Audit fee premium for national and local level of industry expertise in Italy

Purpose – This study aims to examine whether national industry expertise in Italy is more dominant than local expertise. Prior studies from Australia, USA and UK show that audit fees for industry experts are priced at a higher premium at the local level than the national level. These countries have voluntary audit ﬁ rm rotation, while Italy has mandatory audit ﬁ rm rotation (MAFR). The authors predict that Italy has a stronger nationalthanlocal level of industryexpertise,to better retain andtransferindustry expertise. Design/methodology/approach – The authors compare audit fee premiums of national industry experts to local levels,usingquantitative (multivariate tests) andqualitative(interviews)methodology. Findings – Using hand-collected audit fees, the authors ﬁ nd that the audit fee premium for industry expertise is greater at the national level than the local level. The authors ﬁ nd corroborating results with audit hours. To provide further support, the authors conduct analysis for a neighboring country that does not have audit ﬁ rm rotation. Using hand-collected data from Germany, the authors ﬁ nd that audit fee premiums from nationalindustry expertisearenodifferent from localindustryexpertise. Originality/value – The present study study has theoretical and practical implications, for European Unioncountries, whichrecently adopted MAFR andforcountriesconsideringadoptioninthefuture. is a moment in which I have to consider how not to lose the competence in that industry because maybe after nine years the will come back, and I will also have to consider how to gain competence in the new industry. Industry expertise is a conduit for the transfer of auditors ’ industry speci ﬁ c has higher client turnover than the USA, because of a nine-year MAFR requirement, and it also has a smaller publicly listed client audit market. We predict that national industry expertise is stronger than area industry expertise because higher client turnover, and a small audit market, increases the sensitivity of industry-speci ﬁ c knowledge being lost at the local area level. We ﬁ nd that audit fees of industry specialists at the national level and the area level are greater than fees of non-specialists. We do not ﬁ nd the same evidence for city industry specialists. However, unlike prior studies restricted to countries with voluntary audit ﬁ rm rotation (Ferguson et , 2003; Francis et al. , 2005; Basioudis and Francis, 2007), we ﬁ nd that national industry specialization dominates over area industry specialization in audit fee premiums and audit hours, suggesting there is greater centralization of industry specialization in Italy. We conduct the same analysis for companies located in Germany, which has a similar legal and regulatory system, but does not have MAFR. We ﬁ nd no difference between the audit fee premiums of national industry specialists and area industry specialists. However, the sample size for Germany is larger. We conclude that the higher client rotation of MAFR is a contributing factor, as well as a smaller audit market size. Our study has a number of implications for academics, regulators and legislators considering adopting MAFR.


Introduction
This study examines whether the national level of auditor industry expertise is more dominant than the local level of expertise in Italy. Prior studies (Basioudis and Francis, 2007;Ferguson et al., 2003;Francis et al., 2005) find that local industry experts charge a higher audit fee premium, suggesting that local expertise dominates in the USA, Australia and the UK. However, in other contexts, audit expertise may be distributed differently, perhaps because mandatory audit firm rotation (MAFR) is in place, as it is in the Republic of Italy (1975), or perhaps because the local market for audits of publicly traded companies is smaller. To the best of the authors' knowledge, no previous studies have investigated whether these differences involve the organization of audit industry expertise at national or local level. We predict that Italy should have a stronger national level of industry expertise, because higher client turnover under MAFR (Bandyopadhyay et al., 2014) reduces the stability of client portfolios at local level, so more expertise remains at national than local level. The smaller local market size also generates fewer industry experts at local than national level. In turn, the national level of expertise means that expertise is more efficiently transferred to offices in the national network on a just-in-time basis as new clients are accepted. The study may supply indications useful for European Union countries which have recently adopted MAFR in achieving a balance between audit independence and competencies.
We find that Italian auditors focus on national industry expertise rather than local industry expertise to offer differentiated audit quality, as shown by higher industry audit fee premiums at national level than local level. We corroborate this finding with audit hour premiums. We show that industry specialists, a high-quality type of auditor, adapt to the challenges of MAFR by focusing their expertise at national instead of local level.
The research contributes to the literature in several ways. Firstly, it shows that industry expertise is differently organized in countries with small audit markets, and audit fee premiums are at national rather than city level, which is widely found by prior literature. Secondly, it shows how MAFR affects industry expertise. MAFR has been the subject of debate for several decades, and reemerged following the 2007-09 financial crisis (European Commission, 2010;PriceWaterhouseCoopers LLP, 2013;Public Company Accounting Oversight Board (PCAOB), 2011;The Conference Board, 2003). While Italy has had a MAFR requirement in place for more than four decades, the EU recently promulgated MAFR effective for the 2016 fiscal year (European Parliament, 2014). Opponents argue that MAFR impairs industry expertise because a shorter tenure impairs auditor acquisition of knowledge (Jamal, 2012). Recent US-based studies suggest that industry expertise requires sufficient time and in-depth learning (Francis et al., 2017;Gaver and Utke, 2019). Consequently, MAFR can deteriorate office industry expertise when clients mandatorily change audit firms, causing personnel to either change industry specializations or relocate (Daugherty et al., 2012). We contribute to the literature by showing how results differ in a MAFR regime compared to a voluntary rotation regime, where prior studies find the local level is dominant (Basioudis and Francis, 2007;Ferguson et al., 2003;Francis et al., 2005;Numan and Willekens, 2012). The dominance of the local level, shown in prior studies, does not apply to all countries. We corroborate our audit fee premium results by conducting the same analysis for Germany, a neighboring country with a similar legal system (code law) but without MAFR in our sample period, and a closer audit market (Abbott et al., 2018) size than the USA. We find that the audit fee premium for national-level industry leaders is no different for local-level industry leaders, providing further support that MAFR contributes to a stronger national level of industry expertise. We infer causality by using a difference in differences design and using Germany as a control group; we find robust results. Our research design also controls for country-level and local audit market effects.
Thirdly, the study contributes to literature by analyzing whether audit fee premiums of industry experts reflect additional audit effort or are simply economic rents for holding more dominant client market shares. On the one hand, prior studies show that clients of industry experts have higher audit quality (Balsam et al., 2003;Dunn and Mayhew, 2004;Krishnan, 2003;Reichelt and Wang, 2010), suggesting that audit fee premiums reflect additional audit effort. On the other hand, there is also evidence that audit fee premiums reflect economic MAJ rents (Numan and Willekens, 2012). If audit fee premiums reflect effort, rather than economic rents, we should expect to find higher audit hours, as we in fact find in this study. A study by Bae, Choi, and Rho (2016) provides evidence from South Korea that audit fee premiums reflect greater audit hours and thus greater audit effort. However, their study only examines national industry expertise and the sample covers a time period beginning with voluntary audit firm rotation followed by mandatory audit firm rotation. We extend the findings of Bae et al. (2016) by showing not only that national industry experts expend more audit hours than non-experts in a MAFR regime, but also that local industry experts expend more audit hours.
Our paper is organized as follows. Section 2 discusses the background literature, and predictions. Section 3 discusses the empirical design, including the sample and descriptive statistics. Section 4 reports the results, including robustness tests. Section 5 concludes.
2. Literature and hypotheses 2.1 National and office level industry expertise Industry expertise stems from the investment in human capital in accounting professionals and the experience they gain from servicing clients (Ferguson et al., 2003;Francis et al., 2005;Solomon et al., 1999). National industry expertise and local industry expertise (hereafter, office industry expertise) are distinct levels of auditor industry expertise. National industry expertise is characterized by a firm-wide reputation for industry expertise, organization of expertise by industry, and various knowledge-sharing practices (Francis et al., 2005;Hogan and Jeter, 1999;Solomon et al., 1999). Knowledge-sharing practices include internal benchmarking of best practices, standardized audit programs, industry-specific databases, training programs, and the transfer of expert personnel to other offices through travel and consulting arrangements (Francis et al., 2005;Zerni, 2012) [1]. Office-level industry expertise consists of deep industry and client knowledge possessed by audit personnel within a particular office (Francis et al., 2005). Human capital attributes are important to develop audit firm performance (Samagaio and Rodrigues, 2016). Audit personnel gain their expertise from the direct experience of working exclusively on industry-specific engagements (Solomon et al., 1999).
The two levels of industry expertise transfer knowledge in different ways. National industry expertise is transferred from the national headquarters (or through an office expert) to offices within the national network through standardized firm policies and procedures and knowledge-sharing practices. Office industry expertise is transferred within the office by quality control procedures (supervision/coaching) and individual on-the-job experience.
National industry expertise ensures expertise is retained within the audit firm, and is transferred to offices that are in need of it. Mandatory audit firm rotation increases the importance of the national level, because if a sufficiently large number of clients in a particular industry rotate to another firm, the office will lose the knowledge and it will eventually be forgotten (Causholli, 2016). To retain the industry expertise in the audit firm, distinct knowledge-sharing practices and firm-wide policies and procedures are needed at the national level. As new industry-specific clients are accepted, offices will require industry-specific knowledge to be transferred at the national level. While there are distinct differences in national and office-level expertise, more germane to our study is understanding which level dominates by audit fee premium.
Prior studies provide evidence that the office level of industry expertise dominates in audit fee premium over national level (Ferguson et al., 2003;Francis et al., 2005;Basioudis and Francis, 2007;Cahan et al., 2011;Numan and Willekens, 2012;Fung et al., 2012;Bae et al., 2016): audit fee premiums for national industry experts range between statistically insignificant amounts to 8%, and between 13% and 16% for office industry experts. All of these studies are based on data from countries where audit firm rotation is voluntary.
Thus, it is unclear whether audit fee premiums differ between national industry expertise and office industry expertise under MAFR. To explore this matter further, we discuss the nature of audit fee premiums and industry expertise.

Audit fee premiums and industry expertise
Audit fee premiums consist of additional audit effort and economic rents (Okeefe et al., 1994;Simunic, 2014). Audit firms charge a fee premium when they adopt a differentiation strategy (Cahan et al., 2011;Francis et al., 2005;Foster and Shastri, 2016;Scott and Gist, 2013). Arguably, they deliver higher audit quality by expending greater direct labor hours and make greater investments in knowledge sharing. It is well documented that industry expert auditors are associated with higher audit quality, as evident from studies comparing earnings quality, disclosure quality and going concern opinion frequency ( Almutairi et al., 2009;Balsam et al., 2003;Carcello and Nagy, 2004;Dunn and Mayhew, 2004;Krishnan, 2003;Reichelt and Wang, 2010). Kamarudin et al. (2022) find that switching to specialist auditors will result in increased conditional conservatism, while switching from specialist auditors to non-specialist auditors will result in reduced conditional conservatism in Asian countries.
Higher audit quality also indicates that auditors expend greater audit hours (Palmrose, 1986, 108), as evident from audit production theory (Okeefe et al., 1994;Simunic, 2014). Audit effort lowers the likelihood of audit failure and improves accounting information quality, indirectly increasing clients' future equity return performance (Lee et al., 2021). Greater audit hours are expended on more effectively constraining management's attempts to manipulate earnings with discretionary accruals in both magnitude and income-increasing types (Caramanis and Lennox, 2008). In their study using South Korean data, Bae et al. (2016) report that national industry expert auditors expend greater audit fees (and earn higher audit hours) than non-experts. They argue that industry expert auditors will expend more hours because they require additional evidence, and they protect their reputation in a competitive market (Bae et al., 2016).
However, their results may not generalize to our setting because 1) South Korea has weaker investor protection (Wingate, 1997), including lower audit quality and weaker enforcement of accounting standards (Brown et al., 2014), 2) their sample period includes a mandatory audit firm rotation regime followed by a voluntary firm rotation regime, and 3) they do not examine the local level of auditor expertise. Thus, our results are more generalizable to other EU countries which recently adopted MAFR reforms effective from 2016 (European Parliament, 2014). To better understand our setting, we turn our discussion to Italy's regulatory environment.

Mandatory auditor rotation and regulatory developments in Italy
Italy adopted mandatory audit firm rotation in 1975 under Presidential Decree D.P.R. No. 136 (Republic of Italy, 1975). Under this law, publicly listed companies must mandatorily rotate their audit firm every nine years. The European Parliament passed Regulation No. 537 in 2014, requiring a ten-year MAFR term, with an additional ten-year term if the audit engagement is put out for public bid (European Parliament, 2014).
Studies examining audit fees and audit hours within MAFR regimes are limited, and do not examine audit fee premiums of industry specialists. Cameran et al. (2015) find in Italy that after a client rotates to another audit firm, audit fees are lower and audit hours are higher in the first year. Kwon et al. (2014) show in South Korea when the country changed from a voluntary audit firm rotation regime to a mandatory rotation regime, audit fees and audit hours increased. Kwon et al. (2014) compare national industry market shares in South Korea during the time of change to MAFR and find a decline in national industry market MAJ shares after MAFR was implemented. Practitioners from other countries express considerable pushback against the potential efficacy of MAFR to improve audit quality because of various switching costs, notably the loss of client-specific knowledge and expertise upon rotation (Harber and Maroun, 2020).

Hypothesis
Prior literature finds that in the USA, UK and Australia, industry specialization is a joint effect of a national level and a local office level, and that the local level dominates over the national level in audit fee premium (Basioudis and Francis, 2007;Ferguson et al., 2003;Francis et al., 2005). However, the institutional and audit market characteristics of Italy differ from the countries in these studies, in that the national level is arguably dominant over the local level. We argue that Italy's higher client turnover and its smaller publicly listed audit market exacerbate the loss of tacit industry-specific knowledge at the local level, and consequently the seriousness of lost industry expertise. To retain industry expertise in the audit firm, a stronger national level of industry expertise is needed.
Italy's nine-year mandatory audit firm rotation term shortens audit firm tenure and increases the turnover of publicly traded clients. Compared to the USA, Italy's client turnover is higher. In the USA, audit firm tenure averages 12 years (Davis et al., 2009), while in Italy, audit firm tenure averages 4.5 years. Higher client turnover leads to a greater risk that industry-specific knowledge is forgotten in the office, when there are insufficient industry clients to maintain the learning experience (Causholli, 2016). Weakened industry expertise is echoed by opponents of MAFR who claim that the accretion of industry expertise is impaired by a shorter client-auditor engagement term (Jamal, 2012). To retain industry-specific knowledge in the firm and transfer it to offices in need of the expertise, the national level should be more dominant than in countries with voluntary audit firm rotation.
In Italy, higher client turnover from MAFR and a smaller audit market for publicly traded clients arguably make the national level of industry expertise more important for retaining and transferring industry-specific knowledge to offices within the audit firm.
Turning the discussion to audit fee premiums of industry experts, prior studies report that industry experts incur an audit fee premium to compensate for their industry expertise (Basioudis and Francis, 2007;Ferguson et al., 2003;Francis et al., 2005;Reichelt and Wang, 2010). A consequence of stronger national industry expertise is that audit fee premiums are higher than local industry expertise. These studies have defined industry experts as the industry market leaders. Our first hypothesis follows.
H1. Audit fee premiums are higher for national-level industry leaders than for local-level industry leaders.

Industry specialization measures
We use two levels of industry expertise: a national level and a local level. For the national level, we define a national industry specialist as the audit firm that is the leader in market share of client sales for a particular industry and year (Bae et al., 2016). Use of sales, rather than audit fees, mitigates the mechanical relation between audit fee premiums and auditor market share, identified by Minutti-Meza (2013). We also use client sales, rather than audit fees, because the sample is larger, which yields a more accurate market leadership measure. In our robustness tests, we find support for this choice. Industries are classified by the Audit fee premium Global Industry Classification Standard (GICS). GICS is superior to the Standard Industrial Classification (SIC) in capital market research applications (Bhojraj et al., 2003). In Italy, the Big 4 local level is organized by area (4.2.1 Big 4 Partner interviews). The majority of audit fee revenues (untabulated, 56%) are from the two area head offices (Milan -44% and Rome -12%). Rather than using the Metropolitan Statistical Area (MSA), used in the USA (Reichelt and Wang, 2010), or the geographical city as in the UK (Basioudis and Francis, 2007), we use area, which is similar to the MSA. Studies that use US data specify auditor local specialization in terms of the MSA in which the audit office is located. The US Census Bureau defines an MSA as the core of the city and the surrounding areas that have commuter relations with the core. The MSA represents a geographic-economic unit within which individuals have frequent and close contact by commuter relations with the core. The frequent and close contact facilitates the transfer of tacit knowledge. The European Commission (2014) statistics division (EUROSTAT) defines European areas based on the Nomenclature of Territorial Units for Statistics (NUTS). The NUTS area is similar to the US MSA because people who reside in an area typically have close family contacts, and travel by car or train is convenient, inexpensive and does not require an overnight stay.
We choose the two major industrial areas of Italy: The North and the Center-South. The two areas are representative of Italy's industrial activity [2]. At least two other studies concur that many countries have one or two major cities were publicly listed clients are located (Lim and Tan, 2009). Industries tend to cluster geographically (Porter, 1990, p. 154) with the greatest number in the North area (National Observatory of Italian Districts, 2009). Italy's economy is largely driven by the manufacture of high-quality consumer goods, and consists of a developed industrial north and a less developed agricultural south (U.S. Central Intelligence Agency, 2017). We also provide an analysis using the city level, based on the 20 government regions, to support our conjecture that the area level is the most appropriate for this study. Following prior literature (Basioudis and Francis, 2007;Francis et al., 2005;Reichelt and Wang, 2010), a Big 4 audit firm is considered an industry leader if there is at least one other audit firm in the same area, industry and year. This restriction ensures that a single audit firm in a particular area and industry is not classified as an industry expert because of an uncompetitive market (Francis et al., 2005).

Multivariate tests
To test our prediction, we estimate equation (1): Equation (1) is estimated with industry-and year-fixed effects to control for systematic differences in audit fees across industries and across time. We cluster robust standard errors by client firm to control for heteroskedasticity and the lack of temporal independence of residuals within client firms (Petersen, 2009), and to avoid firm-fixed effects estimation when it is not feasible (Wooldridge, 2016, p. 450) [3]. Variable definitions are described in Appendix.

MAJ
LAF is the log of total audit fees. Total audit fees include the audit of the financial statements of the consolidated entity, and the audit of the individual financial statements of the parent and the subsidiaries (group accounts). We include not just the audit of the consolidated entity but also of the group accounts to better measure complexity (size, international presence and dominance of the parent company) and related audit effort [4]. These audit fees do not include non-audit fees for non-audit services because Italy restricts auditors from providing non-audit services. LAH is the log of audit hours for the total audit fees. Equation (1) estimates audit hours using the same model as audit fees, following prior studies (Bae et al., 2016;Cameran et al., 2015).
INDUSTRY LEADER is the variable of interest and denotes three indicator variables for industry specialization: a national industry leader, an area industry leader and a city industry leader [5]. To test our hypothesis, we expect the coefficient on the INDUSTRY LEADER variables to be positive, and we expect that the coefficient on the national leader is greater than that of the area industry leader.
Control variables for estimating audit fees are based on prior studies examining audit fees and earnings quality (Basioudis and Francis, 2007;Francis et al., 2005;Reichelt and Wang, 2010).
[6] Client size (SIZE) is the log of total assets. Several prior studies measure client risk by CATA, QUICK, ROA, LEV and UNCLEAN OPINION. We also control for volatility of operating cash flows (s(CFO)) (Graham et al., 2005), SALES GROWTH for economic life cycle (Dickinson, 2011), TENURE for growth in audit fees with audit firm tenure (Cameran et al., 2015) and total accruals magnitude (jACCR_1j) for earnings quality (Hribar et al., 2014). BIGN controls for audit firm size and Big 4 brand-name. Specific setting characteristics such as state or family ownership or affiliation to business group can influence the direction of the relation. A chief executive officer succession with gender change (Gul et al., 2021) and female directors on audit committees (Alkebsee et al., 2021) lower audit fees in China less pronouncedly or only in specific setting such as in state-owned enterprises and in firms located in more developed regions. Family involvement in management and the supervisory board are negatively related to audit fees (Schierstedt and Corten, 2021). Longer auditor tenure generally enhanced audit quality among Indian firms prior to mandatory rotations, but it changes the direction of the relation for abnormal high fees, particularly if the firm is affiliated with a business group or has chief executive officer duality (Jadiyappa et al., 2021). Thus, we add LABOR_COST specifically for our context.

Sample
Our sample selection starts with all Italian companies listed on the Milan Stock Exchange that are available on Compustat Global. We create a separate sample of audit hours for additional analysis. Our sample period is from 2005 to 2013. This period is chosen because few observations are available before 2005, and also because IFRS was adopted in 2005 (European Parliament, 2002), which allows a cleaner estimation of the audit fee and audit hour models. Table 1 details the selection process for the sample. The sample starts with 2,096 nonfinancial and non-telecommunication firm-year observations from Compustat Global. We hand-collect the names of the audit firms directly from the auditor's report in the consolidated annual report to ensure accurate identification. We delete 126 firm-year observations without an identified audit firm and other variables required to compute industry market share. We require a minimum of two audit firms per area-industry-year combination to identify an auditor industry market leader (Francis et al., 2005). The subtotal is 1,970 firm-year observations. Audit fee premium Next, we hand-collect audit fees and audit hours from publicly available annual shareholder meeting minutes of Italian-listed companies. Two of the authors (who are Faculty members at an Italian university) organized a team of seven research assistants to download the annual shareholder meeting minutes from the client firm's websites. From March to June, 2015, the team manually downloaded the minutes and extracted the audit fees and hours into a spreadsheet, which was checked for accuracy in July, 2015 by different research assistants. The faculty members met with the team every two weeks to plan and control the quantity and quality of the data. After deducting observations with missing audit fees, hours and control variable values, the final sample consists of 1,050 firm-year observations for audit fees and 1,123 firm-year observations for audit hours [7]. Based on the nine-year audit fee (audit hour) sample, there are 58 (62) clients per area per year. At the office level, among the approximately 20 cities in Italy, there are an average of 5.8 (6.2) clients per city per year. At the area level, there are an average of seven clients per area-industry-year combination. Compared to the area level, the office level has fewer industry-year combinations available to compute an auditor industry leader. Table 2 reports descriptive statistics for the Italy sample. The mean client fundamental variables (LAH, LAF, SIZE, LEV, ROA, LOSS and TENURE) are comparable to those used by Cameran et al. (2015), except that mean SIZE is lower in our sample because Cameran et al. (2015) excludes non-Big 4 clients. Other client fundamental variables (CATA, QUICK, s(CFO), SALES GROWTH and BIGN) are comparable to other related studies (Basioudis and Francis, 2007;Carson, 2009;Reichelt and Wang, 2010). The mean National Leader (35.2%) is comparable to Basioudis and Francis (2007) who report 33.6% in the UK. The mean Area Leader (39.7%) and mean City Leader (37.9%) are comparable to the same study's mean city leader variable (46.8%). Table 3 reports descriptive statistics with the number of observations at city level. It shows the small number of observations in several regions (Trentino, Alto Adige and Umbria), which in turn shows that the decision to use the two main areas instead of city level is robust. Tables 4 and 5 show the sample distribution by year and industry. The sample is unbalanced by year, but fairly stable over the period. Sample distribution by industry shows large industries such as industrial, consumer discretionary, IT and utilities. We checked that MAJ there were sufficient observations in each sample group to compute the industry specialization variable before calculating it. We exclude the financial sector from our analysis because of its dissimilar nature, and we exclude the telecommunications sector because there are too few observations to compute auditor industry specialization variables. At the area level, there are an average of seven clients per area-industry-year combination. Tables 6 and 7 show the correlation analysis for the Italian and German samples. The analysis appears to have no multicollinearity issues for either sample. The highest correlation between independent variables is between the total long-term debt scaled by total assets (LEV) and the ratio of current assets (less inventories) to current liabilities (QUICK) À0.572.  Audit fee premium

Main results
The results of estimating equation (1) are reported in Table 8 for testing the relation between national and area levels of auditor industry expertise and audit fees. Table 8, Panel A, reports that the national level of industry leadership has a 20.5% audit fee premium (over non-leaders) (p < 0.01), which is greater than the area-level audit fee premium of 11.4% (p < 0.10, one-tailed) [8]. The city level of industry leadership is not significantly different from zero, suggesting that city-level leadership is virtually non-existent. This result also suggests that the area level is a more appropriate unit of industry leadership than the city level. Panel B reports that the difference between the audit fee premium of the national industry leader and the area industry leader is statistically significant (p < 0.05, one-tailed), which supports H1. It is also economically significantnational industry leaders have a 10.7% greater fee premium for national industry leaders than area industry leaders over non-leaders. Audit fee premiums are comparable to prior studies (Francis et al., 2005), except that Italy's national level is greater. In short, national industry experts have a higher audit fee premium over area industry experts. We do not observe an audit fee premium for the city level of industry expertise, likely because each area is dominated by a major city -Milan in the North, and Rome in the Center-South. In a country where the audit regulatory environment is characterized by mandatory audit firm rotation versus other countries that have voluntary audit firm rotation (Australia, UK, USA and NZ), and a smaller audit market, we observe a more centralized organization of industry expertise at the national level.  Model (1) Model (2) Model (3) Estimate

Additional analysis
To strengthen our results, we add the following further qualitative and quantitative analysis: Big 4 partner interviews; audit hours analysis; audit industry leadership in Germany; auditor industry leadership based on client lagged total assets; audit fees for the audit of parent companies; complexity; hourly rate; distance between client and auditor office; endogeneity; both national and local leader; and auditor change.

Big 4 partners interviews.
We interviewed Big 4 audit partners with two main purposes: (1) to assess the organization of industry expertise in Italy to confirm our choice of local level of audit expertise; and (2) to find out about the effect of the mandatory audit firm rotation, compared with EU countries where rotation is not mandatory.
We interviewed a total of 11 partners from Big 4 audit firm offices in Italy. Four partners were from Milan (the main office of the Northern region), four were from Rome (the main office of the Center-South region) and three were from Parma (a smaller office in the Northern region  (2009). The interviews were tape-recorded, and the Italian-speaking authors prepared typed transcripts from the tape recordings. We provided the partners with the definition of an industry expert to avoid misinterpretation, and informed the participants that their responses would be kept confidential. After the interviews, the two Italian-speaking authors separately interpreted all typed transcripts and then compared their interpretations. The initial agreement level was 85%, indicating a high level of reliability (Beasley et al., 2009). Next, an English-speaking academic did the same, but from the English-translated transcripts. Any disagreements were resolved by relistening to the recordings and by recontacting the audit partner when necessary. For our first aim, we asked the following question: Q1. How does your audit firm organize industry expertise at national and local level?" Big 4 partner answers revealed that industry expertise is frequently transferred, and provided insight into another level of expertise, the area level. The area level complements MAJ the national level, allowing for the transfer of tacit and codified knowledge between offices, and the city level is nonexistent. For instance, one partner said, "We tend to have competencies in-house at the area level of the North (Milan) and the Center-South (Rome)." Offices obtain industry experts by transferring partners from other offices, most often within the same area and country. Industry experts are more frequently requested by energy and media and telecommunications industries than by utilities, consumer industrial, health care and pharmaceutical, construction, information technology, mechanical and retail. Partners transfer expertise to other offices by traveling to and working in multiple offices. Given the small size of the audit market in Italy, it is impossible organize the audit industry expertise at city level, with the exceptions of Rome (for the Center-South area) and Milan (for the Northern area). Rome and Milan are similar to the Metropolitan Statistical Area, described in the USA. For our second purpose, we asked the following question: Q2. How does MAFR affect industry expertise in Italy and compared to other countries?
All partners interviewed have experience working with overseas clients or in overseas offices (subsidiaries of Italian parent companies). They have attended two-to three-day industry-based meetings or annual and bi-annual International-European general meetings and courses in other countries to be up-to-date on the "hottest" topics. They also read audit firm industry studies and have performed quality reviews outside Italy. In Italy, industry expertise is similar to that of other continental European countries and at a lower level than in the US and the UK. One partner said, "In the U.S., expertise is greater. In Italy, the listed firms are in a smaller market, and are comparable with Germany and France." All partners in Rome and Milan believe that MAFR brings higher costs. The reasons for this include learning about a new client (time spent in meetings and reading documentation), learning about a new industry and its differences from the old one, as indicated in the following responses: "I also see a risk due to the learning effect in the first year and a half, mainly for a complex client": When I lose a client and acquire a new one, there is a moment in which I have to consider how not to lose the competence in that industry because maybe after nine years the client will come back, and I will also have to consider how to gain competence in the new industry.
However, a minority of Milan and Rome partners believe that MAFR is beneficial, because it encourages greater auditor independence, and its costs are sustainable. Interviews revealed that Italy's requirement to identify the name of the partner on the audit report incentivizes partner independence. As one partner said: The strongest incentive for independence in Italy is from the name on the signature of the audit opinion. The individual partner suffers this risk, and the incentive for independence is higher than the incentive that could come from [an alternative] regulation.
When an audit firm accepts a new client from a new industry, the firm may have to move staff with industry expertise from other offices or hire from other audit firms, implying greater costs and lower quality. The interviews and subsequent discussions confirm that Italian people typically have a close connection to a particular region and are less likely to relocate. Moreover, audit firms typically do not hire staff from other audit firms when they need industry expertise and tend to train existing office staff to meet the need.

Audit hours analysis.
Audit fees reflect audit hours expected by audit personnel (Cameran et al., 2015;Simunic, 1980;Azzali et al., 2021), so we expect that audit hour premiums should be higher for national industry experts than local industry experts. A Audit fee premium related paper by Bae et al. (2016) finds that audit hours are, in fact, higher for national industry experts.
We substitute audit hours for audit fees in equation (1) and estimate the equation. Table 9, Panel A, reports that audit hours are higher for national industry leaders (than nonleaders) by 18.3% (p < 0.01) compared to 10.1% (p < 0.10, one-tailed) by area industry leaders, and the difference is statistically significant (Panel B) and economically significant. National industry leaders have 9.5% greater hours than area industry leaders compared to non-leaders. These results corroborate our main results with audit fees, and suggest that national industry experts expend greater hours than area-level industry experts. These results imply that industry expert auditors use a differentiation strategy to distinguish higher quality in audit hours in a setting that is characterized by mandatory audit firm rotation, a smaller audit market and more centralized industry expertise. National industry experts appear to be stronger differentiators than area industry experts.

Auditor industry leadership in Germany.
For comparison purposes, we perform analysis of audit fees of a neighboring country -Germany (audit hours are not publicly available). We choose Germany because it is similar with all respects except that it does not have mandatory audit firm rotation. Its population (83 million) is similar to that of Italy (61 million). It did not have MAFR during the sample period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013) (Lennox, 2014). Untabulated results show that our sample of German listed firms has a maximum audit tenure of 20 years and a high frequency of firms audited by the same audit firms for 12 years. As a European Union member (European Parliament, 2002), it follows the same audit standards (IAASB) and accounting standards (IFRS). It has a code law requirement (Brown et al., 2014;La Porta et al., 1998). It has similar accounting and auditing quality. Brown et al. (2014) assign Germany and Italy a similar accounting enforcement rating (44 and 46, respectively), and a similar audit quality rating (23 and 27, respectively) [10]. The next most similar country, France, was not suitable for the comparison because it has a joint audit requirement, which affects industry expertise (Francis et al., 2009). We divide Germany into two areas, East and West, based on the 16 NUTS subdivisions.
Our sample includes all German companies that are available from the EUR business database and Compustat Global. As for the sample of companies from Italy, we exclude the financial and telecommunications sectors from our analysis. The sample of companies from Germany includes publicly listed companies over the sample period 2005-2013.
Table 10 details the selection process for the Germany sample. We use the EUR business database to obtain audit fees, and we integrate it with the Compustat Global database for financial statement and audit variables (tenure and opinion). The final sample has 2,910 firm-year observations, after deducting 1,028 observations with missing values necessary to compute auditor expertise, and 1,307 observations with missing audit fees and control variable data. There are an average of 161 clients per area per year. Among the 16 subdivisions in Germany, there are an average of 20.2 companies per subdivision per year. The Germany sample is larger (2,910) than the Italian sample (1,050) and has more observations per area. Table 11 reports descriptive statistics for the Germany sample. Many of the variables are comparable to Italy, with a few exceptions. The Germany sample has lower financial leverage, a smaller Big N audit market share and less frequent auditor industry leadership. Prior cross-country studies (Kwon et al., 2007;Lim and Tan, 2009) show a similar pattern for these exceptions. Audit firm tenure is greater for the Germany sample because Germany does not have mandatory audit firm rotation.
For the Germany sample, the results of estimating equation (1) for H1 are reported in Table 12. We compare whether audit fee premiums differ between the national and local levels MAJ Dependent variable is the natural logarithm of audit hours (N = 1,123) Model (1) Model (2) Model ( Audit fee premium for Germany which does not have mandatory audit firm rotation. Table 7 reports that the national industry leader has an audit fee premium of 14.5% (p < 0.01), and the area industry leader has an audit fee premium of 12.2% (p < 0.05), but these premiums are not statistically different (p = 0.373). In short, we conclude that the audit fee premium is no different between the national level and the local level. Given that Germany does not have MAFR, and Italy does, MAFR appears to contribute to a higher national level of industry expertise. It may also be the case that the larger size of Germany contributes to the difference. 4.2.4 Auditor industry leadership based on client lagged total assets. We repeat the analysis for the Italy sample by changing the industry leader definition using client lagged total assets (Francis et al., 2005, p. 130), for the purpose of computing auditor industry market share. Using lagged client assets mitigates any potential mechanical relation between market share and audit fee premium (Minutti-Meza, 2013) [11]. The coefficient (untabulated) on the national industry leader is greater than the area industry leader for both audit fees (20.1% vs 8.7%), and the difference is statistically significant for audit fees (p < 0.05) [12]. These results suggest that the main results are robust to an alternative measure of industry leadership.
4.2.5 Audit fees for the audit of parent companies. The main analysis uses audit fees reported in the minutes of the annual shareholder meeting. They include the review of the Model (1) Model (2) Model (3) Estimate Audit fee premium interim financial statements and accounting procedures, the statutory audit of the consolidated financial statements and the statutory audit of the group accounts. Most of the companies in our sample report a total amount for these three items. For robustness, we restrict our analysis to a subsample of firms that report separate audit fees and audit hours for the consolidated financial statement audit (n = 664) and a subsample of firms that separately report the statutory audit of the group accounts (n = 564). For the consolidated financial statement audit subsample, we find that there is an audit fee premium for the national industry leader (37.4%, p = 0.004), which is higher than the area industry leader (10.8%, p = 0.34), and the difference is statistically significant (p < 0.01) (untabulated). We find for the group accounts sub-sample that there is an audit fee premium for the national industry leader (26.9%, p = 0.01), which is higher than the area industry leader (9.4%, p = 0.344), and the difference is statistically significant (p = 0.02) (untabulated). In short, our results are robust to separating audit fees between the consolidated financial statement audit and the group accounts audit [13]. 4.2.6 Complexity. To rule out the possibility that complexity is an omitted variable, we add to equation (1) the number of business segments, the number of subsidiaries and foreign revenue. We hand-collected these variables for the period 2008-2011 for a subsample of 414 observations with available data, following the Cameran et al. (2015) model. We find that the coefficient on the national industry leader variable remains positive and statistically significant (0.260, p < 0.01), the coefficient on the area industry leader variable remains positive and statistically significant (0.145, p = 0.05, one-tailed) and the national leader has a higher fee premium than the area leader (p = 0.04) (untabulated) [14]. In short, our results are robust to controlling for complexity.
4.2.7 Hourly rate. Unlike Bae et al. (2016), we argue that higher audit hours by industry specialists are not driven by a lower hourly rate. We compute the fees per hours for a subsample of firms for which we have both audit fees and audit hours, and use the natural logarithm of the hourly rate as the dependent variable. The coefficients on the industry leader variables are positive and insignificant, confirming that higher audit hours by industry specialists are not driven by a lower hourly rate. 4.2.8 Distance between client and auditor office. Because the audit fee premium may be caused by factors such as the distance between client and auditor office which are omitted in the regression, we downloaded the auditor office location for the audit firms in our sample from their website. We merged the office location with the client location and computed the distance in kilometres from the nearest office. The greatest distance is Pievesestina di Cesena at 297 km from Milan for the non-Big 4 firm Baker, and for the Big 4, Rimini at 118 km from Bologna for Ernst &Young. However, 55.25% of the sample have an office in the same city as the client (Milan and Rome). The average distance in the sample is 18 km. Regression results (un-tabulated) show that distance between client and auditor office is not significant in explaining audit hours and fee premium, and the variable for industry leader remains significant, as in the main analysis including distance as control variable.
4.2.9 Endogeneity. Endogeneity is very difficult to address. Given the difficulty in finding a variable related to industry specialization which is not related to audit fees and audit hours, we use propensity-score matching for this. We used propensity-score matching models developed by Rosenbaum and Rubin (1983), which match observations on the basis of probability of undergoing a treatment, which in this case is the probability of being industry leader. We used logit models, the most frequently used approach (Guo and Fraser, 2010), and also used in previous accounting research (Lawrence et al., 2011;Lennox et al., 2012;DeFond et al., 2016 for a description of matching models). This approached involved estimating a logit regression to predict the probability of being industry leader. Propensity scores were matched for all the independent variables, using the propensity scores that are predicted from the logit regression from the previous step to obtain a propensity-score matched sample with neighbor (1) caliper (0.1). Observations which could not be paired were eliminated from the sample and the regressions were run on subsamples. Results (Table 13, Panels A and B) show a significant reduction of mean bias from 23-26 to 3-5. Our initial multivariate results hold (Table 13, Panels A and B).
4.2.10 Both national and area leaders. We follow prior literature (Francis et al., 2005;Basioudis and Francis, 2007;Carson, 2009), to ascertain whether the auditors are a nationally top-ranked auditor and city-level industry leader at the same time. Our sample includes auditors which were top-ranked for both nation and area (about 31% observations) and for both nation and city (about 24%). Results (un-tabulated) show that the coefficient of national and area level and the coefficient of national and city level are both positive and statistically significant. We conclude that a necessary and sufficient condition to gain an audit fee premium is audit partner leadership at national and area level.
4.2.11 Auditor change. Italy has a stronger national than local level of industry expertise because auditors aim to retain and transfer industry expertise. We control for auditor change in the regression using the variable Tenure. The variable Industry specialization is also time-variant, and changes when the auditors change if the auditors lose the largest market share. In addition, we perform a robustness analysis (untabulated) removing Tenure and adding First year of auditor change. Auditor change is negatively and significantly associated with audit fees but not significantly associated with Audit hours. Using Auditor change and controlling for the difference in the First year of the audit engagement, rather than using Tenure, our results for industry leader are confirmed.

Conclusion
Our paper examines how auditor national industry expertise and local industry expertise is organized in Italy, where mandatory audit firm rotation has been in force since 1975. Industry expertise is a conduit for the transfer of auditors' industry specific knowledge. Italy has higher client turnover than the USA, because of a nine-year MAFR requirement, and it also has a smaller publicly listed client audit market. We predict that national industry expertise is stronger than area industry expertise because higher client turnover, and a small audit market, increases the sensitivity of industry-specific knowledge being lost at the local area level. We find that audit fees of industry specialists at the national level and the area level are greater than fees of nonspecialists. We do not find the same evidence for city industry specialists. However, unlike prior studies restricted to countries with voluntary audit firm rotation (Ferguson et al., 2003;Francis et al., 2005;Basioudis and Francis, 2007), we find that national industry specialization dominates over area industry specialization in audit fee premiums and audit hours, suggesting there is greater centralization of industry specialization in Italy. We conduct the same analysis for companies located in Germany, which has a similar legal and regulatory system, but does not have MAFR. We find no difference between the audit fee premiums of national industry specialists and area industry specialists. However, the sample size for Germany is larger. We conclude that the higher client rotation of MAFR is a contributing factor, as well as a smaller audit market size. Our study has a number of implications for academics, regulators and legislators considering adopting MAFR.

MAJ
Theoretical implications involve both method of research and the debate on costs and benefits of national and local industry specialization of audit partners. Previous research based on audit fees displays limitations and conflicting results, but the present research suggests that the use of audit hours strengthens results and that audit efforts can be used as a measure of audit quality. Our results also show that the distribution of audit expertise depends partly on the size of the audit market and MAFR. Where the audit market is small and MAFR is in place, it is likely that audit firm industry expertise will be mainly organized at national and area level, rather than city level. The organization of audit expertise at national and area level assures a higher level of audit quality in these contexts. MAFR, which is now in force in all countries of the European Union, together with the mandatory rotation of audit partners, was introduced to assure auditor independence. It is, however, costly because it lowers the levels of competency and industry specialization of audit partners. MAFR thus appears to weaken the city level of industry expertise, which can affect audit quality as a whole. More frequent turnover of clients requires industry expert audit personnel to either adopt another industry, or else relocate, when a significant number of clients depart from an industry and are not replaced. This can create a "brain drain" of client and industry-specific knowledge, and shift a greater burden onto national headquarters to nationalize industry expertise, such as through national industry-specific training programs and consulting industry experts. Our results show that audit fee premiums and audit hour premiums are greater for national industry experts than for area industry experts, and the differences are statistically and economically significant. A potential consequence is that with more frequent turnover of clients, auditor quality is weakened at the office level where engagement partners and staff have a more intimate knowledge of the client. However, their industry knowledge has a shorter useful life and can depreciate quickly with the loss of expert personnel to other offices or other industries. Combined national and local levels of audit quality could be lower than in countries without MAFR, if national industry expertise cannot sufficiently compensate for the loss of local level expertise. This study has certain limitations, which may be overcome by future research. First, the introduction of the "area level" as the local level of industry expertise could be questioned. Following the Big 4 partner interviews, we here divide Italy in two main areas, North and Center-South, but the literature suggests the city could be more suitable as the local level of organization of industry expertise. The coefficient of city level in our regression model is not significant, confirming the dominance of national level of expertise. Comparing our results with Germany, we investigated the area level in that country, but future research is required to confirm the effectiveness of the area level of organization of audit industry expertise in other countries. Second, we interview Big 4 partners to corroborate the effects of area level and the effects of MAFR on industry expertise, but our sample may not be completely representative of the Italian population. A different approach including more extensive interviews and/or surveys would provide better insights on this. Future research will also need to use both quantitative and qualitative methods of research. Third, given the controversial results of audit fees as a measure of audit effort and audit quality, we complement our model with audit hours. But a promising next step for future research might be to investigate how audit industry expertise, and specifically the national and local level of industry expertise, influences audit quality using different measures, for example, the level of discretionary accruals and/or the going concern opinions. Fourth, we compute audit fees without taking into account possible omitted variables such as political donations.
Audit fee premium 6. Given that we have a small sample size compared to previous studies, we are careful to select control variables to avoid problems related to the degree of freedoms from adding too many independent variables. We do not include client firm age because the Italian listed firms are on average very old (some are hundreds of years). We do not include years of listing because Italian firms often use loan financing before the firm is listed on a stock exchange. We do not include the fiscal year-end because a very large proportion of our sample (95%) has a December 31 year-end. We do not include the Herfindahl-Hirschman index of auditor area industry concentration (Numan and Willekens, 2012) because the coefficient was not significant (p > 0.10), and there is too little variation over time to be reliable for fixed-effects estimation. We do not include the number of business segments because Compustat does not report it, and we use size to control for complexity. In our robustness tests, we use a hand-collected business segments variable from a smaller sample, and our main results still hold.
7. Some companies disclose only audit hours and no audit fees.
8. Following Craswell, Francis, and Taylor (1995), the audit fee premium in percentage points is (e z À 1) Â 100, where z is the coefficient on the industry leader variable.
9. The interviews include several questions, but for this study, we use only answers to the two questions quoted below.
10. Brown et al. (2014) survey 51 countries, and report for 2008 that the audit quality index has a mean of 18.25 and a standard deviation of 8.27, and the accounting enforcement index as a mean of 12.59 and a standard deviation of 6.65.
11. We repeat the analysis computing industry leadership of the national industry leader and the area industry leader by market share of audit fees and by market share of audit hours. We find that our results are not consistent when compared to using market share of client assets or client sales. We conclude that client sales and client assets are more reliable, likely because there are more observations to accurately compute the auditor's industry market share.
13. For audit hours, the results are robust. There is a premium for the national industry leader for the consolidated financial statements audit subsample (33.1%, p<0.01) which is greater than the area MAJ industry leader (8.2%, p = 0.43), and the difference is statistically significant (p < 0.01). There is also an audit hour premium for the group accounts subsample for the national industry leader (24.8%, p = 0.03), which is greater than the area industry leader (7.3%, p = 0.49), and the difference is statistically significant (p = 0.03) (untabulated).
14. For audit hours, we find that the coefficient on the national industry leader variable is positive and statistically significant (0.202, p=0.04), and the coefficient on the area industry leader variable is positive (0.139, p=0.06, one-tailed) even though the difference is less evident (p=0.28) (untabulated). The less significant results are likely because of the reduced sample size from 1,123 to 414 firm-year observations.