Competitive Procurement of Auditing Services with Limited Information

Abstract

We analyse the auditing procurement process when the client is initially uncertain about her/her auditing needs and when she/he cannot discern the level of assurance provided by the auditor. The client's limited information can lead to under-auditing and substantial rent for auditors, despite the presence of ex ante competition among auditors. The information asymmetry can also make it difficult for new auditors with superior auditing technologies to displace incumbent auditors. In addition, the asymmetry can limit the incentives of auditors to improve their technologies.

Acknowledgements

We thank the editor Laurence Van Lent, the associate editor Jeroen Suijs, two anonymous reviewers, and Martin Wu for very helpful comments and suggestions.

Notes

Dulleck and Kerschbamer (2006, p. 5) observe that a credence good is one for which ‘an expert knows more about the type of good or service [a] consumer needs than the consumer himself’. Automobile repair is a leading example of a credence good. The typical driver is not an expert on the internal workings of her vehicle. When a driver takes her automobile to a repair shop, she/he usually relies on the mechanic to diagnose and fix the problem. When the driver leaves the shop with the ‘repaired’ vehicle, she/he often knows only what the mechanic said she/he did to the automobile, which may differ from what she/he actually did.

Auditor reputation can serve as an informative but imperfect signal of likely audit quality even when audits are credence goods. This will be the case, for example, when realised audit quality is revealed occasionally and imperfectly through the legal or regulatory process.

Simunic (1980) assumes that clients determine how much auditing they need based on their trade-off of internal and external monitoring. In practice, the auditor typically determines how much effort is needed to conduct an audit in the process of assessing a client's risk. The appropriate level of testing in an audit reflects the minimal standards for auditing (GAAS) and the target level of assurance for the specific client, attributes that are best determined by the auditor conducting the engagement.

Kanodia and Mukherji (1994, p. 595) observe that ‘Audits are not standardised products… [T]he efficient design of an audit consistent with GAAS requires detailed client-specific information, much of which is discovered by an auditor in the process of conducting the initial audit for the specific client’. Thus, only the auditor performing the audit can know the extent of auditing that is needed for an engagement even if the client has substantial knowledge of her risk type.

The appropriate level of effort in a given engagement can vary with a variety of factors, including the client's desired level of assurance, professional standards, an inspector's interpretation of standards, and/or the balancing of costs and benefits. The appropriate level of audit effort in our model balances the benefits that accrue to the client and the auditor's costs. This level of effort can differ from the baseline level dictated by professional standards.

Institutional mechanisms that can limit under-auditing include: (i) litigation (Dye, 1993; Willenborg, 1999; Khurana and Raman, 2004); (ii) regulations and standards, including the GAAS, the AICPA Code of Professional Conduct, and the Quality Control Standards; and (iii) ex post examination of auditors' work through peer reviews and inspections by the Public Company Accounting Oversight Board (PCAOB) (Knechel et al., 2007).

We abstract from non-salvageable investments in brand names and advertising that an auditor might make in order to signal that she/he will supply a high-quality audit, and thereby secure repeat business from clients (Klein and Leffler, 1981; Watts and Zimmerman, 1986). Although an auditor may rationally develop and maintain a favourable reputation in a market where delivered quality is ultimately observed, reputation formation and protection can be more difficult when delivered quality is not readily observed (Klein and Leffler, 1981; Dulleck et al., 2011).

Empirical studies suggests that audit quality initially increases as auditor tenure increases, but eventually declines with tenure (Davis et al., 2009; Brooks et al., 2011).

Of course, there may be other costs and benefits of mandatory auditor rotation that our model does not address explicitly.

Although our model does not explicitly consider repeated interaction between the client and her/his auditor, the two potential auditors in our model can be viewed as an incumbent auditor and a new auditor. Therefore, our analysis of the client's choice of an auditor might be viewed as an investigation of a client's decision to switch auditors. Our finding that the client may switch to a new auditor who has more to lose from audit failure than does the incumbent auditor complements other explanations for auditor switching that have been suggested in the literature. For instance, a client may switch auditors when the incumbent auditor makes an error (Dye, 1991; Teoh, 1992).

Consequently, an audit failure results when an auditor performs a routine audit in the presence of fraud.

The presumed constant unit cost of delivering audit service abstracts from capacity constraints, as analysed in Emons (1997). The simplifying assumption that the two auditors incur the same unit cost of delivering audit services facilitates a streamlined exposition of our main conclusions and a focus on other differences of interest. Our findings change in obvious ways when the auditors have different unit costs of delivering audit service.

In practice, a client may be able to observe the number of hours supplied by the staff of an audit firm. However, clients typically are unable to evaluate accurately whether those hours are appropriate or accurately assess the diligence with which the staff operates, the extent of staff expertise, or the quality of the tests (i.e. the analytical procedures) the staff implements.

The SOX requirement that audit committees include a financial expert may help to limit the information problems stemming from the credence nature of the audit. However, auditors typically possess a substantial information advantage over even relatively knowledgeable audit committees.

Smith et al. (2000) and Patterson and Smith (2007) analyse settings in which an auditor first undertakes control testing to assess the likelihood of fraud and then engages in substantive testing to detect fraud. Smith et al. (2000) demonstrate that the two forms of testing together can be more cost effective than substantive testing alone. Patterson and Smith (2007) demonstrate that the increased control testing mandated by SOX may diminish incentives for substantive testing and thereby increase the likelihood of undetected fraud. Our analysis differs from these studies in part by focusing on competition among potential auditors and on the credence nature of audits.

It is readily verified that and .

The expected losses from audit error and auditor negligence can be viewed as being determined stochastically by ex post legal or regulatory investigations. Although the client cannot determine presently whether an auditor has delivered a routine or a complex audit, a court or other investigative body may eventually make such a determination via a costly, potentially imperfect process that is summarised in the expected loss parameters and .

This loss is best viewed as the client's expected net loss after pursuing all appropriate legal avenues to recover relevant damages from the auditor.

The audit fees that are paid in equilibrium are characterised in detail below.

The key qualitative effects emphasised below continue to arise if .

In practice, client staff typically must spend a substantial amount of time helping a new auditor understand the client's operations, systems, and financial reporting practices.

Thus, the client cannot use the threat of a ‘second opinion’ (Pesendorfer and Wolinksy, 2003) to discipline the selected auditor.

Causholli and Knechel (in press) observe that a client can secure the full-information outcomes in the presence of perfect monitoring and disciplining mechanisms.

When she/he is uncertain how each auditor will react to the information she/he observes privately, the client will select an auditor on the basis of her expectations about each auditor's likely behaviour. An auditor who is uncertain about his rival's full expected cost must decide how aggressively to bid for the right to serve the client.

, where is the probability that auditor j observes signal when management is type .

In a dynamic setting, an auditor might offer to conduct an audit in return for a fee below his full expected cost if she/he believed she/he could offset the short-term financial loss with a long-term financial gain derived from future interaction with the client.

For expositional simplicity, we assume the client hires auditor I when she/he is indifferent between hiring the two auditors.

This might be the case, for example, if the auditors are big-four audit firms with substantial reputations to protect.

This reduced loss from hiring auditor is the product of the client's loss from undetected fraud ( ), the probability that fraud is present ( ), the probability that the auditor observes signal when fraud is present ( ), and the probability that the auditor's complex audit uncovers the existing fraud ( ).

In essence, an auditor's advantage in delivering benefits to the client operates like a reduction in the auditor's relative cost when determining which auditor will secure the client's patronage.

The corresponding characterisation of the equilibrium outcomes when the new auditor always delivers a routine audit whereas the incumbent auditor delivers a complex audit if and only if she/he observes signal is similar, and so is omitted.

Zhang and Thoman (1999) demonstrate that auditors increase audit effort when they face higher losses from audit failures.

This stark conclusion reflects the binary effort structure in our model. However, limited ability to commit to use an improved audit technology would seem to limit incentives for marginal improvements in technology more generally. In contrast, an auditor may implement a more costly major improvement in his audit technology if the improvement ensures that she/he will find it profitable to deliver a complex audit upon seeing signal .

This may be the case, for example, if the new auditor faces a higher probability of PCAOB inspection than does the incumbent auditor.

A corresponding conclusion is readily derived in the setting where only auditor I will deliver a complex audit (upon observing signal ). In this setting, the client will hire auditor I if the sum of his incumbency advantages ( ) and the reduction in the client's expected loss from undetected fraud that her/his complex audit provides ( ) exceeds any cost disadvantages that auditor I might face. These disadvantages reflect auditor 's higher expected direct auditing cost ( ) and any excess of her/his expected losses from auditor negligence and audit error ( ) over auditor 's expected loss from negligence ( ).

This loss will tend to be small (and so the auditor is more likely to behave opportunistically), when the risk that under-auditing will be exposed is low (perhaps due to a small chance of litigation or ex post review).

This finding is consistent with empirical research that finds auditors may under-audit when they face fee pressure (Alderman and Deitrick, 1982; Houston, 1999; Jenkins et al., 2008).

Regulation can be a double-edged sword. Excessive regulation can induce over-auditing (Levine, 2009), as the recent controversy surrounding section 404 of SOX indicates.

The client will also continue to select the most efficient auditor when the auditors face different incremental unit costs (c) of supplying the auditing service. In this case, the selected auditor's rent will reflect in part his incremental cost advantage in delivering the auditing service.

For example, the client might opt for a review of financial statements rather than a full audit.

This latter consideration is particularly relevant in light of EU proposals to require joint audits of public interest entities (European Commission, 2011a).

In particular, the substantial economies of scope between assessing risk and conducting an audit typically render the separation of these two functions uneconomic. Furthermore, the cost of engaging multiple auditors simultaneously often is prohibitive.

Our analysis is readily extended to analyse over-auditing. If, for example, inequality (1) does not hold for and for , then both auditors will always deliver a complex audit even though the client prefers a routine audit when the selected auditor observes signal . The auditors do not tailor their actions to their private information in this setting, and so the auditor with the most accurate auditing technology may not be selected. Furthermore, equilibrium fees typically will exceed their full-information levels.

To avoid an unduly lengthy analysis, Lemma 5 focuses on the case in which , so auditor 's full expected cost is higher when than when . Conclusions similar to those reported in Lemma 5 prevail when .

As the proof of Lemma 5 reveals, this will be the case in the setting with more limited client knowledge when and Condition 1 does not hold.

Different types of information asymmetry can prevail in practice. For example, the incumbent auditor may have better knowledge of the client's switching cost than the new auditor. Alternatively or in addition, each auditor may be the only party to be fully informed about his own costs, and the other auditor may have better information about these costs than the client.

Additional information

Notes on contributors

Monika Causholli

Paper accepted by Jeroen Suijs.

W. Robert Knechel

Paper accepted by Jeroen Suijs.

Haijin Lin

Paper accepted by Jeroen Suijs.

David E. M. Sappington

Paper accepted by Jeroen Suijs.