In this paper, we summarize our recent research into the association between abnormal audit report lags (the length of time between the fiscal year-end and the audit report date) and the likelihood of future restatements (Blankley, Hurtt, and MacGregor 2014). We find that as the audit report lag increases, the likelihood of a future restatement increases, and this dynamic appears to be associated with the effect of time pressure on the auditor. We will first summarize the motivation for the study, then discuss the empirical methods used, explain our results, and conclude with a discussion of the practical applications to practitioners.
One of the more interesting questions in auditing research involves how to interpret the length of time it takes to complete an audit. Any professional task, done well, will take time to complete, but we often do not know how long a high-quality job should take, or at what point the length of time begins to signal a problem. This ambiguity regarding how to interpret an “appropriate” length of time is particularly acute in a task as complex as an audit, since there are so many variables that can affect the audit report lag (the length of time between the end of the client's fiscal year and the audit report date) including the complexity of the audit, the nature of the client, and the expertise of the auditor. In our recent study, “The Relationship between Audit Report Lags and Future Restatements” (Blankley, Hurtt, and MacGregor 2014), we investigated whether unusual delays in the audit during a given year were statistically associated with a future restatement of that year's financial statements. We found that there is, in fact, an association, and that association is positive. In other words, when an audit takes an unusually long time to complete, the likelihood of the firm restating its financial statements in the future increases. We also investigated what factors might influence this relationship and found that time pressure appears to undermine the quality of the additional audit effort. We summarize our findings below, briefly discussing our motivation for the study, the research methods, and results of our statistical tests, and then conclude by discussing the implications for managers, auditors, audit committees, and regulators.
Previous academic research into audit report lags (ARLs) finds that they are positively correlated with audit effort.1 An important corollary to this finding is the intuitive assumption that greater audit effort leads to higher audit quality. In fact, prior research also suggests this relationship (see, for example, Knechel and Payne 2001; Knechel, Rouse, and Schelleman 2009; Tanyi, Raghunandan, and Barua 2010). Knechel et al. (2009) proposed an audit production model linking audit activity, audit evidence, and greater levels of assurance as follows:
The [audit production] model assumes labor inputs are used in the audit process to perform audit activities that result in the generation of audit evidence. There are different types of audit activities that may be performed, but they all lead to the generation of audit evidence. Thus, the greater the extent of evidence-gathering activities, the greater the likelihood that an auditor reaches the correct conclusion about whether the financial reports are misstated, leading to a higher achieved level of assurance.
The notion that greater audit effort leads to higher audit quality is intuitive, and has some support in the empirical academic literature. For example, recent work by Bryant-Kutcher, Peng, and Weber (2013) reports that when the SEC modified its 10-K filing deadlines, it reduced auditors' ability to exert additional effort or extend the audit, increased time pressure on the auditors, and ultimately degraded audit quality. Their results are consistent with the notion that longer audits are associated with higher quality. While this claim may frequently be true, it is not clear that this is always the case, especially as the length of the audit increases beyond what may be considered normal. Audits that run unusually long may, in fact, signal a problem. The fact that the market reaction to late filings is negative (Alford, Jones, and Zmijewski 1994; Li and Ramesh 2009; Bartov, DeFond, and Konchitchki 2011), and that the probability of a late filing is driven by audit completion time (Bryant-Kutcher, Peng, and Zvinakis 2007), suggests that lengthy audits do not necessarily reflect good news to the market. In this context, longer audits and additional audit effort may not signal higher audit quality. It is unclear what, if anything, the length of the audit might imply about the quality of the audit.
Because the interpretation of the audit report lag (and, by extension, the efficacy of the audit effort inherent in the lag) is ambiguous, we were interested, first, to find out whether unusual delays in completing the audit were associated with audit quality measures. If they were, then we were also interested to identify factors that might influence the association. To examine the question, we derived the “abnormal” or unusual audit report lag for each sample company, each year, and then tested whether this abnormal or unexpected portion of the audit report lag was associated with an increase in the likelihood of a future restatement associated with that year. This line of reasoning led to our first hypothesis, stated in the null form:
There is no association between the current year's abnormal audit report lags and a subsequent restatement of that year's financial statements.
If unusually long audits reflect higher audit quality, then we would expect to observe a negative association between the abnormal ARL and the likelihood of a future restatement. On the other hand, if unusually long audits reflect issues that potentially reduce audit quality, then we would expect to observe a positive association between the abnormal ARL and the likelihood of a future restatement.
RESEARCH METHODS AND RESULTS
In order to test H1, we utilized a two-stage statistical analysis. In the first stage, we developed a robust regression model to estimate the abnormal or unexpected portion of the audit report lag. In this model we regressed each firm's normalized audit report lag against variables representing audit effort, audit complexity, audit risk, filing deadlines, relevant auditor characteristics, client capital structure, and the quality of the client's internal control system. These variables tend to explain a large portion of each firm's ARL, but they do not explain it all. The regression model also generates a statistical “residual” for each observation, which represents the portion of a firm's actual ARL that differs from the value of the ARL predicted by the statistical model after considering the effect of the other variables on the ARL. We then use these residuals to represent the unexpected or abnormal portion of the audit report lag.
In the second stage, we use a logistic regression model to assess the relationship between the unexpected audit report lag and the future restatement. In this model, the dependent variable is a binary variable that indicates whether the client eventually restates the current year's financial statements. The primary variable of interest in this stage was the unexpected audit report lag (the residual) generated from the first stage. In addition, we also included several independent variables to control for firm characteristics, such as firm size, leverage, free cash flows, and the trend in earnings, as well as for characteristics of the auditor, such as whether the auditor was a specialist in the client's industry, the length of tenure with the client, and other characteristics. We refer interested readers to Blankley et al. (2014) for exact specification of the two different models and definitions of the variables.
Results from both univariate statistical tests and linear regression models indicated that unexpected ARLs have a significant, positive association with future restatements. For example, the average unexpected audit report lag for firms that did not need to restate their financials in the future was −0.002, while the average unexpected ARL for firms that did restate that year's financial statements in the future was 0.024, which is a statistically significant difference in the averages across the two groups. The regression model confirmed this univariate result after controlling for a number of other influences associated with future restatements. The data indicated that the longer an audit runs beyond what might be considered normal, the greater the likelihood is that the firm will ultimately have to restate that year's financial statements. In other words, longer unexpected delays in the audit appear to correlate with an increase in the likelihood of a future restatement. Figure 1 graphically illustrates the association. The chart identifies the percentage of future restatements associated with each of six equal sextiles representing the distribution of audit report lags in our sample. The columns indicate the percent of firms in each sextile that ultimately restated their financial statements. The curved line represents a best-fit trend line of the percent of restatements for each sextile expressed as a second-order polynomial. The mean value for each sextile represents the average audit report lag for that sextile. For example, firms representing the lowest sixth of our distribution (i.e., the first sextile) had a mean audit report lag of 50.8 days. Firms in the sixth sextile had an average audit report lag 79.9 days, indicating that there is nearly a 30-day difference in audit report lags between the average firm in the first sextile compared to the average firm in the sixth sextile. We then calculated the percent of firms in each sextile that ultimately restated that year's financial statements. Again, looking at the two extremes, we observe that 6.69 percent of first-sextile firms ultimately restated their financials, while 10.6 percent of sixth-sextile firms had future restatements. Figure 1 indicates that the percentage of future restatement firms decreases slightly across the first three sextiles, but increases monotonically from the third through sixth sextile.
The chart indicates that additional time spent on the audit does, in fact, contribute to increased audit quality, but that the additional time yields diminishing returns that eventually reverse. As the length of time continues to increase, the risk of a future restatement increases as well. Intrigued by the nature of this association, we next investigated three possible explanations for it: economic bonding, lack of auditor expertise, and time pressure.
Alternative Potential Undermining Influences
We first examined the notion of economic bonding as a way to explain the positive association. In this case, if the auditor felt a strong economic bond with the client, then this dysfunction might be reflected in the observed association between unexpected lags and future restatements. Previous research has suggested that an auditor who feels a strong economic bond with the client may fail to exercise sufficient professional skepticism to interpret audit evidence correctly or judge the evidence impartially (Bazerman, Morgan, and Loewenstein 1997). When faced with a difficult or ambiguous audit issue, the auditor may expend additional effort yet fail to resolve the issue. If there is a strong economic bond between the auditor and client, then the auditor, faced with the loss in revenue if he or she resigns the engagement, as well as the costs to replace the client, may decide to continue the engagement and accept the client's proposed treatment of the issue. In this scenario, the perceived financial benefits to the auditor outweigh the perceived costs associated with the uncertain audit issue.
The second alternative explanation was based on auditor expertise. Since the auditor's expertise affects both the efficiency and effectiveness of the audit, the greater the auditor's expertise is, the more likely it is that the auditor will be able to resolve difficult or high-risk audit issues that arise during the course of the engagement. On the other hand, less knowledgeable auditors may work longer due to their relative inefficiency, but still be less likely to identify or adequately resolve difficult or problematic issues. In this scenario, auditors with lower expertise may experience much longer audits but still have an increased likelihood of a future restatement. If the auditor expertise explanation is true, then that might explain the positive association.
Finally, we investigated the effect time pressure has on the association between abnormal audit report lags and future restatements. Prior academic literature shows that when auditors are under duress from time pressure, audit quality suffers in a number of ways. It may affect the planned level of effort (Maksymov, Nelson, and Kinney 2012), impair auditors' sampling adequacy and processing accuracy (McDaniel 1990), undermine the effectiveness of audit techniques used by managers and partners (Agoglia, Brazel, Hatfield, and Jackson 2010), diminish professional skepticism (Braun 2000), or potentially reduce the effectiveness of review (Lambert, Jones, and Brazel 2011). In addition, with reduced filing deadlines imposed by the SEC and pressure from the client over both cost constraints and completing the audit as quickly as the previous year, it is likely that as the audit begins to take longer than normal, time pressure increases and results in outcomes similar to those listed above. Lopez and Peters's (2012) results suggest that the time pressure and fatigue brought on by heavy workload demands may lead to undesirable behaviors, including impaired professional judgment and accepting weak client explanations. In fact, the notion that time pressure could be undermining audit quality has become a focus of the PCAOB. On December 4, 2012, the PCAOB issued Staff Audit Practice Alert No. 10, Maintaining and Applying Professional Skepticism in Audits. In this alert, the board remarked:
Scheduling and workload demands can put pressure on partners and other engagement team members to complete their assignments too quickly, which might lead auditors to seek audit evidence that is easier to obtain rather than evidence that is more relevant and reliable, to obtain less evidence than is necessary, or to give undue weight to confirming evidence without adequately considering contrary evidence. (PCAOB 2012)
Under the time pressure scenario, as the audit runs longer than normal and the auditor begins to feel intense pressure to resolve any outstanding issues so the audit can be completed, he or she may make suboptimal judgments and/or decisions in response to the pressure. Thus, while the auditor spends more time and effort on the audit than normal, the quality of the work may be undermined by the auditors' dysfunctional response to the pressure, and audit quality suffers. In summary, if the unexpected portion of the audit report lag signals greater auditor effort, and auditor effort is positively associated with higher audit quality, then these unexpected delays would be inversely related to reporting problems. If, however, unexpected delays in the audit create or exacerbate time pressure on the auditors, then we could reasonably expect to observe a positive association between the unexpected portion of the audit report lag and future restatements.
Results of Tests of the Potential Undermining Influences
The results of our primary test of the association between unexpected ARLs and future restatements indicate that, as the audit increases beyond a certain point, the additional time is less effective and the likelihood of a future restatement rises. This result suggests that something undermines the additional time observable in the ARL, resulting in a reduction in the quality of the additional audit time. To investigate the three potential explanations—economic bonding, auditor expertise (or lack of expertise), and time pressure—we modified the logistic regression used in the initial test of the association discussed above.
In this phase of the study, we developed hypotheses for each of the undermining influences, and then modified the original statistical model to include variables representing each of the three undermining influences. In the first modification we tested for the influence of potential economic bonding by including variables for high and low audit fees and high and low audit report lags and the four interactions between them. For example, if the audit runs long and fees are high, so that the auditor does not want to resign from the engagement in spite of an unresolved audit issue, then we would expect to observe a significant coefficient on the interaction of high fees and high lag on future restatements. None of the variables supporting an economic bonding interpretation proved significant. While we do not rule out the possibility that economic bonding may potentially influence the relationship between unexpected ARLs and future restatements, we found no evidence that economic bonding was undermining the additional audit effort in our tests.2
The second modification involved including a variable in the model indicating whether the audit firm was considered a specialist in the client's industry and interacting that variable with the variable representing the abnormal ARL. We expected that if lack of industry expertise caused the audit to run long and ultimately be more likely to result in a future restatement, then the coefficient of the interaction would have a negative sign. In fact, the results were just the opposite. The interaction variable was significant with a positive sign. This result does not support the notion that lack of auditor expertise leads to longer audits associated with future reporting problems. In fact, we speculate that auditor expertise might attract clients with more difficult audit issues that take longer to resolve, in spite of the auditor's expertise in the industry, due to their ambiguity or complexity, and may therefore be associated with a greater likelihood of a future restatement.3
To test the effect of time pressure, we modified the model in two different ways. We first changed the variable for the abnormal ARL to represent the difference from the prior year's abnormal ARL. This variable then represents the unexpected time difference associated with the current year's audit compared to last year's audit. This variable proved highly significant with a positive sign, suggesting that firms whose unexpected audit report lag increased relative to the previous year were significantly more likely to have a future restatement than firms whose unexpected audit report lags did not increase. Given that the variable captures the increase (or decrease) in audit time not explained by changes in audit risk, complexity, or other variables, we believe this metric is an appropriate proxy for time pressure on the auditor.
In a second test of time pressure, we included the original variable representing the abnormal lag, but we also included a dummy variable representing whether the firm completed its SEC filing within five days of its statutory filing deadline.4 We then interacted this dummy variable with the abnormal lag variable. The interaction term isolates the ARLs of firms under the greatest time pressure to meet their filing deadlines. Results of this test were also significant, suggesting that firms with unexpectedly long audits under pressure to meet filing deadlines had a higher likelihood of a future restatement than firms not under such pressure.
From a practical point of view, the first application is the observation that while greater audit effort is associated with higher audit quality, there comes a point in the audit when greater time spent on the audit yields diminishing returns, particularly as pressure to complete the audit increases. Audits that run long, especially those that run unusually long, may be a warning signal rather than an indication of higher audit quality.
Second, the effect of time pressure on the audit cannot be minimized. Reduced statutory filing deadlines, resource constraints caused by the economic recession, and increasing demands by the PCAOB have only exacerbated the pressure auditors face when completing the audit. While it is tempting to conclude that pressure may only exist when firms are near the filing deadlines, our study shows any unexpected delays can undermine audit quality. In response, firms should attempt to develop in-house models for predicting when auditors may face higher levels of pressure than normal. Audits that then meet this threshold could then be subject to additional review and the addition of experienced staff. In addition, in developing the model, firms should consider including certain private information into their models. We could only use publicly available information in our models; by using their own private information, firms can develop even more accurate models. We suggest that firms consider including the following variables:
Changes in key staff from the prior year.
Reductions in staff over the previous year: Reductions in audit staff may make it more difficult to complete the audit on schedule.
Budgeted hours and changes in budget hours.
Increase in control risk assessment: When the auditor decides to place less reliance on internal control, then there may be greater need to do additional post-fiscal year-end audit work.
Reduction in interim testing: When the auditor elects to complete less work throughout the year, then there may be a greater volume of work to be completed after year-end.
Greater or lesser economic pressure on management.
Once the firm has identified audits where time pressure may be especially high, the firm has two options. The firm could assign field auditors with the experience, skills, and behavioral characteristics to more effectively resist the dysfunctional effects associated with pressure. Alternatively, firms could consider allocating more staff to the audit to more effectively reduce the individual workloads in those particular audits. In addition, the audit partner could delay scheduling meetings with the audit committee until the audit manager has a relatively accurate estimate of how long it will take to complete the audit. This recommendation implies that the audit partner may have to discuss the effects of time pressure on audit quality with the client's executive committee.
When the audit firm finds itself in an engagement that has already run unusually long, then it may be necessary to adopt remedial actions. First, the audit firm could allocate more internal quality reviews to audits facing greater pressure. Second, the firm could do more intensive reviews for audit procedures completed near the end of the audit procedures. Time pressure is unlikely to be experienced uniformly over an audit engagement. Since the pressure is likely to be more intense toward the end of the engagement, it seems reasonable that more scrutiny should be placed on procedures completed near the end of the audit. For the same reason, critical decisions made toward the end of the audit should be reviewed for appropriateness within the context of the audit.
Perhaps most importantly, the audit firms should consider in advance how unexpected delays might impact future engagements. This implies that firms take whatever proactive steps to preventing time pressure they can, given the constraints inherent in auditing (e.g., the availability of access to client data) and the inherent pressure created by the various deadlines. Strategies may include planning decisions that intentionally consider the likely pressure involved in specific audit engagements and therefore affect decisions regarding the time budget, the resource allocations, the assessed level of audit risk, the audit fee schedule, and staffing decisions. These decisions—particularly the staffing decisions—should be more strategic than tactical and may involve hiring professional staff who have demonstrated the ability to make sound decisions under pressure, planning more professional staff for what are likely to be more difficult or problematic audit engagements, or assigning staff to clients based on behavioral or experiential characteristics of the staff.
In this paper, we summarize and discuss the results of our recent paper, “The Relationship between Audit Report Lags and Future Restatements” (Blankley et al. 2014). In that study, we evaluated the association between the unusual portion of the audit report lag and the likelihood that the firm would have to restate its financial statement in the future. In other words, we investigated the question, does an unusually long audit have some association with future reporting problems, or does it signal a higher quality audit? We found that there is a positive association between unusual or unexpected ARLs and future restatements. Furthermore, it appears that firms with unexpected delays in their audits that also were subject to increased time pressure had a significantly greater likelihood of a future restatement than other firms, suggesting that time pressure played a role in undermining the effectiveness of the additional audit effort observable in the unusual portion of the audit report lag. Two other potential explanations for the positive association between the unusual ARL and future restatements—economic bonding and the lack of auditor expertise—did not appear to influence it.
There is a relatively extensive literature investigating the audit report lag and its determinants. It is outside the scope of this article to review the literature, but a few examples will illustrate our point. Ashton, Graul, and Newton (1989) investigated the determinants of ARLs and found that variables capturing the complexity of client operations like size, industry, and extraordinary items were positively associated with audit lags. E. Bamber, L. Bamber, and M. Schoderbek (1993) found that lags were positively correlated with the amount of audit work required, and Knechel and Payne (2001) found a positive association between hours worked and the ARL. In each of these examples, there is a strong, positive association between a construct representing audit effort and the length of time taken to complete the audit.
Our failure to confirm an economic bonding influence is, of course, subject to the appropriateness of the variables we used to capture the influence and the construct validity of those variables. Since economic bonding is a broad concept and could have multiple, even conflicting, manifestations in this setting, we do not claim that there is no economic bonding occurring, only that with our statistical tests using the model and variables we did, we did not obtain any evidence of this being an undermining influence.
While we used an established measure of industry expertise in our test, we should also note that measuring this construct reliably is difficult. As with bonding, our conclusion depends on the construct validity of the measure.
We used a five-day window before the filing deadline to capture the pressure on the auditor working to complete the audit requirements during the final workweek immediately prior to the deadline. Our assumption is that audits conducted for clients who filed with one workweek or less until the filing deadline would be more likely to reflect any dysfunction related to time pressure than those of clients who filed with more than one week's cushion.