Do Auditors and Audit Committees Lower Fraud Risk by Constraining Inconsistencies between Financial and Nonfinancial Measures? Free

This article summarizes a recent study titled “Do Auditors and Audit Committees Lower Fraud Risk by Constraining Inconsistencies between Financial and Nonfinancial Measures?” (Brazel and Schmidt 2018). Specifically, I review the study's motivation, research method, and primary results. I then discuss several of the study's more significant implications. Most notably, the study raises the possibility that audit committee (AC) chairs with industry expertise could be quicker to rationalize that fraud risks can exist without posing significant risks to financial reporting quality.

Professional standard-setters, regulators, and academic researchers have discussed the potential for nonfinancial measures (NFMs) to be a powerful and independent benchmark for evaluating the validity of financial statement numbers (PCAOB 2004; Brazel, Jones, and Zimbelman 2009; Messier, Glover, and Prawitt 2012). Revenue-related NFMs disclosed in companies' 10-K filings (e.g., number of retail stores, patents, employee headcount) provide “difficult to manipulate” measures of economic activity (Brazel et al. 2009). Prior research finds that companies committing fraudulent financial reporting (hereafter, fraud) often exhibit large inconsistencies between their revenue growth and growth in revenue-related NFMs (Brazel et al. 2009; Dechow, Ge, Larson, and Sloan 2011).¹ Thus, large inconsistencies between audited financial information and related NFMs can be considered a red flag indicative of high fraud risk (Brazel et al. 2009). For parsimonious purposes, Brazel and Schmidt (2018) (hereafter, “we” or “our”) refer to the difference between revenue and NFM growth as “DIFF,” with a “large” DIFF being indicative of high fraud risk (e.g., revenue growth outpacing NFM growth by more than 20 percent).

The auditor is responsible for providing reasonable assurance that financial statements are free from material misstatements resulting from error or fraud. However, recent research has called into question the auditor's ability to detect fraud (e.g., Dyck, Morse, and Zingales 2010). In addition, evidence exists of an “expectations gap” in which financial statement users expect more from an audit than what the audit profession believes it can provide vis-à-vis fraud detection (e.g., Mock et al. 2013). Thus, there is a need to better understand the auditor's ability to mitigate fraud risk.

The 2013 COSO Framework highlights the importance of fraud risk assessment for those charged with corporate governance (see Principle 8).² The audit committee (AC) chair, as the “CEO of the audit committee” (Ernst & Young 2011), plays an important corporate governance role by overseeing the accounting, financial reporting, and auditing processes of companies (U.S. House of Representatives 2002). Thus, AC chairs should engage in mitigating fraud risk. However, Beasley, Carcello, Hermanson, and Neal (2009) indicate that AC members try to distance themselves from their responsibilities to assess fraud risk, suggesting that AC chairs are not apt to reduce fraud risk.

Because not all auditors or AC chairs are equally adept at identifying and reducing fraud risk via the use of NFMs (Brazel et al. 2009; Beasley et al. 2009; Brazel, Jones, and Prawitt 2014), the objective of our study was to determine whether auditors and AC chairs with certain attributes are more likely to constrain fraud risk. Specifically, we examine whether greater effort, client-specific tenure, and industry expertise on the part of auditors and AC chairs reduce the likelihood that a company's audited financial statements exhibit a “large” DIFF (i.e., high fraud risk).³

We began by utilizing a web-based tool to identify quantitative, revenue-related NFMs in 691 companies'10-K filings for two consecutive years during the 2007–2009 time period. As shown in Table 1, we collected a total of 4,138 NFMs across the 691 companies (an average of 6 NFMs per company). The largest categories are “Employees,” “Facilities,” and “Products and Inventory,” which represent 16.7 percent, 16.3 percent, and 12.6 percent of the NFMs collected, respectively. The average change in NFMs between the two years is then computed for each company and compared to the change in the company's reported revenue across the same two years (what is called DIFF in the study). In Figure 1, we show the distribution of DIFF for our sample of companies. Figure 1 illustrates that financial growth is typically consistent with NFM growth as DIFF for our sample is normally distributed and centered at approximately 0. For the entire sample, DIFF has a mean of 0 percent and a median value of about 1 percent.

We identified revenue growth exceeding the average change in revenue-related NFMs by more than 20 percent as a LARGEDIFF indicative of high fraud risk. We used 20 percent as our threshold because the prior literature documents that fraud risk is high when growth in revenue exceeds the average growth in related NFMs by 20 percent or more (see Brazel et al. [2009, Table 3] for DIFF percentages). As illustrated in Appendix A, LARGEDIFF is then coded 1 if the company's growth in revenues outpaces the average growth in NFMs from year t−1 to year t (t = our latest sample year) by more than 20 percent and coded 0 otherwise. Approximately 11 percent of our sample consisted of companies exhibiting a LARGEDIFF (74/691 companies), where growth in revenues outpaces the average growth in NFMs by greater than 20 percent.

Our first auditor-related variable of interest was AuditorEffort, which was equal to the natural logarithm of total audit fees billed in the current year. We used audit fees as our measure of auditor effort because actual audit effort is not observable and numerous studies document that audit fees are higher when auditors expend more effort to address risks (e.g., Pratt and Stice 1994). Our second auditor-related independent variable was AuditorTenure, which was equal to the number of consecutive years since 1974 the company is reported by Compustat to have retained the auditor as of the end of year t. We used 1974 as it is the year that Compustat began identifying auditors. Our final auditor-related independent variable was AuditorIndExpert, which was an indicator variable set equal to 1 if the auditor had an annual national market share greater than 30 percent within the two-digit Standard Industrial Classification (SIC) industry in year t and set equal to 0 otherwise (Knechel, Naiker, and Pacheco 2007; Reichelt and Wang 2010).

With respect to AC chair characteristics, our first variable of interest was ACEffort, which was equal to the total number of AC meetings that occurred within year t according to the company's annual proxy statement (Farber 2005). Next, we used the BoardEx database to capture the length of the AC chair's tenure on the board and defined ACCTenure, our second variable of interest, as the total number of years the AC chair had been serving on the board of directors as of the end of year t.⁴ Our third AC chair variable of interest was ACCIndExpert, which was an indicator variable set equal to 1 if the AC chair has current or previous work experience within the same two-digit industry SIC code of the company and set equal to 0 otherwise (Cohen, Hoitash, Krishnamoorthy, and Wright 2014).⁵

The study uses logistic regression models to investigate whether companies with greater auditor and AC chair effort, tenure, and industry expertise are less likely to exhibit a “large” DIFF (i.e., high fraud risk). As described in Table 2, we find that certain auditor and AC chair characteristics are associated with lower fraud risk/constrain large inconsistencies between reported revenue growth and related NFMs. In particular, we observe that as auditor tenure and industry expertise increase, the audited financial statements are less likely to exhibit a LARGEDIFF.

Our analyses also demonstrate that as AC chair tenure increases, the audited financial statements are less likely to exhibit a LARGEDIFF. Surprisingly, we observe that companies engaging an AC chair with industry experience are more likely to exhibit a LARGEDIFF compared to companies without an industry expert AC chair. To illustrate this counter-intuitive result, we observe that 26 percent of companies exhibiting a LARGEDIFF have an AC chair with industry expertise, whereas only 12 percent of companies not exhibiting a LARGEDIFF have an AC chair with industry expertise.

Several factors could account for this observed relation. For example, AC chairs with industry expertise are likely to have multiple board of director appointments and be too busy to monitor fraud risk at any one company (Lublin 2016). Alternatively, AC chairs with industry expertise could have strong connections with management or receive equity incentives that would make them more willing to side with management. As such, we measure the AC chair's busyness as the number of board appointments, connectedness using Muckety.com scores, and equity incentives following Campbell, Hansen, Simon, and Smith (2015). We observe that the relation between AC chair industry expertise and our measure of fraud risk is stronger in settings where the AC chair is less busy or more connected with management. In additional tests, we find that the positive relation is driven by instances where the AC chair has industry expertise, but the auditor does not. While we did not ascertain the precise mechanism behind the unexpected AC chair industry expertise result, our evidence suggests that AC chairs sometimes have the incentive, time, and power to influence the financial reporting process such that fraud risk is not constrained. As such, our results indicate that not all forms of AC expertise aid in constraining fraud risk as it is measured in our study. Additional research studies employing alternative research methods are needed to further ascertain if and why AC chair industry expertise does not constrain fraud risk.

Brazel and Schmidt (2018) contributes to the understanding of the fraud audit process in several key ways. First, while prior research suggests that auditors are not adept at reducing fraud risk by constraining large DIFFs (Brazel et al. 2009; Brazel et al. 2014), we provide empirical evidence that certain auditors—those exhibiting longer tenure and greater industry expertise—are able to reduce inconsistencies in audited financial statements. Our results suggest that these auditors can be viewed as a preventative control that audit committees can use as a tool to reduce fraud risk. Second, our results related to tenure indicate that mandatory audit firm rotation may not be the best means of mitigating fraud risk.⁶ Third, we observe very little evidence of a relation between audit fees (effort) and our measure of fraud risk. We therefore illustrate that auditors were not compensated for increased fraud risk during our sample period that coincided with the global financial crisis. Finally, while prior research reports that AC financial and industry expertise are associated with higher quality financial statements (e.g., Abbott, Parker, and Peters 2004; Cohen et al. 2014), our study raises the possibility that audit AC chairs with industry expertise could be quicker to rationalize that fraud risks can exist without posing significant risks to financial reporting quality.

Additional analyses examined in Brazel and Schmidt (2018) suggest that AC chairs siding with management over the auditor may be conditional on AC chairs with industry expertise having more time available and/or connections with management. We also identify that expertise gaps between the AC chair and the auditor may be detrimental to the mitigation of fraud risk. Overall, our results indicate that additional research is needed to determine if and when AC industry expertise is problematic.

The formation of our sample requires the collection of revenue-related NFMs disclosed in companies' 10-Ks. We collected the appropriate NFMs using a website developed for the Financial Industry Regulatory Authority. For each company in our sample of 691 companies, we used the average change in NFMs (%AVGCHGNFM) to proxy for the general direction of changes in company-specific NFMs. The calculation of DIFF then subtracts %AVGCHGNFM from the percentage change in revenue (%CHG_REV). LARGEDIFF is then coded 1 if the growth in revenues outpaces the average growth in NFMs from year t−1 to year t (t = our latest sample year) by more than 20 percent and 0 otherwise. Below, we provide calculations for %AVGCHGNFM, DIFF, and LARGEDIFF from data for Panera Bread Company, a company in our sample where LARGEDIFF = 0. We also provide the same data for Company X, a company in our sample where LARGEDIFF = 1. Given that we propose that a LARGEDIFF = 1 indicates higher fraud risk, we have disguised the name of this company. However, after 2009, Company X was investigated by the SEC related to its revenue recognition practices. A class action lawsuit was also filed. Company X has since restated its financial statements from 2007–2010, and the founder and chairman of Company X has been fired.