Description
Selectinganappropriatemeasurementbasisforfinancialreportingisafundamentalandcontentiousaccountingpolicyissue.Whilemanyarguethatfairvalueisthemostrelevantmeasurementbasisforfinancialreporting,otherobserversexpressconcernsaboutthereliability(or“faithfulrepresentation”),andthustheusefulness,offairvaluemeasurements.BhatandRyan(2015)considertheroleofriskman-agementtechnologies—inparticular,marketandcreditriskmodeling—intheestimationoffairvalues.InlightofourdiscussionofBhatandRyan’sstudy,wearguethatfutureresearchshouldaimtoextendourunderstandingofthefairvalueestimationprocessandthefactorsthatexplainvariationintheeliabil-ityoffairvaluesaswellasthechannelsthroughwhichinvestorslearnaboutfairvaluemeasurementreliability.
Accounting, Organizations and Society 46 (2015) 96–99
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Fair value measurement capabilities, disclosure, and the perceived
reliability of fair value estimates: A discussion of Bhat and Ryan (2015)
Ryan P. McDonough, Catherine M. Shakespeare
?
Ross School of Business, University of Michigan, United States
a r t i c l e i n f o
Article history:
Received 15 May 2015
Accepted 19 May 2015
Available online 10 June 2015
JEL Classi?cation:
G14
G21
G32
M41
Keywords:
Fair value measurement
Risk modeling
Market risk
Credit risk
Value relevance
a b s t r a c t
Selecting an appropriate measurement basis for ?nancial reporting is a fundamental and contentious
accounting policy issue. While many argue that fair value is the most relevant measurement basis for
?nancial reporting, other observers express concerns about the reliability (or “faithful representation”),
and thus the usefulness, of fair value measurements. Bhat and Ryan (2015) consider the role of risk man-
agement technologies—in particular, market and credit risk modeling—in the estimation of fair values. In
light of our discussion of Bhat and Ryan’s study, we argue that future research should aim to extend our
understanding of the fair value estimation process and the factors that explain variation in the reliabil-
ity of fair values as well as the channels through which investors learn about fair value measurement
reliability.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Measuring and reporting fair values of assets and liabilities has
long been a topic of substantial debate among academics, policy-
makers, and practitioners (see, e.g., Laux & Leuz, 2009 and Hodder,
Hopkins, & Schipper, 2014). A central theme of the fair value
debate is the tradeoff between the two fundamental qualitative
characteristics of accounting information: relevance and reliabil-
ity.
1
Advocates argue that fair value is the most relevant measure-
ment attribute for ?nancial reporting purposes because it increases
transparency by providing more timely information. In contrast,
critics contend that some fair value measurements are not useful
to investors because the reliability of these estimates is diminished
?
Corresponding author.
E-mail address: [email protected] (C.M. Shakespeare).
1
The Financial Accounting Standards Board (FASB) recently replaced the term
“reliability” with “faithful representation” (FASB, 2010). We use these terms inter-
changeably throughout our discussion in a way that is intended to be consistent
with prior use of the term “reliability” in the academic literature and the common-
ality between the FASB’s de?nitions of both terms.
when they are susceptible to manipulation, prone to estimation er-
ror, and/or di?cult to verify.
2
Academic researchers have contributed to the fair value debate,
in part, by determining whether and to what extent fair value
measurements are relevant to investors for valuation. In tests of
value relevance, which are joint tests of both relevance and relia-
bility, capital market researchers commonly examine associations
between fair value measurements and equity values (e.g., Barth,
1994; Barth, Beaver, & Landsman, 1996; Beaver & Venkatacha-
lam, 2003; Nelson, 1996; Petroni & Whalen, 1995; Venkatachalam,
1996).
3
Value relevance studies document estimated regression co-
e?cients that are often smaller in magnitude than their theoret-
2
The bias (noise) injected into ?nancial reports as a result of unobservable man-
agerial manipulation (estimation error) in determining fair values is particularly
problematic when investors are unable to discern the direction and magnitude of
misreporting, as in Fischer and Verrecchia’s (2000) analytical model.
3
Value relevance studies examine the relation between accounting information
and stock prices or changes in prices (i.e., returns). An accounting number is con-
sidered value relevant if the regression coe?cient is statistically different from zero
and has the correct sign. Fair values of assets (liabilities) have predicted regres-
sion coe?cients of 1 (?1); these predicted coe?cients are theoretical benchmarks
based on making certain underlying assumptions. We refer readers to Holthausen
and Watts (2001), Barth, Beaver, and Landsman (2001); and Landsman (2007) for
reviews of the value relevance literature.http://dx.doi.org/10.1016/j.aos.2015.05.003
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99 97
ically predicted values, suggesting that investors price some fair
value estimates at a discount. Variation in equity valuation mul-
tiples is generally consistent with differences in the perceived re-
liability of fair value estimates. In particular, pricing discounts are
primarily observed in the context of unveri?able fair value esti-
mates that are sensitive to managerial discretion over valuation
inputs, measurement error, or both. Consequently, fair value es-
timates may not be fully impounded in stock prices because in-
vestors’ assessments of relevance are confounded by their percep-
tions of measurement reliability (e.g., Kadous, Koonce, & Thayer,
2012). Opportunities to contribute to this literature, therefore, in-
clude isolating the underlying sources of variation in the reliability
of fair value measurements. Bhat and Ryan (2015) provide such a
study by attempting to link ?rm-speci?c technologies to the esti-
mation of fair value measurements.
Bhat and Ryan consider risk management technologies as po-
tential inputs in the fair value estimation process. In particular,
they assess whether market and credit risk modeling by ?nan-
cial institutions enhances the relation between stock returns and
estimated unrealized fair value gains and losses on ?nancial in-
struments. Estimated fair value gains and losses are assigned to
one of three categories based on ?nancial reporting treatment: (1)
amounts recorded in net income, (2) amounts recorded in other
comprehensive income, and (3) amounts disclosed in the notes.
Fair value gains and losses recorded in note disclosures (and, to
some extent, in other comprehensive income) are assumed to be
less reliably measured than amounts recorded in net income. Bhat
and Ryan conjecture that banks’ market and credit risk modeling
techniques can improve the usefulness of fair value measurements
suffering the most from reliability concerns, thus attenuating the
pricing discount applied by investors to account for measurement
error and bias. For their sample of 238 banks from 2002 to 2013,
the authors conclude that market and credit risk modeling gener-
ally improves the association between stock returns and estimated
fair value gains and losses recorded in other comprehensive in-
come and in note disclosures.
Below we begin by framing Bhat and Ryan’s study in the con-
text of the extant literature, particularly with respect to studies ex-
amining the value relevance of fair value measurements. We then
discuss some of Bhat and Ryan’s research design choices and pos-
sible alternative explanations to consider when interpreting their
results. Throughout our discussion, we offer suggestions for future
research that can push the fair value literature forward, in part by
addressing some of the inherent limitations encountered in this
study.
2. What is the contribution of the paper?
2.1. Value relevance and fair value measurements
An empirical challenge in the fair value literature is to identify
settings that allow researchers to disentangle the constructs of rel-
evance and reliability. Koonce, Nelson, and Shakespeare (2011) use
experimental methods to isolate and directly study investors’ be-
liefs about the relevance of fair values for ?nancial instruments,
while holding constant measurement reliability and other charac-
teristics. Archival studies, however, generally assess the differential
value relevance of fair value measurements by examining cross-
sectional variation in the perceived reliability of those estimates,
while assuming a minimum level of relevance. Song, Thomas, and
Yi (2010), for example, consider how the value relevance of fair
value estimates varies predictably with respect to the source of es-
timation inputs. According to the hierarchy established by State-
ment of Financial Accounting Standards No. 157 (now Account-
ing Standards Codi?cation 820), Level 1 and 2 fair value mea-
surements are derived from observable valuation inputs based on
quoted prices in active markets (FASB, 2006). In contrast, Level 3
fair value estimates are sensitive to unobservable valuation model
assumptions that are subject to managerial discretion and estima-
tion error. Consistent with Level 3 estimates likely being less reli-
ably measured than Level 1 and 2 estimates, the authors ?nd that
the estimated coe?cients from price level regressions are larger
for Level 1 and 2 fair value measurements than for Level 3 esti-
mates (see also Goh, Li, Ng, & Yong, 2015). Furthermore, they ?nd
that stronger corporate governance improves the association be-
tween stock price and Level 3 fair values, suggesting that effective
corporate governance can mitigate investors’ concerns about mea-
surement reliability.
2.2. Risk modeling, fair value gains and losses, and stock returns
Bhat and Ryan’s study complements the extant literature in sev-
eral ways. First, to proxy for fair value measurement reliability,
they examine variation in the recognition and placement of fair
values in ?nancial statements, rather than variation in the source
of information used to estimate fair values (e.g., estimates based
on Level 1, 2, and 3 valuation inputs). They argue that fair value
gains and losses recorded in note disclosures (and, to some ex-
tent, in other comprehensive income) correspond to ?nancial in-
struments that trade in illiquid (or “thin”) markets and, therefore,
are more likely to be less reliably measured than fair value gains
and losses recorded in net income. Second, the authors consider a
relatively unexplored mechanism for reliability enhancement—i.e.,
market and credit risk modeling techniques. Following the recent
?nancial crisis, risk management technologies, such as risk mod-
eling, have become an increasingly important aspect of corporate
decision making, particularly within ?nancial institutions. Improv-
ing our understanding of the role of risk management technologies
in, for example, the estimation of fair values is an important area
of research. Third, Bhat and Ryan use an extended and more het-
erogeneous sample period than the samples used in prior studies,
providing opportunities for the authors to conduct subsample tests
that emphasize the importance of market and credit risk modeling
in certain contexts such as the 2008 ?nancial crisis.
Some of their conclusions, however, are open to alternative ex-
planations. For example, it is unclear whether their results are
driven by a reduction in information asymmetry resulting from the
choice to disclose information about risk modeling activities or by
the risk modeling activities themselves. In particular, as we will
discuss below, it is di?cult to disentangle the role of risk manage-
ment technologies in the estimation of fair values from the impact
of disclosures on investors’ perceptions of fair value measurements.
3. Comments on empirical tests and on opportunities for
future research
3.1. Risk modeling quality and the reliability of fair value
measurements
Bhat and Ryan’s results provide interesting insights into the
role of internal risk modeling techniques in the measurement of
complex fair value estimates. For instance, based on the results
reported in Panel A of Table 4, Bhat and Ryan conclude that
market risk models improve the measurement of fair value gains
and losses recorded in other comprehensive income, while credit
risk models predominantly improve the measurement of fair value
gains and losses disclosed in the notes. An empirical limitation,
however, is that the authors make assumptions about variation in
the quality (and intensity) of risk modeling and the reliability of
fair value measurements. For example, the authors assume that the
quality of risk modeling is generally uniform across banks with the
98 R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99
same risk modeling score.
4
Speci?cally, variation in risk modeling
quality is restricted to correspond to the quantity of risk modeling
techniques that are disclosed in banks’ 10-K ?lings.
5
Since some
banks may treat risk modeling as a “check the box” compliance
exercise, a more thorough examination is required to better un-
derstand cross-sectional variation in banks’ risk modeling capabil-
ities and the potentially nuanced relation between risk modeling
and the reliability of fair value measurements. Opportunities for
future research include identifying settings to isolate variation in
risk modeling quality that do not rely on the subjective disclosure
of these activities, which will help determine the extent to which
superior risk modeling capabilities are incrementally informative
for fair value measurement.
The second assumption the authors make is that the nature of
fair value gains and losses is homogenous within the three ?nan-
cial reporting categories, which are based on the display of fair
value measurements: amounts reported in net income, other com-
prehensive income, and note disclosures. Variation within these
categories is important to explore because both market and credit
risk modeling systems may have implications for both recognized
and disclosed amounts. We note that in Panel B of Table 4, the au-
thors begin to explore cross-sectional variation in the nature of fair
value gains and losses by examining speci?c amounts recorded in
other comprehensive income. In particular, they document a more
pronounced impact of market risk modeling for banks with above
median percentages of mortgage-backed and asset-backed securi-
ties in their available-for-sale portfolios. We suspect, however, that
models of market risk may also be important for understanding
risks corresponding to demand deposits and ?xed-rate mortgage
loans, which are both disclosed in the notes. Furthermore, compo-
nents of credit risk models may be important for banks that hold a
substantial amount of agency debt (i.e., securities issued by Fannie
Mae or Freddie Mac) or sovereign debt, which are both recorded in
other comprehensive income as available-for-sale securities. Future
research can attempt to link speci?c technologies, such as compo-
nents of risk modeling techniques, to the estimation of fair values
of speci?c ?nancial instruments.
3.2. Disentangling the risk modeling and disclosure effects
How does market and credit risk modeling improve the reliabil-
ity of fair value estimates, thus making them more decision-useful
to investors for valuation purposes? In the context of Bhat and
Ryan, there are two plausible sources of the documented enhance-
ment in the relation between stock returns and fair value gains and
losses. First, risk modeling may improve the faithful representa-
tion of estimated fair values through a reduction in estimation er-
ror.
6
Second, the joint disclosure of risk modeling techniques with
other contemporaneous disclosures related to fair value measure-
ment may result in fair value estimates that are more veri?able
and understandable to investors. Veri?ability and understandability
4
Market and credit risk modeling scores are based on the number of risk mod-
eling techniques a bank discloses. The ?ve market risk modeling techniques are
interest rate gap analysis, interest rate sensitivity analysis, Value-at-Risk analysis,
stress testing, and backtesting. The four credit risk modeling techniques are statis-
tical credit risk measurement, credit scoring, internal credit risk rating, and stress
testing.
5
The authors are careful to state the following two related assumptions. The ab-
sence of risk modeling disclosures does not preclude risk modeling from being un-
dertaken by a bank. However, more disclosure of risk modeling activities is assumed
to correspond to relatively higher intensities of risk modeling than less (or no) dis-
closure of these activities.
6
One way to think about measurement error is in terms of the signal-to-noise
ratio, where an increase in the faithful representation of estimated fair value gains
and losses decreases the noise component (i.e., the denominator) and, therefore,
increases the signal-to-noise ratio, making these estimates more decision-useful to
investors.
are two qualitative characteristics that enhance the relevance and
reliability of accounting information (FASB, 2010). Below we dis-
cuss the potentially confounded roles of risk modeling and disclo-
sure, and the challenge in empirically disentangling these effects.
The potential for risk modeling-related improvements in the
faithful representation of fair value estimates is plausible, but likely
indirect. For instance, both risk models and fair value estimates re-
quire managers to make assumptions. Risk modeling, even if pru-
dently implemented, does not eliminate managerial discretion over
model assumptions. Consequently, it is unclear why risk modeling
should produce inputs for estimating fair values that are consid-
ered more reliable than if risk modeling techniques were not used
to estimate fair values. In addition, risk modeling is concerned
with estimating the risks associated with a distribution of prob-
able outcomes, whereas fair value measurement is concerned with
calculating a point estimate of the price of an asset or liability.
While models used by ?nancial institutions to estimate both risk
and fair value likely rely on similar technologies, capabilities, and
input data, it is not clear how or why risk models should feed di-
rectly into the valuation models that produce fair value estimates.
We argue that the risk modeling proxies developed by Bhat
and Ryan are likely imperfectly capturing a higher-order construct,
which we consider to be discipline over and capabilities in fair
value estimation. Unfortunately, the authors are unable to rule out
the possibility that the disclosure of risk modeling activities is
correlated with unobserved fair value estimation capabilities that
more directly explain cross-sectional variation in the reliability of
fair value measurements. Future research in this area can further
explore the potentially omitted ?rm-speci?c factors that enhance
the usefulness of fair value estimates to investors and the extent
to which these factors interact. Such factors might include ?rm-
speci?c capabilities in and control over the quality of fair value
measurement, certain auditor characteristics, and incentive com-
pensation programs that mitigate managers’ incentives to bias fair
value estimates.
Assuming that risk modeling enhances the reliability of fair
value estimates, how is this enhancement identi?ed by investors?
The market and credit risk modeling techniques considered by the
authors are identi?ed and categorized based on disclosures pro-
vided in 10-K ?lings. Investors may also learn about the reliability
of fair value estimates from these risk modeling disclosures. As the
authors suggest, however, these disclosures may not provide su?-
cient information for investors to discern the implementation and
quality of banks’ risk modeling systems and the corresponding im-
pact on fair value measurement. For instance, Bhat and Ryan (pp.
15–16) note that:
1. Presumably the disclosure of risk-modeling activities implies
banks engage in the activities, because there does not appear
to be any reason for banks to lie in these disclosures, which
generally are too terse to divulge meaningful proprietary infor-
mation.
Therefore, while these disclosures may be useful for identify-
ing the presence of certain risk modeling techniques, they may
not provide information to investors that explain the observed re-
lation between returns and fair value gains and losses. This is
consistent with the insigni?cant estimated coe?cients on the risk
modeling disclosure measures in the regressions reported in Ta-
bles 4 through 6. Another possibility, however, is that investors
learn about the reliability of fair value measurements through con-
temporaneous disclosures that discuss, among other things, the
controls, processes, and procedures designed to mitigate fair value
measurement concerns (see Chung, Goh, Ng, & Yong, 2015). Un-
fortunately, the authors do not control for contemporaneous dis-
closures that could in?uence the relation between returns and fair
value gains and losses.
R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99 99
Absent a more compelling and direct link between risk mod-
eling techniques, fair value estimation, and the channel through
which investors learn about the impact of ?rms’ capabilities and
technologies in the estimation of fair values, we interpret Bhat and
Ryan’s results with caution.
7
In particular, it is di?cult to disen-
tangle the role of risk modeling as a contributor to the reliability
of fair value measurements and the impact of contemporaneous
disclosures on the relation between returns and fair values. Future
research can seek to better understand the fair value estimation
process and the mechanisms through which investors learn about
the reliability of fair values.
4. Concluding remarks
The merits of the fair value measurement attribute continue
to be a controversial policy issue, particularly following the re-
cent ?nancial crisis. A fundamental concern with fair value mea-
surements is ensuring that the estimates represent the underlying
economic transactions they are intended to represent, particularly
when active (or “thick”) markets for the underlying asset or liabil-
ity do not exist. To that end, Bhat and Ryan provide some evidence
on the potential role of risk management technologies in enhanc-
ing the reliability of fair value measurements.
Future research should aim to expand our understanding of the
complex process of estimating fair values. Furthermore, it is im-
portant to better identify the underlying and potentially interre-
lated set of factors and ?rm capabilities that explain variation in
the reliability of fair value measurements. Finally, uncovering the
channels through which investors learn about enhancements to
fair value measurement reliability is important for understanding
the role of improved fair value disclosures, particularly in the con-
text of opaque ?nancial institutions whose disclosure decisions are
not well understood.
Acknowledgements
We appreciate the helpful comments provided by Linda Myers,
Hun-Tong Tan (editor), and participants at the 2014 Accounting, Or-
ganizations, and Society Conference on Accounting Estimates. We
gratefully acknowledge ?nancial support provided by Deloitte &
Touche LLP. Ryan McDonough is grateful for the ?nancial support
provided by the Paton Accounting Fellowship. Catherine Shake-
speare is the Teitelbaum Research Scholar.
7
Participants at the 2014 Accounting, Organizations, and Society Conference on
Accounting Estimates expressed similar concerns about the nature of the links be-
tween risk modeling disclosures, the reliability of fair value estimates, and investors’
perceptions of the reliability of fair value estimates.
References
Barth, M. E. (1994). Fair value accounting: Evidence from investment securities and
the market valuation of banks. The Accounting Review, 69(1), 1–25.
Barth, M. E., Beaver, W. H., & Landsman, W. R. (1996). Value-relevance of banks’ fair
value disclosures under SFAS No. 107. The Accounting Review, 71(4), 513–537.
Barth, M. E., Beaver, W. H., & Landsman, W. R. (2001). The relevance of the value rel-
evance literature for ?nancial accounting standard setting: another view. Journal
of Accounting and Economics, 31(1–3), 77–104.
Beaver, W. H., & Venkatachalam, M. (2003). Differential pricing of components of
bank loan fair values. Journal of Accounting, Auditing, & Finance, 18(1), 41–68.
Bhat, G., & Ryan, S. G., (2015). The impact of risk modeling on the market percep-
tion of banks’ estimated fair value gains and losses for ?nancial instruments.
Accounting, Organizations, and Society, this issue.
Chung, S., Goh, B., Ng, J., & Yong, K. (2015). Voluntary fair value disclosures beyond
SFAS 157’s three-level estimates. Working paper, Singapore Management Univer-
sity.
Financial Accounting Standards Board (FASB) (2006). Fair Value Measurements. State-
ment of Financial Accounting Standards No. 157. Norwalk, CT: FASB.
Financial Accounting Standards Board (FASB) (2010). Conceptual Framework for Fi-
nancial Reporting. Chapter 3, “Qualitative Characteristics of Useful Financial In-
formation”. Statement of Financial Accounting Concepts No. 8. Norwalk, CT:
FASB.
Fischer, P. E., & Verrecchia, R. E. (2000). Reporting Bias. The Accounting Review, 75(2),
229–245.
Goh, B. W., Li, D., Ng, J., & Yong, K. O. (2015). Market pricing of banks’ fair value
assets reported under SFAS 157 since the 2008 ?nancial crisis. Journal of Ac-
counting and Public Policy, 34(2), 129–145.
Hodder, L. D., Hopkins, P. E., & Schipper, K. (2014). Fair Value Measurement in Fi-
nancial Reporting. Foundations and Trends in Accounting, 8(3–4), 143–270.
Holthausen, R. W., & Watts, R. L. (2001). The relevance of the value-relevance lit-
erature for ?nancial accounting standard setting. Journal of Accounting and Eco-
nomics, 31(1–3), 3–75.
Kadous, K., Koonce, L., & Thayer, J. M. (2012). Do ?nancial statement users judge
relevance based on properties of reliability? The Accounting Review, 87(4), 1335–
1356.
Koonce, L., Nelson, K. K., & Shakespeare, C. M. (2011). Judging the relevance of fair
value for ?nancial instruments. The Accounting Review, 86(6), 2075–2098.
Landsman, W. R. (2007). Is fair value accounting information relevant and reli-
able? Evidence from capital market research. Accounting and Business Research,
37(Special Issue), 19–30.
Laux, C., & Leuz, C. (2009). The crisis of fair-value accounting: Making sense of the
recent debate. Accounting, Organizations and Society, 34(6–7), 826–834.
Nelson, K. K. (1996). Fair value accounting for commercial banks: An empirical anal-
ysis of SFAS No. 107. The Accounting Review, 71(2), 161–182.
Petroni, K. R., & Whalen, J. M. (1995). Fair values of equity and debt securities
and share prices of property-liability insurers. The Journal of Risk and Insurance,
62(4), 719–737.
Song, C. J., Thomas, W. B., & Yi, H. (2010). Value relevance of FAS No. 157 fair value
hierarchy information and the impact of corporate governance mechanisms. The
Accounting Review, 85(4), 1375–1410.
Venkatachalam, M. (1996). Value-relevance of banks’ derivatives disclosures. Journal
of Accounting and Economics, 22(1–3), 327–355.
doc_130772355.pdf
Selectinganappropriatemeasurementbasisforfinancialreportingisafundamentalandcontentiousaccountingpolicyissue.Whilemanyarguethatfairvalueisthemostrelevantmeasurementbasisforfinancialreporting,otherobserversexpressconcernsaboutthereliability(or“faithfulrepresentation”),andthustheusefulness,offairvaluemeasurements.BhatandRyan(2015)considertheroleofriskman-agementtechnologies—inparticular,marketandcreditriskmodeling—intheestimationoffairvalues.InlightofourdiscussionofBhatandRyan’sstudy,wearguethatfutureresearchshouldaimtoextendourunderstandingofthefairvalueestimationprocessandthefactorsthatexplainvariationintheeliabil-ityoffairvaluesaswellasthechannelsthroughwhichinvestorslearnaboutfairvaluemeasurementreliability.
Accounting, Organizations and Society 46 (2015) 96–99
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Fair value measurement capabilities, disclosure, and the perceived
reliability of fair value estimates: A discussion of Bhat and Ryan (2015)
Ryan P. McDonough, Catherine M. Shakespeare
?
Ross School of Business, University of Michigan, United States
a r t i c l e i n f o
Article history:
Received 15 May 2015
Accepted 19 May 2015
Available online 10 June 2015
JEL Classi?cation:
G14
G21
G32
M41
Keywords:
Fair value measurement
Risk modeling
Market risk
Credit risk
Value relevance
a b s t r a c t
Selecting an appropriate measurement basis for ?nancial reporting is a fundamental and contentious
accounting policy issue. While many argue that fair value is the most relevant measurement basis for
?nancial reporting, other observers express concerns about the reliability (or “faithful representation”),
and thus the usefulness, of fair value measurements. Bhat and Ryan (2015) consider the role of risk man-
agement technologies—in particular, market and credit risk modeling—in the estimation of fair values. In
light of our discussion of Bhat and Ryan’s study, we argue that future research should aim to extend our
understanding of the fair value estimation process and the factors that explain variation in the reliabil-
ity of fair values as well as the channels through which investors learn about fair value measurement
reliability.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Measuring and reporting fair values of assets and liabilities has
long been a topic of substantial debate among academics, policy-
makers, and practitioners (see, e.g., Laux & Leuz, 2009 and Hodder,
Hopkins, & Schipper, 2014). A central theme of the fair value
debate is the tradeoff between the two fundamental qualitative
characteristics of accounting information: relevance and reliabil-
ity.
1
Advocates argue that fair value is the most relevant measure-
ment attribute for ?nancial reporting purposes because it increases
transparency by providing more timely information. In contrast,
critics contend that some fair value measurements are not useful
to investors because the reliability of these estimates is diminished
?
Corresponding author.
E-mail address: [email protected] (C.M. Shakespeare).
1
The Financial Accounting Standards Board (FASB) recently replaced the term
“reliability” with “faithful representation” (FASB, 2010). We use these terms inter-
changeably throughout our discussion in a way that is intended to be consistent
with prior use of the term “reliability” in the academic literature and the common-
ality between the FASB’s de?nitions of both terms.
when they are susceptible to manipulation, prone to estimation er-
ror, and/or di?cult to verify.
2
Academic researchers have contributed to the fair value debate,
in part, by determining whether and to what extent fair value
measurements are relevant to investors for valuation. In tests of
value relevance, which are joint tests of both relevance and relia-
bility, capital market researchers commonly examine associations
between fair value measurements and equity values (e.g., Barth,
1994; Barth, Beaver, & Landsman, 1996; Beaver & Venkatacha-
lam, 2003; Nelson, 1996; Petroni & Whalen, 1995; Venkatachalam,
1996).
3
Value relevance studies document estimated regression co-
e?cients that are often smaller in magnitude than their theoret-
2
The bias (noise) injected into ?nancial reports as a result of unobservable man-
agerial manipulation (estimation error) in determining fair values is particularly
problematic when investors are unable to discern the direction and magnitude of
misreporting, as in Fischer and Verrecchia’s (2000) analytical model.
3
Value relevance studies examine the relation between accounting information
and stock prices or changes in prices (i.e., returns). An accounting number is con-
sidered value relevant if the regression coe?cient is statistically different from zero
and has the correct sign. Fair values of assets (liabilities) have predicted regres-
sion coe?cients of 1 (?1); these predicted coe?cients are theoretical benchmarks
based on making certain underlying assumptions. We refer readers to Holthausen
and Watts (2001), Barth, Beaver, and Landsman (2001); and Landsman (2007) for
reviews of the value relevance literature.http://dx.doi.org/10.1016/j.aos.2015.05.003
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99 97
ically predicted values, suggesting that investors price some fair
value estimates at a discount. Variation in equity valuation mul-
tiples is generally consistent with differences in the perceived re-
liability of fair value estimates. In particular, pricing discounts are
primarily observed in the context of unveri?able fair value esti-
mates that are sensitive to managerial discretion over valuation
inputs, measurement error, or both. Consequently, fair value es-
timates may not be fully impounded in stock prices because in-
vestors’ assessments of relevance are confounded by their percep-
tions of measurement reliability (e.g., Kadous, Koonce, & Thayer,
2012). Opportunities to contribute to this literature, therefore, in-
clude isolating the underlying sources of variation in the reliability
of fair value measurements. Bhat and Ryan (2015) provide such a
study by attempting to link ?rm-speci?c technologies to the esti-
mation of fair value measurements.
Bhat and Ryan consider risk management technologies as po-
tential inputs in the fair value estimation process. In particular,
they assess whether market and credit risk modeling by ?nan-
cial institutions enhances the relation between stock returns and
estimated unrealized fair value gains and losses on ?nancial in-
struments. Estimated fair value gains and losses are assigned to
one of three categories based on ?nancial reporting treatment: (1)
amounts recorded in net income, (2) amounts recorded in other
comprehensive income, and (3) amounts disclosed in the notes.
Fair value gains and losses recorded in note disclosures (and, to
some extent, in other comprehensive income) are assumed to be
less reliably measured than amounts recorded in net income. Bhat
and Ryan conjecture that banks’ market and credit risk modeling
techniques can improve the usefulness of fair value measurements
suffering the most from reliability concerns, thus attenuating the
pricing discount applied by investors to account for measurement
error and bias. For their sample of 238 banks from 2002 to 2013,
the authors conclude that market and credit risk modeling gener-
ally improves the association between stock returns and estimated
fair value gains and losses recorded in other comprehensive in-
come and in note disclosures.
Below we begin by framing Bhat and Ryan’s study in the con-
text of the extant literature, particularly with respect to studies ex-
amining the value relevance of fair value measurements. We then
discuss some of Bhat and Ryan’s research design choices and pos-
sible alternative explanations to consider when interpreting their
results. Throughout our discussion, we offer suggestions for future
research that can push the fair value literature forward, in part by
addressing some of the inherent limitations encountered in this
study.
2. What is the contribution of the paper?
2.1. Value relevance and fair value measurements
An empirical challenge in the fair value literature is to identify
settings that allow researchers to disentangle the constructs of rel-
evance and reliability. Koonce, Nelson, and Shakespeare (2011) use
experimental methods to isolate and directly study investors’ be-
liefs about the relevance of fair values for ?nancial instruments,
while holding constant measurement reliability and other charac-
teristics. Archival studies, however, generally assess the differential
value relevance of fair value measurements by examining cross-
sectional variation in the perceived reliability of those estimates,
while assuming a minimum level of relevance. Song, Thomas, and
Yi (2010), for example, consider how the value relevance of fair
value estimates varies predictably with respect to the source of es-
timation inputs. According to the hierarchy established by State-
ment of Financial Accounting Standards No. 157 (now Account-
ing Standards Codi?cation 820), Level 1 and 2 fair value mea-
surements are derived from observable valuation inputs based on
quoted prices in active markets (FASB, 2006). In contrast, Level 3
fair value estimates are sensitive to unobservable valuation model
assumptions that are subject to managerial discretion and estima-
tion error. Consistent with Level 3 estimates likely being less reli-
ably measured than Level 1 and 2 estimates, the authors ?nd that
the estimated coe?cients from price level regressions are larger
for Level 1 and 2 fair value measurements than for Level 3 esti-
mates (see also Goh, Li, Ng, & Yong, 2015). Furthermore, they ?nd
that stronger corporate governance improves the association be-
tween stock price and Level 3 fair values, suggesting that effective
corporate governance can mitigate investors’ concerns about mea-
surement reliability.
2.2. Risk modeling, fair value gains and losses, and stock returns
Bhat and Ryan’s study complements the extant literature in sev-
eral ways. First, to proxy for fair value measurement reliability,
they examine variation in the recognition and placement of fair
values in ?nancial statements, rather than variation in the source
of information used to estimate fair values (e.g., estimates based
on Level 1, 2, and 3 valuation inputs). They argue that fair value
gains and losses recorded in note disclosures (and, to some ex-
tent, in other comprehensive income) correspond to ?nancial in-
struments that trade in illiquid (or “thin”) markets and, therefore,
are more likely to be less reliably measured than fair value gains
and losses recorded in net income. Second, the authors consider a
relatively unexplored mechanism for reliability enhancement—i.e.,
market and credit risk modeling techniques. Following the recent
?nancial crisis, risk management technologies, such as risk mod-
eling, have become an increasingly important aspect of corporate
decision making, particularly within ?nancial institutions. Improv-
ing our understanding of the role of risk management technologies
in, for example, the estimation of fair values is an important area
of research. Third, Bhat and Ryan use an extended and more het-
erogeneous sample period than the samples used in prior studies,
providing opportunities for the authors to conduct subsample tests
that emphasize the importance of market and credit risk modeling
in certain contexts such as the 2008 ?nancial crisis.
Some of their conclusions, however, are open to alternative ex-
planations. For example, it is unclear whether their results are
driven by a reduction in information asymmetry resulting from the
choice to disclose information about risk modeling activities or by
the risk modeling activities themselves. In particular, as we will
discuss below, it is di?cult to disentangle the role of risk manage-
ment technologies in the estimation of fair values from the impact
of disclosures on investors’ perceptions of fair value measurements.
3. Comments on empirical tests and on opportunities for
future research
3.1. Risk modeling quality and the reliability of fair value
measurements
Bhat and Ryan’s results provide interesting insights into the
role of internal risk modeling techniques in the measurement of
complex fair value estimates. For instance, based on the results
reported in Panel A of Table 4, Bhat and Ryan conclude that
market risk models improve the measurement of fair value gains
and losses recorded in other comprehensive income, while credit
risk models predominantly improve the measurement of fair value
gains and losses disclosed in the notes. An empirical limitation,
however, is that the authors make assumptions about variation in
the quality (and intensity) of risk modeling and the reliability of
fair value measurements. For example, the authors assume that the
quality of risk modeling is generally uniform across banks with the
98 R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99
same risk modeling score.
4
Speci?cally, variation in risk modeling
quality is restricted to correspond to the quantity of risk modeling
techniques that are disclosed in banks’ 10-K ?lings.
5
Since some
banks may treat risk modeling as a “check the box” compliance
exercise, a more thorough examination is required to better un-
derstand cross-sectional variation in banks’ risk modeling capabil-
ities and the potentially nuanced relation between risk modeling
and the reliability of fair value measurements. Opportunities for
future research include identifying settings to isolate variation in
risk modeling quality that do not rely on the subjective disclosure
of these activities, which will help determine the extent to which
superior risk modeling capabilities are incrementally informative
for fair value measurement.
The second assumption the authors make is that the nature of
fair value gains and losses is homogenous within the three ?nan-
cial reporting categories, which are based on the display of fair
value measurements: amounts reported in net income, other com-
prehensive income, and note disclosures. Variation within these
categories is important to explore because both market and credit
risk modeling systems may have implications for both recognized
and disclosed amounts. We note that in Panel B of Table 4, the au-
thors begin to explore cross-sectional variation in the nature of fair
value gains and losses by examining speci?c amounts recorded in
other comprehensive income. In particular, they document a more
pronounced impact of market risk modeling for banks with above
median percentages of mortgage-backed and asset-backed securi-
ties in their available-for-sale portfolios. We suspect, however, that
models of market risk may also be important for understanding
risks corresponding to demand deposits and ?xed-rate mortgage
loans, which are both disclosed in the notes. Furthermore, compo-
nents of credit risk models may be important for banks that hold a
substantial amount of agency debt (i.e., securities issued by Fannie
Mae or Freddie Mac) or sovereign debt, which are both recorded in
other comprehensive income as available-for-sale securities. Future
research can attempt to link speci?c technologies, such as compo-
nents of risk modeling techniques, to the estimation of fair values
of speci?c ?nancial instruments.
3.2. Disentangling the risk modeling and disclosure effects
How does market and credit risk modeling improve the reliabil-
ity of fair value estimates, thus making them more decision-useful
to investors for valuation purposes? In the context of Bhat and
Ryan, there are two plausible sources of the documented enhance-
ment in the relation between stock returns and fair value gains and
losses. First, risk modeling may improve the faithful representa-
tion of estimated fair values through a reduction in estimation er-
ror.
6
Second, the joint disclosure of risk modeling techniques with
other contemporaneous disclosures related to fair value measure-
ment may result in fair value estimates that are more veri?able
and understandable to investors. Veri?ability and understandability
4
Market and credit risk modeling scores are based on the number of risk mod-
eling techniques a bank discloses. The ?ve market risk modeling techniques are
interest rate gap analysis, interest rate sensitivity analysis, Value-at-Risk analysis,
stress testing, and backtesting. The four credit risk modeling techniques are statis-
tical credit risk measurement, credit scoring, internal credit risk rating, and stress
testing.
5
The authors are careful to state the following two related assumptions. The ab-
sence of risk modeling disclosures does not preclude risk modeling from being un-
dertaken by a bank. However, more disclosure of risk modeling activities is assumed
to correspond to relatively higher intensities of risk modeling than less (or no) dis-
closure of these activities.
6
One way to think about measurement error is in terms of the signal-to-noise
ratio, where an increase in the faithful representation of estimated fair value gains
and losses decreases the noise component (i.e., the denominator) and, therefore,
increases the signal-to-noise ratio, making these estimates more decision-useful to
investors.
are two qualitative characteristics that enhance the relevance and
reliability of accounting information (FASB, 2010). Below we dis-
cuss the potentially confounded roles of risk modeling and disclo-
sure, and the challenge in empirically disentangling these effects.
The potential for risk modeling-related improvements in the
faithful representation of fair value estimates is plausible, but likely
indirect. For instance, both risk models and fair value estimates re-
quire managers to make assumptions. Risk modeling, even if pru-
dently implemented, does not eliminate managerial discretion over
model assumptions. Consequently, it is unclear why risk modeling
should produce inputs for estimating fair values that are consid-
ered more reliable than if risk modeling techniques were not used
to estimate fair values. In addition, risk modeling is concerned
with estimating the risks associated with a distribution of prob-
able outcomes, whereas fair value measurement is concerned with
calculating a point estimate of the price of an asset or liability.
While models used by ?nancial institutions to estimate both risk
and fair value likely rely on similar technologies, capabilities, and
input data, it is not clear how or why risk models should feed di-
rectly into the valuation models that produce fair value estimates.
We argue that the risk modeling proxies developed by Bhat
and Ryan are likely imperfectly capturing a higher-order construct,
which we consider to be discipline over and capabilities in fair
value estimation. Unfortunately, the authors are unable to rule out
the possibility that the disclosure of risk modeling activities is
correlated with unobserved fair value estimation capabilities that
more directly explain cross-sectional variation in the reliability of
fair value measurements. Future research in this area can further
explore the potentially omitted ?rm-speci?c factors that enhance
the usefulness of fair value estimates to investors and the extent
to which these factors interact. Such factors might include ?rm-
speci?c capabilities in and control over the quality of fair value
measurement, certain auditor characteristics, and incentive com-
pensation programs that mitigate managers’ incentives to bias fair
value estimates.
Assuming that risk modeling enhances the reliability of fair
value estimates, how is this enhancement identi?ed by investors?
The market and credit risk modeling techniques considered by the
authors are identi?ed and categorized based on disclosures pro-
vided in 10-K ?lings. Investors may also learn about the reliability
of fair value estimates from these risk modeling disclosures. As the
authors suggest, however, these disclosures may not provide su?-
cient information for investors to discern the implementation and
quality of banks’ risk modeling systems and the corresponding im-
pact on fair value measurement. For instance, Bhat and Ryan (pp.
15–16) note that:
1. Presumably the disclosure of risk-modeling activities implies
banks engage in the activities, because there does not appear
to be any reason for banks to lie in these disclosures, which
generally are too terse to divulge meaningful proprietary infor-
mation.
Therefore, while these disclosures may be useful for identify-
ing the presence of certain risk modeling techniques, they may
not provide information to investors that explain the observed re-
lation between returns and fair value gains and losses. This is
consistent with the insigni?cant estimated coe?cients on the risk
modeling disclosure measures in the regressions reported in Ta-
bles 4 through 6. Another possibility, however, is that investors
learn about the reliability of fair value measurements through con-
temporaneous disclosures that discuss, among other things, the
controls, processes, and procedures designed to mitigate fair value
measurement concerns (see Chung, Goh, Ng, & Yong, 2015). Un-
fortunately, the authors do not control for contemporaneous dis-
closures that could in?uence the relation between returns and fair
value gains and losses.
R.P. McDonough, C.M. Shakespeare / Accounting, Organizations and Society 46 (2015) 96–99 99
Absent a more compelling and direct link between risk mod-
eling techniques, fair value estimation, and the channel through
which investors learn about the impact of ?rms’ capabilities and
technologies in the estimation of fair values, we interpret Bhat and
Ryan’s results with caution.
7
In particular, it is di?cult to disen-
tangle the role of risk modeling as a contributor to the reliability
of fair value measurements and the impact of contemporaneous
disclosures on the relation between returns and fair values. Future
research can seek to better understand the fair value estimation
process and the mechanisms through which investors learn about
the reliability of fair values.
4. Concluding remarks
The merits of the fair value measurement attribute continue
to be a controversial policy issue, particularly following the re-
cent ?nancial crisis. A fundamental concern with fair value mea-
surements is ensuring that the estimates represent the underlying
economic transactions they are intended to represent, particularly
when active (or “thick”) markets for the underlying asset or liabil-
ity do not exist. To that end, Bhat and Ryan provide some evidence
on the potential role of risk management technologies in enhanc-
ing the reliability of fair value measurements.
Future research should aim to expand our understanding of the
complex process of estimating fair values. Furthermore, it is im-
portant to better identify the underlying and potentially interre-
lated set of factors and ?rm capabilities that explain variation in
the reliability of fair value measurements. Finally, uncovering the
channels through which investors learn about enhancements to
fair value measurement reliability is important for understanding
the role of improved fair value disclosures, particularly in the con-
text of opaque ?nancial institutions whose disclosure decisions are
not well understood.
Acknowledgements
We appreciate the helpful comments provided by Linda Myers,
Hun-Tong Tan (editor), and participants at the 2014 Accounting, Or-
ganizations, and Society Conference on Accounting Estimates. We
gratefully acknowledge ?nancial support provided by Deloitte &
Touche LLP. Ryan McDonough is grateful for the ?nancial support
provided by the Paton Accounting Fellowship. Catherine Shake-
speare is the Teitelbaum Research Scholar.
7
Participants at the 2014 Accounting, Organizations, and Society Conference on
Accounting Estimates expressed similar concerns about the nature of the links be-
tween risk modeling disclosures, the reliability of fair value estimates, and investors’
perceptions of the reliability of fair value estimates.
References
Barth, M. E. (1994). Fair value accounting: Evidence from investment securities and
the market valuation of banks. The Accounting Review, 69(1), 1–25.
Barth, M. E., Beaver, W. H., & Landsman, W. R. (1996). Value-relevance of banks’ fair
value disclosures under SFAS No. 107. The Accounting Review, 71(4), 513–537.
Barth, M. E., Beaver, W. H., & Landsman, W. R. (2001). The relevance of the value rel-
evance literature for ?nancial accounting standard setting: another view. Journal
of Accounting and Economics, 31(1–3), 77–104.
Beaver, W. H., & Venkatachalam, M. (2003). Differential pricing of components of
bank loan fair values. Journal of Accounting, Auditing, & Finance, 18(1), 41–68.
Bhat, G., & Ryan, S. G., (2015). The impact of risk modeling on the market percep-
tion of banks’ estimated fair value gains and losses for ?nancial instruments.
Accounting, Organizations, and Society, this issue.
Chung, S., Goh, B., Ng, J., & Yong, K. (2015). Voluntary fair value disclosures beyond
SFAS 157’s three-level estimates. Working paper, Singapore Management Univer-
sity.
Financial Accounting Standards Board (FASB) (2006). Fair Value Measurements. State-
ment of Financial Accounting Standards No. 157. Norwalk, CT: FASB.
Financial Accounting Standards Board (FASB) (2010). Conceptual Framework for Fi-
nancial Reporting. Chapter 3, “Qualitative Characteristics of Useful Financial In-
formation”. Statement of Financial Accounting Concepts No. 8. Norwalk, CT:
FASB.
Fischer, P. E., & Verrecchia, R. E. (2000). Reporting Bias. The Accounting Review, 75(2),
229–245.
Goh, B. W., Li, D., Ng, J., & Yong, K. O. (2015). Market pricing of banks’ fair value
assets reported under SFAS 157 since the 2008 ?nancial crisis. Journal of Ac-
counting and Public Policy, 34(2), 129–145.
Hodder, L. D., Hopkins, P. E., & Schipper, K. (2014). Fair Value Measurement in Fi-
nancial Reporting. Foundations and Trends in Accounting, 8(3–4), 143–270.
Holthausen, R. W., & Watts, R. L. (2001). The relevance of the value-relevance lit-
erature for ?nancial accounting standard setting. Journal of Accounting and Eco-
nomics, 31(1–3), 3–75.
Kadous, K., Koonce, L., & Thayer, J. M. (2012). Do ?nancial statement users judge
relevance based on properties of reliability? The Accounting Review, 87(4), 1335–
1356.
Koonce, L., Nelson, K. K., & Shakespeare, C. M. (2011). Judging the relevance of fair
value for ?nancial instruments. The Accounting Review, 86(6), 2075–2098.
Landsman, W. R. (2007). Is fair value accounting information relevant and reli-
able? Evidence from capital market research. Accounting and Business Research,
37(Special Issue), 19–30.
Laux, C., & Leuz, C. (2009). The crisis of fair-value accounting: Making sense of the
recent debate. Accounting, Organizations and Society, 34(6–7), 826–834.
Nelson, K. K. (1996). Fair value accounting for commercial banks: An empirical anal-
ysis of SFAS No. 107. The Accounting Review, 71(2), 161–182.
Petroni, K. R., & Whalen, J. M. (1995). Fair values of equity and debt securities
and share prices of property-liability insurers. The Journal of Risk and Insurance,
62(4), 719–737.
Song, C. J., Thomas, W. B., & Yi, H. (2010). Value relevance of FAS No. 157 fair value
hierarchy information and the impact of corporate governance mechanisms. The
Accounting Review, 85(4), 1375–1410.
Venkatachalam, M. (1996). Value-relevance of banks’ derivatives disclosures. Journal
of Accounting and Economics, 22(1–3), 327–355.
doc_130772355.pdf