How Auditor Legal Liability Influences the Detection and Frequency of Fraudulent Financial Reporting

In the wake of audit failures around the new millennium, regulators and politicians considered many reforms intended to enhance auditors' efforts to detect fraudulent financial reporting (hereafter referred to as fraud). In addition, the audit profession has been actively seeking protection from extreme liability claims that leaders in the audit profession argue could cause audit firms to fail, which has the potential to lead to an economic shock to the global economy. For example, Jim Turley, CEO of Ernst & Young, provided the following testimony to the U.S. Treasury Advisory Committee:

Similarly, Dennis Nally, CEO of PricewaterhouseCoopers, told the Wall Street Journal that audit firms are pushing for a limitation on legal liability because, “There is a concern that without some form of liability relief over time you run the risk that if one of [the Big 4 auditing] firms ultimately fails, there is a catastrophic loss that can't be dealt with” (Reilly 2007).

This paper summarizes a recent study (“The Impact of Audit Penalty Distributions on the Detection and Frequency of Fraudulent Reporting,” Burton et al. 2011) that reports the results of two experiments that show how the system that determines auditors' legal liabilities when fraud goes undetected can influence both auditor decisions to detect fraud and auditee decisions to commit fraud. Interestingly, the research suggests that a relatively simple change to today's legal system could both eliminate the possibility of an extreme legal liability that would cause an audit firm to fail and help achieve the goal of increasing auditors' efforts to detect fraud.

Auditors must plan the nature, timing, and extent of their effort to search for fraud in an auditee's financial statements. The study reviewed in this summary shows that, although auditors' decisions to search for fraud are affected by the average penalty for failing to detect fraud, they also are affected by other properties of the penalty, including the possibility of an extremely large legal liability (hereafter referred to as a skewed penalty system). Perhaps counter to conventional wisdom, the study shows that a system that automatically penalizes auditors when they fail to detect fraud and also eliminates the possibility of an extremely large legal liability for failing to detect fraud (hereafter referred to as a deterministic penalty system) would increase auditors' effort to detect fraud. This finding holds when the average penalty between the two systems is held constant. The study also shows that these changes to the legal system would yield a decrease in auditees' propensity to commit fraud.

The study explored the amount of effort auditors expend to detect fraud for a hypothetical deterministic penalty system relative to a skewed penalty system. A deterministic penalty system is defined as one that automatically imposes a predetermined auditor legal liability whenever fraud goes undetected. In other words, under a deterministic penalty system, auditors automatically would be liable to shareholders, with the amount of the liability being predetermined, whenever material fraud is discovered. For example, if a fraud was found in an audited financial statement, the auditor could be required to pay a penalty based on a percentage of the outstanding market value of equity as of the date of the audit opinion.

In contrast, a skewed penalty system is one in which auditors can be liable for extreme, potentially unlimited, damage awards because of various factors, such as a biased jury. Using an experimental economics research method, the study reviewed in this summary examined auditor and auditee behavior under these two penalty systems, while holding constant the average penalties that auditors incur between the two systems.

The study predicted and found that auditors facing a skewed penalty system will expend less effort to detect fraud than auditors facing a deterministic penalty system, even though the average penalty that auditors faced was identical. The theory used to explain this behavior relies on the assumption that competition in the audit environment pressures auditors to cut costs. As auditors consider their competitors' actions and the likelihood that competitors will discount the risk of extreme claims, the pressure to cut audit costs increases. Consequently, auditors are motivated to underweight the extreme negative penalties that exist in a skewed penalty system and focus on the more salient and more frequent deterministic penalty, which is not as extreme. In other words, auditors will be motivated to discount or ignore the rare, extreme penalty that could result from factors in the current legal environment, such as a biased jury. In that case, auditors, when planning audits, will consider the typical low penalty cost under the skewed penalty system and perform less work than they might under the higher penalties under the deterministic penalty system, even when the average penalties are the same. Overall, the study shows that, because auditors are not as likely to discount a deterministic penalty, auditors under a deterministic penalty system will expend more audit effort than those under a skewed penalty system.

In addition, the study predicted and found that auditees were less likely to commit fraud under a deterministic audit penalty system when compared to a skewed audit penalty system, which is the system that exists in today's legal environment. The theory used to predict and explain this outcome is based on behavioral game theory, which suggests that client managers who are committing fraud will consider and respond to their personal incentives as well as the incentives that their auditors face. As the auditors increase their effort under a deterministic penalty system, managers will respond by decreasing the rate of fraud because there is a greater likelihood that the auditors will discover the fraud.

This study has implications for public policy. As policymakers grapple with the optimal mix of incentives, penalties, and oversight for the audit profession, it would be easy to interpret prior research as suggesting that caps on audit damages or penalties will result in less audit effort.¹ However, this study suggests that implementing a system of auditor legal liability that eliminates extreme legal penalties and creates automatic penalties for failing to detect a material fraud, and issuing an inappropriate audit report, is likely to increase auditors' efforts to detect fraud and decrease the occurrence of fraud.

The remainder of this article is organized as follows: first, we discuss the research method used in the study. In the next section, we present the study's results. Finally, we offer concluding thoughts.

The study uses an experimental economics approach in which the key economic conditions of a real-world setting (i.e., an audit market) are reproduced in an otherwise abstract setting. The abstract setting is then used to explore how changes to policies in the real world might affect economic decisions. This experimental approach allows researchers to control the many other complicating (but less interesting) factors that occur in natural settings, while observing the direct effects of potential policy changes. As noted by Kachelmeier and King (2002), “[R]eal world data can only inform us of reactions to policies that already exist.” However, an experimental approach allows researchers to provide “insights of likely reactions to policies that could exist” (Kachelmeier and King 2002, 219). Consequently, experiments are ideal for determining how changing one aspect of the real world (e.g., moving from a skewed penalty system to a deterministic penalty system) will impact auditor effort.²

Eighty-five students from a major university participated in the experiment. Participants were randomly assigned to roles within one of four experimental settings: (1) auditee in a deterministic penalty setting; (2) auditee in a skewed penalty setting; (3) auditor in a deterministic penalty setting; and (4) auditor in a skewed penalty setting. Auditors and auditees were randomly and anonymously paired for each round. The pairings changed each time an audit was performed.

Auditees decided whether they would report the true value of a portfolio of assets or fraudulently inflate the portfolio's true value. Auditees earned money based on their reported earnings, which provided an incentive to fraudulently overstate true value. When auditees inflated their reports and were not caught, they earned more money. Auditors received a fixed amount of audit revenue for each audit, which was reduced by (1) the cost of sampling they performed, and (2) any penalty imposed for reaching an incorrect conclusion (such as accepting an inflated value for the assets or rejecting a true value of the assets). Auditors chose a sample size that ranged from 0 to 30 of the 100 assets in the portfolio. Auditors understood that, by increasing their sample sizes, they could have a better estimate of the portfolio's true value, but that larger sample sizes cost them more than smaller sample sizes and, consequently, reduced their profit.

An auditor's sampling decision could yield four possible outcomes with respect to the auditee's reported asset value: (1) correct agreement; (2) correct disagreement; (3) incorrect disagreement; and (4) incorrect agreement. For both “incorrect” outcomes, auditors incurred a penalty, which is described in more detail below. An incorrect agreement is analogous to auditors not detecting fraud and providing a clean opinion on fraudulent financial statements (i.e., a type II error); an incorrect disagreement is analogous to auditors providing a qualified opinion on materially correct financial statements (i.e., a type I error). The penalty for a type I error (i.e., incorrect disagreement) was relatively small and was held constant across participants. The size of the penalty for a type II error (i.e., incorrect agreement, or an undetected fraud) was manipulated between auditor participants and was determined based on the penalty system to which the auditor was assigned. If the auditor was in the skewed penalty system, a software algorithm randomly calculated the amount of the penalty so that 98 percent of the time a small penalty was incurred and 2 percent of the time a very large penalty was incurred. In the deterministic penalty system, the auditor received a fixed penalty for each incorrect agreement, with the amount of the fixed penalty being set equal to the average penalty incurred in the skewed system. Thus, the expected, or average, values of the penalties in both systems were equivalent.³ Both auditees and auditors observed the outcome of the audit and the penalties incurred by the auditors.

Participants indicated the extent to which they agreed with the following question: “Each time my sample results led me to incorrectly agree with an inflated report, I expected a substantial penalty.” Responses suggested that participants in the skewed setting perceived a substantial penalty to be less likely (2.9 on a scale from 1 to 7) than participants in the deterministic setting (5.9), even though the expected value of the penalty was constant across settings. In other words, participants in the skewed setting appeared to discount the likelihood of incurring a large penalty if they failed to detect a fraud.⁴

The study's authors anticipated that because participants in the skewed setting were not as concerned about a large penalty, these participants would sample less than participants in the deterministic setting. As shown in Figure 1, the study's results reflect this pattern of results, with participants in the skewed setting sampling, on average, 10 of the 30 items, while participants in the deterministic setting sampled an average of 16 items. These results suggest that participants facing a skewed penalty system sample less than participants facing a deterministic penalty system, in part because participants under a skewed penalty system underestimate the average penalty for missing a fraud relative to participants operating under a deterministic penalty system.

As mentioned previously, the study also predicted that auditees' reporting choices would be affected by the penalties (or incentives) that their auditors faced and, as such, a deterministic penalty system would yield both greater audit sample sizes and less propensity for auditees to commit fraud. To test this prediction, the authors calculated the percentage of rounds in which each auditee inflated their reports. As depicted in Figure 2, auditees whose auditors faced a skewed penalty system inflated their reports 77 percent of the time, while auditees whose auditors faced a deterministic penalty system inflated their reports only 44 percent of the time. This result was equally as strong in the early rounds as it was in the later rounds of the experiment, suggesting that auditees quickly anticipated how their auditors would respond to their respective penalty systems, and adapted accordingly and consistently.

This study provides insights into a potential misunderstanding regarding the impact of auditors' legal liabilities on audit effort. Some people may think that audit effort would be lower under a deterministic penalty system as opposed to a skewed penalty system because the potential penalties in the deterministic system are not “extreme enough.” In other words, some people believe that, if auditors no longer faced the potential for a rare but substantial penalty, then auditors would not be as vigilant. This study suggests that this notion, likely, is erroneous. Specifically, the study shows that, if the current legal liability system were changed to a system in which auditors incurred an automatic penalty each time that undetected material fraud was later discovered, then auditors would be more vigilant in their search for fraud and auditees would be less likely to commit fraud. This finding is expected to hold even if the expected liabilities that auditors incur for failing to detect fraud are no more than the expected liabilities incurred under the current system. In other words, because the expected penalties would remain the same under both systems, total damage awards provided to the victims of fraud would be held constant, assuming that the number of instances in which auditors fail to detect fraud is the same. However, the study suggests that, because auditors would do more work, auditees would commit less fraud, and these two factors would lead to fewer victims of fraud and lower costs associated with legal liabilities related to fraud.

Of course, this study is limited by its methodology, and the authors recommend additional research on the effects of deterministic penalties. In addition to how auditors and auditees respond to changes in audit penalty systems, investor behavior also may be impacted. While one implication of the study's findings is that the cost of capital could be reduced, because the incidences of undetected fraud would be reduced, this study does not investigate these important effects. Consequently, we encourage additional studies that investigate these issues, and encourage policymakers to support and review such research prior to adopting a deterministic penalty for auditors.