Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231

Contents lists available at ScienceDirect

Advances in Accounting, incorporating Advances in

International Accounting
journal homepage: www.elsevier.com/locate/adiac

Subjectivity in fair-value estimates, audit quality, and informativeness

of other comprehensive income☆,☆☆
Cheol Lee a,⁎, Myung S. Park b,1
a
Department of Accounting, School of Business Administration, Wayne State University, 207 Rands House, 5201 Cass Ave., Detroit, MI 48202, United States
b
Department of Accounting, School of Business, Virginia Commonwealth University, Snead Hall, 301W. Main Street, Box 844000, Richmond, VA 23284, United States

a r t i c l e i n f o a b s t r a c t

JEL classification: This study empirically examines whether difference in audit quality is reflected in the pricing of other com-
M41 prehensive income (OCI). Specifically, we first investigate whether OCI measures of Big 4 clients are more
M42 value-relevant than those of non-Big 4 clients. Considering different degrees of subjective management judg-
ment involved in the OCI reporting process, we then explore whether the differential valuation effect of OCI
Keywords:
between Big 4 and non-Big 4 clients is more pronounced for more subjective OCI components (e.g., minimum
Other comprehensive income
pension liability and foreign currency-translation adjustment) than a less subjective component (e.g., mar-
Audit quality
Subjectivity in fair-value estimate ketable securities adjustment). We predict that the aggregate OCI of a Big 4 client is more value-relevant
than that of a non-Big 4 client. We also hypothesize that the differential valuation effect between Big 4
and non-Big 4 clients can be attributed to the amount of subjective assumption and judgment required in es-
timating OCI. Consistent with our predictions, we find that aggregate OCI audited by a Big 4 auditor has
incremental information content over earnings, compared to OCI audited by a non-Big 4 auditor. More inter-
estingly, our results also show that the differential valuation effect between Big 4 and non-Big 4 clients is
stronger for OCI components of a more subjective nature. Our results are robust even after controlling for
self-selection bias, the potential effect of the financial crisis, and other related effects.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction The current financial reporting environment raises concerns about

The Financial Accounting Standards Board (FASB) requires firms to
report comprehensive income as the sum of net income and other com-
prehensive income (OCI) (see Accounting Standard Codification (ASC)
220-10-45, pre-codification Statement of Financial Accounting Stan-
dards (SFAS) No. 130). This requirement raises debates about the infor-
mativeness of such items (e.g., Biddle & Choi, 2006; Chambers,
Linsmeier, Shakespeare, & Sougiannis, 2007; Dhaliwal, Subramanyam,
& Trezevant, 1999; Jones & Smith, 2011, among others). Moreover, the
reporting of OCI is getting more attention because of the convergence
of U.S. Generally Accepted Accounting Principles (GAAP) to Internation-
al Financial Reporting Standards (IFRS).2
☆ Data used in this study are available from public sources identified in the paper.
☆☆ We appreciate the helpful comments and suggestions from Philip Reckers
(editor), two anonymous reviewers, and participants at the Annual Congress of Euro-
pean Accounting Association and the 2013 Annual Meeting of American Accounting
Association.
⁎ Corresponding author. Tel.: +1 313 577 2242; fax: +1 313 577 2000.
E-mail addresses: cheol@wayne.edu (C. Lee), mspark@vcu.edu (M.S. Park).
1
Tel.: +1 804 828 3161; fax: +1 804 828 8884.
2
Under IFRS, the magnitude and scope of OCI reporting is greater than that provid-
ed under U.S. GAAP, by allowing firms to write-up temporary changes in the fair value
of assets and/or liabilities.
the determination of fair-value measures, creating additional chal-
lenges for auditors and users of financial information (see Christensen,
Glover, & Wood, 2012; Public Company Accounting Oversight Board
(PCAOB), 2007b, 2011; SEC, 2008). OCI includes unrealized holding
gains or losses on available-for-sale securities, gains or losses associated
with pensions or other postretirement benefits, adjustments of foreign-
currency translation, and gains and losses on derivative instruments
that are designated as cash-flow hedges. Since OCI typically includes
several “mark-to-market” types of adjustments, financial reporting of
OCI may be subject to management's assumptions and judgments.
Specifically, managers can exercise discretion over the classification,
timing, and choice of valuation methods, including key assumptions,
when they determine the fair-value estimates of components of OCI.
The Securities and Exchange Commission (SEC)'s Office of the Chief
Accountant and the staff of the FASB provide guidance to investors, pre-
parers, and auditors on the application of fair-value measurements in
the current market environment (see SEC, 2008). According to this
guidance, when an active market for a security does not exist, the use
of management estimates is acceptable. Using judgment and subjectiv-
ity in estimating fair-value estimates significantly affects the quality of
financial information because it may be subject to an inherent degree
of uncertainty, and as a result, actual results could differ materially
from the estimates. For instance, Dhaliwal et al. (1999) argue that

0882-6110/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.adiac.2013.05.003
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
some components of OCI may involve more subjective estimates, there-
by adding noise to financial reporting.
There is a widespread belief that because of the subjectivity
of fair-value estimates, managers have incentives for substantial
reporting judgment and discretion, and thus it is difficult for auditors
to challenge. Recently, the PCAOB (Fall of 2011) issues a detailed re-
port outlining certain high risk-audit areas and challenges that it
identified during its regular inspections of registered accounting
firms. Fair-value measurement is identified as one of the more signif-
icant high-risk audit areas in the PCAOB's report.3 Primary deficien-
cies identified by the PCAOB relating to the fair-value measurement
of financial instruments include (1) whether fair-value measure-
ments are determined using appropriate valuation methods; and
(2) the reasonableness of management's significant assumptions
used to measure fair value, such as discount rates, and credit loss ex-
pectations (see PCAOB, 2011). Especially, the PCAOB's report states
that the attention and additional scrutiny directed towards these
high-risk areas will be considerable. In addition, Mark Olson, the for-
mer chairman of the PCAOB, expresses concern about the challenges
for assurance services related to fair-value estimates, saying that
“The increased use of fair value accounting poses a challenge for audi-
tors and the PCAOB” (PCAOB, 2007a). Considering the nature of OCI,
which is mainly derived from fair-value application and subject to
managerial judgment and discretion, subjective fair-value estimates
of OCI should be of particular concern and scrutinized closely by
auditors. Specifically, more extensive audit work on subjective OCI es-
timates will be necessary for auditors, because a high degree of man-
agement judgment and subjectivity may present risks of material
misstatement, thereby increasing the audit risk (SAS No.99, 2002). In-
tensive scrutiny by auditors, in turn, reduces the effect of subjective
management judgment, thereby enhancing the quality of financial
reporting. Therefore, it is an interesting research question to deter-
mine whether the valuation effect of OCI between Big 4 and non-Big
4 clients is contingent upon the amount of subjectivity in professional
judgment over estimates of OCI measures.
In this study, we empirically examine whether differences in audit
quality are reflected in the pricing of OCI. Specifically, we first investi-
gate whether OCI items for Big 4 clients have incremental information
content in explaining stock returns compared to those of non-Big 4 cli-
ents. The basic premise of this study is that audit quality reflects a form
of service distinction that the capital market values differently. Given
that Big 4 auditors provide higher-quality assurance regarding the cred-
ibility of reported earnings (e.g., Khurana & Raman, 2004), we predict
that the OCI audited by a Big 4 auditor is more value-relevant than
that audited by a non-Big 4 auditor.
Based on the level of subjectivity and degree of management dis-
cretion involved in the OCI reporting process, we then explore wheth-
er the differential valuation effect of OCI between Big 4 and non-Big 4
clients is stronger for more subjective OCI components compared to a
less subjective component (e.g., marketable-securities adjustment).
Some OCI components, such as foreign-currency translation and min-
imum pension-liability adjustments may utilize more subjective
estimates than the marketable-securities adjustment (e.g., Dhaliwal
et al., 1999) and therefore require more assurance from auditors.
When an active market for a security exists, quoted market prices
are the best evidence of fair value and should be used as the basis
for measurement, because investors and creditors regularly rely on
those prices to make their decisions (PCAOB, 2008). The level of man-
agement judgment required in establishing the fair value of financial
instruments may be minimal when a quoted price in an active market
3
Specifically, some of the more significant high-risk audit areas identified by the
PCAOB’s report are: (1) fair-value measurements for financial instruments; (2) fair-
value measurement for non-financial assets; (3) impairment of goodwill, indefinite-
lived intangible assets, and other long-lived assets; and (4) revenue recognition, and
others (e.g., valuation of inventory, income taxes).
219
is available. Thus, we argue that among OCI components, marketable-
securities adjustments are less subject to management judgment and
assumption than other components of OCI.4
Given that perceived audit quality is valued by the capital market,
we predict that Big 4 auditors in general provide higher-quality audits
with respect to the amount of subjectivity and management judg-
ment in fair-value estimates compared to non-Big 4 auditors, for sev-
eral reasons. First, Big 4 auditors are more prone to litigation risk than
smaller auditors (see Becker, DeFond, Jiambalvo, & Subramanyam,
1998; Dye, 1993; Francis & Wang, 2008; Kim, Chung, & Firth, 2003).
Dye (1993) argues that auditors with more wealth at risk from litiga-
tion have greater incentive to issue accurate reports. It is also known
that Big 4 auditors are more likely to be sued in case of misstatements
because of their “deeper pockets” (Kim et al., 2003) and that audit
failures produce greater reputation losses, especially for Big 4 audi-
tors, because they possess greater reputational capital (Becker et al.,
1998). Second, Big 4 firms mitigate information asymmetry more
than non-Big 4 auditors. Prior studies (Francis, Maydew, & Sparks,
1999; Willenborg, 1999) document that, by providing higher quality
auditing services, Big 4 auditors help reduce information asymmetry
between shareholders and managers. Khurana and Raman (2004)
provide empirical evidence that the cost of equity capital is lower
for Big 4 clients than for non-Big 4 clients. Third, large accounting
firms are subject to a regular annual inspection by the PCAOB (see
PCAOB inspection rule 4003). Although four second-tier audit firms
(see 2010 PCAOB inspection report) are also inspected annually by
the PCAOB, Big 4 firms are more likely to be subject to the inspection
than non-Big 4 auditors, including the second-tier firms, because Big 4
audit firms have far more issuer clients than second-tier firms and a
higher number of issuers inspected.5 Closer monitoring by the PCAOB
induces large auditors to increase their audit effort and make more con-
servative decisions in their audit engagements. That is, large auditors
are likely to be more diligent and watchful (e.g., DeFond, 2010). Fourth,
Big 4 auditors have a greater ability to constrain questionable account-
ing decisions, because they in general have better audit technology,
superior knowledge, and a strong negotiation stance with clients in
terms of financial-statement requirements (DeFond & Jiambalvo,
1993; Francis et al., 1999).6 Finally, the market values a perceived differ-
ence in audit quality between Big 4 and non-Big 4 (Knechel, Naiker, &
Pacheco, 2007; Menon & Williams, 1991; Palmrose, 1988; Teoh &
Wong, 1993; Watts & Zimmerman, 1986).
We partition OCI components into two groups based on the degree
of subjective judgment: (1) a less subjective component—the change
in unrealized gains and losses on available-for-sale marketable securi-
ties (ΔSEC), and (2) more subjective components—the sum of other
OCI components, such as change in pension or other postretirement
benefits (ΔPEN), change in adjustments for foreign currency translation
(ΔFOR), and change in adjustments for derivative instruments that are
designated as cash-flow hedges (ΔDER).7 Following prior research
(e.g., Becker et al., 1998; Craswell, Francis, & Taylor, 1995; DeAngelo,
1981; Khurana & Raman, 2004), we rely on auditor type (i.e., Big N) as
4
Not all available-for-sale securities are classified as level 1 estimates. Some of such
financial instruments can be grouped as level 2 or level 3. This is even more severe for
financial institutions, because available-for-sale securities are often classified as level 1,
level 2, or level 3 estimates. In this study, we exclude financial institutions from our
sample to mitigate such a potential impact. We thank an anonymous reviewer for this
insight.
5
Church and Shefchik (2012) report that, in 2009 client portfolios, Big 4 firms have
many more issuer clients (6648) than second-tier firms (1432), and 267 issuers of Big
4 audit firms (compared to 104 issuers of the second-tier audit firms) are inspected in
2009.
6
DeFond and Jiambalvo (1993) find that Big 8 (now Big 4) auditors are more likely
to oppose questionable accounting methods that increase reported earnings.
7
We also examine each component of OCI separately and show the results. Details
are discussed in a later section.
220 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
a surrogate for audit quality.8 We also use auditor industry specializa-
tion as an alternative proxy for audit quality.
Consistent with our predictions, we find that aggregate OCI audited
by Big 4 auditors has incremental information content in explaining
stock returns relative to OCI audited by non-Big 4 auditors. This implies
that investors account for audit quality in their valuation processes for
reported OCI items. Interestingly, our results also show that the differ-
ential valuation effect of OCI between Big 4 and non-Big 4 clients is
more pronounced for more subjective OCI components, such as mini-
mum pension liability and foreign-currency translation adjustments,
compared to the less subjective OCI item, marketable-securities adjust-
ments. These results are robust after controlling for self-selection bias,
potential effect of financial crisis, and other related effects.
This study contributes to the literature in several ways. Our study
is the first to provide evidence that audit quality plays an important
role in determining the reporting quality of OCI. Extant empirical
research (e.g., Bamber, Jiang, Petroni, & Wang, 2010; Biddle & Choi,
2006; Chambers et al., 2007; Choi, Das, & Zang, 2007; Kanagaretnam,
Mathieu, & Shehata, 2009; Louis, 2003) reports that market prices infor-
mation embedded in OCI measures, which is consistent with the com-
prehensive income argument. Although the aforementioned studies
highlight the value-relevance of OCI, the effect of audit quality on the
usefulness of OCI remains unexplored. To the best of our knowledge,
there is little evidence on the valuation implication of OCI conditioned
on the degree of management subjectivity and judgment. Therefore,
by extending prior work, this study fills a gap in the prior literature. Fur-
thermore, by showing that our findings still hold even after controlling
for potential effects of the financial crisis, we add value to the literature
on fair-value audits. Evidence from this study can also help standard
setters, auditors, and regulators better understand the reporting impli-
cations for OCI items in light of the amount of subjectivity, management
judgment, and audit quality.
The remainder of this paper proceeds as follows. We discuss back-
ground and develop hypotheses in Section 2. In Section 3, we present
the research design. Section 4 discusses the empirical results. A sum-
mary and conclusion follow in Section 5.
2. Background and hypotheses development
2.1. Audit quality, market perception, and earnings quality
Audit quality is defined as the probability that the auditor will both
discover and report a breach in the client's accounting process
(DeAngelo, 1981). Investors are unable to directly observe audit quality
and determine whether the reported information is an unbiased indica-
tor of firms' financial performance. It is well-known that reputable audi-
tors perform higher-quality audits and better certify the reliability of
financial statements (e.g., Balvers, McDonald, & Miller, 1988; Beatty,
1989; Datar, Feltham, & Hughes, 1991; DeAngelo, 1981; Francis &
Wilson, 1988; Krishnan, 2003; Palmrose, 1988; Teoh & Wong, 1993).9
Prior studies (Knechel et al., 2007; Menon & Williams, 1991;
Palmrose, 1988) document that the market reacts differently to per-
ceived audit-quality differences. Menon and Williams (1991) find that
Big 8 clients get better pricing of stock issues compared to those hiring
non-Big 8 auditors. A number of prior studies (Eichenseher, Hagigi, &
Shields, 1989; Johnson & Lys, 1990; Knechel et al., 2007; Nichols &
Smith, 1983) also present evidence that the market response to an
8 tion adjustment is associated with a loss of value and that such a
Lawrence et al. (2011) measure audit quality using discretionary accruals, the ex
ante cost-of-equity capital, and analyst forecast accuracy. They find that there are no
differences in audit quality between Big 4 and non-Big 4 auditors. They conclude that
observed differences in quality are likely driven by client attributes, such as client size.
We conduct additional analyses after controlling for clients’ size and discuss the results
in the sensitivity analysis section.
9 11
Prior studies use auditor brand name to proxy for audit quality and examine the as-
sociation between brand name and earnings quality (e.g., Balsam et al., 2003; Becker et
al., 1998; Reynolds & Francis, 2000).
auditor change is positive (negative) when a client switches to (from)
a high-quality auditor because prestigious auditors offer better moni-
toring capabilities. Recently, Knechel et al. (2007) find that firms
switching between Big 4 auditors experience significant positive (nega-
tive) abnormal returns when the successor auditor is an industry spe-
cialist (not a specialist). They also find that such market reactions are
more likely to be caused by changes in perceived audit quality rather
than differential costs of using specialist auditors.
Numerous studies (e.g., Balsam, Krishnan, & Yang, 2003; Becker et
al., 1998; Francis et al., 1999; Reynolds & Francis, 2000; Teoh & Wong,
1993) find that audit quality is associated with earnings quality. For
instance, Becker et al. (1998) and Reynolds and Francis (2000) assert
that high-quality auditors (e.g., Big 6 auditors in their cases) are able
to detect earnings management better because of their superior
knowledge and act to constrain clients' opportunistic accounting de-
cisions. Specifically, these studies find that clients of high-quality au-
ditors (e.g., Big 4) have lower discretionary accruals than clients of
lower-quality auditors.
In addition, high-quality auditors can reduce the perceived uncer-
tainty and noise in reported earnings, thereby resulting in a higher
earnings-response coefficient (ERC). Holthausen and Verrecchia (1988)
contend that investors' responses to an earnings surprise depend on
the perceived credibility of the earnings report. In their theoretical
framework, Teoh and Wong (1993) predict a negative association be-
tween the noise in earnings and the ERC (see Eq. (5)). They also show
empirical evidence of higher ERCs for clients who are audited by Big 8
audit firms. Balsam et al. (2003) report that high-quality auditors,
proxied by industry specialization, result in a higher ERC than
lower-quality auditors. Hackenbrack and Hogan (2002) show that dis-
closures about reasons for auditor changes signal information about
earnings precision, leading to changes in ERC following auditor changes.
In sum, these studies suggest that auditors are an integral part of the
financial-reporting process and that perceived differences in audit qual-
ity explain variation in investors' responses to financial reports.
2.2. Value-relevance of OCI reporting
Prior research provides mixed evidence of the value-relevance of
OCI. With the exception of available-for-sale securities adjustments,
Dhaliwal et al. (1999) find little evidence that comprehensive income
is associated with stock returns. Their findings suggest that among
the components of OCI, only the marketable securities adjustment
improves the association between income and returns. 10 Using a
sample of U.K. firms, O'Hanlon and Pope (1999) report that investors
do not price OCI and its components consistently. In contrast, several
prior studies document that investors price OCI differently (e.g.,
Biddle & Choi, 2006; Carroll, Linsmeier, & Petroni, 2003; Chambers
et al., 2007; Choi et al., 2007; Kanagaretnam et al., 2009; Soo & Soo,
1994). For instance, Soo and Soo (1994) find that the market incorpo-
rates information about foreign-translation adjustments reported in
stockholders' equity. Biddle and Choi (2006) find that comprehensive
income dominates both traditional net income and fully comprehen-
sive income, measured by the change in retained earnings plus the
common stock dividend, in explaining stock returns. Chambers et al.
(2007) provide empirical evidence that as-reported OCI components,
such as unrealized gains and losses on marketable securities and
foreign-currency translation adjustments, are priced positively.11 In
addition, Louis (2003) finds that a positive foreign-currency transla-
10
Dhaliwal et al. (1999) use the return/earnings association approach to examine
whether adding the foreign-translation adjustment to net income increases the associ-
ation of earnings with returns. They find no evidence that adding the translation ad-
justment to net income affects the return/earnings association.
Chambers et al. (2007) claim that previous studies fail to find the value-relevance
of OCI items because they use an estimated as-if measure of OCI that is subject to mea-
surement error.
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
negative association between the translation adjustment and change
in value is largely attributable to those firms that are the most
labor-intensive.
Based on those observations, if OCI captures all sources of value
creation and investors incorporate audit quality in their valuation
processes, we expect OCI of a Big 4 client to show an incrementally
positive association with stock returns. Thus, given that OCI is priced
by the market (e.g., Biddle & Choi, 2006; Carroll et al., 2003;
Chambers et al., 2007; Louis, 2003), we posit the following hypothesis
in the alternative form:
H1. The aggregate OCI of a Big 4 client is more value-relevant than
that of a non-Big 4 client.
2.3. Subjectivity in OCI components
Under ASC 820 (“Fair value measurements and disclosures”), a cli-
ent must determine the appropriate level in the fair-value hierarchy
for each fair-value measurement from level 1 to level 3. The level 1 in-
puts are quoted prices in active markets for identical assets or liabili-
ties, and the level 3 inputs indicate unobservable inputs for the asset
or liability. Since different consequences associated with each of the
three levels of inputs may appear, managers have an incentive to in-
appropriately classify fair-value measurements within the hierarchy.
For example, an asset or liability with level-2 estimates is measured
using observable market inputs other than quoted prices, which
would be included in level 1. Thus, level-2 measurements might re-
quire management to make more subjective assumptions or judg-
ments than level-1 measurements (see PCAOB, 2007b). As a result,
the estimate of fair value tends to be highly subjective and requires
significant judgment regarding the estimate of the magnitude and
timing of future reporting of unit cash flows. The reporting of OCI is
subject to fair-value measurement because OCI items are in general
a “mark-to-market” type of adjustments, which are derived from
changes in interest rates, exchange rates, and other random walk pro-
cesses (Smithson, Smith, & Wilford, 1995).
2.3.1. Unrealized gains or losses on available-for-sale v securities
The fair value of equity securities, such as available-for-sale secu-
rities, is determined based on quoted prices in active marketplaces.
That is, the determination of fair value of marketable securities is
based on direct observations of transactions, rather than judgments
or assumptions, thereby providing superior reliability. Not all
available-for-sale securities are classified as level-1 estimates; some
can be grouped as either level 2 or level 3. Nonetheless, there is still
less room for management subjectivity and judgment in determining
the fair value of marketable securities compared to other components
of OCI, because an active market for a security exists, and quoted mar-
ket prices are usually available as evidence of fair value. Therefore, we
consider unrealized holding gains or losses on available-for-sale secu-
rities as a less subjective OCI item.
2.3.2. Foreign-currency translation adjustment
Foreign-currency translation adjustments require management's
judgment to determine the functional currency in which the foreign
entity primarily conducts its operations and generates/expends its
cash (ASC 830-10-45-6). Managers determine the functional currency
of the firm and its subsidiaries in foreign countries (IAS No. 21). De-
termination of functional currency depends on judgments about the
firm's current and future cash flows and could change in the future
if there is a significant change in the firm's economic environment.
Specifically, managers make the choice between the two translation
methods, temporal and current-rate methods, based on GAAP criteria.
Holt (2011) contends that managers generally manipulate the func-
tional currency choice. Pinto (2002) finds that managers opportunis-
tically determine whether the temporal method or the current-rate
221
method will be used for translation. Thus, the foreign-currency trans-
lation adjustment can be a subjective measure, thereby requiring
more assurance from auditors.
2.3.3. Pensions/other postretirement benefits
The provisions of ASC 820 (“Fair value measurements and disclo-
sures”) apply to the measurement and valuation of pensions/other
postretirement benefits plan assets in financial statements. It is well
known that option valuation models require highly subjective inputs.
Several assumptions are considered in determining a firm's pension
obligation, such as the discount rate, expected return on plan assets,
and rate of compensation increase. The expected rate of return on
plan assets is a particularly important element in the determination
of pension expense (PCAOB, 2008). Actuarial assumptions used to de-
termine future values of plan assets and liabilities are also subject to
management judgment. Accordingly, pension-related estimates are
likely to be classified as either level 2 or 3 inputs.
2.3.4. Derivative instruments
Derivatives depend on the value of other instruments or events.
Derivatives may be valued through the use of complex models and as-
sumptions (e.g., credit default probabilities) or the use of pricing
specialists. The assumptions used in these models can be highly sub-
jective, and even a slight difference in assumptions could result in sig-
nificant valuation discrepancies (PCAOB, 2008).12 In particular, the
measurement of the fair value of interest-rate-related derivative in-
struments is based on estimates of future interest rates, which may
change based on economic or other factors. When market data are
not available, key inputs to the fair-value measurement include
quotes from counterparties and other market data. Hence, a firm's de-
rivative fair-value measurements consider assumptions about coun-
terparty and its own non-performance risk, such as credit-default
probabilities. Changes in the fair value of these derivative instruments
are recorded either through current earnings or as other comprehen-
sive income, depending on the type of hedge designation. In particu-
lar, gains and losses on derivative instruments designated as cash-
flow hedges are reported in OCI and reclassified into earnings in
the periods during which earnings are impacted by the hedged item
(ASC 815). The assurance about derivatives may require the auditor
to use a considerable amount of effort and time, because the valuation
of derivatives is based on highly subjective assumptions or is par-
ticularly sensitive to changes in underlying circumstances (see AU
Section 332.41).
Based on our discussions above, we argue that the fair-value mea-
surement of marketable securities are less subjective but the
fair-value estimates of other OCI components are more subjective
(i.e., level 2 or 3 inputs).
2.4. Auditing of fair-value measurements for OCI
The PCAOB's auditing-practice standards provide guidance to
auditors on auditing fair-value and other accounting estimates.13
For instance, AU section 328, Auditing Fair Value Measurements and
Disclosures, applies to auditing fair-value measurements and disclo-
sures in financial statements. It states that the auditor should evaluate
whether the fair-value measurements and disclosures in the financial
statements are in conformity with GAAP.
During the audit, the auditor obtains an understanding of the
client's process for determining fair-value measurements and dis-
closures. The auditor should also evaluate whether management's
12
In September 2008, the FASB issued a staff position aimed at improving disclosures
about credit derivatives.
13
See AU section 328, Auditing Fair Value Measurements and Disclosures; AU section
332, Auditing Derivative Instruments; and AU section 342, Auditing Accounting
Estimates.
222 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
assumptions are reasonable, and whether the client's method for
determining fair-value measurements is applied consistently (see
PCAOB, 2007b).
Audit firms are subject to a regular annual inspection by the PCAOB
(see PCAOB inspection rule 4003), which is intended to identify the
quality of financial statements, including fair-value audit deficiencies.
Palmrose (2006) argues that closer monitoring by the PCAOB may in-
duce auditors to increase their audit effort and make more conservative
judgments and decisions in their audit engagements. Thus, considering
the subjective nature of the fair-value estimates of OCI, the role of audi-
tors in evaluating and assuring management's estimates and judgment
on OCI should also be a main concern to investors, because they reflect
the perceived audit-quality difference in valuing the firm (see Knechel
et al., 2007; Menon & Williams, 1991; Palmrose, 1988; Teoh & Wong,
1993).
We predict that Big 4 auditors in general provide higher-quality
audits regarding the audit of fair-value estimates compared to
non-Big 4 auditors, for several reasons. First, Big 4 auditors are likely
to be more concerned about litigation risk than smaller auditors. Big 4
auditors have a higher probability of being sued in case of misstate-
ments because of their “deeper pockets” and higher insurance cover-
age (Kim et al., 2003). In addition, auditors lose their reputation
capital when audit-market participants perceive that they have
allowed misreporting (Abbott, Gunny, & Zhang, 2008; Skinner &
Srinivasan, 2012; Weber, Willenborg, & Zhang, 2008). Reputation
capital is a primary factor that motivates auditors to provide
high-quality auditing (DeAngelo, 1981). Thus, audit failures are likely
to produce greater reputation losses, especially for Big 4 auditors, be-
cause they possess greater reputational capital (Becker et al., 1998,
p.8). Francis and Wang (2008) find that Big 4 auditors impose ac-
counting conservatism on their clients in order to mitigate their expo-
sure to litigation and reputation risks.
Second, Big 4 auditors tend to mitigate information asymmetry more
than non-Big 4 auditors. Prior studies (Francis et al., 1999; Willenborg,
1999) document that Big 4 auditors provide higher-quality auditing ser-
vices than non-Big 4 auditors and that such high-quality service enables
firms to decrease information asymmetry between shareholders
and managers. In a similar vein, Khurana and Raman (2004) report a
lower cost of equity capital for Big 4 clients compared to the clients of
non-Big 4 firms.
Third, large accounting firms are subject to a regular annual
inspection by the PCAOB (see PCAOB inspection rule 4003). The
PCAOB conducts inspections of registered public accounting firms as
required by the Sarbanes–Oxley Act of 2002. Specifically, the PCAOB
performs annual inspections of audit firms that regularly provide
their audit reports with respect to more than 100 issuers. Although
four second-tier audit firms (see 2010 PCAOB inspection report) are
also inspected annually by the PCAOB, Big 4 firms are more likely to
be subject to the inspection than non-Big 4 auditors, including the
second-tier firms, because Big 4 audit firms have far more issuer cli-
ents than second-tier firms and a higher number of issuers inspected.
Closer monitoring by the PCAOB induces large auditors to increase
their audit effort and make more conservative decisions in their
audit engagements. That is, large auditors are likely to be more dili-
gent and watchful (e.g., DeFond, 2010). In contrast, for smaller
auditing firms that audit public registrants, the PCAOB conducts in-
spections at least once every three years.
Fourth, Big 4 auditors have a greater ability to constrain question-
able accounting decisions, thus reducing uncertainty about reported
earnings.14 This is because Big 4 auditors in general have better
audit technology and superior knowledge, interpretations of GAAP
that reduce the scope for management subjectivity and a strong ne-
gotiation stance with clients in terms of requirements for financial
14
The term Big 4 to denote the four largest audit firms is referred to previously as the
Big 8, Big 6, or Big 5.
statements (DeFond & Jiambalvo, 1993; Francis et al., 1999). DeFond
and Jiambalvo (1993) find that Big 8 (now Big 4) auditors tend to op-
pose questionable accounting methods to increase reported earnings.
Francis et al. (1999) argue that hiring a Big 4 auditor by a client sig-
nals that the client's current and future earnings are likely to be less
subject to managers' discretion than earnings of non-Big 4 clients.
Finally, the market values the perceived difference in audit quality
between Big 4 and non-Big 4 auditors (Knechel et al., 2007; Menon &
Williams, 1991; Palmrose, 1988; Teoh & Wong, 1993; Watts &
Zimmerman, 1986). For instance, Menon and Williams (1991) find
that clients employing Big 8 auditors get better pricing of stock issues
compared to those hiring non-Big 8 auditors. In a similar vein, Teoh
and Wong (1993) find that Big 4 auditors are more likely to reduce
the perceived uncertainty and noise in reported earnings, thereby
resulting in a higher earnings response coefficient (ERC).15
Dhaliwal et al. (1999) assert that some components of OCI, such
as the foreign-currency translation and minimum pension liability
adjustments may bear more on subjective estimates than the
marketable securities adjustment and thus may be noisier than the
marketable-securities adjustment. Accordingly, these components re-
quire more assurance from auditors compared to marketable securi-
ties adjustments. That is, more extensive audit work on such
components is required for auditors, because a high degree of subjec-
tive assumptions, judgments, and estimates may present increased
risk of material misstatement, thereby increasing the audit risk (SAS
No.99, 2002). The auditor's greater scrutiny, in turn, reduces the effect
of management's subjective judgment, thereby enhancing the quality
of financial reporting. Therefore, given that the capital market values
audit quality, we would expect the differential valuation effect be-
tween Big 4 and non-Big 4 clients to be more pronounced for those
more subjective OCI components than for less subjective marketable
securities adjustments. We posit the following hypothesis in an alter-
native form:
H2. The differential valuation effect of OCI between Big 4 and non-Big
4 clients is stronger for more subjective OCI components than for
marketable securities adjustments.
3. Research design
3.1. Data and sample
We obtain our sample from three sources: (1) Compustat Funda-
mentals Annual for financial data, (2) CRSP Monthly Stock File for
stock price and computing returns, and (3) AuditAnalytics for auditor
information. Because Compustat only provides financial data for OCI
and components of OCI since 2001, our sample period covers eight
years, 2002–2009.16 First, we eliminate observations for which audi-
tor or audit-fee data are not available. We also exclude firms in the fi-
nancial services industry.17 To gather financial data, we require that
all firms have a non-zero value of OCI, non-negative book value of eq-
uity, and non-negative value of sales. Further, we require firms to
have return data over twelve consecutive months, eight months
15
Specifically, Teoh and Wong (1993) show empirical evidence of higher ERCs for cli-
ents who are audited by Big 8 audit firms. In addition, Balsam et al. (2003) report that
high-quality auditors, proxied by industry specialization, result in a higher ERC than
lower-quality auditors. Hackenbrack and Hogan (2002) show that disclosures about
reasons for auditor changes signal information about earnings precision, leading to
changes in ERC following auditor changes.
16
Since 2001, Compustat has reported financial data for OCI and components of OCI.
17
We delete observations that belong to the financial services industry for the fol-
lowing two reasons: First, operations of firms in the financial industry are not compa-
rable to those of firms in other industries. Second, the available-for-sale securities can
be primary holding assets of financial firms. Thus, to avoid the potential effects by any
particular industry or particular assets composition, we exclude financial firms. We,
however, conduct additional analysis including financial industry in our sample. We
find that the results are unchanged.
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231 223

before fiscal-year-end through four months after fiscal-year-end and Table 1

market value of equity at eight months before fiscal-year-end. Finally, Sample.

we truncate earnings and return variables at the top and bottom 0.5% Panel A: Sample selection
to mitigate the effects of extreme values, yielding 17,878 firm-year
Procedures Number
observations as the final sample.18 Panel A of Table 1 presents the of obs.
sample-selection process.
Firms with non-zero other comprehensive income on the Compustat 47,868
Panel B of Table 1 presents the sample distribution by the Annual Database for 2002–2009
two-digit SIC code industry. The most heavily represented industry Less firms with missing auditor information on the AuditAnalytics 6200
is manufacturing (52.76%, SIC codes 20–39), followed by Services Less firms in financial sectors (two-digit SIC code with 60–67) 9799
(19.75%, SIC codes 70–89), and Transportation, Communication, Elec- Less firms with missing 12-month buy and hold returns on the CRSP 10,245
Less firms with negative value of equity, non-positive value of sale, and 3040
tric, and Gas (10.75%, SIC codes 40–49). Panel C of Table 1 reports the missing variables in regression model
annual distribution of our sample observations. As shown, the sample Less firms with truncated at top and bottom 0.5% for variables of change 706
observations are equally distributed over the sample period. in net income, change in other comprehensive income and 12-month
holding returns.
Final sample 17,878
3.2. Components of OCI
Panel B: Sample distribution by industry
OCI includes several components. Following the FASB ASC Glossa-
Two-digit SIC Industry # of % of
ry, firms are required to report unrealized gains and losses on code Obs. sample
available-for-sale marketable securities (SEC), gains or losses associ-
01–09 Agriculture, Forestry, and Fishing 80 0.45%
ated with pension or other postretirement benefits (PEN), adjust- 10–14 Mining 978 5.48%
ments for foreign-currency translation (FOR), and gains and losses 15–17 Construction 182 1.02%
on derivative instruments that are designated as cash-flow hedges 20–39 Manufacturing 9416 52.76%
(DER).19 We compute OCI as the change in accumulated OCI 40–49 Transportation, Communication, 1919 10.75%
Electric, and Gas
(Compustat variable name—ACOMINC). Similarly, the components 50–51 Wholesale 626 3.62%
of OCI are also computed by the change in accumulated unrealized 52–59 Retail 1024 5.74%
gains or losses on available-for-sale securities (Compustat variable 70–89 Services 3525 19.75%
name—MSA), the change in the accumulated balance of minimum 91–99 Public Administration 78 0.44%
Total 17,878 100.00%
pension liabilities (Compustat variable name—AOCIPEN), the change
in the accumulated adjustment of foreign-currency translation Panel C: Sample distribution by year
(Compustat variable name—RECTA), and the change in accumulated
Year # of Obs. % of sample
derivatives' unrealized gains or losses (Compustat variable name—
2002 1931 10.80%
AOCIDERGL). Relying on these components, we examine both the ag-
2003 2104 11.77%
gregate measure of OCI and its individual components by replacing 2004 2238 12.52%
total OCI with its components. 2005 2376 13.29%
2006 2441 13.65%
3.3. Empirical models 2007 2433 13.61%
2008 2349 13.14%
2009 2006 11.22%
With respect to the information content of OCI, prior studies (e.g., Total 17,878 100.00%
Chambers et al., 2007; Choi et al., 2007; Dhaliwal et al., 1999) express
stock returns as a function of OCI and its components. To test our H1
and H2, we estimate the following OLS regression models:
ΔNSEC change in adjustments of OCI components other than
RET it ¼ a0 þ a1 ΔNIit þ a2 ΔOCI it þ a3 BIG4it þ a4 ΔNI  BIG4it available-for-sale securities, measured as the sum of change
ð1Þ
þa5 ΔOCI  BIG4it þ eit in the accumulated balance of minimum pension liability
(ΔPEN), plus change in accumulated adjustment of foreign
RET it ¼ a0 þ a1 ΔNI it þ a2 ΔSEC it þ a3 ΔNSEC it þ a4 BIG4it currency translation (ΔFOR), plus change in accumulated
þa5 ΔNI  BIG4it þ a6 ΔSEC  BIG4it þ a7 ΔNSEC  BIG4it þ eit ; derivatives unrealized gains or losses that are designated
ð2Þ as cash-flow hedges (ΔDER).

where, for firm i in year t:

All earnings-related variables are scaled by average total assets. In
Eq. (1), if ΔOCI of Big 4 clients provides more valuable information
RET buy-and-hold return measured over a one-year window
about firm value than that of non-Big 4 clients, we will observe a pos-
(eight months before to four months after the fiscal year
itive and significant estimate on a5 (i.e., a5 > 0). Similarly, in Eq. (2),
end);
if non-SEC components of OCI (ΔNSEC) assured by Big 4 auditors are
ΔNI change in net income excluding extraordinary items and
more value-relevant than the adjustment of available-for-sale securi-
discontinued operations;
ties (ΔSEC), a positive and significant estimate on a7 is predicted.
ΔOCI change in accumulated other comprehensive income;
BIG4 an indicator variable that takes a value of 1 if the auditor is a
4. Empirical results
Big 4, and 0 otherwise;
ΔSEC change in accumulated unrealized gains and losses on
4.1. Descriptive statistics
available-for-sale securities;

Panel A of Table 2 reports descriptive statistics on the variables for

18
the full sample. The mean and median values of RET (mean = 0.1702
Our sample excludes Arthur Andersen clients.
19
Pension accounting standards (ASC 715) require that when additional pension lia-
and median = 0.0476) show that firms, on average, experience pos-
bility exceeds the amount of unrecognized prior service cost, the excess amount must itive buy-and-hold returns during the sample period. The mean value
be reported as a direct reduction in OCI. of change in reported net income, ΔNI, is positive (mean = 0.0096),
224 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231

Table 2 Panel C of Table 2 presents RET by quartile of ΔOCI. We examine

Descriptive statistics. whether differences in RET between Big 4 and non-Big 4 clients ap-
Panel A: Full sample (N = 17,878) pear in the clients with higher OCI. We find that mean and median
differences in RET between Big 4 and non-Big 4 clients are significant
Variable Mean Median St. Dev. P1 Q1 Q3 P99
at the conventional levels for the higher OCI quartile groups (i.e., Q3
RET 0.1702 0.0476 0.7043 −0.8128 −0.2497 0.3930 3.0450
and Q4). Thus, the results suggest that auditor quality is more likely
ΔNI 0.0096 0.0070 0.1353 −0.3897 −0.0267 0.0374 0.5403
ΔOCI 0.0005 0.0004 0.0185 −0.0708 −0.0027 0.0059 0.0515 to be reflected in stock price for clients whose OCI changes are high.
ΔSEC −0.0001 0.0000 0.0030 −0.0109 0.0000 0.0000 0.0103 We observe similar results using the ΔNSEC quartile. Panel D of
ΔNSEC 0.0006 0.0001 0.0181 −0.0705 −0.0016 0.0051 0.0509 Table 2 reports RET by ΔNSEC quartile. The results show that clients
BIG4 0.8378 1.0000 0.3686 0.0000 1.0000 1.0000 1.0000 in higher ΔNSEC quartiles tend to experience a significant difference
ΝΙ −0.0019 0.0368 0.1774 −0.7242 −0.0212 0.0809 0.2825
SIZE 6.3617 6.3351 1.9268 2.0962 5.0523 7.6168 11.1112
in RET between Big 4 and non-Big 4 clients.
ΜΒ 2.9888 2.0834 3.2802 0.3392 1.3427 3.4266 17.4071 Table 3 presents Pearson correlation coefficients for selected vari-
LEV 0.1963 0.1697 0.1840 0.0000 0.0099 0.3211 0.6886 ables. RET is positively correlated with ΔOCI, ΔSEC, and ΔNSEC variables.
LOSS 0.3012 0.0000 0.4588 0.0000 0.0000 1.0000 1.0000 Other OCI components, ΔPEN, ΔFOR, and ΔDER, are also positively corre-
Panel B: Big 4 versus Non-Big 4
lated with the RET variable. This evidence suggests that the market pos-
itively values OCI and its components. Turning to control variables, we
Variable Big 4 Non-Big 4 Difference tests
observe that NI and SIZE (LOSS) are positively (negatively) correlated
(N =14,978) (N = 2,900)
with RET.
Mean Median Mean Median t-test Wilcoxon Z

RET 0.1777 0.0611 0.1312 −0.0216 3.14 *** 6.92 ***

4.2. Association between return and OCI conditioned on audit quality
ΔNI 0.0101 0.0074 0.0069 0.0050 0.98 3.01 ***
ΔOCI 0.0005 0.0004 0.0006 0.0003 0.18 1.37
ΔSEC -0.0001 0.0000 −0.0001 0.0000 0.28 0.84 To begin, we explore whether aggregate OCI audited by Big 4 au-
ΔNSEC 0.0006 0.0001 0.0007 0.0000 0.28 2.23 ** ditors is more value-relevant than that assured by non-Big 4 auditors
NI 0.0038 0.0388 −0.0314 0.0211 8.14 *** 9.82 ***
(i.e., H1). Table 4 reports the results of Eq. (1), with buy-and-hold re-
SIZE 6.7078 6.6735 4.5751 4.5409 69.06 *** 54.72 ***
MB 3.0619 2.1346 2.6120 1.7881 7.12 *** 13.08 ***
turn measured over a one-year window as the dependent variable.20
LEV 0.2047 0.1844 0.1528 0.0912 14.43 *** 14.67 *** In the first column, model (1), the coefficient on ΔOCI is positive
LOSS 0.2796 0.0001 0.4128 0.0000 13.44 *** 14.22 *** and significant at the 1-percent level (t-stat. = 18.34, p b 0.001),
indicating that the aggregate measure of OCI has incremental explan-
Panel C: OCI quartile—return difference
atory power over earnings in explaining stock returns. This is consis-
Quartile RET tent with findings in prior studies (e.g., Biddle & Choi, 2006; Carroll et
Big 4 Non-Big 4 Difference tests al., 2003; Chambers et al., 2007; Choi et al., 2007; Kanagaretnam et al.,
(N = 14,978) (N = 2900) 2009; Soo & Soo, 1994). In the second column, model (2) includes the
Mean Median Mean Median t-test Wilcoxon Z audit-quality variable, BIG4, and its interaction terms, ΔNI ∗ BIG4, and
ΔOCI ∗ BIG4. As shown, the coefficient on ΔNI ∗ BIG4 is positive and
Lowest 0.1269 0.0153 0.1208 −0.0154 0.20 1.65 *
Q2 0.1537 0.0309 0.1265 −0.0368 0.93 2.42 ** significant at the 10-percent level. This evidence indicates the differ-
Q3 0.1963 0.0702 0.1235 −0.0247 2.57 ** 4.22 *** ential information content of ΔNI between BIG 4 and non-BIG 4 cli-
Highest 0.2341 0.1231 0.1541 −0.0053 2.60 *** 5.63 *** ents. Interestingly, the coefficient on ΔOCI ∗ BIG4 is also positive and
Panel D: NSEC quartile—return difference significant at the 1-percent level (t-value = 3.16, p b 0.001), indicat-
ing that aggregate OCI of Big 4 clients is more value-relevant than that
Lowest 0.1112 0.0251 0.1528 −0.0116 1.30 0.81
Q2 0.1603 0.0183 0.1166 −0.0439 1.54 2.25 **
of non-Big 4 clients. This, in turn, implies that investors account for
Q3 0.1990 0.0963 0.0674 −0.0324 3.76 *** 5.22 *** the audit quality in their valuation of the OCI item. This result sup-
Highest 0.2405 0.1327 0.1693 −0.0019 2.23 ** 5.41 *** ports our H1.
All variables are deflated by the average total asset excepting for the variable of RET.
Variables are defined in Appendix A. In Panels B–D, significances of means and 4.3. Association between return and OCI conditioned on audit quality and
medians are evaluated based on the t-test and Wilcoxon test, respectively (p-values
for the t-statistic and Z-statistic are two-tailed). *, **, *** indicate statistical significance
level of subjectivity
at the 0.10, 0.05, and 0.01 levels (two-tailed), respectively.
Our H2 predicts that the differential valuation effect of OCI be-
tween Big 4 and non-Big 4 clients is more pronounced for subjective
OCI components than for the less-subjective marketable-securities
indicating that sample firms report a net income of approximately 1
adjustment. To test our hypothesis, we decompose our sample OCI
percent of average total assets. We also observe that ΔOCI is, on aver-
components into two groups based on the level of subjectivity and
age, 0.05% of average total assets. The results also show that while the
degree of management discretion: (1) less subjective component:
mean value of available-for-sale securities adjustment, ΔSEC, is nega-
change in unrealized gains and losses on available-for-sale market-
tive, mean and median values of non-securities components, ΔNSEC,
able securities (SEC); and (2) more subjective components: sum of
are positive (mean = 0.0006 and median = 0.0001). Approximately
other components, such as pension or other postretirement benefits
84% of our sample firms are audited by Big 4 auditors.
(PEN) adjustment, adjustments for foreign-currency translation (FOR)
Panel B of Table 2 presents descriptive statistics of Big 4 and
non-Big 4 clients. As shown, mean and median values of RET and 20
All reported t-statistics are computed using White’s (1980) heteroscedasticity-
ΔNI of Big 4 clients tend to be larger than those of non-Big 4 clients.
consistent standard error. Recently, Gow, Ormazabal, and Taylor (2010) claim that ex-
The median differences of these variables are statistically significant tant methods commonly used in accounting literature produce misspecified test statis-
at the one-percent level. With respect to OCI, the mean ΔOCI of Big tics (i.e., overstated t-statistics) because of the year-specific fixed effect that drives
4 clients is smaller than that of non-Big 4 clients. The difference in cross-sectional correlation and the firm fixed effect that drives the time-series correla-
mean value, however, is insignificant. We observe that mean value tion. To correct these problems, they suggest well-specified test statistics using a two-
way clustering approach, which is robust standard error clustered by firm and time. For
of ΔNSEC for non-Big 4 clients is larger than that of Big 4 clients. a robustness check, we also repeat our regression analyses using the two-way cluster-
The mean difference in ΔNSEC between Big 4 and non-Big 4 clients ing and calculate t-statics based on cluster-robust standard errors by firm and year.
is also insignificant. Untabulated results are qualitatively similar to our results reported in this study.
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231 225

Table 3
Correlation matrix for selected variables.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)

(1) RET
(2) ΔΝΙ 0.216
(3) ΔΟCI 0.139 0.046
(4) ΔSEC 0.039 0.005 0.303
(5) ΔNSEC 0.134 0.046 0.953 0.015
(6) ΔPEN 0.083 0.031 0.567 0.007 0.596
(7) ΔFOR 0.113 0.039 0.749 0.018 0.783 0.070
(8) ΔDER 0.017 −0.001 0.291 0.069 0.313 0.019 −0.002
(9) BIG4 0.022 0.006 −0.000 0.061 −0.004 −0.035 0.021 −0.004
(10) NI 0.091 0.301 0.009 0.141 0.007 −0.027 0.034 −0.011 0.073
(11) SIZE 0.058 0.024 0.008 0.115 0.001 −0.052 0.043 −0.008 0.408 0.368
(12) MB 0.007 0.002 −0.004 0.101 −0.004 0.001 0.001 −0.013 −0.015 −0.001 0.010
(13) LEV −0.005 −0.052 −0.038 -0.089 −0.042 −0.043 −0.007 −0.044 0.104 0.018 0.145 0.028
(14) LOSS −0.066 −0.208 −0.002 -0.003 −0.014 0.018 −0.037 0.010 −0.107 −0.671 −0.388 −0.001 −0.029

Pearson correlation coefficients are presented. The coefficients in bold are statistically significant at the 5% level (two-tailed test). All variables are defined in Appendix A.

and adjustment for derivative instruments that are designated as
cash-flow hedges (DER). As discussed earlier, we assume that some OCI
components, such as foreign-currency translation, minimum pension-
liability adjustment, and adjustment of derivatives, contain more sub-
jective management judgments and estimates (e.g., Dhaliwal et al.,
1999), thereby requiring more audit work and increased auditor scrutiny.
We include two variables, ΔSEC and ΔNSEC, as well as their interaction
terms with the Big 4 variable in our regression models. Table 5 reports
the results.
As shown in the first column, the coefficients on ΔSEC and ΔNSEC are
positive and significant at the 1-percent level. This indicates that the
market values OCI components. Interestingly, as presented in the sec-
ond column, the coefficient on ΔNSEC ∗ BIG4 is positive and significant
at the 1-percent level (t-value = 3.63, p b 0.001), indicating that
other OCI components of Big 4 clients have stronger valuation implica-
tions, compared to those of non-Big4 clients. In contrast, we find little
evidence on differential-valuation effect of the marketable securities
adjustment (ΔSEC ∗ BIG4) between Big 4 and non-Big 4 auditors.
The coefficient difference between ΔSEC ∗ BIG4 and ΔNSEC ∗ BIG4 is sta-
tistically significant at the 5-percent level. Next, we decompose ΔNSEC
into its individual components: ΔPEN, ΔFOR, and ΔDER. We repeat our
analysis by replacing ΔNSEC with these three components as indepen-
dent variables. The last column in Table 4 reports the results using
these individual components. As shown, the coefficient on ΔFOR ∗ BIG4
(ΔPEN ∗ BIG4) is positive and significant at the 1 (10)-percent level, in-
dicating that a Big 4 premium plays an important role in pricing, espe-
cially for adjustments of both pension liability and foreign-currency
translation.21
Furthermore, to delve into the differential valuation effect of OCI
components, we partition our sample into high- and low-ΔNSEC
groups based on median value of ΔNSEC. Panel B of Table 5 presents
the results by the level of ΔNSEC. In the last column, the coefficient
on ΔNSEC ∗ BIG4 is positive and significant at the 1-percent level
(t-value = 3.19, p b 0.001), indicating that the audit-quality differ-
ence appear in the client group with a high level of non-securities
OCI components. This is not the case for the Big 4 clients with a low
level of non-securities OCI components. Overall, our findings reported
in Table 5 suggest that the differential valuation effect of OCI compo-
nents between Big 4 and non-Big 4 clients can be attributed to the
amount of subjective judgment. This is consistent with our H2.

4.4. Control for potential self-selection bias

Table 4
Association between returns and other comprehensive income conditioned on audit
Theoretical studies (Datar et al., 1991; Titman & Trueman, 1986)
quality.
present models to explore clients' self-selection of auditors. Thus, it
Panel A: OLS Regression results of Big 4 vs. Non-Big 4 is widely accepted in principle that clients self-select their auditors
Dependent variable = RET (Chaney, Jeter, & Shivakumar, 2004). Chaney et al. (2004) contend
Variable Predicted sign Model 1 Model 2
that self-selection introduces a bias in standard OLS regressions.
Coefficient (t-statistic) Coefficient (t-statistic) Thus, to examine whether the results are sensitive to specifications
that consider the self-selection bias, we estimate Heckman's (1979)
Intercept ? 0.157*** 0.123***
(31.59) (9.35) two-stage procedure.22 In the first stage, we obtain consistent esti-
ΔNI + 1.129*** 0.946*** mates from a probit regression of an indicator variable, BIG4, on exog-
(18.56) (8.70) enous variables. These estimates are used to compute the inverse
ΔOCI + 4.904*** 2.762***
Mills ratios (IMR). Then, in the second stage, the RET equations are
(18.34) (3.69)
BIG4 ? 0.039***
estimated by OLS, including the inverse Mills ratio as an additional
(2.81)
ΔNI ∗ BIG4 + 0.244*
(1.87) 21
Chambers et al. (2007) document that unrealized gains and losses on available for sale
ΔOCI ∗ BIG4 + 2.533***
securities are valued at a greater than dollar-for-dollar compared to foreign-currency
(3.16)
translation, which is valued at dollar-to-dollar. Our results reported in Table 5 are consis-
Adjusted R2 6.64% 6.78%
tent with findings in Chambers et al. (2007). As shown in model 3 of Table 5, joint tests
N 17,878 17,878
reveal that combined coefficient of ΔSEC (i.e., ΔSEC + ΔSEC ∗ BIG4 = 11.747) is larger
RET is buy-and-hold return, measured over one-year window (eight months before to than that of ΔFOR (i.e., ΔFOR + ΔFOR ∗ BIG4 = 5.627). We confirm that the coefficient
four months after the fiscal year end).Other variables are defined in Appendix A. difference between these two variables is statistically significant at the 5-percent level
t-statistics are calculated using heteroscedasticity-consistent standard error. The vari- (p = 0.0165).
22
ables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom 0.5 percentile levels. Lawrence et al. (2011) estimate the inverse Mills ratio controlling for client size using
*, **, *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively, total assets, total sales, and market value variables. Following Lawrence et al. (2011), we
based on one-tailed tests if the coefficient has a predicted sign, and two-tailed tests also estimate the inverse Mills ratio controlling for client size, using various measures.
otherwise. We find that untabulated results remain unchanged.
226 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231

Table 5 explanatory variable. Including this ratio helps control for the likeli-
Association between returns and components of other comprehensive income condi- hood of self-selecting into the auditor. Specifically, we estimate the
tioned on audit quality.
following probit regression in the first stage:
Panel A: OLS regression results of Big 4 vs. Non-Big 4

Dependent variable = RET

BIG4 ¼ f ðACC SHORT; ACC LONG; LEV; PE; LOSSÞ ð3Þ

Variable Predicted Model 1 Model 2 Model 3

sign Coefficient Coefficient Coefficient
where
(t-statistics) (t-statistics) (t-statistics)
ACC_SHORT short-term accruals, measured as net change in working
Intercept ? 0.157*** 0.124*** 0.124***
(33.74) (9.46) (10.89) capital accruals, deflated by sales;
ΔNI + 1.129*** 0.938*** 0.935*** ACC_LONG long-term accruals, measured as the sum of depreciation,
(18.57) (8.63) (8.68) change in deferred tax, and change in deferred charges, de-
ΔSEC + 12.878*** 16.442*** 16.395*** flated by sales;
(6.82) (5.55) (5.53)
ΔNSEC + 4.667*** 2.151***
LEV debt-to-asset ratio, measured as long-term liabilities divided
(17.25) (2.80) by total assets;
ΔPEN + 2.364* PE the price earnings ratio; and
(1.73) LOSS an indicator variable that takes the value of 1 if current net
ΔFOR + 2.429***
income is less than 0, and 0 otherwise.
(2.66)
ΔDER + −0.703
(−0.19) Table 6 reports the results from the second-stage OLS regressions,
BIG4 ? 0.039*** 0.037*** including IMR as an additional control.
(2.74) (2.61)
As shown in the first column of Table 6, we find a positive and sig-
ΔNI ∗ BIG4 + 0.254* 0.255**
(1.95) (1.97) nificant coefficient on ΔOCI ∗ BIG4. In the second column, the coeffi-
ΔSEC ∗ BIG4 + −4.662 −4.648 cient on ΔNSEC ∗ BIG4 is positive and significant at the 1-percent
(−1.24) (−1.23) level (t-value = 3.61, p b 0.001). As shown in model (2), the coeffi-
ΔNSEC ∗ BIG4 + 2.972***
cient difference between ΔSEC ∗ BIG4 and ΔNSEC ∗ BIG4 is statistically
(3.63)
ΔPEN ∗ BIG4 + 2.612**
significant at the 5-percent level. In the last column, the results show
(1.78) that differential-valuation implication of OCI between Big 4 and
ΔFOR ∗ BIG4 + 3.198*** non-Big 4 clients appears in both the pension-liability adjustment
(3.21) (ΔPEN) and the foreign-currency translation adjustment (ΔFOR). To-
ΔDER ∗ BIG4 + 3.418
gether, these results are consistent with those previously reported,
(0.90)
Adjusted R 2
6.76% 6.92% 6.95% even after controlling for potential self-selection bias. Therefore,
N 17,878 17,878 17,878 self-selection bias does not pose a problem in our study.
Test of Coefficients:
ΔSEC ∗ BIG4 = ΔNSEC*BIG4 χ2 = 3.91(p = 0.0479)
4.5. Analysis based on the propensity-score matching model
Panel B: OLS regression results of Big 4 vs. Non-Big 4 conditioned on the level of
ΔNSEC Recently, Lawrence, Minutti-Meza, and Zhang (2011) utilize a
Dependent variable = RET propensity-score matching model to control for differences in client
characteristics across Big 4/non-Big 4 audit firms.23 They find that
Variable Predicted Low-ΔNSEC High-ΔNSEC
sign Coefficient Coefficient
the effects of Big 4 auditors are insignificantly different from those
(t-statistic) (t-statistic) of non-Big 4 auditors in terms of three audit-quality proxies (i.e., dis-
cretionary accruals, ex ante cost-of-equity capital, and analyst forecast
Intercept ? 0.167*** 0.115***
(2.64) (5.38) accuracy). They conclude that the Big 4 distinction may reflect client
ΔNI 0.942*** 0.937*** characteristics, specifically, client size.
(16.83) (6.17) To investigate whether the results using the propensity-score
ΔSEC + 14.704*** 17.676***
matching model are the same as those previously documented in our
(3.33) (4.85)
ΔΝSEC + 4.637*** 0.779 main analyses, we match Big 4 auditors to non-Big 4 auditors based on
(3.97) (0.60) the propensity scores (predicted probabilities). Specifically, the propen-
BIG4 + 0.010*** 0.035 sity scores are obtained from the Big 4 auditor choice model as follows:
(0.43) (1.47)
ΔNI ∗ BIG4 0.186 0.326**
BIG 4 ¼ f ðROA; SIZE; LEV; CURR; TURN; SIC2; YEARÞ ð4Þ
(1.03) (1.72)
ΔSEC ∗ BIG4 + −4.972 −3.008
(−0.90) (−0.63) where ROA = returns on earnings, measured as net income deflated by
ΔNSEC ∗ BIG4 + 0.973 4.615*** average total assets; SIZE = firm size, measured as natural logarithm of
(0.79) (3.19) total assets; LEV = debt-to-asset ratio, measured as long-term liabilities
Adjusted R2 7.21% 6.30%
divided by total assets; CURR = current ratio; TURN = turnover ratio,
N 8994 8884
measured by sales divided by lagged total assets; SIC2 = two-digit
In Panel A, RET is buy-and-hold return, measured over one-year window (eight months
SIC code industry dummy controlling for industry fixed effects; and
before to four months after the fiscal year end). ΔSEC is change in accumulated
unrealized gains and losses on available-for-sale securities. ΔΝSEC is change in OCI YEAR = a year dummy controlling for year fixed effects. We replicate
items other than AFS securities, measured as the sum of foreign currency translation our analyses using this propensity score-matching approach.
adjustment, minimum pension liability, and derivatives. Other variables are defined Table 7 reports the results based on the propensity score-
in Appendix A. t-statistics are calculated using heteroscedasticity-consistent standard matching model. We find that the propensity-score matched results
error. The variables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom 0.5 per-
are qualitatively similar to those previously reported, except for the
centile levels. In Panel B, Low (High)—ΔNSEC is constructed if firms belong to below
(above) median annual rank of ΔNSEC. *, **, *** indicate statistical significance at the result of minimum pension adjustment. Specifically, the results
0.10, 0.05, and 0.01 levels, respectively, based on one-tailed tests if the coefficient
has a predicted sign, and two-tailed tests otherwise. 23
Lawrence et al. (2011) also employ attribute-based matching model in their study.
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231 227

Table 6 Table 7
Two-stage regressions controlling for self-selection of auditors. Regression results based on propensity score matching approach.

Two-stage regression results by Big 4 versus Non-Big 4 Dependent variable = RET

Dependent variable = RET Variable Predicted Model 1 Model 2 Model 3

sign Coefficient Coefficient Coefficient
Variable Predicted Model 1 Model 2 Model 3
(t-statistics) (t-statistics) (t-statistics)
sign Coefficient Coefficient Coefficient
(t-statistics) (t-statistics) (t-statistics) Intercept ? 0.158*** 0.159*** 0.165***
(8.93) (9.00) (9.10)
Intercept ? 0.115*** 0.115*** 0.114***
ΔNI + 1.318*** 1.322*** 1.321***
(6.06) (6.09) (5.96)
(8.36) (8.40) (8.41)
ΔNI + 0.945*** 0.936*** 0.933***
ΔOCI + 2.699***
(8.69) (8.62) (8.70)
(2.17)
ΔOCI + 2.767***
ΔSEC + 16.123*** 15.956***
(3.69)
(3.90) (3.81)
ΔSEC + 16.448*** 16.402***
ΔNSEC + 1.995*
(5.55) (5.52)
(1.56)
ΔNSEC + 2.157***
ΔPEN + 8.027**
(2.81)
(1.86)
ΔPEN + 2.352**
ΔFOR + 1.024
(1.72)
(0.72)
ΔFOR + 2.443***
ΔDER + 4.055
(2.67)
(0.57)
ΔDER + −0.711
BIG4 ? 0.013 0.011 0.006
(−0.20)
(0.59) (0.53) (0.28)
BIG4 ? 0.045*** 0.044*** 0.043***
ΔNI ∗ BIG4 + −0.187 −0.190 −0.194
(2.62) (2.60) (2.53)
(−0.94) (−0.96) (−0.98)
ΔNI ∗ BIG4 + 0.244** 0.254** 0.255***
ΔOCI ∗ BIG4 + 3.564***
(1.87) (1.94) (1.97)
(2.06)
ΔOCI ∗ BIG4 + 2.767***
ΔSEC ∗ BIG4 + −8.865 −8.709
(3.69)
(−1.23) (−1.21)
ΔSEC ∗ BIG4 + −4.638 −4.621
ΔNSEC ∗ BIG4 + 4.503***
(−1.23) (−1.23)
(2.57)
ΔNSEC ∗ BIG4 + 2.959***
ΔPEN ∗ BIG4 + −1.636
(3.61)
(−0.52)
ΔPEN ∗ BIG4 + 2.596**
ΔFOR ∗ BIG4 + 5.048***
(1.77)
(2.52)
ΔFOR ∗ BIG4 + 3.415***
ΔDER ∗ BIG4 + 2.528
(3.20)
(1.05)
ΔDER ∗ BIG4 + 3.415
Adjusted R2 6.67% 6.79% 6.88%
(0.90)
5800 5800 5800
IMR ? 0.014 0.016 0.018
Test of Coefficient:
(0.57) (0.66) (0.72)
2 ΔSEC ∗ BIG4 = ΔNSEC ∗ BIG4 χ2 = 4.43 (p = 0.0353)
Adjusted R 6.77% 6.92% 6.95%
N 17,878 17,878 17,878 Big 4 auditors are matched to non-big 4 auditors based on the propensity scores (predicted
Test of Coefficients: probabilities) of engaging with Big 4 auditors. The propensity scores are obtained from the
ΔSEC ∗ BIG4 = ΔNSEC ∗ BIG4 χ2 = 3.87(p = 0.0491) auditor choice model as follows: BIG 4 = f(ROA, SIZE, LEV, CURR, TURN, SIC2, YEAR), where
ROA = returns on earnings, measured as net income deflated by average assets; SIZE =
IMR is the Inverse Mills ratio based on Heckman (1979) procedure obtained from
firm size, measured as natural logarithm of total assets, LEV = debt-to-asset ratio, mea-
logistic regression of the following auditor choice model: BIG 4 = f(ACC_SHORT,
sured as long-term liabilities divided by total assets; CURR = current ratio; TURN = turn-
ACC_LONG, LEV, PE, LOSS), where ACC_SHORT = short-term accruals, measured as net
over ratio measured by sales divided by lagged total assets; SIC2 is two-digit sic code
change in working capital accruals, deflated by sales; ACC_LONG = long-term accruals,
industry dummy controlling for industry fixed effect; and YEAR = a year dummy control-
measured as the sum of depreciation, change in deferred tax, and change in deferred
ling for year fixed effect. The t-statistics are calculated using heteroscedasticity-consistent
charges, deflated by sales; LEV = debt-to-asset ratio, measured as long-term liabilities
standard error. The variables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom 0.5
divided by total assets; PE = the price earnings ratio; and LOSS = an indicator variable
percentile levels. All other variables are defined in Appendix A.
that takes the value of 1 if current period net income is less than 0, and 0 otherwise.
*, **, *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively,
The t-statistics are calculated using heteroscedasticity-consistent standard error. The
based on one-tailed tests if the coefficient has a predicted sign, and two-tailed tests
variables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom 0.5 percentile levels.
otherwise.
All other variables are defined in Appendix A.
*, **, *** indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively,
based on one-tailed tests if the coefficient has a predicted sign, and two-tailed tests Hackenbrack and Hogan (2002) find that losses are significantly relat-
otherwise. ed to stock returns. Following these studies, to see if our results are
robust even after controlling for potentially correlated omitted vari-
show that the coefficients on ΔOCI ∗ BIG4 (in model 1), ΔNSEC ∗ BIG4 ables, we include the level of earnings (NI), firm size (SIZE), leverage
(in model 2), and ΔFOR ∗ BIG4 (in model 3) are all positive and signif- (LEV), market-to-book ratio (MB), loss indicator (LOSS), and year
icant at the conventional level. The coefficient on ΔPEN ∗ BIG4, how- dummy variables as control variables. In addition, we also include
ever, is negative and insignificant. ΔOTHEROCI to control for all other items included in OCI. These vari-
ables are defined in Table 8. Panel A of Table 8 presents the results.
4.6. Additional analyses Our tabulated results are robust even after taking into account poten-
tial confounding effects of additional control variables.24
4.6.1. Analysis controlling for other variables
Easton and Harris (1991) find that both the level and change in
earnings play a role in security valuation. Prior studies (Fama &
24
French, 1992; Khurana & Raman, 2004; Teoh & Wong, 1993) docu- We also attempt to control for firm-fixed effects as well as year-fixed effects. Spe-
cifically, to control for time-series and cross-sectional dependences, we utilize the two-
ment several variables affecting the earnings-return association, way clustering approach (Gow et al., 2010) and calculate cluster-robust standard er-
such as size, beta, leverage, and book-to-market equity as well as rors. The results remain unchanged even after controlling for both within-firm and
earnings variable (i.e., ratio of total earnings to market equity). within-year dependences.
228 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231

Table 8 Table 8 (continued)

Additional analyses. Dependent variable = RET
Dependent variable = RET Variable Predicted Model 1 Model 2 Model 3
Variable Predicted Model 1 sign Coefficient Coefficient Coefficient
Model 2 Model 3
sign Coefficient (t-statistics) (t-statistics) (t-statistics)
Coefficient Coefficient
(t-statistics) (t-statistics) (t-statistics) Panel B: Regression results excluding financial crisis periods
ΔOCI ∗ BIG4 + 1.742**
Panel A: Regressions results controlling for other variables
(1.73)
Intercept ? 0.021 0.024 0.029
ΔSEC ∗ BIG4 + −4.193 −4.356
(0.87) (0.99) (1.16)
(−0.84) (−0.87)
ΔNI + 0.947*** 0.941*** 0.907***
ΔNSEC ∗ BIG4 + 3.743***
(8.32) (8.13) (8.15)
(3.34)
ΔOCI + 2.685***
ΔPEN ∗ BIG4 + 3.459**
(3.95)
(1.79)
ΔSEC + 17.407*** 17.357***
ΔFOR ∗ BIG4 + 3.941***
(5.04) (5.01)
(2.61)
ΔNSEC + 1.776***
ΔDER ∗ BIG4 + 1.234
(2.20)
(0.26)
ΔPEN + 4.091***
Adjusted R2 4.61% 4.71% 4.81%
(2.56)
N 13,523 13,523 13,523
ΔFOR + 1.898**
Test of Coefficients:
(1.73)
ΔSEC ∗ BIG4 = ΔNSEC ∗ BIG4 χ2 = 5.96 (p = 0.0147)
ΔDER + 11.294***
(2.04) Panel C: Regression results using non-zero observations of ΔSEC
BIG4 ? 0.002 0.001 −0.003 Intercept ? 0.103*** 0.104*** 0.105***
(0.12) (0.09) (−0.20) (5.21) (5.23) (5.22)
ΔNI ∗ BIG4 + 0.208* 0.218* 0.220** ΔNI + 0.549*** 0.557*** 0.572***
(1.57) (1.64) (1.70) (3.64) (3.67) (3.74)
ΔOCI ∗ BIG4 + 2.370*** ΔOCI + 3.367***
(3.24) (3.19)
ΔSEC ∗ BIG4 + −5.211 −5.354 ΔSEC + 6.659*** 6.791***
(−1.21) (−1.24) (3.59) (3.65)
ΔNSEC ∗ BIG4 + 3.464*** ΔNSEC + 2.217
(4.04) (1.26)
ΔPEN ∗ BIG4 + 2.492* ΔPEN + 5.017***
(1.48) (2.02)
ΔFOR ∗ BIG4 + 4.181*** ΔFOR + −0.167
(3.56) (−0.07)
ΔDER ∗ BIG4 + 0.113 ΔDER + −15.331
(0.02) (−0.72)
NI + 0.059 0.053 0.071 BIG4 ? 0.029 0.028 0.029
(1.13) (1.33) (1.26) (1.40) (1.33) (1.33)
ΔOTHEROCI + 0.577 −2.076 ΔNI ∗ BIG4 + 0.340** 0.331** 0.310**
(0.18) (−0.42) (1.99) (1.92) (1.79)
SIZE + 0.017*** 0.017*** 0.017*** ΔOCI ∗ BIG4 + 1.782*
(5.47) (5.32) (5.21) (1.59)
MB + 0.001*** 0.001*** 0.002*** ΔSEC ∗ BIG4 + −0.216 −0.524
(3.70) (3.63) (3.98) (−0.09) (−0.22)
LEV ? 0.017 0.021 0.029 ΔNSEC ∗ BIG4 + 3.303**
(0.56) (0.68) (0.92) (1.82)
LOSS - 0.014 0.012 0.013 ΔPEN ∗ BIG4 + 1.698
(0.81) (0.70) (0.71) (0.64)
YEAR ? Included Included Included ΔFOR ∗ BIG4 + 5.346***
Adjusted R2 6.77 6.95 7.17 (2.15)
N 17,878 17,878 17,878 ΔDER ∗ BIG4 + 28.985
Test of Coefficient: (1.25)
2
ΔSEC ∗ BIG4 = ΔNSEC ∗ BIG4 χ = 4.27 (p = 0.0387) Adjusted R2 5.75% 5.77% 5.79%
Panel B: Regression results excluding financial crisis periods N 7952 7952 7952
Intercept ? 0.108*** 0.115*** 0.114*** Test of Coefficients:
(7.75) (7.76) (7.76) ΔSEC ∗ BIG4 = ΔNSEC ∗ BIG4 χ2 = 2.85 (p = 0.0914)
ΔNI + 0.906*** 0.900*** 0.909*** Panel D: Regression results using auditor industry specialization as an alternative
(7.02) (6.92) (6.96) proxy for audit quality
ΔOCI + 0.341 Intercept ? 0.142*** 0.143*** 0.141***
(0.36) (15.22) (15.29) (14.70)
ΔSEC + 9.369*** 9.547*** ΔNI + 1.038*** 1.037*** 1.034***
(2.45) (2.49) (10.66) (10.66) (10.70)
ΔNSEC + −0.148* ΔOCI + 3.708***
(−1.40) (7.08)
ΔPEN + 0.652 ΔSEC + 13.432*** 13.403***
(0.35) (5.47) (5.46)
ΔFOR + −1.489 ΔNSEC + 3.324***
(−1.05) (6.19)
ΔDER + 2.562 ΔPEN + 3.321***
(0.58) (3.28)
BIG4 ? 0.035** 0.033** 0.031* ΔFOR + 3.788***
(2.33) (2.09) (1.90) (5.41)
ΔNI ∗ BIG4 + 0.152 0.165 0.153
(1.00) (1.08) (0.99)
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231 229

Table 8 (continued) quartile. This may cause the lower coefficient of ΔSEC due to the pre-
Dependent variable = RET ponderance of zero values. To investigate this possibility, we utilize
Variable Predicted Model 1 Model 2 Model 3 non-zero ΔSEC observations only and repeat our analyses. Panel C
sign Coefficient Coefficient Coefficient of Table 8 displays the results using non-zero values of ΔSEC. We
(t-statistics) (t-statistics) (t-statistics) find that the results using non-zero ΔSEC observations only are qual-
Panel D: Regression results using auditor industry specialization as an alternative itatively similar to those previously reported, except for the result of
proxy for audit quality minimum pension adjustment. While the coefficients on ΔOCI ∗ BIG4
ΔDER + 0.220 (in model 1), ΔNSEC ∗ BIG4 (in model 2), and ΔFOR ∗ BIG4 (in model
(0.11)
3) are all positive and significant at the conventional level, the coeffi-
ISA ? 0.020** 0.020** 0.020**
(1.84) (1.84) (1.77) cient on ΔPEN ∗ BIG4 is positive but insignificant. Such an insignifi-
ΔNI ∗ ISA + 0.141 0.142 0.145 cant result of minimum pension adjustment may be caused by the
(1.13) (1.14) (1.17) small number of observations.
ΔOCI ∗ ISA + 1.682***
(2.77)
ΔSEC ∗ ISA + −0.848 −0.853 4.7. Analysis using auditor industry specialization as an alternative proxy
(−0.23) (−0.24) for audit quality
ΔNSEC ∗ ISA + 1.878***
(3.03)
Prior literature (e.g., Craswell et al., 1995; DeFond, 1992; Deis &
ΔPEN ∗ ISA + 1.757*
(1.52) Giroux, 1992; Hogan & Jeter, 1999; Owhoso, Messier, & Lynch, 2002;
ΔFOR ∗ ISA + 1.857*** Solomon, Shields, & Whittington, 1999) suggests that auditor industry
(2.27) specialization is a good proxy for auditor quality. To examine whether
ΔDER ∗ ISA + 2.831 our results are robust to an alternative measure for audit quality, we
(1.20)
Adjusted R2 6.70% 6.83% 6.85%
conduct an additional analysis by replacing BIG4 with ISA (i.e., auditor
N 17,878 17,878 17,878 industry specialization). Panel D of Table 8 reports the results.
Test of Coefficients: Consistent with the results reported earlier, the coefficients on
2
ΔSEC ∗ ISA = ΔNSEC ∗ ISA χ = 0.66 (p = 0.4158) ΔOCI ∗ ISA (in model 1), ΔNSEC ∗ ISA (in model 2), and ΔPEN ∗ ISA
Panel A reports the results of OLS regressions after including control variables, where and ΔFOR ∗ ISA (in model 3) are all positive and significant at conven-
NI = the level of net income before extraordinary item deflated by average total assets; tional levels. Therefore, our results are robust even with an alterna-
ΔOTHER = component of ΔOCI other than ΔSEC, ΔPEN, ΔFOR, and ΔDER; SIZE = firm's tive proxy for audit quality.
size, measured as natural logarithm of market value of equity; MB = the market to
book ratio; LEV = debt-to-asset ratio, measured as long-term liabilities divided by
total assets; LOSS = an indicator variable that takes the value of 1 if current period 4.7.1. Analysis using Big 6 Auditors as an alternative proxy for audit
net income is less than 0, and 0 otherwise; and YEAR = a firm's fiscal year dummy var- quality
iable. The t-statistics are calculated using heteroscedasticity-consistent standard error.
Recent studies (e.g., Boone, Khurana, & Raman, 2010; Lawrence
The variables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom 0.5 percentile
levels. All other variables are defined in Appendix A. *, **, *** indicate statistical signif- et al., 2011) show significant similarities in audit quality between
icance at the 0.10, 0.05, and 0.01 levels, respectively, based on one-tailed tests if the co- Big 4 and non-Big 4, especially second-tier auditors. For instance,
efficient has a predicted sign, and two-tailed tests otherwise. Panel B reports the results Boone et al. (2010) report that although there is a more pronounced
of OLS regression after excluding financial crisis periods of 2008–2009. Panel C presents difference in perceived audit quality between Big 4 and second-tier
the results of OLS regression based on sample with positive SEC. In Panel D, ISA is an
indicator of industry specialized auditor (ISA) as an alternative proxy for audit quality,
auditors, little difference in actual audit quality is observed. Hence,
measured by market share threshold of 15% (based on audit fees). All other variables by adding two second-tier audit firms, Grant Thornton and BDO
are defined in Appendix A. t-statistics are calculated using heteroscedasticity-consistent Seidman, to Big 4 audit firm group, we utilize Big 6 auditors as an al-
standard error. The variables of RET, ΔNI, and ΔOCI are trimmed at the top and bottom ternative proxy for audit quality and replicate our analyses.27
0.5 percentile levels. *, **, *** indicate statistical significance at the 0.10, 0.05, and 0.01
Untabulated results confirm that our findings are robust to Big 6 audi-
levels, respectively, based on one-tailed tests if the coefficient has a predicted sign, and
two-tailed tests otherwise. tors. This evidence supports the view of similarities in audit quality
between Big 4 and second-tier auditors.28

4.6.2. Analysis excluding financial crisis periods

Our findings may be sensitive to the market condition, especially
during the financial crisis of 2008. This is especially important, be-
cause fair-value accounting could be significantly influenced by the
crisis.25 The financial crisis overlaps the last two years of our sample
period. Therefore, to capture the different valuation effect of audit
quality after controlling for the potential effect of the financial crisis,
we exclude observations in the financial-crisis period (i.e., 2008–2009
in our sample) and repeat our analysis. Panel B of Table 8 reports the re-
sults. As shown, the results remain unchanged even after excluding ob-
servations from the financial crisis.26
4.6.3. Analysis using non-zero value of ΔSEC
As observed in Table 2, many of the OCI components in our sample
have zero values. For instance, ΔSEC has a value of zero at the third
25
We thank an anonymous reviewer for bringing up this important insight.
26
In addition, having a Big 4 auditor may be especially valuable during the financial
crisis because investors could heavily discount fair value estimates during the crisis.
Thus, to see if whether OCI was still value-relevant in general over this time period,
we conduct additional analysis using the sample over 2008–2009 period only.
Untabulated results show that the coefficients of ΔOCI ∗ BIG4 and ΔNSEC ∗ BIG4 are
positive and significant at the 1% level, respectively, supporting our H1 and H2.
5. Conclusions
It is known that audit quality matters in financial reporting, thereby
affecting the pricing (e.g., Knechel et al., 2007; Menon & Williams, 1991;
Palmrose, 1988) and quality of earnings (e.g., Balsam et al., 2003; Becker
et al., 1998; Francis et al., 1999; Reynolds & Francis, 2000; Teoh & Wong,
1993). The current financial reporting environment raises concerns
about the determination of fair-value measures that are challenging
27
Recently, four second-tier audit firms, Grant Thornton, BDO Seidman, McGladrey &
Pullen, and Crowe Horwath, are also inspected annually by the PCAOB (see 2010
PCAOB inspection report). Therefore, we utilize Big 8 auditors as another alternative
proxy for audit quality and repeat our analysis. We also conduct an additional analysis
to see if OCI components audited by Big 4 auditors are more value-relevant for smaller
firms than for large firms. By using two subsamples (small firms vs. large firms), which
are annually ranked by firm size, we find the value-relevance of OCI audited by Big 4
auditors only for small firms. Overall, untabulated results are qualitatively similar to
those reported in this study.
28
Furthermore, we conduct additional tests to examine how the direction of an audi-
tor change affects the informativeness of OCI estimates. We predict that an upward au-
ditor change (i.e. non-Big 4 to Big 4 auditor) may result in a higher value relevance of
OCI than a downward auditor change (i.e., Big 4 to non-Big 4 auditor). Untabulated re-
sults, however, show no differential valuation effect between the two types of auditor
changes. These results may be caused by the small sample sizes of the two groups.
230 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
for auditors and users of financial information (see Christensen et al.,
2012; PCAOB, 2007b, 2008; SEC, 2008). In particular, financial reporting
of OCI may be subject to management's assumptions and judgments.
Extending prior literature, this study empirically explores whether
OCI items of Big 4 clients have incremental information content in
explaining stock returns relative to non-Big 4 clients. Assuming that
the higher the level of subjectivity and professional judgment involved
in the financial reporting process, the more auditor scrutiny increases,
we then investigate whether the differential valuation effect of OCI be-
tween Big 4 and non-Big 4 clients is attributable to the degree of subjec-
tive judgment and estimates required in estimating OCI components.
We find that OCI of Big4 clients is more value-relevant than that of
non-Big4 clients. Our results also show that the differential valuation
effect of OCI between Big 4 and non-Big 4 clients is more pronounced
for the more subjective OCI components, such as minimum pension
liability and foreign-currency translation adjustment, compared to
the less subjective marketable-securities adjustments. These results
are robust after controlling self-selection bias, potential effect of fi-
nancial crisis, and other related effects.
Our findings provide implications for accounting professionals, reg-
ulators, and future research. Since the financial reporting paradigm is
being moved toward fair-value accounting, accounting professionals
and regulators have cast doubt on auditors' ability or willingness to ex-
ercise professional judgment for handling fair-value audits (e.g., Jamal
et al., 2010). Given that OCI is largely measured from the fair-value ap-
plication, our results imply that Big 4 auditors provide greater assurance
in the context of fair-value audits, especially for OCI components with
more subjective judgment. Since our findings portray only a limited
scope of fair-value audits, future research may provide additional in-
sights of the effectiveness of Big 4 auditors on fair-value accounting by
extending the scope.
Appendix A. Variable definitions
RETit buy-and-hold return measured over a one-year window
(eight months before to four months after the fiscal year
end);
ΔNI it change in net income after extraordinary of firm i in year t,
scaled by average total assets;
ΔOCIit change in accumulated other comprehensive income of
firm i in year t, scaled by average total assets;
ΔSECit change in accumulated unrealized gains or losses on
available-for-sale securities of firm i in year t, scaled by
average total assets;
ΔNSECit change in adjustments of OCI components other than
available-for-sale securities, measured as the sum of change
in the accumulated balance of minimum pension liability
(ΔPEN), plus change in accumulated adjustment of foreign
currency translation (ΔFOR), plus change in accumulated
derivatives unrealized gains or losses (ΔDER), scaled by
average total assets;
ΔPENit change in accumulated balance of minimum pension ad-
justment of firm i in year t, scaled by average total assets;
ΔFORit change in accumulated adjustment of foreign currency
translation of firm i in year t, scaled by average total assets;
ΔDERit change in accumulated derivatives unrealized gains or
losses of firm i in year t, scaled by average total assets;
BIGit an indicator variable that takes a value of 1 if firm i's auditor
in year t is a Big 4, and 0 otherwise.
References
Abbott, L. J., Gunny, K., & Zhang, T. (2008). When the PCAOB talks, who listens?
Evidence from client firm reaction to adverse, GAAP deficient PCAOB inspection re-
ports. Working paper. University of Memphis, University of Colorado at Boulder,
and Singapore Management University.
American Institute of Certified Public Accountants (AICPA) (2002). Consideration of
fraud in a financial statement audit. Statement on Auditing Standards No. 99. New
York, NY: AICPA.
Balsam, S., Krishnan, J., & Yang, J. S. (2003). Auditor industry specialization and the
earnings response coefficient. Auditing: A Journal of Practice & Theory, 22(2), 71–97.
Balvers, R., McDonald, B., & Miller, R. (1988). Underpricing of new issues and the choice
of auditors as a signal of investment banker reputation. The Accounting Review,
63(4), 605–622.
Bamber, L., Jiang, J., Petroni, K., & Wang, I. (2010). Comprehensive income: Who's afraid
of performance reporting? The Accounting Review, 85(1), 97–126.
Beatty, R. (1989). Auditor reputation and the pricing of initial public offerings. The
Accounting Review, 64(4), 693–709.
Becker, C. L., DeFond, M. L., Jiambalvo, J., & Subramanyam, K. R. (1998). The effect of
audit quality on earnings management. Contemporary Accounting Research, 15(1),
1–24.
Biddle, G., & Choi, J. (2006). Is comprehensive income useful? Journal of Contemporary
Accounting and Economics, 2(1), 1–32.
Boone, J. P., Khurana, I., & Raman, K. (2010). Do the Big 4 and the Second-tier firms
provide audits of similar quality? Journal of Accounting and Public Policy, 29(4),
330–352.
Carroll, T., Linsmeier, T., & Petroni, K. (2003). Fair value vs. historical cost accounting:
Evidence from closed-end mutual funds. Journal of Accounting, Auditing and Finance,
18, 1–23.
Chambers, D., Linsmeier, T., Shakespeare, C., & Sougiannis, T. (2007). An evaluation of
SFAS No. 130 comprehensive income disclosures. Review of Accounting Studies,
12(4), 557–593.
Chaney, P., Jeter, D., & Shivakumar, L. (2004). Self-selection of auditors and audit pric-
ing in private firms. The Accounting Review, 79(1), 51–72.
Choi, J., Das, S., & Zang, Y. (2007). Comprehensive income, future earnings and market
mispricing. Working Paper. Seoul National University.
Christensen, B., Glover, S., & Wood, D. A. (2012). Extreme estimation uncertainty in fair
value estimates: Implications for audit assurance. Auditing: A Journal of Practice &
Theory, 31(1), 127–146.
Church, B., & Shefchik, L. (2012). PCAOB inspections and large accounting firms.
Accounting Horizons, 26(1), 43–63.
Craswell, J., Francis, R., & Taylor, S. L. (1995). Auditor brand name reputations and in-
dustry specializations. Journal of Accounting and Economics, 20(3), 297–322.
Datar, S., Feltham, G., & Hughes, J. (1991). The role of audits and audit quality in valuing
new issues. Journal of Accounting and Economics, 14(1), 3–49.
DeAngelo, L. (1981). Auditor independence, ‘low-balling’, and disclosure regulation.
Journal of Accounting and Economics, 3(2), 113–127.
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.
DeFond, M. L. (2010). How should the auditors be audited? Comparing the PCAOB in-
spections with the AICPA peer reviews. Journal of Accounting and Economics,
49(1–2), 104–108.
DeFond, M. L., & Jiambalvo, J. (1993). Factors related to auditor–client disagreements
over income-increasing accounting methods. Contemporary Accounting Research,
9(2), 415–431.
Deis, D. R., & Giroux, G. (1992). Determinants of audit quality in the public sector. The
Accounting Review, 67(3), 462–479.
Dhaliwal, D., Subramanyam, K., & Trezevant, R. (1999). Is comprehensive income supe-
rior to net income as a measure of firm performance? Journal of Accounting and
Economics, 26(1–3), 43–67.
Dye, R. A. (1993). Auditing standards, legal liability, and auditor wealth. Journal of
Political Economy, 101(5), 887–914.
Easton, P. D., & Harris, T. S. (1991). Earnings as an explanatory variable for returns.
Journal of Accounting Research, 29(1), 19–36.
Eichenseher, J., Hagigi, M., & Shields, D. (1989). Market reaction to auditor changes by
OTC companies. Auditing: A Journal of Practice & Theory, 9(1), 29–40.
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal
of Finance, 47(2), 427–465.
Francis, J. R., Maydew, E. L., & Sparks, H. C. (1999). The role of Big 6 auditors in the
credible reporting of accruals. Auditing: A Journal of Practice & Theory, 18(2), 17–34.
Francis, J., & Wang, D. (2008). The joint effect of investor protection and Big 4 audits on
earnings quality around the world. Contemporary Accounting Research, 25(1),
157–191.
Francis, J. R., & Wilson, E. R. (1988). Auditor changes: A joint test of theories relating to
agency costs and auditor differentiation. The Accounting Review, 63(4), 663–682.
Gow, I. D., Ormazabal, G., & Taylor, D. (2010). Correcting for cross-sectional and time-
series dependence in accounting research. The Accounting Review, 85(2), 483–512.
Hackenbrack, K. E., & Hogan, C. E. (2002). Market response to earnings surprises condi-
tional on reasons for an auditor change. Contemporary Accounting Research, 19(2),
195–223.
Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47(1),
153–161.
Hogan, C. E., & Jeter, D. C. (1999). Industry specialization by auditors. Auditing: A Journal
of Practice & Theory, 18(1), 1–17.
Holt, P. (2011). Some effects of alternative foreign currency translation methodologies
on two short-term liquidity ratios. American Journal of Economics and Business
Administrations, 3(4), 645–651.
Holthausen, R., & Verrecchia, R. (1988). The effects of sequential information releases
on the variance of price changes in an intertemporal multi-asset market. Journal
of Accounting Research, 26(1), 82–106.
Jamal, K., Bloomfield, R., Christensen, T., Colson, R., Moehrle, S., Ohlson, J., et al. (2010).
A research-based perspective on the SEC's proposed rule—Roadmap for the
 C. Lee, M.S. Park / Advances in Accounting, incorporating Advances in International Accounting 29 (2013) 218–231
potential use of financial statement prepared in accordance with International
Financial Reporting System (IFRS) by U.S. issuers. Accounting Horizons, 24(1),
139–147.
Johnson, W., & Lys, T. (1990). The market for audit services: Evidence from voluntary
auditor changes. Journal of Accounting and Economics, 12(1–3), 281–308.
Jones, D., & Smith, K. (2011). Comparing the value relevance, predictive value, and per-
sistence of other comprehensive income and special items. The Accounting Review,
86(6), 2047–2073.
Kanagaretnam, K., Mathieu, R., & Shehata, M. (2009). Usefulness of comprehensive in-
come reporting in Canada. Journal of Accounting and Public Policy, 28(4), 349–365.
Khurana, I., & Raman, K. (2004). Litigation risk and the financial reporting credibility
of Big 4 versus non-Big 4 audits: Evidence from anglo-american countries. The
Accounting Review, 79(2), 473–495.
Kim, J. B., Chung, R., & Firth, M. (2003). Auditor conservatism, asymmetric monitoring
and earnings management. Contemporary Accounting Research, 20(2), 323–359.
Knechel, R., Naiker, V., & Pacheco, G. (2007). Does auditor industry specialization
matter? Evidence from market reaction to auditor switches. Auditing: A Journal of
Practice & Theory, 26(1), 19–45.
Krishnan, G. V. (2003). Audit quality and the pricing of discretionary accruals. Auditing:
A Journal of Practice & Theory, 22(1), 109–126.
Lawrence, A., Minutti-Meza, M., & Zhang, P. (2011). Can Big 4 versus Non-Big 4 differences
in audit-quality proxies be attributed to client characteristics? The Accounting Review,
86(1), 259–286.
Louis, H. (2003). The value relevance of the foreign translation adjustment. The
Accounting Review, 78(4), 1027–1047.
Menon, K., & Williams, D. (1991). Auditor credibility and initial public offerings. The
Accounting Review, 66(2), 313–332.
Nichols, D., & Smith, D. (1983). Auditor credibility and auditor changes. Journal of
Accounting Research, 21(2), 534–544.
O'Hanlon, J., & Pope, P. (1999). The value-relevance of U.K. dirty surplus accounting
flows. The British Accounting Review, 31(4), 459–482.
Owhoso, V. E., Messier, W. F., & Lynch, J. G. (2002). Error detection by industry-
specialized teams during sequential audit review. Journal of Accounting Research,
40(3), 883–900.
Palmrose, Z. -V. (1988). An analysis of auditor litigation and audit service quality. The
Accounting Review, 63(1), 55–73.
Palmrose, Z. -V. (2006). Maintaining the value and viability of auaditors as gatekeepers
under SOX: An auditing master proposal. In Y. Fuchita, & R. E. Litan (Eds.), Financial
gatekeepers: Can they protect investors? (pp. 103–135). Washington, D.C.: Brookings
Institution Press and the Nomura Institute of Capital Markets Research.
Pinto, J. (2002). Foreign currency translation method choice: insights from Game
Theory. Journal of Applied Business Research, 18(4), 25–34.
231
Public Company Accounting Oversight Board (PCAOB) (2007a). Speech by Chairman of
Board Mark W. Olson at AICPA National Conference on Current SEC and PCAOB De-
velopments. Washington, DC. http://pcaobus.org/News/Speech/Pages/12102007_
OlsonAICPASECandPCAOBDevelopments.aspx
Public Company Accounting Oversight Board (PCAOB) (2007b). Staff Audit Practice
Alert No. 2: Matters related to auditing fair value measurements of financial instru-
ments and the use of specialists Washington, DC.
Public Company Accounting Oversight Board (PCAOB) (2008). Auditing fair value mea-
surements and disclosures. AU Section 328 (Washington, D.C.).
Public Company Accounting Oversight Board (PCAOB) (2011). Report on observations
of PCAOB inspectors related to audit risk areas affected by the economic crisis.
Washington, D.C.
Reynolds, J. K., & Francis, J. R. (2000). Does size matter? The influence of large clients on
office-level auditor reporting decisions. Journal of Accounting and Economics, 30(3),
375–400.
Securities and Exchange Commission (2008). SEC Office of the Chief Accountant and
FASB Staff Clarifications on Fair Value Accounting-For immediate Release 2008-234
Washington, D.C.
Skinner, D., & Srinivasan, S. (2012). Audit quality and auditor reputation: Evidence
from Japan. The Accounting Review, 87(5), 1737–1765.
Smithson, C., Smith, C., & Wilford, D. (1995). Managing financial risk: A guide to derivative
products, financial engineering and value maximization. New York: Irwin Professional
Publishing.
Solomon, I., Shields, M. D., & Whittington, O. R. (1999). What do industry-specialist
auditors know? Journal of Accounting Research, 37(1), 191–208.
Soo, B. S., & Soo, L. G. (1994). Accounting for multinational firms: Is the translation pro-
cess valued by the stock market? The Accounting Review, 69(4), 617–637.
Teoh, S., & Wong, T. (1993). Perceived auditor quality and the earnings response coef-
ficient. The Accounting Review, 68(2), 346–366.
Titman, S., & Trueman, B. (1986). Information quality and the valuation of new issues.
Journal of Accounting and Economics, 8(2), 159–172.
Watts, R. L., & Zimmerman, J. L. (1986). Positive accounting theory. Upper Saddle River,
NJ: Prentice Hall.
Weber, J., Willenborg, M., & Zhang, J. (2008). Does auditor reputation matter? The case
of KPMG Germany and ComROAD AG. Journal of Accounting Research, 46(4),
941–972.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a
direct test for heteroskedasticity. Econometrica, 48(4), 817–838.
Willenborg, M. (1999). Empirical analysis of the economic demand for auditing in the
initial public offering market. Journal of Accounting Research, 37(1), 225–238.