Strategic pricing by Big 4 audit firms in private client segments
Abstract
We examine how Big 4 auditors compete for new private clients. We find evidence suggesting that Big 4 auditors offer fee discounts to attract non-Big 4 private clients to experience attributes of their brand name audit services. We also find that to attract clients from competing Big 4 suppliers, Big 4 auditors target fee discounts at clients in industries where they are the market leader. Our results further indicate that the Big 4 industry leaders target fee discounts to fast-growing clients and are able to charge these clients significant price fee increases in the second mandate period (after 3 years).
1. Introduction
Numerous studies investigate determinants of audit pricing and document fee discounting on initial engagements in public client market segments (see Hay et al., 2006). Other studies examine auditor strategy (e.g. Mayhew and Wilkins, 2003; Casterella et al., 2004) and auditing in private company settings (e.g. Hope et al., 2012). However, we are unaware of studies that examine how Big 4 auditors strategically compete for new clients in private client market segments. The market for private audit clients is less concentrated than the public client market as the Big 4 auditors compete not only with other Big 4 auditors but also to a large extent with non-Big 4 audit suppliers. In this study, we examine pricing strategies of the Big 4 audit firms to gain market share in (more) competitive private client market segments.
There is evidence that Big 4 auditors compete on price through offering fee discounts to new clients. DeAngelo (1981) and Chan (1999) argue that Big 4 auditors will offer fee discounts on initial engagements to become the incumbent. The US public company evidence is consistent with fee cutting on initial engagements (e.g. Francis and Simon, 1987; Ettredge and Greenberg, 1990; Turpen, 1990; Sankaraguruswamy and Whisenant, 2005). Basioudis and Francis (2007) also report fee discounts on initial engagements in the UK. However, a study by Ghosh and Lustgarten (2006) indicates that there is more intense fee cutting on initial engagements by non-Big 4 auditors than by Big 4 auditors in the United States.1
Big 4 auditors not only compete through price but also on product differentiation. More specifically, Big 4 firms differentiate themselves from non-Big 4 firms through their investments in their brand name and their reputation as high-quality auditors.2 In addition, there is also differentiation within the group of Big 4 auditors through industry specialization. These firms invest in industry-specific audit methodologies, additional staff training, industry knowledge and the like to differentiate themselves from the non-specialists.3 Auditors who differentiate themselves from their competitors through industry specialization have been shown to charge fee premiums in listed client segments of the market (e.g. Craswell et al., 1995; Ferguson et al., 2003, 2006; Francis et al., 2005; Numan and Willekens, 2012). Although the evidence is more limited, Big N premiums and industry specialist premiums have also been reported in private client studies (Willekens and Gaeremynck, 2005; Dutillieux and Willekens, 2009).
In this study, we combine these two strategies of competing on price and differentiation and investigate how the Big 4 compete for new private clients. Drawing on experience good theory (Nelson, 1970; Shapiro, 1983; Craswell and Francis, 1999; Ferguson et al., 2006) and utilizing proprietary data from the Belgian Institute of Registered Auditors (IBR/IRE) for all statutory audit engagements for the year 2004, we first examine whether Big 4 auditors offer fee discounts on initial private client engagements. We argue that private companies that are considering an auditor upgrade and hence are potential first-time buyers of Big 4 and/or industry specialist audit services have imperfect information about such an audit service's superior value. Therefore, Big 4 auditors offer fee discounts to switchers that upgrade to allow the client to experience the value of such an upgrade at lower cost. We find that this is the case for new engagements that were previously audited by non-Big 4 auditors (upward switchers), but find no fee discounts when the client switches from another Big 4 competitor (lateral switchers). We build further on experience good theory by arguing that to attract new clients from competing Big 4 auditors (lateral switchers), the Big 4 target their fee discounting efforts at those industries where they are already specialists. If private clients learn to value the services of an industry specialist gained through an ‘experience discount’, they are more likely to be willing to pay a fee premium after the initial mandate has been completed. Furthermore, industry specialist premiums earned on continuing engagements could also allow for a larger fee discount, which is expected to make the auditor more successful at attracting new clients.4 Consistent with this theory, our results show that discounting on initial engagements also applies to a switch from a Big 4 auditor (non-industry specialist) to a Big 4 industry specialist auditor.
Next, we examine whether the Big 4 pursue a pricing strategy to further target particular (valuable) clients into their portfolios. We conjecture that they also seek to attract clients expected to rapidly grow and who are therefore more likely to value their services once they experience a Big 4 and/or industry specialist audit. In particular, we investigate whether Big 4 industry specialist auditors grant fee discounts to attract these fast-growing clients. We find that a fee discount is offered by Big 4 industry specialists to initial clients who are expected to experience rapid growth to attract them to switch from their non-Big 4 auditors to a specialist Big 4 auditor.
Finally, in a supplemental analysis, we look at what happens in the second fee mandate period for these clients.5 These fee discounts are targeted at the clients expected to be faster growing and are provided on the basis that audit fees in future periods will increase. Our results are consistent with the theory as they indeed show that industry specialist Big 4 auditors earn fee increases in the second mandate period from those clients that were expected to grow faster.
The remainder of this article is organized as follows. In section 2., we develop our hypotheses based on experience good theory. Next, in section 3., we provide some background on the Belgian private client market. Section 4. describes the research design. In section 5., the sample, data, main regression results, supplemental analyses and robustness checks are discussed. Finally, section 6. provides conclusions, limitations and contributions.
2. Theory and hypotheses
In this section we develop our hypotheses about how Big 4 auditors compete for new clients in the private client segment of the audit market. As indicated above, future clients of Big 4 auditors can be classified as either potential upward switchers (previous non-Big 4 clients) or lateral switchers (previous Big 4 clients).
2.1. Experience good theory, audit upgrades and audit fee discounts
Although rational consumers need information on price and quality prior to the purchase of a good or service, there can be information asymmetry between buyers and sellers about certain attributes of the good or service. Unlike the ‘search’ attributes of goods or services which can be readily assessed and/or observed prior to purchase, ‘experience’ attributes are not readily determinable as they are rather bundled with the good itself and learned through experience (Nelson, 1970; Shapiro, 1983).6 It is generally recognized that audit services are not tangible as they involve a complex process (Causholli and Knechel, 2012).7 Although an auditor's brand name can be observed, the actual level of audit assurance supplied is unknown. In addition, the actual value/usefulness of certain audit(or) characteristics in a specific client setting is ex ante also not certain. In this study, we build on the premise that audit services have certain experience attributes and that therefore quality-differentiated audit services, such as Big 4 audits and industry specialist audits, are ‘experience goods’. We further argue that the potential value of quality-differentiated audits is even harder to assess ex ante for private audit clients.8 Although the Big 4 have an international reputation, this reputation mainly relates to publicly listed clients and ex ante it is not obvious that what is valuable in a public client setting is also valuable in private client settings. This is also related to the fact that agency and information asymmetry problems are different for public and private companies (see for example, Chaney et al., 2004; Hope et al., 2012).9 Note further that, the empirical evidence on auditor quality differences, measured through, for example earnings quality or audit opinion proxies, mainly relates to samples of publicly listed clients. These studies document consistently higher Big 4 or industry specialist quality effects (Francis, 2004). In contrast, the evidence on Big 4 or industry specialist quality effects from private client studies is both scarcer and mixed (Eilifsen and Willekens, 2008).
Overall, we argue that private companies that are considering an auditor upgrade and hence are potential first-time buyers of Big 4 and/or industry specialist audit services have imperfect information about such an audit service's superior value. Due to this imperfect information, the true (but ex ante unknown) value is likely to exceed the perceived value. Hence, Big 4 and industry specialist auditors need to do reputation building in the private client market segment to convince these clients of the value of their acclaimed brand name services. Shapiro (1983) analytically shows that discount pricing will occur on initial purchases of experience goods when consumers underestimate ex ante the value of a good. In such a situation, the optimal way to build a reputation for a good is to use a low introductory price followed by a higher regular price in subsequent periods as soon as the consumer has experienced the higher (than initially perceived) value of the good. Applied to the private client audit market setting, brand name auditors will have to do reputation building by offering discounts on initial engagements of clients that are ‘experiencing’ a brand name upgrade.10
This reasoning leads to the following two related hypotheses:
- Hypothesis 1A: Relative to continuing engagements, Big 4 auditors are more likely to grant an audit fee discount to private clients switching from a non-Big 4 supplier than to private clients switching from a Big 4 supplier.
- Hypothesis 1B: Relative to continuing engagements, Big 4 industry specialist auditors are more likely to grant an audit fee discount to private clients switching from a Big 4 non-industry specialist supplier than to private clients switching from a Big 4 industry specialist supplier.
The focus of hypotheses 1A and 1B is on supply-side arguments of strategic audit fee discounting. Note that various demand factors that increase the potential value of switching to a higher brand name auditor could of course also be relevant in an auditor switching setting. In the next subsection, we will focus on such demand-side arguments.
2.2. Audit client targeting and initial price discounts in the private market
Shapiro (1983) also shows that an important element for discounting to be part of an optimal pricing regime for experience goods is that the price in the first period is followed by a higher price in subsequent periods. However, only clients that are economically strong enough and to whom the audit quality upgrade will eventually be valuable will be able and willing to pay for a higher audit fee in subsequent periods. As clients could appoint a Big 4 auditor because of the fee discount without intending to retain the auditor after the initial mandate period (i.e. when the fee is raised), Big 4 auditors are likely to be selective about identifying those able and willing to pay higher audit fees in subsequent periods.11
We therefore argue, consistent with H1A and H1B, that Big 4 auditors (and those Big 4 that are industry specialists) focus on those non-Big 4 clients (and industries for other Big 4 clients) where they are valued (and have most experience and which allowed them to be recognized as industry specialists). Furthermore, the strategy of focusing on the non-Big 4 clients (industries of other Big 4 clients where they are specialists) also facilitates fee discounting.12 We further argue that fee discounts are targeted at certain groups of clients who will benefit from learning to value the services of the Big 4 (Big 4 industry specialists) and therefore be willing and able to pay a premium in the future. As profit maximizers, Big 4 (and Big 4 industry specialist) auditors will target a selection of those clients with the highest likelihood of becoming a profitable engagement. Prior studies examining demand- and supply-side alignments of auditor and clients consistently find that expected growth is significantly positively associated with switches from smaller audit firms to larger audit firms (e.g. Palmrose, 1984; Healy and Lys, 1986; Francis and Wilson, 1988; Eichenseher and Shields, 1989; Johnson and Lys, 1990).13 We argue that expected growth proxies for client firms' future demand for the services that can be supplied by larger (and more specialized) audit firms. Such clients are more likely to benefit from ‘switching up’ to a Big 4 (‘switching across’ to a Big 4 industry specialist) auditor because growth in their organization increases demand for the services of a Big 4 (Big 4 industry-specific knowledge) in which the Big 4 audit firms have invested. We expect that clients who are expected to grow faster will be more likely to retain their auditor after the initial mandate period.
As a result, we hypothesize that Big 4 (and Big 4 industry specialists) will target non-Big 4 clients (and other Big 4 clients) with expected rapid growth materializing in greater demand for the services by offering fee discounts. Hence, these clients will receive fee discounts in ‘switching up’ to a Big 4 (and ‘switching across’ to industry specialist Big 4 auditors), beyond any Big 4 discount on initial engagements that could be offered and any industry specialist premium for services currently in demand. This suggests the following hypotheses:
- Hypothesis 2A: Big 4 auditors are more likely to grant fee discounts to previous non-Big 4 private clients that are expected to be faster growing.
- Hypothesis 2B: Big 4 industry specialists are more likely to grant fee discounts to previous Big 4 private clients that are expected to be faster growing.
3. The Belgian private client market
Our setting for this study is the Belgian audit market. This setting is chosen for four reasons. First, size thresholds rather than listing status determine whether an audit is mandatory in Belgium.14 The audit market therefore primarily consists of private clients with a small number of listed clients, and there is extensive variation in their size and complexity. It is likely that many small Belgian private firms appoint an auditor to comply with the legal requirement but would not do so otherwise (Willekens and Achmadi, 2003), and thus, price is likely the most important criterion when they select an auditor to comply with the mandatory audit requirement. As a result, the demand for the (more expensive) services of a Big 4 auditor is lower than it would be in public client settings. Consequently, competition for the clients with a (potential) demand for a Big 4 auditor is expected to be more intense.
Second, although all EU countries apply similar size thresholds for mandatory audits, Belgium is of particular interest because of its relatively low market concentration. This is evidenced by the high average number of auditors in each industry and the low Herfindahl concentration indices suggesting that the Belgian audit market is more competitive than the US public client market.15,16 Furthermore, the Big 4 firms account for just above 50 per cent of all audit engagements although their market shares in terms of audit fees and client total assets are higher. Previous studies have also indicated that the Belgian audit market is less concentrated than other audit markets in Europe (e.g. Willekens and Achmadi, 2003).
A third reason for utilizing the Belgian setting is that auditors are appointed by the general assembly of shareholders for a period of three years and dismissal or resignation during this mandate is severely restricted.17 The audit fees are determined at the beginning of each three-year mandate. Therefore, potential fee discounts in the first year of the mandate cannot be compensated in the second or third year because the total fee is already known at the start of the mandate period.18 Solicitation is prohibited, and advertising is subject to some restrictions. We assume that these provisions increase switching costs for clients because it is up to the client firm to contact audit firms and ask for a bid. When solicitation is allowed, clients can more easily compare offers and select the best bid.
Finally, an additional advantage of our setting is that we can use proprietary data from the Belgian Institute of Registered Auditors (IBR/IRE). This enables examining competition across all statutory audit engagements for the year 2004.
4. Research design
This section describes the audit fee models that are used to test our hypotheses about Big 4 initial engagement pricing in private client market segments. These models contain an indicator variable for initial engagements (Initial) and a proxy for industry specialization (IS) that is in line with prior studies.19 Most prior studies use two-digit SIC codes to define industries, but Gramling and Stone (2001) discuss a number of weaknesses of this approach. In addition to that, Casterella et al. (2004) mention that audit firms have structured their activities around broad industry sectors and they illustrate this point with the way that KPMG promotes its industry expertise. Therefore, we believe it is more appropriate to use broader industry groups instead of two-digit SIC codes. In particular, we will use eleven of the twelve industry groups as defined by Campbell (1996). More information about these groups can be found in Appendix I.
We determine whether an engagement is continuing or initial by comparing the current audit firm with that of the previous mandate period. To test H2A and H2B, we need to identify firms that are expected to grow faster and define a variable SGP for that purpose. We compare each firm's sales growth rate (over the period 2004–2007) with the median values in their industry. We use a forward-looking variable as we have argued that Big 4 auditors will focus on clients who are likely to increase their demand for their services in the near future (i.e. after the initial mandate). For firms with an expected sales growth that exceeds (is lower than) the median growth in their industry and hence are identified as fast-growing (slow-growing) clients, SGP is set equal to one (zero). Note that we also calculate the Hirschman–Herfindahl index (HHI) at the industry-decile level as a control for the effects of market concentration on initial pricing.20 The following models are then tested.
(1)
(2)Note that the above two models will each be tested on three (sub) samples: (i) the total sample of all switching firms and continuing engagements (reported below as Model 1 in Table 4 and Model 2 Table 5, see infra); (ii) a subsample of upward switching firms (i.e. switching from a non-Big 4 to a Big 4 firm) and continuing engagements (reported as Model 1A in Table 4 and Model 2A in Table 5, see infra); and (iii) a subsample of lateral switching firms (from a Big 4 to another Big 4) and continuing engagements (reported as Model 1B in Table 4 and Model 2B in Table 5, see infra).
To test H1A, our experimental variable of interest is the Initial variable in Model 1A capturing the effect of a predicted fee discount for an upward switch from a non-Big 4 supplier to a Big 4 supplier. To test H1B, the variable of interest is the interaction term InitialxIS in Model 1B capturing the effect of a predicted fee discount for a lateral switch from a Big 4 non-industry specialist supplier to a Big 4 industry specialist supplier.
For hypotheses 2A and 2B, our experimental variables of interest are the interaction variable InitialxSGP capturing the effect of a predicted additional fee discount for previous non-Big 4 private clients that are expected to be faster growing (H2A), and the interaction variable InitialxISxSGP capturing the effect of a predicted additional fee discount for previous non-industry specialist Big 4 private clients that are expected to be faster growing (H2B).
Our audit fee models follow Simunic (1980) and regress the natural logarithm of audit fees (Ln(Audit Fee)) on proxies for client size, client risk and client complexity. Our choice of control variables is based on the meta-analysis of Hay et al. (2006) and on a large-scale Belgian fee-study by Willekens and Gaeremynck (2005). The natural logarithm of total assets (Ln(Total Assets)), of sales (Ln(Sales)) and of the square of the natural logarithm of sales and whether the audit mission included reporting to the works council (Works Council) are included to proxy for size.21 We include both assets and sales in our model because previous studies on fee-setting in Belgium (Willekens and Gaeremynck, 2005; Dutillieux and Willekens, 2009) suggest that they capture different aspects that are relevant for fee-setting in the private client segment. Consistent with Dutillieux and Willekens (2009), we also include the square of the natural logarithm of sales as a control variable to capture non-linear effects of client size on audit fees.22 We proxy for complexity using a dummy variable that indicates whether the financial statements were consolidated (Consolidation) and the natural logarithm of the number of subsidiaries of the audited firm (Ln(#Subsidiaries)). The ratio of receivables and inventories to total assets proxies for the inherent risk of the audit ((Rec+Inv)/TA). Client profitability is measured using the gross sales margin (GSM) and with a loss dummy (Loss). The ratio of total equity to total assets (EQ/TA) and the acid test ratio (Acid Ratio) are added to proxy for leverage and for liquidity, respectively. Because it is possible that the Big 4 have different pricing strategies, we include auditor fixed effects in the models.23 Furthermore, our models also include industry fixed effects and the standard errors are clustered per audit firm to correct for unobserved within-auditor correlation.
5. Data and results
5.1. Sample and data
As noted, the fee data used in this study are proprietary and were made available by the Belgian Institute of Registered Auditors (IBR/IRE). As part of a quality monitoring programme, the Institute requires that all registered auditors file a yearly statement that contains audit fees from all their engagements. These data enable us to calculate market shares based on all audit fees for the year 2004. We complement the auditor information with financial statement information from the BELFIRST database.24 These additional data are necessary to estimate the fee models. Our sample consists of all audits for listed and private companies in which a statutory auditor is appointed.25 Public institutions (such as hospitals and social security organizations) and not-for-profit organizations are excluded from the sample.
The starting sample contains fees for 16,321 audit engagements. This can be considered as an almost complete representation of the Belgian audit market for statutory audit engagements in 2004. We remove the petroleum industry group because it consists of only 29 engagements and each size decile contains only a few observations. This yields extremely high concentration ratios, and therefore, we consider this industry as an outlier. We also exclude 1,218 engagements from two-digit SIC industries that cannot be designated to one of the twelve industry groups defined by Campbell (1996). We argue that it is better to exclude these engagements rather than forming a new industry group because it concerns very different activities.26 This yields a sample of 15,074 observations to calculate market shares based on audit fees for the remaining eleven industry groups.
Before testing the audit fee models, we again need to remove a number of observations. First, 85 listed firms are dropped from the sample because we focus on competition in the private client segment. Second, we remove 7,302 non-Big 4 engagements because the focus of this article is on which strategies the Big 4 use to penetrate a highly competitive market. Third, we lose 1,690 observations for which financial statement data for the year 2004 are missing. Fourth, we use auditor data from the years 2001 and 2003 to determine which engagements are initial and which are continuing. Because these data are incomplete, we lose 1,298 observations for which we cannot identify the previous auditor. We consider an engagement as initial when the auditor is serving their first three-year mandate with the client because prices are fixed at the start of each mandate period. This leaves a final sample of 4,699 observations for our tests of hypotheses 1A and 1B. About 26.1 per cent of this sample (i.e. 1,230 observations) consists of initial engagements. The majority (i.e. 68.8 per cent) of these new engagements consist of lateral switches (i.e. from one Big 4 to another Big 4). For our test of hypotheses 2A and 2B, we require sales data for the year 2007 (to calculate our growth proxy). Because of missing data, our sample for these analyses is reduced to 3,717 observations. Table 1 provides an overview of the sample selection procedure.
| Starting sample with available fee data | 16,321 |
| Less firms from the petroleum industry | 29 |
| Less firms not designated to an industry group | 1,218 |
| Sample for determination of market shares | 15,074 |
| Less listed firms | 85 |
| Less non-Big 4 engagements | 7,302 |
| Less firms with missing financial statement data | 1,690 |
| Less clients for which previous auditor is unknown | 1,298 |
| Sample for test of hypothesis 1 | 4,699 |
| Of which continuing engagements | 3,469 |
| Of which upward switchers | 383 |
| Of which lateral switchers | 847 |
| Less firms with missing sales growth data | 982 |
| Sample for test of hypothesis 2 | 3,717 |
| Of which continuing engagements | 2,775 |
| Of which upward switchers | 294 |
| Of which lateral switchers | 648 |
| Less firms with data for 2007 | 156 |
| Less firms that switched between 2004 and 2007 | 860 |
| Sample for test of pricing in the next mandate period | 2,701 |
| Of which continuing engagements | 2,002 |
| Of which upward switchers | 214 |
| Of which lateral switchers | 485 |
5.2. Descriptive statistics
In panel A of Table 2, we report descriptive statistics on all client-level variables that are included in our models for both the sample of all engagements and for the three subsamples (i.e. continuing engagements, lateral switchers and upward switchers). The average audit fee for clients is 18,634 Euro (median of 10,500 Euro). The average client has 130 million Euro worth of total assets (median of 10.2 million). About 24 per cent of the sample firms reported a loss in 2004. In the median company, equity accounts for almost 30 per cent of total assets, while 57 per cent of the balance sheet consists of receivables and inventories. The auditor has to report to the works council in about 39 per cent of the engagements, while only 7.7 per cent of the financial statements are consolidated. The median company has a sales growth of 16 per cent over the period 2004–2007. The average Herfindahl index is 0.143, which suggests moderate concentration of the audit market at the industry-decile level.27 Panel B of Table 2 provides the industry composition of the full sample and each of the three subsamples. The largest three industries for the full sample are the consumer durables, textiles and trade, and the finance and real estate industry.
| Mean | Standard deviation | Minimum | Q1 | Median | Q3 | Maximum | |
|---|---|---|---|---|---|---|---|
| Panel A | |||||||
| All engagementsa | |||||||
| Ln(Audit Fee) | 9.221 | 1.084 | 6.346 | 8.517 | 9.259 | 9.967 | 11.333 |
| Audit fee (in Euro) | 18,634 | 28,234 | 138 | 5,000 | 10,500 | 21,320 | 510,000 |
| IS | 0.290 | 0.454 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Initial | 0.262 | 0.440 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Sales Growth Post | 0.339 | 1.105 | −0.945 | −0.039 | 0.160 | 0.402 | 8.275 |
| SGP | 0.497 | 0.500 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Ln(Total Assets) | 9.368 | 1.913 | 3.638 | 8.119 | 9.230 | 10.512 | 14.066 |
| Total Assets (in 000 Euro) | 130,199 | 728,847 | 3 | 3,356 | 10,198 | 36,747 | 18,874,757 |
| Equity/TA | 0.251 | 0.702 | −6.363 | 0.126 | 0.298 | 0.535 | 0.999 |
| Loss | 0.244 | 0.429 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| (Rec+Inv)/TA | 0.542 | 0.287 | 0.000 | 0.308 | 0.573 | 0.788 | 1.000 |
| GSM | 0.077 | 0.321 | −2.310 | 0.020 | 0.068 | 0.144 | 0.898 |
| Acid Ratio | 3.103 | 16.183 | 0.005 | 0.640 | 1.052 | 1.686 | 271.100 |
| Sales (in 000 Euro) | 74,517 | 356,233 | 1 | 3,551 | 12,289 | 37,759 | 14,803,099 |
| Consolidation | 0.077 | 0.267 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| Works Council | 0.388 | 0.487 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| HHI | 0.143 | 0.059 | 0.043 | 0.100 | 0.124 | 0.179 | 0.342 |
| Ln(#Subsidiaries) | 0.730 | 1.095 | 0.000 | 0.000 | 0.000 | 1.000 | 6.313 |
| Audit fee change 2004–2007 | 0.604 | 1.306 | −0.811 | 0.000 | 0.133 | 0.598 | 4.997 |
| Continuing engagementsb | |||||||
| Ln(Audit Fee) | 9.207 | 1.084 | 6.346 | 8.509 | 9.253 | 9.956 | 11.333 |
| Audit fee (in Euro) | 18,257 | 26,383 | 249 | 4,958 | 10,432 | 21,080 | 483,000 |
| IS | 0.271 | 0.445 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Initial | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Sales Growth Post | 0.362 | 1.168 | −0.945 | −0.039 | 0.156 | 0.415 | 8.275 |
| SGP | 0.499 | 0.500 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Ln(Total Assets) | 9.290 | 1.875 | 3.638 | 8.075 | 9.149 | 10.436 | 14.066 |
| Total Assets (in 000 Euro) | 107,809 | 603,401 | 3 | 3,213 | 9,409 | 34,063 | 18,874,757 |
| Equity/TA | 0.263 | 0.671 | −6.363 | 0.130 | 0.298 | 0.537 | 0.999 |
| Loss | 0.235 | 0.424 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| (Rec+Inv)/TA | 0.546 | 0.287 | 0.000 | 0.317 | 0.578 | 0.792 | 1.000 |
| GSM | 0.080 | 0.317 | −2.310 | 0.020 | 0.067 | 0.143 | 0.898 |
| Acid Ratio | 3.339 | 17.217 | 0.005 | 0.642 | 1.069 | 1.734 | 271.100 |
| Sales (in 000 Euro) | 73,311 | 380,309 | 1 | 3,379 | 11,238 | 36,613 | 14,803,099 |
| Consolidation | 0.070 | 0.256 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| Works Council | 0.389 | 0.488 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| HHI | 0.141 | 0.058 | 0.043 | 0.100 | 0.124 | 0.176 | 0.342 |
| Ln(#Subsidiaries) | 0.700 | 1.055 | 0.000 | 0.000 | 0.000 | 1.000 | 6.313 |
| Audit fee change 2004–2007 | 0.597 | 1.309 | −0.811 | 0.000 | 0.128 | 0.579 | 4.997 |
| Lateral switchersc | |||||||
| Ln(Audit Fee) | 9.353 | 1.091 | 6.346 | 8.613 | 9.393 | 10.117 | 11.333 |
| Audit fee (in Euro) | 21,859 | 37,152 | 259 | 5,500 | 12,000 | 24,750 | 510,000 |
| IS | 0.362 | 0.481 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Initial | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |
| Sales Growth Post | 0.308 | 0.987 | −0.945 | −0.030 | 0.164 | 0.395 | 8.275 |
| SGP | 0.492 | 0.500 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Ln(Total Assets) | 9.618 | 2.049 | 3.638 | 8.177 | 9.424 | 10.859 | 14.066 |
| Total Assets (in 000 Euro) | 228,522 | 1,150,761 | 16 | 3,559 | 12,379 | 52,001 | 12,863,812 |
| Equity/TA | 0.210 | 0.768 | −6.363 | 0.104 | 0.270 | 0.522 | 0.997 |
| Loss | 0.274 | 0.446 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| (Rec+Inv)/TA | 0.548 | 0.285 | 0.000 | 0.297 | 0.585 | 0.792 | 1.000 |
| GSM | 0.071 | 0.328 | −2.310 | 0.017 | 0.070 | 0.139 | 0.898 |
| Acid Ratio | 2.703 | 15.064 | 0.005 | 0.680 | 1.034 | 1.638 | 271.100 |
| Sales (in 000 Euro) | 89,142 | 313,281 | 1 | 4,044 | 15,307 | 46,128 | 3,532,510 |
| Consolidation | 0.086 | 0.281 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| Works Council | 0.406 | 0.491 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| HHI | 0.152 | 0.061 | 0.043 | 0.104 | 0.138 | 0.196 | 0.342 |
| Ln(#Subsidiaries) | 0.740 | 1.159 | 0.000 | 0.000 | 0.000 | 1.000 | 5.317 |
| Audit fee change 2004–2007 | 0.609 | 1.309 | −0.811 | 0.000 | 0.137 | 0.607 | 4.997 |
| Upward switchersd | |||||||
| Ln(Audit Fee) | 9.056 | 1.031 | 6.346 | 8.294 | 9.068 | 9.808 | 11.333 |
| Audit fee (in Euro) | 14,918 | 19,742 | 138 | 4,000 | 8,676 | 18,180 | 135,993 |
| IS | 0.303 | 0.460 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Initial | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |
| Sales Growth Post | 0.195 | 0.631 | −0.945 | −0.055 | 0.171 | 0.351 | 7.162 |
| SGP | 0.486 | 0.501 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| Ln(Total Assets) | 9.522 | 1.889 | 3.638 | 8.510 | 9.346 | 10.459 | 14.066 |
| Total Assets (in 000 Euro) | 115,564 | 518,408 | 17 | 4,965 | 11,457 | 34,874 | 7,055,082 |
| Equity/TA | 0.234 | 0.817 | −6.363 | 0.134 | 0.347 | 0.548 | 0.999 |
| Loss | 0.256 | 0.437 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| (Rec+Inv)/TA | 0.488 | 0.280 | 0.000 | 0.249 | 0.491 | 0.701 | 1.000 |
| GSM | 0.058 | 0.345 | −2.310 | 0.020 | 0.072 | 0.157 | 0.898 |
| Acid Ratio | 1.844 | 4.986 | 0.005 | 0.593 | 0.928 | 1.451 | 71.943 |
| Sales (in 000 Euro) | 53,097 | 171,212 | 1 | 4,302 | 13,732 | 30,942 | 1,710,149 |
| Consolidation | 0.117 | 0.322 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| Works Council | 0.342 | 0.475 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
| HHI | 0.142 | 0.060 | 0.043 | 0.096 | 0.131 | 0.173 | 0.342 |
| Ln(#Subsidiaries) | 0.982 | 1.265 | 0.000 | 0.000 | 0.000 | 1.693 | 5.635 |
| Audit fee change 2004–2007 | 0.660 | 1.277 | −0.811 | 0.008 | 0.146 | 0.714 | 4.997 |
| All engagements | Continuing engagements | Lateral switchers | Upward switchers | |||||
|---|---|---|---|---|---|---|---|---|
| # Obs. | % | # Obs. | % | # Obs. | % | # Obs. | % | |
| Panel B | ||||||||
| Industry groupe | ||||||||
| Finance/real estate (FIN) | 560 | 11.92 | 400 | 11.53 | 115 | 13.58 | 45 | 11.75 |
| Consumer durables (CDR) | 1,131 | 24.07 | 844 | 24.33 | 220 | 25.97 | 67 | 17.49 |
| Basic (BAS) | 381 | 8.11 | 271 | 7.81 | 75 | 8.85 | 35 | 9.14 |
| Food/tobacco (FTB) | 251 | 5.34 | 203 | 5.85 | 38 | 4.49 | 10 | 2.61 |
| Construction (CNS) | 250 | 5.32 | 169 | 4.87 | 35 | 4.13 | 46 | 12.01 |
| Capital goods (CAP) | 223 | 4.75 | 148 | 4.27 | 46 | 5.43 | 29 | 7.57 |
| Transportation (TRN) | 433 | 9.21 | 343 | 9.89 | 59 | 6.97 | 31 | 8.09 |
| Utilities (UTI) | 154 | 3.28 | 100 | 2.88 | 43 | 5.08 | 11 | 2.87 |
| Textiles/trade (TEX) | 652 | 13.88 | 501 | 14.44 | 94 | 11.10 | 57 | 14.88 |
| Services (SVS) | 484 | 10.30 | 367 | 10.58 | 87 | 10.27 | 30 | 7.83 |
| Leisure (LSR) | 180 | 3.83 | 123 | 3.55 | 35 | 4.13 | 22 | 5.74 |
| Total | 4,699 | 100 | 3,469 | 100 | 847 | 100 | 383 | 100 |
- a The number of observations for the sample of all engagements is 4,699. Because of missing data, the number of observations for the variable Sales Growth Post is only 3,717. For the audit fee change variable, the number of observations is limited to 2.701.
- b The number of observations for the sample of continuing engagements is 3,469. Because of missing data, the number of observations for the variable Sales Growth Post is only 2,775. For the audit fee change variable, the number of observations is limited to 2,002.
- c The number of observations for the sample of lateral switchers is 847. Because of missing data, the number of observations for the variable Sales Growth Post is only 648. For the audit fee change variable, the number of observations is limited to 485.
- d The number of observations for the sample of upward switchers is 383. Because of missing data, the number of observations for the variable Sales Growth Post is only 294. For the audit fee change variable, the number of observations is limited to 214.
- e The industry composition is based on the number of engagements in the different industry groups.
In Table 3, we report Pearson and Spearman correlations. We note that total assets and sales are highly correlated which is not surprising since both proxy for size. We note that our proxies for size have relatively high correlations with the HHI measure for concentration (i.e. the Spearman correlation between HHI and Ln(Total Assets) is 0.711). This is likely due to the fact that the HHI is mostly increasing with client size (i.e. the industry-deciles of the largest clients are also the most concentrated ones). We decide to retain all variables in our models because all variance inflation factors are below 5, which suggests that multicollinearity is not a severe issue.
| Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Ln(Audit Fee) | 0.022 | 0.402 | 0.075 | −0.101 | −0.039 | 0.470 | −0.041 | 0.009 | 0.037 | −0.058 | −0.070 | 0.564 | 0.569 | 0.110 | 0.243 | 0.178 | |
| 2 | Initial Engagement | 0.022 | 0.057 | 0.070 | −0.035 | −0.007 | 0.069 | −0.028 | 0.034 | −0.026 | −0.018 | −0.026 | 0.052 | 0.049 | 0.042 | −0.003 | 0.046 | |
| 3 | HHI | 0.395 | 0.056 | 0.113 | −0.019 | −0.031 | 0.721 | 0.072 | −0.066 | −0.192 | 0.022 | 0.005 | 0.514 | 0.575 | 0.267 | 0.267 | 0.373 | |
| 4 | IS | 0.075 | 0.070 | 0.106 | −0.018 | −0.025 | 0.099 | −0.013 | 0.035 | −0.040 | −0.012 | 0.011 | 0.092 | 0.099 | 0.054 | −0.052 | 0.017 | |
| 5 | Sales Growth Post | −0.034 | −0.004 | −0.022 | −0.012 | 0.432 | −0.051 | −0.019 | 0.033 | −0.017 | −0.104 | 0.025 | −0.207 | −0.174 | 0.018 | −0.016 | −0.003 | |
| 6 | SGP | −0.036 | −0.007 | −0.030 | −0.025 | 0.851 | −0.013 | 0.005 | −0.018 | 0.003 | −0.007 | −0.026 | −0.016 | −0.021 | −0.038 | −0.032 | −0.021 | |
| 7 | Ln(Total Assets) | 0.501 | 0.064 | 0.711 | 0.087 | −0.018 | −0.010 | 0.197 | −0.118 | −0.227 | 0.035 | 0.013 | 0.682 | 0.713 | 0.306 | 0.294 | 0.459 | |
| 8 | Equity/TA | −0.108 | −0.022 | 0.021 | −0.009 | −0.013 | −0.029 | 0.019 | −0.267 | −0.115 | 0.133 | 0.123 | 0.059 | 0.054 | 0.080 | 0.026 | 0.085 | |
| 9 | Loss | 0.004 | 0.034 | −0.072 | 0.035 | −0.023 | −0.018 | −0.112 | −0.288 | −0.008 | −0.222 | −0.042 | −0.115 | −0.113 | −0.019 | −0.012 | −0.035 | |
| 10 | (Rec+Inv)/TA | 0.016 | −0.028 | −0.187 | −0.041 | 0.004 | 0.002 | −0.232 | −0.140 | −0.008 | −0.143 | −0.024 | 0.055 | 0.032 | −0.158 | −0.053 | −0.277 | |
| 11 | GSM | −0.095 | −0.003 | 0.088 | −0.005 | 0.003 | −0.006 | 0.108 | 0.197 | −0.423 | −0.296 | −0.079 | 0.126 | 0.091 | −0.051 | 0.002 | −0.035 | |
| 12 | Acid Ratio | −0.120 | −0.034 | −0.120 | −0.009 | −0.010 | −0.034 | −0.164 | 0.540 | −0.183 | 0.173 | 0.062 | −0.172 | −0.149 | 0.000 | −0.004 | −0.025 | |
| 13 | Ln(Sales) | 0.577 | 0.052 | 0.506 | 0.089 | −0.029 | −0.014 | 0.726 | −0.112 | −0.117 | 0.011 | −0.034 | −0.204 | 0.985 | 0.165 | 0.276 | 0.259 | |
| 14 | Ln(Sales)² | 0.577 | 0.052 | 0.506 | 0.089 | −0.029 | −0.014 | 0.726 | −0.112 | −0.117 | 0.011 | −0.034 | −0.204 | 1.000 | 0.199 | 0.303 | 0.297 | |
| 15 | Consolidation | 0.113 | 0.042 | 0.254 | 0.054 | −0.021 | −0.038 | 0.299 | 0.111 | −0.019 | −0.154 | 0.029 | −0.063 | 0.174 | 0.174 | 0.058 | 0.453 | |
| 16 | Works Council | 0.253 | −0.003 | 0.239 | −0.052 | −0.026 | −0.032 | 0.303 | −0.007 | −0.012 | −0.059 | 0.000 | −0.062 | 0.294 | 0.294 | 0.058 | 0.112 | |
| 17 | Ln(#Subsidiaries) | 0.193 | 0.035 | 0.351 | 0.010 | −0.026 | −0.030 | 0.457 | 0.058 | −0.025 | −0.248 | 0.042 | −0.141 | 0.312 | 0.312 | 0.377 | 0.122 |
- Pearson correlations shown above the diagonal, Spearman correlations shown below the diagonal. The correlations are calculated for the sample of 4,699 observations that are used to test the main models (i.e. for H1 and H2). Correlations with the industry dummies are not reported for brevity. All Pearson and Spearman correlations for the industry dummies are below 30 per cent. The correlations were also calculated for the sample of 2,701 observations that is used to test the fee premiums in the next audit engagement but are not reported for brevity. No abnormal correlations were found.
5.3. Regression results
In this section, we report OLS regression results from the models that were used to test our hypotheses. As indicated above, each of the models in equations 1 and 2 is estimated for (i) all initials versus continuing engagements, (ii) upward switchers versus continuing engagements and (iii) lateral switchers versus continuing engagements. Standard errors are clustered per audit firm to correct for unobserved within-auditor correlation patterns (Petersen, 2009).
In Table 4, we report the analyses performed to test H1A and H1B. As expected, we find different results for upward and lateral switchers. As can be seen from the results reported for Model 1 in Table 4, the coefficient on Initial is not significant (p-value = 0.4753), indicating that there is no evidence of a general discount for all types of initial audit engagements as compared to continuing engagements. However, the coefficient on Initial is significant and negative (p-value = 0.0555) in Model 1A, which is in line with fee discounting by non-specialist Big 4 auditors to clients switching from a non-Big 4 auditor. This result is supportive of hypothesis 1A. Note further that Initial is not significant in Model 1B (p-value = 0.1487). We further also see that InitialxIS is significant and negative (p-value = 0.0347) in Model 1, indicating that industry leaders grant fee discounts on initial engagements. However, a closer look reveals that this is driven by the subsample of lateral switchers as InitialxIS is not significant in Model 1A (subsample of upward switchers), but is significant and negative (p-value = 0.0143) in Model 1B, which indicates that industry leaders do not offer fee discounts to upward switchers but do so to lateral switchers. Former non-Big 4 clients may not be of interest for the industry specialist and/or may not have a demand for the specialist's services. These results are consistent with hypothesis 1B. The fee discount for lateral switchers can be seen as evidence consistent with experience good arguments. Finally, in all three models, we find significant fee premiums for industry specialization (IS) and higher fees in more concentrated industry-deciles (HHI). All significant control variables have signs in line with prior audit fee research.
| Dependent variable: Ln(Audit Fee) | Model 1 (All initials) | Model 1A (Upward) | Model 1B (Lateral) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | |
| Intercept | 5.3894 | 12.42 | 0.0011 | 5.3066 | 12.45 | 0.0011 | 5.3297 | 13.57 | 0.0009 |
| Initial | 0.0459 | 0.81 | 0.4753 | −0.1730 | −3.05 | 0.0555 | 0.1488 | 1.93 | 0.1487 |
| IS | 0.1685 | 13.42 | 0.0009 | 0.1872 | 6.66 | 0.0069 | 0.1516 | 8.63 | 0.0033 |
| InitialxIS | −0.3082 | −3.68 | 0.0347 | −0.1351 | −0.88 | 0.4427 | −0.3984 | −5.13 | 0.0143 |
| Ln(Total Assets) | 0.1140 | 2.41 | 0.0948 | 0.1051 | 2.44 | 0.0924 | 0.1165 | 2.72 | 0.0728 |
| Equity/TA | −0.1308 | −7.58 | 0.0048 | −0.1436 | −4.91 | 0.0162 | −0.1405 | −7.52 | 0.0049 |
| Loss | 0.1054 | 9.50 | 0.0025 | 0.1150 | 8.09 | 0.0039 | 0.0981 | 6.34 | 0.0079 |
| (Rec+Inv)/TA | 0.0446 | 1.93 | 0.1486 | 0.0032 | 0.16 | 0.8807 | 0.0433 | 1.42 | 0.2504 |
| GSM | −0.2944 | −15.93 | 0.0005 | −0.3563 | −11.81 | 0.0013 | −0.2906 | −27.51 | 0.0001 |
| Acid Ratio | 0.0006 | 1.19 | 0.3206 | 0.0001 | 0.11 | 0.9208 | 0.0006 | 1.30 | 0.2854 |
| Ln(Sales) | 0.2616 | 3.57 | 0.0377 | 0.3020 | 3.72 | 0.0339 | 0.2732 | 3.86 | 0.0307 |
| Ln(Sales)² | −0.0040 | −1.05 | 0.3707 | −0.0063 | −1.50 | 0.2304 | −0.0047 | −1.23 | 0.3057 |
| Consolidation | −0.0635 | −0.31 | 0.7772 | 0.0535 | 0.35 | 0.7478 | −0.0719 | −0.33 | 0.7648 |
| Works Council | 0.1640 | 2.00 | 0.1395 | 0.1505 | 1.99 | 0.1401 | 0.1714 | 1.92 | 0.1502 |
| HHI | 2.1031 | 4.81 | 0.0171 | 2.5977 | 5.75 | 0.0105 | 1.9563 | 5.30 | 0.0131 |
| Ln(#Subsidiaries) | −0.0091 | −0.48 | 0.6654 | −0.0123 | −0.45 | 0.6823 | −0.0088 | −0.60 | 0.5885 |
| Auditor and industry dummies | Included | Included | Included | ||||||
| Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | |
|---|---|---|---|---|---|---|---|---|---|
| Standard errors are clustered per audit firm (i.e. 4 clusters) | 4,699 | 0.3906 | 0.3869 | 3.852 | 0.4051 | 0.4007 | 4.316 | 0.3953 | 0.3914 |
To test hypotheses 2A and 2B, we investigate whether Big 4 auditors target high-growth firms by offering fee discounts to attract such firms. The results from this analysis are included in Table 5. For hypothesis 2A, the variable of interest is InitialxSGP capturing the effect of a predicted additional fee discount for previous non-Big 4 private clients that are expected to be faster growing. From Model 2A in Table 5, we see that the coefficient on InitialxSGP is positive and not significant, which does not support H2A.
| Dependent variable: Ln(Audit Fee) | Model 2 (All initials) | Model 2A (Upward) | Model 2B (Lateral) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | |
| Intercept | 5.2515 | 13.22 | 0.0009 | 4.9694 | 12.73 | 0.0010 | 5.1908 | 15.94 | 0.0005 |
| Initial | −0.0278 | −0.56 | 0.6158 | −0.2654 | −2.77 | 0.0697 | 0.0973 | 1.13 | 0.3413 |
| IS | −0.0379 | −1.24 | 0.3017 | −0.0244 | −0.57 | 0.6101 | −0.0570 | −1.60 | 0.2089 |
| SGP | −0.1435 | −4.14 | 0.0255 | −0.1455 | −4.00 | 0.0279 | −0.1433 | −4.18 | 0.0250 |
| InitialxIS | 0.0534 | 0.40 | 0.7145 | 0.0726 | 0.29 | 0.7900 | 0.0048 | 0.04 | 0.9722 |
| InitialxSGP | 0.1478 | 2.93 | 0.0608 | 0.1864 | 2.11 | 0.1255 | 0.1162 | 1.73 | 0.1824 |
| ISxSGP | 0.3427 | 7.90 | 0.0042 | 0.3478 | 8.10 | 0.0039 | 0.3433 | 7.85 | 0.0043 |
| ISxSGPxInitial | −0.5056 | −3.97 | 0.0286 | −0.4118 | −2.04 | 0.1340 | −0.5291 | −3.81 | 0.0319 |
| Ln(Total Assets) | 0.0916 | 2.05 | 0.1322 | 0.0953 | 3.04 | 0.0559 | 0.0937 | 2.48 | 0.0894 |
| Equity/TA | −0.0968 | −6.68 | 0.0068 | −0.1177 | −5.31 | 0.0130 | −0.0948 | −5.00 | 0.0154 |
| Loss | 0.0754 | 3.68 | 0.0348 | 0.0699 | 2.02 | 0.1366 | 0.0675 | 2.94 | 0.0605 |
| (Rec+Inv)/TA | 0.0618 | 1.59 | 0.2096 | 0.0111 | 0.39 | 0.7220 | 0.0714 | 2.18 | 0.1177 |
| GSM | −0.2778 | −3.89 | 0.0301 | −0.3480 | −5.77 | 0.0103 | −0.2838 | −3.85 | 0.0309 |
| Acid Ratio | 0.0019 | 4.83 | 0.0169 | 0.0013 | 1.25 | 0.3006 | 0.0018 | 4.05 | 0.0271 |
| Ln(Sales) | 0.3291 | 3.78 | 0.0324 | 0.3971 | 4.31 | 0.0231 | 0.3428 | 4.11 | 0.0261 |
| Ln(Sales)² | −0.0061 | −1.35 | 0.2700 | −0.0105 | −2.12 | 0.1238 | −0.0069 | −1.59 | 0.2094 |
| Consolidation | 0.0233 | 0.15 | 0.8883 | 0.0409 | 0.23 | 0.8332 | 0.0259 | 0.17 | 0.8736 |
| Works Council | 0.1728 | 2.03 | 0.1348 | 0.1844 | 2.89 | 0.0628 | 0.1638 | 1.78 | 0.1729 |
| HHI | 2.1838 | 5.81 | 0.0102 | 2.5787 | 5.66 | 0.0109 | 2.0888 | 6.03 | 0.0091 |
| Ln(#Subsidiaries) | 0.0057 | 0.21 | 0.8461 | −0.0083 | −0.24 | 0.8233 | 0.0098 | 0.45 | 0.6856 |
| Auditor and industry dummies | Included | Included | Included | ||||||
| Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | |
|---|---|---|---|---|---|---|---|---|---|
| Standard errors are clustered per audit firm (i.e. 4 clusters) | 3,717 | 0.4142 | 0.4091 | 3.069 | 0.4197 | 0.4136 | 3.423 | 0.4202 | 0.4147 |
To test H2B, we inspect the results in Model 2B regarding the interaction variable InitialxISxSGP as it is capturing the effect of a predicted additional fee discount for previous non-industry specialist Big 4 private clients that are expected to be faster growing. This coefficient is as predicted negative and significant (p-value = 0.0319). Taken together with the insignificant coefficient on InitialxIS (p-value = 0.0079), these results suggest that industry specialists only offer fee discounts to fast-growing lateral switchers, which supports our hypothesis 2B. Finally, we conclude from all three models in Table 5 that premiums for industry specialization are only paid by fast-growing firms (as evidenced by the insignificant coefficients on IS and the very significant positive coefficients on ISxSGP).
5.4. Supplemental analysis of audit fees in the next mandate period
As noted in section 2., the initial fee discounts to private audit clients that purchase a quality upgrade are provided on the basis that the auditor will be able to increase audit fees in future periods (Shapiro, 1983). Hence, as an additional test of the validity of the experience argument underlying our hypotheses, we test whether the fees were actually increased in the subsequent mandate period. To test whether fees increased in the next mandate period, we specify a fee change model controlling for changes in underlying fee drivers over the period 2004 to 2007. In particular, we construct a new dependent variable being the Audit Fee2007-Audit Fee2004)/Audit Fee2004 to capture the fee increases at the start of the second mandate period (i.e. after the first period from 2004 to 2006). For the control variable equivalents used in equations 1 and 2, we calculated changes in the same way consistent with Huang et al. (2009). A detailed specification of the fee change model that we adopted is given in Appendix II. We investigate whether there are fee increases at the start of the following fee mandate period (2007) both for clients that upgraded from a non-Big 4 auditor to a Big 4 (model 3A in Table 6) as well as those that laterally switched in 2004 (model 3B in Table 6). The main variables of interest are InitialxSGP for upward switchers and InitialxISxSGP for lateral switchers.
| Dependent variable: (Audit Fee2007 − Audit Fee2004)/Audit Fee2004 | Model 3 (All initials) | Model 3A (Upward) | Model 3B (Lateral) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | Coeff. | t-stat. | p-value | |
| Intercept | 0.4311 | 4.18 | 0.0249 | 0.4104 | 2.40 | 0.0959 | 0.4388 | 5.85 | 0.0099 |
| Initial | −0.0378 | −0.94 | 0.4151 | −0.1075 | −1.18 | 0.3230 | −0.0085 | −0.11 | 0.9226 |
| IS | 0.0696 | 0.81 | 0.4787 | 0.0055 | 0.05 | 0.9648 | 0.0851 | 0.99 | 0.3961 |
| SGP | −0.0513 | −0.98 | 0.3994 | 0.0101 | 0.11 | 0.9183 | −0.0609 | −1.47 | 0.2380 |
| InitialxIS | 0.0070 | 0.06 | 0.9593 | 0.5859 | 2.53 | 0.0854 | −0.2027 | −1.41 | 0.2524 |
| InitialxSGP | −0.1053 | −1.00 | 0.3893 | 0.0130 | 0.05 | 0.9621 | −0.1510 | −0.90 | 0.4328 |
| ISxSGP | −0.5061 | −8.16 | 0.0039 | −0.5100 | −7.38 | 0.0052 | −0.5102 | −8.81 | 0.0031 |
| ISxSGPxInitial | 0.6591 | 6.04 | 0.0091 | −0.0121 | −0.04 | 0.9711 | 0.9147 | 5.76 | 0.0104 |
| (TA2007-TA2004)/TA2004 | 0.0343 | 0.55 | 0.6204 | 0.0598 | 0.72 | 0.5239 | 0.0484 | 0.81 | 0.4759 |
| (EQ/TA2007-EQ/TA2004)/(EQ/TA2004) | 0.0299 | 4.72 | 0.0180 | 0.0247 | 1.03 | 0.3775 | 0.0426 | 10.10 | 0.0021 |
| (GSM2007-GSM2004)/GSM2004 | 0.0156 | 0.51 | 0.6437 | 0.0319 | 0.72 | 0.5233 | 0.0152 | 0.53 | 0.6336 |
| (CR2007-CR2004)/CR2004 | 0.0189 | 1.32 | 0.2771 | 0.0115 | 0.48 | 0.6672 | 0.0133 | 1.05 | 0.3707 |
| (RECINVTA2007-RECINVTA2004)/RECINVTA2004 | −0.0427 | −0.80 | 0.4835 | −0.0698 | −2.28 | 0.1072 | −0.0382 | −0.94 | 0.4180 |
| (SALES2007-SALES2004)/SALES2004 | 0.3454 | 3.17 | 0.0507 | 0.1850 | 2.59 | 0.0812 | 0.3585 | 2.82 | 0.0665 |
| ((SALES2007-SALES2004)/SALES2004)² | 0.0482 | 0.68 | 0.5436 | 0.1175 | 1.55 | 0.2196 | 0.0433 | 0.71 | 0.5297 |
| (HHI2007-HHI2004)/HHI2004 | 0.2974 | 3.10 | 0.0532 | 0.3313 | 2.87 | 0.0639 | 0.2529 | 2.75 | 0.0709 |
| (#SUBS2007-#SUBS2004)/#SUBS2004 | −0.0309 | −1.22 | 0.3087 | −0.0363 | −1.39 | 0.2597 | −0.0258 | −0.92 | 0.4262 |
| Auditor and industry dummies | Included | Included | Included | ||||||
| Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | Obs. | R² | Adj. R² | |
|---|---|---|---|---|---|---|---|---|---|
| Standard errors are clustered per audit firm (i.e. 4 clusters). | 2,701 | 0.0735 | 0.0634 | 2.216 | 0.0845 | 0.0724 | 2.487 | 0.0689 | 0.0580 |
To test the fee change models in Table 6, we collected fee and other data for the year 2007 to calculate change variables, and we exclude any further switches at the start of the new mandate period so as to focus on the outcomes for those clients that made the switches back in 2004 and stayed on in the new mandate period. Missing data account for the reduced sample size of 2,701 (see also Table 2). The results from this analysis are reported in Table 6. In the case of upward switches from a non-Big 4 auditor to a Big 4 auditor (model 3A), fast-growing firms do not experience significant fee increases in 2007 (p-value = 0.9621). This is actually not surprising as hypothesis 2A was also not confirmed for these types of client firms as we did not find significant fee discounting for fast-growing upward switchers in 2004. Note, however, that the results in model 3A in Table 6 indicate that significant fee increases only occur for upward switchers that appointed an industry leader audit firm, as InitialxIS is positive and significant in model 3A.
In the case of lateral switches involving switches from a non-specialist Big 4 to a specialist Big 4 (model 3B), those targeted for their expected growth over 2004–2007 (consistent with H2B) do experience significant audit fee increases (the coefficient on InitialxISxSGP is positive and significant at p = 0.0104). These results support that faster growing firms targeted by the Big 4 having received the largest discounts in the initial mandate period generate (and are willing to pay) the largest fee increases for the Big 4 in the next mandate after the clients have experienced the value of the higher audit quality.
5.5. Robustness tests
In this subsection, we report untabulated results of robustness checks that we performed. As we ran a large number of tests, we focus on reporting whether a robustness checks confirms the main findings from models 1A and B, 2A and B and 3A and B. In particular, we check whether Initial remains significantly negative in model 1A (support of H1A); InitialxIS remains significantly negative in model 1B (support of H1B); InitialxSGP is not significant in model 2A (lack of support of H2A); InitialxISxSGP remains significantly negative in model 2B; InitialxSGP is not significant in model 3A (lack of evidence of fee increase for fast-growing upward switchers); InitialxISxSGP remains significantly positive in model 3B (support of fee increase for fast-growing lateral switchers with an industry specialist).
5.5.1. Self-selection concerns
The design of our tests could be subject to self-selection concerns, as it mixes fee models with switching variables. We therefore re-ran all our tests adopting a two-stage model in which we model auditor switching in stage one and then run all the fee models in stage two including the probability of switching obtained from stage one in these models. The specification of our stage one switch model is based on López and Peters (2011). As the explanatory power of our stage 1 model is not very high (R2 = 0.0177), we report this test as a robustness test instead of our main analysis. Note that all results remain robust to this two-stage specification of our research design (in particular: p-value for H1A = 0.0658, p-value for H1B = 0.0137; p-value for H2A = 0.1164; p-value for H2B = 0.0421).
5.5.2. Specification of industry specialization
We examine the sensitivity of our findings with regard to the definition for industry specialization. In our main analyses, both specialization and concentration measures are based on fee market shares. However, we also repeated all our analyses with measures based on client assets market shares. This yields very similar findings, and all our main conclusions hold in this alternative specification (in particular: p-value for H1A = 0.0555, p-value for H1B = 0.0143; p-value for H2A = 0.1255; p-value for H2B = 0.0319).
5.5.3. Ownership concentration as an additional control variable
Ownership concentration is likely to be an important control variable for the demand for auditing. We were able to collect data on the number of shareholders in private firms have. As we lose an important number of our observations, we decided to re-run all the tests on the reduced sample and include Ln(shareholders), defined as the natural log of the number of shareholders, as a control variable in all our models (Models 1, 1A, 1B, 2, 2A, 2B). All our results remain robust to this new model specification (in particular: p-value for H1A = 0.0700, p-value for H1B = 0.0143; p-value for H2A = 0.1445; p-value for H2B = 0.0353). Note that Ln(shareholders) is significant in some but not all models.
5.5.4. Alternative specification of growth measure
In Models 2 (A and B) and 3 (A and B), we use a growth measure. To test the robustness of our results, we re-ran all these models using an alternative specification of the growth proxy. Instead of defining fast-growing firms as those for which the sales growth rate (over the period 2004–2007) exceeds the median values in the industry, we define fast-growing firms as those for which sales growth rate is in the top quartile within the industry. Both the results for hypotheses H2A (positive coefficient with p-value = 0.0984) and H2B (p-value = 0.0785) as well the fee effects results in the next mandate period hold (p-value = 0.9401 for upward switchers and 0.0037 for lateral switchers).
5.5.5. Results without Ln(sales) and Square of Ln(sales) in the fee model
Client size is the most important determinant of audit fees, and typically, Ln(Assets) is included to control for size in audit fee models (Hay et al., 2006). In our analysis, we not only include Ln(Assets) as a control variable for client size, but based on prior fee research in private client settings, we also include Ln(Sales) and the square of Ln(Sales). To test the sensitivity of our results to including these two additional size variables, we re-ran all models in Tables 4, 5 and 6 without Ln(Sales) and the square of Ln(Sales). All our results remain robust to this alternative model specification (in particular: p-value for H1A = 0.0742, p-value for H1B = 0.0174; for H2A, we even find – opposite to what was predicted – a positive coefficient with p-value = 0.0695; p-value for H2B = 0.0312; p-value for H3A = 0.9462 and H3B = 0.0080).
5.5.6. Split-sample regressions instead of three-way interactions
The regression analyses in Tables 5 and 6 make use of three-way interactions that may complicate the interpretation of the results. Therefore, we re-ran all models in Tables 5 and 6 using a split-sample approach instead, splitting the sample between faster and slower growing firms. As expected, our results continue to hold when we use a split-sample approach. In particular, for model 2A (upward switchers), we find significant negative coefficients on the Initial variable in the both fast-growing (p-value = 0.0032) and slow-growing clients (p-value = 0.0810) subsample. For model 2B, (lateral switchers) the coefficient on the InitialxIS interaction is only significant (and negative) in the subsample of fast-growing clients (p-value = 0.0009) in line with our hypothesis 2B. For model 3A (upward switchers), we find no significant coefficients on the Initial variable in either of the subsamples. For model 3B (lateral switchers), the coefficient on the InitialxIS interaction is only significant (and positive) in the subsample of fast-growing clients (p-value = 0.0020) in line with our main results and consistent with our theory.
5.5.7. Alternative tests for the tests of fee increases in next mandate period
To test whether fee increases occur in the next mandate period for those switchers that received a discount to experience an upgrade and/or because they are fast-growing, we estimated alternative models to those in Table 6. On top of including change variables on the right-hand side of the fee equation, we also included level variables as an additional control. All results remain robust. In addition, we also re-ran the models in Table 6 by specifying the dependent variable as the unscaled change in the natural logarithm of audit fees (as in Huang et al., 2009) between 2007 and 2004. All results remain robust (in particular: p-value = 0.9524 for upward switchers and p-value = 0.0225 for lateral switchers).
6. Conclusions
This study examines how Big 4 auditors compete for new clients to gain market share in the private client audit market. Private clients are the largest market segment targeted by auditors and about which little is known in comparison with public listed client segments. The market for private audit clients is less concentrated than the public client market as the Big 4 auditors compete not only with other Big 4 auditors but also to a large extent with non-Big 4 audit suppliers.
We consider two strategies of competing on price and product differentiation and investigate whether quality differentiation affects price competition. In line with Craswell and Francis (1999) and Ferguson et al. (2006), we argue that Big 4 audit firms offer fee discounts on initial engagements to let clients who are unfamiliar with their services experience the ‘experience’ attributes of their brand name audits at a lower initial price. Using a very comprehensive sample of Big 4 audits of private companies in Belgium in 2004, we find that this is the case for new engagements that were previously audited by non-Big 4 auditors (upward switchers), but find no fee discounting when the client switches from another Big 4 competitor (lateral switchers). We build further on experience good theory by arguing that to attract new clients from competing Big 4 auditors (lateral switchers), the Big 4 target their fee discounting efforts at those industries where they are already specialists. Our results are consistent with this.
Next, we also examined whether the Big 4 target growth clients through fee discounting. We do not find fee discounts for upward switches that are expected to be fast-growing. Our results do indicate that Big 4 industry specialists indeed offer fee discounts on initial engagements that are expected to grow faster than the median firm in their industry.
Finally, we provide evidence of an additional test of our experience attribute arguments used to motivate our hypotheses and extend our analysis to look at what happens in the second fee mandate period for clients that received a discount in the initial mandate period. These fee discounts are targeted at the clients expected to be faster growing and are provided on the basis that audit fees in future periods will increase (Shapiro, 1983). Our results indeed show fee increases in the next mandate period for Big 4 clients that switched to an industry leader and are fast-growing.
This article contributes to the literature by providing evidence that suggests that Big 4 auditors use fee discounting on initial engagements strategically to attract new clients in the private client audit market. In particular, the findings of this study suggest that auditors' product differentiation through Big 4 brand name reputation and industry specialization facilitates fee discounting. To our knowledge, no prior study has shown this relation between fee discounting and industry specialization. Additionally, we document that only industry specialist auditors offer fee discounts to new clients who we argue are more likely to retain their auditor after the initial mandate. Prior research on fee discounting reports no significant association between fee discounts and client characteristics (e.g. Ettredge and Greenberg, 1990; Turpen, 1990). Finally, we believe that research on fee-setting and industry specialization in a private client setting is particularly useful as the competitive nature of such a market differs significantly from the public client settings that were used in prior studies.
The results of this study are subject to a number of limitations. First, industry specialization is difficult to measure and it is unclear whether expertise is developed by auditing many clients or by auditing the largest clients. Second, although Belgium is not the only setting with mandatory audits for private firms and with lower market concentration, it is possible that the results are not valid in other countries even if they have similar characteristics. Because of three-year audit mandates and higher switching costs, it is possible that Belgian audit firms offer more fee discounts than auditors in countries without such regulation.
Finally, there is scope for further research to examine whether there are detrimental effects, such as less competent audits because of any squeeze in margins, from the fee discounting behaviour documented for the initial mandate period.
Notes
Appendix 1
Campbell (1996) industry classification
| Description | US SIC industries | Mnemonic | |
|---|---|---|---|
| 1 | Petroleum industry | 13, 29 | PET |
| 2 | Finance/real estate industry | 60–69 | FIN |
| 3 | Consumer durables industry | 25, 30, 36, 37, 50, 55, 57 | CDR |
| 4 | Basic industry | 10, 12, 14, 24, 26, 28, 33 | BAS |
| 5 | Food/tobacco industry | 1, 20, 21, 54 | FTB |
| 6 | Construction industry | 15–17, 32, 52 | CNS |
| 7 | Capital goods industry | 34, 35, 38 | CAP |
| 8 | Transportation industry | 40–42, 44, 45, 47 | TRN |
| 9 | Utilities industry | 46, 48, 49 | UTI |
| 10 | Textiles/trade industry | 22–23, 31, 51, 53, 56, 59 | TEX |
| 11 | Services industry | 72, 73, 75, 80, 82, 89 | SVS |
| 12 | Leisure industry | 27, 58, 70, 78, 79 | LSR |
The following two-digit SIC codes are not designated to one of the twelve industry groups: 2 (agricultural production – livestock and animal specialties), 7 (agricultural services), 8 (forestry), 9 (fishing, hunting and trapping), 39 (miscellaneous manufacturing industries), 76 (miscellaneous repair services), 81 (legal services), 83 (social services), 84 (museums, art galleries, and botanical and zoological gardens), 86 (membership organizations), 87 (engineering, accounting, research, management and related services), 91 (executive, legislative, and general government, except finance) and 96 (administration of economic programmes). The petroleum industry is excluded from the analyses because it is considered an outlier.
Appendix 2
Specification of fee changes model
