develop a more comprehensive model which captures all levels of

industry expertise and its related audit quality. This analysis should include industry expertise on the firm, office and individual level simultaneously in order to study the relationships between industry expertise and earnings quality in a more complete setup. Another element is the

development of new proxies for industry expertise and earnings quality that measure the variables more accurately. An example is provided by Bernard and Skinner (1996) who find situations where discretionary accruals and non-discretionary accruals are misclassified by the Jones model, which has many similarities with the cross-sectional modified Jones model used in this thesis. Last area for future research is clarifying the relationship between industry expertise and audit quality. We are assuming in this thesis that the industry expertise of an office or engagement partner is a possible determinant of audit quality. Krishnan (2003) states that clients who report lower discretionary accruals might be tended to self-select industry experts as auditors. This is an important area for future research because the motives of choosing an industry specialist or non-industry specialist are unknown (Craswell, 1995).

6 References

Authoriteit Financiele Markten(AFM). 2014. Uitkomsten onderzoek kwaliteit wettelijke controles Big 4-accountantsorganisaties

Balsam, S., Krishnan, J., & Yang, J. S. (2003). Auditor industry specialization and earnings quality. Auditing: A Journal of Practice & Theory, 22(2), 71-97.

Bartov, E., Gul, F. A., & Tsui, J. S. (2000). Discretionary-accruals models and audit qualifications. Journal of accounting and economics, 30(3), 421-452.

Becker, C. L., DeFond, M. L., Jiambalvo, J., & Subramanyam, K. R. (1998). The effect of audit quality on earnings management. Contemporary accounting research, 15(1), 1-24.

Bernard, V. L., & Skinner, D. J. (1996). What motivates managers' choice of discretionary accruals?. Journal of Accounting and Economics, 22(1), 313-325.

Bonner, S. E., & Walker, P. L. (1994). The effects of instruction and experience on the acquisition of auditing knowledge. Accounting Review, 157-178.

Browning, E. S., & Weil, J. (2002). Accounting woes roil stock markets as nervous investors stampede exits. Wall Street Journal.

Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and losses. Journal of accounting and economics, 24(1), 99-126.

Chen, H., Chen, J. Z., Lobo, G. J., & Wang, Y. (2011). Effects of audit quality on earnings management and cost of equity capital: Evidence from China. Contemporary Accounting

Research, 28(3), 892-925.

Chi, H. Y., & Chin, C. L. (2011). Firm versus partner measures of auditor industry expertise and effects on auditor quality. Auditing: A Journal of Practice & Theory, 30(2), 201-229.

Choi, J. H., Kim, C., Kim, J. B., & Zang, Y. (2010). Audit office size, audit quality, and audit pricing. Auditing: A Journal of practice & theory, 29(1), 73-97.

Choi, J. H., Kim, J. B., Qiu, A., & Zang, Y. (2008). Auditor locality, audit quality and audit pricing.

Craswell, A. T., Francis, J. R., & Taylor, S. L. (1995). Auditor brand name reputations and industry specializations. Journal of accounting and economics, 20(3), 297-322.

Danos, P., Eichenseher, J. W., & Holt, D. L. (1989). Specialized knowledge and its communication in auditing. Contemporary Accounting Research, 6(1), 91-109.

Davidson, S., Stickney, C., & Weil, R. (1987). Accounting: The language of business. Thomas Horton and Daughter. Sun Lakes Arizona.

DeAngelo, L. E. (1981). Auditor size and audit quality. Journal of accounting and

economics, 3(3), 183-199.

DeAngelo, H., DeAngelo, L., & Skinner, D. J. (1994). Accounting choice in troubled companies. Journal of accounting and economics, 17(1), 113-143.

Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings management. Accounting review, 193-225.

DeFond, M. L., & Francis, J. R. (2005). Audit research after sarbanes-oxley. Auditing: A Journal

of Practice & Theory, 24(s-1), 5-30.

DeFond, M. L., & Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals. Journal of accounting and economics, 17(1), 145-176.

DeFond, M. L., & Park, C. W. (2001). The reversal of abnormal accruals and the market valuation of earnings surprises. The Accounting Review, 76(3), 375-404.

Fama, E. F., & Jensen, M. C. (1983). Agency problems and residual claims. The Journal of Law

and Economics, 26(2), 327-349.

Ferguson, A., Francis, J. R., & Stokes, D. J. (2003). The effects of firm-wide and office-level industry expertise on audit pricing. The accounting review, 78(2), 429-448.

Francis, J. R., Stokes, D. J., & Anderson, D. (1999). City Markets as a Unit of Analysis in Audit Research and the Re‐Examination of Big 6 Market Shares. Abacus, 35(2), 185-206.

Francis, J. R., Reichelt, K., & Wang, D. (2005). The pricing of national and city-specific

reputations for industry expertise in the US audit market. The accounting review, 80(1), 113-136. Francis, J. R., & Yu, M. (2009). The effect of Big Four office size on audit quality. The

Accounting Review, 84(5), 1521–1552.

Frankel, R. M., Johnson, M. F., & Nelson, K. K. (2002). The relation between auditors' fees for nonaudit services and earnings management. The accounting review, 77(s-1), 71-105.

Goodwin, J., & Wu, D. (2014). Is the effect of industry expertise on audit pricing an office-level or a partner-level phenomenon?. Review of Accounting Studies, 19(4), 1532-1578.

Gramling, A. A., & Stone, D. N. (2001). Audit firm industry expertise: A review and synthesis of the archival literature. Journal of accounting literature, 20, 1.

Gul, F. A., Fung, S. Y. K., & Jaggi, B. (2009). Earnings quality: Some evidence on the role of auditor tenure and auditors’ industry expertise. Journal of Accounting and Economics, 47(3), 265-287.

Gul, F. A., Wu, D., & Yang, Z. (2013). Do individual auditors affect audit quality? Evidence from archival data. The Accounting Review, 88(6), 1993-2023.

Hay, D., & Jeter, D. (2011). The pricing of industry specialisation by auditors in New Zealand. Accounting and business research, 41(2), 171-195.

Healy, P. M. (1985). The effect of bonus schemes on accounting decisions. Journal of accounting

and economics, 7(1-3), 85-107.

Jiang, L., Jeny-Cazavan, A., & Audousset-Coulier, S. (2012). Who are Industry Specialist Auditors?.

Jones, J. J. (1991). Earnings management during import relief investigations. Journal of

accounting research, 193-228.

Kilgore, A., R. Radich, and G. Harrison. 2012. Drivers of audit quality: The importance of audit-team and auditfirm attributes. Working paper. Macquarie University.

Knechel, W. R., Niemi, L., & Zerni, M. (2013). Empirical evidence on the implicit determinants of compensation in Big 4 audit partnerships. Journal of Accounting Research, 51(2), 349-387. Krishnan, G. V. (2003). Does Big 6 auditor industry expertise constrain earnings

management?. Accounting horizons, 17, 1.

Kwon, S. (1996). The impact of competition within the client's industry on the auditor selection decision. Auditing, 15(1), 53.

Narayanan, V. G. (1995). Moral hazard in repeated professional partnerships. Contemporary

Accounting Research, 11(2), 895-917.

Neal, T. L., & Riley Jr, R. R. (2004). Auditor industry specialist research design. Auditing: A

Journal of Practice & Theory, 23(2), 169-177.

Peasnell, K. V., Pope, P. F., & Young, S. (2000). Detecting earnings management using cross-sectional abnormal accruals models. Accounting and Business research, 30(4), 313-326. Public Company Accounting Oversight Board (PCAOB). 2013. PCAOB Reproposes

Amendments to Improve Transparency by Requiring Disclosure of the Engagement Partner and Certain Participants in the Audit. Available at:

Reichelt, K. J., & Wang, D. (2010). National and office‐specific measures of auditor industry expertise and effects on audit quality. Journal of Accounting Research, 48(3), 647-686.

Solomon, I., Shields, M. D., & Whittington, O. R. (1999). What do industry-specialist auditors know?. Journal of accounting research, 37(1), 191-208.

Vera-Munoz, S. C., Ho, J. L., & Chow, C. W. (2006). Enhancing knowledge sharing in public accounting firms. Accounting Horizons, 20(2), 133-155.

Wallman, S. M. (1996). The future of accounting, part III: Reliability and auditor independence. Accounting Horizons, 10(4), 76.

Watts, R. Zimmerman (1986). Positive accounting theory. Edgewood Cliff, NJ. Zerni, M. (2012). Audit partner specialization and audit fees: Some evidence from Sweden. Contemporary Accounting Research, 29(1), 312-340.