lowensohn et al - 2007 - an empirical investigation of auditor rotation requirements

ABSTRACT We add to the literature on mandatory auditor rotation by examining the Florida government audit environment in which there exist both rotation and non-rotation regimes and in

An Empirical Investigation

of Auditor Rotation Requirements

SUZANNE LOWENSOHN,* JACQUELINE RECK,**

JEFF CASTERELLA,* AND BARRY LEWIS*

September 2007

Preliminary draft

Please do not copy or quote

ABSTRACT

We add to the literature on mandatory auditor rotation by examining the Florida government audit environment in which there exist both rotation and non-rotation regimes and in which there exists an independent measure of the joint quality of the audit and of the financial statements of the reporting entity In this environment we compare rotating entities with entities that have established long-term relationships with their auditors We find that after controlling for other factors believed to be associated with report quality, governments that rotate receive higher quality reports

1 Introduction

The issue of mandatory audit firm rotation is once more in the forefront of the effort

to improve auditor independence and to increase investor confidence in audited financial statements The Sarbanes-Oxley Act of 2002 directed the Comptroller General to study the possible implementation effects of mandatory rotation Although the GAO (2003) study indicates that mandatory rotation may not be the most efficient way to improve auditor independence, it did not rule out the possibility if Sarbanes-Oxley requirements

do not have the desired effects As happened in the past, parties are lining up on both sides of the issue In a recent policy statement, the International Chamber of Commerce (2005) cited several adverse consequences of compulsory rotation of auditors,

“…including negative effects on audit quality.” More recently, Glass Lewis & Co (2007) issued a research report in which it recommended that audit committees adopt a 10-year rotation policy

The arguments for and against rotation are well-known by now, but evidence in support of those arguments has been anecdotal and/or flawed in several respects There is

no empirical support for the argument that rotation increases auditor independence by removing the incentive to sacrifice current judgments for the promise of long-term revenues from a client And while there is substantial empirical evidence that audit problems are more prevalent in new or early audits, that evidence is from a regime that does not include a policy of periodic rotation

This paper provides an empirical analog to the Dopuch, King and Schwartz (2001) experimental finding that rotation can increase auditor independence We analyze the government audit market in Florida to assess the effects of rotation on audit quality This market has two key advantages for examining rotation The first advantage is that some jurisdictions in Florida require rotation of auditors and still others have informally adopted such a policy This means that the population of government reporting entities includes both rotators and entities that have opted to retain auditors on a long-term basis The second advantage in this market is that the State Auditor General evaluates the quality of the audited financial statements of the reporting entities This is a joint measure

of the adherence to government auditing standards, generally accepted accounting principles, and Rules of the Auditor General Using these data, we are able to compare financial reporting quality across the rotation and non-rotation regimes Our results indicate that, after controlling for several variables thought to influence reporting quality, rotation is associated with a higher quality financial reporting system

The remainder of this paper is organized as follows We begin with a brief review of recent research most relevant to the issues addressed in this paper This is followed by a discussion of our data and the models and variables used to examine the effects of rotation We end with a presentation of our results and some concluding comments

2 Prior Research

Numerous recent empirical studies have provided evidence that short-term auditor-client relations are fraught with problems and that long-term relations are either not problematic or are actually beneficial Johnson, Khurana and Reynolds (2002) divided their sample of Big-6 audit clients into three audit firm tenure classification: short (2-3

years), medium (4-8 years) and long (9+ years) Using accruals-based measures of earnings quality, they find that short audit firm tenures are associated with lower quality

of earnings relative to the medium group but find no evidence that long audit firm tenure leads to diminishing earnings quality These results hold, as well, when the persistence of the accrual component of earnings is used as the dependent variable Also using accruals

as a proxy for earnings quality, Myers, Myers and Omer (2003) provide evidence that earnings quality actually increases with audit firm tenure Ghosh and Moon (2005) add to this line of literature by using the earnings response coefficient as a proxy for the market’s perception of audit firm tenure Their results indicate a positive relation between the market’s perception of earnings quality and tenure They also find that as tenure increases, the influence of prior years’ earnings on earnings forecasts increases

Using a slightly different approach, other studies have documented problems with short audit firm tenure In line with the accounting profession’s arguments against mandatory rotation, these studies examine audit failures, defined variously as fraud, SEC enforcement actions, and restatements An early paper by Raghunathan, Lewis and Evans (1994) finds a higher incidence of SEC actions against companies who had recently changed auditors Interestingly, they also find a slightly, but significantly, higher incidence for long audit firm tenures A similar study by Walker, Lewis and Casterella (2001) finds that while short-term audits are more likely to be associated with audit failures, long-term failures are significantly more costly More recently, Carcello and Nagy (2004) find that violations of Rule 10(b)-5 are significantly more likely in the early years of the auditor-client relationship Similarly, using restatements as a measure of audit failure, Stanley and DeZoort (2007) also find a disproportionate number of failures

in the early years Only Deis and Giroux (1992) report conflicting results Their study of Texas school districts finds that audit quality decreases with length of the auditor-client tenure In summary, nearly all of these studies are consistent with the auditing profession’s claim that new audits are problematic because the auditors lack the institutional memory and experience with the client

We question the extent to which the existing body of literature can provide useful insights about the potential impact of moving toward mandatory rotation Extant evidence

in the United States comes from a regime that does not require reporting entities to change auditors periodically Walker et al (2001) document the fact that the short-term tenure group in their sample is comprised largely of voluntary switchers who differ significantly from a control sample on dimensions that would predict trouble Most of the results above might be explained by a self-selection bias which has troubled companies either choosing to switch auditors or being dropped by their auditors

Two recent studies have bypassed this issue in clever ways and have produced a different set of insights into the rotation issue Nagy (2005) examines auditor changes caused by the demise of Arthur Andersen Nagy notes that this is not a perfect analogy to mandatory rotation because it is a one-time switch which has no pre-defined time horizon Furthermore, the notoriety of the dissolution of Andersen put a spotlight of sorts

on this group of clients On the other hand, the sample of firms is much more likely to be similar to a control group than the sample of voluntary switchers discussed in Walker et

al above In stark contrast to other studies using accruals as a measure of earnings quality, Nagy finds earnings quality improves after the switch, although for only the smaller half of the firms in the sample Nagy attributes this finding to an increased

vigilance by the new auditor For larger clients, he theorizes that the lack of a finite horizon may have given the clients more leverage over the new auditor While not exactly a rotation regime, this natural experiment offers a unique opportunity to study a large number of mandatory auditor changes

Recognizing the difficulty of making cross-regime comparisons in the real world, Dopuch et al (2001) use a laboratory experiment to examine the effect of mandatory rotation and/or retention on the independence of auditor-subjects across four different regimes Independence in this case is operationalized as the auditor’s willingness to issue reports biased in the favor of the manager The laboratory environment allows a much cleaner, but necessarily more abstract, test of the effects of rotation policies The environment also allows the authors to precisely define the economic incentives facing both the auditors and the managers in the different regimes Their results clearly indicate

a higher level of auditor independence in the two regimes that included mandatory rotation

To summarize, other than Nagy (2005) and Dopuch et al (2001), prior literature on rotation has lacked sufficient data to test empirically the effect of rotating some entities, but not others The purpose of this study is to test directly the arguments for and against rotation Regulators often suggest that reporting quality would improve if the number of long term auditor-client relationships were reduced by forcing a change in audit firm thereby increasing the number of newer, short term auditor-client relationships On the other hand, the profession believes that newer, short term auditor-client relationships are problematic and have lower reporting quality Therefore, they believe that long term auditor-client relationships should be allowed With the data described next, we are able

to inform the debate since we can isolate entities into the two distinct groups debated by regulators and the profession – entities with short term auditor-client relationships resulting from rotation policies, and entities with long term auditor-client relationships Once the groups are identified we are able to contrast the financial reporting quality of the groups

3 Research Approach

3.1 DATA

We surveyed 451 finance directors of Florida local municipalities and counties regarding their current auditor, audit firm tenure, and auditor rotation policies A total of

222 usable survey responses were received.1 We then obtained a measure of the entities’ financial report quality and audit quality for the fiscal year ending September 30, 2003 from the State of Florida Auditor General’s Office In accordance with Florida statutes, State Auditor General Office personnel use a checklist to review audited financial statements and audit reports annually for compliance with government auditing standards, generally accepted accounting principles, and Rules of the Auditor General

1

Generally speaking, entities with a rotation policy would rotate before reaching an auditor-client relationship of 7 years

We received 7 survey responses indicating that a rotation policy was in place, yet the entities had auditor-client relationships of 7 or more years In some cases, entities may have available a same auditor “renewal option” which can extend the length of the auditor-client relationship Since it is not clear whether or not these 7 entities are clearly rotators they have been dropped from our analyses However, inclusion of these entities does not significantly change the reported results

We were provided access to the checklist and entity financial records to gather the joint quality measure, which we reconciled to publicly-available State Auditor General Office reports

Other variables, described below, were obtained from publicly available sources such as financial statements and information submitted to the Florida Auditor General’s Office, the Municipal Year Book (2005), and the Florida Department of Financial Services website

3.2 MODEL

The purpose of this research is to provide a better test of the effects of auditor rotation on report/audit quality Carcello and Nagy (2004) acknowledged an important caveat to their results and the results reported in earlier studies The caveat is that all

prior empirical studies relied on auditor changes that were voluntary rather than auditor

changes that were based on a rotation policy In this study, we do not have that constraint since our data includes entities that have adopted a rotation policy

Table 1 describes our data in terms of rotation policy [yes/no] and auditor tenure [short/medium/long] The proponents of audit firm rotation argue that entities should be rotated away from long term audit relationships to new short term audit relationships In other words, they argue that rotation will shorten auditor-client relationships, and thus result in better audits Rotating entities with short term auditor-client relationships are in Cell 1 On the other hand, the opponents of audit firm rotation argue that entities should not be forced to rotate audit firms but rather should be allowed to remain in long term audit relationships In other words, they argue that long term auditor-client relationships are better Entities with long term relationships are in Cell 5 To test these opposing arguments, our model compares the financial reporting quality of Cell 1 to that of Cell 5

- Insert Table 1 about here -

The model we use is similar to that of Carcello and Nagy (2004) with adaptations made for government data:

Comments = FC + Mgr + Growth + LnRev + B4 + ACom + RSHORT_LONG + e

Where,

Comments, the dependent variable, is the number of reporting deficiencies associated

with the entity’s 2003 financial report from the Florida State Auditor General’s Office review checklist A lower number indicates higher financial reporting quality

FC, financial condition, is an indicator variable denoting the existence of financial

problems specified in the Florida statutes related to meeting long or short-term requirements; failure to pay creditors; failure to make employee-related payroll, tax or

benefit transfers; or fund balance and retained earnings deficits Baber, Brooks, Ricks (1987) suggest that audit risk varies inversely with financial strength and, therefore, we expect financial problems to be associated with lower reporting quality Given the nature

of the dependent variable measurement, the predicted sign of the coefficient is positive

Mgr, is an indicator variable for entities which have a council-manager form of

government Prior research suggests that a council-manager form of government is consistent with greater demand for external monitoring (Raman and Wilson 1994, Zimmerman 1977) and higher reporting quality than other forms of government (Jensen and Payne 2005) Therefore, we predict a negative coefficient on this variable indicating that the council-manager form of government is associated with higher report quality

Growth measures the average annual growth in total revenue on a percentage basis for

the period 1998-2003 We predict a positive sign, since rapid growth entities are more complex to audit (Francis and Wilson 1988) and should be associated with lower quality reports

LnRev, entity size, is measured as the natural logarithm of the entity’s revenues for fiscal

year end 2003 We do not predict a sign, since there is conflicting evidence Although Baber (1983) and Evans and Patton (1987) argue that demand for auditing increases with size, others find that audit quality as measured by oversight review is a decreasing function of size (Deis and Giroux 1992; O'Keefe and Westort 1992)

B4 indicates that the entity is audited by a Big4 accounting firm While Big 4 firms are

generally associated with higher levels of reporting quality (DeAngelo 1981), this has not always been the case in governmental samples (Lowensohn, Johnson, Elder, and Davies 2007), therefore we do not predict a sign on this coefficient

ACom denotes the existence of an audit committee and is expected to be associated with

higher quality reporting, consistent with Jensen and Payne (2005)

RSHORT_LONG is the primary test variable which contrasts the reporting quality of

recent rotators with that of non-rotator entities in long-term auditor relationships This is done by assigning a value of +1 to rotators with auditor tenure of three years or less; a value of -1 to non-rotators with auditor tenure of seven years or more; and a value of zero

to all other entities Based on the experimental results of Dopuch et al (2001), we expect short term rotators to be associated with better reporting quality than non-rotators with long tenures

4 Analysis and Results

4.1 PRIMARY TEST

Table 2 outlines the cell sample sizes for three levels of tenure and two levels of rotation A traditional modeling approach would test the main effect of rotation [all rotators vs all non-rotators], the main effect of tenure [for example all short term relationships vs all long term relationships], and perhaps an interaction term However, such modeling does not answer the long-standing argument about auditor rotation For

example, the interpretation of the result for the main effect of rotation would be: across all levels of tenure, how do all rotators compare to all non-rotators But the central argument is not about all rotators, it is about rotators in the newer auditor-client relationships Furthermore, the argument is not about all rotators, but rather the non-rotators in long term relationships Therefore, a rotation main effect variable is too coarse of a test Similarly, the interpretation of the result for the main effect of tenure [short vs long] would be: Regardless of rotation policy, how do all short term relationships compare to all long term relationships? Again, the central argument is not about all short versus all long relationships, it is about short relationships created by rotation versus the status quo, which are long relationships not affected by rotation Lastly, an interaction term that crosses rotation [yes/no] by tenure [short/medium/long]

would tell us how the effect of rotation is moderated, overall, by the effect of tenure

Here too, the interaction term is too broad based and is ill-equipped to test the policy effects of adopting a rotation policy

As indicated earlier, the key test of interest is focused squarely on Cells 1 and 5 Adoption of a rotation policy manifests itself by shifting audits from long tenure [Cell 5]

to short tenure audits [Cell 1] (see Table 1) An a priori specific test of this type is best analyzed using a contrast coded variable Buckless and Ravenscroft (1990) discuss the advantages of contrast coded variables Other recent accounting studies that have used them for planned comparisons include Hales (2007); Hirst, Koonce, and Venkataraman (2007); Hammersley (2006); and Kennedy, Kleinmuntz and Peecher (1997) To compare

Cell 1 to Cell 5 we use a contrast coded variable RSHORT_LONG Cell 1 contains short

tenure rotators and is coded +1 and Cell 5 contains long tenure non-rotators and is coded -1 All remaining cells (2,3,4) are coded as zero, and included in the analysis to improve the statistical power of our test

Using ordered logistic regression we regress the number of reporting deficiencies

(Comments) on five control variables and the test variable RSHORT_LONG Results are

presented in Table 3

- Insert Tables 2 and 3 about here -

The overall model is significant at p=.052 and our primary test variable

(RSHORT_LONG) is negative and significant at p=.02 (one tailed) The interpretation is

that entities that have recently rotated audit firms have significantly fewer deficiencies when compared to entities that do not rotate auditors and are in long tenure audits This finding is consistent with regulators’ concern that long tenure audits have lower audit quality

In addition, two control variables are significant We find that larger entities

(Ln_Rev) are associated with fewer deficiencies at p<.004 Also, entities with a Big 4

auditor tend to have more deficiencies (p<.037), consistent with prior government research

4.2 ADDITIONAL TESTS

Several diagnostic tests were performed All variance inflation factors are less than 1.4, and all condition indices are less than the cut-off of 30 recommended by Belsley, Kuh, and Welsch (1980), indicating that multicollinearity is not a problem Outlier observations were identified for undue influence on our primary results Removal of outliers does not change the result for RSHORT_LONG A test for heteroskedasticity was also performed The null of normally distributed residuals could not be rejected Last, we increased the cutoff for long tenure to 9 years (consistent with Carcello and Nagy 2004) The results for RSHORT_LONG were consistent with the results in Table

3

We also ran a simple non-ordered logistic regression model where the dependent variable was coded as one if there were 1 or more deficiencies; zero otherwise The results for RSHORT_LONG are the same as shown in Table 3 and significant at a p-value <.05 (one tailed)

As a follow up test, we analyzed the same model after replacing the primary test variable with a new variable that contrasts short tenure non-rotators in Cell 3 (i.e voluntary changers) with long tenure non-rotators in Cell 5 The results (not shown) indicate no significant difference between voluntary changers and long tenure non-rotators The interpretation is that not all newer audits have higher financial reporting quality, only those under a rotation policy have higher financial reporting quality

5 Conclusions

By examining the Florida government environment, we were able to provide a real-world comparison of different audit regimes In this empirical analog to Dopuch et al (2001), we found support for their experimental results Using measures of reporting quality provided by reviews of the Florida State Auditor General’s Office, we found that entities that have adopted a policy of rotation have fewer financial reporting and auditing deficiencies than entities with long-term associations with their auditors

We recognize that the government audit environment is not the ideal proxy for the public company audit environment, but we believe that the evidence provided by this study, in conjunction with the Dopuch et al (2001) experiment and Nagy’s (2005) examination of former Andersen clients casts significant doubt on both the risks associated with new audits and the benefits associated with long-term auditor tenure

REFERENCES

BABER, W R “Toward Understanding the Role of Auditing in the Public Sector.”

Journal of Accounting and Economics 5 (1983): 213-227

BABER, W.R.; R E BROOKS; AND W.E RICKS “An Empirical Investigation of the

Market for Audit Services in the Public Sector.” Journal of Accounting Research 25

(1987): 293-305

BELSLEY, D.A.; E KUH; AND R WELSCH “Regression Diagnostics: Identifying Influential Data and Sources of Collinearity.” (1980) New York: John Wiley and Sons

BUCKLESS, F.A.; and S.P RAVENSCROFT “Contrast Coding: A Refinement of

ANOVA in Behavioral Analysis.” The Accounting Review 65 (1990): 933-945

CARCELLO, J V., and A L NAGY “Audit Firm Tenure and Fraudulent Financial

Reporting.” Auditing: A Journal of Practice and Theory 23 (2004): 55-69

DEANGELO, L.E “Auditor Size and Audit Quality.” Journal of Accounting & Economics 3 (1981): p183-199

DEIS, D.R., and G A GIROUX “Determinants of Audit Quality in the Public Sector.”

The Accounting Review 67 (1992): 462-479

DOPUCH, N.; R.R KING; and R SCHWARTZ “An Experimental Investigation of

Retention and Rotation Requirements.” Journal of Accounting Research 39 (2001):

93-117

EVANS, J H., and J M PATTON “Signaling and Monitoring in Public-Sector

Accounting.” Journal of Accounting Research 25 (1987): 130-158

FRANCIS, J., and E WILSON “Auditor Changes: A Joint Test of Theories Relating to

Agency Costs and Auditor Differentiation.” The Accounting Review 63 (1988):

663-682

GHOSH, A., and D MOON "Does Auditor Tenure Impair Audit Quality?" The Accounting Review 80 (2005): 585-612

GLASS LEWIS & CO “Speak No Evil.” Yellow Card Trend Alert published by Glass

Lewis & Co LLC May 21, 2007: 1-40

HALES, J “Directional Preferences, Information Processing, and Investors' Forecasts of

Earnings.” Journal of Accounting Research 45 (2007): 607-628

HAMMERSLEY, J “Pattern Identification and Industry-Specialist Auditors.” The

Accounting Review 81 (2006): 309-336

HIRST, E D.; KOONCE, L.; and V SHANKAR “How Disaggregation Enhances the

Credibility of Management Earnings Forecasts.” Journal of Accounting Research 45

(2007): 811-837

INTERNATIONAL CHAMBER OF COMMERCE “The Adverse Effects of Compulsory Audit Firm Rotation.” (March 4, 2005) Available at http://www.iccwbo.org/home/statements_rules/statements/2005/audit_rotation_ICC% 20_statement.asp

JENSEN, K L., and J L PAYNE “Audit Procurement: Managing Audit Quality and

Audit Fees in Response to Agency Costs.” Auditing: A Journal of Practice and Theory 24 (2005): 88-98

JOHNSON, V.E.; I.K KHURANA; and K.J REYNOLDS “Audit-Firm Tenure and the

Quality of Financial Reports.” Contemporary Accounting Research 19 (2002):

637-660

KENNEDY, J.; D N KLEINMUNTZ; and M E PEECHER “Determinants of the

Justifiability of Performance in Ill-Structured Audit Tasks.” Journal of Accounting Research 35 (1997): 105-123

LOWENSOHN, S.; L E JOHNSON; R E ELDER; and S DAVIES “Auditor Specialization, Perceived Audit Quality, and Audit Fees in the Local Government

Audit Market.” Journal of Accounting and Public Policy (2007) Forthcoming

MUNICIPAL YEAR BOOK (2005) International City/County Management Association: Washington, D.C

MYERS, J.; L MYERS; and T OMER "Exploring the Term of the Auditor-Client Relationship and the Quality of Earnings: A Case for Mandatory Auditor Rotation?"

The Accounting Review 78 (2003): 779-99

NAGY, A.L “Mandatory Audit Firm Turnover, Financial Reporting Quality, and Client

Bargaining Power: The Case of Arthur Andersen.” Accounting Horizons 19 (2005):

51-68

O'KEEFE, T B., and P J WESTORT “Conformance to GAAS Reporting Standards in Municipal Audits and the Economics of Auditing: The Effects of Audit Firm Size,

CPA Examination Performance, and Competition.” Research in Accounting Regulation 6 (1992): 39-77

RAGHUNATHAN, B., B.L LEWIS; and J.H EVANS III “An Empirical Investigation

of Problem Audits.” Research in Accounting Regulation 8 (1994): 33-58

RAMAN, K K., and E R WILSON “Governmental Audit Procurement Practices and

Seasoned Bond Price.” The Accounting Review 69 (1994): 517-538

STANLEY, J.D., and T.F DEZOORT “Audit Firm Tenure and Financial Restatements:

An Analysis of Industry Specialization and Fee Effects.” Journal of Accounting & Public Policy 26 (2007): 131-159

U.S GOVERNMENT ACCOUNTING OFFICE (GAO) 2003 Public Accounting Firms: Required Study on the Potential Effects of Mandatory Audit Firm Rotation

Report to the Senate Committee on Banking, Housing, and Urban Affairs and the House Committee on Financial Services November Washington, D.C.: Government Printing Office

WALKER, P.L., B.L LEWIS; AND J.R CASTERELLA “Mandatory Auditor

Rotation: Arguments and Current Evidence.” Accounting Enquiries 10 (2001):

209-242

ZIMMERMAN, J L “The Municipal Accounting Maze: An Analysis of Political

Incentives.” Journal of Accounting Research 15 (1977): 107-144