Discussion of “Short Interest as a Signal of Audit Risk”*

In their paper "Short Interest as a Signal of Audit Risk" in this issue Cassell, Drake, and Rasmussen (hereafter CDR) aim at investigating whether short interest provides a signal of the degree of audit risk, where audit risk is defined as "the risk that the auditor issues an unqualified opinion when the client's financial statements contain material misstatements" (CDR: 1278). They do so by analyzing the association between audit fees and short interest, which they hypothesize and document to be positive. From the analyses presented in the paper it seems that the positive association between audit fees and short interest only holds in the Sarbanes-Oxley (SOX) era. Based on their findings, CDR conclude that "the auditors' attention to risk factors increased after 2002, which implies that efforts by regulators and standard-setters to improve auditors' client risk assessments have been successful" (CDR: 1295). Throughout the paper the authors seem to assume that the higher audit fees for audit clients with high short interest are attributable to a reaction by the auditor based on a risk signal embedded in short interest. CDR offer a fresh perspective on the association between audit risk and audit pricing. Looking for new determinants beyond the traditional measures based on research findings outside the auditing literature is a valuable contribution to that literature. The paper is also well-motivated from the viewpoint of these other literatures. My major critique, however, is that the paper is less well-integrated in the extant auditing literature. This leaves the reader with the rather fundamental question of why the documented positive association between short interest and audit fees is so strong (note that the correlation between Rank of Short Interest and Log of Audit Fees is equal to 0.37; see CDR: Table 3). If such a significant (both economically and statistically) determinant of audit fees has not been picked up in prior studies, what has been ignored then in prior research? One explanation could be that it is a rather recent phenomenon. Another that short interest is correlated with other drivers of effort such as client size. In the remainder of my discussion I will focus on five issues. In the next section my discussion will address the question of why short interest could or could not be a determinant of audit risk. In section 3, I will offer my view on likely drivers of the positive association between short interest and audit fees based on experimental evidence from prior auditing research. I then briefly discuss in section 4 why the test variable, Rank of Short Interest, might not be an ideal measure. Subsequently I elaborate in section 5 on the observation that short interest is only positively associated with short interest in the SOX era. Finally, I question in section 6 whether the evidence provided by CDR really shows that auditors perceive short interest as a signal of audit risk. Section 7 concludes my discussion. The motivation for this paper is based on evidence in prior research outside the auditing literature that short sellers are sophisticated users of financial information. In particular, this evidence pertains to studies documenting that short interest provides incremental information about future firm performance and returns (see, for example, Boehmer, Jones, and Zhang 2008; Boehmer, Huszar, and Jordan 2010; Drake, Rees, and Swanson 2011), and that short sellers target firms with suspect financial reporting (see, e.g., Dechow, Sloan, and Sweeney 1996; Desai, Krishnamurthy, and Venkataraman 2006). This offers without a doubt a fresh angle to the concept of audit risk, but it is not clear from the motivation in the paper how the (supposedly) incremental information embedded in (publicly available) short interest affects the audit process. The authors use the rather "convenient" argument that short interest serves as a signal of audit risk to motivate "a" link between short interest and auditing. Furthermore, they define audit risk as "the risk that the auditor issues an unqualified opinion when the client's financial statements contain material misstatements" (CDR: 1299), that is, the probability of Type II error. Hence, they seem to perceive increased audit risk as somewhat equivalent to an increased likelihood of financial statement errors and irregularities. A rather obvious related question is whether the increased probability of questionable financial reporting in firms with high short interest originates from inherent risk factors (for example, performance pressure or financial distress) and/or control risk factors (for example, inadequate corporate governance and/or internal controls). A description of the conceptual link between short interest and the components of the audit risk model is essential to address a key question in this paper, that is, whether short sellers and auditors respond to common information risk (correlated information sets) or whether auditors use short interest as a source of information. If the inherent and/or control risk factors are systematically different (more risky) in firms with high short interest, short interest could be picking up these effects. A related observation is that CDR motivate the link between audit fees and short interest solely from an audit production perspective, thereby ignoring potential audit demand factors. It is presumed that audit clients are willing to spend more resources on auditing and pay higher audit fees, ceteris paribus, when there is high short interest. However, short interest is publicly available information. If there is informational value in short selling, aren't short interest and auditing substitutes from an information demand perspective? Wouldn't we then expect to observe a lower rather than a higher audit fee, ceteris paribus? Why then would a firm with (relatively) high short interest be willing to pay a higher audit fee, or stated differently, why would this lead to an increase in audit demand, ceteris paribus? One explanation could be that firms with high short interest compensate for the embedded negative signal to the market by demanding more and/or higher-quality audit services. Finally, if auditors would indeed act upon short interest as a signal of audit risk and increase their audit effort on an engagement, why then do prior studies find that financial information — which is audited — in firms with higher short interest is of lesser quality? If auditors act on short interest as a risk signal, wouldn't we then expect that audited financial statements of firms with large short interest be at least of the same (or even higher) quality, on average, since auditors dedicated more effort to such statements? And since this does not seem to be the case, wouldn't we then expect a higher incidence of going-concern opinions in companies with high short interest?1 Note the latter does not hold either for the data used in the study, as the correlation between Rank of Short Interest and Going Concern in Table 3 of the study is negative. To motivate a positive association between short interest and audit fees, CDR refer to the seminal work on audit pricing in Simunic 1980 as a theoretical basis. They do not establish a link with other prior auditing literature that has addressed how and why audit risk affects audit effort and fees. However, the more interesting question in my view is not so much whether there is a positive association between short interest and audit fees, but why there is one. Addressing the latter question also helps to rule out some alternative explanations for what is observed. Basically, Simunic (1980) argues that in a competitive audit market audit fees equal marginal cost, which results in a production (cost) oriented audit fee model. He further argues that an auditor's cost function consists of two components: a resource cost component which is increasing in the level of auditor effort and an expected future loss component. CDR simply claim short interest to be a signal of audit risk, but do not offer an explanation whether the conjectured positive association is attributable to more audit effort or a higher litigation risk premium. Though both increased audit effort and a litigation risk premium would result in a positive association between short interest and audit fees, it is an important distinction as it could lead to different audit outcomes and hence different levels of audit assurance supplied. There is experimental evidence about the associations between audit-related risks, audit effort, and audit fees which could be useful in conjecturing some of the effects. Let's turn to only a few examples. Pratt and Stice (1994) document from a field experiment that a client's overall financial condition is the primary consideration in the auditor's assessment of litigation risk and corresponding recommendations for the audit plan and audit fees. As client firms with high short interest are typically risky companies with bad financial performance (see, e.g., Drake et al. 2011), short interest could pick up a higher probability of future bankruptcy (increased litigation risk). To rule out this explanation, a control variable for bankruptcy risk based on the Altman Z-score, as in Griffin and Lont 2007, seems warranted. In another experimental study, Houston, Peters, and Pratt (1999) identify conditions under which the audit risk model does and does not describe audit investment and pricing. The results indicate that the audit risk model dominates auditor effort and pricing decisions when the likelihood of an error is high, whereas business risk considerations dominate when the likelihood of an irregularity is high. In the latter case the audit fee also contains a risk premium. Given prior evidence that short sellers anticipate corrective disclosures that ultimately lead to class action suits (Griffin 2003), it is reasonable to believe that short interest is associated with increased business risk and a higher risk premium (on top of more effort), rather than (only) with increased audit risk and effort. Teasing out business risk versus audit effort effects associated with short interest audit risk seems a valuable avenue for future research. A concern could be that the positive association between short interest (risk) and audit fees is picking up a client size effect. CDR use Rank of Short Interest as the test variable in the audit fee model. This variable is computed as the annual decile rank of the short interest ratio, scaled to range between zero and one. Using the rank instead of the raw short interest observation is done to correct for skewness in the short interest distribution. However, ranking this variable (per year) also results in a different (increased) correlation between short interest and client size. From the correlation matrix in Table 3 we see that Rank of Short Interest has a high correlation with Log of Audit Fees (0.37) but also with Log of Assets (0.37). The table unfortunately does not report the correlation between raw short interest and client size, but it is very likely that this correlation would be much lower.2 Why is this important? Audit fees are typically very highly correlated with client size, as client size triggers more audit effort. Rank of Short Interest could be picking up a nonlinear client size effect.3 Note that some prior audit fee studies illustrate that the association between audit fees and client size is nonlinear. For example, Carson, Fargher, Simon, and Taylor (2004) and Carson and Fargher (2007) illustrate that the interaction between Log of Assets and the size partition to which a client belongs is significantly associated with audit fees. They argue that fee models that fail to account for this effect could be misspecified. In general, future research is needed about the client size effects in audit pricing. CDR also hypothesize that the association between short interest and audit fees becomes stronger after 2002, in the SOX era. The motivation provided for this additional hypothesis is based on new regulations aimed at curbing corporate fraud, such as SOX in July 2002 and SAS No. 99 issued by the American Institute of Certified Public Accountants. From the results in columns 2 and 3 in Table 5 and footnote 19 we see, however, that the documented positive association between short interest and audit fees is in fact driven by the observations from the SOX era. This seems to disconfirm Hypothesis 1 which predicts an overall positive short interest effect and triggers the question of why the effect is only picked up after 2002. Is it possible that auditors' changed focus on short interest as an audit risk signal is attributable to changes in the audit process in the SOX era? Or does Post2002 × Rank of Short Interest Ratio in Table 5 pick up (at least part of) the additional audit cost associated with an increase in audit hours to comply with SOX Section 404, instead of auditors starting to see short interest as a signal of audit risk after 2002? It is clear that SOX changed the overall scope, quality, and cost of an audit substantially by requiring an integrated audit of company financial statements and internal controls. In addition, SOX also changed the oversight of the auditing profession by creating the Public Company Accounting Oversight Board. Griffin and Lont (2007) document that audit fee increases following SOX can be explained both by increased risk and audit effort. Future research about changes in the audit process and risk assessment practices of audit firms in the SOX era would be useful from this perspective. Finally, a key question in the paper is whether auditors use short interest as a risk signal, or whether auditors and short sellers use a correlated underlying information set. This is a very interesting but also extremely challenging question. CDR try to address it by running a changes analysis examining whether changes in audit fees are associated with past versus future changes in short interest. They find that past changes in short interest are positively associated with changes in audit fees, but future changes of short interest are not. More tests seem warranted before one could conclude that there is a causal link between short interest and audit decision making. Various alternative research designs are possible to address this question. A simple test within the context of the design adopted by CDR could be to estimate an audit fee regression model including the same publicly available underlying fundamental signals as in Drake et al. 2011 (and some controls of course) and see whether this model outperforms the model tested in CDR which includes the Rank of Short Interest variable. If it does, it would be hard to conclude that auditors act on short interest information. The question opens many possibilities for future research and various other designs (beyond the audit fee literature) are thinkable to test whether auditors use short interest as a signal of audit risk. One approach could be to test the association between short interest and audit outcomes. As CDR define audit risk as Type II error risk, a direct test of whether short interest is a signal of audit risk could be to test whether there is indeed a positive association between short interest and type II errors by auditors. Here the prediction would be that short interest is positively associated with Type II errors. CDR address an interesting yet challenging research question and offer the first study examining a relationship between short interest and auditing. As with many studies in empirical accounting research, it is difficult to establish a causal link. Therefore it remains unclear whether auditors indeed directly rely on short interest in their assessment of audit risk, or whether short sellers and auditors use correlated information sets. Future research can build on the findings in CDR and examine how and why CDR observe a positive association between short interest and audit fees in the Sarbanes-Oxley era. A few interesting related questions include: Do firms with higher short interest demand higher quality auditors? Is short interest associated with higher inherent and/or control risk factors, and what are these? Is the higher audit fee associated with short interest the result of more audit effort and/or a short interest litigation risk premium? Did changes in the audit process in the Sarbanes-Oxley era trigger a role for short interest in audit decision making, and if so which changes?

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.