Accepted Manuscript

Title: Audit firm tenure and independence: A comprehensive

investigation of audit qualifications in Spain

Author: Josep Garcia-Blandon Josep Ma Argiles

PII: S1061-9518(15)00003-8
DOI: http://dx.doi.org/doi:10.1016/j.intaccaudtax.2015.02.001
Reference: ACCAUD 194

To appear in: Journal of International Accounting, Auditing and Taxation

Received date: 31-10-2013

Revised date: 23-1-2015
Accepted date: 12-2-2015

Please cite this article as: Garcia-Blandon, J., and Argiles, J. M.,Audit firm
tenure and independence: A comprehensive investigation of audit qualifications
in Spain, Journal of the Chinese Institute of Chemical Engineers (2015),
http://dx.doi.org/10.1016/j.intaccaudtax.2015.02.001

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
*Title Page (including author details)

AUDIT FIRM TENURE AND INDEPENDENCE: A COMPREHENSIVE INVESTIGATION OF AUDIT

QUALIFICATIONS IN SPAIN

Authors:

Josep Garcia-Blandon

IQS School of Management. Universitat Ramon Llull

t
ip
cr
Josep M Argiles

Universitat de Barcelona (Department of Accounting)

us
Address for correspondence:
an
Josep Garcia-Blandon. IQS School of Management, Universitat Ramon Llull.
Via Augusta, 390, 08017 Barcelona. Spain.
Email: josep.garcia@iqs.edu
M
ed
pt
ce
Ac

Page 1 of 39
*BLINDED Manuscript (no author details)

1
2
3
4
5 Audit Firm Tenure and Independence:
6
7 A Comprehensive Investigation of Audit Qualifications in Spain
8
9
10
11

t
12

ip
13 Josep Garcia-Blandon
14
IQS School of Management
15

cr
16 Universitat Ramon Llull
17 Via Augusta, 390, 08017 Barcelona. Spain.
18 josep.garcia@iqs.edu

us
19
20
21
22
23 Josep M Argiles
24
25
26
an
Universitat de Barcelona
Department of Accounting
27
M
28
29
30
31
32
ed

33
34
35
36
pt

37
38
39
ce

40
41
42
43
44
Ac

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 2 of 39
65
 Audit Firm Tenure and Independence:
1
2 A Comprehensive Investigation of Audit Qualifications in Spain
3
4 ABSTRACT
5
6
7
The European Commissions Green Paper on Audit Policy has raised serious concerns
8
9 about external auditor independence and also encouraged further research about the
10
11

t
implications of long audit firm tenures. We examine the effects of audit firm tenure on
12

ip
13
14 independence, while measuring independence using the audit report opinion with a
15

cr
16 sample of Spanish public companies for the period 2002-2009. Generally, prior research
17
18
has limited the examination of modified going-concern opinions to financially

us
19
20
21 distressed firms. This approach limits the generalization of results as well as the ability
22
23
24
25
26
an
to address the traditional auditor role in the classical corporate governance scheme. To

address these issues, we propose a multinomial approach which not only includes all
27
M
28
29
types of audit qualifications in the analysis, but also considers the implications of going-
30
31 concern opinions for the auditor-client relationship. Our results show that auditors seem
32
ed

33 willing to sacrifice independence in lengthy engagements, but only for non-going-

34
35
36 concern modified opinions. This finding is robust to various proxies of accounting
pt

37
38 quality.
39
ce

40
41
42
43 Keywords: Auditor independence; audit firm tenure; accounting quality; litigation risk;
44
Ac

45
46
multinomial logistic model
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 2
63
64 Page 3 of 39
65
 1. Introduction
1
2 Auditing plays an important role in the classical approach to corporate
3
4
5 governance, as it provides an external and supposedly, independent certification of the
6
7 accounting information produced by managers. However, auditors face a conflict of
8
9
10 interest that might undermine credibility and ultimately destroy investors confidence.
11

t
12 Over five decades ago, Mautz and Sharaf (1961) pointed out that extended auditor-

ip
13
14
client relationships could reduce auditors objectivity thus impairing independence.
15

cr
16
17 Since then, lengthy auditor-client relationships have been considered a major issue in
18

us
19 the auditors conflict of interest. Shockley (1982) and Myers et al. (2003) have
20
21
22 suggested that long-term relationships might cause auditor complacency and possible
23
24
25
26
an
complicity in decisions that management makes regarding financial statement

27 presentation.
M
28
29 The main motivation of this study relies on the current regulatory discussion of a
30
31
32
mandatory audit firm rotation rule. Empirical research on this issue is not only
ed

33
34 inconclusive but mainly refers to the U.S. Since litigation risk plays a major role as a
35
36 determinant of the auditor reporting decision (e.g., Melumad and Thoman, 1990;
pt

37
38
39 Narayanan, 1994), the relatively high-litigation risk of the U.S. would make the
ce

40
41 generalization of results difficult. In addition, those studies using the opinion of the
42
43
44 audit report as the measure of independence generally have been limited to the analysis
Ac

45
46 of the issuance of going-concern modified opinions (GCMOs) to financially distressed
47
48
49 firms. This approach might be subject to, at least two shortcomings: a lack of
50
51 generalization problem and the inability to account for the main role of the auditors as
52
53
information verifiers in the classical corporate governance scheme (Simunic, 1984).
54
55
56 Finally, the scarce research addressing all types of firms and audit qualifications has
57
58
59
60
61
62 3
63
64 Page 4 of 39
65
 failed to account for the potentially different implications of GCMOs and non-going-
1
2 concern modified opinions (NGCMOs).
3
4
5 This article contributes to the literature on auditor independence by providing a
6
7 new framework which might overcome these shortcomings. The multinomial logistic
8
9
10 approach we propose allows for the inclusion of all types of qualifications into the
11

t
12 analysis, while at the same time acknowledging the potentially different implications of

ip
13
14
varying types of audit qualifications. Unlike previous research examining the tenure-
15

cr
16
17 audit qualifications relationship, our model includes accounting quality as a determinant
18

us
19 of audit qualifications. We consider this an important issue, because without including
20
21
22 accounting quality in the analysis, any reported negative effect of tenure on audit
23
24
25
26
an
qualifications (as for example in Vanstraelen, 2000) could mean either loss of

27 independence with tenure or higher accounting quality in longer tenures (and thus,
M
28
29 lower likelihood of audit qualifications). Evidently, the regulatory implications of each
30
31
32
situation would be opposite.
ed

33
34 The unique institutional framework of this research also enhances the relevance
35
36 of the reported results. The Spanish audit market is characterized by unusually lengthy
pt

37
38
39 audit firm tenures, and thus the potentially negative implications of long tenures on
ce

40
41 independence can be better addressed. Also, for the first time the analysis of the effects
42
43
44 of audit firm tenure is performed under mandatory partner rotation. This is useful to
Ac

45
46 regulators and policy makers, because if a firm rotation rule were established, the
47
48
49 mandatory rotation of the lead audit partner would likely be maintained. Finally, during
50
51 our research period some audit firms had multiyear contracts (between three and nine
52
53
years for the first contract with the client) while others were under annual
54
55
56 reappointments (any reappointment by the client after the first contract has to be done
57
58
59
60
61
62 4
63
64 Page 5 of 39
65
 annually). Therefore, we can also address the effects of the auditor contract regime on
1
2 the potential loss of independence with tenure.
3
4
5 Our results show that long tenures increase the likelihood of NGCMOs without
6
7 affecting the issuance GCMOs. Moreover, the lower likelihood of NGCMOs in lengthy
8
9
10 audit engagements cannot be explained by the supposedly higher accounting quality
11

t
12 achieved in these engagements. This finding highlights the limitations of the general

ip
13
14
approach used to investigate the tenure-independence relationship when only distressed
15

cr
16
17 firm GCMOs are examined.
18

us
19 The remainder of the paper is organized as follows. In section two, we review
20
21
22 the literature on the association between tenure and audit quality, with a focus on those
23
24
25
26
an
papers addressing the tenure-audit qualifications relationship. Section three summarizes

27 the regulation of the auditor-client relationship in the Spanish market. In section four we
M
28
29 define our model and describe our dataset. Results are discussed in section five. Finally,
30
31
32
in the last section the conclusions and implications of our findings are drawn.
ed

33
34 2. Literature Review
35
36 The audit report opinion constitutes the usual proxy for auditor independence
pt

37
38
39 and therefore is an indicator of audit quality. Since the probability of switching the audit
ce

40
41 firm increases after a qualified report (e.g., Krishnan, 1994; Lennox, 2000), qualified
42
43
44 opinions denote an exercise of independence by the auditor. Following concerns about
Ac

45
46 the negative effects of long tenures on independence, we would expect a lower
47
48
49 likelihood of qualified reports in lengthy auditor-client relationships. However,
50
51 empirical research has mostly refuted that independence could be threatened by long
52
53
tenures. Neither Louwers (1998) nor Carcello and Neal (2000) found negative effects of
54
55
56 tenure on the likelihood of GCMOs to financially distressed firms in the U.S. A similar
57
58 conclusion was reached by Vanstraelen (2002) and Knechel and Vanstraelen (2007) for
59
60
61
62 5
63
64 Page 6 of 39
65
 the Belgian market, and by Ruiz-Barbadillo et al. (2004) and (2006) for Spain.
1
2 However, more recent research has provided somewhat contradictory results. While
3
4
5 Lim and Tan (2010) reported a positive effect of tenure on audit quality in the U.S., Gul
6
7 et al. (2011) in the U.S. and Firth et al. (2012) in China concluded that auditors were
8
9
10 willing to forgo independence by issuing fewer GCMOs in longer tenures. Without
11

t
12 limiting the analysis to financially distressed firms or GCMOs, Vanstraelen (2000)

ip
13
14
found that long-term auditor-client relationships significantly increased the likelihood of
15

cr
16
17 unqualified audit reports in the Belgian market.
18

us
19 With the exception of Vanstraelen (2000), the effects of tenure on the likelihood
20
21
22 of qualified reports have been studied in the context of GCMOs and financially
23
24
25
26
an
distressed firms. This approach has serious shortcomings regarding the generalization of

27 the results either to the whole population of firms or to the world of audit qualifications.
M
28
29 Financially distressed firms are not representative of the whole population of audited
30
31
32
firms and the same holds for GCMOs regarding the whole world of audit qualifications.
ed

33
34 On the other hand, results reported by Vanstraelen (2000), although more generalizable
35
36 as the author includes all types of audit qualifications and firms in the analysis, do not
pt

37
38
39 differentiate between GCMOs and NCCMOs. In our view, this distinction is meaningful
ce

40
41 as both categories represent different dimensions of the audit activity and also, because
42
43
44 GCMOs involve particularly serious implications for the audited firm. Regarding the
Ac

45
46 first issue, Simunic (1984) posed that the auditors traditional role in the corporate
47
48
49 governance scheme is to verify the information produced by managers. However, when
50
51 auditors issue a qualified report for reasons of going concern, they do not perform such
52
53
role, but instead act as substitutes of bankruptcy prediction models. Regarding the
54
55
56 particularly serious implications of GCMOs, the decision of issuing a qualified opinion
57
58 is influenced by the perceived consequences in the economic trade-off between the
59
60
61
62 6
63
64 Page 7 of 39
65
 expected cost of the potential loss of a client, on the one hand, and the probability of
1
2 being exposed to third-party lawsuits and loss of reputation on the other. The risk of
3
4
5 litigation faced by the audit firm will be particularly high when it fails to issue GCMOs
6
7 to clients that subsequently go bankrupt (e.g., Palmrose 1987; Carcello and Palmrose,
8
9
10 1994; Krishnan and Krishnan, 1997; Francis, 2004). Therefore, Vanstraelens (2000)
11

t
12 approach fails to account for the potentially different implications of GCMOs and

ip
13
14
NGCMOs. If auditors are willing to forego independence regarding NGCMOs but not
15

cr
16
17 regarding comparatively riskier GCMOs, the inclusion of the two types of audit
18

us
19 qualifications into a single variable could cause misleading results.
20
21
22 There also is abundant evidence on the effects of tenure on other measures of
23
24
25
26
an
audit quality: restatements (e.g., Myers et al., 2005; Stanley and DeZoort, 2007), audit

27 failures (Geiger and Raghunandan, 2002), earnings quality (e.g., Myers et al., 2003;
M
28
29 Johnson et al., 2002; Davis et al., 2009), and investors perceptions of earnings quality
30
31
32
(Ghosh and Moon, 2005). These studies generally support a positive or neutral effect of
ed

33
34 tenure on audit quality.
35
36 3. Audit Regulation in Spain
pt

37
38
39 The Ley de Auditora de Cuentas (Audit Law) was enforced in 1988 as a
ce

40
41 consequence of the implementation of the 8th Directive on Company Law. The Law
42
43
44 established a three-to-nine-year contract with the audit firm. In order to strengthen
Ac

45
46 independence, a mandatory rotation of the audit firm at the end of the initial contract
47
48
49 was established. Nevertheless, as a result of a legal reform in 1995, the mandatory
50
51 rotation of the audit firm was abolished. After this reform, auditors could be engaged for
52
53
the same three-to-nine-year initial period but, after the expiration of the contract, any
54
55
56 renewal had to be done on an annual basis.
57
58
59
60
61
62 7
63
64 Page 8 of 39
65
 Similar to SOX in the U.S., in 2002 the Ley de Medidas para la Reforma del
1
2 Sistema Financiero (Measures for Reforming the Financial System Act) commonly
3
4
5 known as Ley Financiera (Financial Law) was passed as a reaction to the corporate
6
7 financial scandals of the 1990s. During the Laws approval process an amendment
8
9
10 imposing mandatory rotation of the audit firm after twelve years was included. As with
11

t
12 the 1995 reform revoking mandatory rotation, strong criticism from the auditing

ip
13
14
profession caused its eventual withdrawal. Mandatory rotation was finally limited to a
15

cr
16
17 seven-year period for the audit team without affecting the audit firm. The maximum
18

us
19 seven years of tenure for the lead partner has been maintained by the 2010 reform of the
20
21
22 Audit Law without imposing rotation of the whole audit team. Thus, our research period
23
24
25
26
an
(2002-2009) would be characterized by voluntary audit firm rotation, mandatory partner

27 rotation and tacit year by year reappointments of the audit firm.

M
28
29 According to the Spanish law, the audit report has to include the opinion of the
30
31
32
auditor about the firms accounting reports. This opinion can be unqualified, qualified,
ed

33
34 unfavorable or disclaimer of opinion. Nevertheless, audit reports with unfavorable or
35
36 disclaimer of opinion are very rare in Spain, at least for quoted companies. The
pt

37
38
39 supervisor of the Spanish stock market, Comisin Nacional del Mercado de Valores
ce

40
41 (CNMV), classifies audit qualifications into two major groups: quantified and
42
43
44 unquantified. Some examples of quantified qualifications would be inadequate
Ac

45
46 depreciation, understatement of bad debt expenses, non-recognition of severance
47
48
49 payments, overstatement of the long-term investments, and inaccurate estimation of the
50
51 taxes payable. These quantified qualifications are subsequently classified into two
52
53
subgroups, depending on whether they affect the profit and losses statement or equity.
54
55
56 Similarly, unquantified qualifications are also classified into uncertainty and others
57
58 and limitations (CNMV, 2009). Among qualifications due to uncertainties, the most
59
60
61
62 8
63
64 Page 9 of 39
65
 serious would be those concerning the continuity of business, the so-called GCMOs.
1
2 However, uncertainties also can have much less dramatic effects; for example, they
3
4
5 might be associated with the firms ability to recover some tax credits. On the other
6
7 hand, qualifications for limitations of scope would show that the auditor has not had
8
9
10 enough information to apply the procedures required by technical auditing standards.
11

t
12 Therefore, while GCMOs are issued when the auditor has doubts about the future of the

ip
13
14
firm, NGCMOs are issued when the auditor believes that the financial statements do not
15

cr
16
17 adequately represent the current situation of the firm.
18

us
19 4. Research Design
20
21
22 4.1. Model and Hypotheses Development
23
24
25
26
an
Prior studies generally have measured independence through the issuance of

27 GCMOs to financially distressed firms. The election of GCMOs as a proxy for

M
28
29 independence has been justified on the grounds that available evidence (e.g., Chow and
30
31
32
Rice, 1982 and Krishnan, 1994 for the U.S.; Craswell, 1988 for Australia) shows higher
ed

33
34 probability of audit firm switches after the issuance of a qualified report. Similarly, Ball
35
36 et al. (1979) argued that auditors are reluctant to qualify their reports, since it would
pt

37
38
39 adversely affect the interests of corporate participants. In the same vein, Levinthal and
ce

40
41 Fichman (1988) claimed that a qualified opinion is an indicator of conflict in the
42
43
44 auditor-client relationship. Besides, audit qualifications have negative effects for the
Ac

45
46 client in terms of negative stock price reactions (Chow and Rice, 1982b; Dopuch et al.,
47
48
49 1987; Loudder et al., 1992; and Ameen et al., 1994), lower market responses to earnings
50
51 announcements (Choi and Jeter, 1992) and higher debt costs (Chen et al., 2012).
52
53
However, these papers do not limit the analysis to the implications of GCMOs, but
54
55
56 include all types of audit qualifications. Therefore, empirical evidence supports the use
57
58 of audit qualifications (not only GCMOs) as a proxy for independence. This view also
59
60
61
62 9
63
64 Page 10 of 39
65
 was implicitly supported by Vanstraelen (2000) and Firth et al. (2012), as they studied
1
2 all types of audit qualifications, although without differentiating between GCMOs and
3
4
5 NGCMOs. Given the particularly serious consequences of GCMOs for the auditors
6
7 client, we suggest that GCMOs and NGCMOs should not be included as a single
8
9
10 category of a bivariate dependent variable. Supporting this view, Krishnan (1994) found
11

t
12 that while auditor switching significantly increases after GCMOs and NGCMOs, the

ip
13
14
increase is almost double for GCMOs than for NGCMOs.
15

cr
16
17 Accordingly, our research question is: Does auditor independence, measured by
18

us
19 the propensity to issue qualified reports, decrease with tenure? Based on results reported
20
21
22 by Vanstraelen (2000) for a low litigation risk country, and considering the relatively
23
24
25
26
an
low litigation risk profile of the Spanish audit market, we expect a lower likelihood of

27 NGCMOs in longer tenures. Therefore, we hypothesize that:

M
28
29 H1: Ceteris paribus, longer auditor-client relationships will make audit reports with
30 NGCMOs less likely.
31
32
ed

33 However, if H1 is not rejected, it could mean either that independence is

34
35 impaired in lengthy auditor-client relationships or, alternatively, that the lower
36
pt

37
38 likelihood of audit qualifications in longer tenures would be the consequence of higher
39
accounting quality achieved in lengthy engagements. Clearly, both situations would
ce

40
41
42
43
involve opposite implications for regulators and policy makers.
44
Ac

45 Since prior research has generally supported a positive or neutral effect of tenure
46
47 on the likelihood of GCMOs for the Belgian (Vanstraelen, 2002 and Knechel and
48
49
50 Vanstraelen, 2007) and Spanish (Ruiz-Barbadillo et al., 2004 and 2006) low litigation
51
52 risk audit markets, we hypothesize that:
53
54
55 H2: Ceteris paribus, the likelihood of audit reports with GCMOs does not depend on the
56 tenure with the audit firm.
57
58
59
60
61
62 10
63
64 Page 11 of 39
65
 Previous studies in the tenure-audit report literature have used the classical
1
2 logistic regression approach, with a dependent binary variable coded zero in case of an
3
4
5 unqualified audit report and one in case of an unclean report. Consequently, they did
6
7 not explicitly address the potentially different implications of GCMOs and NGCMOs.
8
9
10 We propose to overcome this limitation by using a multinomial logistic model with a
11

t
12 dependent variable coded zero in case of an unqualified report; coded one in case of

ip
13
14
NGCMOs; and coded two in case of GCMOs. When the audit report contains both types
15

cr
16
17 of audit qualifications, due to the particularly serious implications of GCMOs, it also
18

us
19 has been coded two. Following prior research, the experimental variable in our model is
20
21
22 the audit firm tenure, while control variables are similar to those first used by DeFond et
23
24
25
26
an
al. (2002) and later by Carey and Simnett (2006) to estimate the issuance of GCMOs,

27 and basically comprise indicators of financial health. In addition, we also include an

M
28
29 independent variable accounting for periods of economic downturns since the likelihood
30
31
32
of qualified opinions should be higher during these periods.
ed

33
34 A major issue in this research is that we use the financial ratios traditionally
35
36 proposed to explain GCMOs, as the determinants of NGCMOs. We base our decision
pt

37
38
39 on three points. Firstly, the rationale for including financial rations to explain GCMOs
ce

40
41 relies on the fact that auditors litigation risk increases when clients show higher levels
42
43
44 of financial distress. However, Krishnan and Krishnan (1997) studied the determinants
Ac

45
46 of lawsuits against auditors, among them, client financial distress and the issuance of
47
48
49 modified audit opinions, and argued that auditors can offset litigation risk by increasing
50
51 the issuance of modified opinions. Thus, litigation risk also increases when the auditor
52
53
fails to issue NGCMOs to a company that subsequently goes bankrupt and whose
54
55
56 financial statements were potentially misleading to users. Therefore, financial ratios
57
58 should explain not only GCMOs but all types of audit qualifications. Secondly, as firms
59
60
61
62 11
63
64 Page 12 of 39
65
 with solvency problems also are expected to be more willing to manipulate financial
1
2 statements (Butler et al., 2004), situations of financial distress should make NGCMOs
3
4
5 more likely. Finally, many articles (e.g., Gul et al., 1992; Laitinen and Laitinen, 1998;
6
7 Vanstraelen,2000; Hudaib and Cooke, 2005; Meyer et al., 2007; and Firth et al., 2012)
8
9
10 have studied the issuance of audit qualifications, in most cases without differentiating
11

t
12 between GCMOs and NGCMOs. These studies have used financial ratios to explain the

ip
13
14
issuance of qualified reports. As audit reports with NGCMOs generally represent about
15

cr
16
17 90 percent of the total number of qualified reports, these investigations are in fact using
18

us
19 financial ratios to explain NCMOs.
20
21
22 We propose the following model to perform the multinomial analysis.
23
24
25
26
an
OPINION = f (PBANK, SIZE, AGE, LEVERAGE, CHLEVERAGE LIQUIDITY,
STOCKS, LOSSES, AUDFIRM, CRISIS, DISACC, TENURE) (1)
27
M
28 where:
29
30 OPINION = the audit report opinion: unqualified (0), NGCMO (1), and GCMO (2).
31
32
ed

33 TENURE = number of consecutive years the client has been audited by the same firm.
34
35 For the control variables:
36
pt

37
38 PBANK = the probability of bankruptcy as measured by the adjusted Zmijewski (1984)
39 score, with the weights proposed by Carcello et al. (1995).
ce

40
41 SIZE = the natural log of the firms total assets in book values at the end of the year.
42
43
44 AGE = the natural log of the number of years a firm has been listed in the Spanish stock
Ac

45 market.
46
47 LEVERAGE = the firms level of financial leverage calculated as total debt divided by
48
49 total assets, both in book values, at the end of the year.
50
51 CHLEVERAGE = change in LEVEVERAGE during the year.
52
53
LIQUIDITY =the sum of the firms cash positions divided between current liabilities.
54
55
56 STOCKS = the firms inventories divided by total assets, both in book values.
57
58 LOSSES = 1 if the firms net profit in year t-1 is negative, 0 otherwise.
59
60
61
62 12
63
64 Page 13 of 39
65
 AUDFIRM = 1 if the company is audited by a Big 4 audit firm, 0 otherwise.
1
2 CRISIS = 1 for years 2002, 2008 and 2009, and 0 otherwise. The year 2002 corresponds
3
4
to the dotcom crisis and 2008 and 2009 to the beginning of the financial crisis.
5
6 DISACC = discretionary accruals in absolute values, as defined by DeFond and Park
7 (2001), and later used by Carey and Simnett (2006).
8
9
10 PBANK measures the probability of bankruptcy, where higher values indicate a
11

t
12 greater probability of bankruptcy, and a higher litigation risk for the auditor. Thus, we

ip
13
14
expect a positive effect of PBANK on the likelihood of GCMOs and NGCMOs.
15

cr
16
17 The positive relationship between client size and litigation costs (e.g., Lys and
18

us
19 Watts, 1994; Shu, 2000) would make qualified reports more likely for large firms.
20
21
22 However, according to DeAngelo (1981) auditors incentives to compromise
23
24
25
26
an
independence would depend on the importance of the client. Also, large clients would

27 have more negotiating power to avoid audit qualifications. As the risk of litigation is
M
28
29 relatively low in the Spanish audit market, we predict a negative effect of SIZE on the
30
31
32
probability of a qualified report.
ed

33
34 AGE captures the fact that financial distress is more likely for younger
35
36 companies. Therefore, we predict a negative effect of AGE on both types of audit
pt

37
38
39 qualifications.
ce

40
41 Financial leverage may make bankruptcy more likely and possibly raise
42
43
44 auditors litigation risk. Accordingly, we expect a positive effect of LEVERAGE and
Ac

45
46 CHLEVERAGE on both types of audit qualifications.
47
48
49 A lack of liquidity is usually considered a significant determinant of bankruptcy
50
51 prediction (e.g., Hopwood et al., 1989). Therefore, poor liquidity would make qualified
52
53
reports more likely as it increases litigation risk. In addition, firms with liquidity
54
55
56 problems might be more willing to manipulate financial statements (e.g., Butler et al.,
57
58
59
60
61
62 13
63
64 Page 14 of 39
65
 2004), thus making NGCMOs more likely. Therefore, we predict a negative effect of
1
2 LIQUIDITY on GCMOs and NGCMOs.
3
4
5 Auditing inventories can present serious difficulties, because it involves two
6
7 audit assertions: valuation and completeness (McDaniel, 1990). This would explain the
8
9
10 higher audit fees for those firms with relatively large amounts of inventories (Simunic,
11

t
12 1980). In addition, audit errors (Firth, 2002) and lawsuits against auditors (St. Pierre and

ip
13
14
Anderson, 1984) also are often caused by inventories. Accordingly, we hypothesize a
15

cr
16
17 positive effect of STOCKS on the probability of NGCMOs.
18

us
19 Following previous research (e.g., Dopuch et al., 1987; Firth, 2002) firms with
20
21
22 losses would face higher probabilities of audit qualifications. The explanation would be
23
24
25
26
an
similar to the one proposed to justify the expected negative relationship between

27 liquidity and audit qualifications. As litigation risk is higher when auditing firms with
M
28
29 losses, so should be the probability of audit qualifications. In addition, firms
30
31
32
experiencing losses would more likely engage in earnings management activities,
ed

33
34 therefore making NGCMOs more likely. Thus, we predict a positive effect of LOSSES
35
36 on GCMOs and NGCMOs.
pt

37
38
39 We include AUDFIRM to account for a potentially higher propensity by big 4
ce

40
41 auditors to issue qualified reports. Accordingly, we expect a positive effect of
42
43
44 AUDFIRM on the likelihood of GCMOs and NGCMOs.
Ac

45
46 Since the risk of bankruptcy can be particularly high during economic downturns, we
47
48
49 predict a positive effect of CRISIS on GCMOs and NGCMOs. However, as GCMOs are
50
51 directly related to the survival of the company, we expect a stronger effect of CRISIS on
52
53
the model with GCMOs compared to the model with NGCMOs.
54
55
56 Finally, audit qualifications, particularly NGCMOs, should be more likely for
57
58 those companies showing higher levels of discretionary accruals, i.e. lower accounting
59
60
61
62 14
63
64 Page 15 of 39
65
 quality. Thus, we predict a positive coefficient for DISACC on both types of audit
1
2 qualifications.
3
4
5 4.2. Sample and Data
6
7 This study analyzes companies quoted in the Spanish Stock Exchange (Sistema
8
9
10 de Interconexin Burstil Espaol) during the research period 2002-2009. We choose
11

t
12 2002 as the first year of this study because it was then that the Financial Law was

ip
13
14
enacted which imposed a maximum partner tenure of seven years. Thus, our research
15

cr
16
17 period is characterized by homogeneous regulation of firm and partner rotation. While
18

us
19 the length of the initial audit contract can range between three and nine years, those
20
21
22 firms renewing a contract with an audit firm are subject to year by year tacit
23
24
25
26
an
reappointment. Although we cannot know the exact number of firms under multiyear

27 contract and year by year reappointment, all firms with nine years or more of tenure will
M
28
29 necessarily be subject to annual reappointments. However, it should be noted that not all
30
31
32
firms with less than nine years will be under a multiyear contract, because the year by
ed

33
34 year reappointment is mandatory after the end of the first audit contract independent of
35
36 its length. For example, a firm with six years of tenure will be under a multiyear
pt

37
38
39 contract if it has signed a first nine-year contract with its current auditor, while it will be
ce

40
41 under year by year reappointment if it signed a five-year initial contract with its current
42
43
44 auditor.
Ac

45
46 Information about audit firm tenure was obtained from corporate governance
47
48
49 reports, and the audit report opinion was provided by financial statements registered in
50
51 the CNMV. Thomson Reuters Knowledge provided the independent variables data.
52
53
Since our model includes liquidity and debt ratios as control variables, we removed
54
55
56 banks and financial companies from the sample as is usual in the literature. Our sample
57
58 was initially comprised of 112 firms. However, consistent with prior research, we only
59
60
61
62 15
63
64 Page 16 of 39
65
 analyzed firms quoted in the Spanish Stock Exchange during the entire research period.
1
2 This yielded 83 firms with 664 firm-year observations over the eight-year research
3
4
5 period. Twelve cases were removed for missing variable information resulting in a final
6
7 sample of 652 firm-year observations. Of these, 544 cases have unqualified reports, 87
8
9
10 cases show NGCMOs and the remaining 21 cases have GCMOs.
11

t
12 Table 1 shows the 652 audit reports classified by year and type of opinion. The

ip
13
14
percentage of unqualified reports steadily increased during 2002-2006, but decreased
15

cr
16
17 afterwards as a result of the international financial crisis. The maximum value
18

us
19 corresponds to 2006 and 2007 when 90 percent of reports were unqualified. Since
20
21
22 GCMOs were very infrequent before the start of the financial crisis, the rise of
23
24
25
26
an
unqualified reports in the 2002-2006 subperiod was caused by a fall in the number of

27 reports with NGCMOs. However, during the 2006-2009 subperiod, audit reports with
M
28
29 GCMOs showed a dramatic increase.
30
31
32
INSERT TABLE 1 ABOUT HERE
ed

33
34 Table 2 provides descriptive statistics about model independent variables. The
35
36 average company SIZE is 6.73, corresponding to total assets of about five billion euros.
pt

37
38
39 The statistics for LEVERAGE would indicate that, on average, the amount of debt
ce

40
41 represents 63 percent of total assets. Firms cash positions represent 16 percent of
42
43
44 current liabilities, while inventories account for 15 percent of total assets. The mean
Ac

45
46 value of 0.12 for LOSSES indicates net losses for 12 percent of our sample. The table
47
48
49 also shows the extreme level of concentration of the Spanish audit market by Big 4
50
51 auditors, as they issue 92 percent of the audit reports in the sample. Our experimental
52
53
variable for auditor-client relationships, TENURE, shows 10 years of average duration,
54
55
56 with a maximum of 23 years. As noted previously, the particularly long audit firm
57
58 tenures in Spain constitute an important advantage for our study. The distribution of
59
60
61
62 16
63
64 Page 17 of 39
65
 TENURE is far from normal, as it shows a relatively high concentration of firms in the
1
2 two extremes of the variable.
3
4
5
6
7 INSERT TABLE 2 ABOUT HERE
8
9
10 A correlation matrix (not reported) shows relatively low levels of correlation
11

t
12 between the independent variables, with a maximum value of 0.54 between PBANK and

ip
13
14
LEVERAGE. Thus, we do not expect serious multicollinearity in the data.
15

cr
16
17 5. Results
18

us
19 5.1. Univariate Analysis
20
21
22 Before estimating the multinomial logistic model, we performed a univariate
23
24
25
26
an
analysis of differences of means across groups of firms according to the opinion of the

27 audit report: unqualified, NGCMOs, and GCMOs. Since the Shapiro-Wilk test rejects
M
28
29 the hypothesis of normality for each independent variable, we use the Mann-Whitney
30
31
32
test of differences of medians to analyze statistical significance. In table 2 we report
ed

33
34 mean and median values of independent variables across subsamples as well as
35
36 significance levels from the Mann-Whitney test for the continuous variables and the
pt

37
38
39 Pearson chi-square test for the dichotomous variables: LOSSES, AUDFIRM and CRISIS.
ce

40
41 Results from table 2 strongly fit our expectations. Focusing on TENURE, firms
42
43
44 with qualified reports (either GCMOs or NGCMOs) show shorter audit tenures
Ac

45
46 compared to firms with unqualified reports. These differences are statistically
47
48
49 significant at one percent (NGCMOs) and 10 percent (GCMOs) levels. Although firms
50
51 with GCMOs show a shorter median tenure than firms with NGCMOs, the statistical
52
53
significance for this group is achieved only at marginal levels. This apparent
54
55
56 contradiction can be explained by the small number of observations for GCMOs. We
57
58 also observe significantly higher discretionary accruals associated to NGCMOs and
59
60
61
62 17
63
64 Page 18 of 39
65
 GCMOs. However the latter shows only marginally significant differences, as was the
1
2 case with TENURE. The negative correlation between TENURE and DISACC, jointly
3
4
5 with the lower levels of discretionary accruals reported for firms with unqualified audit
6
7 reports, could explain the longer median tenures of firms with unqualified reports as a
8
9
10 consequence of higher accounting quality. Results also indicate that NGCMOs and
11

t
12 GCMOs are significantly associated to PBANK and LOSSES. In addition, NGCMOs are

ip
13
14
also significantly related to SIZE and LIQUIDITY, while GCMOs are also associated to
15

cr
16
17 AGE, LEVERAGE, CHLEVERAGE, AUDFIRM, and CRISIS. It should be noted that
18

us
19 except for AGE regarding GCMOs, in all cases differences in mean and median values
20
21
22 have the expected sign. Finally, STOCKS is the only variable showing no significant
23
24
25
26
differences for either NGCMOs or GCMOs. an
27 5.2. Multivariate Analysis
M
28
29 Using a multinomial logistic regression model, we address the joint effect of
30
31
32
tenure and the proposed control variables on the likelihood of audit qualifications. Due
ed

33
34 to the nature of some independent variables, we find significant heteroscedasticity in the
35
36 data. Thus, reported z-values are calculated with robust standard errors. Table 3 shows
pt

37
38
39 estimation results for NGCMOs and GCMOs, where firms with unqualified audit
ce

40
41 reports are the comparison group.
42
43
44 INSERT TABLE 3 ABOUT HERE
Ac

45
46 According to the likelihood ratio chi-square test, the null hypothesis that all
47
48
49 predictors regression coefficients in the model are simultaneously zero is rejected, and
50
51 the model also shows a McFadden's pseudo R2 of 0.25. A main issue in multinomial
52
53
logistic models is whether some categories of the dependent variable could be combined
54
55
56 into a single one. To address this issue, we perform a log-likelihood ratio test for
57
58 combining dependent variable categories. Our main interest is whether the following
59
60
61
62 18
63
64 Page 19 of 39
65
 categories of audit qualifications could be combined into one: unqualified versus
1
2 NGCMOs; unqualified versus GCMOs and NGCMOs versus GCMOs. Our test
3
4
5 results reveal that in all three possible combinations, the null hypothesis that the
6
7 specified categories are indistinguishable with respect to the variables in the model is
8
9
10 rejected (P-value <0.01). This result supports the choice of a multinomial logistic model
11

t
12 over the traditional bivariate approach. Although correlation coefficients did not suggest

ip
13
14
serious multicollinearity, we calculated variance inflation factors (VIF) to rule out the
15

cr
16
17 negative potential effects of multicollinearity on the results. As expected, VIF (not
18

us
19 reported) are rather low (the average value is 1.31 with a maximum of 1.71 for variable
20
21
22 LEVERAGE), thus supporting our initial view that multicollinearity would not affect our
23
24
25
26
results. an
27 Table 3 reports relative risk ratios (RRR) in addition to coefficients of the
M
28
29 estimation and z-values. The RRR of a coefficient indicates how the risk of falling in the
30
31
32
comparison group (GCMOs or NGCMOs) compared to the risk of falling in the referent
ed

33
34 group (unqualified reports) changes with each variable. We are primarily concerned
35
36 with the sign and statistical significance of the coefficients of TENURE. If, as H1 states,
pt

37
38
39 longer auditor-client relationships make NGCMOs less likely, the coefficient of
ce

40
41 TENURE should be negative and statistically significant. Table 3 shows for the model
42
43
44 with NGCMOs, that the coefficient of TENURE is negative and significant (P-value
Ac

45
46 <0.01). Therefore, H1 cannot be rejected. Moreover, the RRR indicates that one more
47
48
49 year of tenure would decrease the relative risk of NGCMOs regarding an unqualified
50
51 report by a factor of 0.94, given that the other variables in the model are held constant.
52
53
In other words, one more year of tenure would increase the likelihood of being in the
54
55
56 unqualified group by six percent. Similarly, H2, the likelihood that GCMOs do not
57
58 depend on tenure, cannot be rejected. Taken together, these results suggest that external
59
60
61
62 19
63
64 Page 20 of 39
65
 auditors could be willing to sacrifice independence in lengthy engagements, but only
1
2 regarding the issuance of NGCMOs. As noted previously, the results for GCMOs could
3
4
5 be explained by the particularly high litigation risk associated with this type of opinion.
6
7 It should be noted that our main result that NGCMOs are less likely in longer
8
9
10 tenures is not fully comparable to prior research, as this is the first attempt to address
11

t
12 the tenure-auditors opinion relationship by differentiating between GCMOs and

ip
13
14
NGCMOs. However, prior research with all types of qualifications (of which NGCMOs
15

cr
16
17 roughly represent about 90%) can provide some basis for comparing results. For
18

us
19 example, the negative effect of tenure on NGCMOs would support Vanstraelens (2000)
20
21
22 findings for the Belgian market. On the other hand, evidence reported about the lack of
23
24
25
26
an
a negative effect of tenure on the likelihood of GCMOs is consistent with most previous

27 research, particularly results for Spain (Ruiz-Barbadillo et al., 2004 and 2006), and
M
28
29 evidence for the Belgian low litigation risk audit market (Vanstraelen, 2002 and
30
31
32
Knechel and Vanstraelen, 2007).
ed

33
34 Control variable results indicate that whenever a significant effect is reported, it
35
36 has the predicted direction. PBANK shows significant positive effects on GCMOs and
pt

37
38
39 NGCMOs (P-value <0.01). As expected, the RRR suggests that the importance of
ce

40
41 clients solvency is higher for GCMOs than for NGCMOs. SIZE shows a negative effect
42
43
44 on both types of qualifications, though it is only significant for NGCMOs (P-value
Ac

45
46 <0.01). Thus, the likelihood of NGCMOs would be lower for larger firms, while the
47
48
49 issuance of GCMOs would not depend on the size of the client. As noted previously,
50
51 SIZE could have opposite effects on the likelihood of audit qualifications. While larger
52
53
clients involve higher litigation risk for the audit firm, these clients also have more
54
55
56 negotiating power to avoid audit qualifications. Our results suggest that auditors could
57
58 be willing to impair independence with large and, presumably, more rewarding clients,
59
60
61
62 20
63
64 Page 21 of 39
65
 although they would only do it regarding NGCMOs. However, they would not do it for
1
2 GCMOs, as they involve particularly high levels of litigation risk (e.g., Carcello and
3
4
5 Palmrose, 1994; Francis, 2004). However, an alternative explanation could be that the
6
7 likelihood of NGCMOs would be lower for larger companies because they have higher
8
9
10 accounting quality, as the negative and significant correlation between SIZE and
11

t
12 DISACC would suggest. Interestingly, accounting quality, proxied through DISACC,

ip
13
14
shows significant effects on GCMOs and NGCMOs (P-value <0.01). This indicates that
15

cr
16
17 the likelihood of qualified reports is lower under conditions of higher accounting
18

us
19 quality. Audit qualifications are less likely for those firms with stronger cash positions,
20
21
22 although results for GCMOs are only marginally significant.
23
24
25
26
an
The reporting of losses makes NGCMOs more likely without significantly

27 affecting the issuance of GCMOs. This result is rather surprising, as we had predicted a
M
28
29 stronger effect of LOSSES on GCMOs than that on NGCMOs. A possible explanation
30
31
32
could be that, given the relatively low number of audit reports with GCMOs in the
ed

33
34 sample, the effect of LOSSES had already been incorporated in the variable PBANK,
35
36 which shows a highly significant effect on GCMOs.
pt

37
38
39 Being audited by a Big 4 firm makes NGCMOs more likely but it does not affect
ce

40
41 the issuance of GCMOs. This result suggests that non-Big 4 auditors could be willing to
42
43
44 impair independence regarding the issuance of NGCMOs, but they would not do it for
Ac

45
46 GCMOs as the latter would involve particularly high levels of litigation risk.
47
48
49 GCMOs are more likely during economic downturns (P-value <0.05). However,
50
51 the effect of CRISIS on NGCMO, although positive, is non-significant. We do not report
52
53
significant effects of AGE, LEVERAGE, CHLEVERAGE and STOCKS on either
54
55
56 GCMOs or NGCMOs.
57
58 5.2.1. Sensitivity Analyses
59
60
61
62 21
63
64 Page 22 of 39
65
 We tested the robustness of our reported results to a different measure of audit
1
2 tenure. We defined the dichotomous variable LONGTENURE as one when tenure is 10
3
4
5 years or more and, zero otherwise. We chose 10 years as the cutoff point because the
6
7 median tenure in our sample is nine years. We then re-estimate equation (1) with
8
9
10 LONGTENURE instead of the original variable TENURE. Results (not reported) remain
11

t
12 largely unchanged. The new variable LONGTENURE shows a negative and significant

ip
13
14
coefficient in the NGCMOs model (P-value <0.01) while in the GCMOs model the
15

cr
16
17 coefficient is non-significant. Moreover, control variable coefficients show the same
18

us
19 signs and levels of significance as those reported in table 3.
20
21
22 There is some evidence (Levinthal and Fichman, 1988; Vanstraelen, 2000) that
23
24
25
26
an
auditors would be more willing to issue unqualified reports during the first two years of

27 engagement (during the so-called honeymoon period). To test this possible effect, we
M
28
29 re-estimated equation (1) after including the new variable NEWCLIENT, a dichotomous
30
31
32
variable with score of one for the first and second year of the audit contract, and zero
ed

33
34 otherwise. Under a honeymoon effect, the coefficient of NEWCLIENT would be
35
36 negative and statistically significant. Results (not reported) show a negative effect,
pt

37
38
39 though non-significant for either GCMOs or NGCMOs. Thus, results do not support a
ce

40
41 honeymoon effect in the Spanish audit market. Finally, following Davis et al. (2009),
42
43
44 we check the robustness of our findings to a potential non-monotonic effect of tenure on
Ac

45
46 audit quality. The authors argued that audit quality could increase in the early years of
47
48
49 engagement through a learning effect, but it would decrease in later years due to a
50
51 bonding effect. Accordingly, we estimate a quadratic model with TENURE and the new
52
53
variable TENURE2, defined as the square of TENURE. Under the non-monotonic effect
54
55
56 of tenure, the coefficient of TENURE should be positive and significant while the
57
58 coefficient of TENURE2 should be negative and significant. However, results (not
59
60
61
62 22
63
64 Page 23 of 39
65
 reported) do not support a non-monotonic effect of tenure on audit quality, since the
1
2 coefficient of TENURE2 is non-significant.
3
4
5 5.2.2. Effects of Accounting Quality and Auditor-Client Contract Regime
6
7 In this section we address two additional issues. First, could the lower likelihood
8
9
10 of NGCMOs in longer tenures previously reported be explained by the higher
11

t
12 accounting quality achieved in lengthy engagements? Also, what are the potential

ip
13
14
effects of the auditor-client contract regime, not only on auditor independence, but also
15

cr
16
17 on the loss of independence in longer tenures?
18

us
19 The first question is of major importance, particularly from a regulatory
20
21
22 viewpoint. In longer tenures auditors will more likely detect accounting misstatements
23
24
25
26
an
as they have deeper client knowledge, consequently, the reported negative effect of

27 tenure on NGCMOs could be the result of higher accounting quality achieved in longer
M
28
29 tenures. If this is the case, regulatory concerns regarding about long firm tenures for
30
31
32
audit quality would not be justified. Conversely, if the lower likelihood of audit
ed

33
34 qualifications in longer tenures is not explained by higher accounting quality, it would
35
36 suggest an impairment of independence in longer tenures. To address this issue we
pt

37
38
39 estimate the following model given by equation (2):
ce

40
41 OPINION = f (PBANK, SIZE, AGE, LEVERAGE, CHLEVERAGE, LIQUIDITY,
42
43
STOCKS, LOSSES, AUDFIRM, CRISIS, DISACC, TENURE,
44 DISACC*TENURE) (2)
Ac

45
46 Since this analysis refers to NGCMOs, we use a bivariate logistic model. While
47
48
49 similar to equation (1), the dependent variable OPINION is now defined as one, if the
50
51 audit report has NGCMOs, and zero otherwise. In addition, we include the tenure-
52
53
accounting quality interaction variable DISACC*TENURE. We estimate equation (2)
54
55
56 with robust standard errors, after removing sample observations for GCMOs. Results of
57
58 in table 4 show that the coefficient of TENURE remains negative and statistically
59
60
61
62 23
63
64 Page 24 of 39
65
 significant (P-value <0.01) as in table 3, while the new interaction variable
1
2 DISACC*TENURE shows no significant effects on the issuance of NGCMOs.
3
4
5 Accordingly, we conclude that the lower likelihood of NGCMOs in longer tenures
6
7 reported in table 3 cannot be explained by higher accounting quality. Instead, our
8
9
10 findings support loss of independence in long audit engagements. As expected, results
11

t
12 for control variables are very similar to those reported in table 3.

ip
13
14
15
INSERT TABLE 4 ABOUT HERE

cr
16
17 To check the robustness of our results to other indicators of accounting quality,
18

us
19 we proxied accounting quality using discretionary accruals as defined in Carey and
20
21
22 Simnett (2006). Thus, we performed sequential estimations of equation (2) after
23
24
25
26
an
substituting DISACC with each of the following: CURACC (total current accruals);

27 SMPROF (the reporting of small profits); and SMIPROF (the reporting of a small
M
28
29 increase in profits). CURACC is the change in current assets (excluding cash) minus
30
31
32
change in current liabilities (excluding change in short-term notes and current portion of
ed

33
34 long-term debt. SMPROF is a dichotomous variable coded one when profit is less than
35
36 two percent of total assets, and zero otherwise. SMIPROF is a dichotomous variable
pt

37
38
39 coded one when the increase in profit (decrease in loss) over last years profit/loss is
ce

40
41 less than two percent of total assets, and zero otherwise. We also included the
42
43
44 subsequent restatement of financial statements (RESTAT) as an indicator of poor
Ac

45
46 accounting quality. In each of these four new estimations, the tenure-accounting quality
47
48
49 interaction variable is defined according to the proxy of accounting quality used in the
50
51 analysis. For simplicity, table 5 only reports results for variables related to either tenure
52
53
or accounting quality. In each of the four estimations TENURE shows a negative effect
54
55
56 on the likelihood of NGCMOs, while the coefficient of the interaction variable is always
57
58 non-significant. Thus, we conclude that our prior finding about NGCMOs tenure and
59
60
61
62 24
63
64 Page 25 of 39
65
 accounting quality is robust to the most usual measures of accounting quality. Finally, it
1
2 also should be noted that only two of the four measures of accounting quality, CURACC
3
4
5 and SMIPROF, show significant effects (P-value <0.05 and <0.1, respectively) on the
6
7 likelihood of NGCMO (in both cases with the predicted positive sign). This suggests a
8
9
10 rather weak relationship between the issuance of NGCMOs and accounting quality.
11

t
12 INSERT TABLE 5 ABOUT HERE

ip
13
14
Finally, we address the potential effects of the auditor contract regime on our
15

cr
16
17 results. As noted previously, some firms in our sample are subject to multiyear contracts
18

us
19 while others are under year by year tacit reappointments. While all firms with nine or
20
21
22 more years of tenure will necessarily be under the year by year reappointment regime,
23
24
25
26
an
not every firm with less than nine years of tenure will necessarily be under a multiyear

27 contract. However, despite the serious limitation of our dataset to address this issue, we
M
28
29 have performed a preliminary analysis about the implications of the auditor contract
30
31
32
regime on independence. Accordingly, we estimate the following model given by
ed

33
34 equation (3):
35
36 OPINION = f (PBANK, SIZE, AGE, LEVERAGE, CHLEVERAGE LIQUIDITY,
pt

37
38 STOCKS, LOSSES, AUDFIRM, CRISIS, DISACC, TENURE, YBYREAP,
39 TENURE*YBYREAP) (3)
ce

40
41
42
43
This model is similar to equation (2) but we removed the interaction term
44
Ac

45 DISACC*TENURE and included YBYREAP (one if tenure is nine or more years, and
46
47 zero otherwise), as well as the interaction term TENURE*YBYREAP. If auditor
48
49
50 independence is lower under year by year reappointments than under multiyear
51
52 contracts, the coefficient of YBYREAP should be negative and significant. Moreover, the
53
54
55 interaction variable TENURE*YBYREAP allows us to capture differential effects of
56
57 tenure on independence depending on the auditor-client contract regime. Accordingly, if
58
59
60
the loss of independence with tenure reported in tables 3 and 4 is more severe for those
61
62 25
63
64 Page 26 of 39
65
 auditors under year by year reappointments, the coefficient of the interaction term
1
2 should be negative and significant. Equation (3) results (not reported) show that neither
3
4
5 YBYREAP nor the interaction term TENURE*YBYREAP significantly affect the
6
7 likelihood of NGCMOs. Thus, we conclude that the auditor-client contract regime does
8
9
10 not seem to affect independence. However, as we have already pointed out, this result
11

t
12 should be interpreted with caution due to potential measurement errors in our variable

ip
13
14
YBYREAP.
15

cr
16
17 6. Concluding remarks
18

us
19 The main conclusion of this study is that auditor independence would be
20
21
22 impaired in lengthy engagements. The lower likelihood of NGCMOs with tenure cannot
23
24
25
26
an
be explained by higher accounting quality supposedly achieved in these engagements.

27 Thus, our results support the concerns expressed by the EC in the Green Paper on Audit
M
28
29 Policy about the negative effects of long audit firm tenures on the independence of
30
31
32
external auditors. Also, auditors seem willing to compromise independence in lengthy
ed

33
34 engagements regarding NGCMOs, but not for GCMOs, the latter involving particularly
35
36 high levels of litigation risk. These findings for the low litigation risk Spanish audit
pt

37
38
39 market, together with U.S. research evidence, which mostly fails to report a negative
ce

40
41 effect of tenure on independence, would support litigation risk as an appropriate
42
43
44 framework for addressing the auditor reporting decision.
Ac

45
46 Our results have methodological implications for the study of the tenure-audit
47
48
49 qualifications relationship. Prior research has mostly addressed the issue through the
50
51 analysis of the issuance of GCMOs to financially distressed firms. However, the same
52
53
evidence reported to justify GCMOs as a proper measure of independence would also
54
55
56 advocate the use of the issuance of NGCMOs to the whole population of firms. Our
57
58 results strongly support this view since whenever a significant effect is reported in the
59
60
61
62 26
63
64 Page 27 of 39
65
 model with NGCMOs, it has always the predicted sign. Our results also show that
1
2 GCMOs and NGCMOs need to be separately addressed, thus supporting a multinomial
3
4
5 approach.
6
7 Finally, this study has some interesting implications for regulators and policy
8
9
10 makers, particularly within the current discussion of mandatory audit firm rotation.
11

t
12 While most previous research has examined high litigation risk settings, we report

ip
13
14
evidence for a low litigation risk country. Next, the Spanish market, characterized by
15

cr
16
17 unusually long audit tenures, provides an ideal context to address the tenure-
18

us
19 independence relationship. Also, previous research has been performed under
20
21
22 conditions of voluntary partner rotation. As many countries have already enforced
23
24
25
26
an
mandatory partner rotation rules, the available evidence on the effects of tenure on

27 independence needs to be updated. Although our findings would support a mandatory

M
28
29 rotation rule at the firm level in order to strengthen the value of audit reports for
30
31
32
external users, the negative potential effects of such a rule, for example on the costs of
ed

33
34 audits, would require a more careful analysis. Moreover, the auditor-client contract
35
36 regime does not seem to affect either the propensity to issue audit qualifications or the
pt

37
38
39 loss of independence with tenure. However, due to potential measurement errors in the
ce

40
41 variable accounting for the contract regime, this result should be viewed as highly
42
43
44 preliminary and thus, further research would be undoubtedly welcome. Nevertheless,
Ac

45
46 the most important limitation of our study results from the small number of GCMOs in
47
48
49 our sample. Since our results regarding GCMOs are based on a low number of qualified
50
51 reports, they should be carefully taken into consideration.
52
53
54
55
56
57
58
59
60
61
62 27
63
64 Page 28 of 39
65
 References
1
2 Ameen, E.C., Chan, K. and Guffey, D.M., 1994. Information content of qualified audit
3
4
opinions for over-the-counter firms. Journal of Business Finance and
5 Accounting, 21, 9971011.
6
7 Ball, R., Walker, R.G. and Whittred, G.P., 1979. Audit qualifications and share prices.
8 Abacus, 15 (1), 1979, 23-34.
9
10
11 Butler, M., Leone, A. and Willenborg, M., 2004. An empirical analysis of auditor

t
12 reporting and its association with abnormal accruals. Journal of Accounting and

ip
13 Economics, 37, 139165.
14
15

cr
16 Carcello, J.V., Hermanson, D.R. and Huss, H.F., 1995. Temporal changes in
17 bankruptcy- related reporting. Auditing: A Journal of Practice and Theory, 14
18 (2), 133-143.

us
19
20
21 Carcello, J.V., and Neal, T., 2000. Audit committee composition and auditor reporting.
22 The Accounting Review, 75 (4), 453467.
23
24
25
26
an
Carcello, J.V., and Palmrose, Z.V., 1994. Auditor litigation and modified reporting on
bankrupt clients. Journal of Accounting Research, 32 (supplement), 1-30.
27
M
28 Carey, P.J. and Simnett, R., 2006. Audit partner tenure and audit quality. The
29 Accounting Review, 81 (3), 653-676.
30
31
32
Chen. P.F. He, S. Ma, Z. and Stice, D.E., 2012. Qualified audit opinions and debt
ed

33 contracting. Working Paper (Kobe University).

34
35 Choi, S. and Jeter, D., 1992. The effects of qualified audit opinions on earnings
36 response coefficients. Journal of Accounting and Economics, 15 (2-3), 229-247.
pt

37
38
39 Chow, C.W., and Rice, S.J., 1982. Qualified audit opinions and auditor switching. The
Accounting Review, 57 (2), 326-335.
ce

40
41
42
43
Chow, C.W., and Rice, S.J., 1982b. Qualified audit opinions and share prices - An
44 investigation. Auditing: A Journal of Practice and Theory, 1 (2), 35-53.
Ac

45
46 CNMV (Comisin Nacional del Mercado de Valores), 2009, Report on the Review of
47 the Annual Financial Reports Filed with the CNMV, (Madrid).
48
49
50 Craswell, A.T., 1988. The association between qualified opinions and auditor switches.
51 Accounting and Business Research, 19, 23-31.
52
53
54
Davis, L.R., Soo, B. S. and Trompeter, G. M., 2009. Auditor tenure and the ability to
55 meet or beat earnings forecasts. Contemporary Accounting Research, 26 (2),
56 517-548.
57
58 DeAngelo, L., 1981. Auditor size and audit quality. Journal of Accounting and
59
60 Economics, 3 (3), 183-199.
61
62
63
64 Page 29 of 39
65
 1 DeFond, M. and Park, C.W., 2001. The reversal of abnormal accruals and the market
2 valuation of earnings surprises. The Accounting Review 76 (3): 375404.
3
4
5 DeFond, M., Raghunandan, K. and Subramanyam, K.R., 2002. Do nonaudit service fees
6 impair auditor independence? Evidence from going-concern audit opinions.
7 Journal of Accounting Research, 40 (4), 12471274.
8
9
10 Directive 2006/43/EC of the European Parliament and of the Council of 17 May 2006
11 on statutory audits of annual accounts and consolidated accounts, OJ L157/2006

t
12 p.87 (2006).

ip
13
14
15
Dopuch, N., Holthausen, R. and Leftwich, R., 1987. Predicting audit qualifications with

cr
16 financial and market variables. The Accounting Review, 62 (3), 431-454.
17
18 European Commission, 2010. Audit policy: Lessons from the crisis. Green Paper.

us
19 (Brussels).
20
21
22 Firth, M., 2002. Auditor-provided consultancy services and their associations with audit
23 fees and audit opinion. Journal of Business Finance and Accounting, 29 (5-6),
24
25
26
661-693. an
27 Firth, M., Oliver, M. and Wu, X., 2012. How do various forms of auditor rotation affect
M
28 audit quality? Evidence from China. The International Journal of Accounting, 47
29 (1), 109-138.
30
31
32
Francis, J.R., 2004. What do we know about audit quality? The British Accounting
ed

33 Review, 36 (4), 345-368.

34
35 Geiger, M.A., and Raghunandan, K., 2002. Auditor tenure and audit reporting failures.
36 Auditing: A Journal of Practice and Theory, 21 (1), 67-78.
pt

37
38
39 Ghosh, A. and Moon, D., 2005. Auditor tenure and perceptions of audit quality. The
Accounting Review, 80 (2), 585612.
ce

40
41
42
43
Gul, F.A., Basioudis, I. and Ng, A., 2011. Non audit fees, auditor tenure and auditor
44 independence. International Symposium on Audit Research (ISAR).
Ac

45
46 Gul, F.A., Lee, D.S., and Lynn, M., 1992. A note on audit qualifications and switches:
47 some further evidence from a small sample study. Journal of International
48
49 Accounting, Auditing and Taxation, 1 (1), 111-120.
50
51 Hopwood, W., McKeown, J.C. and Mutchler, J., 1989. A test of the incremental
52 explanatory power of opinions qualified for consistency and uncertainty. The
53
54
Accounting Review, 64 (1), 28-48.
55
56 Hudaib, M. and Cooke, T., 2005. The impact of managing director changes and
57 financial distress on audit qualification and auditor switching. Journal of
58 Business Finance and Accounting, 32 (9-10), 1703-1739.
59
60
61
62
63
64 Page 30 of 39
65
 Johnson, V.E., Khurana, I. and Reynolds, J.K., 2002. Audit firm tenure and the quality
1 of financial reports. Contemporary Accounting Research, 19 (4), 637660.
2
3
4
Knechel, W. and Vanstraelen, A., 2007. The relationship between auditor tenure and
5 audit quality implied by going-concern opinions. Auditing: A Journal of Practice
6 and Theory, 26 (1), 113131.
7
8 Krishnan, J., 1994. Auditor switching and conservatism. The Accounting Review, 69
9
10 (1), 200-215.
11

t
12 Krishnan, J., and Krishnan, J., 1997. Litigation risk and auditor resignations. The

ip
13 Accounting Review, 72 (4), 539 60.
14
15

cr
16 Laitinen, E.K., and Laitinen, T., 1998. Qualified audit reports in Finland: Evidence from
17 large companies. European Accounting Review, 7(4), 639-653.
18

us
19 Lennox, C., 2000. Do companies successfully engage in opinion-shopping? Evidence
20
21 from the UK. Journal of Accounting and Economics, 29 (3), 321-337.
22
23 Levinthal, D.A., and Fichman, M., 1988. Dynamics of interorganizational attachments:
24
25
26
an
auditor client attachments. Administrative Science Quarterly, 33 (3), 345-369.

27 Lim, C. and Tan, H., 2010. Does auditor tenure improve audit quality? Moderating
M
28 effects of industry specialization and fee dependence. Contemporary Accounting
29 Research, 27(3), 923-957.
30
31
32
Loudder, M., Khurana, I., Sawyers, R. Cordery, C., Johnson, C., Lowe, J. and
ed

33 Wunderle, R., 1992. The information content of audit qualifications. Auditing: A

34 Journal of Practice and Theory, 11 (1), 6982.
35
36 Louwers, T., 1998. The relation between going-concern opinions and the auditor's loss
pt

37
38 function. Journal of Accounting Research, 36 (1), 143-156.
39
Lys, T., and Watts, R.L. 1994. Lawsuits against auditors. Journal of Accounting
ce

40
41 Research, 32, 65-93.
42
43
44 Mautz, R.K., and Sharaf, H.A. 1961. The Philosophy of Auditing. Sarasota, FL:
Ac

45 American Accounting Association.

46
47 McDaniel, L.S., 1990. The effects of time pressure and audit program structure on audit
48
49 performance. Journal of Accounting Research, 28 (2), 267-285.
50
51 Melumad, N., and Thoman, L., 1990. On auditors and the courts in an adverse selection
52 setting. Journal of Accounting Research, 28 (1), 77120.
53
54
55 Meyer, M., Rigsby, J. and Boone, J. (2007). The impact of auditor-client relationships
56 on the reversal of first-time audit qualifications. Managerial Auditing Journal, 22
57 (1), 53-79.
58
59
60
61
62
63
64 Page 31 of 39
65
 Myers, J., Myers, L. and Omer, T., 2003. Exploring the terms of the auditor-client
1 relationship and the quality of earnings: A case for mandatory auditor rotation?.
2 The Accounting Review, 78 (3), 779-799.
3
4
5 Myers, J., Myers, L., Palmrose, V. and Scholz, S. 2005. The length of auditor-client
6 relationships and financial statement restatements. Working Paper, Texas A&M
7 University.
8
9
10 Narayanan, V., 1994. An analysis of auditor liability rules. Journal of Accounting
11 Research, 32 (supplement), 3964.

t
12

ip
13 Palmrose, V., 1987, Litigation and independent auditors: the role of business failures
14
15
and management fraud. Auditing: A Journal of Practice and Theory, 6 (2), 90-

cr
16 102.
17
18 Ruiz-Barbadillo, E., Gomez Aguilar, N. and Carrera, N., 2006. Evidencia emprica

us
19 sobre el efecto de la duracin del contrato en la calidad de la auditora: anlisis
20
21 de las medidas de retencin y rotacin obligatoria de auditors. Investigaciones
22 Econmicas, 30 (2), 283-316.
23
24
25
26
an
Ruiz-Barbadillo, E., Gomez Aguilar, N., De Fuentes-Barber, C. and Garca Benau,
M.A., 2004. Audit quality and the going-concern decision making process:
27 Spanish evidence. European Accounting Review, 13 (4), 597-620.
M
28
29 St. Pierre, K., and Anderson, J., 1984. An analysis of the factors associated with
30 lawsuits against public accountants. The Accounting Review, 59 (2), 242-263.
31
32
ed

33 Shockley, R.A., 1982. Perceptions of audit independence: a conceptual model. Journal

34 of Accounting, Auditing and Finance, 5, 126 -143.
35
36 Shu, S., 2000. Auditor resignations: clientele effects and legal liability. Journal of
pt

37
38 Accounting and Economics, 29 (2), 173-205.
39
Simunic, D.A., 1980. The pricing of audit services: theory and evidence. Journal of
ce

40
41 Accounting Research, 18 (1), 161-190.
42
43
44 Simunic, D.A., 1984. Auditing, consulting, and auditor independence. Journal of
Ac

45 Accounting Research, 22 (2), 679-702.

46
47 Stanley, J.D. and DeZoort, F.T. (2007). Audit firm tenure and financial restatements:
48
49 An analysis of industry specialization and fee effects. Journal of Accounting and
50 Public Policy, 26 (2), 131-159.
51
52 Vanstraelen, A., 2000. Impact of renewable long-term audit mandates on audit quality.
53
54
European Accounting Review, 9 (3), 419-442.
55
56 Vanstraelen, A., 2002. Auditor economic incentives and going-concern opinions in a
57 limited litigious continental European business environment: empirical evidence
58 from Belgium. Accounting and Business Research, 32 (3), 171186.
59
60
61
62
63
64 Page 32 of 39
65
 Zmijewski, M.E., 1984. Methodological issues related to the estimation of financial
1 distress prediction models. Journal of Accounting Research, 22 (Supplement),
2 50-82.
3
4
5
6
7
8
9
10
11

t
12

ip
13
14
15

cr
16
17
18

us
19
20
21
22
23
24
25
26
an
27
M
28
29
30
31
32
ed

33
34
35
36
pt

37
38
39
ce

40
41
42
43
44
Ac

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 33 of 39
65
 Table 1
1 Classification of Audit Reports by Year and Auditor Opinion
2
3
4 Auditor
2002 2003 2004 2005 2006 2007 2008 2009
5 Opinion
6 Unqualified 80% 85% 84% 87% 90% 90% 82% 84%
7
8 NGCMOs 20% 15% 16% 12% 9% 8% 13% 11%
9 GCMOs 0% 0% 0% 1% 1% 2% 5% 5%
10
11 Number of
80 82 83 82 82 82 81 80

t
12 reports

ip
13
14 GCMO = going-concern modified opinions
15

cr
16
NGCMO = non-going-concern modified opinions
17
18

us
19
20
21
22
23
24
25
26
an
27
M
28
29
30
31
32
ed

33
34
35
36
pt

37
38
39
ce

40
41
42
43
44
Ac

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 34 of 39
65
 Table 2
1 Descriptive Statistics for Independent Variables by Auditor Opinion.
2
3 Mean [Median]
4 Variable Unqualified NGCMOs GCMOs Overall St. Dev. Max. Min.
5 -1.87 -1.03 -0.02 -1.73
PBANK 2.90 12.08 -67.55
6 [-1.53] [-1.27] *** [0.25] *** [-1.51]
7 6.91 5.56 6.93 6.73
SIZE 1.92 11.6 0.10
8 [6.64] [5.43] *** [6.92] [6.47]
9 1.18 1.17 1.26 1.18
10 AGE 0.18 1.38 0.00
[1.23] [1.23] [1.31] ** [1.23]
11

t
0.62 0.68 0.81 0.63
12 LEVERAGE 0.20 2.52 0.00
[0.66] [0.66] [0.84] ** [0.66]

ip
13 CHLEVERA 0.12 0.14 2.37 0.01
14 0.13 1.18 -1.86
GE [0.01] [-0.03] [0.67] ** [0.01]
15 0.17 0.08 0.08 0.16

cr
16 LIQUIDITY 0.22 2.99 0.00
[0.09] [0.03] *** [0.05] [0.08]
17 0.14 0.19 0.21 0.15
18 STOCKS 0.15 0.97 0.00
[0.11] [0.13] [0.16] [0.11]

us
19
0.08 0.32 0.33 0.12
20 LOSSES 0.32 1.00 0.00
[0.00] [0.00] *** [0.00] ** [0.00]
21
0.92 0.92 0.75 0.92
22 AUDFIRM 0.28 1.00 0.00
[1.00] [1.00] [1.00] ** [1.00]
23
24
25
26
CRISIS

DISACC
0.24
[0.00]
0.08
0.23
[0.00]
0.39
0.67
[1.00] ***
0.23
an 0.25
[0.00]
0.12
0.43

1.00
1.00

24.02
0.00

0.00
27 [0.03] [0.07] *** [0.05] * [0.04]
M
28 10.21 6.98 7.33 9.73
TENURE 5.88 23 1.00
29 [10.00] [6.00] *** [4.00] * [9.00]
30
31 *; **; *** Significant at the 10 percent, five percent and one percent levels, respectively.
32
ed

33 Mann-Whitney test of differences of medians for variables: PBANK, SIZE, AGE, LEVERAGE,
34 CHLEVERAGE, LIQUIDITY, STOCKS DISACC and TENURE.
35 Pearsons chi-square test for the dichotomous variables: LOSSES, AUDFIRM and CRISIS.
36
pt

37 GCMO = going-concern modified opinions;

38 NGCMO = non-going-concern modified opinions;
39 PBANK = probability of bankruptcy as measured by adjusted Zmijewski (1984) score, with the weights
proposed by Carcello et al. (1995);
ce

40
41 SIZE = the natural log of the firms total assets in book values at the end of the year;
42 AGE = natural log of the number of years a company has been listed by in the Spanish stock market;
43 LEVERAGE = the firms level of financial leverage calculated as total debt divided by total assets, both in
44 book values, at the end of the year.
Ac

45 CHLEVERAGE = change in LEVEVERAGE during the year;

46 LIQUIDITY = the sum of the firms cash positions divided between current liabilities;
47 STOCKS = the firms inventories divided by total assets at the end of the year, both in book values;
48 LOSSES = a binary variable with score 1 if the companys net profit in year t-1 is negative and 0
49 otherwise;
50 AUDFIRM = 1 if the company is audited by a Big 4 audit firm and 0 otherwise;
51 CRISIS = 1 for years 2002, 2008 and 2009, and 0 otherwise. The year 2002 corresponds to the dotcom
52 crisis and 2008 and 2009 to the beginning of the financial crisis;
53 DISACC = a measure of discretionary accruals, in absolute values, as defined by DeFond and Park (2001)
54 and later used by Carey and Simnett (2006);
55 TENURE = the number of consecutive years the company has been audited by the same firm.
56
57
58
59
60
61
62
63
64 Page 35 of 39
65
 Table 3
1 Multinomial Logistic Estimation Results: Firms with Unqualified Reports as Comparison Group
2
3 OPINION = f (PBANK, SIZE, AGE, LEVERAGE, CHLEVERAGE LIQUIDITY,
4 STOCKS, LOSSES, AUDFIRM, CRISIS, DISACC, TENURE)
5
6 N 652
7 Pseudo R2 0.25
8 Wald Chi (2) 150.22***
9
10
11

t
NGCMOs GCMOs
12

ip
Coefficient Coefficient
13
Relative Risk Relative Risk
14
Predicted Ratio Sig. Predicted Ratio Sig.
15 Variable

cr
sign (z-value) Level sign (z-value) Level
16
0.57 *** 0.84 ***
17
PBANK + 1.76 + 2.31
18
(3.79) (4.68)

us
19
20 -0.64 *** -0.72
21 SIZE - 0.53 - 0.93
22 (-5.31) (-0.30)
23 -0.96 1.32
24
25
26
AGE - 0.38
(-1.18)
0.03
an - 3.75
(0.49)
0.01
27 LEVERAGE + 1.03 + 1.01
M
28 (0.86) (0.24)
29 -0.12 0.22
CHLEVER
30 + 0.89 + 1.24
AGE
31 (-0.62) (1.50)
32 -4.47 *** -3.36 *
ed

33 LIQUIDITY - 0.01 - 0.03

34 (-3.02) (-1.58)
35 -1.49 -0.31
36 no
STOCKS + 0.23 0.73
pt

37 significant
(-1.45) (-0.23)
38 0.72 ** 0.39
39 LOSSES + 2.05 + 1.47
ce

40 (2.04) (0.45)
41 1.28 *** -0.12
42 AUDFIRM + 3.60 + 1.13
43 (2.80) (-0.18)
44
Ac

0.32 1.23 **
45 CRISIS + 1.38 + 3.43
46 (0.94) (1.99)
47
2.65 *** 2.66 ***
48
DISACC + 1.42 + 1.43
49
(3.22) (3.21)
50
-0.65 *** no -0.04
51
TENURE - (H1) 0.94 significant 0.96
52
53 (-2.69) (H2) (-0.56)
54 5.03 *** -4.14
Constant
55 (3.55) (-0.86)
56
57 *; **; *** Significant at 10 percent, five percent and one percent levels, respectively.
58
59 See Table 2 for variable definitions.
60
61
62
63
64 Page 36 of 39
65
 Table 4.
1 Bivariate Logistic Estimation Results
2
3
4
OPINION = f (PBANK, SIZE, AGE, LEVERAGE, CHLEVERAGE, LIQUIDITY,
5 STOCKS, LOSSES, AUDFIRM, CRISIS, DISACC, TENURE, DISACC*TENURE
6
7 631
8
Pseudo R2
Wald Chi (2) 0.24
9
10 96.89***
11

t
12

ip
13 Predicted Coefficient Sig.
Variables
14 sign (z-value) Level
15 0.45 ***

cr
16 PBANK +
(2.94)
17
18 -0.59 ***
SIZE -
(-3.90)

us
19
20 -0.82
21 AGE -
(-0.97)
22
23 0.06
24
25
26
LEVERAGE

CHLEVERAGE
+

+
an (1.39)
-0.09
(-0.39)
27
-4.39 ***
M
28 LIQUIDITY -
29 (-3.01)
30 -1.65
31 STOCKS +
(-1.55)
32
ed

33 0.77 **
LOSSES +
34 (2.15)
35 1.31 ***
36 AUDFIRM +
(2.84)
pt

37
0.37
38 CRISIS +
39 (1.08)
2.85 ***
ce

40
41 DISACC +
(3.15)
42
-0.07 ***
43 TENURE -
44 (-2.71)
Ac

45 DISACC*TENU -0.00
46 ?
RE (-0.98)
47
4.42 ***
48 Constant
49 (2.83)
50
51 Dependent variable coded 1 for NGCMOs and 0 for unqualified reports. Observations
52 with GCMOs are removed from the sample.
53
54
55 *, **, *** Significant at 10 percent, five percent and one percent levels, respectively.
56
57 See Table 2 for variable definitions.
58
59
60
61
62
63
64 Page 37 of 39
65
 Table 5.
1 Bivariate Logistic Estimation Results: Alternative Proxies for Accounting Quality
2
3 CURACC SMPROF SMIPROF RESTAT
4
5 N 631 631 631 565
6 Pseudo R2 0.24 0.23 0.23 0.23
7
Wald Chi (2) 96.89 90.07 90.25 87.66
8
9 Level of significance *** *** *** ***
10
11

t
Predicted Coefficient Sig. Coefficient Sig. Coefficient Sig. Coefficient Sig.
12

ip
sign (z-value) Level (z-value) Level (z-value) Level (z-value) Level
13
14 3.27 **
CURACC +
15 (2.10)

cr
16 0.23
17 SMPROF +
(0.40)
18

us
19 1.23 *
SMIPROF +
20 (1.75)
21 -0.10
22 RESTAT +
(-0.18)
23
24
25
26
TENURE -
-0.06
(-2.48)
-0.00
** an-0.06
(-2.36)
** -0.07
(-2.50)
** -0.07
(-2.38)
**

27 CURACC*TENURE ?
(-0.49)
M
28
-0.02
29 SMPROF*TENURE ?
30 (-0.24)
31 -0.00
32 SMIPROF*TENURE ?
ed

(-0.10)
33
0.00
34 RESTAT*TENURE ?
35 (0.05)
36
pt

37
38 *; **; *** Significant at 10 percent, five percent and one percent levels, respectively.
39
CURACC = the change in current assets (excluding cash) minus change in current liabilities (excluding
ce

40
41 change in short-term notes and current portion of long-term debt.
42 SMPROF = 1 when profit is less than two percent of total assets, and 0 otherwise.
43 SMIPROF = 1 when the increase in profit (decrease in loss) over last years profit/loss is less than two
44 percent of total assets, and 0 otherwise.
Ac

45 RESTAT= 1 if the company restates accounts the next year, and 0 otherwise.
46 See Table 2 for other variable definitions.
47
48 Dependent variable coded 1 for NGCMOs, and 0 for unqualified reports. Observations with GCMOs are
49 removed from the sample. Information about control variables is omitted.
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 38 of 39
65
Figure

Figure 1
Histogram of Audit Tenure

.08
.06

t
ip
Density

.04

cr
us
.02

an
0

0 5 10 15 20 25
Years audited by the same firm
M
ed
pt
ce
Ac

Page 39 of 39