delay
-
Upload
greatyanto -
Category
Documents
-
view
4 -
download
0
description
Transcript of delay
Managerial Auditing JournalDo audit delays affect client retention?Vivek Mande Myungsoo Son
Article information:To cite this document:Vivek Mande Myungsoo Son, (2011),"Do audit delays affect client retention?", Managerial Auditing Journal,Vol. 26 Iss 1 pp. 32 - 50Permanent link to this document:http://dx.doi.org/10.1108/02686901111090826
Downloaded on: 02 September 2015, At: 21:37 (PT)References: this document contains references to 45 other documents.To copy this document: [email protected] fulltext of this document has been downloaded 2492 times since 2011*
Users who downloaded this article also downloaded:Mai Dao, Trung Pham, (2014),"Audit tenure, auditor specialization and audit report lag", ManagerialAuditing Journal, Vol. 29 Iss 6 pp. 490-512 http://dx.doi.org/10.1108/MAJ-07-2013-0906Sharad Asthana, (2014),"Abnormal audit delays, earnings quality and firm value in the USA", Journal ofFinancial Reporting and Accounting, Vol. 12 Iss 1 pp. 21-44 http://dx.doi.org/10.1108/JFRA-09-2011-0009Shahed Imam, Zahir Uddin Ahmed, Sadia Hasan Khan, (2001),"Association of audit delay and audit firms’international links: evidence from Bangladesh", Managerial Auditing Journal, Vol. 16 Iss 3 pp. 129-134http://dx.doi.org/10.1108/02686900110385669
Access to this document was granted through an Emerald subscription provided by emerald-srm:501757 []
For AuthorsIf you would like to write for this, or any other Emerald publication, then please use our Emerald forAuthors service information about how to choose which publication to write for and submission guidelinesare available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.comEmerald is a global publisher linking research and practice to the benefit of society. The companymanages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well asproviding an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committeeon Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archivepreservation.
*Related content and download information correct at time of download.
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Do audit delays affect clientretention?
Vivek Mande and Myungsoo SonCalifornia State University, Fullerton, California, USA
Abstract
Purpose – The purpose of this study is to examine whether lengthy audit delays lead to auditorchanges in the subsequent year. The paper hypothesizes that a lengthy interaction between clients andtheir auditors reflects high audit risk factors relating to management integrity, internal controls, andthe financial reporting process. It argues that auditors are more likely to drop clients with long auditdelays because they would like to avoid these types of audit risks.
Design/methodology/approach – Using logistic regressions, the paper first tests whether alengthy audit delay leads to an auditor change. It then examines whether as audit delays increase,auditor changes are more likely to be downward than lateral.
Findings – The results support the hypothesis that Big N auditor-client realignments occur followinglong audit delays. Further, as the length of the delay increases, the paper finds that there are moredownward changes.
Research limitations/implications – An implication of our study is that a long audit delayrepresents a publicly observed proxy for the presence of audit risk factors that lead to an auditor change.
Practical implications – This study suggests that all else constant, investors should consider alengthy audit delay as indicating that there has been deterioration in the quality of the client-auditorinteraction. An audit delay also presents an observable proxy for successor auditors to consider whileevaluating risks associated with a new client.
Originality/value – The results of our study increase our understanding of how Big N auditorsmanage their client portfolios to mitigate their exposure to risk factors.
Keywords Auditing, Risk assessment, Customer retention
Paper type Research paper
1. IntroductionAgainst the backdrop of Sarbanes-Oxley (SOX), this study examines whether lengthyaudit delays, used as a proxy for audit risk, increase the likelihood of Big N auditorresignations in the subsequent year.
In the years immediately following the passage of the SOX Act, audit firmsexperienced sharp increases in the demand for assurance staff (Rama and Read, 2006;Cenker and Nagy, 2008). Section 404 of SOX, in particular, led to shortages of skilledauditors (at least in the short run) and increases in audit fees (GAO, 2006). It isestimated that a typical audit took 40-60 percent more time to complete in the post-SOXthan in the pre-SOX years (Koehn and DelVecchio, 2006).
The paucity of staff resources was more pronounced in the Big N firms becauseSection 404 of SOX initially affected only the large (accelerated) filers, whose auditorswere mostly the Big N firms. Facing severe staffing constraints, the Big N auditorsadopted a more conservative stance on client retention by resigning from high riskengagements (Rama and Read, 2006)[1], and, as a result, Non-Big N audit firms gainedsignificant market share. Ettredge et al. (2007), for example, find that the majority of
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0268-6902.htm
MAJ26,1
32
Received 7 October 2009Revised 30 April 2010Accepted 21 June 2010
Managerial Auditing JournalVol. 26 No. 1, 2011pp. 32-50q Emerald Group Publishing Limited0268-6902DOI 10.1108/02686901111090826
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Big N clients who changed their auditors after SOX hired non-Big N firms as their newauditors – a pattern contrary to pre-SOX trends in auditor changes (Landsman et al.,2009; Turner et al., 2005; Public Accounting Report, 2004)[2].
Audit firms conduct routine risk reviews of their client portfolios and resign fromclients regarded as high risk (Jones and Raghunandan, 1998; Johnstone and Bedard,2004; Winograd et al., 2000; Bell et al., 2002; Beneish et al., 2005; Krishnan andKrishnan, 1997; Shu, 2000). The above cited studies use litigation risk and financialdistress variables to proxy for risk factors considered by auditors in their clientretention decisions. The proxy variables are computed using financial and market datathat are publicly available on research data bases.
An exception to this stream of research is a study by Johnstone and Bedard (2004)who examine auditor resignations using proprietary auditors’ assessments of theirclients’ audit risks[3]. They report that risk factors relating to clients’ internal controls,financial reporting quality, and management integrity provide more useful measuresof audit risks than do financial risk factors. Johnstone and Bedard argue that theirproprietary proxies are more comprehensive risk measures than financial riskmeasures because they include qualitative risk factors. They find that their riskmeasures are more strongly associated with auditors’ client retention decisions thanare financial and litigation risk variables. Their findings suggest that more research isneeded on how auditors incorporate risks relating to client controls, financial reportingquality, and management integrity in their client retention decisions.
This paper argues that audit delays, the time between the fiscal year end and theaudit completion date, can be a potential candidate for measuring risk factors relatingto clients’ internal controls, financial reporting quality, and management integrity.A lengthy audit delay often occurs due to problems in the audit, disagreementsbetween the auditor and client on accounting issues, and/or a general deterioration inthe quality of auditor-client interaction. A long delay could also occur when a clientfirm has high inherent and/or control risk requiring more work by the auditor (Ireland,2003). Thus, a lengthy audit delay could represent an observable proxy for the abovenon-public audit risk factors that affect client retention decisions.
To investigate whether client risk factors are associated with Big N auditorresignations, we estimate a logistic regression that models auditor resignations as afunction of audit delays and control variables identified by prior research. We alsoconsider the direction of auditor switches: lateral (Big N to Big N) or downward (Big N tonon-Big N). Finding another Big N auditor to replace the current auditor might not be afeasible option for all client firms, in particular for client firms with high audit risks.As the severity of audit risks increases, these client firms may have no alternative but tofind a non-Big N auditor (i.e. downward change). Therefore, we expect that there will bean ordered correlation between audit risks (i.e. audit delays) and auditor changes.
We use different proxies for measuring audit delays. Because Section 404 alteredBig N incentives to retain risky clients, we first use the actual delays, proxying for theexisting quality of auditor-client interaction, for predicting auditor changes. However,we also perform analyses using two alternative proxies: unexpected audit delays andindustry-adjusted audit delays.
Our empirical results are consistent with auditors managing their client portfoliosto reduce their risk exposure. Specifically, we find results consistent with thehypothesis that auditor resignations occur in the year following lengthy audit delays.
Audit delays
33
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Furthermore, we find that auditor resignations are more likely to result in downwardrather than lateral changes as the length of audit delays increases. Our results alsosuggest that more than financial distress, risk factors (proxied by audit delays)concerning clients’ internal controls, financial reporting quality, and managementintegrity are related to subsequent auditor changes. These findings increase ourunderstanding of how Big N auditors use audit risk factors – generally unobservableto the investing public – in their client management strategy. From an investor’sperspective, a lengthy audit delay could suggest that there has been a deterioration inthe quality of the client-auditor interaction, which could translate into a downwardauditor change, and a negative stock price reaction (Krishnamurthy et al., 2006).
While the client-auditor realignments examined in this study reduce Big N auditors’risk, they may not necessarily reflect a socially desirable outcome, since non-Big Nfirms must now audit the risky clients dropped by the Big N firms. Because Big Nauditors possibly provide better monitoring of the financial reporting process (Cassellet al., 2007), following the realignments, there could be a reduction in audit quality andan increased likelihood of audit failures. This could represent an unintended costassociated with Section 404 that has not been examined in the literature. It isnoteworthy, however, that the Securities and Exchange Commission (SEC)’s view onthis issue is that audit quality is not impaired when dropped Big N clients are pickedup by non-Big N auditors (Taub, 2004). While our focus is limited to the study ofwhether audit delays proxy for unobserved audit risks, whether or not overall auditquality is impaired due to these realignments is a topic that needs to be pursued byfuture work.
The remainder of this study comprises five sections. Section 2 reviews the relevantliterature. Section 3 presents the research design, while Sections 4 and 5 discuss resultsof main tests and additional analyses, respectively. Section 6 concludes the study.
2. Literature and hypotheses developmentPrior research suggests that auditors take actions to mitigate their risk exposure whenmanaging their client portfolios[4]. Jones and Raghunandan (1998), for example, findthat client portfolios of the Big 6 audit firms shifted towards having fewerfinancial-distressed clients and fewer clients in high-tech industries during a period ofincreasing litigation. Choi et al. (2004) also report that there is a decrease in themagnitudes of financial risk measures of Big 6 audit firms’ clients during a period whenauditors’ exposure to litigation increased. Schwartz and Menon (1985) find that clientswho are in financial distress are more likely to be dropped by a Big N audit firm whencompared to a matched-pair sample of non-failing firms. Shu (2000) constructs two riskmeasures proxying for the likelihood of litigation and the probability of bankruptcy.Shu finds that there is a positive association between auditor resignations and theproxies. Finally, Krishnan and Krishnan (1997) find that litigation risk, proxied byStice’s (1991) litigation score, is positively associated with auditor resignations.The above-mentioned studies use publicly available financial data to constructempirical proxies for the financial and litigation risk factors.
Following the passage of the Private Securities Litigation Reform Act of 1995, anauditor cannot be held liable in a class action lawsuit for bankruptcy or mismanagementby officers of the company, if there is no wrongdoing on the part of the auditor. Thus,financial distress by itself does not constitute an audit risk factor. Consistent with this
MAJ26,1
34
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
idea, Johnstone and Bedard (2004) suggest that, more than financial distress, audit riskfactors relating to management’s integrity, internal controls, and financial reportingquality have a greater impact on auditors’ decisions to retain or dismiss clients.The authors support their hypothesis with tests using proprietary risk assessmentsobtained from auditing firms. However, despite the significance of their findings forauditing research, there has been little additional work on this issue.
This study argues that audit delays can potentially provide an observable proxy forthe proprietary audit risk factors used by Johnstone and Bedard in their study.Specifically, we suggest that audit delays often occur when there are concerns aboutpoor internal controls, low-quality financial reports, lack of management integrity, andclients’ lack of attention to the external audit. For example, a lengthy audit delay canresult when an auditor uncovers a problem in the financial statements and needs to doadditional work to give an opinion on the financial statements. A long audit delaycould also occur if a client-firm is attempting to apply aggressive or non-generallyaccepted accounting principles accounting treatments that the auditor is unwilling toaccept. These risk factors result in auditors executing additional audit proceduresand/or conducting longer negotiations with their clients.
Consistent with this, a few prior studies have used audit delays as a proxy for riskfactors. Ireland (2003) documents that the longer the audit lag, the more likely a companyis to receive a modified audit report. Ireland argues that a long audit lag occurs whenauditees have a high level of inherent and/or control risk which requires more audit workoften resulting in a modified audit opinion. Furthermore, when auditors wish to modifytheir opinion, the result is often longer negotiations between the auditor and the clientover the form of the final accounts and the associated audit report that is the subject oftheir disagreement (Ireland, 2003)[5]. Another study is by Schloetzer (2007) whoexamines whether a lengthy filing delays lead client firms to switch from a Big N to anon-Big N auditor. Schloetzer uses the Form 10-K filing delay – the time between thefiscal year end and SEC filing date – as a proxy for the quality of client-auditorinteraction[6]. He argues that the longer the filing delay, the lower is the overall quality ofthe client-auditor interaction which in turn leads to a change in auditors, from Big N toNon-Big N[7].
Schwartz and Soo (1996) also find evidence suggesting that a lengthy audit delayconstitutes a risk factor for auditors. Using a sample of auditor changes, they find thatclients who switched auditors late in the fiscal year (late switchers) are associated withmore disagreements with the auditor than early switching firms and, in turn,experience lengthier audit delays. Knechel and Payne (2001) provide direct evidencethat a lengthy audit delay is due to more audit hours being expended on anengagement. Using a proprietary database, they document that a lengthy audit delayoccurs because more audit effort and audit hours than normally required are needed tocomplete the audit.
Based on the discussion above, we formulate our first hypothesis. H1 predicts thatBig N auditors make client retention decisions based on the quality of the interactionwith their clients during the preceding year’s audit, suggesting that long audit delaysincrease the likelihood of an auditor change in the following year:
H1. There is a positive association between audit delays and auditor resignationsin the following year.
Audit delays
35
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
H1 does not distinguish between types of auditor changes: lateral (Big N to anotherBig N) or downward (Big N to a non-Big N). Research suggests that once a risky clientleaves a Big N auditor, that firm may have no choice but to seek a non-Big N firm toserve as its auditor (Bockus and Gigler, 1998). Big N auditors have more to lose fromlitigation and suffer a greater loss of reputation from an audit failure than do non-Big Nfirms (Jones and Raghunandan, 1998). Big N auditors are also less dependent on anyone client’s fees (Watkins et al., 2004). This line of research suggests that Big Nauditors are less likely to accept a client dropped by another Big N auditor because ofthe associated client risks that all Big N auditors are anxious to avoid.
However, there is also some prior research which suggests that auditees dropped by aBig N auditor may be accepted as clients by another Big N auditor (i.e. a lateral auditorswitch). Johnstone and Bedard (2004) suggest that otherwise “unacceptable” clients aresometimes accepted into another Big N audit firm’s portfolio conditional on theassignment of specialist personnel or the collection of higher engagement fees. Big Nauditors also have a larger client base than non-Big N auditors which allows themspread a given client’s risk over a well-diversified client portfolio ( Johnstone andBedard, 2004). Finally, the risks facing the old and new auditors can differ (Landsmanet al., 2009). For example, a Big N auditor who is also an industry expert may be able toaccept a higher level of audit risk than another Big N auditor who has no such industryexpertise.
We empirically investigate whether risky firms dropped by a Big N auditorfind another Big N or a non-Big N auditor. If risky clients are more likely to be excludedfrom the Big N audit market, then downward switches will be more highly associatedwith client risk characteristics than lateral switches (Landsman et al., 2009). Further,while another Big N auditor may be willing to accept a risky client dropped by a Big Nauditor, as the level of the audit risk increases, proxied using audit delays, the switch ismore likely to be downward than lateral:
H2. As audit delays increase, auditor changes in the following year are more likelyto be downward than lateral.
3. Research design3.1 DataOur initial sample includes all firms for which audit fees and auditor report dates wereavailable on Audit Analytics for the period 2002-2006 (59,180 firm-years). FollowingSengupta (2004), we delete observations (8,985 firm-years) where the audit report datewas either within seven days of the fiscal year end or more than 90 days after the fiscalyear end, in order to eliminate outliers and/or potential errors in report dates[8].We require firms to have at least two consecutive years of financial data because weuse lagged independent variables in our tests. This criterion results in the elimination of30,996 firm-years. We then only consider client firms audited by a Big N auditor in theprevious year. As discussed, this is because resource constraints (e.g. shortage ofassurance staff) were more pronounced for Big N auditors since some of the SOXprovisions applied only to large clients, who make up the majority of the Big N auditors’client base (GAO, 2006). We deleted an additional 2,412 observations consisting ofnon-accelerated filers who have a longer period to file their financial statements and can,therefore, be expected to have a longer audit report lag. Finally, we deleted 665observations where the auditor change was reported as a dismissal. The final sample
MAJ26,1
36
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
consists of 11,307 firm-years for the period 2003-2006[9]. It is comprised of threemutually exclusive groups:
(1) firms continuing with a Big N auditor (no change firms);
(2) firms with lateral auditor changes in the following year; and
(3) firms with downward auditor changes in the following year.
Panel B of Table I exhibits the industry distribution of sample firms where industryis defined as in Frankel et al. (2002). Sample firms are widely distributed amongindustries, with some clustering of firms in the durable manufacturers and computerindustries. Mean audit report lags (ARLs) vary across industries, with longest foragriculture (64 days) and the shortest for textile and printing (52 days), suggesting aneed to control for industry effects. Panel C of Table I reports the distribution ofsample firms by year. The first column shows the number of firms which switchedtheir auditors from Big N to non-Big N, while the second column reports the switchesfrom Big N to another Big N. It is noteworthy that in all years, there are moredownward auditor switches than lateral auditor switches in the post-SOX years.Consistent with our conjecture that Section 404 triggered Big N resignations, thepanel shows that the largest number of resignations took place in the years 2004 and2005. The last column represents the number of firms that retained their incumbentBig N auditors.
3.2 Regression modelWe model the likelihood of a Big N auditor resignation using ARLs and a set ofcontrol variables identified by prior studies. The following logistic regression is used(Table II):
RESIGNt ¼ b0 þ b1ARLt21 þ b2GCt21 þ b3MODOPt21 þ b4TENUREt21
þ b5AUDFEEt21 þ b6ROAt21 þ b7LOSSt21 þ b8LVRGt21
þ b9GROWTHt21 þ b10DAt21 þ b11SIZEt21 þ year dummiesþ industry dummiesþ 1
ð1Þ
3.3 Control variables usedThe control variables used have been identified by prior research as determinants ofauditor resignations.
Companies that do not receive a clean audit opinion are more risky, which increasesthe likelihood that there will be an auditor resignation (Krishnan and Krishnan, 1997;Johnstone and Bedard, 2004). Therefore, we predict positive coefficients on thevariables GC and MODOP. Prior research suggests that client firms with long-tenuredauditors are less risky than those with short-tenured auditors (Stice, 1991; Krishnanand Krishnan, 1997; Landsman et al., 2009). Therefore, we expect that TENURE willhave a negative relation with the likelihood of an auditor resignation. The nature of therelationship between audit fees (AUDFEE) and auditor resignations is debatable.Bockus and Gigler (1998) contend that incumbent auditors might increase audit fees forriskier clients but once the level of risk crosses a certain threshold, the auditors willresign. This suggests that there could be a positive association between prior year’saudit fees and auditor resignations. On the other hand, it is possible that incumbent
Audit delays
37
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
No. of firm-yearsPanel A: sample selection procedureFirm-years with audit fees and ARLsin Audit Analytics (2002-2006) 59,180Less: ARLs that are less than 7 daysor more than 90 days 8,985Missing lagged ARLs, audit fee data,and financial dataon Compustat 30,996Firm-years having non-Big N as aprevious auditor 4,815Non-accelerated filers 2,412Auditor dismissals 665Final sample (2003-2006) 11,307
Panel B: distribution of sample by industry
Industry Total %Auditor
resignationMeanARL
Agriculture 18 0.00 0 64.06Mining and construction 247 0.02 1 58.07Food 263 0.02 2 53.78Chemicals 281 0.02 3 55.43Computers 1,731 0.15 58 53.25Durable manufacturers 2,248 0.20 33 54.65Extractive 423 0.04 3 61.06Financial 1,481 0.13 19 58.44Pharmaceuticals 765 0.07 11 56.93Retail 1,102 0.10 14 54.92Services 993 0.09 15 59.51Textile and printing/publishing 481 0.04 8 52.07Transportation 774 0.07 7 60.74Utilities 462 0.04 5 58.62Not classified 38 0.00 1 59.95Total 11,307 1.00 180
Panel C: type of auditor change by year
YearBig N to non-
Big NBig N to Big
N No change2003 31 9 2,7212004 56 12 2,9562005 42 8 2,8202006 18 4 2,630
Total 147 33 11,127
Notes: Agriculture (0100-0999), mining/construction (1000-1999, excluding 1300-1399), food (2000-2111), chemicals (2800-2824, 2840-2899), computers (7370-7379, 3570-3579, 3670-3679), durablemanufacturers (3000-3999, excluding 3570-3579 and 3670-3679), extractive (2900-2999, 1300-1399),financial (6000-6999), pharmaceuticals (2830-2836), retail (5000-5999), services (7000-8999, excluding7370-7379), textiles (2200-2799), transportation (4000-4899), utilities (4900-4999), and not classified(2112-2199, 2837-2839, 2825-2829); ARL is the number of calendar days from fiscal year end to date ofthe auditor’s report
Table I.Sample selection andsample distribution
MAJ26,1
38
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
auditors are less likely to resign from clients paying higher audit fees, suggesting anegative association between the two variables.
We also include controls for clients’ financial risks. Because less profitable and moreleveraged clients are considered more risky (Schloetzer, 2007), we predict a higherlikelihood of an auditor resignation in these companies. We use the following threevariables, ROA, LOSS, and LVRG, to proxy for financial risks. These variables areexpected to have negative, positive, and positive coefficients, respectively.
Stice (1991) argues that high-growth firms pose additional risks for auditorsbecause these firms tend to have less-effective internal control systems. Therefore,we include GROWTH in our regressions and expect this variable to be positivelyassociated with the likelihood of an auditor resignation. Since positive discretionaryaccruals suggest that there is a higher risk of earnings management being present,we could expect more auditor resignations from client firms with larger positivediscretionary accruals. This predicts that there will be a positive associationbetween RESIGN and DA. Because large clients are less likely to fail, we predict anegative coefficient on SIZE. Finally, we include year and industry dummies in themodels[10].
Variable Expected sign Definition
Dependent variableRESIGN 1 if a Big N auditor resigns and 0 otherwiseTest variableARL þ Number of days between the client’s fiscal year end
and the completion date of the auditAuditor characteristicsGC þ 1 if the audit opinion is a going concern opinion and 0
otherwiseMODOP þ 1 if the audit opinion is modified for any reason other
than going concern and 0 if unqualifiedTENURE 2 Number of years audited by the incumbent (old)
auditorAUDFEE þ /2 Natural logarithm of the preceding year’s audit feeFirm characteristicsROA 2 Return on assets, defined as net income before
extraordinary items divided by total assetsLOSS þ 1 if earnings before extraordinary items is ,0 and 0
otherwiseLVRG þ Ratio of long-term debt to total assetsGROWTH þ Percentage changes in total assetsDA þ Performance-matched discretionary accrualsa
SIZE 2 Natural logarithm of the market value of equity
Notes: aWe calculate performance-matched discretionary accruals following Kothari et al. (2005).First, we obtain discretionary accruals from the Jones model, which requires regressing total accrualson variables that are expected to vary with normal accruals, such as changes in revenues and capitalintensity; to obtain performance-matched discretionary accruals, we match each firm-year observationwith an observation belonging to a firm from the same two-digit SIC code and year, with the closestreturn on assets (net income divided by total assets); then, we define performance-matcheddiscretionary accruals as each firm’s discretionary accruals minus its matched counterpart’sdiscretionary accruals Table II.
Audit delays
39
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
4. Empirical test results4.1 Descriptive statisticsTable III presents descriptive statistics for the variables used in the regressionsreported separately by the three categories: downward changes (Big N to Non-Big N),lateral changes (Big N to another Big N), and no changes. We winsorize allnon-indicator variables at the 1 and 99 percent levels to reduce the effects of extremevalues on the test results. Table III presents the means, medians, and SD of thevariables according to each group. There are more downward switches whencompared to lateral switches, which is consistent with Schloetzer (2007) who finds thatauditor changes from a Big N to a non-Big N auditor became more frequent followingthe passage of SOX.
The mean value of ARL for the downward switch group is 67.116, which indicatesthat it took a Big N auditor, on average, about 67 days to complete the audit ofa client-firm that switched its auditor to a non-Big N auditor in the following year.The mean value of ARL of downward switch firms (67.116 days) is statisticallysignificantly longer than that for lateral switch firms (58.969 days). While the meanARL of no switch firms (56.205 days) is shorter than that of lateral switch firms (58.969days), the difference is not statistically significant. The above univariate comparison ofmeans is consistent with our hypothesis, H2, that firms with longer ARLs are likely toswitch from a Big N to a non-Big N auditor, rather than change from a Big N to anotherBig N auditor or stay with incumbent Big N auditor.
Firms that switch auditors downward (compared to other groups) tend to have moregoing concern opinions but fewer types of other modified opinions and relativelysmaller audit fees in the year preceding the switch. They are also less profitable andare smaller in size[11]. With the exception of “other modified opinions” (MODOP),these univariate results are consistent with expectations.
Table IV presents correlations for the variables used in the regressions. Consistentwith our prediction, RESIGN is positively correlated with one-year-lagged ARL.In addition, RESIGN is statistically significantly correlated with all independentvariables except for TENURE and LVRG. Interestingly, contrary to our expectation,RESIGN is negatively correlated with GROWTH and MODOP (modified opinion).Also, with the exception of MODOP, ARL is significantly correlated with the followingaudit risk factors: going concern opinions, tenure, profitability, leverage, and firm size.The direction of the correlations supports our argument that audit delays aresignificantly related to audit risks. The largest correlation in the table is between firmsize and audit fees (0.70).
Tables III and IV, however, only show results obtained from univariate testing.Below we present results from estimation of multivariate models which are morepowerful than the univariate tests.
4.2 Regression analysesTable V presents empirical results of estimations of our logistic regression models.The table reports results using data pooled over all years and separately using yearsbefore and after Section 404 of SOX went into effect[12]. For the post-Section 404sample (2005-2006), we include a dummy variable that is coded 1 if a firm reports aninternal control weakness and 0 otherwise. This is due to Ettredge et al. (2007) who
MAJ26,1
40
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Big
Nto
non
-Big
N(A
)(n
¼14
7)B
igN
toB
igN
(B)
(n¼
33)
No
chan
ges
(C)
(n¼
11,1
27)
t-te
st(A
¼B
)t-
test
(B¼
C)
Var
iab
leM
ean
Med
ian
SD
Mea
nM
edia
nS
DM
ean
Med
ian
SD
t-st
atis
tic
t-st
atis
tic
ARL
67.1
1673
.000
19.9
0758
.969
64.0
0019
.574
56.2
0559
.000
18.6
602.
13*
*0.
85GC
0.21
10.
000
0.40
90.
061
0.00
00.
242
0.01
10.
000
0.10
62.
03*
*2.
64*
**
MODOP
0.28
60.
000
0.45
30.
576
1.00
00.
502
0.43
40.
000
0.49
62
3.26
**
*1.
64TENURE
9.95
99.
000
6.10
18.
606
8.00
04.
527
9.77
97.
000
8.62
31.
202
0.78
AUDFEE
12.6
1612
.378
1.10
513
.354
13.5
410.
967
13.5
5913
.474
1.22
32
3.54
**
*0.
96ROA
20.
240
20.
073
0.36
12
0.10
10.
011
0.25
90.
001
0.03
50.
169
22.
09*
*2
3.43
**
*
LOSS
0.68
01.
000
0.46
80.
455
0.00
00.
506
0.25
00.
000
0.43
32.
47*
*2.
71*
*
LVRG
0.20
10.
101
0.25
70.
217
0.13
90.
223
0.22
30.
183
0.21
72
0.35
20.
14GROWTH
0.06
02
0.04
60.
515
0.13
20.
029
0.50
20.
159
0.08
70.
339
20.
732
0.47
DA
0.02
00.
011
0.21
62
0.03
02
0.00
80.
157
20.
011
20.
006
0.13
31.
272
0.85
SIZE
4.35
94.
245
1.57
65.
973
6.08
71.
829
6.77
66.
641
1.69
62
5.16
**
*2
2.71
**
*
Notes:
Sig
nifi
can
ceat
:* 1
0,*
* 5,a
nd
**
* 1p
erce
nt;
var
iab
led
efin
itio
ns:RESIGN
,1if
the
aud
itor
resi
gn
sfr
omit
sau
dit
oran
d0
oth
erw
ise;ARL
,nu
mb
erof
day
sb
etw
een
the
clie
nt’
sfi
scal
yea
ren
dan
dth
eau
dit
com
ple
tion
dat
e;GC
,1if
the
aud
itop
inio
nis
ag
oin
gco
nce
rnan
d0
oth
erw
ise;MODOP
,1if
the
aud
itop
inio
nis
mod
ified
for
any
reas
onot
her
than
goi
ng
con
cern
and
0if
un
qu
alifi
ed;T
ENURE
,nu
mb
erof
yea
rsau
dit
edb
yth
ein
cum
ben
t(o
ld)a
ud
itor
;AUDFEE
,n
atu
ral
log
arit
hm
ofth
ep
rece
din
gy
ear’
sau
dit
fee
(old
aud
itor
);ROA
,re
turn
onas
sets
,d
efin
edas
net
inco
me
bef
ore
extr
aord
inar
yit
ems
div
ided
by
tota
las
sets
;LOSS
,1if
earn
ing
sb
efor
eex
trao
rdin
ary
item
sis,
0an
d0
oth
erw
ise;LVRG
,rat
ioof
deb
tto
tota
las
sets
;GROWTH
,per
cen
tag
ech
ang
esin
tota
las
sets
;DA
,p
erfo
rman
ce-m
atch
edd
iscr
etio
nar
yac
cru
als;SIZE
,n
atu
ral
log
arit
hm
ofth
em
ark
etv
alu
eof
equ
ity
Table III.Descriptive statistics
Audit delays
41
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
ARL
GC
MODOP
TENURE
AUDFEE
ROA
LOSS
LVRG
GROWTH
DA
SIZE
RESIGN
0.06
3(0
.000
)0.
182
(0.0
00)
20.
024
(0.0
11)
20.
001
(0.9
15)
20.
082
(0.0
00)
20.
153
(0.0
00)
0.11
1(0
.000
)2
0.01
1(0
.243
)2
0.03
2(0
.000
)0.
020
(0.0
31)
20.
155
(0.0
00)
ARL
0.08
8(0
.000
)2
0.07
0(0
.000
)2
0.08
8(0
.000
)0.
142
(0.0
00)
20.
065
(0.0
00)
0.07
2(0
.000
)0.
092
(0.0
00)
0.00
1(0
.946
)0.
008
(0.4
05)
20.
129
(0.0
00)
GC
20.
105
(0.0
00)
20.
030
(0.0
01)
20.
031
(0.0
00)
20.
237
(0.0
00)
0.17
3(0
.000
)0.
042
(0.0
00)
20.
026
(0.0
06)
0.01
9(0
.049
)2
0.16
4(0
.000
)MODOP
0.02
8(0
.023
)0.
122
(0.0
00)
0.05
7(0
.000
)2
0.03
9(0
.000
)0.
136
(0.0
00)
20.
076
(0.0
00)
0.00
1(0
.896
)0.
111
(0.0
00)
TENURE
0.24
6(0
.000
)0.
108
(0.0
00)
20.
121
(0.0
00)
0.00
9(0
.326
)2
0.10
3(0
.000
)0.
028
(0.0
03)
0.25
1(0
.000
)AUDFEE
0.18
6(0
.000
)2
0.17
6(0
.000
)0.
171
(0.0
00)
20.
119
(0.0
00)
20.
019
(0.0
39)
0.70
3(0
.000
)ROA
20.
664
(0.0
00)
20.
010
(0.2
69)
0.03
6(0
.000
)0.
021
(0.0
29)
0.34
6(0
.000
)LOSS
0.01
9(0
.041
)2
0.08
1(0
.000
)2
0.02
6(0
.005
)2
0.35
7(0
.000
)LVRG
20.
077
(0.0
00)
0.02
9(0
.002
)0.
089
(0.0
00)
GROWTH
0.01
6(0
.094
)2
0.03
7(0
.000
)DA
20.
043
(0.0
00)
Notes:
Th
ep-
val
ues
are
inp
aren
thes
es;
see
Tab
leII
Ifo
rd
efin
itio
ns
ofth
ev
aria
ble
s;n¼
11,3
07
Table IV.Pearson correlationmatrix of variables usedin our tests
MAJ26,1
42
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
argue that SOX 404 reports provide a new measure of a client’s audit risk[13] thatpotentially could lead to a realignment of the auditor-client relationship.
All models are statistically significant, explaining 27.11-31.54 percent of thevariation in the dependent variable. In all models, we find that the variable of interest,ARL, is positive and statistically significant, which is consistent with H1 that auditorresignations occur more frequently following lengthy ARLs.
Also, in all models, compared with firms that keep their auditor, audit firms aremore likely to resign from client firms that are smaller in size and have going concernopinions (significant at least at the 10 percent level of testing). Excepting for the period2005-2006, auditors are less likely to resign from firms with high return on assets(ROA). However, a similar result holds for the years 2005-2006 where we find thatauditors are more likely to resign from firms with losses (LOSS). The variable MODOP(modified opinion) which, contrary to expectations, was significantly and negativelycorrelated to audit delays in the univariate tests (Table IV) does not have a significantassociation with ARL in the multivariate regressions. The coefficients on AUDFEEand GROWTH are statistically significant only during 2005-2006 and 2003-2004,respectively. Importantly, in the post-Section 404 period (2005-2006), we find astatistically positive coefficient of MW, which is consistent with Ettredge et al. (2007).It is interesting that our measure of audit risks, ARL, has a significant explanatorypower, over and above MW, for predicting auditor resignations. The coefficients on allother variables are statistically not significant.
To gauge economic significance, we also estimate the logistic regression using adummy variable that takes a value of 1 if the ARL is longer than the median ARL and 0otherwise (results not reported). In the pooled sample, the coefficient on the dummy is0.3539 (t-value: 3.56) which suggests that (compared to the control group) there is anapproximately 42 percent greater chance of an auditor resignation taking place forfirms with lengthier ARLs[14].
Pooled years 2003-2004 2005-2006Variable Coefficient p-value Coefficient p-value Coefficient p-value
Intercept 0.9918 0.4618 20.7841 0.6820 4.3635 0.0555ARL 0.0170 ,0.0001 0.0134 0.0199 0.0249 0.0256GC 1.1248 0.0002 1.1351 0.0028 1.0410 0.0516MODOP 0.0086 0.9630 20.0901 0.7028 0.0570 0.8529TENURE 0.0485 0.3547 0.0486 0.6824 0.0545 0.2275AUDFEE 20.1067 0.3218 0.0219 0.8815 20.3895 0.0286ROA 21.1029 0.0026 21.4152 0.0018 20.9271 0.1788LOSS 0.3091 0.1587 20.0368 0.9013 0.7212 0.0311LVRG 20.1791 0.6230 20.2658 0.5572 0.2348 0.7099GROWTH 20.3393 0.1074 20.8389 0.0082 0.3084 0.3074DA 0.6114 0.1900 0.5650 0.3583 0.5042 0.5145SIZE 20.7294 ,0.0001 20.8013 ,0.0001 20.6015 0.0001MW 1.5871 ,0.0001Wald x2 387.89 * * * 230.20 * * * 167.89 * * *
Pseudo-R 2 0.2711 0.2959 0.3154n 11,307 5,785 5,522
Notes: Significance at: *10, * *5, and * * *1 percent; see Table III for definitions of variables
Table V.Results of logistic
regressions of auditorresignations
Audit delays
43
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Next, we test H2 by ordering our sample into three groups: downward auditorswitches (coded 3), lateral auditor switches (coded 2), and no switches (coded 1). In theordered logistic regression analyses in Table VI, we find that the coefficient on ARL ispositive and statistically significant in all regressions which supports H2, namely thatas audit risks increase not only does the likelihood of an auditor change increase butthat clients are more likely to engage a smaller, non-Big N auditor in the subsequentyear. The results on the remaining variables are virtually identical to those presentedin Table V.
5. Additional analyses5.1 Unexpected ARLAn advantage of using audit delays in our tests is that they constitute a simple proxyfor the quality of an auditor-client relationship. A more sophisticated proxy that couldbe used instead is unexpected audit delay. Unexpected audit delays can be interpretedas reflecting changes in client-auditor interactions which in turn could result in auditorswitches. (See Appendix for the model used to obtain predicted values of audit delays.)Unexpected ARLs get at the idea that something has changed for the client firms thathas motivated the auditor change. However, a disadvantage of this proxy from aninvestor’s perspective is its complexity; in addition, measurement error from modelmisspecification could affect the test results. While we find positive and statisticallysignificant coefficients on unexpected audit delays using both logistic and orderedlogistic models (significant at the 1 percent level in both models), it is worth noting thatthe simple proxy, audit delay does just as well in predicting auditor changes – which isconsistent with our conjecture that Section 404 of SOX altered auditors’ incentives toretain risky clients, as proxied by lengthy audit delays (results of these tests areuntabulated).
Pooled sample 2003-2004 2005-2006Variable Coefficient p-value Coefficient p-value Coefficient p-value
Intercept1 2.7748 0.0037 0.6164 0.6534 7.1352 ,0.0000Intercept2 2.9886 0.0018 0.8402 0.5405 7.3600 ,0.0001ARL 0.0177 ,0.0000 0.0120 0.0015 0.0317 ,0.0001GC 1.1118 ,0.0000 1.1274 ,0.0000 1.0932 0.0046MODOP 20.0895 0.4774 20.1587 0.3129 20.0318 0.8857TENURE 0.0441 0.5621 0.0403 0.2354 0.0545 0.8326AUDFEE 20.1151 0.1277 20.0076 0.9419 20.4549 0.0004ROA 21.1438 ,0.0001 21.5429 ,0.0001 20.9053 0.0741LOSS 0.2268 0.1243 20.0593 0.7621 0.5857 0.0139LVRG 20.1291 0.6042 20.0741 0.8076 0.0304 0.9469GROWTH 20.3737 0.0063 20.9264 ,0.0001 0.3376 0.0992DA 0.5296 0.0943 0.4090 0.3286 0.6818 0.2263SIZE 20.8779 ,0.0000 20.9125 ,0.0001 20.7758 ,0.0000MW 1.8587 ,0.0000Wald x2 886.49 * * * 522.99 * * * 361.63 * * *
Pseudo-R 2 0.3694 0.3835 0.4392n 11,307 5,785 5,522
Notes: Significance at: *10, * *5, and * * *1 percent; see Table III for definitions of variables
Table VI.Results of ordered logisticregressions ofauditor resignations
MAJ26,1
44
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
5.2 Industry-adjusted ARLsWe also replace ARL with an industry-adjusted ARL. We calculate median values ofARL for each industry as defined in Frankel et al. (2002) and obtain industry-adjustedARLs which are ARLs minus the median of the industry ARLs. We obtain qualitativelysimilar results to those documented (results of these tests are untabulated).
5.3 Non-audit service feesIt is possible that auditors will be less likely to resign from clients that provide largeamounts of non-audit service revenue (e.g. fees for tax services). We find (results nottabulated) that the coefficient on non-audit service fees is negative and statisticallysignificant, suggesting that auditor changes are less likely to occur for clients requiringlarge amounts of non-audit services. The results on the other variables arequalitatively unchanged.
6. ConclusionThis study uses audit delays to proxy for audit risk factors that measure the quality ofauditor-client interactions. Our results are consistent with those of previous workshowing that auditors respond to high audit risk factors by dropping risky clients.Our results show that more than financial distress, the quality of the auditor-clientinteraction determines whether an auditor will stay with a client or resign.
Auditor changes are important events that markets want to know about. However,companies often do not fully disclose reasons for auditor changes. Our researchsuggests that a long audit delay represents a publicly observed proxy for possibledisagreements between auditors and their clients or the presence of audit risk factorsthat lead to an auditor change in the following year. This study suggests that all elseconstant, investors should regard a lengthy audit report lag as indicating that there hasbeen a deterioration in the quality of the client-auditor interaction, which couldtranslate into a downward auditor change in the following year. An audit delay alsopresents an observable proxy for successor auditors to consider while evaluating theirrisk exposure from the acceptance of a client whose auditor has resigned. As a proxyfor client-auditor disagreement, audit delays are able to predict an auditor change justas well as our more sophisticated proxy (unexpected audit delays) that controls forknown factors that affect audit delays.
Finally, there is an important concern associated with auditor changes that followlong audit delays. Our results suggest that the successor auditors are likely to be oflower quality. Research suggests that capital markets react negatively to downwardauditor changes (Krishnamurthy et al., 2006). This raises an important question forregulators as to whether the smaller accounting firms will be able to perform qualityaudits for these high-risk firms (Turner et al., 2005). An interesting question for futureresearch would be to examine how non-Big N auditors manage their risk exposurefollowing the acceptance of risky clients dropped by Big N firms.
Notes
1. There were 2,514 auditor changes during 2003 and 2004 affecting more than one-fourth of allUS publicly traded companies (Taub, 2005).
2. For example, in 2004, BDO Seidman, LLP had a net gain of 71 clients, while Grant Thornton,LLP gained 17 clients, net (Koehn and DelVecchio, 2006).
Audit delays
45
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
3. Audit risk is the risk that an auditor may unknowingly fail to appropriately modify his/heropinion on financial statements that are materially misstated (AICPA, 1983, AU 312.02).
4. Prior studies also suggest that auditors are proactive in managing their client portfolios.For example, Krishnan and Krishnan (1997) and Shu (2000) document that auditors are morelikely to resign from riskier clients rather than simply wait to be dismissed.
5. That is, when there is a going concern issue, auditors may delay expressing an audit opinionor finalizing their report, in the hopes that the problem will be resolved and a modifiedopinion will be avoided (Ireland, 2003).
6. There are several differences between this study and Schloetzer’s (2007) study. First, wemeasure the reporting lag as the time between the fiscal year end and the audit completiondate (audit report lag), while Schloetzer (2007) measures delays as the time between the fiscalyear end and the SEC filing date (filing lag). In our sample, the correlation between these twolags is 0.418, which suggests that the two lags potentially represent different underlyingevents. Lee et al. (2009) argue that upon completion of an audit, managers exercise theirdiscretion by determining the optimal time for announcing earnings or filing their financialstatements by considering the costs and benefits of their timing decisions. For example,firms with good news are likely to announce earnings or file their financials soon after thecompletion of an audit. The filing lag used by Schloetzer, therefore, captures both the natureof the auditor-client interaction (proxied by the audit report lag) and the client-manager’sdiscretionary timing decision (proxied by the discretionary lag). Second, we focus only onauditor resignations, while Schloetzer’s sample includes both resignations and dismissals.Client firms might dismiss their auditors for lengthy delays because lengthy audit delayscould imply inefficiencies in an audit. Therefore, focusing only on resignations mightprovide for a cleaner sample for testing our hypothesis. We thank an anonymous reviewerfor this suggestion. Third, unlike Schloetzer (2007), we control for alternative explanationsfor an auditor change, including audit fees, growth opportunity, auditor tenure, anddiscretionary accruals (DeFond and Subramanyam, 1998; Johnson and Lys, 1990; Bockusand Gigler, 1998). Most importantly, we control for internal control weaknesses that havebeen shown to significantly increase audit delay (Ettredge et al., 2007). Finally, we providean additional new result showing that auditor-client realignments follow an orderedsequence as audit delays increase.
7. Another proxy for audit risk is a material weakness in internal control which is a requireddisclosure under SOX 404. Ettredge et al. (2006) show that the presence of a materialweakness in internal controls over financial reporting is associated with longer audit delays.This supports our assertion that audit delays proxy for audit risks. In a later section, weexplore this issue in more detail.
8. While we lose a significant number of firms (8,985 observations or 15 percent of the sample)due to this criterion, we found that a large number of these firms are not covered byCompustat. For example, we found that approximately half of these firms do not have dataon total assets on Compustat (Krishnan and Yang, 2009). Coverage on Audit Analytics ismuch wider than on Compustat whose coverage is restricted to larger firms showing asurvivorship bias. The results are qualitatively the same when we use 180 days for deletingextreme ARLs.
9. We lose the year 2002 because we use lagged variables in our tests.
10. We also include two more control variables: EXPERT, a dummy variable coded 1 if anauditor has 30 percent or more market share in an industry and 0 otherwise and MNA, adummy variable coded 1 if the client had a merger or acquisition in the two previous years,and 0 otherwise. The above two variables are used to model auditor changes due toclient-initiated reasons. These variables are included due to Lee et al. (2004) who argue that
MAJ26,1
46
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
sometimes the distinction between resignations and dismissals is blurry. Inclusion of thesevariables does not alter our inference.
11. Other than those discussed here, there are no statistically significant differences in the othervariables used in our tests.
12. The SEC reduced the 10-K filing period for large accelerated filers (i.e. equity public float$700 M or more) to 60 days for year ends on or after December 15, 2006 (SEC, 2005). Tocontrol for this change, we include a dummy variable for large accelerated filers and aninteraction term with audit delay for the years 2005-2006. Our main results (untabulated)remain unchanged.
13. The reports reveal auditors’ judgments about internal controls and accounting-related issuesthat are largely independent of a client’s financial conditions and provide a strong indicator(in contrast to bankruptcy or litigation risk metrics) for explaining client retention decisions(Ettredge et al., 2007).
14. The odds are computed as the exponent of 0.3539, which is 1.42 or a 42 percent larger chance.
References
AICPA (1983), Audit Risk and Materiality in Conducting an Audit, Statement on AuditingStandards No. 47, American Institute of Certified Public Accountants, New York, NY.
Bamber, E., Bamber, L. and Schoderbek, M. (1993), “Audit structure and other determinants ofaudit report lag: an empirical analysis”, Auditing: A Journal of Practice & Theory, Vol. 12,pp. 72-80.
Bell, T.B., Bedard, J.C., Johnstone, K.M. and Smith, E.F. (2002), “RiskSM: a computerized decisionaid for client acceptance and continuance risk assessments”, Auditing: A Journal ofPractice & Theory, Vol. 21, pp. 97-113.
Beneish, M.D., Hopkins, P.E., Jansen, I.P. and Martin, R.D. (2005), “Do auditor resignations reduceuncertainty about the quality of firms’ financial reporting?”, Journal of Accounting &Public Policy, Vol. 24, pp. 357-90.
Bockus, K. and Gigler, F. (1998), “A theory of auditor resignation”, Journal of AccountingResearch, Vol. 36, pp. 191-208.
Cassell, C.A., Giroux, G., Myers, L.A. and Omer, T.C. (2007), “The effect of the Sarbanes-OxleyAct of 2002 on the relation between the strength of corporate governance andauditor-client alignments”, working paper, Texas A&M University, College Station, TX.
Cenker, W.J. and Nagy, A.L. (2008), “Auditor resignations and auditor industry specialization”,Accounting Horizons, Vol. 22 No. 3, pp. 279-95.
Choi, J.-H., Doogar, R. and Ganguly, A. (2004), “The riskiness of large audit firm client portfoliosand changes in audit liability regimes: evidence from the US audit market”, ContemporaryAccounting Research, Vol. 21 No. 4, pp. 747-85.
DeFond, M.L. and Subramanyam, K.R. (1998), “Auditor changes and discretionary accruals”,Journal of Accounting and Economics, Vol. 25, pp. 35-67.
Ettredge, M., Li, C. and Sun, L. (2006), “The impact of SOX section 404 internal control qualityassessment on audit delay in the SOX era”, Auditing: A Journal of Practice & Theory,Vol. 25 No. 2, pp. 1-23.
Ettredge, M., Heintz, J., Li, C. and Scholz, S. (2007), “Auditor realignments accompanyingimplementation of SOX 404 reporting requirements”, working paper, University of Kansas,Lawrence, KS.
Audit delays
47
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
Frankel, R.M., Johnson, M.F. and Nelson, K.K. (2002), “The relation between auditors’ fees fornonaudit services and earnings management”, The Accounting Review, Vol. 77, pp. 71-105.
GAO (2006), “Sarbanes-Oxley Act: consideration of key principles needed in addressingimplementation for smaller public companies”, United States Government AccountabilityOffice, available at: www.gao.gov/cgi-bin/getrpt?GAO-06-361
Henderson, B. and Kaplan, S. (2000), “An examination of audit report lag for banks: a panel dataapproach”, Auditing: A Journal of Practice & Theory, Vol. 19, pp. 160-74.
Ireland, J.C. (2003), “An empirical investigation of determinants of audit reports in the UK”,Journal of Business Finance & Accounting, Vol. 30 Nos 7-8, pp. 975-1015.
Johnson, W.B. and Lys, T. (1990), “The market for audit services: evidence from voluntaryauditor changes”, Journal of Accounting and Economics, Vol. 12, pp. 281-308.
Johnstone, K.M. and Bedard, J.C. (2004), “Audit firm portfolio management decisions”, Journal ofAccounting Research, Vol. 42 No. 4, pp. 659-90.
Jones, F.L. and Raghunandan, K. (1998), “Client risk and recent changes in the market for auditservices”, Journal of Accounting & Public Policy, Vol. 17, pp. 169-81.
Knechel, W. and Payne, J. (2001), “Additional evidence on audit report lag”, Auditing: A Journalof Practice & Theory, Vol. 20, pp. 137-46.
Koehn, J.L. and DelVecchio, S.C. (2006), “Revisiting the ripple effects of the Sarbanes-Oxley Act”,The CPA Journal, Vol. 76 No. 5, pp. 32-9.
Kothari, S.P., Leone, A.J. and Wasley, C.E. (2005), “Performance matched discretionary accrualmeasures”, Journal of Accounting Economics, Vol. 39, pp. 163-97.
Krishnamurthy, S., Zhou, J. and Zhou, N. (2006), “Auditor reputation, auditor independence, andthe stock market impact of Andersen’s indictment on its client firms”, ContemporaryAccounting Research, Vol. 23 No. 2, pp. 465-90.
Krishnan, J. and Krishnan, J. (1997), “Litigation risk and auditor resignations”, The AccountingReview, Vol. 72, pp. 539-60.
Krishnan, J. and Yang, J.S. (2009), “Recent trends in audit report and earnings announcementlags”, Accounting Horizons, Vol. 23, pp. 265-88.
Landsman, W.R., Nelson, K.K. and Rountree, B.R. (2009), “Auditor switches in the pre- andpost-Enron eras: risk or realignment?”, The Accounting Review, Vol. 84 No. 2, pp. 531-58.
Lee, H., Mande, V. and Ortman, R. (2004), “The effect of audit committee and board of directorindependence on auditor resignation”, Auditing: A Journal of Practice & Theory, Vol. 23No. 2, pp. 131-46.
Lee, H., Mande, V. and Son, M. (2009), “Do lengthy auditor tenure and the provision of non-auditservices by the external auditor reduce audit report lags?”, International Journal ofAuditing, Vol. 13 No. 2, pp. 87-104.
Public Accounting Report (2004), “Big Four shed smaller clients in Big Numbers”, September 30.
Rama, D.V. and Read, W.J. (2006), “Resignations by the Big N and the market for audit services”,Accounting Horizons, Vol. 20 No. 2, pp. 97-109.
Schloetzer, J.D. (2007), “Arthur Andersen, SOX section 404 and auditor turnover: theory andevidence”, working paper, University of Pittsburgh, Pittsburgh, PA.
Schwartz, K. and Menon, K. (1985), “Auditor switches by failing firms”, The Accounting Review,Vol. 60, pp. 248-61.
Schwartz, K. and Soo, B.S. (1996), “The association between auditor changes and reporting lags”,Contemporary Accounting Research, Vol. 13 No. 1, pp. 353-70.
MAJ26,1
48
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
SEC (2005), “Revisions to accelerated filer definition and accelerated deadlines for filing periodicreport”, Release No. 33-8644, Securities and Exchange Commission, Washington, DC,available at: www.sec.gov/rules/final/33-8644.htm
Sengupta, P. (2004), “Disclosure timing: determinants of quarterly earnings release dates”,Journal of Accounting & Public Policy, Vol. 23, pp. 457-82.
Shu, S. (2000), “Auditor resignations: clientele effects and legal liability”, Journal of Accountingand Economics, Vol. 29, pp. 173-205.
Stice, J.D. (1991), “Using financial market information to identify pre-engagement factorsassociated with lawsuits against public accountants”, The Accounting Review, Vol. 66,pp. 516-33.
Taub, S. (2004), “Big Four seen shedding small clients: Sarbox rules stretch audit firms’resources, E&Y Chief Executive claims”, available at: www.cfo.com/ (accessed24 September 2004).
Taub, S. (2005), “Auditors rotating at a dizzying pace”, available at: www.cfo.com/ (accessed18 February 2005).
Turner, L.E., Williams, J.P. and Weirich, T.R. (2005), “An inside look at auditor changes”,The CPA Journal, pp. 12-21.
Watkins, A.L., Hillison, W. and Morecroft, S.E. (2004), “Audit quality: a synthesis of theory andempirical evidence”, Journal of Accounting Literature, Vol. 23, pp. 153-93.
Winograd, B.N., Gerson, J.S. and Berlin, B.L. (2000), “Audit practices of PricewaterhouseCoopers”,Auditing: A Journal of Practice & Theory, Vol. 19 No. 2, pp. 175-82.
Further reading
DeAngelo, L.E. (1981), “Auditor size and audit quality”, Journal of Accounting and Economics,Vol. 3, pp. 183-99.
Elder, R., Zhang, Y., Zhou, J. and Zhou, N. (2007), “Internal control weaknesses and client riskmanagement”, working paper, Syracuse University, New York, NY.
Ettredge, M., Li, C. and Scholz, S. (2007), “Audit fees and auditor dismissals in theSarbanes-Oxley era”, Accounting Horizons, Vol. 21 No. 4, pp. 371-86.
SEC (2002), Acceleration of Periodic Report Filing Dates and Disclosure Concerning WebsiteAccess to Reports, Release No. 33-8128, Securities and Exchange Commission,Washington, DC, available at: www.sec.gov/rules/final/33-8128.htm
AppendixFollowing prior studies (Bamber et al., 1993; Henderson and Kaplan, 2000; Knechel and Payne,2001), we estimate the following ARL model for obtaining the predicted or expected values ofaudit delays. Unexpected ARLs are actual ARLs minus their predicted values:
ARLit ¼ a0 þ b1TENit þ b2LNNAFit þ b3FCit þ b4SEGNUMit þ b5ABFEEit
þ b6EXTRAit þ b7LOSSit þ b8AUOPit þ b9YENDit þ b10BIG4itþ b11LNSIZEit þ b12NEWSit þ b13INSOWNRit þ industry dummiesþ year dummiesþ 1it
ðA1Þ
where (expected sign is shown in parentheses):
ARL : number of calendar days from fiscal year end to date of the auditor’s report;
TEN : number of years working with current auditors (2 );
Audit delays
49
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
LNNAF : natural log of non-audit service fees paid to incumbent auditors (2 );
FC : financial condition measured as Zmijewski’s financial condition index (þ );
SEGNUM : reportable segments of a client (þ );
ABFEE : abnormal audit fees (þ );
EXTRA : 1 if a firm reports extraordinary items and 0 otherwise (þ );
LOSS : 1 if a firm reports negative earnings and 0 otherwise (þ );
AUOP : 1 if the auditor’s opinion is modified or qualified and 0 otherwise (þ );
YEND : 1 if a firm’s fiscal year ends in December and 0 otherwise (þ );
BIG4 : 1 if an auditor is one of the Big N auditing firms and 0 otherwise (þ /2 );
LNSIZE : client firm size measured as the natural log of total assets (2 );
NEWS : client’s unexpected earnings defined as the difference between current andprior year’s earnings per share (EPS), divided by the absolute value of prioryear’s EPS (2 ); and
INSOWNR : ranked value of institutional ownership (2 ).
About the authorsVivek Mande is a Professor of Accounting at the Mihaylo College of Business and Economics atthe California State University, Fullerton. He is also the Director of the Center for CorporateReporting and Governance. His work has been published in Contemporary Accounting Research,Auditing: A Journal of Practice and Theory, Journal of Accounting and Public Policy, andnumerous other journals.
Myungsoo Son is an Associate Professor of Accounting at the Mihaylo College of Businessand Economics at the California State University, Fullerton. His research interests include auditquality, auditor fees, and earnings management. Myungsoo Son is the corresponding author andcan be contacted at: [email protected]
MAJ26,1
50
To purchase reprints of this article please e-mail: [email protected] visit our web site for further details: www.emeraldinsight.com/reprints
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)
This article has been cited by:
1. SAEED RABEA BAATWAH, Zalailah Salleh, Norsiah Ahmad. 2015. CEO characteristics and auditreport timeliness: Do CEO tenure and financial expertise matter?. Managerial Auditing Journal 30:8/9. .[Abstract] [PDF]
2. Wan Nordin Wan-Hussin, Hasan Mohammed Bamahros. 2013. Do investment in and the sourcingarrangement of the internal audit function affect audit delay?. Journal of Contemporary Accounting &Economics 9, 19-32. [CrossRef]
Dow
nloa
ded
by D
IPO
NE
GO
RO
UN
IVE
RSI
TY
At 2
1:37
02
Sept
embe
r 20
15 (
PT)