The Last Chance to Improve Financial Reporting Reliability: Evidence from. Recorded and Waived Audit Adjustments

Transcription

1 The Last Chance to Improve Financial Reporting Reliability: Evidence from Recorded and Waived Audit Adjustments Preeti Choudhary* University of Arizona Kenneth Merkley Cornell University Katherine Schipper Duke University First Version: October 2017 Abstract: We provide evidence on auditors proposed audit adjustments and management s disposition of those adjustments using a large sample of audits of SEC registrants carried out during by eight large accounting firms and inspected by the PCAOB. We find that proposed audit adjustments are pervasive, 81% of sample audit engagements receive at least one proposed adjustment, and that managers appear to frequently disagree with auditors about the disposition of these adjustment, at least one adjustment is waived in 72% of sample engagements. We find that firms with earnings management incentives and those with poor quality financial reporting systems are more likely to benefit from financial statement audits because these factors are associated with more proposed audit adjustments relative to waived adjustments. Finally, we provide evidence consistent with management s disposition of audit adjustments having consequences for financial reporting reliability. We find that management s decision to waive audit adjustments is associated with a higher likelihood that financial statements are subsequently restated. * Corresponding author. This paper was written while Preeti Choudhary was a Senior Economic Research Fellow at the PCAOB. The PCAOB as a matter of policy disclaims responsibility for any private publication or statement by any of its Economic Research Fellows, consultants and employees. The views expressed in this paper are the views of the authors and do not necessarily reflect the views of the Board, individual Board members, or staff of the PCAOB. We thank Michael Gurbutt, Patricia Ledesma, Christian Leuz, Luigi Zingales, PCAOB staff and seminar participants at the PCAOB for helpful discussions. 1

2 1. Introduction This paper provides evidence on proposed audit adjustments 1 and management s disposition of those adjustments, one of the final steps in the financial reporting process and a key determinant of financial reporting reliability. After auditors conduct tests and gather audit evidence, they present information about the financial reporting discrepancies they uncovered to client managers, who then decide whether to make (record) or waive (not record) the proposed adjustments. Recorded adjustments change financial statement amounts (for example, revenues) reported to market participants from the amounts presented to the auditor by management to the amounts proposed by the auditor. That is, the auditor identifies items to be corrected and management is responsible for correcting discrepancies and affirming that waived (unrecorded) adjustments are immaterial. The reliability of a firm s financial statements is thus jointly determined by the pre-audited quality of the financial statements, audit quality the misstatements uncovered by the auditor and management s disposition decisions whether to waive the proposed adjustments. Our study sheds light on three issues with respect to the importance of audit adjustments and their disposition as determinants of financial reporting reliability. First, we provide large sample archival evidence on the frequency, nature and magnitude of audit adjustments. Second, we identify characteristics of audit clients whose financial reports are more likely to benefit from 1 In our analysis, audit adjustment is used to refer to a proposed correction of the financial statements that, in the auditor s judgment, may not have been detected except through the auditing procedures performed (para.09, AU 380, Communication with Audit Committees, superseded by AS 16 currently codified as AS 1301, Communications with Audit Committees.) As discussed in more detail in Section 2, our reading of the authoritative guidance suggests the PCAOB has replaced audit adjustments with misstatements, retaining the idea that the entity may or may not record the correction (adjustment) proposed by the auditor. 2

3 a financial statement audit through an analysis of factors related to proposed and waived audit adjustments, including indicators of financial reporting system reliability, financial reporting complexity, and incentives to manage earnings. Third, we provide evidence on the consequences of management s disposition of audit adjustments, specifically, whether decisions to waive audit adjustments detract from, improve, or have no meaningful impact on financial reporting reliability. Management s decisions to waive adjustments would be expected to detract from reporting reliability if, for example, those decisions derive from incentives to manage earnings toward external targets. Waive decisions would have no meaningful impact if the waived adjustments are truly inconsequential. Finally, waive decisions could increase financial reporting reliability if management responds to a proposed adjustment by explaining how and why its original judgment or estimate is preferable to the auditor s treatment, and the auditor agrees. We analyze a broad sample of proposed, waived and recorded adjustments from audits chosen for inspection 2 by the Public Company Accounting Oversight Board (PCAOB) between 2005 and We combine proprietary data from inspection documents submitted by audit firms as part of the PCAOB inspection process with data from public databases. Relative to previous research on audit adjustments, our analysis of PCAOB data has three advantages for providing evidence on whether and how the audit-adjustment component of the assurance process increases the reliability of financial reporting outcomes. First, our sample contains audit adjustment data from eight large audit firms across more than a decade, while previous research 2 Our sample includes 3,144 audit-year observations and 1,681 distinct audit clients. Because inspected audits are by design a non-random sample of audits by registered public accounting firms, our results may not apply to the population of U.S. listed firms. Section 4.4 provides evidence that selection bias is not a major factor in our analyses. 3

4 has often used data from a single audit firm and/or a small number of years (e.g., Houghton and Fogarty 1991; Wright and Wright 1997; Joe et al. 2011). Second, our data are from the current reporting and assurance environment which differs substantially from the environments examined in prior research (e.g., Wright and Wright 1997; Nelson et al. 2002), because of changes in auditing standards (e.g., Statement of Auditing Standards (SAS) 89, Audit Adjustments, issued December 1999, superseded by AS 2810 and AS 1301 effective after Dec 15, 2010 and 2012, respectively) and legal/regulatory changes affecting auditing and auditors (e.g., the Sarbanes-Oxley Act of 2002 and the establishment of the PCAOB). Third, our sample is from the U.S. financial reporting environment, which differs from non-u.s. environments with respect to reporting standards/regulations and oversight (e.g., Ruhnke and Schmidt 2014; Lennox et al. 2014, 2016). Our aim is to provide evidence on the role of proposed, waived and recorded audit adjustments in achieving reliable financial statements. We start by providing detailed archival evidence on audit adjustments, comparing our results with those reported in prior literature to highlight similarities and differences that may tie to regulatory changes. To provide a sense of economic magnitudes, we present data on adjustments scaled by quantitative materiality, an audit-engagement-specific measure of the magnitude of adjustment needed to affect the mix of information provided to investors. We find that proposed audit adjustments are pervasive: approximately 81% of our sample engagements receive at least one proposed audit adjustment, suggesting that a substantial majority of audits provide opportunities for management to improve reporting reliability by correcting misstatements identified by the auditor. With regard to management s disposition of 4

5 proposed adjustments, we find that 40% of sample audits record at least one adjustment and 72% waive at least one adjustment. However, and perhaps surprisingly, as shown in Figure 1, the distribution of dispositions of proposed adjustments is concentrated in all-or-none outcomes; in approximately 50.5% of sample audits management makes no proposed adjustments (waives 100% of proposed adjustments) and in 11.6% of sample audits management makes all the proposed adjustments (waives 0% of proposed adjustments). Between these two extremes, the remainder of the distribution is roughly uniformly distributed. The descriptive evidence on proposed, recorded and waived audit adjustments suggests audit clients vary in the degree to which the reliability of their financial reports could benefit and do benefit from the audit process. To provide evidence on this point, we examine the relation between proposed and waived adjustments and three categories of firm-specific characteristics shown by prior research to be related to financial reporting outcomes: (1) earnings management incentives, including those discussed in SEC Staff Accounting Bulletin No. 99, Materiality (SAB 99); (2) indicators of client financial reporting system quality, for example, the presence of a material weakness, and indicators of client financial reporting complexity, for example, foreign income, number of segments, reporting a loss and restructuring activity; (3) auditor and auditcommittee characteristics. Our results suggest factors intended to measure earnings-management incentives are associated with more income-related proposed adjustments, which capture the opportunity to improve financial statement reliability, but not with a greater propensity to waive income-related adjustments, which captures management s response to that opportunity. That is, we do not find evidence that earnings management incentives are of first-order importance in determining 5

6 management dispositions of proposed audit adjustments to income. One interpretation of these results is that the guidance in SAB 99 pointing to the importance of contextual factors in materiality judgments involving income is effective; that is, because of SAB 99 the presence of earnings-management incentives increases some combination of auditors detection of and insistence on income adjustments and management s willingness to make those adjustments. We also find that certain earnings management incentives, for example income that is close to breakeven, are associated with management s decision to waive adjustments, when we aggregate across all categories of proposed adjustments. A comparison of statistically significant coefficients on earnings management incentives across proposed and waived adjustments shows that the latter are reliably smaller in magnitude, suggesting that auditor adjustments are associated with a discernible reduction in earnings management, even if they do not completely eliminate it. Turning to indicators of client reporting system quality and reporting complexity, we find that clients with poorer quality financial reporting systems are more likely to benefit from a financial statement audit (i.e., they have more proposed adjustments relative to waived adjustments). We interpret these results as supporting regulatory changes directing auditors to focus attention on internal controls to diagnose the quality of the client s financial reporting system. We find mixed evidence for an association between audit adjustments and measures of reporting complexity such as foreign income. We fail to find evidence that auditor characteristics or audit committee characteristics explain proposed adjustments or waive decisions. This null result suggests there may be little or no variation in the ways auditors apply authoritative guidance to identify audit adjustments and 6

7 little or no variation in the ways audit committees oversee management s disposition of proposed audit adjustments. If correct, this conjecture supports the view that the regulatory structure in which auditors and audit committees operate during our sample period has the intended effect of inducing consistent behavior in the assurance/financial reporting process both across-clients and across-auditors. This conjecture assumes the auditor and audit committee characteristics we consider, based on prior research, capture most or all potential sources of variation; we acknowledge that consistent behavior does not imply there is no room for improvement. Taking the perspective that audit quality, captured by our measures of proposed adjustments and controlling for client characteristics, combined with management s judgment as to how to respond to proposed adjustments, is a key determinant of financial reporting reliability, we analyze the implications of waived audit adjustments. First, we consider whether waived adjustments are associated with future audit outcomes. We find that waived adjustments predict future proposed, recorded and waived audit adjustments, with a year-over-year persistence of about 35%, and that revenue has higher waiving persistence relative to assets, equity, working capital, and income statement categories. These results are consistent with waived adjustments, especially those related to revenues, indicating unresolved reporting issues. We pursue these findings by considering whether waived adjustments affect firms financial reporting (un)reliability as measured by restatements, and find that waiving adjustments is associated with a higher likelihood of restatement for the period of the adjustments. A one standard deviation increase in the magnitude of waived net income adjustments scaled by materiality is associated with a 2.4% increase in the likelihood of restatements; the unconditional sample restatement incidence is 4.0%. An alternative interpretation of these results is that waived 7

8 adjustments capture poor financial reporting quality, as opposed to leading directly to future restatements. Under this interpretation, we would expect to find a similar relation between restatements and recorded adjustments as long as these adjustments are also a signal of poor reporting quality (i.e., proposed adjustments is the sum of both waived and recorded adjustments). However, we find no evidence that recorded adjustments are linked to restatements, despite their ability to predict future proposed adjustments. This finding casts doubt on the idea that the link between waived adjustments and restatements arises because waived adjustments indicate clients with poorer quality financial reporting systems. These results suggest that management s decision to waive adjustments stems, on average, from disagreements with auditors that are resolved with outcomes that detract from financial reporting reliability. The governance implication is that audit committees, charged with oversight of the reporting process, should review management s disposition of audit adjustments carefully, including an evaluation of the costs and benefits of recording vs waiving proposed adjustments, as this represents their last chance to improve the reliability of financial statements prior to their public issuance. The rest of this paper contains four sections. Section 2 provides background by reviewing the relevant research literature and the relevant authoritative guidance. Section 3 describes the data and analyses. Section 4 reports the results. Section 5 concludes. 2. Background: Authoritative guidance and academic research on audit adjustments Section 2.1 describes the current authoritative guidance in generally accepted auditing standards for recording vs waiving audit adjustments. Section 2.2 describes prior research relevant to our analysis of audit adjustments. 8

9 2.1 Authoritative guidance for audit adjustments. AS 2810, Evaluating Audit Results, effective December 15, 2010, specifies the auditor should accumulate misstatements 3 identified during the audit other than those that are clearly trivial, where the latter refers to a smaller order of magnitude than the materiality level of the audit (para 10). The accumulation of misstatements should include a best estimate of the total misstatement in the accounts/disclosures tested, not just the amount of misstatements specifically identified, including amounts related to estimates and projected misstatements from substantive procedures that involve audit sampling (para 12). The auditor should take into account relevant quantitative and qualitative factors in evaluating whether uncorrected misstatements are material, individually or in combination, and with respect to specific accounts or the financial statements as a whole (para 17). AS 1301, Communications with the Audit Committees, effective December 15, brings the audit committee explicitly into the process of management s disposition of the auditor s proposed audit adjustments by, among other things, specifying that the auditor should provide the audit committee with the schedule of uncorrected misstatements presented to management, including the basis for determining the uncorrected errors were immaterial (para 18). AS 1301 Appendix C, para.c1.c.6 clarifies that management is responsible for adjusting 3 Appendix A2 defines and discusses misstatement and clarifies that a misstatement can arise from (unintentional) error or from fraud and can relate to reported amounts, classification, presentation or disclosure. 4 The guidance in AS 1301 is broadly similar to that in SAS 89, Audit Adjustments, issued December 1999 and codified as AU Section 333. SAS 89 is viewed by some, e.g., Ratcliffe (2000), as a response to SEC Staff Accounting Bulletin No. 99, Materiality (SAB 99, issued August 1999). SAB 99, which provides guidance for preparers of financial statements, states, in part, that SEC registrants and by implication their auditors should not use quantitative thresholds alone to judge materiality and provides a list of contextual factors, such as nearness to earnings targets, that should be considered. Our reading of SAS 89 and AS 1301 suggests that the later guidance substitutes the term misstatement for the term audit adjustment, and distinguishes uncorrected misstatements from corrected misstatements. 9

10 the financial statements to correct material misstatements relating to accounts or disclosures and for affirming to the auditor in the representation letter that the effects of any uncorrected misstatements aggregated by the auditor are immaterial, both individually and in the aggregate, to the financial statements taken as a whole. Taken together, AS 2810 and AS 1301 affirm the auditor is responsible for detecting misstatements and reporting them to management, while management, overseen by the audit committee, is responsible for the disposition of those errors. Management s disposition of detected errors is ultimately a financial reporting decision not an auditing decision; the decision is to be based on the materiality criterion including the guidance in SAB 99. The perspective taken in authoritative guidance is not entirely consistent with academic research that views the treatment of audit adjustments as the outcome of a negotiation (e.g., Antle and Nalebuff 1991; Brown and Wright 2008). We adopt the perspective of authoritative guidance and assume that auditors and preparers of financial statements apply the same authoritative guidance including the same materiality criterion, so that disagreements reflect either differences in judgments about the accounting for a given item or differences in how to apply the materiality criterion, particularly its qualitative aspects. If disagreements between the auditor and management and/or the audit committee about the disposition of audit adjustments cannot otherwise be resolved, the auditor could issue a modified audit opinion or even resign from the engagement. 2.2 Prior academic research on audit adjustments. Because the authoritative guidance for financial reporting and auditing and the regulatory regime, specifically the guidance for the disposition of audit adjustments, differs between the US and international contexts, we separate 10

11 our discussion accordingly. Prior research on audit adjustments for US clients consists mostly of two broad streams: (1) descriptive evidence on audit adjustments and (2) analysis of factors related to variation in waived audit adjustments. Previous archival research on the nature, magnitude and type of audit adjustments mostly analyzes small samples of proprietary data, often predating important changes in authoritative guidance relevant to decisions about the disposition of audit adjustments, for example, SAS 89 SAB 99, AS 1301 and AS For example, Bell and Knechel (1994) analyze 28 audit engagements of property and casualty insurers by a single large firm and identify 259 audit adjustments; Kinney and Martin (1994) describe 13 studies using nine data sets covering 1,526 distinct audit engagements during ; and Icerman and Hillison (1991) analyze 147 engagements during We compare our descriptive evidence with results from past research in Sections 3 and 4. A second strand of audit adjustment research on US clients has analyzed factors related to management s disposition of audit adjustments. For example, Brown-Liburd and Wright (2011) find that managements disposition decisions are affected by auditor cooperation and audit committee oversight. Nelson et al. (2002) report survey evidence in which auditors are asked to recall past audits; the survey indicates that auditors recall managers are less likely to waive adjustments that are material, income increasing, or related to a precise accounting standard (defined as a standard that requires little judgment to implement). More recently, Joe et al. (2011) report that management is more likely to waive adjustments as auditor tenure increases, if the client s internal controls do not have material weaknesses, and if a similar adjustment was previously waived; in contrast to prior research, they do not find an association for materiality or client size. 11

12 Researchers have also studied adjustments in an international context, subject to different authoritative guidance and regulation of auditing procedures and management s disposition of audit adjustments. For example, Ruhnke and Schmidt (2014) analyze 1,148 waived and recorded audit adjustments in the 2007 audits of 255 clients of a large German audit firm and find that both the number and magnitude of adjustments vary in predictable ways with client factors such as management quality assessed by the auditor, inherent risk as measured by the client s economic position and control risk as assessed by the auditor and as measured by the presence of an internal audit function and an audit committee. A stream of related and recent research analyzes Chinese data on differences between pre-audit earnings and post-audit earnings (and in some cases, total assets, total liabilities, owners equity and tax liabilities) as a proxy for audit adjustments. 5 These data do not distinguish between errors identified and corrected by management during the audit process (management adjustments) and errors identified and corrected because of the auditor s efforts (audit adjustments) and do not provide information on either total adjustments or waived adjustments separately from recorded adjustments. This research (e.g., Chen et al. 2015) identifies factors that explain variation in pre- vs post-audit balances, including auditor attributes such as industry specialization, client attributes such as profitability and contextual factors such as proximity to a profitability threshold to qualify for stock listing or stock offering. These studies use differences in pre- vs post-audit balances as a measure of audit quality (e.g., Lennox 5 Licensed Chinese audit firms are required to provide certain pre-and post-audit balance amounts to the Chinese Institute of Certified Public Accountants (starting 2001, per Chen et al. 2009). Starting 2006, the reporting is under the purview of the Ministry of Finance, which requires the reporting of pre- and post-audit earnings and total assets (e.g., Lennox et al. 2014, 2016). Some academic researchers have applied for and obtained permission to access these data for research purposes. 12

13 et al. 2014) or compare earnings attributes such as persistence, smoothness, accruals quality, close-to-zero earnings and signed/absolute accruals based on pre- vs post-audit earnings (e.g., Lennox et al. 2016). While substantial differences between the US and Chinese audit, oversight and reporting environments limit the applicability of this research to a US setting, these studies suggest managers ascribe value to audits, including adjusting financial statement amounts during the audit process. We expand this literature and contribute to the understanding of audit adjustments in several ways. First, we provide large sample evidence of audit adjustments in companies that apply US GAAP during a period following important regulatory changes. Hence, our study is more relevant to today s auditing environment, as compared to prior research. AS 1301 and SAB 99, combined, clarify and distinguish the roles of management and auditors in proposing audit adjustments and deciding what to do about those adjustments, and shift the focus of materiality judgments and auditor-client dynamics toward consideration of qualitative and contextual factors. Accordingly, it is an empirical question as to whether results from research using data preceding 2000 generalize to today s reporting environment. 6 Second, we examine factors related to management s disposition of audit adjustments, as has prior research, and extend this line of research by analyzing factors related to proposed adjustments. This extension provides evidence on what types of companies are most likely to benefit from the financial statement audit. Specifically, we evaluate the role of earnings management incentives in both proposed audit adjustments and management s disposition 6 Direct assessments of the effects of SAS 89 on auditor judgments are sparse. An exception is Libby and Kinney (2000) who present behavioral-experimental evidence that SAS 89 led to more immaterial error corrections under limited circumstances. 13

14 decisions, thereby shedding light on the effectiveness of SAB 99. We also consider other factors that might be related to proposed and waived audit adjustments, such as the quality of the client s financial reporting system and its financial reporting complexity, auditor characteristics and audit committee characteristics. Most importantly, our study contributes by investigating the implications of managements disposition decisions for financial reporting reliability. That is, we shed light on the implications of a key preparer financial reporting decision that occurs late in the overall reporting process a last-chance opportunity to improve financial reporting reliability. The implications of management s waive decisions are particularly relevant for audit committees who have responsibility for overseeing the firm s financial reporting, to users of audited financial reports and to securities regulators charged with enforcing compliance with US GAAP. All these groups would be interested in knowing if auditor proposed adjustments are likely to improve financial reporting reliability, even when management disagrees. 3. Data and descriptive analysis We obtained audit adjustment data and other audit-related data from the PCAOB, collected as part of the inspection process mandated by the Sarbanes-Oxley Act. To access these data, we provided a research proposal describing the nature, research question and expected contribution of our study and the data we wanted to access. As a condition of data access, the PCAOB reviews our analysis and must approve the release of any nonpublic information. After receiving PCAOB approval, we collected data on audit adjustments from inspection documents for engagements inspected between 2005 and 2014 from the eight largest US audit firms (Deloitte & Touche (Deloitte); Ernst & Young (EY); Pricewaterhouse Coopers (PwC); KPMG; 14

15 BDO; Grant Thornton; Crowe and RSM (formerly McGladrey). This process resulted in 2,522 inspection observations. The audit adjustment data contain the net effect (the unsigned magnitude) of audit adjustments that exceed a de mimimus threshold (typically set to 3-5% of materiality) classified into seven financial statement categories commonly used in ratio analysis: revenue, operating income, pretax income, net income, working capital, assets and equity. Some types of proposed adjustments are not in the PCAOB data; for example, an adjustment to reclassify an item from the operating section to the financing section on the statement of cash flows. Individual adjustments identified in audit engagements may be more frequent or larger in magnitude than our data indicate, because our data are summed net magnitudes, not the magnitudes of each individual adjustment. Because the PCAOB data do not typically distinguish between adjustments to correct overstatements versus understatements, our analysis uses absolute values (magnitudes) of net audit adjustments, classified into seven financial statement categories. We combine PCAOB audit adjustment data with PCAOB data on auditor materiality judgments collected by Choudhary et al. (2017a) and with data from public databases such as COMPUSTAT, CRSP and Audit Analytics. Because inspection documents request prior year summary adjustment data, we include the year prior to inspection as a separate firm-year observation when possible. Table 1, Panel A displays the ways our sample is affected by data sources and data requirements. The final sample contains 3,144 firm-year observations from 1,681 distinct audit clients. 15

16 Table 1, Panel B reports the number of sample observations by audit firm. 7 As expected, larger audit firms account for much of the sample, with PWC, Deloitte, and KPMG each comprising about 20% of the sample observations followed by Grant Thornton (13.84%), EY (8.72%), BDO (7.38%), RSM (formerly McGladrey) (5.22%) and Crowe (3.78%). About 70% of sample observations are from the four largest audit firms, while 62% of observations in the Audit Analytics database for our sample period represent audits from these firms. Because our sample is skewed towards larger public accounting firms, our analysis may not generalize to all auditors of SEC registrants. Panel C of Table 1 reports observations by fiscal year; both early years and the last year of our sample period contain relatively fewer observations, with 151, 213, and 248 observations from 2005, 2006 and 2007, respectively (cumulatively, about 19.5% of the sample) and 232 observations from 2014 (about 7.4% of the sample). Table 2 reports descriptive statistics for the summary audit adjustment variables. Panel A reports the frequency of audit engagements with proposed, recorded and waived audit adjustments overall and the frequency of adjustments classified into seven financial statement categories. The auditor proposes adjustments for about 81% of our sample observations; Kinney and Martin (1994) report that between 60% and 90% of audits have proposed adjustments. In 40% of sample engagements management records some or all of the proposed adjustments, and in 72% of sample engagements management waives at least some of the proposed adjustments. The relative frequencies of these adjustments are similar for income statement and balance sheet 7 We extract data from PCAOB inspection documents completed by audit firms. The frequency of observations for each audit firm does not reflect the frequency of inspections for that firm due to data limitations, for example, information missing in public databases (e.g. Compustat) and information missing in inspection documents that is conveyed to inspectors in other ways. 16

17 items overall, with variation across the financial statement categories. 8 The least-frequent proposed adjustment is to revenue (32%) and this is also the least-often recorded (9%). In contrast, Kinney and Martin (1994) report that revenues is among the accounts most commonly affected by adjustments. For all financial statement categories, the percentages of audits that record some or all of the proposed adjustments are smaller, sometimes substantially smaller, than the percentages that waive some or all of the adjustments. As shown in Figure 1, in approximately 50.5% of our sample audits management recorded none of the proposed audit adjustments. Panels B, C, and D describe the magnitudes (unsigned amounts) of net adjustments summed across the seven financial statement categories (Sum) as well as separately by category. The amounts are scaled by the auditor s engagement-specific overall quantitative materiality assessments to adjust for variation in audit-client sizes. The overall (Sum) measure is intended to capture the pervasiveness of the audit adjustments, albeit at the possible cost of including the same adjustment twice. 9 To address this limitation we also analyze the net magnitude of audit adjustments to net income. While net income adjustments do not double-count, they have the disadvantage of not capturing balance sheet adjustments. In untabulated analysis, we perform robustness tests using the maximum magnitude of adjustment across the seven financial statement categories and find results similar to those reported in the tables. 8 The number of non-missing adjustment values for the operating income and working capital categories is smaller, as compared to the other categories because not all audit clients calculate these subtotals on their financial statements (e.g. non-classified balance sheets do not report a subtotal for current assets or current liabilities). If the client does not report these subtotals on financial statements auditors tend to leave the adjustment amounts as missing. 9 For example, an adjustment that affects revenue and net income will be included twice (although at different amounts because of tax effects); an adjustment that affects both assets and income will also be included twice. 17

18 Panel B reports the distribution of proposed audit adjustments scaled by engagementspecific materiality. Mean (median) summed proposed audit adjustments are three times (0.88 times) quantitative materiality; Kinney and Martin (1994) report aggregate adjustments in their sample ranging from two to eight times their estimate of materiality. 10 There is significant variation in the magnitude of proposed adjustments across the seven financial statement categories, with the sum of income statement proposed adjustments and the sum of balance sheet proposed adjustments both approximately 1.5 times materiality, on average (the medians are 0.32 and 0.36, respectively); the averages for other financial statement categories are less than 1.0 times materiality, sometimes substantially so. Panels C and D report the distributions of the magnitudes of recorded and waived adjustments, respectively. On average, summed recorded (waived) misstatements are 1.57 (1.43) times audit-engagement-specific materiality with a median of zero (0.47). Panel E reports the ratio of waived adjustments to proposed adjustments. Consistent with Wright and Wright (1997) and Icerman and Hillison (1991) who report 65% and nearly 50% of audit adjustments are waived, respectively; 11 we find that on average 68% of proposed audit adjustments (measured by magnitude of adjustments) are waived with a standard deviation of Income statement adjustments are more likely to be waived (73%) relative to balance sheet adjustments (65%; p<0.001; untabulated). The most frequently waived adjustments are to revenue (mean and first quartile are both 79%) and the least frequently waived adjustments are related to working capital (63% mean) and assets (65% mean). 10 Kinney and Martin estimate materiality as 5% of net income, 0.5% of revenue, and 0.5% of assets. 11 However, Joe et al. (2011) report that only 24.2% of adjustments are waived. Their sample period (2002) follows some prominent financial scandals, perhaps contributing to more conservative disposition decisions. 18

19 These descriptive statistics suggest substantial variation in management s disposition of audit adjustments. In particular, results in Panel E of Table 2 show that at least half the time, management waives the proposed adjustments for every one of the seven financial statement categories we consider. Our data do not allow us to examine the reasons for these decisions, for example, a determination by management that the misstatement identified for adjustment is immaterial or a disagreement between management and auditor judgments or estimates that is resolved without adjustment. 4. Results and discussion 4.1 Which types of clients are most likely to benefit from a financial statement audit? Earnings management incentives. We consider earnings management factors associated with the magnitudes of proposed and waived audit adjustments using the following regression model; the unit of observation is an audit engagement of client i in period t.: Proposed (Waived) Adjustments i,t = α0 + α1 Earnings Management Factorsi,t + αi Year + αj Audit Firm + αk Industry + et (1) Based on prior research (e.g., Healy and Wahlen, 1999) and authoritative guidance (e.g., SAB 99) we predict auditors will propose more adjustments in the presence of earnings management incentives, proxied by Break Even, an indicator set to one if the absolute value of current period pretax income is less than 25% of its three-year historical average; Near EPS, an indicator set to one if current period EPS is with two cents of last year s EPS; Small Profit, an indicator set to one if ROA is positive and less than 3%; Small Positive Streak, an indicator set to one if the change in income before extraordinary items for the last three years is positive and less than 3%; Going Concern, an indicator set to one if the audit report includes a going concern modification; 19

20 and Capital Raising, an indicator set to one if the sum of debt and equity issuances exceeds 20% of total assets. We also consider Near Beat Analyst, an indicator set to one if current period EPS is within two cents of exceeding the most recent analyst consensus forecast; and F Score, the fraud prediction score from Dechow et al. (2011); because the data requirements for these tests reduce the sample by 30% we report, but do not tabulate the results. Because our interest is the association between audit adjustments and their disposition and overall earnings management incentives, we create a composite score (EM Incentive) equal to the sum of the following indicators: Break Even, Near EPS, Small Profit, Small Positive Streak, Going Concern, and Capital Raising. Table 3 reports descriptive statistics for these variables; we see that no single earnings management incentive is pervasive in our sample; the most frequently occurring incentives are Break Even (27%) and Small Profit (24.6%). The least frequently occurring incentives are Going Concern (1.1%) and Near EPS (2.7%). However, nearly 87% of the sample has at least one earnings management incentive factor. Results of estimating equation (1) are reported in Table 4. Panels A and B report results for proposed and waived adjustments, respectively; the dependent variable is either NI Proposed Adjust (the magnitude of proposed income-related adjustments, scaled by materiality) or Sum Proposed Adjust (the magnitude of total proposed adjustments). Sum Proposed Adjust captures adjustment pervasiveness and NI Proposed Adjust captures income adjustments only. We expect more income adjustments in the presence of earnings management incentives; we report balance sheet adjustments because income statement adjustments can affect balance sheet accounts. Columns 1 and 2 of Table 4, Panel A show that the magnitude of income statement proposed adjustments is positively associated with three earnings management incentives 20

21 (Breakeven, Going Concern, and Capital Raising; p<0.01) and with overall earnings management incentives (EM Incentive; p<0.01). Small Positive Streak is negatively associated with the magnitude of proposed income statement adjustments (p<0.01). Results for Sum Adjustments, presented in Columns 3 and 4, show that total proposed adjustments are positively associated with Break Even, Capital Raising, and EM Incentive. In untabulated results we find larger magnitudes of proposed audit adjustments are associated with greater fraud risk as measured by F-score (p <0.01). We do not find an association between Near Beat Analyst and proposed income adjustments or total adjustments. Overall, these results are consistent with Nelson et al. s (2002) survey evidence that auditors believe earnings management incentives affect unaudited financial statement amounts. In Panel B the dependent variable is waived adjustments, using the income-based and total-adjustments definitions in Panel A. As managers can waive only adjustments the auditor has already proposed, the number of sample observations in Panel B is smaller than in Panel A: 2,263 (2,551) for net income (sum adjustments) in Columns 1-2 (3-4). Positive coefficients on earnings management factors suggest more waiving, consistent with client incentives to manage earnings, while negative coefficients suggest less waiving. We find evidence, reported in Columns 1 and 2, that more waiving of income-related adjustments is associated with Capital Raising (p<0.10) and Going Concern (p<0.05), less waiving is associated with Small Profit (p<0.10); we find no evidence of a statistically reliable association between waiving of incomerelated adjustments and EM Incentive. We interpret these results as suggesting that management does not respond to (overall) earnings management incentives by waiving more income adjustments, possibly because of some combination of auditor insistence and management 21

22 attention to the guidance in SAB Results in Column (4) show a positive association between Sum Waived Adjustments and both Break Even and EM Incentives (p<0.01), likely because of waived balance sheet adjustments. We use a seemingly unrelated regression to compare coefficient estimates on earnings management factors across Panels A (proposed adjustments) and B (waived adjustments) of Table 4. Reliably smaller coefficient estimates in Panel B as compared to Panel A would suggest that auditor proposals for adjusting financial statements combined with management dispositions reduce earnings management that is, the amount remaining unadjusted is statistically smaller in magnitude than the amount detected. Results (not tabulated) are that for every reliably nonzero earnings management coefficient in Panel A, the Panel B coefficient is smaller in magnitude; differences are significant at the 0.05 level or better. We interpret these differences as evidence that auditor adjustments reduce earnings management by a discernible amount even if they do not eliminate it. We conjecture that contextual earnings management factors will matter more in decisions to waive smaller proposed adjustments, based on reasoning that quantitative considerations will dominate dispositions of large proposed adjustments and SAB 99 specifies that materiality judgments should consider contextual factors. To shed light on this conjecture, we re-estimate the association between NI Waived Adjust and EM Incentive (Table 4, Panel B, Column 2) on the subsample of observations with proposed income-related adjustments below the sample median 12 This result contrasts with, for example, Nelson et al. s (2002) results based on a survey in which auditors recalled instances of managerial attempts to manage earnings. Nelson et al. s survey data were gathered in 1998, predating the guidance in SAB 99.The data are based on auditor recollections of client characteristics, not archival data on those characteristics. 22

23 (results not tabulated). We find a negative coefficient on EM Incentive (p<0.10). This result is consistent with the importance of qualitative earnings management factors when proposed adjustments are relatively smaller, as suggested by the guidance in SAB 99. Viewed as whole, the results suggest factors intended to measure earnings-management incentives are associated with auditor behavior, in the form of more income-related proposed adjustments, but not with management behavior, in the form of propensity to waive incomerelated adjustments. To the extent this outcome is due to the guidance in SAB 99, which focuses attention on contextual earnings-management factors in reaching income-related materiality judgments, we infer that both auditors and managers are responding to this guidance as intended Client Characteristics. We next consider whether client characteristics are associated with the magnitudes of proposed and waived audit adjustments. Because client characteristics such as financial reporting system quality and financial reporting complexity tend to change slowly, we first consider the persistence of proposed audit adjustments, that is, the ability of proposed audit adjustments to predict themselves. In untabulated analysis we regress one-period-ahead proposed audit adjustments on current period proposed adjustments with the unit of observation an adjustment to a financials statement category (not a firm-year); results suggest the persistence of proposed adjustments is about 40% across the seven financial statement categories. These findings are consistent with the idea that in responding to proposed audit adjustments, management may not fully address the underlying, possibly-firm-specific, issue(s) that led the auditor to identify a misstatement. Based on this finding, we explore the ability of financial-reporting-related client characteristics to explain cross-sectional variation in audit adjustments. We estimate equation (2) 23

24 to evaluate the associations between our measures of proposed and waived adjustments (Sum Proposed Adjust and NI Proposed Adjust; Sum Waived Adjust and NI Waived Adjust) and three client characteristics: financial reporting system quality, financial reporting complexity, and accounting performance: 13 Proposed (Waived) Adjustments i,t = α0 + α1 Financial Reporting System Quality + α2 Financial Reporting Complexity + α3 Accounting Performance + αi Year + αj Audit Firm + αk Industry + et (2) We proxy for financial reporting system quality using MW (indicator set to one if there is a material weakness in internal controls), Audit Time (difference between the 10-K filing statutory due date and the date the audit firm signed its audit opinion), Earn Ann Time (difference between the audit opinion signature date and the earnings announcement date) and Asset (log of assets). We expect higher (lower) values of Audit Time, Earn Ann Time, and Assets (MW) to capture better financial reporting system quality. We proxy for financial reporting complexity using Earn Vol (standard deviation of income before extraordinary items for the current year and the last three fiscal years), Intangibles (sum of R&D and advertising expense scaled by assets), Foreign Income (ratio of pretax foreign income to pretax income), Segments (natural logarithm of number of geographic and business segments), Restructure (ratio of restructuring costs after-tax to lagged total assets), Merger (an indicator set to one if the client had a business combination that affected sales and zero otherwise), and Accruals (earnings before extraordinary items less operating cash flows divided by total assets). We expect higher values to correspond with more 13 We do not expect associations between client characteristics and adjustments to vary depending on whether adjustments are measured as income-related only or in total; we estimate equation (2) using both measures because they capture different aspects of adjustments. We include audit firm, industry (defined as 2 digit SIC code), and year fixed effects to capture variation in audit engagements. 24

25 complex financial reporting, except that higher values of Intangibles correspond to less asset capitalization and therefore less estimation. We measure accounting performance using Sales Growth (year-over-year percentage change in sales) and ROA (income before extraordinary items divided by total assets). We expect auditors to propose more audit adjustments when the client s financial reporting system is of poorer quality and/or more complicated. With respect to waiving adjustments, auditors might resist management s waive decisions more strongly when the financial reporting system is of poorer quality or more complicated because of greater misstatement risk. On the other hand, management of clients with complex reporting systems may resist recording adjustments related to judgment or extrapolation of errors (i.e., the amount is statistically implied not directly detected) which are more likely in more complex financial reporting systems. Table 5 reports the results of estimating equation (2) for income-related adjustments, denoted NI (Columns 1 and 3 for proposed and waived adjustments, respectively) and for total adjustments, denoted Sum (Columns 2 and 4 for proposed and waived adjustments, respectively). In Columns (1) and (2) we find a positive (negative) coefficient on MW (Audit Time, Earn Ann Time, and Log Assets), significant at the 0.01 level, suggesting auditors propose more adjustments when the client s financial reporting system is poorer. In terms of economic magnitudes, a client material weakness is associated with net income (sum) proposed adjustments that exceed 50% (400%) of quantitative materiality. Dropping the financial reporting system quality variables from the regression decreases the R-squared by 50% (results not 25

26 tabulated), highlighting the importance of financial reporting system quality for the reliability of financial statements. In contrast to the evidence of a link between both proposed and waived audit adjustments and the quality of the audit client s financial reporting system, we find that only two indicators of the client s financial reporting complexity are reliably associated with proposed adjustments; Foreign Income (Intangibles) is associated with a greater magnitude (smaller magnitude) of proposed adjustments (p<0.10 or better). Finally, better performing clients have fewer proposed adjustments as ROA loads negatively (p<0.01). Overall, these results suggest that financial reporting system quality is a more important factor in determining proposed audit adjustments than financial reporting complexity. The implication is that the auditor has greater potential to improve the reliability of financial reporting outcomes when its client has a poorer financial reporting system. Table 5 also presents analysis of client characteristics associated with waived incomerelated adjustments (Column 3) and total adjustments (Column 4). With regard to financial system reporting quality, Log Assets, a measure of both client size and reporting system quality, is negatively associated with waived adjustments (p<0.01), while clients with material weaknesses (MW) are more likely to waive adjustments (p<0.01). In contrast, Joe et al. (2011) report that management is more likely to waive adjustments when the client s internal controls do not have material weaknesses. With regard to financial reporting complexity and accounting performance, Intangibles (an inverse measure of complexity) and ROA (a measure of performance) are negatively and positively associated with waiving adjustments, respectively. 26

27 Estimating a seemingly unrelated regression allows us to compare the MW coefficient estimates from specifications based on proposed and waived adjustments (results not tabulated). We find the MW coefficient for waived adjustments ( in Column 3) is smaller than the MW coefficient for proposed adjustments ( in Column 1), consistent with both proposed audit adjustments and management s disposition of those adjustments improving the reliability of financial reporting in the presence of material weaknesses. Generally these results suggest that clients with poorer quality financial reporting systems are most likely to benefit from a financial statement audit supporting regulatory changes directing auditors to focus on internal controls (e.g. AS 5) Client Audit Committee Characteristics and Auditor Characteristics. In untabulated analyses we consider the associations between proposed and waived audit adjustments and both client audit committee characteristics and auditor characteristics. For the subsample with governance data in Boardex (approximately 90% of total observations), we construct the following audit committee measures, based on Badolato et al. (2014): Busy AC (indicator set to one if 50% or more of audit committee members serve on three or more boards simultaneously), Accounting Exp (ratio of audit committee accounting experts to total audit committee members in a given year), Finance Exp (ratio of audit committee finance experts to total audit committee members in a given year), Supervisory Exp (ratio of audit committee supervisory finance experts to total audit committee members in a given year) 14, and Industry 14 As explained in Badaloto et al. s Appendix A, Boardex defines expertise based on prior employment or a professional distinction. Accounting experts have accounting experience as evidenced by a job title, or auditing experience at a designated list of firms, or professional certification; finance experts have work experience using financial reports; supervisory finance experts have work experience supervising individuals involved in financial reporting, for example, a CEO or president. 27

28 Exp (ratio of audit committee industry experts to total audit committee members in a given year). Despite the requirement that audit committees review waived audit adjustments (AU Section 380), we do not find evidence that these audit committee characteristics are associated with variation in decisions to waive adjustments. In contrast, in an experiment involving 63 auditors and strong vs weak audit committee and past-client-relationship manipulations, Brown-Liburd and Wright (2011) find that strong vs weak audit committees affect auditor judgments. One possible explanation is that in our archival sample all audit committees, regardless of their characteristics, review waived adjustments, so there is no variation to be explained; alternatively, the finding of no associations might be due to noise in the audit committee measures. To shed light on whether auditor characteristics are associated with proposed or waived adjustments, we create several auditor-characteristics variables, measured at the clientengagement level using Audit Analytics data: New Auditor, an indicator set to one if this engagement is the auditor s first audit of the client; Expert Auditor (indicator variable set to one if the auditor s market share measured by audit fees exceeds 30% of the total audit fees in the client s two-digit SIC and Metropolitan Statistical Area (MSA)); Auditor Office Size (natural log of total audit fees collected by the client s auditor from all clients in the same MSA by year); Important Client (ratio of the client s total assets to the total assets of all audit clients of the client s auditor in the same MSA and year), Log Auditor Tenure (natural logarithm of number of years this auditor has audited the client). We also include Log # Part 1 (natural logarithm of the number of Part 1 findings identified by the PCAOB in the audit inspection, representing audit quality (Aobdia 2016)). In some specifications, we find evidence of a link between waived adjustments and new clients (p<0.10), and a link between longer auditor tenure and smaller 28

29 magnitudes of proposed audit adjustments (p<0.05). This evidence is sensitive to measurement choices. Based on these findings, we conclude that after controlling for audit firm fixed effects, neither proposed audit adjustments nor waived adjustments vary meaningfully with the auditor characteristics we measure. Our evidence contrasts with Joe et al. (2011) report that management is more likely to waive adjustments as auditor tenure increases. Viewed as a whole, the results of these untabulated analyses suggest that audit committee characteristics and auditor characteristics have little ability to explain cross-sectional variation in proposed and waived audit adjustments. Given that guidance for both audit committees and auditors is applicable to all committees and auditors regardless of their circumstances and characteristics, these findings are consistent with the inference that those responsible for implementing guidance in proposing audit adjustments and overseeing management s disposition of those adjustments are doing so consistently. We acknowledge, however, that consistency in behavior does not necessarily imply that there is no room for improvement. 4.2 Implications of Management s Decisions to Waive Audit Adjustments Our final analyses consider how waived adjustments are relate to future audit adjustments (section 4.2.1) and how waived adjustments are related to an inverse measure of reporting reliability, restatements of financial reports (section 4.2.2) Waived Adjustments and Future Audit Adjustments. To examine the audit implications of waived adjustments we consider the relation between last-period waived adjustments and current-period proposed, recorded and waived audit adjustments. We estimate the following regression equation: Proposed, Recorded, or Waived Adjustment i,t = α + α1 Waived Adjustment i,t-1 (3) 29

30 + αm Category i,t-1 + αn (Audit Adjustment i,t-1 x Category i,t-1 ) + αi Year + αj Audit Firm + αk Industry + ei,t-1 The results shed light on whether waived adjustments are associated with auditor-identified misstatements in the next period (as opposed to being eliminated, for example, by accrual reversals). We estimate equation (3) for both the total amount of waived adjustments and waived adjustments for each of seven financial statement categories described in Table 2. The unit of observation is an adjustment to a category (not a firm-year). The regressions include industry, year, and audit firm fixed effects. Standard errors are clustered by firm. Table 6 reports the results. Analyses based on proposed adjustments are presented in Columns 1, 2 and 3. We find that waived adjustments in period t-1 are positively associated with period t proposed adjustments (p<0.01. The coefficient on Waived Adjustments in Column (1) suggests that for every dollar of period t-1 waived adjustments (scaled by materiality), dollars of period t adjustments are proposed by the auditor, consistent with the idea that the underlying causes of proposed adjustments recur or continue to pose problems for at least one period. Results in Column 2 do not provide evidence that this association differs based on whether the adjustments are related to the income statement versus the balance sheet (the coefficient on Income Statement x Waived Adjustmentst-1 is not significant at the 0.10 level). Results in Column 3 suggest some differences across financial statement categories: recurring proposed adjustments are more likely for adjustments related to assets or revenue (p <0.01). If period t-1 waived adjustments recur as proposed adjustments in period t, it is possible that these adjustments are simply passed on to, and recorded in, period t. To examine this possibility, we evaluate the relation between waived adjustments in period t-1 and recorded 30

31 adjustments in period t. Results, presented in Columns 4, 5 and 6 of Table 6, indicate that waived adjustments have predictive ability for future recorded adjustments (p<0.01), consistent with clients recording some adjustments one period after they are proposed. Noting that for an adjustment to be waived it must first be proposed, we interpret the coefficient of on Waived Adjustments in Column (4) to suggest that for every dollar of period t-1 waived adjustments (adjusted for materiality), approximately 0.11 dollars of adjustments (1*0.6029*0.1857) is recorded in period t. A comparison of coefficients in Column 1 (proposed adjustments) with those in Column 4 (recorded adjustments) based on a seemingly unrelated regression estimation (results not tabulated), shows that waived adjustments are more predictive of proposed adjustments than of recorded adjustments (p<0.01), suggesting that some but not all period t-1 waived adjustments are recorded in period t. Comparing Columns 3 and 6 using seemingly unrelated regression (results not tabulated) we find evidence of a stronger relation between period t-1 waived adjustments and period t proposed adjustments, relative to recorded adjustments, with the disparity more pronounced in asset and revenue accounts (p<0.01). This evidence, which suggests that errors in revenues and asset accounts tend to recur, is a possible contributing explanation for the greater frequency of restatements in revenue accounts relative to other financial statement categories (e.g., Choudhary, et al. 2017b). Given the evidence that the relation between waived adjustments in period t-1 and proposed adjustments in t is greater than that between waived adjustments in period t-1 and recorded adjustments in period t, we consider the relation between waived adjustments in periods t and t-1. Columns 7, 8, 9 report these results. As expected from the results in the other columns, 31

32 the coefficient on waived adjustments in period t-1 in Column 7 is positive and highly significant, consistent with waived adjustments continuing to pose problems for financial reporting for at least one period. Similarly, we find that the persistence of waived adjustments is highest for revenue accounts (p<0.05; Column 9), consistent with concurrent research (e.g. Choudhary et al. 2017b) that finds evidence of recurring material errors in revenues Waived Adjustments and Restatements. Our final analyses consider the implications of waived adjustments for financial reporting (un)reliability, measured by restatements. We use equation (4) to examine the relation between waived and recorded period t adjustments in a given period t and restatements (material error corrections) of period t financial statements: Restated = α0 + α1 Waived Adj $ + α2 Recorded Adj $ + α3 Client Characteristics + αi Year + αj Audit Firm + αk Industry + et (4) Table 7 reports the results of estimating equation (4) using a linear probability model. 15 As in previous analyses, in Panel A Columns 1 and 2 (Columns 3 and 4) the test variables are waived income adjustments (waived summed adjustments across financial statement categories). In Columns 1 and 3, larger magnitudes of waived adjustments are associated with a higher likelihood of restatement (p < 0.01). This results suggests that waiving not recording proposed adjustments reduces financial reporting reliability as measured (inversely) by restatement incidence. The coefficient on Waived Adjustments in Column 1 suggests that a one standard deviation increase in waived adjustments is associated with a 2.4% increase in the likelihood of restatement (0.0911*0.26=0.024). The effect represents a 60% increase in 15 Inferences are unchanged if we estimate a logistic regression model. We report results based on the linear probability model because it better accommodates fixed effects and facilitates interpretation of marginal effects. 32

33 restatement likelihood, relative to the 4.0% unconditional sample probability of restatement. We obtain similar results (not tabulated) when we replace the dependent variable with restatements of period t or t+1 or period t, t+1 or t+2, or with an indicator equal to one if the period was either restated or revised; a revision is identified using the Audit Analytics indication that financial statements were corrected without an Item 4.02 disclosure. In the second and fourth columns we create indicator variables, Large Waived Adjustment and Large Recorded Adjustment, for whether the waived and recorded misstatement are above the sample median level of proposed adjustments scaled by quantitative materiality. We find that large waived adjustments are associated with 1.3% (t= 1.59; p<0.12; Column 2) and 2.1% (p<0.05; Column 4) increases in the probability of a restatement. We find no evidence that large recorded adjustments are linked to restatements (p > 0.10). Because data limitations do not allow us to link the topic area of a restatement to the topic area of an adjustment we cannot show whether a specific waived adjustment led to a specific restatement.. Thus, an alternative interpretation of our results is that waived adjustments relate to restatements because they indicate poor financial reporting quality (i.e., waived adjustments represent errors in the client accounting system). However, recorded adjustments also represent errors in the client s accounting system (that is, proposed adjustments include both waived and recorded adjustments). The lack of a statistically significant association between recorded adjustments and restatements in our analyses casts doubt on the idea that the link between waived adjustments and restatements is due only to waived adjustments providing 33

34 a signal of a poor quality financial reporting system, as we would expect to find similar results for recorded adjustments under this explanation. 16 In Panel B of Table 7, we repeat the analysis on the subsample in which management either waived all adjustments or recorded all adjustments, and continue to find evidence that waived adjustments are linked to future restatements; results are significant at p<0.10 or better. In Column 3, we find evidence that recorded adjustments in this subsample are negatively associated with future restatements (p<0.01), suggesting that auditors improve financial reporting reliability by proposing audit adjustments, conditional on management agreeing to record these adjustments. As in Panel A, large (above the sample median) waived adjustments are positively associated with restatement incidence (p<0.01). Overall, results of our tests of the financial reporting implications of waiving suggest that management s decision to waive adjustments stems, on average, from disagreements with auditors that detract from financial reporting reliability, as measured by increased restatement incidence. The governance implication is that audit committees, who are charged with oversight of the reporting process, should review management s disposition of audit adjustments carefully, including an evaluation of the costs and benefits of recording vs waiving proposed adjustments. 4.3 Selection Bias Our sample is based on audits the PCAOB selected for inspection using a risk-based approach and is therefore a non-random sample of SEC-registrant audits. Prior research has evaluated whether PCAOB-inspected samples exhibit selection bias. First, Aobdia, et al. (2017) 16 In untabulated analyses, we confirm that both waived and recorded adjustments in period t-1 have statistically indistinguishable abilities to predict next period s proposed adjustments (p>0.10), suggesting that both have similar ability to signal the quality of the client s financial reporting system. 34

35 use a seemingly unrelated regression to evaluate if the association between material weakness and restatement differs in a PCAOB-inspected sample versus the same-time-period population of SEC-registrant audits with available data. They find no evidence of a difference between the coefficients on material weakness in the full sample versus the PCAOB-inspected sample, supporting a conclusion that selection bias in the restatement setting does not affect their conclusions. Second, Aobdia (2016 and Choudhary, et al. (2017a) create selection models to predict inspected engagements; the predictive ability of these models ranges from 24% to 48% better than random prediction. Neither paper finds evidence that selection bias affects their analysis. Assuming the validity and comprehensiveness of the variables included in the selection models, their modest predictive ability supports the view that PCAOB-inspected samples are not sufficiently different from other audit engagements to affect inferences. In untabulated analyses, we estimate a Heckman selection model for our sample with an area under the ROC curve of After applying the Heckman procedure to the analysis reported in Table 7, Panel A, we find no evidence of selection bias (i.e., Rho, the test for independent equations, is insignificant at conventional levels; p > 0.10), and we continue to find evidence that waived adjustments are positively associated with restatements. Also, while selection bias could result if the PCAOB chooses audits for inspection on the basis of audit adjustments or reporting reliability (for example, restatements) or both we think this is unlikely. For our sample period, the PCOAB obtains information on both materiality and proposed adjustments from inspection documents that audit firms provide after their audits are selected for inspection, and most restatements have not been announced when the PCAOB selects audits for inspection. Finally, the rough similarity in our sample descriptive statistics and descriptive 35

36 statistics reported in previous research, as discussed in Sections 3 and 4, suggests that our sample is not highly atypical. An omitted variable that is correlated with both waived adjustments and reporting reliability could also influence our results. Following Frank (2000) and Larcker and Rusticus (2010), we evaluate how much the omitted variable must be correlated with both the dependent variable and the variable of interest to overturn our reported results. Using a significance level of 0.05 (0.10), we estimate that an omitted variable must be positively correlated with both waived income-related [summed] adjustments and restatements by (0.137) [0.125 (0.145)] after controlling for other factors. This seems implausible given the difficulties of predicting restatements and because most known predictors do not correlate at this level. 5. Conclusion We provide evidence on how audits, specifically the auditor s identification of misstatements (proposed audit adjustments) and management s decisions as to whether to waive those adjustments, affect reporting outcomes. Our analysis is motivated by the perspective that audit adjustments provide an opportunity for management to improve the reliability of financial reports as a near-final step in the overall reporting process. We focus first on the auditor s task of identifying misstatements and provide descriptive evidence on the frequency and other properties of audit adjustments for a broad sample of US audit engagements during Our analysis updates and extends previous research that typically uses small samples, often from a single audit firm and often from a time period preceding the current regulatory environment, or samples from non-u S audits. We find that over 80% of audits have proposed adjustments, an indicator of the extent to which the assurance 36

37 process could improve the quality of financial reporting outcome. When we turn to management s role, we find that the client s disposition of these adjustments shows a wide variation in responses, with a portion (approximately 12%) recording all proposed adjustments and a much larger proportion (just over 50%) waiving all proposed adjustments. Our analyses of determinants of proposed adjustments and management s disposition of those adjustments indicates that an important determinants of proposed adjustments is the quality of the client s financial reporting system, as would be expected if the auditor extends substantive procedures when the client s financial reporting system is poorer, including in particular a material weakness. We also find that earnings management incentives have reliably positive associations with proposed income-related and total adjustments, as would be expected if auditors follow the guidance in SAB 99 to take account of contextual factors in assessing errors detected in the course of the audit. With regard to management s disposition of proposed adjustments, we find that earnings management incentives do not explain waiving of incomerelated adjustments but do have explanatory power for total adjustments. Finally, we find little or no statistically reliable evidence of links between waived adjustments and either audit committee governance-characteristics or auditor characteristics, as would be expected if both audit committee procedures and audit procedures consistently follow authoritative materiality guidance. Our perspective is that the purpose of audit adjustments is to increase the quality of financial reporting outcomes and the extent to which this happens is a function of both the auditor s effectiveness in detecting errors (audit adjustments) and management s disposition of those adjustments (waive vs record). When we focus on management s disposition, taking as 37

38 predetermined the auditor s proposed adjustments, we find that waived adjustments, especially larger amounts, are associated with both next-period proposed adjustments (suggesting that underlying causes of misstatements recur) and with restatements (suggesting that waiving adjustments reduces the quality of reporting outcomes). 38

39 References Antle, R. and B. Nalebuff Conservatism and auditor-client negotiations. Journal of Accounting Research 29: supplement, Aobdia, D The validity of publicly available measures of audit quality: Evidence from the PCAOB inspection data. Working paper, Northwestern University. Aobdia, D., P. Choudhary and G. Sadka Do auditors correctly identify and assess internal control deficiencies? Evidence from the PCAOB data. Working paper, University of Arizona, Northwestern University and University of Texas at Dallas. Badolato, P., D. Donelson and M. Ege Audit committee financial expertise and earnings management: the role of status. Journal of Accounting and Economics 58 (2-3): Bell, T. and R. Knechel Empirical analyses of errors discovered in audits of property and casualty insurers. Auditing: A Journal of Practice and Theory 13:1, Brown, H. and A. Wright Negotiation research in auditing. Accounting Horizons 22:1, Brown-Liburd, H., and A. Wright The effect of past client relationship and strength of audit committee on auditor negotiations. Auditing: A Journal of Practice and Theory 30:4, Brown-Liburd, H., A. Wright and V. L. Zamora Managers strategic reporting judgments in audit negotiations. Auditing: A Journal of Practice and Theory 35:1, Chen, C.J.P, X. Liu, X. Su and X. Wu Auditor-client interdependence and audit quality: partner-level evidence. Working paper, China Europe International Business School, University of Hong Kong, City University of Hong Kong and Central University of Finance and Economics, China Chen, S., G. Krishnan, W. Li and Y. Zhang What client and auditor attributes are associated with auditors decision to require adjustments to pre-audit financial statements? Working paper, Beijing Institute of Technology and American University. Choudhary, P., K. Merkley and K. Schipper. 2017a. Direct measures of auditors quantitative materiality judgments: properties, determinants and consequences for audit characteristics and financial reporting reliability, working paper, University of Arizona and Cornell University. 39

40 Choudhary, P., K. Merkley and K. Schipper. 2017b. Qualitative characteristics of financial reporting errors deemed immaterial by managers, working paper, University of Arizona and Cornell University. Dechow, P., W. Ge, C. Larson and R. Sloan Predicting material accounting misstatements. Contemporary Accounting Research 28:1, DeFond, M. and J. Jiambalvo Incidence and circumstances of accounting errors. The Accounting Review 66:3, Frank, K Impact of a confounding variable on a regress coefficient. Sociological Methods and Research 29: Healy, P. and J. Wahlen A review of the earnings management literature and its implications for standard setting. Accounting Horizons 13:4, Houghton, C. and J. Fogarty Inherent risk. Auditing: A Journal of Practice and Theory, Spring, Icerman, R. and W. Hillison Disposition of audit-detected errors: Some evidence on evaluative materiality. Auditing: A Journal of Practice and Theory, 10:1, Joe, J. A. Wright and S. Wright The impact of client and misstatement characteristics on the disposition of proposed audit adjustments. Auditing: A Journal of Practice and Theory 30:2, Keune, M. and K. Johnstone Materiality judgments and the resolution of detected misstatements: The role of managers, auditors and audit committees. The Accounting Review 87:5, Kinney, W. and R. Martin Does auditing reduce bias in financial reporting? A review of audit-related adjustment studies. Auditing: A Journal of Practice and Theory 13:1, Larcker, D. and T. Rusticus On the use of instrumental variables in accounting research. Journal of Accounting and Economics 49:3: Lennox, C., X. Wu and T. Zhang Does mandatory rotation of audit partners improve audit quality? The Accounting Review 89:5, Lennox, C., X. Wu and T. Zhang The effect of audit adjustments on earnings quality: Evidence from China. Journal of Accounting and Economics 61:2-3,

41 Libby, R. and W. Kinney Does mandated audit communication reduce opportunistic corrections to manage earnings to forecasts? The Accounting Review 75:4, Nelson, M., J. Elliott and R. Tarpley Evidence from auditors about managers and auditors earnings management decisions. The Accounting Review 77: supplement, Ratcliffe, T Understanding audit adjustments. The CPA Journal April, Ruhnke, K. and M. Schmidt Misstatements in financial statements: The relationship between inherent and control risk factors and audit adjustments. Auditing: A Journal of Practice and Theory 33:4, Trotman, K., A. Wright and S. Wright Auditor negotiations: an examination of the efficacy of intervention methods. The Accounting Review 80:1, Wright, A. and S. Wright An examination of factors affecting the decision to waive audit adjustments. Journal of Accounting, Auditing and Finance 12:1,

42 Variable Appendix Recorded Adjustments Waived Adjustments Proposed Adjustments Materiality Income Statement Balance Sheet Working Capital Assets Revenue Operating Income Pretax Income Net Income NI Proposed Adjust Sum Proposed Adjust The absolute dollar amount of audit adjustments recorded by the client to a particular financial statement category (i.e. revenue, operating income, pretax income, net income, assets, equity, or working capital) scaled by quantitative materiality from PCAOB proprietary inspection documents The absolute dollar amount of audit adjustments waived by the client to a particular financial statement category (i.e. revenue, operating income, pretax income, net income, assets, equity, or working capital) scaled by quantitative materiality from PCAOB proprietary inspection documents The sum of Recorded Adjustment and Waived Adjustment from PCAOB proprietary inspection documents The dollar amount of quantitative materiality reported by the client s audit firm to the PCAOB. Indicator equals one if the adjustment affects a subtotal on the income statement (i.e., revenue, operating income, pretax income, or net income) from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects a subtotal on the balance sheet (i.e., assets, equity, or working capital) from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects working capital from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects assets from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects revenue from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects operating income from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects pretax income from PCAOB proprietary inspection documents Indicator equals one if the adjustment affects net income from PCAOB proprietary inspection documents The absolute value of income-related adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents The sum of the absolute values of proposed adjustments across seven financial statement categories reported in PCAOB inspection documents (working capital, assets, equity, revenue, operating income, pretax income, and net income) scaled by quantitative financial statement materiality 42

43 NI Waived Adjust Sum Waived Adjust Large Waived NI Adjustment Large Recorded NI Adjustment Large Waived Sum Adjustment Large Recorded Sum Adjustment The absolute value of waived income-related adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents The sum of the absolute values of waived adjustments across seven financial statement categories reported in PCAOB inspection documents (working capital, assets, equity, revenue, operating income, pretax income and net income) scaled by quantitative financial statement materiality An indicator set to one if NI Waived Adjust exceeds the sample median level of proposed adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents. An indicator set to one if NI Record Adjust exceeds the sample median level of proposed adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents. An indicator set to one if Sum Waived Adjust exceeds the sample median level of proposed adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents. An indicator if Sum Record Adjust exceeds the sample median level of proposed adjustments scaled by quantitative materiality from PCAOB proprietary inspection documents. Breakeven An indicator set to one if either (1) EPS (epspx) is between - $0.05 and $0.05 or (2) if current period pretax income (pi) from COMPUSTAT is less than 25% of 3 year historical average of pretax income, and zero otherwise Near EPS Small Profit Small Positive Streak Going Concern Capital Raising EM Incentive Near Beat Analyst An indictor set to one if the EPS from COMPUSTAT in period t is within 2 cents of EPS in period t-1 An indicator set to one if return on assets (ROA) from COMPUSTAT is between 0 and 3%. An indicator set to one if the change in income before extraordinary items for the last three years (from COMPUSTAT) is greater than zero and less than a 3% increase relative to the prior year, and zero otherwise. An indicator set to one if the audit opinion from COMPUSTAT includes a going concern modification. An indicator set to one if COMPUSTAT dltix + sstk is greater than 20% of assets (at) Sum of earnings management incentives from COMPUSTAT (Breakeven, Near EPS, Small Profit, Small Positive Streak, Going Concern, Capital Raising) An indicator set to one if the EPS is within 2 cents of beating consensus EPS estimate by analyst from IBES 43

44 F-score Fraud risk measured computed following Dechow et al. (2011) using data from COMPUSTAT. MW An indicator set to one if the company reported a material weakness (IC_IS_EFFECTIVE) and zero otherwise in COMPUSTAT. Audit Time The difference between the 10-K filing statutory due date and the date the audit opinion was signed by the audit firm (SIG_DATE_OF_OP_S) from Audit Analytics. Earnings Announce Time The difference between the audit signature date ((SIG_DATE_OF_OP_S from Audit Analytics) and the earnings announcement date (rdq from COMPUSTAT). Assets COMPUSTAT at Earnings Volatility The standard deviation of income before extraordinary items for the current year and the last three fiscal years from COMPUSTAT. Intangibles The sum of R&D (COMPUSTAT xrd) and advertising (COMPUSTAT xad) expense scaled by Assets. Foreign Income Ratio of pretax foreign income (COMPUSTAT pifo) divided by pretax income (COMPUSTAT pi). Segments The natural logarithm of the number of geographic and business segments from COMPUSTAT. Restructure Ratio of restructuring costs after-tax (COMPUSTAT rca) to lagged total assets. Merger An indicator set to one if the company had an acquisition that contributed to sales and zero otherwise from COMPUSTAT. Accruals Earnings before extraordinary items less operating cash flows divided by total assets from COMPUSTAT. Sales Growth The percentage year-over-year change in sales from COMPUSTAT. ROA The ratio of income before extraordinary items to total assets from COMPUSTAT. Net Income Total net income reported for the fiscal year (COMPUSTAT ib). Pre-tax Income Total income before taxes reported for the fiscal year (COMPUSTAT pi). Revenue Total revenue reported for the fiscal year (COMPUSTAT revt). Restated An indicator set to the one if the company restated the current year financial statements in a subsequent year from Audit Analytics. Log (# Part One Findings) The natural log of 1+ the number of part 1 PCAOB inspection findings from PCAOB proprietary inspection documents Capital Expenditures Ratio of capital expenditures (COMPUSTAT capx) to total assets. 44

45 Loss BM Multinational An indicator set to one if income before extraordinary items (COMPUSTAT ib) is less than zero. The ratio of the book value of common equity to the market value of equity from COMPUSTAT. An indicator set to one if a client reports non-zero foreign income taxes (COMPUSTAT txfo). 45

46 Figure 1. Frequency Distribution of the Ratio of Waived Audit Adjustments to Proposed Audit Adjustments The figure shows the frequency distribution of the ratio of the total magnitude of waived audit adjustments to the magnitude of proposed adjustments for our sample observations. The sample contains 3,144 audits (1,681 unique audit clients) by the eight largest accounting firms, chosen for inspection by the PCAOB during The frequency of 100% waived adjustments is 50.5% and the frequency of 0% waived adjustments is 11.6%. 46