Auditor Ratification and Shareholders’ Perception of
External Financial Reporting Quality✩
Jacob Justus Leidnera,1 , Sven Hörnera,2
a Julius-Maximilians-Universität
Würzburg
Abstract
Auditor ratification by shareholders is typically a routine matter and non-binding.
This study uses a returns-earnings design and demonstrates that the auditor ratification vote also constitutes market-relevant information. The empirical evidence
reveals that higher auditor approval ratios lead to greater market reactions to
earnings surprises. Moreover, this effect appears to be larger for firms with higher
amounts of free float. In addition, there are moderate indications that the time lag
between the auditor ratification date and the date of the earnings announcement
also influences the association. In light of these results, it appears appropriate
and economically reasonable for shareholder activists—and the Final Report of the
Advisory Committee on the Auditing Profession to the U.S. Department of the
Treasury—to call for mandatory shareholder ratification of auditors. The results
support our theoretical reasoning that auditor approval ratios signal shareholders’
satisfaction with the expected level of (perceived) audit quality. Concurrently, auditor
ratification also indirectly enables shareholders to signal their satisfaction with the
expected (perceived) external financial reporting quality.
JEL classification: G14, G38, M41, M42
Keywords: audit quality, auditor ratification, earnings response coefficient,
external financial reporting quality, shareholder voting
1. Introduction
On balance, prior research indicates that the opportunity for shareholders
to ratify an auditor might be important, particularly because such a vote is one
of the few—if not the only—situation in which shareholders can express their
assessment of the auditor (Marshall, 2005, 41; Sainty et al., 2002, 111). In this
✩ We wish to thank Ralf Ewert, Hansrudi Lenz and Andrea Szczesny and the participants
at the 2015 DART Mini Graduate Workshop—University of Graz, Center for Accounting
Research in Graz for their helpful suggestions.
1 Jacob Justus Leidner is a Ph.D. student and research assistant at the Julius-MaximiliansUniversität Würzburg, Chair of Financial Accounting, Auditing and Consulting.
2 Sven Hörner is a Ph.D. student and research assistant at the Julius-MaximiliansUniversität Würzburg, Chair of Financial Accounting, Auditing and Consulting and at the
Chair of Business Management, Controlling and Accounting.
context, it is not surprising that various key actors have called for mandatory
shareholder vote on this matter (Liu et al., 2009, 227), although shareholder
ratification of auditors is normally a routine and non-binding action.3 Nonetheless, one of the primary objectives of the Securities and Exchange Commission
(SEC) is, inter alia, to ensure the provision and disclosure of important information to shareholders.4 Indeed, previous research hints at a linkage between
shareholders’ perceptions of audit topics and their voting decisions and reactions to reported earnings. However, whether there is a relation between the
decision-usefulness of reported earnings and auditor approval voting results remains an open question. If auditor approval voting results are associated with
earnings’ decision-usefulness, mandatory and/or binding ratification of auditors
at the shareholders’ meeting might be advisable.
Previous studies have considered shareholders’ perceptions of audit- or
auditor-related issues. On one hand, some studies focus on shareholder votes
ratifying an auditor. This research shows that shareholders’ perception of audit quality determines the results of auditor ratification votes (Dao et al., 2008,
309; Raghunandan, 2003, 162; Sainty et al., 2002, 134). Furthermore, findings
in this literature stream demonstrate that shareholder approval of auditors is
meaningful because it seems to enhance audit quality (Dao et al., 2012, 162,
167; Mayhew and Pike, 2004, 817). On the other hand, a series of studies
seeks to determine whether perceived audit quality—e.g., studies referring to
Big N auditors (Gul et al., 2002, 46, 48; Teoh and Wong, 1993, 364) or the
provision of highly paid non-audit services (Eilifsen and Knivsfla, 2013, 109;
Francis and Ke, 2006, 520–521)—enhances the reliability of announced earnings, leading to greater market reactions to unexpected earnings as a result.5
As a consequence, we hypothesize that shareholder voting on auditor selection might be associated with the market’s response to unexpected earnings. We
3
For an example, see the petition for rulemaking (File No. 4–570) submitted to the SEC by
the California State Teachers’ Retirement System (https://www.sec.gov/rules/petitions/
2008/petn4-570.pdf; accessed on 2nd April 2015).
4 For example, the SEC states on its website: “The SEC oversees the key participants in
the securities world [...]. Here the SEC is concerned primarily with promoting the disclosure
of important market-related information, maintaining fair dealing, and protecting against
fraud” (http://www.sec.gov/about/whatwedo.shtml; accessed on 2nd April 2015). See also
Saul (1996, 135), SEC (2000b) and SEC (2000a).
5 When we use the word “market” (as in “market” reaction) we are referring to equity
markets and their investors and not to debt markets and debt investors.
2
argue that the influence of shareholders’ perceptions of firm- and auditor-specific
characteristics in the determination of perceived external financial reporting
quality (EFRQ) is indirectly captured by the results of shareholder voting on
auditor ratification. This argument is based on the notion that EFRQ can be
described as the conclusion of the interaction between financial reporting quality before the audit and audit quality. Against this backdrop, auditor approval
ratios can signal shareholders’ satisfaction with the expected level of (perceived)
audit quality. Simultaneously, auditor ratification also indirectly enables shareholders to signal their satisfaction with expected (perceived) EFRQ. Therefore,
we hypothesize that shareholders’ expectations regarding (perceived) EFRQ at
the time of the voting decision provides information about perceived EFRQ at
the date of the earnings announcement. Therefore, the auditor ratification vote
would be market-relevant information.
Using a returns-earnings design, this study’s results reveal that earnings’
decision-usefulness is associated with shareholder approval of auditors; the
higher the percentage of votes supporting an auditor’s engagement, the higher
the earnings response coefficient (ERC). Moreover, this effect appears to be
greater when firms have more free float. We also hypothesize that the time
lag between the date of the shareholders’ auditor ratification vote and that of
the earnings announcement influences this association. However, the results
concerning the latter two hypotheses are not fully statistically robust.
This study contributes to the literature in several ways. First, we extend
the shareholder voting literature by showing that auditor ratification results are
associated with earnings’ decision-usefulness, which indicates that such information might also benefit prospective shareholders. Second, we examine how the
ERC is influenced by a comprehensive variable—shareholder votes in support
of the auditor—which captures shareholders’ perceptions regarding the interaction of firm characteristics (e.g., board composition, internal controls) and the
auditor’s quality attributes (e.g., Big N auditor, specialization). Third, if perceived EFRQ is associated with shareholder approval, it is entirely legitimate
to debate policy recommendations regarding shareholder ratification of auditors
more intensively. Thus, the few years old report of the Advisory Committee on
the Auditing Profession (ACAP, 2008, VIII:20–VIII:21), which made a foray
3
into this domain, is a natural focal point of emphasis in this regard.
The remainder of this article is structured as follows: Section 2 provides
an overview of the previous literature regarding shareholder voting on auditor
ratification and shareholders’ perception of EFRQ and audit quality. In Section
3, we develop three hypotheses and explain the model specification (Section 3.1),
we describe both the sample selection procedure and the descriptive statistics
and discuss the empirical findings (Section 3.2), and we perform robustness
checks (Section 3.3). The article closes with a brief summary and an examination
of the study’s limitations (Section 4).
2. Theoretical Background and Prior Literature
Shareholder Voting and Auditor Ratification
Although studies of shareholders’ auditor ratifications have been performed
in the past, the topic has received substantial attention from investors, the
SEC and researchers over the last decade (Mishra et al., 2005, 10; SEC, 2000b).
Glezen, G. William and Millar (1985, 861, 868–869) study whether disclosure of
non-audit services—which is an indicator of auditor independence—is of particular interest to shareholders; notably, in making their determination, these authors use voting ratios supporting auditors’ ratification as a proxy for shareholders’ perception of auditor independence. Glezen, G. William and Millar (1985,
869) find that non-audit services are not related to ratification results in general;
thus, they conclude either that auditor independence is not important to shareholders or that the level of non-audit services provided is not perceived to limit
auditor independence. Raghunandan (2003, 155) finds that the high rate (approximately 97%) of auditor ratification by shareholders in his 2001 sample of
172 Fortune 1000 companies supports the latter argument; nonetheless, he also
finds that the non-audit to audit fee ratio has a positive and significant effect on
the percentage of shareholder votes against auditor’s ratification (Raghunandan,
2003, 160). Thus, the disclosure of non-audit services helps certain shareholders
determine how they vote their shares. In a similar vein, Mishra et al. (2005, 20–
21) find that various categories of non-audit services are perceived differently by
shareholders. Another study notes that such empirical observations depend on
the composition of the audit committee (Raghunandan and Rama, 2003, 260).
4
In addition to offering insights regarding auditor independence and shareholders’ voting behavior, other studies examine what additional factors determine
the results of auditor ratification votes. For example, among other findings that
they make, Sainty et al. (2002, 128) reveal that engagements of less credible
auditors and going concern opinions affect the proportion of votes that do not
support auditor’s ratification (against and abstain votes). By contrast, no effect is demonstrated by variables relating to auditors’ industry specialization
and audit tenure. More recent findings suggest that audit tenure is positively
associated with votes against an auditor (Dao et al., 2008, 305). There is also
evidence that financial restatements (Liu et al., 2009, 233–235) and an adverse
Section 404 internal control opinion (Hermanson et al., 2009, 403–405) influence
shareholders’ votes.
The literature also discusses which U.S. firms are likely to implement shareholder ratification of auditor selections at all. Accordingly, firms that pay higher
total fees and that have a qualitative audit committee (as measured by the proportion of financial experts) are more likely to seek shareholder votes on auditor
ratification. Dissatisfaction with the board negatively influences such decisions
(Krishnan and Ye, 2005, 248–249). Moreover, firms asking shareholders to ratify
the auditor (approximately 75% of the analyzed U.S. sample in fiscal year 2006)
are not only more likely to pay higher audit fees but also less likely to issue subsequent restatements and have lower abnormal accruals (Dao et al., 2012, 157,
159, 162, 167). These results demonstrate that shareholder participation in auditor selection enhances auditor independence and—as a result—audit quality;
this notion has also been supported experimentally (Mayhew and Pike, 2004,
817).
Studies examining shareholder ratification of auditors generally draw two
broad conclusions.
First, auditor approval by shareholders appears to
strengthen auditor independence and audit quality. Second, more recent evidence suggests that shareholders’ perceptions of auditor-related issues affect
their decisions regarding the auditor. Simultaneously, it remains debatable
whether perceptions of audit quality—including auditor independence—are related to shareholders’ investment decisions and whether such perceptions constitute important market-related information.
5
Perceived External Financial Reporting Quality, Perceived Audit
Quality, and the Earnings Response Coefficient
The objective of external audits is not an end in itself but to ensure a sufficient level of EFRQ, which implies that audited financial reports should provide
decision-useful information, whose two fundamental requirements are relevance
and reliability (FASB, 1978, Para. 16; FASB, 2010, Para. QC5).6 Thus, assuming a given level of relevance, an audit’s purpose is to safeguard an adequate
degree of reliability (FASB, 1978, Para. 8).
Even if the qualitative characteristics of decision-useful information and
EFRQ are not directly observable, studies of the market’s reaction to reported
earnings make it possible to measure shareholders’ related perceptions.7 In addition, studies find an association between audit quality and EFRQ (Francis,
2004, 353, 360), which also means that it is indirectly possible to examine perceived audit quality. Hence, a series of studies focus on shareholders’ perception of audit quality, as measured by the ERC. For instance, Teoh and Wong
(1993, 349, 364) examine the question regarding whether higher perceived audit
quality—represented by a Big 8 dummy—has a positive influence on the reliability of reported earnings. The authors’ results highlight increased ERCs for Big 8
clients in comparison with non-Big 8 clients. This result might be interpreted to
mean that the reported earnings of Big 8 clients are more reliable and, therefore,
that the perceived audit quality of the Big 8 is higher.8 Although, Big N auditors and its office size might essentially define audit quality (Francis and Yu,
2009, 1547–1548), Balsam et al. (2003, 89) provide evidence that the stock market’s reaction to earnings surprises is positively affected when the auditors are
industry specialists. From these results, it might be concluded that perceived
auditors’ expertise leads to higher perceived audit quality and, therefore, to
higher perceived earnings quality.
6
Hereafter, we use the terms “external financial reporting quality (EFRQ)” and “earnings
quality” interchangeably (cf. DeFond and Zhang, 2014, 281). The term “reliability”, which is
used throughout this article, is not entirely accurate under the current nomenclature of the
FASB. For a brief discussion of this topic, see FASB (2010, Para. BC3.25–3.26).
7 For additional information regarding different measures of earnings quality, see
Dechow et al. (2010) and Perotti and Wagenhofer (2014).
8 The authors of a study of Big 6 auditors in Australia (Gul et al., 2002, 40, 48) posit
that the findings of Teoh and Wong (1993) and their reasoning can mostly be explained by a
reduction in potential agency problems.
6
Another stream of the literature addresses shareholders’ perceptions of potentially compromised auditor independence, which is a sign of impaired audit
quality. In these analyses, researchers commonly focus on the economic bond
between auditor and auditee: audit and non-audit fees. Examining the earnings announcements of SEC-registered firms in the year immediately prior to
(fiscal year-ends from June 30, 2000 to December 31, 2001) and immediately
after the first mandatory fee disclosure, Higgs and Skantz (2006, 13–19) observe higher ERCs for firms with abnormally high total fees and audit fees. By
contrast, the statistical evidence demonstrating a negative relationship between
abnormally high non-audit services and the ERC is not convincing. Nevertheless, Francis and Ke (2006, 509) find that firms with high levels of non-audit
fees in the post-disclosure period have lower quarterly ERCs; thus, shareholders’ perceptions that an auditor is charging high non-audit fees might lead to
inferior perceived EFRQ because of greater shareholder skepticism regarding
auditor independence. This conclusion is supported by Krishnan et al. (2005,
131). However, Ghosh et al. (2009, 377–379) present contrary evidence when
controlling for client importance (total fees from an auditee as a percentage
of the auditor’s total revenue).9 Another study documents that the relationship between non-audit services and shareholders’ perceptions of audit quality
is also influenced by auditor specialization (Lim and Tan, 2008, 233). Finally,
Eilifsen and Knivsfla (2013, 101–107) find further evidence for these latter two
mentioned empirical findings. Overall, there is some evidence to suggest that
high levels of non-audit fees are associated with shareholders’ perceptions of
impaired auditor independence, which leads to a lower perceived EFRQ.
In summary, studies show that perceived audit quality affects shareholders’
assessment of the reliability of earnings information and, therefore, the perceived EFRQ. Nevertheless, the evidence is not entirely homogeneous and is
conditional on various combinations of firm characteristics (e.g., board composition, the audit committee, internal controls) and auditor characteristics (e.g.,
Big N auditor, specialization, independence).
9 Nonetheless, Gul et al. (2006, 806–808) observe a negative association between the ratio
of non-audit fees to total audit fees and a firm’s ERC only for non-BIG 6 clients in Australia.
7
3. Research Design, Sample, and Empirical Results
3.1. Formation of Hypotheses and Model Specification
Hypothesis 1
A regulatory change (NYSE Rule 452) in 2010 led more companies to
seek shareholder ratification of auditors because this typical routine matter helps firms reach quorums for their annual meetings. Thus, procedural
technicalities—as opposed to boards’ interest in shareholders’ opinions—might
have increased the importance of auditor ratification vote in recent years.10 In
addition, shareholder approval of auditors is not mandatory nor is the result
binding (Brown, Jr, J. Robert, 2012, 527–528). So, there are several reasons
why shareholders might not attach great importance to auditor approval ratios (cf. Hermanson et al., 2009, 394). Notwithstanding these issues, the SEC
assumes (SEC, 2000b; SEC, 2000a) and the empirical findings in the outlined
literature suggest that shareholders’ perceptions regarding audit-related issues
determine certain of their voting and investment actions, particularly because
this vote is one of the few or possibly the only opportunity for shareholders to
express their views about the auditor or their perceptions of the audit quality
(Marshall, 2005, 41; Sainty et al., 2002, 111; Saul, 1996, 135). It is thus not
surprising that shareholder activists argue that this shareholder vote should be
mandatory (Liu et al., 2009, 227–228; Mishra et al., 2005, 10–11). Thus, we
assume that there is an association between shareholder votes and the market’s
reaction to earnings surprises, i.e., perceived EFRQ.
Further, we believe that this association can be described in more detail.
Theoretical studies (Holthausen and Verrecchia, 1988, 83–87; Lev, 1989, 186–
187) reveal that the extent of the price reaction due to a single earnings announcement depends, inter alia, on the quality of the earnings signal, i.e., the
variance in earnings noise or the EFRQ.11 However, what is behind the variance
of the earnings noise? First, it can be assumed that this earnings noise consists of a random amount of material misstatements multiplied by the financial
10
http://ww2.cfo.com/risk-compliance/2010/06/
See,
for
example,
more-shareholder-say-on-auditors/
and
http://scgg.parkerpoe.com/
corporate-governance/whats-market-are-non-binding-auditor-ratification-votes-required/;
accessed on 2nd April 2015.
11 The theoretical framework and a simplified example are briefly outlined in the Appendix
on p. 30 et seqq.
8
reporting risk that the audited financial statements contain material misstatements. Such material misstatements might occur, for example, if the financial
statements did not correctly represent the substance of a firm’s economic transactions (cf. FASB, 2010, Para. BC3.26). Considering the logic behind the Audit
Risk Model (AICPA, 1984), the financial reporting risk that audited financial
statements contain material misstatements is again defined by both the auditor’s detection risk and the financial reporting risk that the unaudited financial
statements contain material misstatements.12 Together, the expected value and
the variance of the latter—which is a combined factor of the inherent risk and
the control risk—represent the financial reporting quality before the audit.13 By
contrast, audit quality is characterized by auditor’s detection risk and its variance.14 Both aspects of quality determine the variance of the earnings noise, i.e.,
the EFRQ. In sum, EFRQ is, inter alia, the result of a firm’s financial reporting
quality before the audit and the audit quality.
Insert Figure 1 here.
In addition, we suppose that a firm’s shareholders demand a certain
minimum level of EFRQ. If auditor ratification votes enables shareholders
to signal shareholders’ satisfaction with the level of perceived audit quality,
it also indirectly enables shareholders to signal their satisfaction with the
perceived EFRQ. This also means that the auditor ratification captures
shareholders’ perceptions regarding the interaction of firm characteristics and
the auditor’s quality attributes. The voting decision is made on the voting
date (V Dt ). As shown in Figure 1, the vote occurs before the audit of the
financial statements begins (ASt ). Thus, the vote represents an expression of
shareholders’ satisfaction or dissatisfaction with expected (perceived) audit
12 It is notable that the Audit Risk Model originally referred to the audit planning process. However, we interpret detection risk as the probability that the auditor does not detect
material misstatements. This probability is affected by auditor’s incentives, motivation, professional skepticism, expertise, chosen audit effort, etc. (cf. Knechel et al., 2013, 404–405).
Detection risk could also be understood to mean one minus the “market-assessed joint probability that a given auditor will both (a) discover a breach in the client’s accounting system,
and (b) report the breach” (DeAngelo, 1981, 186).
13 We interpret financial reporting quality before the audit in a fashion similar to
DeFond and Zhang (2014, 281–282). However, in our reasoning, the achievable level of EFRQ
is not constrained by the financial reporting quality before the audit.
14 For further discussions regarding the definition of audit quality, see DeFond and Zhang
(2014), Francis (2004), Francis (2011) and Knechel et al. (2013).
9
quality and, as a result, with expected (perceived) EFRQ. Consequently, a
high percentage of votes supporting an auditor’s engagement (VOTESFOR)
might indicate shareholders’ satisfaction with the expected (perceived) EFRQ.
On a later date (i.e., the earnings announcement date, EADt ), the market’s
reaction to earnings surprises (investment decisions) is observable and, so,
shareholders’ perception of the EFRQ. If the reasoning described above holds
and shareholders’ investment decisions are made in accordance with their
voting decisions, then shareholders’ expectations about the (perceived) EFRQ
on the date of the voting decision will yield information regarding perceived
EFRQ on the earnings announcement date.
This vote would therefore be
market-relevant information. The alternative form of Hypothesis 1 is stated as
follows:
H1a: Ceteris paribus, the higher the percentage of votes for (supporting)
the auditor’s engagement, the higher the ERC, i.e., the stock market’s reaction
to earnings surprises.
Hypotheses 2 & 3
A wide variety of studies have focused on the question of measuring the
extent to which audits or different levels of audit quality can effectively mitigate agency costs due to information asymmetries (DeAngelo, 1981, 185–187;
DeFond, 1992, 30; Francis and Wilson, 1988, 665–668; Watts, 1977, 57–59).15
The possible existence and complexity of information asymmetries depends
on a firm’s ownership structure, among other characteristics.16
Therefore,
it might be argued that owners of non-strategic holdings face higher levels
of information asymmetries than major shareholders (Jensen and Meckling,
1976, 312–330, 338–339). Major shareholders might have access to non-public
information sources (e.g., via appointed board members) and, as a result,
are not as reliant on published audited financial reports.
Moreover, some
authors have recently expressed general doubts about whether accounting
15 Of course, in addition to agency conflicts, other issues also determine the demand for
audits. See, for example, Francis et al. (2011).
16 There is also evidence that calls such a relationship into question; see Barton (2005, 577).
10
reports provide new information to shareholders (Ball, 2013, 848–850) and
whether EFRQ may have direct effects on firm’s value (Zimmerman, 2013,
888–889). For instance, Ball et al. (2012, 138–140, 146–150) show that audited
financial reports and other disclosed private information—such as voluntary
management earnings forecasts—are complements rather than substitutes.17
However, if there is at least a second- or third-order effect of different levels
of EFRQ on firm values—as posited by Zimmerman (2013)—we assume that
the following reasoning holds: higher levels of dispersed ownership mean that
there are higher levels of information asymmetries in principle, and, as a result,
shareholders will demand that published financial reports have higher levels
of reliability. Consequently, the audit process and the auditor become more
important. Finally, there is also evidence that shareholder votes in general
can be affected by a firm’s ownership structure (Raghunandan, 2003, 158). In
sum, the linkage in H1a is assumed to be stronger when higher percentages of
total shares have been issued to ordinary investors, i.e., non-strategic holdings
(FREEFLOAT). Thus, we propose our second hypothesis as follows:
H2a: Ceteris paribus, the higher the percentage of total shares issued to
ordinary investors (free float), the greater the effect of the percentage of votes
for (supporting) the auditor’s engagement on the ERC, i.e., the stock market’s
reaction to earnings surprises.
Using the ERC research methodology, the present analysis addresses the
perception of audit quality. We interpret the variable VOTESFOR to indicate
shareholders’ level of satisfaction regarding expected (perceived) audit quality,
given a certain expected level of (perceived) financial reporting quality before
the audit. However, shareholders’ satisfaction can change over time, and the
relationship of H1a can thus be amplified, diminished or even disappear because
the longer the time lag between the audit ratification date and the date of
earnings announcement, the greater the possibility that issues concerning the
17 Therefore, it is questionable whether audited reported earnings have more a “confirmation” function and are not a primary information source (Ball and Shivakumar, 2008, 1012;
Gigler and Hemmer, 1998, 138). Contradicting this perspective, Basu et al. (2013, 221) argue
that reported earnings represent a crucial source of new information.
11
ratified auditor might arise and directly influence shareholders’ expectations
about the quality of the audit. For example, Weber et al. (2008, 950–957)
demonstrate that KPMG’s clients suffered from negative abnormal returns
on event dates in the German ComROAD scandal, in which KPMG was the
auditor. Chaney and Philipich (2002, 1243) note similar outcomes regarding
the Enron case and Arthur Andersen’s clients. Of course, there might also be
situations that favor a particular auditor. Thus, variances in auditor reputation
might cause changes in the effect of H1a. In the end, we cannot identify what
auditor-related events can either confirm and strengthen or refute and weaken
a recent ratification vote. In addition, even if we could identify all such events,
it is almost impossible to determine whether the confirming effects would
outweigh the refuting effects or vice versa. Thus, the third hypothesis in its
alternative form reads as follows:
H3a: Ceteris paribus, the time lag between the date of the auditor ratification and the earnings announcement date affects the effect of the percentage
of votes for (supporting) the auditor’s engagement on the ERC, i.e., the stock
market’s reaction to earnings surprises.
Model Specification
We implement a returns-earnings methodology to answer our research question. It is chosen, because shareholders’ perception of EFRQ is reflected in
the extent of stock price responses to unexpected earnings and it determines
the capital allocation. Following prior research (Lev, 1989), the price reaction
around a firm’s fiscal year-end earnings release is measured by the cumulative
abnormal return (CAR)—i.e., the stock’s cumulative excess return over the
Standard and Poor’s 500 Composite return—aggregated over a 3-day window
(-1 day to +1 day) relative to the earnings announcement date.18 To test the
hypotheses that follow, we introduce the variable SURP, which is defined as the
18 The calculation of the CAR is based on the market model estimated over the 180day window ending 21 trading days before the earnings announcement date. Following
Bergh and Gibbons (2011, 552), we choose a long enough event window to capture the market’s price response to unexpected earnings. However, the window should also remain as
short as possible to protect against confounding events (McWilliams and Siegel, 1997, 636).
To address this trade-off, we check the results’ robustness using different event windows.
12
earnings surprise for a respective fiscal year. SURP is calculated as the reported
earnings for a respective fiscal year minus the mean earnings forecast for that
fiscal year one week before the earnings announcement date, as scaled by the
firm’s stock price one day before the event window. The variable SURP and
its interaction terms with the preceding introduced variables and the control
variables described below determine the ERC.
The model to test H1a is specified as follows:
CAR = α0 + α1 SU RP + α2 V OT ESF OR + α3 M V E + α4 M B + α5 M BN EG
+ α6 LEV + α7 BET A + α8 SU RP N EG + α9 AN ALY ST + α10 LIT
+ α11 V OT ESF OR × SU RP + α12 M V E × SU RP + α13 M B × SU RP
+ α14 M BN EG × SU RP + α15 LEV × SU RP + α16 BET A × SU RP
+ α17 SU RP N EG × SU RP + α18 AN ALY ST × SU RP + α19 LIT
× SU RP + α20 -α22 Y EAR20XX + α23 -α25 Y EAR20XX × SU RP + ε
(1)
In contrast to the Equation 1, the model to test H2a and H3a has to
include additional two- and three-way interactions. These are excluded in the
regression of H1a, because first we want to focus on the marginal effect of
VOTESFOR on the ERC. This effect can be analyzed directly in Equation 1
and is not depend on other regressors, i.e., three-way interaction terms, as it is
in Equation 2. We briefly discuss this issue again in Section 3.3. Based on the
above discussion, the following model is tested:
CAR = β0 + β1 SU RP + β2 V OT ESF OR + β3 F REEF LOAT + β4 T IM ELAG
+ β5 M V E + β6 M B + β7 M BN EG + β8 LEV + β9 BET A
+ β10 SU RP N EG + β11 AN ALY ST + β12 LIT + β13 V OT ESF OR
×F REEF LOAT +β14 V OT ESF OR×T IM ELAG+β15 V OT ESF OR
× SU RP + β16 F REEF LOAT × SU RP + β17 T IM ELAG × SU RP
+ β18 M V E × SU RP + β19 M B × SU RP + β20 M BN EG × SU RP
+ β21 LEV × SU RP + β22 BET A × SU RP + β23 SU RP N EG × SU RP
+ β24 AN ALY ST × SU RP + β25 LIT × SU RP + β26 V OT ESF OR
× F REEF LOAT × SU RP + β27 V OT ESF OR × T IM ELAG
× SU RP + β28 -β30 Y EAR20XX + β31 -β33 Y EAR20XX × SU RP + ε
(2)
In accordance with the prior literature (e.g., Fama and French, 1992), we introduce a further set of regressors in both models to control for additional firm
13
characteristics. Firm size is measured as the natural log of the market value of
equity (MVE; Atiase, 1985, 21–22). The market-to-book value of equity (MB)
proxies a firm’s growth opportunities (Hackenbrack and Hogan, 2002, 207, 213).
Following Higgs and Skantz (2006, 7), an indicator variable (MBNEG) controls
for a negative MB. A negative MB value is replaced with zero because MB ratios
that are less than zero are not economically reasonable. With respect to firm’s
risk, two independent variables are included in the regression. On the one hand,
a firm’s financing structure is represented by its leverage ratio (LEV), which is
calculated as total debt to total capital plus short-term debt plus the current
portion of long-term debt (Baber et al., 2014). On the other hand, the beta factor (BETA) captures a firm’s systematic risk (Collins and Kothari, 1989, 157).
Further, a dummy variable (SURPNEG) equals one for negative values of SURP;
this variable is introduced because shareholders capitalize unexpected negative
and positive earnings differently (Basu, 1997, 23). We also test for variations in
a firm’s predisclosure environment (Bhushan, 1989, 255; Teoh and Wong, 1993,
359, 364), which is calculated as the natural log of one plus the number of earnings estimates by analysts following the firm (ANALYST). Lastly, an indicator
variable (LIT) is set to one for industries characterized by a high exposure to
litigation risk (Rogers and Stocken, 2005, 1257; Zhan Shu, 2000, 187).
Insert Table 1 here.
3.2. Sample and Descriptive Statistics
The data for the sample are taken from four databases: Audit Analytics,
Datastream, I/B/E/S and Worldscope. First, we use Audit Analytics. As our
main variables of interest refer to shareholder ratification of auditors, we initially obtain 15,703 firm-years for SEC registrants for fiscal years 2010, 2011,
2012 and 2013. In addition, Audit Analytics provides information on other variables regarding auditors and formal information about financial statements (e.g.,
fiscal year-end date).19 Because this information is taken from sub-databases
of Audit Analytics, we eventually obtain 10,424 firm-years. Datastream is the
data origin for all the financial market-related variables, such as daily stock
19
The information from Audit Analytics is not necessarily restricted to inclusion in the
final regression. With respect to several variables, it is essential to accurately merge all the
data from the four different databases.
14
prices. Balance sheet and income statement data are collected from Worldscope. Using both databases, the sample decreases by 136 observations. It
is commonly acknowledged that I/B/E/S typically causes the biggest drop in
sample size because its coverage tends to be biased toward larger companies.20
In our case, the problem concerns some information that is relevant to calculating an earnings surprise, i.e., earnings per share and forecasts, in addition to
the calculation of the ANALYST variable. The sample consists of 8,245 firmyears after merging all four databases. Subsequently, the sample drops to 8,103
firm-years of 10-K-fillers. On the one hand, we control for significant inconsistencies in the dataset (e.g., overlapping dates regarding the vote date for the
fiscal year and the earnings announcement date for the previous fiscal year).
On the other hand, firm-years with time lags greater than 365 days between the
auditor ratification vote date and the earnings announcement date are deleted.
It should be ensured that the auditor ratification-related data is still timely
relevant. In addition, 50 observations concerning penny stocks are deleted because the literature shows that those are frequently linked to price anomalies
(Bali et al., 2005, 922; Ball et al., 1995, 104–105; Bhardwaj and Brooks, 1992,
558–559). Finally, the forecasts must be economically meaningful and approximate market opinion. Hence, earnings forecasts are only employed if at least
three analysts follow a firm (Barron et al., 2002, 829; Choi et al., 2006, 195–
196; Imhoff, Jr, Eugene A. and Lobo, 1992, 431).21 The final sample consists
of 6,928 firm-years and 2,455 different firms, respectively.
Finally, all continuous regression variables are winsorized (1st and 99th percentile) to protect the results against the possible influence of outliers.
Insert Table 2 here.
Certain aspects of the summary statistics warrant highlighting. The CAR
and the SURP are both close to zero, whether focusing on the mean or the median. Each variable is slightly left-skewed. On average, approximately 98% of
20 We are aware that there are further problems regarding I/B/E/S; see, for example,
Zhang, X. Frank (2006, 572).
21 This step in the sample selection process also indirectly addresses possible problems of
stale forecasts. Nevertheless, we are aware that this procedure enhances the sample’s big
company bias that is already present from using I/B/E/S forecast data. We briefly discuss
this issue again during the robustness checks.
15
all shareholders vote for (supporting) the auditor’s engagement, which is comparable with previous research. Although there is evidence that acceptance levels
decreased at the beginning of the 2000s (Hermanson et al., 2009, 394, 400) and
that auditor ratification has gained increasing importance in the aftermath of
Enron (Raghunandan and Rama, 2003, 262), our sample does not confirm such
trends with respect to current and prior periods. The free float lies between
31% and 100%, whereby circa three-quarters of all the observations are characterized by free floats of at least 73%. The time lag between the date of auditor
ratification and the earnings announcement date is a minimum of 158 days and
a maximum of 315 days (median: 273 days). The mean observation exhibits
an untransformed market value of equity of approximately $1.60 billion. The
highest noticeable skewness and kurtosis concern the variables MB (median:
1.96) and MBNEG (median: 0.00), which signifies the possible influence of outliers. Less than three percent of all the market-to-book ratios are negative and
are therefore replaced with the value of zero for MB. LEV ranges from 0.00 to
1.45, indicating that the pooled sample contains firms financed solely by equity
and clearly indebted firms. Further, circa 35% of all the observations show a
negative earnings surprise. The median observation has nine analysts following
the firm. Finally, approximately 23% of all firm-years belong to firms in highlitigation-risk industries. This category of firms is more than ten percentage
points smaller than the applicable category in the sample used by Ball et al.
(2012, 143). Table 3 presents the summary statistics of the pooled sample.
Insert Table 3 here.
In addition, Table 4 shows the Pearson product-moment correlation coefficients. Except for the correlation between MVE and ANALYST, an analysis
of these values does not indicate potential collinearity problems. However, this
simple procedure may be insufficient. Because our two regressions (Equation 1
and 2) include two- and three-way interactions, collinearity is present by construction.22 But, that is not problematic as long as the collinear variables
are significant and the F-statistic hint towards rejection of the null hypothe22
Indeed, the Variance Inflation Factors (VIF) indicate that possible collinearity problems
might be present. However, if we exclude all the interaction terms in Model 1 (Equation 1)—
and, therefore, the “constructed collinearity”—the highest VIF is 2.43 for the MVE variable.
16
sis that all coefficient estimates are jointly zero (Brambor et al., 2006, 70–71;
Eilifsen and Knivsfla, 2013, 92).
Insert Table 4 here.
3.3. Test of Hypotheses and Robustness Checks
The regression results for the pooled sample are shown in Table 5 and Table 6. Each regression is an OLS estimation that includes year dummies, and
standard errors are clustered by firm. The adjusted R2 values (0.065 for Model
1 and 0.069 for Model 2) are modest. However, they are at the higher end of
the normally observed range in this area of empirical research (Francis and Ke,
2006, 515; Krishnan et al., 2005, 128; Lev, 1989, 163–164).
Hypothesis 1—Regression Result
Simply put, the question of interest in H1a is technically whether VOTESFOR has an effect on the ERC. The ERC is given as the first derivative of
Equation 1 with respect to SURP.
∂CAR
= α1 + α11 V OT ESF OR + α12 M V E + α13 M B
∂SU RP
+ α14 M BN EG + α15 LEV + α16 BET A + α17 SU RP N EG (3)
+ α18 AN ALY ST + α19 LIT + α23 -α25 Y EAR20XX
= ERCE1
Finally, the effect of VOTESFOR on the ERC is mathematically determined by
the derivation of the ERC with respect to VOTESFOR.
∂ERCE1
= α11
∂V OT ESF OR
(4)
The empirical outcome of α11 is positive (coef. of 6.48) and highly significant. If
a one-tailed test is calculated, the p-value totals 0.005. Thus, VOTESFOR has a
positive impact on the ERC. As illustrated above (H1a), this result implies that
a higher percentage of votes for (supporting) an auditor’s engagement indicates
the shareholders’ satisfaction with the perceived audit quality and, therefore,
with the perceived EFRQ. An increased market response to earnings surprises
results because shareholders rely more on reported information. Therefore, there
appears to be evidence that shareholders are interested in auditor ratification
because investment decisions are associated with the voting decision regarding
17
auditor ratification. Because the result of auditor ratification voting—i.e., shareholders’ perceptions regarding the interaction of firm characteristics and auditor’s quality attributes—is positively associated with the decision-usefulness of
reported earnings, this information might also benefit prospective shareholders.
From a policy perspective, it seems to be worth discussing whether shareholder
approval of auditors should be mandatory and/or binding.
Insert Table 5 here.
Hypotheses 2 & 3—Regression Results
The same procedure—i.e., derivatives of Equation 2 with respect to SURP,
then VOTESFOR and finally with respect to FREEFLOAT or TIMELAG,
respectively—is also used to analyze H2a & H3a: Do ownership structure and
the time between the auditor ratification date and the date of the earnings announcement influence the relationship in H1a? However, prior answering the
question, one further (technical) note regarding the relationship of H1a: To capture the effect of VOTESFOR on the ERC, Equation 5 must be now considered
(Brambor et al., 2006, 71–77).
(∂CAR/∂SU RP )
= β15 + β26 F REEF LOAT + β27 T IM ELAG
∂V OT ESF OR
(5)
Here, the marginal effect of VOTESFOR on the ERC depends, in addition,
on the firm’s free float and the TIMELAG variable. The three coefficients are
jointly different from zero (P rob > F of 0.021), and the calculation of the
marginal effect of VOTESFOR on the ERC for an average firm results in a
value of 2.80.
Referring to H2a, we assume that the free float is positively related to the
effect of VOTESFOR on the ERC. This assumption is based on the economic
reasoning that a higher amount of free float is accompanied by a higher demand
for reliability in published financial reports. Therefore, the audit, auditor, and
auditor ratification are emphasized, which should result in a stronger effect for
H1a. For this purpose, β26 is examined (coef. of 30.17) because it is the result
of the necessary derivatives of Equation 2.
(∂CAR/∂SU RP )/∂V OT ESF OR
= β26
∂F REEF LOAT
18
(6)
The non-existence of the hypothesized effect can be rejected at a 0.030 significance level (one-tailed test). In light of higher levels of information asymmetry,
this result provides evidence that the auditor ratification vote is particularly
meaningful for firms characterized by higher free floats. The result also suggests
that the debate regarding mandatory and/or binding shareholder approval of
auditors is especially worthwhile for firms with high percentages of non-strategic
holdings.
Insert Table 6 here.
Referring to Equation 7, the coefficient β27 is highly significant (two-tailed
p-value of 0.013). In the first instance, this result demonstrates that the effect
of VOTESFOR on the ERC is also determined by the time elapsed after auditor
ratification.
(∂CAR/∂SU RP )/∂V OT ESF OR
= β27
∂T IM ELAG
(7)
In the present sample, a positive effect (coef. of 0.22) is observable. This result
can be interpreted to mean that the effects of confirming the latest ratification
vote outweigh the non-confirming effects. Furthermore, if the overall circumstances support shareholders’ voting results over time, the shareholders’ vote
should have a greater effect on the ERC. This result possibly indicates that
shareholders execute their auditor ratification vote thoroughly and give weight
to it. Nevertheless, in our opinion, such reasoning is not universal and depends
on the specific setting and sample examined.
Robustness Checks
Qualitatively similar results are found when the CARs are summed over
other event windows, i.e., -1 to 0, -2 to +2 and -3 to +3. However, for the -3 to
+3 period, the three-way interaction of V OT ESF OR×F REEF LOAT ×SU RP
is only barely significant (one-tailed p-value of 0.104). Further, the empirical
evidence is insensitive to the selected market return index (Standard and Poor’s
500 Composite versus Dow Jones Industrials). The same applies when the
respective variables refer to median forecasts rather than mean forecasts.
Some control variables are also replaced to provide another impression of the
robustness of the results. First, to proxy firm size the natural log of total asset is
used (Balsam et al., 2003, 76). Second, leverage can be measured as total debt
19
to common equity (Francis and Ke, 2006, 502). Third, the values of the beta
factors also depend on the estimation period (Dimson and Marsh, 1983, 756,
773). Hence, we include beta factors calculated over five years with monthly
data on the fiscal year-end dates instead of those from the market model regression. Fourth, we introduce a dummy variable that is related to negative
net income rather than SURPNEG (Krishnan et al., 2005, 118). In all four
cases, the regression results are robust. In addition, we include FREEFLOAT
and TIMELAG in Equation 1 as control variables since they are introduced in
Equation 2; the results regarding H1a are qualitatively unchanged. We argue
that VOTESFOR indicates shareholders’ level of satisfaction regarding expected
(perceived) audit quality, given a certain expected level of (perceived) financial
reporting quality before the audit, which, therefore, captures shareholders’ perceptions regarding the interaction of firm characteristics and the auditor characteristics. If we nonetheless control for auditor-specific variables—namely Big
4, audit fees, non-audit to audit fee ratio and auditor change—the hypotheses
interpretations are not altered, which indicates auditor approval ratios comprise
additional market-relevant information.
Moreover,
we
control
for
potential
(Chenhall and Moers, 2007, 175–186;
time-invariant
endogeneity
Roberts and Whited, 2012, 8–24),
and a fixed effects analysis does not alter our stated conclusions. If we use
non-winsorized data, the results—except for the three-way interaction of
V OT ESF OR × F REEF LOAT × SU RP —are qualitatively similar for the
fixed effects regressions (Dyckman and Zeff, 2014, 702).
The procedure of
dropping observations with fewer than three analysts following (Table 2) might
strengthen the sample’s big company bias. Therefore, both regressions are
re-estimated based on a sample that includes the 1,125 firm-years in question.
The results are almost the same except for the interaction term regarding
H3a—its two-tailed p-value equals 0.154.
4. Summary and Limitations
Today, shareholder ratification of auditors is frequently a routine, nonbinding matter, which may seem surprising because it is one of the few ways
for shareholders to express their views about a firm’s auditor and, as a re-
20
sult, expected (perceived) audit quality (Marshall, 2005, 41; Sainty et al., 2002,
111). Therefore, it appears appropriate for shareholder activists as well as
the ACAP to demand mandatory shareholder ratification of auditors (ACAP,
2008, VIII:20–VIII:21; Liu et al., 2009, 227). Indeed, such an approach might
be sound because research suggests that shareholder approval of auditors enhances audit quality (Dao et al., 2012, 162, 167; Krishnan and Ye, 2005, 248–
249). Moreover, the extent of stock price responses to unexpected earnings is
conditional on shareholders’ perception of audit quality (Eilifsen and Knivsfla,
2013, 101–107; Francis and Ke, 2006, 509).
The empirical evidence presented in this study demonstrates that shareholders’ auditor ratification votes—which captures shareholders’ perceptions regarding the interaction of firm characteristics and auditor’s quality attributes—are
associated with the decision-usefulness of reported earnings; in other words, the
higher the percentage of votes supporting an auditor’s engagement, the higher
the ERC. In addition, this effect seems to be positively influenced by higher
levels of dispersed ownership. Finally, there are some moderate indications that
the time lag between the date of the auditor ratification vote and the earnings announcement date affects the relationship between shareholder approval
of the auditor and the ERC. To summarize, the results suggest that it might
be worth discussing the implementation of auditor ratification as a mandatory
and/or binding agenda item at shareholders’ meetings—which comports with
the views of Hermanson et al. (2009, 407) and Liu et al. (2009, 238). In particular, these results apply to firms with high percentages of total shares issued to
ordinary investors, i.e., non-strategic holdings. Finally, the results support the
idea that such shareholder votes are “more than a symbolic act” (Saul, 1996,
135). Whether it is economically meaningful to regulate this matter is beyond
the scope of this study, and further research is required to answer this question.
Although our results are largely robust, all empirical research has limitations,
including that conducted in this study. The ERC framework is used to examine
shareholders’ perceptions. Even if our adjusted R2 values are relatively high
compared with prior research (Francis and Ke, 2006, 515; Higgs and Skantz,
2006, 14–15; Lev, 1989, 163–164), returns-earnings regressions are apparently associated with an omitted variable problem (Balsam et al., 2003, 95;
21
Dechow et al., 2010, 370). Another approach to examine the market’s perception might be to measure it via the cost of capital (e.g., Mansi et al., 2004).
Even though the use of this methodology may be debatable (DeFond and Zhang,
2014, 288–289), it would nonetheless enable the analysis of the perceptions of
debt investors as well. Strictly speaking, the findings presented are valid for
the U.S. sample for the fiscal years we considered (Dyckman and Zeff, 2014,
703). Moreover, our results are constrained to firms that have shareholder votes
to ratify auditors. As demonstrated in prior research (Dao et al., 2012, 156;
Krishnan and Ye, 2005, 247), these firms may differ from those that do not seek
such votes. In the end, there might be situations in which shareholders do not
provide the required instructions to their brokers regarding how to vote on this
matter, i.e., the cases of so-called “broker non-votes”. However, such cases of
reported broker non-votes concerning shareholder auditor ratification are rare.23
These cases might be significant for shareholder voting-related research questions such as ours, which makes this topic an interesting one to examine in
future studies.
References
ACAP (October 6, 2008). Final Report of the Advisory Committee on the
Auditing Profession to the U.S. Department of the Treasury.
AICPA (1984). Statement on Auditing Standards No. 47 – Audit Risk and Materiality in Conducting an Audit. The Journal of Accountancy, 157(2):143–
146.
Atiase, R. K. (1985). Predisclosure Information, Firm Capitalization, and Security Price Behavior Around Earnings Announcements. Journal of Accounting Research, 23(1):21–36.
Baber, W., Krishnan, J., and Zhang, Y. (2014). Investor Perceptions of the
Earnings Quality Consequences of Hiring an Affiliated Auditor. Review of
Accounting Studies, 19(1):69–102.
23 In our sample, 128 out of 6,928 firm-years report a value for “broker non-votes”. The
median of those 128 observations equals 5.65%.
22
Bali, T. G., Cakici, N., Yan, X. S., and Zhang, Z. (2005). Does Idiosyncratic
Risk Really Matter? The Journal of Finance, 60(2):905–929.
Ball, R. (2013). Accounting Informs Investors and Earnings Management is
Rife: Two Questionable Beliefs. Accounting Horizons, 27(4):847–853.
Ball, R., Jayaraman, S., and Shivakumar, L. (2012). Audited financial reporting and voluntary disclosure as complements: A test of the Confirmation
Hypothesis. Journal of Accounting and Economics, 53(1–2):136–166.
Ball, R., Kothari, S. P., and Shanken, J. (1995). Problems in measuring portfolio
performance: An application to contrarian investment strategies. Journal
of Financial Economics, 38(1):79–107.
Ball, R. and Shivakumar, L. (2008). How Much New Information Is There in
Earnings? Journal of Accounting Research, 46(5):975–1016.
Balsam, S., Krishnan, J., and Yang, J. S. (2003). Auditor Industry Specialization
and Earnings Quality. Auditing: A Journal of Practice & Theory, 22(2):71–
97.
Barron, O. E., Byard, D., and Kim, O. (2002). Changes in Analysts’ Information
around Earnings Announcements. The Accounting Review, 77(4):821–846.
Barton, J. (2005). Who Cares about Auditor Reputation?
Contemporary
Accounting Research, 22(3):549–586.
Basu, S. (1997). The conservatism principle and the asymmetric timeliness of
earnings. Journal of Accounting and Economics, 24(1):3–37.
Basu, S., Duong, T. X., Markov, S., and Tan, E.-J. (2013). How Important are
Earnings Announcements as an Information Source? European Accounting
Review, 22(2):221–256.
Bergh, D. D. and Gibbons, P. (2011). The Stock Market Reaction to the Hiring
of Management Consultants: A Signalling Theory Approach. Journal of
Management Studies, 48(3):544–567.
Bhardwaj, R. K. and Brooks, L. D. (1992). The January Anomaly: Effects of
Low Share Price, Transaction Costs, and Bid-Ask Bias. The Journal of
Finance, 47(2):553–575.
23
Bhushan, R. (1989). Firm characteristics and analyst following. Journal of
Accounting and Economics, 11(2–3):255–274.
Brambor, T., Clark, W. R., and Golder, M. (2006). Understanding Interaction
Models: Improving Empirical Analyses. Political Analysis, 14(1):63–82.
Brown, Jr, J. Robert (2012). The Politicization of Corporate Governance: Bureaucratic Discretion, the SEC, and Shareholder Ratification of Auditors.
Harvard Business Law Review, 2(1):501–534.
Burt, O. R. and Finley, R. M. (1968). Statistical Analysis of Identities in Random Variables. American Journal of Agricultural Economics, 50(3):734–
744.
Chaney, P. K. and Philipich, K. L. (2002). Shredded Reputation: The Cost of
Audit Failure. Journal of Accounting Research, 40(4):1221–1245.
Chenhall, R. H. and Moers, F. (2007). The Issue of Endogeneity within TheoryBased, Quantitative Management Accounting Research. European Accounting Review, 16(1):173–196.
Choi, J.-H., Kang, T., and Yoo, Y. K. (2006). The Association between Analysts’ Earnings Forecast Dispersion and the Magnitude of Earnings Response Coefficient: “Noise” or “Uncertainty”? Journal of Contemporary
Accounting & Economics, 2(2):190–207.
Collins, D. W. and Kothari, S. P. (1989). An analysis of intertemporal and
cross-sectional determinants of earnings response coefficients. Journal of
Accounting and Economics, 11(2–3):143–181.
Dao, M., Mishra, S., and Raghunandan, K. (2008). Auditor Tenure and Shareholder Ratification of the Auditor. Accounting Horizons, 22(3):297–314.
Dao, M., Raghunandan, K., and Rama, D. V. (2012). Shareholder Voting on
Auditor Selection, Audit Fees, and Audit Quality. The Accounting Review,
87(1):149–171.
DeAngelo, L. E. (1981). Auditor size and audit quality. Journal of Accounting
and Economics, 3(3):183–199.
24
Dechow, P., Ge, W., and Schrand, C. (2010). Understanding earnings quality:
A review of the proxies, their determinants and their consequences. Journal
of Accounting and Economics, 50(2–3):344–401.
DeFond, M. and Zhang, J. (2014). A review of archival auditing research.
Journal of Accounting and Economics, 58(2–3):275–326.
DeFond, M. L. (1992). The Association Between Changes in Client Firm Agency
Costs and Auditor Switching. Auditing: A Journal of Practice & Theory,
11(1):16–31.
Dimson, E. and Marsh, P. R. (1983). The Stability of UK Risk Measures and
The Problem of Thin Trading. Journal of Finance, 38(3):753–783.
Dyckman, T. R. and Zeff, S. A. (2014). Some Methodological Deficiencies in Empirical Research Articles in Accounting. Accounting Horizons, 28(3):695–
712.
Eilifsen, A. and Knivsfla, K. H. (2013). How Increased Regulatory Oversight
of Nonaudit Services Affects Investors’ Perceptions of Earnings Quality.
Auditing: A Journal of Practice & Theory, 32(1):85–112.
Fama, E. F. and French, K. R. (1992). The Cross-Section of Expected Stock
Returns. The Journal of Finance, 47(2):427–465.
FASB (1978). Concepts Statement No. 1: Objectives of Financial Reporting by
Business Enterprises.
FASB (2010). SFAC No. 8: Conceptual Framework for Financial Reporting:
Chapter 1, The Objective of General Purpose Financial Reporting, and
Chapter 3, Qualitative Characteristics of Useful Financial Information.
Francis, J. R. (2004). What do we know about audit quality?
The British
Accounting Review, 36(4):345–368.
Francis, J. R. (2011). A Framework for Understanding and Researching Audit
Quality. Auditing: A Journal of Practice & Theory, 30(2):125–152.
Francis, J. R. and Ke, B. (2006). Disclosure of Fees Paid to Auditors and the
Market Valuation of Earnings Surprises. Review of Accounting Studies,
11(4):495–523.
25
Francis, J. R., Khurana, I. K., Martin, X., and Pereira, R. (2011). The Relative
Importance of Firm Incentives versus Country Factors in the Demand for
Assurance Services by Private Entities. Contemporary Accounting Research,
28(2):487–516.
Francis, J. R. and Wilson, E. R. (1988). Auditor Changes: A Joint Test of Theory Relating to Agency Costs and Auditor Differentiation. The Accounting
Review, 63(4):663–682.
Francis, J. R. and Yu, M. D. (2009). Big 4 Office Size and Audit Quality. The
Accounting Review, 84(5):1521–1552.
Ghosh, A. A., Kallapur, S., and Moon, D. (2009). Audit and non-audit fees and
capital market perceptions of auditor independence. Journal of Accounting
and Public Policy, 28(5):369–385.
Gigler, F. and Hemmer, T. (1998). On the Frequency, Quality, and Informational Role of Mandatory Financial Reports. Journal of Accounting Research, 36(3):117–147.
Glezen, G. William and Millar, J. A. (1985). An Empirical Investigation of
Stockholder Reaction to Disclosures Required by ASR No. 250. Journal of
Accounting Research, 23(2):859–870.
Gul, F. A., Lynn, S. G., and Tsui, Judy S. L. (2002). Audit Quality, Management
Ownership, and the Informativeness of Accounting Earnings. Journal of
Accounting, Auditing & Finance, 17(1):25–49.
Gul, F. A., Tsui, J., and Dhaliwal, D. S. (2006). Non-audit services, auditor quality and the value relevance of earnings. Accounting & Finance,
46(5):797–817.
Hackenbrack, K. E. and Hogan, C. E. (2002). Market Response to Earnings
Surprises Conditional on Reasons for an Auditor Change. Contemporary
Accounting Research, 19(2):195–223.
Hermanson, D. R., Krishnan, J., and Ye, Z. S. (2009). Adverse Section 404
Opinions and Shareholder Dissatisfaction toward Auditors. Accounting
Horizons, 23(4):391–409.
26
Higgs, J. L. and Skantz, T. R. (2006). Audit and Nonaudit Fees and the Market’s
Reaction to Earnings Announcements. Auditing: A Journal of Practice &
Theory, 25(1):1–26.
Holthausen, R. W. and Verrecchia, R. E. (1988). The Effect of Sequential
Information Releases on the Variance of Price Changes in an Intertemporal
Multi-Asset Market. Journal of Accounting Research, 26(1):82–106.
Imhoff, Jr, Eugene A. and Lobo, G. J. (1992). The Effect of Ex Ante Earnings
Uncertainty on Earnings Response Coefficients. The Accounting Review,
67(2):427–439.
Jensen, M. C. and Meckling, W. H. (1976). Theory of the firm: Managerial
behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4):305–360.
Knechel, W. R., Krishnan, G. V., Pevzner, M., Shefchik, L. B., and Velury, U. K.
(2013). Audit Quality: Insights from the Academic Literature. Auditing:
A Journal of Practice & Theory, 32(Supplement 1):385–421.
Krishnan, J., Sami, H., and Yinqi Zhang (2005). Does the Provision of Nonaudit
Services Affect Investor Perceptions of Auditor Independence? Auditing:
A Journal of Practice & Theory, 24(2):111–135.
Krishnan, J. and Ye, Z. S. (2005). Why Some Companies Seek Shareholder
Ratification on Auditor Selection. Accounting Horizons, 19(4):237–254.
Lev, B. (1989). On the Usefulness of Earnings and Earnings Research: Lessons
and Directions from Two Decades of Empirical Research. Journal of Accounting Research, 27(3):153–192.
Lim, C.-Y. and Tan, H.-T. (2008). Non-audit Service Fees and Audit Quality:
The Impact of Auditor Specialization. Journal of Accounting Research,
46(1):199–246.
Liu, L.-L., Raghunandan, K., and Rama, D. (2009). Financial Restatements and
Shareholder Ratifications of the Auditor. Auditing: A Journal of Practice
& Theory, 28(1):225–240.
27
Mansi, S. A., Maxwell, W. F., and Miller, D. P. (2004). Does Auditor Quality
and Tenure Matter to Investors? Evidence from the Bond Market. Journal
of Accounting Research, 42(4):755–793.
Marshall, J. (2005).
Testing the Winds of Reform.
Financial Executive,
21(4):39–41.
Mayhew, B. W. and Pike, J. E. (2004). Does Investor Selection of Auditors
Enhance Auditor Independence? The Accounting Review, 79(3):797–822.
McWilliams, A. and Siegel, D. (1997).
Event Studies in Management Re-
search: Theoretical and Empirical Issues. Academy of Management Journal, 40(3):626–657.
Mishra, S., Raghunandan, K., and Rama, D. V. (2005). Do Investors’ Perceptions Vary with Types of Nonaudit Fees? Evidence from Auditor Ratification Voting. Auditing: A Journal of Practice & Theory, 24(2):9–25.
Perotti, P. and Wagenhofer, A. (2014). Earnings Quality Measures and Excess
Returns. Journal of Business Finance & Accounting, 41(5-6):545–571.
Raghunandan, K. (2003). Nonaudit Services and Shareholder Ratification of
Auditors. Auditing: A Journal of Practice & Theory, 22(1):155–163.
Raghunandan, K. and Rama, D. V. (2003). Audit Committee Composition
and Shareholder Actions: Evidence from Voting on Auditor Ratification.
Auditing: A Journal of Practice & Theory, 22(2):253–263.
Roberts, M. R. and Whited, T. M. (2012). Endogeneity in Empirical Corporate
Finance.
Rogers, J. L. and Stocken, P. C. (2005). Credibility of Management Forecasts.
The Accounting Review, 80(4):1233–1260.
Sainty, B. J., Taylor, G. K., and Williams, D. D. (2002). Investor Dissatisfaction
toward Auditors. Journal of Accounting, Auditing & Finance, 17(2):111–
136.
Saul, R. S. (1996). What Ails the Accounting Profession? Accounting Horizons,
10(2):131–137.
28
SEC (June 30, 2000b). Proposed Rule: Revision of the Commission’s Auditor
Independence Requirements.
SEC (November 21, 2000a). Final Rule: Revision of the Commission’s Auditor
Independence Requirements.
Teoh, S. H. and Wong, T. J. (1993). Perceived Auditor Quality and the Earnings
Response Coefficient. The Accounting Review, 68(2):346–366.
Watts, R. L. (1977). Corporate Financial Statements, A Product of the Market
and Political Processes. Australian Journal of Management, 2(1):53–75.
Weber, J., Willenborg, M., and Zhang, J. (2008). Does Auditor Reputation
Matter? The Case of KPMG Germany and ComROAD AG. Journal of
Accounting Research, 46(4):941–972.
Zhan Shu, S. (2000). Auditor resignations: clientele effects and legal liability.
Journal of Accounting and Economics, 29(2):173–205.
Zhang, X. Frank (2006). Information Uncertainty and Analyst Forecast Behavior. Contemporary Accounting Research, 23(2):565–590.
Zimmerman, J. L. (2013). Myth: External Financial Reporting Quality Has a
First-Order Effect on Firm Value. Accounting Horizons, 27(4):887–894.
29
Appendix
Theoretical Framework and Simplified Example
The theoretical model developed by Lev (1989, 186–187), which is first illustrated, considers the revision of a firm’s market price due to a single earnings
announcement. The extent of the price reaction depends, inter alia, on the
quality of this earnings signal, i.e., the EFRQ.
Starting at date 0, the price of a firm, P0 , equals the present value of the ung ),
known random future cash flows to the firm’s risk-neutral shareholders, E(CF
which follows a normal distribution.
g)
P0 = E(CF
g ∼ N (E(CF
g ), σ 2 )
CF
(8)
At date 1, the firm releases an earnings signal, e1 , before any cash flow to
the firm’s shareholders is observable. As a result of this signal, shareholders can
revalue the firm because the expected future cash flows are linked to the firm’s
earnings. The earnings signal corresponds to a scale factor, a, multiplied by
g , plus a random noise term,
the present value of random future cash flows, CF
ǫ̃, which is independent of these cash flows. Moreover, the noise is normally
distributed with mean zero and variance σǫ2 .
g + ǫ̃
ẽ = aCF
ǫ̃ ∼ N (0, σǫ2 )
(9)
g , ǫ̃) = 0
cov(CF
As it is assumed that the shareholders use Bayes’ rule to update their expectations about the present value of the unknown random future cash flows, the
price of the firm after the announcement of the signal is represented by Equation
10.
g |e1 ) =
P1 = E(CF
e1 /a
σǫ2
1
σǫ2
+
+
f)
E(CF
a2 σ 2
1
a2 σ 2
(10)
g ). For further simplifiConsidering Equation 9, it follows that E(ẽ) = aE(CF
cation, the scale factor, a, is set to 1.
P1 − P0 =
σ2
σ2
(e1 − E(e˜1 ))
+ σǫ2
30
(11)
In summary, Equation 11 highlights that the change in the stock price is determined by the earnings signal, e1 , and its expectation, E(e˜1 ). In addition, the
change in the stock price also depends on variances in the value of the company
and the earnings noise—the ERC, σ 2 /(σ 2 + σǫ2 ).
∂ERC
>0
∂σ 2
∂ERC
<0
∂σǫ2
(12)
The variance σ 2 (σǫ2 ) has a positive (negative) influence on the ERC (see Equation 12). Notably, the EFRQ is represented by the variance in the noise term,
σǫ2 . Thus, higher quality means higher reliability and, so, a lower variance; it
results, ceteris paribus, in a higher ERC and, therefore, a greater price reaction.
However, what is behind the variance in the earnings noise or the EFRQ,
respectively? The noise, ǫ̃, represents perceived deficiencies in reported earnings
due, for instance, to insufficient reliability. So, it might be the case that the
financial statements do not correctly represent the substance of an economic
transaction (cf. FASB, 2010, Para. BC3.26); such a material misstatement results in a non-faithful or non-reliable, respectively, representation. Therefore,
we assume that the earnings noise, ǫ̃, consists of a random amount of material
misstatements, ω̃, multiplied by the risk that the audited financial statements
contain material misstatements, κ]
post . The latter term is further denoted as the
financial reporting risk after the audit. Thus, ǫ̃ can be interpreted as the total
amount of material misstatements that remain undetected. Related to the logic
of the Audit Risk Model (AICPA, 1984), the financial reporting risk after the
audit, κ]
post , consists of the financial reporting risk before the audit (which summarizes the inherent and control risks), κg
pre , and the auditor’s detection risk,
τ̃ .24 The expected value and the variance of the former express the financial reporting quality before the audit. Accordingly, the audit quality is characterized
by the auditor’s expected detection risk and its variance. Thus, both attributes
of quality have an impact on the variance in the financial reporting risk after
2
the audit, κ]
post , and, moreover, on σǫ , i.e., the EFRQ.
A simplified example can be described that allows a deeper understanding
24
Please also refer to Footnote 12 and Footnote 13, p. 9.
31
of the reasoning outlined above. It is assumed that the capital market forms its
expectation based only on the means of the financial reporting risk before the
audit, κpre , and the auditor’s detection risk, τ . Therefore, the variances are set
to zero. The constant expectation values are interpreted as the perceived quality
of the financial reporting before the audit and of the audit; the assumption is
made for simplicity. In fact, expectations should also be formed regarding the
variances. Nevertheless, with a sufficient number of audit areas, the variances
might be comparatively small and, thus, their impact—relative to the influence
of the means—might be marginal. Hence, in this simplified case, the capital
market decides based on a heuristic.
ǫ̃ = ω̃κpost
ω̃ ∼ N (0, σω2 )
(13)
Further, the financial reporting risk after the audit, κpost , depends solely on the
means of the financial reporting risk before the audit and the auditor’s detection
risk.
κpost = κpre τ
(14)
Thus, the variance of σǫ2 can be derived using Taylor’s series expansion
(Burt and Finley, 1968, 736–737).
σǫ2 = (κpre τ )2 σω2
(15)
Considering Equations 11 and 15, the price reaction formula is now defined as:
P1 − P0 =
σ2
σ2
(e1 − E(e˜1 ))
+ [(κpre τ )2 σω2 ]
(16)
Differentiating the ERC of Equation 16 with respect to the variables of interest
leads to the additional conclusions provided below.
∂ERC
<0
∂τ
∂ERC
<0
∂κpre
(17)
Equation 17 demonstrates that it can be claimed that the expected auditor’s
detection risk, τ , as well as the expected financial reporting risk before the
audit, κpre , are inversely related to the ERC. Consequently, the same causal
link applies for the financial reporting risk after the audit, κpost . One further
32
aspect to note is that investors can achieve any level of EFRQ with different
combinations of financial reporting quality before the audit and audit quality.
If it is assumed that financial reporting quality before the audit is constant in
the short-term, the audit quality demanded can be used as an instrument to
achieve a certain desired level of financial reporting risk after the audit and,
thus, a certain EFRQ.
33
Figures
Figure 1: Time Line of Shareholder Voting Date, Beginning of the Audit Process, Fiscal Year-End Date, and Earnings Announcement Date
TIMELAG
F Y Et−1
EADt−1
V Dt
voting decision
ASt
F Y Et
EADt
V Dt+1
investment decision
Note:
34
V Dt
Shareholder voting date for fiscal year t.
ASt
Beginning of the audit process for fiscal year t.
F Y Et
Fiscal year-end date of fiscal year t.
EADt
Earnings announcement date for fiscal year t.
TIMELAG Period measured in days between the date of the auditor ratification and the earnings announcement date.
ASt+1 F Y Et+1
Tables
Table 1: Variables and Definitions
Variable
CAR
SURP
VOTESFOR
FREEFLOAT
TIMELAG
MVE
MB
MBNEG
LEV
BETA
SURPNEG
ANALYST
LIT
Definition
Dependent Variable
A stock’s cumulative excess return over the Standard and Poor’s 500
Composite return aggregated over a 3-day window, i.e., -1 day to +1
day relative to the earnings announcement date. The calculation is
based on the market model estimated over the 180-day window ending
21 trading days before the earnings announcement date.
Variables of Interest
Reported earnings for a respective fiscal year minus the mean earnings
forecast for this fiscal year one week before the earnings announcement
date, scaled by the firm’s stock price one day before the event window.
Shareholders’ voting result in percentage terms with respect to the votes
for (supporting) the auditor’s engagement.
Percentage of total shares in issue available to ordinary investors, i.e.,
non-strategic holdings (percentage of total shares in issue less than 5%
held not strategically).
Period measured in days between the date of the auditor ratification
and the earnings announcement date.
Control Variables
Natural logarithm of the market value of equity.
Market-to-book value, defined as the market value of the common equity
divided by the balance sheet value of the common equity. Negative
values are replaced with zeros.
A dummy variable where one indicates a firm has a negative marketto-book value and zero otherwise.
Leverage, defined as long-term debt plus short-term debt plus current
portion of long-term debt divided by total capital plus short-term debt
plus the current portion of long-term debt.
Beta factor from the market model regression, which is calculated over
the 180-day window ending 21 days before the earnings announcement
date.
A dummy variable where one indicates that a firm has a negative earnings surprise (SURP) and zero otherwise.
Natural logarithm of one plus the number of earnings per share estimates made by analysts.
A dummy variable where one indicates a firm operates in a high litigation risk industry (SIC code 2833-2836, 3570-3577, 3600-3674, 52005961, 7370-7374 and 8731-8734) and zero otherwise.
35
Table 2: Sample Selection
FirmYears
Initial sample of SEC registrants with shareholder voting results for the ratification of
auditors for fiscal years 2010, 2011, 2012 or 2013 in Audit Analytics.
Less:
Firm-years with more than one shareholder voting (date) for the ratification of an auditor
in a respective fiscal year.
15,703
Firm-years with no data regarding VOTESFOR and TIMELAG in Audit Analytics.
5,083
196
10,424
Less:
Firm-years with no data in Datastream.
130
10,294
Less:
Firm-years with no data in Worldscope.
6
10,288
Less:
Firm-years with no data in I/B/E/S.
2,043
8,245
Less:
Firm-years with inconsistent data; e.g., a negative time lag between the voting date and
the earnings announcement date or filing date.
134
8,111
Less:
Firm-years with TIMELAG greater than 365 days.
8
8,103
Less:
Firm-years referring to penny stocks, i.e., the price three days before the earnings announcement date is less than $1.
50
8,053
Less:
Firm-years with less than three analysts following.
1,125
Final sample
6,928
Note: This table outlines the sample selection procedure. Variable definitions: VOTESFOR represents shareholders’ voting result in percentage terms with respect to the votes for (supporting) the auditor’s engagement.
TIMELAG denotes the period measured in days between the date of the auditor ratification and the earnings
announcement date.
36
Table 3: Summary Statistics
Mean
Std.
Dev.
25%
50%
75%
Min.
Max.
CAR
0.0007
SURP
-0.0001
VOTESFOR 0.9830
FREEFLOAT 0.8049
TIMELAG 270.1817
MVE
21.1910
MB
3.1936
MBNEG
0.0292
LEV
0.3513
BETA
1.2079
SURPNEG
0.3492
ANALYST
2.3540
LIT
0.2330
0.0679
0.0128
0.0221
0.1482
24.1647
1.6399
3.9674
0.1683
0.2874
0.4268
0.4767
0.6122
0.4228
-0.0324
-0.0012
0.9803
0.7300
259.0000
20.0127
1.2500
0.0000
0.1113
0.9103
0.0000
1.7918
0.0000
0.0012
0.0004
0.9897
0.8300
273.0000
21.1091
1.9600
0.0000
0.3361
1.1665
0.0000
2.3026
0.0000
0.0353
0.0024
0.9953
0.9200
286.0000
22.2506
3.4400
0.0000
0.5216
1.4802
1.0000
2.8332
0.0000
-0.2161
-0.0708
0.8580
0.3100
158.0000
17.7414
0.0000
0.0000
0.0000
0.2860
0.0000
1.3863
0.0000
0.1957
0.0482
0.9999
1.0000
315.0000
25.4702
27.1400
1.0000
1.4456
2.3777
1.0000
3.6376
1.0000
n
6, 928
Note: This table presents the summary statistics for the pooled data for fiscal years 2010 to 2013. Variable
definitions: CAR represents a stock’s cumulative excess return over the Standard and Poor’s 500 Composite
return aggregated over a 3-day window, i.e., -1 day to +1 day relative to the earnings announcement date.
The calculation is based on the market model estimated over the 180-day window ending 21 trading days
before the earnings announcement date. SURP equals reported earnings for a respective fiscal year minus
the mean earnings forecast for this fiscal year one week before the earnings announcement date, scaled by the
firm’s stock price one day before the event window. VOTESFOR represents shareholders’ voting result in
percentage terms with respect to the votes for (supporting) the auditor’s engagement. FREEFLOAT is the
percentage of total shares in issue available to ordinary investors, i.e., non-strategic holdings (percentage
of total shares in issue less than 5% held not strategically). TIMELAG denotes the period measured in
days between the date of the auditor ratification and the earnings announcement date. MVE is the natural
logarithm of the market value of equity. MB equals market-to-book value, defined as the market value of the
common equity divided by the balance sheet value of the common equity. Negative values are replaced with
zeros. MBNEG is a dummy variable where one indicates a firm has a negative market-to-book value and zero
otherwise. LEV represents leverage, defined as long-term debt plus short-term debt plus current portion
of long-term debt divided by total capital plus short-term debt plus the current portion of long-term debt.
BETA is the beta factor from the market model regression, which is calculated over the 180-day window
ending 21 days before the earnings announcement date. SURPNEG is a dummy variable where one indicates
that a firm has a negative earnings surprise (SURP) and zero otherwise. ANALYST is the natural logarithm
of one plus the number of earnings per share estimates made by analysts. LIT is a dummy variable where
one indicates a firm operates in a high litigation risk industry (SIC code 2833-2836, 3570-3577, 3600-3674,
5200-5961, 7370-7374 and 8731-8734) and zero otherwise.
37
Table 4: Pearson Product-Moment Correlation Coefficients
CAR
CAR
SURP
VOTESFOR
FREEFLOAT
TIMELAG
MVE
MB
MBNEG
LEV
38
BETA
SURPNEG
ANALYST
LIT
SURP
VOTESFOR
FREEFLOAT
TIMELAG
MVE
MB
MBNEG
LEV
BETA
SURPNEG
ANALYST
LIT
1.000
0.166
(0.000)
-0.005
(0.701)
0.005
(0.694)
-0.031
(0.010)
0.016
(0.194)
0.004
(0.765)
-0.008
(0.494)
0.002
(0.890)
-0.004
(0.770)
-0.242
(0.000)
0.010
(0.426)
0.002
(0.862)
1.000
0.021
(0.079)
0.006
(0.612)
-0.004
(0.720)
0.059
(0.000)
-0.006
(0.605)
-0.002
(0.857)
-0.048
(0.000)
-0.003
(0.829)
-0.491
(0.000)
0.040
(0.001)
0.048
(0.000)
1.000
-0.139
(0.000)
-0.014
(0.238)
0.021
(0.077)
0.002
(0.860)
0.020
(0.096)
0.012
(0.300)
-0.023
(0.054)
-0.015
(0.220)
-0.001
(0.905)
-0.007
(0.542)
1.000
-0.010
(0.429)
0.241
(0.000)
-0.050
(0.000)
0.001
(0.954)
0.061
(0.000)
-0.062
(0.000)
-0.006
(0.607)
0.235
(0.000)
-0.114
(0.000)
1.000
-0.027
(0.026)
-0.048
(0.000)
0.009
(0.445)
0.085
(0.000)
0.028
(0.018)
0.037
(0.002)
-0.059
(0.000)
-0.127
(0.000)
1.000
0.085
(0.000)
-0.048
(0.000)
0.150
(0.000)
-0.128
(0.000)
-0.103
(0.000)
0.740
(0.000)
-0.111
(0.000)
1.000
-0.140
(0.000)
0.020
(0.097)
0.014
(0.236)
-0.015
(0.218)
0.077
(0.000)
0.202
(0.000)
1.000
0.460
(0.000)
0.024
(0.049)
0.021
(0.086)
-0.019
(0.112)
0.059
(0.000)
1.000
-0.003
(0.796)
0.065
(0.000)
0.094
(0.000)
-0.214
(0.000)
1.000
0.038
(0.002)
-0.016
(0.178)
0.008
(0.485)
1.000
-0.083
(0.000)
-0.071
(0.000)
1.000
-0.004
(0.712)
1.000
Note: This table presents the Pearson product-moment correlation coefficients for the pooled data for fiscal years 2010 to 2013. The numbers in parentheses below the correlation coefficients indicate p-values
(two-tailed test). Variable definitions: CAR represents a stock’s cumulative excess return over the Standard and Poor’s 500 Composite return aggregated over a 3-day window, i.e., -1 day to +1 day
relative to the earnings announcement date. The calculation is based on the market model estimated over the 180-day window ending 21 trading days before the earnings announcement date. SURP
equals reported earnings for a respective fiscal year minus the mean earnings forecast for this fiscal year one week before the earnings announcement date, scaled by the firm’s stock price one day before
the event window. VOTESFOR represents shareholders’ voting result in percentage terms with respect to the votes for (supporting) the auditor’s engagement. FREEFLOAT is the percentage of total
shares in issue available to ordinary investors, i.e., non-strategic holdings (percentage of total shares in issue less than 5% held not strategically). TIMELAG denotes the period measured in days between
the date of the auditor ratification and the earnings announcement date. MVE is the natural logarithm of the market value of equity. MB equals market-to-book value, defined as the market value of the
common equity divided by the balance sheet value of the common equity. Negative values are replaced with zeros. MBNEG is a dummy variable where one indicates a firm has a negative market-to-book
value and zero otherwise. LEV represents leverage, defined as long-term debt plus short-term debt plus current portion of long-term debt divided by total capital plus short-term debt plus the current
portion of long-term debt. BETA is the beta factor from the market model regression, which is calculated over the 180-day window ending 21 days before the earnings announcement date. SURPNEG is
a dummy variable where one indicates that a firm has a negative earnings surprise (SURP) and zero otherwise. ANALYST is the natural logarithm of one plus the number of earnings per share estimates
made by analysts. LIT is a dummy variable where one indicates a firm operates in a high litigation risk industry (SIC code 2833-2836, 3570-3577, 3600-3674, 5200-5961, 7370-7374 and 8731-8734) and
zero otherwise.
Table 5: OLS Regression—Hypothesis 1
H1
Variable
Coefficient
Robust
Standard
Error
p-value
-5.9131
-0.0075
-0.0002
0.0001
-0.0073
0.0054
-0.0002
-0.0294
-0.0006
-0.0019
6.4755
-0.0125
0.0094
0.1023
-0.4933
0.0212
-0.2966
0.2312
0.0069
0.0261
2.6002
0.0393
0.0008
0.0003
0.0068
0.0038
0.0021
0.0020
0.0021
0.0025
2.4854
0.0761
0.0206
0.4191
0.3199
0.1726
0.2148
0.2302
0.2623
0.0414
0.023
0.849
0.843
0.767
0.282
0.154
0.929
0.000
0.777
0.444
0.009
0.869
0.650
0.807
0.123
0.902
0.167
0.315
0.979
0.528
SURP
VOTESFOR
MVE
MB
MBNEG
LEV
BETA
SURPNEG
ANALYST
LIT
VOTESFOR×SURP
MVE×SURP
MB×SURP
MBNEG×SURP
LEV×SURP
BETA×SURP
SURPNEG×SURP
ANALYST×SURP
LIT×SURP
INTERCEPT
Year Dummies:
Clustered by:
Yes
Firm
n
Adjusted R2
F -stat
P rob > F
6,928
0.065
17.47
0.0000
Note: This table presents the results of a cross-sectional OLS regressions based on the pooled data
for fiscal years 2010 to 2013. Standard errors and t-statistics are adjusted and clustered by firm.
p-values are based on two-tailed tests.
CAR
=
For Hypothesis 1 the following regression model is tested:
α0 + α1 SU RP + α2 V OT ESF OR + α3 M V E + α4 M B + α5 M BN EG + α6 LEV +
α7 BET A + α8 SU RP N EG + α9 AN ALY ST + α10 LIT + α11 V OT ESF OR × SU RP + α12 M V E ×
SU RP + α13 M B × SU RP + α14 M BN EG × SU RP + α15 LEV × SU RP + α16 BET A × SU RP +
α17 SU RP N EG × SU RP + α18 AN ALY ST × SU RP + α19 LIT × SU RP + α20 -α22 Y EAR20XX +
α23 -α25 Y EAR20XX × SU RP + ε, where CAR represents a stock’s cumulative excess return over the
Standard and Poor’s 500 Composite return aggregated over a 3-day window, i.e., -1 day to +1 day
relative to the earnings announcement date. The calculation is based on the market model estimated
over the 180-day window ending 21 trading days before the earnings announcement date. SURP equals
reported earnings for a respective fiscal year minus the mean earnings forecast for this fiscal year one
week before the earnings announcement date, scaled by the firm’s stock price one day before the
event window. VOTESFOR represents shareholders’ voting result in percentage terms with respect
to the votes for (supporting) the auditor’s engagement. MVE is the natural logarithm of the market
value of equity. MB equals market-to-book value, defined as the market value of the common equity
divided by the balance sheet value of the common equity. Negative values are replaced with zeros.
MBNEG is a dummy variable where one indicates a firm has a negative market-to-book value and
zero otherwise. LEV represents leverage, defined as long-term debt plus short-term debt plus current
portion of long-term debt divided by total capital plus short-term debt plus the current portion of
long-term debt. BETA is the beta factor from the market model regression, which is calculated over
the 180-day window ending 21 days before the earnings announcement date. SURPNEG is a dummy
variable where one indicates that a firm has a negative earnings surprise (SURP) and zero otherwise.
ANALYST is the natural logarithm of one plus the number of earnings per share estimates made
by analysts. LIT is a dummy variable where one indicates a firm operates in a high litigation risk
industry (SIC code 2833-2836, 3570-3577, 3600-3674, 5200-5961, 7370-7374 and 8731-8734) and zero
otherwise.
39
Table 6: OLS Regression—Hypotheses 2 & 3
H2 & H3
Variable
Coefficient
Robust
Standard
Error
p-value
82.4888
0.5342
-0.0155
0.0020
-0.0002
0.0001
-0.0076
0.0057
-0.0003
-0.0294
-0.0007
-0.0026
0.0161
-0.0021
-80.5371
-31.3680
-0.2190
-0.0389
0.0073
0.1856
-0.5083
0.0154
-0.3391
0.3414
-0.0086
30.1684
0.2186
-0.4859
29.4483
0.5031
0.3308
0.0014
0.0008
0.0003
0.0067
0.0038
0.0021
0.0020
0.0021
0.0025
0.3345
0.0014
30.1522
15.6011
0.0855
0.0764
0.0202
0.4063
0.3099
0.1669
0.2147
0.2313
0.2626
15.9710
0.0878
0.4953
0.005
0.288
0.963
0.155
0.845
0.822
0.258
0.129
0.903
0.000
0.734
0.303
0.962
0.139
0.008
0.044
0.010
0.611
0.719
0.648
0.101
0.927
0.114
0.140
0.974
0.059
0.013
0.327
SURP
VOTESFOR
FREEFLOAT
TIMELAG
MVE
MB
MBNEG
LEV
BETA
SURPNEG
ANALYST
LIT
VOTESFOR×FREEFLOAT
VOTESFOR×TIMELAG
VOTESFOR×SURP
FREEFLOAT×SURP
TIMELAG×SURP
MVE×SURP
MB×SURP
MBNEG×SURP
LEV×SURP
BETA×SURP
SURPNEG×SURP
ANALYST×SURP
LIT×SURP
VOTESFOR×FREEFLOAT×SURP
VOTESFOR×TIMELAG×SURP
INTERCEPT
Year Dummies:
Clustered by:
Yes
Firm
n
Adjusted R2
F -stat
P rob > F
6,928
0.069
14.40
0.0000
Note: This table presents the results of a cross-sectional OLS regressions based on the pooled data for
fiscal years 2010 to 2013. Standard errors and t-statistics are adjusted and clustered by firm. pvalues are based on two-tailed tests. For Hypotheses 2 and 3 the following regression model is tested:
CAR = β0 + β1 SU RP + β2 V OT ESF OR + β3 F REEF LOAT + β4 T IM ELAG + β5 M V E + β6 M B +
β7 M BN EG+β8 LEV +β9 BET A+β10 SU RP N EG+β11 AN ALY ST +β12 LIT +β13 V OT ESF OR×
F REEF LOAT +β14 V OT ESF OR×T IM ELAG+β15 V OT ESF OR×SU RP +β16 F REEF LOAT ×
SU RP + β17 T IM ELAG × SU RP + β18 M V E × SU RP + β19 M B × SU RP + β20 M BN EG × SU RP +
β21 LEV × SU RP + β22 BET A × SU RP + β23 SU RP N EG × SU RP + β24 AN ALY ST × SU RP +
β25 LIT × SU RP + β26 V OT ESF OR × F REEF LOAT × SU RP + β27 V OT ESF OR × T IM ELAG ×
SU RP + β28 -β30 Y EAR20XX + β31 -β33 Y EAR20XX × SU RP + ε, where CAR represents a stock’s
cumulative excess return over the Standard and Poor’s 500 Composite return aggregated over a 3day window, i.e., -1 day to +1 day relative to the earnings announcement date. The calculation is
based on the market model estimated over the 180-day window ending 21 trading days before the
earnings announcement date. SURP equals reported earnings for a respective fiscal year minus the
mean earnings forecast for this fiscal year one week before the earnings announcement date, scaled
by the firm’s stock price one day before the event window.
VOTESFOR represents shareholders’
voting result in percentage terms with respect to the votes for (supporting) the auditor’s engagement.
FREEFLOAT is the percentage of total shares in issue available to ordinary investors, i.e., nonstrategic holdings (percentage of total shares in issue less than 5% held not strategically). TIMELAG
denotes the period measured in days between the date of the auditor ratification and the earnings
announcement date. MVE is the natural logarithm of the market value of equity. MB equals marketto-book value, defined as the market value of the common equity divided by the balance sheet value of
the common equity. Negative values are replaced with zeros. MBNEG is a dummy variable where one
indicates a firm has a negative market-to-book value and zero otherwise. LEV represents leverage,
defined as long-term debt plus short-term debt plus current portion of long-term debt divided by
total capital plus short-term debt plus the current portion of long-term debt.
BETA is the beta
factor from the market model regression, which is calculated over the 180-day window ending 21 days
before the earnings announcement date. SURPNEG is a dummy variable where one indicates that a
firm has a negative earnings surprise (SURP) and zero otherwise. ANALYST is the natural logarithm
of one plus the number of earnings per share estimates made by analysts. LIT is a dummy variable
where one indicates a firm operates in a high litigation risk industry (SIC code 2833-2836, 3570-3577,
3600-3674, 5200-5961, 7370-7374 and 8731-8734) and zero otherwise.
40