chow and rice - 1982 - qualified audit opinions and auditor switching

If auditors do differ in their pro- pensity to issue qualified opinions, do firms after receiving qualified opinions switch to auditors issuing proportionally fewer qualified opin-

Author(s): Chee W Chow and Steven J Rice

Source: The Accounting Review, Vol 57, No 2 (Apr., 1982), pp 326-335

Stable URL: http://www.jstor.org/stable/247018

Accessed: 28/04/2014 23:50

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive We use information technology and tools to increase productivity and facilitate new forms

of scholarship For more information about JSTOR, please contact support@jstor.org

American Accounting Association is collaborating with JSTOR to digitize, preserve and extend access to The Accounting Review.

http://www.jstor.org

Vol LVII, No 2

April 1982

NOTES

Chee W Chow and Steven J Rice

ABSTRACT: This study focuses on the influence of qualified opinions on auditor switches Results from a random sample of SEC-registrants support the contention that firms switch auditors more frequently after receiving qualified opinions However, it was not found that firms that have received qualified opinions switch systematically to audit firms with a history of rendering proportionally fewer qualified opinions Also, limited results suggest that qualified firms which switch auditors do not tend subsequently to receive more clean opinions

I INTRODUCTION

LARGE-SCALE auditing scandals like

Continental Vending, Equity Fund-

ing, and National Student Market-

ing have focused the public's attention on

auditor "independence." Critics fre-

quently charge that the asymmetry of

power between auditors and managers

may seriously weaken the auditor's

ability to exercise freely his professional

judgment A statement by Sterling [1973,

p 66] is representative:

The major problem facing public account-

ing today is its lack of power First, in com-

far outweighs the authority The public ac-

countant must act judicially but he has not

been given the power to enforce his rulings

The authority is lessened further by the

firms Resignation from an engagement might

were not for the fact that other firms may

take the engagement and issue an opinion

When one considers the fact that ac-

countants must judge management, [the

imbalance between management's power and the accountant's power] is not only un- desirable, it is intolerable

The auditor-client relationship has also been a concern of accounting policy makers The AICPA has promulgated voluminous guidelines for auditing and reporting procedures and commissioned

a study of auditor responsibility The SEC has been increasing its disclosure requirements for auditor-client relation- ships.' Even the U.S Congress has voiced its concern for auditor "independence"

194 [SEC, 1976]

The authors gratefully acknowledge the assistance of Professors John Chiu, Jim Jiambalvo, Robert Bowen, and Jamie Pratt

Chee W Chow is Lecturer in Account- ing and Steven J Rice is Assistant Profes- sor of Accounting, both at the University

of Washington

Manuscript received November 1980

Revision received April 1981

Accepted May 1981

326

in the Moss Committee Report [U.S

House of Representatives, 1976] and the

Metcalf Committee Staff Report [U.S

Senate, 1976]

Policy makers attach importance to

the ability of managers to "shop" for

auditors The concern is that managers

can pressure auditors into giving clean

opinions by threatening to switch to a

new auditor It is argued that managers

wish to avoid qualified opinions because

they may affect (a) the market price of

the firm's common stock and (b) the

manager's compensation Moreover, the

manager often holds some of the firm's

shares in his personal portfolio either

through open market purchases or

through stock options granted as part of

his compensation Compensation may

be tied to the reported earnings and the

reliability of those reports comes into

question when there is a qualification If

(a) qualified opinions are seen (by man-

agers) to reduce security prices and/or

manager compensation and (b) managers

control the auditor selection decision,

then we should be able to observe that

qualified opinions are followed by audi-

tor switches The purpose of this paper is

to present empirical evidence relating to

the following four questions concerning

auditor switching:

1 Do firms switch auditors more often

after receiving qualified opinions

than after receiving clean opinions?

2 Do auditors differ in their propen-

sity to issue qualified opinions?

3 If auditors do differ in their pro-

pensity to issue qualified opinions,

do firms after receiving qualified

opinions switch to auditors issuing

proportionally fewer qualified opin-

ions?

4 When firms switch auditors after

receiving a qualified opinion, are

they more likely to receive clean

opinions the following year?

Positive answers to these questions would imply that firm managements may use auditor switches to pressure auditors into issuing clean opinions Such evidence could prompt the SEC and/or the FASB

to further consider their policies regard- ing firms changing auditors We present our analysis as follows: Section II reviews some prior research regarding auditor switching Section III describes our data collection procedures and the tests per- formed We draw some conclusions in Section IV and discuss the implications of our results for future research and public policy

II SOME PREVIOUS PUBLISHED STUDIES

ON AUDITOR SWITCHING

A number of authors have discussed the role of accounting disputes in firms' choice of auditors.2 However, only one study has provided systematic evidence

on this issue Burton and Roberts [1967] examined 83 auditor switches made by Fortune 500 firms between 1952 and 1965 Using questionnaires as their primary tool, the authors determined that ac- counting standards disputes were the major causes for only six switches The other switches were related to changes

in management, the demand for addi- tional services or needs arising out of new financing This led Burton and Roberts to conclude that competition among audi- tors was not a cause of loose application

of accounting standards This conclusion

is suspect because if firms switch auditors

so as to gain more control over the con- tents of issued financial statements, it is not in the firms' own interest to reveal this action to outsiders Further, if auditors were already forced by compe- tition to apply loose accounting stan- dards, firms would have no need to

2

Goldman and Barlev [1974], Nichols and Price [1976], Burton and Roberts [1967], and Fried and Schiff [1981 ]

328 The Accounting Review, April 1982

switch auditors due to accounting dis-

putes

Even though the Burton and Roberts

study fails to resolve whether accounting

standards play a major role in auditor

switching, it does identify other variables,

such as management changes and new

financing, which affect firms' choice of

auditors These findings are supported

by Carpenter and Strawser [1971] and a

study by an ad hoc committee of the

AICPA [1971]

III DATA COLLECTION AND TESTS

A The Association Between Qualified Au-

dit Opinions and Auditor Switches

One way of testing whether firms

switch auditors after receiving qualified

opinions would be to classify all of the

firms reporting to the public over a given

time period into four categories:

1 Received a qualified opinion and

switched auditors subsequently,3

2 Received a qualified opinion and

kept the same auditor,

3 Received a clean opinion and

switched auditors subsequently, and

4 Received a clean opinion and kept

the same auditor

Then a four-cell contingency table could

be set up and tested using the Chi-square

test for independence of classification A

significant value for Chi-square would

mean that switching was not independent

of receiving a qualified opinion

Our data were collected from the 1973

and 1974 volumes of the Leasco Disclo-

sure Journal.4 All companies that had a

qualified opinion (excluding consistency

exceptions) in 1973 and all companies

that changed auditors between the 1973

and 1974 fiscal year-ends were selected

This procedure yielded a count for the

first three of the categories listed above

The count for the fourth category (clean

opinion and did not switch) was arrived

at indirectly by counting the total number

TABLE 1

THE ASSOCIATION BETWEEN QUALIFIED AUDIT OPINIONS AND AUDITOR SWITCHING

Qualified Unqualified Row totals

Actual x2 (1 d.f.)= 196.81 Significant x2 (1 d.f.)=6.635 for o=0.0

of firms reporting and subtracting the total of the first three categories Table 1 shows the contingency table with the count for each cell and the results of the Chi-square test for independence of While this overall test provides inter- esting information, it ignores the effect of classification The test clearly supports the notion that changing auditors is not independent of receiving a qualified audit opinion.'

other variables that might influence a firm to change auditors In order to address this issue, we collected data for various subsets of the firms in Table 1 For each firm in a subset, we collected additional data from the Disclosure Jour- nal on several variables which had been identified in previous studies as related to changing auditors.6 The following vari- ables were recorded simply as being

3 We assume that, in the absence of other causes, an auditor switch due to a qualified opinion will be made during the next fiscal year If this is not true, the omission

significant relationship

4 The Disclosure Journal includes all firms that file with the SEC

To allow for the possibility that some firms may switch auditors in anticipation of a qualified opinion,

we reconstructed Table 1 by including in Cell 1 57 addi- tional firms which switched auditors before their 1973 fiscal year-end rather than after The results were not significantly altered

[1971], and AICPA [1971]

present or not present for each firm in

fiscal 1974: a management change, some

merger activity, new financing, an ac-

counting dispute with the auditor, and

any other item identified by management

as the reason for changing auditors

Data on the name of the auditor, the

name of the new auditor (if switched),

the type of opinion, the industry (three-

digit SIC code), and revenues earned in

1973 were also recorded

The approach to forming the subsets

and testing them was to randomly select

141 firms from cell 4 in Table 1 (about

1.75 percent of the total cell 4 firms), and

to randomly select the same percentage

of firms from each of the other three

cells.7 This sample size was considered to

be small enough to enable us to gather the

additional information needed in a rea-

sonable amount of time The same per-

centage in each cell was used in order to

preserve the proportions that existed in

the Leasco population The relation of

these variables to auditor switching was

tested in an equation of the form :8

S = a + bjQ + b2Mg + b3Mr

+ b4N + b5X (1)

where:

S=switched auditors (1) or did not

switch (0)

Q = received a qualified opinion (1)

or unqualified (0)

Mg = management change took place

(1) or did not take place (0)

Mr = merger activity present (1) or not

present (0)

N=new financing arranged (1) or

not mentioned (0)

X= some reason (other than those

above) was given for switching

(1) or no other reason listed (0)

It may appear that ordinary least

squares (OLS) would be a reasonable tool

to evaluate Equation (1) However, such

is not the case Theil [1972, sec 12.5]

points out that there are three major problems with OLS when the dependent variable is dichotomous First, the esti- mating equation is unbounded and there

is no guarantee that the predicted value

of the dependent variable (a probability) will be in the (0, 1) interval Second, the residuals will be heteroscedastic and OLS would yield inefficient estimators Finally, because the dependent variable

is dichotomous, the residuals cannot be assumed normal This means that the typical F and t tests of significance cannot

be used

We also rejected multivariate discrimi- nant analysis (MDA) because of its dis- tributional requirements Eisenbeis [1977] and Ohlson [1980] point out that MDA requires the variance-covariance matrices of the independent variables to

be the same for both groups (switched and non-switched) We found that this does not hold for our sample Also, MDA assumes normally distributed indepen- dent variables, which does not hold for our case

Theil [1971] and McFadden [1973] point out that conditional logit analysis essentially avoids all of the problems discussed above This is a maximum- likelihood estimation procedure which applies a transformation to the dependent variable McFadden [1973, p 119] indi- cates that this method yields estimators that are asymptotically efficient and normally distributed He also presents simulation results which suggest that the approximation is reasonably good even

in quite small samples This means that one can construct approximate large- sample confidence bounds and tests of

responding to the number of firms in cell 1

8 We did not include "accounting dispute" as an in- dependent variable The SEC 8-K requirement of ac- counting-dispute disclosure applies only to firms that switched auditors Our data source did not reveal whether some firms that did not switch auditors also had had accounting disputes

330 The Accounting Review, April 1982

TABLE 2

VARIABLES RELATED TO AUDITOR SWITCHING

Asymptotic

X2 (5 d.f.) =9.242

t-score of 1.96 =significance level of 05 and

t-score of 2.33 =significance level of 0 1

hypotheses for parameters Because of

this, we chose a conditional logit analysis

for Equation (1).9 The results appear in

Table 2

McKelvey and Zavoina [1975] show

that the significance of the logit model can

be tested by taking -2 times the log-

likelihood ratio The resulting statistic is

distributed as a Chi-square with degrees

of freedom equal to the number of inde-

pendent variables This statistic is 9.242

for our model and is significant at the 1

level

The t-scores in Table 2 show clearly

that qualification is the only significant

variable in explaining switching How-

ever, given the independent variables

used, the coefficients may have been

biased by multicollinearity We did not

conduct a formal test for this Table 3

presents the correlations among the inde-

pendent variables Even though a few

are statistically significant, none of the

values seems high enough to cause

concern)'0

B Difference Among Auditors' Percent-

ages of Qualified Opinions

An incentive for switching auditors

after receiving a qualified opinion can be

TABLE 3

CORRELATIONS AMONG THE INDEPENDENT VARIABLES

(Numbers in Parenthesis are significance levels.)

(.023)

(.373) (.602)

inferred if there is a different "tendency

to qualify" across auditors A proxy for this tendency may be the observed per- centage of qualified opinions given by an auditor However, a raw calculation of this percentage is not sufficient There is much evidence that auditors specialize

propriate tool We conducted the same analysis with probit and obtained almost indistinguishable results from the logit analysis

10 Farrar and Glauber [1967] provides a detailed explanation of tests for multicollinearity Kmenta [1971] points out that some degree of multicollinearity almost always exists, but the point at which it becomes

"harmful" has never been satisfactorily determined According to one criterion, if the overall equation is significant, but none of the t statistics for the regression coefficients (other than the regression constant) is sig- nificant, then multicollinearity is regarded as harmful According to this criterion, we do not have "harmful" multicollinearity Further, several statisticians who examined our data unanimously declared that the num- bers were so low that multicollinearity was not likely

to be a problem We tried to assess the potential sig- nificance of the multicollinearity problem by generating

estimate Equation 1 three more times Merger had a significant coefficient in one of the new regressions Otherwise the results were unchanged This suggests that multicollinearity is not a significant problem in our study

by client size and industry " If operating

conditions, legal constraints, and firm

characteristics were to differ across indus-

tries, this could cause observed differ-

ences in the percentages of qualified

opinions across auditors which really

reflect industry differences Indeed, War-

ren [1975, 1980] reports a significant

relationship between industry class, firm

size, and the receipt of a qualified opin-

ion Warren [1980] also finds auditor

identity to be significantly related to

qualified opinions However, he does not

investigate which auditors are more

likely to issue such opinions More im-

portant, Warren uses the parametric

analysis of variance (ANOVA) to analyze

his data Unfortunately, the dependent

variable in his studies is a dichotomous

variable, which seriously violates the

basic ANOVA assumption of normally

distributed dependent variables, and

casts doubt on his results.'2

Shank and Murdock [1978] also in-

vestigate the differences among auditors'

percentages of qualified opinions Unlike

Warren, they do not find auditors to be

significantly different after accounting

for systematic risk Also, they do not find

either size or industry to have significant

explanatory power for the incidence of

qualified opinions Shank and Murdock

apparently also use ANOVA on a di-

chotomous dependent variable If so,

their results suffer from the same prob-

lems as Warren [1975, 1980] In any case,

the divergent results obtained by these

researchers suggest the need for further

investigation

To take into account the dichotomous

nature of the dependent variable, the

approach used for testing Question II

was again a logit analysis, with qualifica-

tion as the dependent variable and size,

industry, and auditor as the independent

variables in a linear equation Size was

measured by revenues earned in 1973 The industry was classified as one of seven, depending on the three-digit SIC

code as follows :13

0-149 Mining 200-399 Manufacturing 400-499 Transportation and public

utilities 500-519 Wholesale trade 520-599 Retail trade 600-699 Finance, insurance, real es-

tate 700-899 Services The auditor was classified as one of nine: the Big Eight firms and all others In order to have enough data points, 141 firms were randomly selected from cells

2, 3, and 4 to match the number of firms which comprise the whole population of cell 1 in Table 1 Thus, the analysis was run on 564 firms with 15 independent variables: size, six industry class dummy variables, and eight auditor dummy variables The effect of the service indus- try and "other" auditors are included in the intercept The results are reported in Table 4

The Chi-square for this model is sig- nificant at the 001 level Of the inde- pendent variables, size is significantly and negatively related to qualifications Also, significant industry effects exist for transportation/public utilities, retail, and perhaps also mining Having taken into

I IZeff and Fossum [1967] and Rhode, Whitsell, and Kelsey [1974] report such evidence Another factor which may limit the usefulness of the "observed per- centage qualified" proxy is potential auditor specializa- tion by auditor quality and self-selection by audit clients Chow and Rice [1981 ] provide a discussion of this issue

tions is available in Neter and Wasserman [1974]

13 With the exception of the mining industry, these industry definitions are identical to Warren's [1975, 1980] Our definition of the mining industry is more en- compassing than Warren's However, only one firm's in- dustry class was affected by this difference in definition

332 The Accounting Review, April 1982

TABLE 4

VARIABLES RELATED TO RECEIPT OF A

QUALIFIED OPINION

Asymptotic

Transportation/public

Finance, insurance,

Arthur Andersen

Arthur Young &

Deloitte Haskins &

Peat, Marwick,

Price Waterhouse &

Estimated R2 =.1745

X2 (15 d.f.)=79.1398

* t-score of 1.96= significance level of 05 and

t-score of 2.33 =significance level of 01

account size and industry, three auditors

still have significantly positive coeffi-

cients: Arthur Young & Company,

Coopers & Lybrand, and Touche Ross &

Co These auditors can be interpreted as

having a higher tendency to qualify,

ceteris paribus Again, multicollinearity

does not appear to be of major concern

Of the 105 pairwise correlations among

the independent variables, one was

slightly above 3 (significant beyond 00 1)

and two were between 2 (significant

beyond 001) and 3 The rest were all

much smaller (and usually insignificant)

Thus, Question II has been answered

affirmatively; there may be some in-

centive to switch auditors in seeking

clean opinions

C The Direction of Auditor Switches

by Qualified Firms Can it be shown that firms, after re- ceiving qualified opinions, tend to switch more away from Arthur Young & Com- pany, Coopers & Lybrand, and Touche Ross & Co.? To test this question, a Chi- square analysis was run on the 141 firms which received qualified opinions and then switched- auditors The three signifi- cant auditors in Table 4 were classed as High Tendency (HT) and the others were classed as Lower Tendency (LT) Table 5 shows the results The numbers in paren- theses below the actual frequencies repre- sent the expected frequencies of those cells if there were no relationship between old and new auditor in terms of tendency

to qualify If the answer to Question III

is positive, we would expect more than

120 firms to have switched from HT to

LT (cell 2) and less than five to have switched from HT to HT (cell 4) Clearly, there is no significant relationship here, and the Chi-square value of 2561 (sig- nificance=.6128) bears this out.'4

D Auditor Switches and Subsequent Audit Opinions

The last question was whether a switch

in auditors actually results in receiving a clean opinion To answer this question properly, it would be necessary to hold other factors, such as the firm's financial condition, constant We used the quali- fied-and-no-switch group as a control

As we discuss later, this may not provide

14 We repeated this analysis by including in the HT group auditors from Table 4 whose t-scores were close to significance, i.e., Ernst & Whinney and Peat, Marwick, Mitchell & Co The results were quite similar We also repeated this analysis by re-grouping the data into HT and LT groups based on "folklore" rather than on empirical data "Folklore" might indicate that in the minds of management a certain group of auditors, say, non-Big Eight auditors, had a lower tendency to qualify

We investigated this using a Big Eight versus non-Big Eight dichotomy and again found no significant rela- tionship

TABLE 5

DIRECTION OF AUDITOR SWITCHING BY

QUALIFIED FIRMS

New Auditor

Old

Auditor

X2 (1 d.f.)= 2561

adequate control for other variables

Thus, our results can only be viewed as a

preliminary look at this question

Recall that in testing Question II we

drew a sample of 141 firms for each cell

and looked up the auditor, size, industry,

etc for fiscal 1973 Now, for each firm in

the qualified-and-switched and the quali-

fied-and-no-switch cells, we collected

data on the audit opinion for fiscal 1974

Using both the Disclosure Journal and

microfiche copies of 10-K reports, we

located the 1974 audit opinion for all 141

firms in the qualified-and-no-switch

group and for 132 firms in the qualified-

and-switched group A Chi-square test

was applied to these firms to determine if

a relationship exists between auditor

switching and the subsequent audit opin-

ion Table 6 reports these results

If switching helps reduce the incidence

of qualifications, we would expect fewer

than 98 firms in cell 1 and more than 34

firms in cell 3 Just the opposite is the

case Although the Chi-square value of

1.92 was not significant at the 1 level, the

data suggest that those firms that did not

switch tended to get more clean opinions

TABLE 6

AUDITOR SWITCHES AND SUBSEQUENT

AUDIT OPINIONS

Type of Opinion in Year Following Qualification

Not

Actual X2 (1 d.f.) = 1.924

Significant x2 (1 d.f.) = 2.71 for a = 10

in the following year than those firms that switched 1

One possible reason for this tendency

is the failure to control for the effects of extraneous variables like financial con- dition But perhaps a more basic issue is involved: if a firm knew that the reason for the qualified opinion would not exist the following year, it would not have an incentive to switch auditors for this rea- son Thus, the qualified-and-no-switch group might be expected to have a higher percentage of clean opinions in the fol- lowing year Another possible explana- tion relates to the type of qualification If managers view certain types of qualified opinions as "benign," again there would

be no incentive to switch auditors We did not pursue this potential explanation with

15

We could not find 1974 audit opinions for nine of the firms in the qualified-and-switched group In order

to evaluate the potential impact of these nine firms on the Chi-square test, we did the test assuming that all nine firms received qualified opinions in 1974 The Chi- square value was 2.74 which was significant at the 10 level We did the test again assuming that all nine firms received clean opinions in 1974 There was no significant relationship

334 The Accounting Review, April 1982 our data Our preliminary results sug-

gest, however, that further research is

warranted

IV CONCLUSIONS AND IMPLICATIONS

Our results support the contention that

firms tend to switch auditors after re-

ceiving a qualified opinion The incentive

for such switching may exist due to

observed differences among auditors'

percentages of qualified opinions issued

However, our analysis of switching firms

does not indicate that qualified firms

tend to switch to lower percentage-quali-

fied auditors We also find that firms that

switched auditors after a qualified opin-

ion, compared to the qualified firms that

did not switch, are not more likely to

receive a clean opinion the following

year The evidence tends to support the

opposite While our results are prelimi-

nary and based on a limited time span,

they suggest that further work in this area

may be fruitful

Even if our major results are replicated

in more extensive studies, it does not

necessarily imply that there is cause for

concern over auditor lack of indepen-

dence Recently, Jensen and Meckling [1976], Watts [1977], Smith and Warner [1979], and others have explored the role

of external auditing in business firms They postulate that the manager can benefit from letting shareholders and bondholders monitor his allocation of the firm's resources The choice of an external auditor is part of this monitoring contract, and it is in the manager's self- interest to select an auditor who has es- tablished a reputation for integrity The implication of this analysis is that when a firm manager chooses (or switches) an auditor, he is guided by the expected reactions from shareholders and bond- holders Thus, auditor switching per se, regardless of whether it is associated with

a qualified opinion, may not be contrary

to investors' interests

We believe that future research should explore the roles played by management and external investors in the auditor selection decision When such evidence

is combined with findings on the extent

of auditor switching, we will have a better basis to determine the need for regulatory intervention in this area of accounting

REFERENCES AICPA, Report of the Ad Hoc Committee on Auditor Displacement, summarized in C Carpenter and

R Strawser, "Displacement of Auditors When Clients Go Public," Journal of Accountancy (June 1971),

pp 55-58

Burton, J C., and W Roberts, "A Study of Auditor Changes," Journal of Accountancy (April 1967),

pp 31-36

Carpenter, C G., and R H Strawser, "Displacement of Auditors When Clients Go Public," Journal of Accountancy (June 1971), pp 55-58

Chow, C., and S Rice, "Qualified Audit Opinions and the Measurement of Auditing Standard Uni- formity," Unpublished manuscript, University of Washington (May 1981)

Eisenbeis, R A., "Pitfalls in the Application of Discriminant Analysis in Business, Finance, and Eco- nomics," Journal of Finance (June 1977), pp 875-900

Farrar, D., and R Glauber, "Multicollinearity in Regression Analysis: The Problem Revisited," Review

of Economics and Statistics (February 1967), pp 92-107

Fried, D., and A Schiff, "CPA Switches and Associated Market Reactions," THE ACCOUNTING REVIEW (April 1981), pp 326-341

Goldman, A., and B Barlev, "The Auditor-Firm Conflict of Interests: Its Implications for Independence," THE ACCOUNTING REVIEW (October 1974), pp 707-718

Jensen, M., and W H Meckling, "Theory of the Firm: Managerial Behavior, Agency Costs and Owner- ship Structure," Journal of Financial Economics (October 1976), pp 305-360