Internal Control, Audit Committee and Audit Delay

Internal Control, Audit Committee and Audit Delay Internal Control Quality and Audit Delay in the SOX Era Michael Ettredge* University of Kansas Chan Li University of Kansas (doctoral student) Lili Sun Rutgers University *Contact Author School of Business 1300 Sunnyside Ave. University of Kansas Lawrence, KS 66045-7585 mettredge@ku.edu 785-864-7537 August 2005 1 Internal Control Quality and Audit Delay in the SOX Era SUMMARY: This paper studies the impact of internal control quality on audit delay in the SOX era. We hypothesize and empirically find the following. First, to fulfill the requirements of SOX 404 reporting, companies experience significantly longer audit delay in 2004 compared to 2003. Second, companies with material weakness in ICOFR experience longer audit delays than companies with effective ICOFR. Third, general material weaknesses cause longer delays than specific material weakness. Additional analyses investigate the impact of internal control quality upon audit delay based upon the specific types of control weakness exhibited. We devise categories of control weakness based on the COSO framework. The results contribute to the literature dealing with the relation between audit delay and internal control quality. More importantly, such empirical evidence is timely and important for better understanding the impact of SOX act upon the auditing process and the financial reporting. Keywords: Audit Delay, Internal Control, SOX 404. 2 Internal Control Quality and Audit Delay in the SOX Era INTRODUCTION This paper investigates the effect of companies’ internal control quality on audit delay. The term “audit delay” refers to the length of time from a company’s fiscal year-end to the date the auditors sign their report. The determinants of timeliness of audit reports are interesting since audit delay affects the timeliness of both the annual earnings information release and the Form 10-K filing date. It is well documented that late announcements of earnings are more often associated with lower abnormal returns than are early announcements (Givoly and Palmon 1982; Chambers and Penman 1984; Kross and Schroeder 1984). Delayed disclosure potentially compromises equal access to information among investors (Hakansson 1977). It allows some investors to acquire costly private information and thus trade on their private information at the expense of other, less 1informed investors (Bamber et al. 1993). Finally, understanding the determinants of audit delays may provide some insights into audit efficiency, and possibly improve our understanding of market reactions to earning releases (Bamber et al. 1993; Ashton et al. 1989). Prior research reveals various client and audit firm factors that potentially influence audit delay. Among client-related variables, audit delay is a decreasing function of client size (Ashton et al. 1989; Newton and Ashton 1989; Bamber et al. 1993; Ettredge et al. 2000), of client industry, if the client is in a financial industry (Ashton et al. 1989; Newton and Ashton 1989; Bamber et al. 1993), and of client ownership concentration (Bamber et al. 1993). Audit delay is an increasing function of client extraordinary items (Ashton et al. 1 As a result, the SEC has phased-in accelerated deadlines for financial reporting over a three year period starting in 2003 (SEC 2002). 1989; Newton and Ashton 1989; Bamber et al. 1993), of client net losses (Ashton et al. 1989; Bamber et al. 1993; Ettredge et al. 2000), of client financial condition (Bamber et al. 1993), of modified auditor opinions on the financial statements (Ashton et al. 1989; Bamber et al. 1993; Ettredge et al. 2000), and of the client’s correction of previously reported interim earnings (Kinney and McDaniel 1993). Among auditor-related factors, audit delay is a decreasing function of the proportion of audit work accomplished at interim dates (Ashton et al. 1987; Knechel and Payne 2001), and of the percentage of total audit hours related to partner and manager time (Knechel and Payne 2001). Audit delay is an increasing function of a structured audit approach (Newton and Ashton 1989; Bamber et al. 1993; Ettredge et al. 2000) and of incremental audit effort (Knechel and Payne 2001). The effect of client internal control quality on audit delay is largely unstudied. Newton and Ashton (1989) use survey data to include client internal control quality as a factor in their study of audit delay, but they find no associations for public companies. It is likely that during the post-Enron period, with heightened expectations and greater scrutiny of auditor performance, auditors increase the scope of audit work when material weakness in the internal control over financial reporting (ICOFR) is identified. Recent reports in the business press assert that investors tend to punish companies with material weakness in their ICOFR, and that file their annual reports late (Leone 2005). As a result, weakness in ICOFR is more likely to result in audit delay in the post-SOX period. This paper extends prior studies by examining the effect of internal control quality on audit delay. Sarbanes-Oxley (SOX) Section 404 requires managers of public companies to assess and report on the effectiveness of their firms’ internal control structure, and on procedures for financial reporting. The auditor is required to attest and report on the 2 assessment made by management, and to provide his/her own assessment and report on the internal control over financial reporting. The core of Section 404 emphasizes the relationship between financial reports issued by a company and the underlying processes and controls that are an integral part of producing those reports. The 404 report is a huge task for both management and external auditors. If there is material weakness in the company’s ICOFR, auditors need to perform additional substantive procedures to compensate for the weakness identified (Doss 2004; Leech 2004). This study provides empirical evidence on whether material weakness in the ICOFR affects audit delay, and if it does, whether different types of material weakness have different effects. This study contributes to audit delay and internal control literature by reporting the following important findings. Our results indicate that the 404 report significantly increases audit delay. Even under the SEC’s new accelerated filing rule, companies filed their annual reports about 20 days later in 2004 compared to fiscal year 2003. Companies with material weakness in their ICOFR have longer delays than companies with effective ICOFR. Also, general material weaknesses cause longer delays than specific material weakness. Finally, the study provides evidence that not only weak ICOFR causes delays, but also the 404 procedure itself. The rest of the paper is organized into four sections. The second section comprises a literature review followed by hypotheses development. The third section discusses research method and data selection. The fourth section presents results, and the last section is conclusions and discussions. 3 BACKGROUND AND HYPOTHESES Background After Enron and the WorldCom scandal, Congress passed the landmark SOX Act in 2002 to restore investor confidence. One of the most significant provisions of the SOX is Section 404, which requires management to assess and report on the company’s ICOFR, and requires external auditors to attest and report on the assessment made by the management, as well as providing their own report on the ICOFR. The main purpose of SOX Section 404 is to satisfy the need of investors to have confidence not only in the financial reports issued by a company, but also in the underlying processes and controls that are an integral part of producing those reports. Based on the severity of control problems, Public Company Accounting Oversight Board (PCAOB) designates the following types of issues identified in the assessment of ICOFR: a control deficiency, a significant deficiency, or a material weakness (PCAOB Standard No. 2). It defines a control deficiency as “the design or operation of a control that does not allow management or employees, in the normal course of performing their assigned functions, to prevent or detect misstatements in a timely manner” (para. 8, p. 143). A significant deficiency is defined as “a control deficiency, or combination of control deficiencies, that adversely affects the company’s ability to initiate, authorize, record, process, or report external financial data reliably in accordance with generally accepted accounting principles such that there is more than a remote likelihood that a misstatement of the company’s annual or interim financial statements that is more than inconsequential will not be prevented or detected” (para. 9, p. 143). A material weakness is “a significant deficiency, or combination of significant deficiencies, that results in more than a remote 4 likelihood that a material misstatement of the annual or interim financial statements will not be prevented or detected” (para.10, p.144). When one or more material weaknesses exist in the ICOFR, the ICOFR is considered to be ineffective, and the management is required to disclose the material weaknesses in its 10-K. Hypotheses Under SOX Section 404, all public companies need to design, document, and analyze their ICOFR. In effect, they must create elaborate internal control procedural manuals and update them whenever processes change (Calabro 2004). Before reporting on the ICOFR, both management and external auditors must test the controls for design and operating effectiveness. The PCAOB requires auditors to conduct “walk-throughs” of important control stages. The PCAOB also limits how much external auditors can rely on the work of others, even though internal auditors may already have tested the processes (PCAOB, Standard No. 2). Moreover, because the auditors are required to test any controls that have materially significant impact on companies’ financial statements, they must be vigilant for weakness that may appear in a variety of processes, ranging from how journal entries are consolidated and adjusted, to what information technology controls are implemented to protect the company’s information systems (Calabro 2004). Therefore the requirements of section 404 reporting not only significantly increase management’s workload, but also the auditor’s. The extended audit work should lead to audit delays (Knechel and Payne 2001). Anecdotal evidence from the business press states that 404 requirements lengthened the time companies took to file their fiscal year 2004 reports, so the number of companies missing the filing deadline for annual reports has jumped in 2005 (Richardson 2005; Hadi 2005). 5 A recent SEC rule supports the argument that SOX 404 requirements are a burden to both companies and auditors. Under that rule, accelerated filers that have a fiscal year ending between and including Nov. 15, 2004 and Feb. 28, 2005, and that had public equity floats of less than $700 million at the end of their second fiscal quarter in 2004, will be granted 45 additional days to file their complete 404 reports (SEC, 2005 “Exemptive Order”). Based on the above arguments, our first hypothesis, in alternative form, is: H1: The new internal control over financial reporting (ICOFR) requirements increase audit delays, ceteris paribus. Few studies have examined the relationship between internal control quality and audit delays. Ashton et al. (1987) use survey data from an audit firm’s engagement partners to examine the association between the audit delay and the strength of companies’ internal controls. For their overall sample, they find that internal control quality has a significantly negative relation with audit delay. However, when they examine public and nonpublic companies separately, internal control quality has no significant relationship with audit delay for public companies, while the result still holds for nonpublic companies. Kinney and McDaniel (1993) examine the relation between audit delay and the correction of previously reported interim earnings. They argue that restatement of interim earnings can be a proxy for typically unobservable internal control quality problems because the existence of accounting error may indicate unexpectedly weak internal controls. Therefore, the discovery of accounting error might require the auditor to extend substantive tests in other areas. As a result, the audit report could be delayed (Kinney and McDaniel 1993). Their results show a significant relation between overstatement corrections and audit delay. 6 The existence of material weakness in the ICOFR potentially allows accounting errors to occur and to go undetected. In the post-Enron environment, auditors’ litigation risks have significantly increased, and auditors’ performances are under great scrutiny. Material weakness in the ICOFR indicates auditors need to extend their scope of work and perform additional substantive tests to compensate for the control weakness (Doss 2004; Leech 2004). The extended audit effort should lead to audit delay. Therefore, our second alternative hypothesis is: H2: Companies with material weaknesses in their ICOFR experience longer audit delays than companies with effective ICOFR, ceteris paribus. The material weakness is measured as an adverse auditor opinion on ICOFR. An Auditor is required to issue an adverse opinion on the ICOFR if one or more material weaknesses exist in the company’s ICOFR (PCAOB Standard No. 2). Standard No. 2 identifies two groups of controls: company-level controls and specific controls. Company-level controls refer to the controls that “might have a pervasive effect on the achievement of many overall objectives of the control criteria” (para.52, p. 163). We refer to control problems at the company level as “general material weaknesses”, which include situations such as an ineffective control environment, an ineffective audit committee, an inadequate internal audit or risk assessment function, and an ineffective financial reporting process (PCAOB Standard No. 2). Specific controls are those that are “designed to achieve specific objectives of the control criteria” (para.50, p. 163). We refer to controls at a specific level or account level as “specific material weaknesses”. These relate to control problems over transaction-level processes or specific account balances, such as inventory, accounts receivable, and legal proceedings (Doss, 2004). When an 7 account-level material weakness is identified, auditors can effectively audit around the material weakness by performing additional substantive procedures for the area in which the material weakness is identified. Therefore, the account-level material weakness should have a small impact on the audit delay. A company-level material weakness, however, impacts the financial reporting process so pervasively, that the scope of audit work must be expanded and audit effort must be increased in each area (PCAOB Standard No. 2). As a result, the audit delay should be more substantial in presence of a company-level general weakness. The above arguments lead to our third alternative hypothesis: H3: Companies with general material weaknesses in the ICOFR experience longer audit delays than companies with specific material weaknesses, ceteris paribus. RESEARCH METHOD AND DATA SELECTION Research Method We use the following approach (model 1) to test H1: + bYEAR + bSIZE + bFININD + bHIGHTECH+bOWN AUDELAY= a 1234 5 + bROA + bLEVERAGE + bGOCERN + bEXT + bSEGNUM 678910 + bLOSS + bRESTATE + bAFEE + bAOPIN + bAUDCHG 1112131415 14 +YEARCONTROLVAR，_ (1) ,i,i1 AUDELAY Refers to the number of calendar days from fiscal year-end to the date of auditor’s report. YEAR Equals 1 for fiscal year 2003; 0 for 2004. (used to test H1) 14 YEARCONTROLVAR，_ Represents the 14 interaction terms between YEAR ,i,i1 and 14 control variables. 8 The following regression model (model 2) is used to test H2 and H3: + bMWIC or GMWIC+ bSIZE + bFININD + bHIGHTECHAUDELAY = b01234 +bOWN + bROA + bLEVERAGE + bGOCERN + bEXT 56789 + bSEGNUM + bLOSS + bRESTATE + bAFEE + bAOPIN 1011121314 + bAUDCHG (2) 15 AUDELAY Refers to the number of calendar days from fiscal year-end to the date of auditor’s report. MWIC Identifies whether a company has a material weakness in its (used to test H2) ICOFR (1 = material weakness; 0 = otherwise). GMWIC Identifies whether a company has general material weakness (used to test H3) or specific material weakness in its ICOFR (1 = general material weakness; 0 = specific material weakness). The variables in the above two models are defined in table 1, and are discussed below. Based on previous research, we control for the effects of other factors that likely affect the audit delay: client size, client industry, client ownership concentration, client financial condition, client extraordinary items, client business complexity, client net losses, restatement, audit fee and auditor’s opinion on the financial statements. Client size (SIZE): the client preparation theory and client service theory suggest that larger clients have shorter audit delays because they can prepare their financial statements more quickly, and because auditors are more willing to complete the audit quickly to retain larger clients (Cullinan 2003). Consistent with these theories, several empirical studies find audit delay is negatively related to the client size (Ashton et al. 1989; Newton and Ashton 1989; Bamber et al. 1993; Ettredge et al. 2000). In this study, SIZE is 9 measured by the natural logarithm of the companies’ total assets. Based on the previous studies, we expect the coefficient of SIZE to be negative. Client industry type: We use two measures to examine the impact of client industry membership on the audit delay. One is whether the client belongs to a financial industry; the other is whether the client belongs to a high-tech industry. Companies in the financial industry hold little inventory or fixed assets (Bamber et al. 1993). The financial assets they do hold are easier to audit than non-financial assets. Therefore, audits of financial companies are expected to require less time than audits of non-financial companies. Studies Newton and Ashton (1989), and Bamber et al. (1993) consistently by Ashton et al. (1989), find that financial companies have shorter audit delays compared to non-financial companies. We use a dichotomous variable (FININD) to measure whether the client is in a financial industry (1 = financial industry; 0 = otherwise). The FININD is set to 1 for firms with SIC codes in the 6000-6999 range, and we expect the coefficient of FININD to be negative. Prior studies document that the high-technology industry is associated with high litigation risk (Behrendt et al. 1994; Byrnes 1995; Kasznik and Lev 1995; Bonner et al. 1998). With increased probability of litigation risk, auditors might increase their scope of work and cause longer audit delays. Following Kasznik and Lev (1995), we include the following industries in the high-technology categories: Drugs (SIC codes 2833-2836), Computers (3570-3577), Electronics (3600-3674), Programming (7371-7379) and R&D services (8731-8734). A dichotomous variable (HIGHTECH) is used to measure whether the client is in high-technology industry (1 = high-tech industry; 0 = otherwise). We expect the coefficient of HIGHTECH to be positive. 10 Client ownership concentration (OWN): auditor business risk affects the acceptable audit risk in the engagement, and hence the extent of audit work required (Bamber et al. 1993). The increased extent of audit work can result in audit delay (Knechel and Payne 2001). Audit literature suggests that the extent to which the client’s shares are widely held is one of the factors related to audit business risk (Brumfield et al. 1983; Arens and Loebbecke 1991; Arens et al. 2004). Using the average number of shares per shareholder as the proxy for the client’s ownership concentration, Bamber et al. (1993) empirically show it is negatively related to audit delay. Ettredge et al. (2000) also find a similar result for a quarterly earnings release lag. In this study, the client’s concentration of ownership is measured by the natural logarithm of client’s number of common shares outstanding divided by the number of common shareholders (i.e. natural logarithm of average number of shares per shareholder). Based on the previous studies, we expect the coefficient of OWN to be negative. Client’s financial condition: another factor related to audit business risk is the strength of the client’s financial condition (Brumfield et al. 1983; Arens and Loebbecke 1991; Arens et al. 2004). Bamber et al. (1993) find the audit delay is an increasing function of the vulnerability of the client’s financial position. Our measures of financial condition include ROA (net earnings / total assets), LEVERAGE (total debt / total assets), and GOCERN (1 = a company receives a going concern opinion; 0 = otherwise). The coefficients of LEVERAGE and GOCERN are expected to be positive, while the coefficient of ROA is expected to be negative. Client extraordinary items (EXT): extraordinary items, by definition, reflect nonrecurring events arising from something other than the company’s normal operations. 11 Empirical studies consistently show the presence of extraordinary items is positively associated with audit delay (Ashton et al. 1989; Newton and Ashton 1989; Bamber et al. 1993). A dichotomous variable is used to measure the presence of extraordinary items (1 = client reports an extraordinary item; 0 = otherwise). We expect the coefficient of EXT to be positive. Client business complexity (SEGNUM): the more complex the client business is, the more time the auditor should take to audit the financial reports. Prior research finds no relationship between client operational complexity and audit delay (Ashton et al. 1987; Bamber et al. 1993). In the post-SOX environment, this relationship may be significant given the extensive 404 requirements that must be undertaken in each different line of business. We use the number of companies’ reportable segments (SEGNUM) as the proxy for the client business complexity and expect the relationship to be positive. Client net losses (LOSS): losses raise concerns about the possible existence of material misstatements (AICPA 1992, AU 316), which may require additional audit work. Also, the possibility of lawsuits might increase in the face of a loss. A risk reverse auditor might conduct additional testing to reduce the risk. Previous empirical studies find clients’ net losses are positively related to the audit delay (Ashton et al. 1989; Bamber et al. 1993; Ettredge et al. 2000). We use a dichotomous variable to represent whether the client reports negative earnings for the current year. (1 = report negative earnings; 0 = otherwise). The coefficient of LOSS is expected to be positive. Restatement (RESTATE): correction of previously issued financial reports may significantly increase audit effort or increase auditor’s discussion with the client about what should be restated, how many reports should be restated, and what are the proper 12 disclosures. Kinney and McDaniel (1993) find significant increases in audit delay for firms correcting misstatements in previously reported interim earrings. RESTATE is measured by a dichotomous variable (1= restatement in the current year; 0 = otherwise), and its coefficient is expected to be positive. Audit fee (AFEE): auditors receiving higher audit fees probably allocate more audit manpower and other resources. Thus, the audit work might be done more quickly. However, the higher fees could also be due to complexity of the audit, and complexity increases audit delay. Therefore, we do not expect a specific relationship between audit fee and audit delay. Auditor’s opinion on the financial statements (AOPIN): Modified opinions are usually issued after the auditor has spent considerable time and effort in conducting additional audit procedures (Bamber et al. 1993), or are issued after lengthy negotiations with the management who try to avoid the modifications. Thus, a modified opinion on the financial statements should be a factor delaying the audit report. Several empirical studies support this argument (Ashton et al. 1989; Bamber et al. 1993; Ettredge et al. 2000). In this study, the auditor’s opinion on the financial statements is represented by a dichotomous variable (1= modified opinions for the current year, other than going concern opinions; 0 = otherwise). We expect the coefficient of AOPIN to be negative. Auditor change (AUDCHG): when a company changes its auditor, the new auditor will usually take some time to get to know the company’s business and communicate with the predecessor auditor. The initial versus repeat engagement is also considered as an important factor in assessing inherent risk. With all the extra assessment, the financial statement reporting lag is expected to increase. We use a dichotomous variable to measure AUDCHG (1= a client change auditors in the current year; 0 = otherwise). 13 Data Selection The sample for testing H1 and H2 includes companies that filed 404 reports from January 2005 to June 2005. The sample for testing H3 includes only the subset of companies with material weaknesses in the ICOFR. The companies’ ICOFR status is identified using Audit Analytics Database. Our initial sample size is 3,098, including 2,706 companies with effective ICOFR, 390 companies with material weakness, and 2 companies with no disclosure. We then require the sample companies to have the necessary financial statement variables available from Compustat. This procedure yields a final sample of 2,465 firm-observations for fiscal year 2003, and 2,350 firm-observations for fiscal year 2004, including 2,041 with effective ICOFR and 309 with material weakness. EMPIRICAL RESULTS Descriptive Statistics Table 2 reports descriptive statistics. Panel A shows the audit delay in 2004 is significantly longer than that in 2003. The mean delay in 2003 is 50 days, while the mean delay in 2004 is 70 days. This supports H1. The audit delay for companies with material weakness in their ICOFR is also significantly longer than companies with effective ICOFR. While the mean delay for companies with effective ICOFR is 67 days, the mean delay for companies with material weakness is 86 days, which supports H2. In Panel B, three sets of data are reported: the full sample of 2,350 companies that reported ICOFR in 2004, the sub-sample of 309 companies that have material weakness in their ICOFR, and the sub-sample of 2,041 companies that have effective ICOFR. Univariate analysis indicates that the companies with material weakness in their ICOFR are significantly smaller, are more 14 likely to be in high-tech industries, had higher concentration of ownership, lower ROA, reported more losses, had more restatements, were more likely to change auditors, and were more likely to get modified auditor opinions other than for going concern. Companies having material weaknesses are less likely to be in the financial industry. There are no significant differences for other variables (LEVERAGE, GOCERN, EXT, SEGNUM and AFEE). Panel C in table 2 also reports three sets of data: the full sample of 309 companies that have material weakness in their ICOFR in 2004, the sub-sample of 108 companies that have general material weakness, and the sub-sample of 201 companies that have specific material weakness. Univariate analysis suggests that companies with general material weakness are significantly smaller, are more likely to be in the high-tech industries, had lower ROA, reported more losses, paid smaller audit fees, were less likely to restate and were less likely to receive modified opinions, other than for going concern. Other variables do not differ (FININD OWN, LEVERAGE, GOCERN, EXT, SGENUM and AUDCHG). Regression Results To test whether the intercept differs using pre-404 and post-404 data, based on model (1), we add a dummy variable representing the YEAR (1 = fiscal year 2003, 0 = fiscal year 2004), and interact each independent variable with YEAR (Model 1a). If there is no difference between the intercepts for the two years, the coefficient of YEAR should be zero. Table 3 reports the results of the model. The coefficient of YEAR is -19.643 and is significant (p = 0.000), suggesting the new 404 requirements increased audit delay for fiscal year 2004 by almost 20 days, which provides support for H1. When we examine the two years separately (Model (1b) and Model (1c)), we find the following: consistent with 15 previous studies, companies with longer delays are more likely to be smaller, have higher leverage, report more losses, are more likely to restate, and receive more modified auditor opinions in both years. For the new variables not used in prior studies, consistent with our expectations, auditor changes are more likely to increase delays. The coefficient of audit fee is significantly positive, which may indicate that greater complexity of companies’ operations leads to more audit hours. Contrary to prior research, companies in the financial industry are more likely to have longer delays in both years. And contrary to prior research, higher concentration of ownership is associated with longer delay. This positive association is significantly higher in Year 2004 compared to Year 2003, based upon the significant coefficient of the interaction between YEAR and OWN. Corporate governance literature (e.g. Jensen 1993; Beasley 1996) suggests that higher concentration of ownership, such as the existence of large block shareholders, is associated with stronger internal controls. It is possible that since companies with higher concentration of ownership are better governed, they are more careful in fulfilling the SOX requirements, and therefore take longer time. Contrary to our expectation, companies in the high-tech industries have shorter delays. Going concern opinions are positively associated with audit delay. Extraordinary items, and the number of segments, are positively related to audit delay only in 2003. Table 4 reports the results for testing the relationship between audit delay and internal control quality. Model (2a) uses the full available sample of companies that reported their ICOFR status, and Model (2b) uses a reduced sample of companies with material weakness in their ICOFR. The results of Model (2a) show companies with material weakness in their ICOFR experience significantly longer delays (about 15 days more, p = 0.000), providing support to H2. Among other control variables, the results 16 remain the same as those of Model (1c), except RESTATE and AUDCHG become insignificant, and LOSS becomes only marginally significant. The change of significance for the above three variables, after adding MWIC to the model, suggests they are associated with MWIC. Further analysis shows 22% of the companies reporting negative net earnings had material weaknesses in the ICOFR. 43% of the companies that restated their financial reports also reported material weakness in the ICOFR. 31% of the companies that changed their auditors had material weakness in the ICOFR. The results of Model (2b) suggest companies with general material weakness in their ICOFR experience longer audit delays than companies with specific material weakness (about 7 days more, p = 0.002), which supports H3. Among control variables, ROA now becomes negative and significant, suggesting for companies with material weakness in their ICOFR, higher ROA decreases audit delays. SIZE and the high-tech industry indicator are still negatively related to delay, and audit fees are positively related to delay. Other variables are not significant. Further Analysis Since material weakness is significantly related to audit delay, it is interesting to see whether the impact varies along with the type of weaknesses. We conduct an additional analysis on the relationship between the specific types of material weakness and the audit delays. Starting with model (2a), we replace MWIC with variables representing specific types of material weaknesses. Based on the Committee of Sponsoring Organization’s (COSO) framework, we categorize internal control material weakness into eight major types: Personnel (PERSONNEL), Process and Procedure (PROCESS), Documentation (DOCUMENT), Segregation of Duties (SEGREGATE), Information Systems Process 17 (ISPROCESS), Risk Assessment (RISKASSESS), Closing Process (CLOSING), and Control Environment (CONTRENV). Table 5 shows the results of the relationship between audit delay and specific types of internal control weakness. Companies with internal control problems in personnel, process and procedure, segregation of duties, and closing process have the longest delays. As documented above, when we add MWIC into the model for companies in 2004, the relations between audit delays and some variables become either weaker or insignificant, which shows strong correlations between those variables and MWIC. To further investigate this phenomenon, we regress MWIC on the rest of the independent variables in model (2a), for both a “Late-filer” group and an “On-Time-filer” group, to see whether the two groups have different determinants of material weakness in ICOFR. 2 Companies are categorized as late-filers if their audit report lags are more than 75 days. Table 6 reports the results. In general, companies with internal control material weakness are significantly smaller, receive less going concern opinions, report more losses, are more likely to restate financial statements, have higher audit fees, received more modified audit opinions, and are more likely to change auditors. When we examine the Late-filer group and On-time-filer groups separately, we find companies with internal control material weakness in the Late-filer group are smaller, have higher concentration of ownership, are more likely to report extraordinary items, report more losses, are more likely to restate financial statements, have higher audit fees, and are more likely to change auditors. Companies with internal control material weakness in the On-time-filer group are smaller, are more likely to restate financial statements, have higher audit fees, and are more 2 In 2002 the SEC issued a ruling on Acceleration of Periodic Report Filing Dates (SEC 2002), requiring companies to file Form 10-Ks within 75 days after fiscal year end for fiscal year 2004. 18 likely to change auditors. However, extraordinary items, ownership concentration and loss are not significant for the non-late-filer group. While other recent studies of the determinants of weakness in ICOFR (Ge and McVay 2005; Doyle et al. 2005) tend to find similar results (e.g. companies with weak controls are smaller, report more losses, are more complex, have more special items), our results further suggest that some of the significant results in prior studies (i.e. report more losses, have more special items) are mainly driven by those companies with material weakness in their ICOFR that file their financial reports late. DISCUSSION AND CONCLUSIONS This study examines the relation between audit delay and internal control quality in the SOX era. We find that both the SOX 404 procedure itself and weak ICOFR lengthen audit delay. Our specific findings are summarized as follow. First, to fulfill the heavy requirements of the SOX 404 report, companies experience significantly longer audit delay. According to our results, the length of time from companies’ fiscal year-end to the date the auditors sign their reports has increased by about 20 days from fiscal year 2003 to 2004. Second, material weakness in ICOFR causes still longer audit delays. On average, the presence of material weakness leads to an increase of about 15 days in audit delay. Third, general material weaknesses cause longer delays than specific material weakness. The audit delay for a company with general material weakness is roughly 8 days longer than that for a company with specific material weakness. Additional analyses suggest that the impact of internal control quality upon audit delay varies along with the specific types of control weakness defined under the COSO framework. Specifically, companies with internal control problems in personnel, process 19 and procedure, segregation of duties, and closing process tend to experience longer delays. Finally, we examine whether the determinants of material weakness differ between companies with longer delays (late-filer group) and companies with shorter delays (non-late-filer group). We find that ownership concentration, extraordinary items, and losses are significant determinants of material weakness for the firms that file their Form 10-Ks late, but not for firms that file on time. Some results for the control variables are also interesting enough to be noted here. Consistent with prior literature, we document that companies with longer delays are more likely to be smaller, have higher leverage, report more losses, are more likely to restate earnings, and receive more modified auditor opinions. Different from prior literature, our results suggest that companies in the financial industry are more likely to have longer delays, and that companies with higher concentration of ownership are more likely to have longer delays. In addition, we also experiment with a few variables not studied by prior literature. Consistent with our expectations, auditor changes are more likely to increase delays. Higher audit fee is associated with longer delays, which may indicate that greater complexity of companies’ operations leads to more audit hours. Finally, contrary to our expectation, companies in the high-tech industries have shorter delays. To summarize, empirical evidence provided in this study sheds light on whether and how audit delays are affected by internal control quality in SOX era. It helps us better understand the impact of the SOX act upon the auditing process, and the timeliness of financial reporting. To ensure the timeliness of the study, we only examine one year’s data in the post-SOX period. Future research can study whether and how the impact changes as more data become available. 20 21 Table 1 Definition of Variables and Expected Signs Variable Expected Definition Sign AUDELAY Number of calendar days from fiscal year-end to date of the auditor’s report. MWIC + Whether a client has material weaknesses in the ICOFR (1 = material weaknesses; 0 = otherwise). GMWIC + Whether the material weaknesses are general or specific (1 = general material weakness; 0 = specific material weakness). SIZE - The size of the client. It is measured by the natural logarithm of total assets. FININD - The client’s industry. 1 = financial industry; 0 = otherwise. HIGHTECH + The client’s industry. 1 = high-tech industry, 0 = otherwise. OWN - The client’s concentration of ownership. It is measured by natural logarithm of the client’s number of common shares outstanding divided by the number of common shareholders. ROA - Net earnings divided by total assets. LEVERAGE + Total debt divided by total assets. GOCERN + Whether the client receives a going concern opinion (1 = going concern opinion; 0 = otherwise). EXT + Whether the client reports extraordinary items for the current year. 1 = client reports an extraordinary item; 0 = otherwise. SEGNUM + Number of reportable segment of a client. LOSS + Whether the client reports negative earnings for the current year. (1 = report negative earnings; 0 = otherwise). RESTATE + Whether the client restated its financial report for the current year (1 = restated in the current year; 0 = otherwise). AFEE ? Total audit fee for the current year. It is measured by the natural logarithm of total audit fee. AOPIN + Auditor’s opinion on the financial statements. (1= modified opinions other than going concern; 0 = unmodified opinion). AUDCHG + Whether a client changed auditor during the current year (1= client changed auditor, 0 = otherwise). 22 Table 2 Descriptive Statistics for Sample Companies that Reported ICOFR in 2005. Panel A: Descriptive Statistics for Audit Delay (mean in days) 2004 2004 Material Effective 2004 2003 t-stat p-value Weakness ICOFR t-stat p-value N = 2,350 2,465 309 2,041 Variable: AUDELAY 69.9 50.0 39.716 0.000 *** 85.8 67.4 20.590 0.000 *** Panel B: Descriptive Statistics for Companies that Reported ICOFR in 2004 Material Weakness Effective ICOFR Total sample N = 309 2,041 2,350 Variables: Mean Std. dev. Mean Std. dev. t-stat. p-value Mean Std. dev. MWIC 13% SIZE 19.982 1.705 20.672 1.885 -6.074 0.000 *** 20.581 1.876 FININD 9% 17% -3.263 0.001 *** 16% HIGHTECH 27% 22% 1.945 0.052 * 23% OWN 10.449 1.937 10.137 1.852 2.749 0.006 *** 10.178 1.866 ROA -0.041 0.259 0.026 0.384 -2.970 0.003 *** 0.017 0.371 LEVERAGE 0.226 0.264 0.228 0.229 -0.114 0.909 0.228 0.234 GOCERN 1% 1% 0.246 0.806 1% EXT 3% 4% -0.772 0.440 4% SEGNUM 2.407 1.757 2.505 1.793 -0.890 0.373 2.492 1.788 LOSS 39% 21% 6.884 0.000 *** 23% RESTATE 55% 11% 20.838 0.000 *** 17% AFEE 14.032 1.112 14.025 1.099 0.110 0.912 14.026 1.100 AOPIN 44% 34% 3.690 0.000 *** 36% AUDCHG 16% 6% 6.930 0.000 *** 7% (continued) 23 Table 2 (continued) Panel C: Descriptive Statistics for Companies with Material Weakness in ICOFR, 2004 General Weakness Specific Weakness Total sample N = 108 201 309 Variables: Mean Std. dev. Mean Std. dev. t-stat. p-value Mean Std. dev. GMWIC 35% SIZE 19.688 1.708 20.140 1.686 -2.237 0.026 ** 19.982 1.705 FININD 12% 8% 1.171 0.243 9% HIGHTECH 33% 23% 1.886 0.060 * 27% OWN 10.575 1.795 10.382 2.010 0.836 0.404 10.449 1.937 ROA -0.082 0.259 -0.020 0.257 -2.016 0.045 ** -0.041 0.259 LEVERAGE 0.221 0.244 0.229 0.275 -0.272 0.786 0.226 0.264 GOCERN 2% 0% 1.156 0.248 1% EXT 5% 2% 1.013 0.312 3% SEGNUM 2.514 1.840 2.350 1.712 0.779 0.437 2.407 1.757 LOSS 45% 35% 1.820 0.070 * 39% RESTATE 46% 60% -2.357 0.019 ** 55% AFEE 13.812 1.275 14.151 0.998 -2.582 0.010 *** 14.032 1.112 AOPIN 32% 51% -3.091 0.002 *** 44% AUDCHG 16% 16% -0.154 0.878 16% See Table 1 for variable definitions *** p-values are significant at .01 level, ** p-values are significant at .05 level, * p-values are significant at .10 level. 24 Table 3 Regression Models of Audit Delay Difference between 2003 and 2004 Test of H1, Model (1a) Year = 2003, Model (1b) Year = 2004, Model(1c) Variables: Estimate t-value p-value Estimate t-value p-value Estimate t-value p-value Intercept 97.454 18.095 0.000 *** 77.810 14.949 0.000 *** 97.454 20.034 0.000 *** YEAR (1=03,0=04) -19.643 -2.723 0.007 *** SIZE -4.653 -13.488 0.000 *** -4.106 -11.459 0.000 *** -4.653 -14.933 0.000 *** SIZE * YEAR 0.547 1.146 0.252 FININD 3.071 2.994 0.003 *** 5.600 5.236 0.000 *** 3.071 3.315 0.001 *** FININD * YEAR 2.530 1.778 0.076 * HIGHTECH -3.174 -3.510 0.000 *** -3.321 -3.454 0.001 *** -3.174 -3.886 0.000 *** HIGHTECH * YEAR -0.147 -0.116 0.908 OWN 0.401 2.165 0.030 ** -0.067 -0.338 0.735 0.401 2.397 0.017 ** OWN * YEAR -0.469 -1.794 0.073 * ROA 0.517 0.518 0.604 0.873 0.914 0.361 0.517 0.574 0.566 ROA * YEAR 0.357 0.268 0.789 LEVERAGE 6.699 4.262 0.000 *** 13.945 8.339 0.000 *** 6.699 4.719 0.000 *** LEVERAGE * YEAR 7.246 3.290 0.001 *** GOCERN 5.562 1.493 0.136 9.791 2.995 0.003 *** 5.562 1.653 0.099 * GOCERN * YEAR 4.228 0.882 0.378 EXT 0.420 0.238 0.812 2.595 2.403 0.016 ** 0.420 0.264 0.792 EXT * YEAR 2.175 1.074 0.283 SEGNUM -0.208 -0.954 0.340 0.694 3.002 0.003 *** -0.208 -1.056 0.291 SEGNUM * YEAR 0.902 2.959 0.003 *** LOSS 2.236 2.364 0.018 ** 2.660 2.889 0.004 *** 2.236 2.617 0.009 * LOSS * YEAR 0.425 0.334 0.738 RESTATE 6.005 6.551 0.000 *** 8.956 3.793 0.000 *** 6.005 7.253 0.000 *** RESTATE * YEAR 2.952 1.249 0.212 AFEE 4.338 7.930 0.000 *** 3.678 6.698 0.000 *** 4.338 8.780 0.000 *** AFEE * YEAR -0.661 -0.886 0.376 AOPIN 1.885 2.460 0.014 ** 2.475 3.111 0.002 *** 1.885 2.724 0.007 *** AOPIN * YEAR 0.590 0.557 0.578 AUDCHG 2.287 1.710 0.087 * 4.655 2.547 0.011 ** 2.287 1.893 0.059 * AUDCHG * YEAR 2.369 1.101 0.271 N = 4,815 2,465 2,350 F-statistic 86.617 0.000 *** 21.582 0.000 *** 31.225 0.000 *** 2Adj. R 0.340 0.105 0.153 25 Table 4 Regression Models of Relationship between Audit Delay and Internal Control Quality Test of H2, Model(2a) Test of H3, Model (2b) Companies that Reported ICOFR Companies with Material Weakness in ICOFR Variables: Estimate t-value p-value Estimate t-value p-value Intercept 94.164 20.309 0.000 *** 103.662 5.999 0.000 *** MWIC 14.857 15.531 0.000 *** N.A. N.A. GMWIC N.A. N.A. 7.559 3.296 0.001 *** SIZE -3.914 -13.024 0.000 *** -5.594 -4.544 0.000 *** FININD 2.688 3.046 0.002 *** 0.320 0.075 0.940 HIGHTECH -3.022 -3.886 0.000 *** -7.373 -2.825 0.005 *** OWN 0.305 1.914 0.056 * 0.559 0.979 0.328 ROA 0.886 1.033 0.302 -13.127 -2.028 0.044 ** LEVERAGE 6.426 4.754 0.000 *** 7.077 1.534 0.126 GOCERN 6.799 2.121 0.034 ** 3.590 0.248 0.804 EXT 0.578 0.381 0.703 2.722 0.446 0.656 SEGNUM -0.225 -1.201 0.230 -0.495 -0.722 0.471 LOSS 1.512 1.856 0.064 * -1.980 -0.728 0.467 RESTATE 1.191 1.406 0.160 0.029 0.012 0.990 AFEE 3.523 7.442 0.000 *** 6.125 3.650 0.000 *** AOPIN 1.278 1.936 0.053 * 0.872 0.373 0.710 AUDCHG 0.165 0.142 0.887 3.549 1.214 0.226 N = 2,350 N = 309 F-statistic 48.226 0.000 *** 4.381 0.000 *** 2 0.232 0.142 Adj. R 26 Table 5 Regression Model of Relationship between Audit Delay and Different Types of Internal Control Weakness Variables: Estimate t-value p-value Intercept 95.440 20.422 0.000 *** SIZE -3.980 -13.199 0.000 *** FININD 2.476 2.793 0.005 *** HIGHTECH -3.286 -4.196 0.000 *** OWN 0.315 1.968 0.049 ** ROA 0.934 1.082 0.279 LEVERAGE 6.648 4.892 0.000 *** GOCERN 5.607 1.736 0.083 * EXT -0.030 -0.020 0.984 SEGNUM -0.259 -1.375 0.169 LOSS 1.438 1.755 0.079 * RESTATE 2.256 2.659 0.008 *** AFEE 3.543 7.442 0.000 *** AOPIN 1.576 2.377 0.018 ** AUDCHG 0.525 0.450 0.652 PERSONNEL 4.887 2.793 0.005 *** PROCESS 8.375 6.921 0.000 *** DOCUMENT -1.495 -0.603 0.546 SEGREGATE 4.120 2.036 0.042 ** ISPROCESS 1.681 0.489 0.625 RISKASSESS 8.352 1.326 0.185 CLOSING 10.989 5.670 0.000 *** CONTRENV -2.030 -0.633 0.527 N = 2,350 F-statistic 32.124 0.000 *** 2 0.226 Adj. R 27 Table 6 Logistic Regression Model of the Probability of Disclosing a Material Weakness in the ICOFR for the Late-filer Group and Non-late-filer Group, 2004 Dependent variable = 1 if companies have material weakness in ICOFR, 0 otherwise All Companies that Reported ICOFR Late-filer group On-time-filer group Variables: Estimate p-value Estimate p-value Estimate p-value SIZE -0.598 0.000 *** -0.296 0.046 ** -0.396 0.000 *** FININD 0.154 0.533 0.065 0.886 -0.014 0.966 HIGHTECH -0.104 0.556 0.231 0.501 0.237 0.314 OWN 0.084 0.034 ** 0.154 0.047 ** 0.041 0.429 ROA -0.224 0.298 -0.422 0.240 0.244 0.517 LEVERAGE 0.228 0.469 0.131 0.831 0.150 0.731 GOCERN -1.137 0.152 -1.337 0.224 -19.023 0.998 EXT -0.185 0.652 2.046 0.084 * -1.184 0.121 SEGNUM 0.033 0.478 0.071 0.405 0.024 0.698 LOSS 0.397 0.023 ** 0.674 0.045 ** 0.274 0.261 RESTATE 2.255 0.000 *** 1.835 0.000 *** 2.341 0.000 *** AFEE 0.605 0.000 *** 0.423 0.028 ** 0.370 0.025 ** AOPIN 0.453 0.004 *** 0.363 0.203 0.341 0.107 AUDCHG 1.111 0.000 *** 1.278 0.001 *** 0.705 0.023 ** Intercept -0.306 0.793 -3.087 0.164 -1.177 0.447 N = 2,350 336 2,014 Correctly classified 88% 71% 93% Chi-Square 426.717 0.000 *** 98.192 0.000 *** 198.933 0.000 2 Nagelkerke R 0.308 0.340 0.236 28 Reference: American Institute of Certified Public Accountants (AICPA). 1992. AICPA Professional Standards. Volume 1: U.S. Auditing Standards. Chicago: Commerce Clearing House. Arens, A., R. Elder, and M. Beasley. 2004. Auditing and Assurance Services: An Integrated Approach. Tenth Edition. Upper Saddle River, New Jersey: Prentice Hall. Arens, A., and J. Loebbecke. 1991. Auditing: An Integrated Approach. Fifth Edition. Englewood Cliffs, New Jersey: Prentice Hall. Ashton, Robert H., Paul R. Graul, and James D. Newton. 1989. Audit Delay and the Timeliness of Corporate Reporting. Contemporary Accounting Research 5, no. 2:657-673. Ashton, Robert H., John J. Willingham, and Robert K. Elliott. 1987. An empirical analysis of audit delay. Journal of Accounting Research v25 n2, 275-292 (18 pages). Bamber, E. M., Linda S. Bamber, and Michael P. Schoderbek. 1993. Audit structure and other determinants of audit report lag: An empirical analysis. Auditing: A Journal of Practice & Theory v12n1, 1-23 (23 pages). Beasley, M.S. 1996. An empirical analysis of the relation between the board of director composition and financial statement fraud. The Accounting Review v71 n3, 443-465. Brumfield, C., R. Elliott, and P. Jacobson. 1983. Business risk and the audit process. The Journal of Accountancy (April): 60-68. Calabro, L., 2004. Looking under the hood. CFO Magazine. May 01. Chambers, Anne E., and Stephen H. Penman. 1984. Timeliness of Reporting and the Stock Price Reaction to Earnings Announcements. Journal of Accounting Research v22n1, 21-47 (27 pages). Cullinan, C. P. 2003. Competing size theories and audit lag: evidence from mutual fund audits. Journal of American Academy of Business. v3n1/2. 183-193. Doss, M. 2004. Section 404 reports on internal control: impact on ratings will depend on nature of material weakness reported (October). Moody’s Special Comment. Moody’s Investors Service, Inc. 29 Doyle, J.; W. Ge, and S. McVay 2005. Determinants of weakness in internal control over financial reporting and the implications for earnings quality. Working Paper, University of Utah. Ettredge, Mike, Dan Simon, David B. Smith, and Mary Stone. 2000. Would Switching to Timely Reviews Delay Quarterly and Annual Earnings Releases? Review of Quantitative Finance & Accounting 14, no. 2:111-130. Ge, W. and S. McVay 2005. The disclosure of material weakness in internal control after the Sarbanes-Oxley Act. Accounting Horizon, forthcoming. Givoly, Dan, and Dan Palmon. 1982. Timeliness of annual earnings announcements: some empirical evidence. Accounting Review v57n3, 486-508 (23 pages). Hadi, M. 2005. More companies are filing late. Wall Street Journal, April 11:C3. Hakansson, Nils H. 1977. Interim disclosure and public forecasts - an economic analysis and a framework for choice. Accounting Review v52 n2, 396-416. Jensen, M. C. 1993. The modern industrial revolution, exit, and the failure of internal control systems. The Journal of Finance XLVIII (July): 831-880. Kinney, William R.,Jr, and Linda S. McDaniel. 1993. Audit delay for fims correcting quarterly earnings. Auditing: A Journal of Practice & Theory v12n2, 135-142 (8 pages). Knechel, W. R., and Jeff L. Payne. 2001. Additional evidence on audit report lag. Auditing 20, no. 1:137-146. Kross, William, and Douglas A. Schroeder. 1984. An Empirical Investigation of the Effect of Quarterly Earnings Announcement Timing on Stock Returns. Journal of Accounting Research v22n1, 153-176 (24 pages). Leech, T. 2004. Moody’s questions ability to audit around deficiencies. Compliance Week. Dec. 7, 2004. Leone, M. 2005. How markets punish material weaknesses. CFO.com, July 21. Newton, James D., and Robert H. Ashton. 1989. The Association Between Audit Technology and Audit Delay. Auditing: A Journal of Practice & Theory 8, no. Supplement:22-37. Public Company Accounting Oversight Board (PCAOB) 2004. AUDITING 30 STANDARD No. 2 – An audit of internal control over financial reporting performed in conjunction with an audit of financial statements. Richardson, K. 2005. Big board proposes crackdown on late filers. Wall Street Journal, February, 11: C3. U.S. Securities and Exchange Commission (SEC). 2002. Acceleration of periodic report filing dates and disclosure concerning website access to reports. September 5. U.S. Securities and Exchange Commission (SEC). 2005. Exemptive order on management’s report on internal control over financial reporting and related auditor report. Frequently Asked Questions. January 21. 31