Description
Weexaminewhetherandhowmeasuresofmarketandcreditriskmodelingidentifiedfrombanks’financialreportsenhancethereturns-relevanceoftheirestimatedannualunrealizedfairvaluegainsandlossesforfinancialinstruments.Tocapturedifferencesinmarketliquidityandfairvaluationdifficultiesacrosstypesoffinancialinstruments,wedistinguishunrealizedgainsandlossesthatarerecordedinnetincomeversusrecordedinothercomprehensiveincomeversuscalculableusingfinncialstatementnotedisclosures.Wepredictandgenerallyfindthatbanks’market(credit)riskmodelingenhancesthereturns-relevanceoftheirunrealizedfairvaluegainsandlosses,moresoforlessliquidinstrumentssubjecttogreatermarket-risk-related(credit-risk-related)valuationdifficultiesandduringperiodsforwhichmarket(credit)riskishigher.
Accounting, Organizations and Society 46 (2015) 81–95
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
The impact of risk modeling on the market perception of banks’
estimated fair value gains and losses for ?nancial instruments
R
Gauri Bhat
a,1
, Stephen G. Ryan
b,?
a
Cox School of Business, Southern Methodist University, 6212 Bishop Blvd., Dallas, TX 75275-0333, United States
b
Leonard N. Stern School of Business, New York University, 44 West 4th Street, Suite 10-73, New York, NY 10012-1118, United States
a r t i c l e i n f o
Article history:
Received 5 June 2014
Revised 15 April 2015
Accepted 16 April 2015
Available online 7 May 2015
a b s t r a c t
We examine whether and how measures of market and credit risk modeling identi?ed from banks’
?nancial reports enhance the returns-relevance of their estimated annual unrealized fair value gains and
losses for ?nancial instruments. To capture differences in market liquidity and fair valuation di?culties
across types of ?nancial instruments, we distinguish unrealized gains and losses that are recorded in net
income versus recorded in other comprehensive income versus calculable using ?nancial statement note
disclosures. We predict and generally ?nd that banks’ market (credit) risk modeling enhances the returns-
relevance of their unrealized fair value gains and losses, more so for less liquid instruments subject to
greater market-risk-related (credit-risk-related) valuation di?culties and during periods for which market
(credit) risk is higher. We obtain these ?ndings both for banks’ unadjusted risk modeling measures and
for the portions of these measures that we model as attributable to banks’ risk modeling activities, but
not for the portions we model as attributable to banks’ disclosure of these activities.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
We examine whether and how banks’ risk modeling enhances
the returns-relevance of their estimated annual unrealized fair
value gains and losses (FVGL) on ?nancial instruments. FVGL are
changes in fair value during periods that are not yet realized
through cash received or paid. When the markets for banks’ ?-
nancial instruments are su?ciently illiquid that observable market
inputs do not su?ce to determine the fair values of those instru-
ments, banks must estimate FVGL by developing valuation mod-
els and identifying the inputs necessary to implement those mod-
els. These activities require risk modeling both to predict uncertain
future cash ?ows and to determine appropriate rates to discount
those cash ?ows. To conduct risk modeling effectively, banks must
invest in adequate personnel and information systems and apply
managerial judgment appropriately and with discipline, with inad-
R
This paper was presented at the Accounting, Organizations & Society Confer-
ence on Accounting Estimates on October 23–25, 2014 sponsored by the Deloitte &
Touche LLP. We are grateful for helpful comments from the editor, Hun Tong Tan,
two anonymous reviewers, and conference participants, as well as accounting work-
shop participants at Rice University.
?
Corresponding author. Tel.: +1 (212) 998 0020.
E-mail addresses: [email protected] (G. Bhat), [email protected] (S.G. Ryan).
1
Tel.: +1 (214) 768 2964.
equate investment (self-interested application of judgment) intro-
ducing unintentional (intentional) noise and bias in FVGL. Banks’
investment in risk modeling and other risk management activities
that discipline fair value estimation appears to vary considerably
across banks and time.
2
Banks and their ?nancial instruments exhibit two primary types
of risk, market risk and credit risk. Banks engage in two corre-
sponding types of risk modeling activities, market risk modeling
(MRM) and credit risk modeling (CRM). Market risk is variability
in the value of a position attributable to changes in market prices.
Interest rate risk is the primary market risk for most banks. This
risk manifests through: (1) discounting effects, which are larger for
longer duration positions; (2) prepayment of ?xed-rate mortgage-
related assets (both securities and loans); and (3) the exercise of
other interest rate options, which may be standalone derivatives
or embedded in traditional ?nancial instruments. MRM involves
analyzing the durations of banks’ ?nancial instruments and the
resulting sensitivity of their net interest income and value of eq-
2
For example, Mikes (2011) discusses detailed case studies of two banks and ex-
tensive interviews at ?ve additional banks indicating that the quality of risk man-
agement varies considerably across banks and time. Only 28% (2%) of our sample
banks disclose in their Form 10-K ?lings that they employed a chief risk o?cer in
2013 (2002).http://dx.doi.org/10.1016/j.aos.2015.04.004
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
82 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
uity to interest rate movements. It also involves simulating the ef-
fects of interest rate movements on the prepayment of ?xed-rate
mortgages and exercise of other interest-rate options. Credit risk
is variability of the value of funded assets attributable to uncer-
tainty about default probabilities, losses given default, and timing
of default, as well as variability of the value of unfunded loan com-
mitments due to uncertainty about draws on those commitments,
which are more likely to occur during worse economic times. CRM
involves analysis of these parameters based on attributes of the
borrowers, borrowing contracts, borrowers’ performance to date on
the contracts, and relevant economic conditions.
We argue that banks’ MRM and CRM activities enhance the
quality of their estimates of FVGLs on ?nancial instruments when
two conditions hold: (1) the relevant markets for those instru-
ments are su?ciently illiquid that prices or other information from
these markets do not substantially determine the instruments’ fair
values; and (2) the instruments exhibit features, such as embed-
ded options or complex structuring, that increase the di?culty of
estimating the instruments’ fair values. As a ?rst cut to capture
the applicability of these conditions, we distinguish FVGL that are
recorded in net income versus recorded in other comprehensive in-
come versus calculable using ?nancial statement note disclosures
(c?disclosedd?). Fig. 1 summarizes current relevant U.S. generally
accepted accounting principles, under which FVGL are recorded in
net income for most trading and risk management instruments
and in other comprehensive income for available-for-sale securi-
ties and cash-?ow-hedge derivatives. FVGL are disclosed for most
of banks’ other primary types of ?nancial instruments, including
their largest asset, loans, and largest liability, deposits. We propose
three main hypotheses below that we test by examining whether
and how MRM and CRM enhance the returns-relevance of these
three types of FVGL from 2002 to 2013.
Our ?rst and most general hypothesis is that banks’ MRM and
CRM enhance the returns-relevance of their FVGL, more so for
FVGL on less liquid and more di?cult-to-fair-value ?nancial instru-
ments. In testing this hypothesis, we exploit the fact that banks’
?nancial instruments for which FVGL are disclosed, such as loans
and deposits, usually are less liquid and more di?cult to fair
value than their other instruments. Our second hypothesis is that
banks’ MRM also enhances the returns-relevance of their FVGL
recorded in other comprehensive income. Available-for-sale securi-
ties and cash-?ow-hedge derivatives typically are near credit risk-
less. Moreover, to the limited extent that banks experience credit
losses on these instruments, banks typically record these losses
in net income under impairment accounting rules. Hence, interest
rate risk is the primary risk re?ected in FVGL recorded in other
comprehensive income. We expect this hypothesis to hold only for
available-for-sale securities and cash-?ow-hedge derivatives that
are both less than highly liquid and exhibit fair valuation di?cul-
ties, such as mortgage-backed and asset-backed securities, so that
MRM is essential to estimate the fair values of the instruments
accurately. Our third hypothesis is that banks’ CRM primarily im-
pacts the returns-relevance of their disclosed FVGL, because banks
assume credit risk primarily through their funded loans and un-
funded loan commitments.
To test these hypotheses, we identify banks’ risk modeling ac-
tivities from disclosures in their Form 10-K ?lings. As described in
the Appendix c?Risk modeling measures and chief risk o?cer indi-
catord?, we hand collect disclosures of ?ve MRM activities (inter-
est rate gap analysis, interest rate sensitivity analysis, Value-at-Risk
analysis, stress testing, and backtesting) and four CRM activities
(statistical credit risk measurement, credit scoring, internal credit
risk rating, and stress testing). We equally weight these activities
to construct indices of banks’ MRM and CRM. This approach raises
the issue that many bank-year ?nancial reports include little about
risk modeling activities, particularly CRM early in our sample pe-
riod. Since all banks must engage in at least minimal levels of
MRM and CRM to make investment and ?nancing decisions and to
estimate the fair values of ?nancial instruments for which market
data do not su?ce for the task, it appears that some banks do not
disclose these activities. Hence, non-disclosure of a risk modeling
activity does not imply absence of the activity. We assume, how-
ever, that our MRM and CRM measures capture meaningful varia-
tion in risk modeling intensity across banks and time.
We test all hypotheses using both one-stage and two-stage ap-
proaches. The one-stage approach regresses returns for the twelve
months ending four months after the ?scal year end on net in-
come before FVGL recorded in net income
3
and the three types
of FVGL (recorded in net income, recorded in other comprehen-
sive income, and disclosed), separately and interacted with the un-
adjusted MRM and CRM measures, as well as control variables.
We frame and test our hypotheses as restrictions on the one or
more coe?cients on the interactions of banks’ unadjusted MRM
and CRM measures with speci?c types of FVGL. Empirical results
using this approach support our main hypotheses with one ex-
plainable exception.
We use the two-stage approach to help ensure that the one-
stage approach results are attributable to banks’ risk modeling ac-
tivities rather than to their choice to disclose these activities. In
this approach, we ?rst regress banks’ unadjusted MRM and CRM
measures on proxies for their discipline over risk modeling, techni-
cal sophistication, risk exposures, and risk tolerance, which we ex-
pect primarily indicate banks’ risk modeling activities rather than
their disclosure of those activities. We use the explained (unex-
plained) portions of banks’ unadjusted MRM and CRM measures
from these ?rst-stage models as measures of banks’ risk model-
ing activities (disclosure of these activities) in second-stage re-
turns models. The estimated coe?cients on the MRM and CRM
activity measures in the two-stage approach yield the same in-
ferences as the estimated coe?cients on the unadjusted measures
in the one-stage approach, whereas the estimated coe?cients on
the MRM and CRM disclosure measures generally are insigni?cant.
These results are consistent with the one-stage approach results
being driven by banks’ risk modeling rather than their disclosure
of that modeling.
We further hypothesize that MRM more strongly impacts the
returns-relevance of FVGL that are recorded in other comprehen-
sive income or disclosed in years with high interest rate volatility,
and that CRM more strongly impacts the returns-relevance of dis-
closed FVGL during the ?nancial crisis. To test these predictions,
we interact the primary test variables with indicator variables for
years with above-median interest rate volatility or the crisis period
2007–2009. Empirical results for the unadjusted MRM and CRM
measures and the MRM and CRM activity measures generally sup-
port these further hypotheses.
Our study contributes to the extensive literature beginning
with Barth (1994) that empirically examines the extent and de-
terminants of the value-relevance of fair values and the returns-
relevance of FVGL for ?nancial instruments. Our study is most re-
lated to recent papers examining disclosures of fair valuation in-
puts and other measures of the reliability of recognized fair value
estimates under Statement of Financial Accounting Standards (FAS)
157 (2006, Accounting Standards Codi?cation (ASC) 820), which
became effective in 2008. In particular, Chung, Goh, Ng, and Yong
3
Net income before FVGL recorded in net income includes realized gains and
losses that are distinct from FVGL except for two types of impairment write-down
that are included in net income and thus are accounted for in the same fashion
as realized losses. First, net income includes all or the credit loss portion of other-
than-temporary impairment write-downs of available-for-sale and held-to-maturity
securities. Second, net income includes losses on loans held for sale recognized at
fair value under the lower-of-cost-or-fair-value measurement basis.
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 83
Fig. 1. Summary of three ?nancial reporting treatments for fair value gains and losses on banks’ various types of ?nancial instruments.
(2014) develop measures of the reliability of recognized fair value
estimates from textual analysis of banks’ and insurers’ ?nancial
statement notes. Chung et al. provide evidence that these measures
are associated with enhanced market pricing and lower informa-
tion risk for recognized level 3 fair values. Our study complements
Chung et al. by examining an alternative measure of the reliability
of fair value estimates, a longer and more diverse sample period,
disclosed as well as recognized FVGL, and returns-relevance rather
than value-relevance.
Financial report information about estimated fair values of
?nancial instruments, literature review, and hypothesis
development
In this section, we ?rst describe the various bases in account-
ing standards, accounting research, and other sources for our main
hypotheses about the effects of MRM and CRM on the returns-
relevance of FVGL. We then support and state our further hypothe-
ses about the distinct conditions under which we expect MRM and
CRM to be particularly useful.
Required information about estimated fair values and FVGL
Under current U.S. generally accepted accounting principles,
?rms recognize certain types of ?nancial instruments at fair value
on the balance sheet. They record FVGL in net income for some of
these instruments and in other comprehensive income otherwise.
Banks’ primary types of ?nancial instruments currently recognized
at fair value with unrealized gains and losses recorded in net in-
come are: trading securities under FAS 115 (1993, ASC 320); non-
accounting-hedge and fair value hedge derivatives as well as fair
value hedged items under FAS 133 (1998, ASC 815); most other
trading instruments under industry accounting principles or prac-
tices; and ?nancial instruments for which the fair value option
is selected under FAS 155 (2006, ASC 815.15) and FAS 159 (2007,
ASC 825.10). The primary types of ?nancial instruments that cur-
rently are recognized at fair value with unrealized gains and losses
recorded in other comprehensive income are available-for-sale se-
curities under FAS 115 and cash-?ow-hedge derivatives under FAS
133.
4
We denote FVGL that are recorded in net income (other com-
prehensive income) by NIGL (OCIGL).
FAS 107 (1991, ASC 825.10.50) requires ?rms to disclose the fair
and carrying values of most types of ?nancial instruments in the
notes to ?nancial statements. FAS 107 disclosures are the only in-
formation in banks’ ?nancial reports about the fair values of their
on-balance sheet loans, deposits (excluding core deposit intangi-
bles), and debt as well as off-balance sheet instruments such as
loan commitments; FAS 115 requires these and other disclosures
for held-to-maturity securities. We denote FVGL that can only be
calculated from disclosures by DISCGL. FAS 107 also requires ?rms
to disclose the method(s) and signi?cant assumptions used to es-
timate the fair value of ?nancial instruments. Banks ful?lled this
requirement in boilerplate and minimally informative fashions un-
til the effective date of FAS 157.
FAS 157 de?nes fair value as c?the price that would be received
to sell an asset or paid to transfer a liability in an orderly trans-
action between market participants at the measurement date,d?
and it creates a three-level hierarchy of fair value inputs.
5
Level
1 (highest quality) inputs are quoted prices in active markets for
the identical item. Level 2 inputs are quoted prices in markets that
are not active for the identical item or in active markets for similar
items and most other observable information. Level 3 (lowest qual-
ity) inputs are unobservable reporting-?rm-supplied inputs. The
level of a fair value estimate is determined by the level of its low-
est quality signi?cant input. FAS 157, as amended by FASB Staff Po-
sition (FSP) FAS 157-4 (2009, ASC 820.10.50) and Accounting Stan-
dards Update (ASU) 2010-06 (2010, ASC 820.10.50), requires quar-
terly disclosures of the amounts of each major category of assets
and liabilities that is recognized at fair value on the balance sheet
4
Other standards require certain types of ?nancial instruments to be recognized
at fair value at either: (1) inception, e.g., FAS 166 (2009, ASC 860) for retained inter-
ests in securitizations; or (2) the time of (other than temporary) impairment write-
downs, e.g., FAS 65 (1982, ASC 948.310.35) and SOP 01-6 (2001, ASC 310.10.35) for
held-for-sale loans and FAS 115, as amended by FSP FAS 115-2 & FAS 124-2 (2009,
ASC 320.10.35), for available-for-sale and held-to-maturity securities.
5
Paragraphs C21–C22 of FAS 157 indicates that generally accepted accounting
principles include various practicability exceptions to fair value measurement that
FAS 157 did not change. While banks do not often invoke these exceptions in their
?nancial reports, the exceptions may have some effect on our FVGL measures or be
correlated with our risk modeling measures; e.g., a bank with better risk modeling
may have less need to invoke a practicability exception.
84 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
in each of the three levels, separately for items fair valued each
reporting period versus on a non-recurring basis. For level 2 and
level 3 fair values, ?rms must disclose their valuation techniques
and inputs. For level 3 fair values, ?rms must disclose rollforwards
of the fair values from the beginning to the end of the period, dis-
tinguishing total gains and losses, purchases, sales, issuances, set-
tlements, and transfers in and out of level 3.
6
Prior empirical research on the differential value-relevance of fair
values and returns-relevance of FVGL by type of ?nancial instrument
and by recognition versus disclosure
An extensive literature beginning with Barth (1994) empirically
examines the value-relevance of estimated fair values and returns-
relevance of FVGL on ?nancial instruments. Most of these studies
limit their samples to banks because ?nancial instruments dom-
inate banks’ balance sheets, banks’ ?nancial instruments exhibit
varying levels of liquidity and other factors associated with the
need or ability to exercise discretion over fair value estimation, and
analysis of a single industry mitigates concerns about unmodeled
heterogeneity. Given that several recent and extensive surveys of
this literature exist (e.g., Ryan, 2011, Section 4.5), we brie?y sum-
marize the four main ?ndings in these studies that bear on the
hypotheses developed below.
First, for investment securities and other similarly liquid assets,
fair values are value-relevant and FVGL are returns-relevant, al-
though the returns-relevance results are more sensitive to model
speci?cation and sample (Ahmed & Takeda, 1995; Barth, 1994; Car-
roll, Linsmeier, & Petroni, 2003; Danbolt & Rees, 2008). Second,
inconsistent (essentially no) evidence has been generated as to
whether the fair values of loans and derivatives (other ?nancial
instruments) are value-relevant (Barth, Beaver, & Landsman, 1996;
Eccher, Ramesh, & Thiagarajan, 1996; Nelson, 1996; Venkatacha-
lam, 1996). Third, banks’ disclosed fair values of ?nancial instru-
ments appear to be both noisy (i.e., unexplainable based on con-
temporaneous economic events) and managed to make less sol-
vent banks appear more so. Surprisingly, the value-relevance of
disclosed fair values does not appear to be affected by noise, but
it is reduced by discretionary behavior (Beaver & Venkatchalam,
2003; Nissim, 2003). Fourth, fair values that are recognized (i.e.,
more prominent) are more value-relevant than those that are dis-
closed (Ahmed, Kilic, & Lobo, 2006; Badertscher, Burks, & Eas-
ton, 2014; Chambers, Linsmeier, Shakespeare, & Sougiannis, 2007;
Dong, Ryan, & Zhang, 2014; Hirst, Hopkins, & Wahlen, 2004). Col-
lectively, these ?ndings suggest that estimated fair values and
FVGL for which the signal-to-noise ratio is (perceived to be) higher
have higher value-relevance and returns-relevance, respectively.
We examine returns-relevance rather than value-relevance be-
cause the latter is both more extensively studied in the prior lit-
erature and more subject to omitted variables (but less subject
to measurement error). Barth, Beaver, and Landsman (2001) and
Holthausen and Watts (2001) discuss these methodological trade-
offs.
6
While a massive improvement over FAS 107 disclosures, FAS 157 disclosures ex-
hibit three limitations that make them unsuitable for use in our study. First, these
disclosures are available for the bulk of our sample beginning in 2008, less than half
our sample period and a period dominated by the ?nancial crisis and its aftermath.
Hence, controlling for these disclosures would essentially eliminate our ability to
compare the crisis and non-crisis periods. Second, the most important FAS 157 dis-
closures pertain to level 3 recognized fair values for ?nancial instruments, which
constitute only about 2% of banks’ recognized ?nancial instruments. Third, these
disclosures do not encompass disclosed fair values for the bulk of banks’ ?nancial
instruments.
Prior empirical research on the differential value-relevance of fair
values of ?nancial instruments by FAS 157 level and other reliability
disclosures
Using FAS 157-required disclosures, several studies provide ev-
idence that banks’ recognized fair values of ?nancial instruments
estimated using higher quality inputs are more value-relevant
(Goh, Li, Ng, & Yong, 2015; Kolev, 2009; Song, Thomas, & Yi, 2010).
These studies all ?nd that level 1 and 2 fair values are more value-
relevant than level 3 fair values. Goh et al. ?nd that level 1 fair
values are more value-relevant than level 2 fair values. Song et al.
?nd that level 3 fair values are more value-relevant for banks with
better corporate governance.
Chung et al. (2014) develop binary (disclosure made versus not)
and continuous (number of words in the disclosure divided by to-
tal number of words in the Form 10-K ?ling) measures of FAS 157-
encouraged disclosures in banks’ and insurers’ ?nancial statement
notes about the controls, processes, and procedures used to assure
the reliability of their recognized fair value estimates. Chung et al.
(2014) provide evidence that these measures are associated with
enhanced market pricing (higher share price and lower priced risk)
and lower information risk (higher analyst consensus) for recog-
nized fair values estimated using signi?cant level 3 inputs.
Collectively, these ?ndings suggest that disclosures indicating
that banks more reliably estimate the fair value of their ?nancial
instruments are associated with enhanced value-relevance of these
estimates.
Hypotheses
We discuss banks’ MRM and CRM activities in the introduc-
tion and repeat this discussion here only insofar as it pertains di-
rectly to speci?c hypotheses. All of our hypotheses re?ect the view
that risk modeling typically improves fair value estimation. We ac-
knowledge that risk modeling may instead deteriorate fair value
estimation if it crowds out the appropriate use of judgment or de-
volves into a compliance exercise, as Mikes (2011) and Kaplan and
Mikes (2012) discuss occurred at speci?c banks.
Our ?rst and most general hypothesis is that banks’ MRM and
CRM enhance the returns-relevance of their FVGL, particularly for
less liquid ?nancial instruments that exhibit lower quality mar-
ket information (i.e., less c?price transparencyd?) and thus greater
need for risk modeling to estimate FVGL reliably. This hypothe-
sis is supported by four prior theoretical and empirical literatures:
(1) the theoretical literature indicating that market prices respond
more strongly to more precise information (e.g., Holthausen & Ver-
recchia, 1988); (2) the empirical literature on the value-relevance
of fair values and returns-relevance of FVGL by type of ?nancial
instruments and by recognition versus disclosure discussed above;
(3) the empirical literature that demonstrates that FAS 157 fair
value input level and voluntary fair value reliability disclosures af-
fect this value-relevance discussed above; and (4) Bhat, Ryan, and
Vyas’s (2014) ?ndings that banks’ CRM is positively associated with
the timeliness of their loan loss provisions. We formally state this
hypothesis and all subsequent hypotheses in alternative form.
H1. MRM and CRM enhance the sensitivity of returns to FVGL,
more so for FVGL on less liquid and more di?cult-to-fair-value ?-
nancial instruments.
In testing Hypothesis H1, we exploit the fact that banks’ ?nan-
cial instruments generating DISCGL, such as loans and deposits,
usually are less liquid and more di?cult to fair value than their
other ?nancial instruments. As discussed below, the ?nancial in-
struments that generate OCIGL are a mixed bag of highly liquid
and thus easy-to-fair-value instruments and less liquid and more
di?cult-to-fair-value instruments. The instruments that generate
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 85
NIGL usually are highly liquid and easy to fair value. Hence, we ex-
pect Hypothesis H1 to hold most strongly for DISCGL, less strongly
for OCIGL, and not at all for NIGL.
Our second hypothesis is that banks’ MRM enhances the
returns-relevance of their OCIGL, particularly for mortgage-backed
and asset-backed securities. This hypothesis has three interrelated
bases. First, interest rate risk is the primary risk re?ected in OCIGL,
which are generated by available-for-sale securities and cash-?ow-
hedge derivatives. Banks’ most common types of available-for-sale
securities are U.S. Treasuries, other U.S. federal government se-
curities, and agency-guaranteed mortgage-backed securities. Their
most common type of cash-?ow-hedge derivatives are interest rate
derivatives with highly creditworthy counterparties. These instru-
ments typically are near credit riskless. Moreover, to the limited
extent that banks experience credit losses on these instruments,
they typically record these losses in net income under impairment
accounting rules.
Second, interest rate risk manifests through: (1) discounting,
which more strongly affects longer duration positions; (2) pre-
payment of ?xed-rate mortgages underlying mortgage-backed and
other asset-backed securities; and (3) the exercise of other interest
rate options, which may be standalone cash-?ow-hedge derivatives
or embedded in available-for-sale securities (Ryan, 2011, Section
2.4). Banks must use considerably more sophisticated MRM (e.g.,
interest rate simulation) to determine the effects of prepayment
and other interest rate options than to determine discounting ef-
fects. Relatedly, it is considerably more di?cult for banks to hedge
options than discounting effects.
Third, compared to banks’ other types of available-for-sale secu-
rities, mortgage-backed securities and other types of asset-backed
securities typically are both less liquid and subject to greater
fair valuation di?culties, due to their embedded prepayment
options and complex structuring (e.g., waterfalls). Even agency-
guaranteed mortgage-backed securities generally exhibit low price
transparency unless and until speci?c securities are identi?ed in
to-be-announced trades (Vickrey & Wright, 2013).
7
H2. MRM enhances the sensitivity of returns to OCIGL, especially
OCIGL on less liquid and more di?cult-to-fair-value mortgage-
backed and asset-backed available-for-sale securities.
Our third hypothesis is that banks’ CRM primarily impacts the
returns-relevance of their DISCGL. This hypothesis is motivated
by the fact that banks assume credit risk primarily through their
funded loans and unfunded loan commitments for which esti-
mated fair values are disclosed (Ryan, 2011, Section 2.5).
H3. CRM primarily enhances the sensitivity of returns to DISCGL.
Our fourth and ?fth hypotheses are motivated by the intuition
that banks’ risk modeling should be more important during peri-
ods when the modeled risk is higher. In particular, MRM (CRM)
should be more important during years when interest rates are
more variable (credit losses are unexpectedly higher).
H4. The effect of MRM is stronger during years with above-median
interest rate volatility than in other years.
7
Vickrey and Wright (2013) provide evidence that to-be-announced agency
mortgage-backed securities trades (which are accounted for as derivatives, not in-
vestment securities, until settled) exhibit high volume and price transparency. They
also explain, however, that these trades exhibit various features (e.g., settlement
at a single date each month and the cheapest-to-deliver option) that reduce the
price transparency these trades provide for existing pools of agency-guaranteed
mortgage-backed securities.
H5. The effect of CRM is stronger during the ?nancial crisis than
in other years.
Measurement of variables and methodology
Fair value gains and losses
We measure NIGL as trading revenue reported on line 5.c of
Schedule HI and OCIGL as other comprehensive income reported
on line 4 of Schedule HC-R of Y-9C ?lings. Unfortunately, both NIGL
and OCIGL include items other than FVGL; NIGL includes fee in-
come and realized gains and losses and OCIGL includes other com-
prehensive income from pensions and foreign currency.
8
For ?nan-
cial assets (liabilities) for which fair values are disclosed, we cal-
culate DISCGL as (minus) the change during the year of the differ-
ence between the fair value and carrying value of the instrument.
9
We obtain disclosed fair and carrying values from SNL Financial.
All FVGL components are after taxes calculated using a statutory
tax rate of 35% and divided by beginning-of-year market value of
equity.
Unadjusted risk modeling measures
To construct our MRM and CRM measures, we had to decide
which risk modeling activities to include in the measures and also
the disclosure medium to examine for these activities. To make
the ?rst decision, we considered all of the market and credit risk
modeling activities (including related risk management processes)
mentioned in a 2001 public disclosure survey of large interna-
tional banks conducted by the Basel Committee on Banking Su-
pervision.
10
To accommodate the U.S. banks in our sample, we ex-
cluded risk modeling activities for which disclosures are required
for U.S. banks or that are not provided by any sample bank in any
sample year. This yields ?ve market risk modeling activities (inter-
est rate gap analysis, interest rate sensitivity analysis, Value-at-Risk
analysis, stress testing, and backtesting) and four credit risk mod-
eling activities (statistical credit risk measurement, credit scoring,
internal credit risk rating, and stress testing). We describe these
activities in detail in the Appendix c?Risk modeling measures and
chief risk o?cer indicatord?.
We choose banks’ annual Form 10-K ?lings as the sole disclo-
sure medium for three reasons. First, banks provide the most ex-
tensive disclosures of their risk modeling activities in these ?lings,
primarily due to SEC requirements. For example, SEC FRR 48 re-
quires that banks make extensive disclosures of their market risk
in these ?lings. Second, the extents of disclosure in Form 10-K ?l-
ings and other media are positively correlated (Lang & Lundholm,
1993). Third, our analysis of samples of banks’ ?nancial analyst re-
ports and conference call transcripts yielded no mention of risk
modeling.
We hand collected banks’ risk modeling disclosures from their
Form 10-K ?lings from 2001 to 2013. We electronically searched
8
In untabulated analysis, we tried to remove fee income from NIGL by subtract-
ing its average over the prior three or ?ve years. This transformation of NIGL should
mostly remove fee income, which is much more persistent than fair value gains and
losses. No inferences are affected by this alternative speci?cation of NIGL.
9
Because provisions for credit losses on loans and loan commitments are
recorded in net income, these recognized accruals are not included in DISCGL. This
likely reduces the power of our tests of hypotheses H1, H3, and H5, because prior
research shows that CRM is associated with these accruals (Bhat et al., 2014).
10
While these documents also consider operational risk modeling activities, we
do not deem this type of activity to be directly related to the estimation of fair
values and FVGL for ?nancial instruments. We acknowledge that Chernobai, Jorion,
and Yu (2011) provide evidence that operational risk is positively correlated with
credit risk.
86 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
those ?lings for a large number of relevant words or phrases, such
as c?risk management,d? c?risk model,d? c?scoringd? and c?risk
ratingd?. We read each identi?ed disclosure to verify that the risk
modeling activity was disclosed. We construct an indicator for each
activity that takes a value of one if it is disclosed and zero other-
wise. Our unadjusted market risk modeling measure, MRM (credit
risk modeling measure, CRM), is the simple average of the indica-
tors for the ?ve market risk activities (four credit risk activities).
As averages of indicator variables, MRM and CRM take values from
zero to one.
Several caveats are in order regarding our risk modeling mea-
sures. First, we observe banks’ disclosures of risk modeling activ-
ities, not the activities themselves. Presumably the disclosure of
risk-modeling activities implies banks engage in the activities, be-
cause there does not appear to be any reason for banks to lie in
these disclosures, which generally are too terse to divulge mean-
ingful proprietary information.
11
However, the non-disclosure of an
activity does not imply that the bank does not engage in the activ-
ity. As discussed in the introduction and below, we employ a two-
stage approach to help ensure that our results are attributable to
banks’ risk modeling activities rather than their disclosure of those
activities. Second, our risk modeling measures capture the exis-
tence of particular activities more than the quality of those activi-
ties. Third, and relatedly, despite robustness checks to incorporate
various different weighting approaches, simple averaging of the in-
dicators for risk modeling activities might not re?ect the weights
investors assign to these activities. Lastly, despite careful reading
of each disclosure, our coding of risk modeling activities might be
subjective. The large number of disclosures involved makes coding
by multiple individuals costly, however.
Chief risk o?cer indicator
We use an indicator variable for whether banks employ a chief
risk o?cer, CRO, as our measure of banks’ discipline over risk mod-
eling. We considered using standard proxies for corporate gover-
nance for this purpose, but these proxies do not appear to cor-
respond strongly, if at all, with the estimation of fair values and
FVGL.
12
Consistent with this view, Aebi, Sabato, and Schmid (2012)
?nd that these standard proxies c?are mostly insigni?cantly or
even negatively related to the banks’ performance during the cri-
sis,d? whereas banks that employ chief risk o?cers that report di-
rectly to the board of directors exhibit superior performance dur-
ing the crisis. Ellul and Yerramilli (2013) ?nd similar results for
their risk management index, which depends strongly on the em-
ployment and status of the chief risk o?cer.
CRO takes a value of one if the bank discloses that it employs a
chief risk o?cer in its Form 10-K ?ling and zero otherwise. We
searched ?lings to determine whether the words c?chiefd? and
c?riskd? appear within ?ve words of each other and, if so, whether
the words c?riskd? and either c?o?cerd? or c?directord? appear
within ?ve words of each other. For ?lings that met both ?lters,
we read the relevant passages to con?rm the existence of a chief
11
In addition, disclosures of risk modeling are subject to internal controls, to CEO
and CFO certi?cation of ?nancial reports under Section 302 of the Sarbanes–Oxley
Act, and to audit (reading for consistency) by banks’ auditors if the disclosures
appear in the ?nancial statement notes (MD&A) sections of ?nancial reports. The
observed frequency of disclosures is fairly low, inconsistent with easily imitated
c?cheap talk.d?
12
More generally, Larcker, Richardson, and Tuna (2007) argue that construct valid-
ity issues with corporate governance measures cause mixed and hard-to-interpret
results in many studies, in part due to the absence of theory indicating which cor-
porate governance variables matter in what contexts.
risk o?cer. CRO takes a value of one for these bank-year observa-
tions and zero otherwise.
13
First-stage models explaining risk modeling
In the two-stage approach, the ?rst-stage models regress each
of the unadjusted MRM and CRM measures (collectively denoted
XRM) on seven proxies for banks’ discipline over risk modeling,
technical sophistication, risk exposures, and risk tolerance, all of
which we expect primarily to affect the extent of banks’ risk mod-
eling activities rather than their disclosure of those activities. These
proxies are: CRO, a proxy for banks’ discipline over risk mod-
eling; log of total assets (SIZE) and a trading portfolio indica-
tor (TRADD), proxies for banks’ technical sophistication; 0–1 year
repricing gap (INTSEN), the proportion of assets that are commer-
cial loans (COMMLOAN), and the standard deviation of fair value
income de?ated by beginning market value of equity over the past
ten years (FVISTD), proxies for banks’ interest rate risk, credit risk,
and overall risk, respectively; and tier 1 risk-based capital ratio
(TIER 1), a proxy for banks’ solvency or risk tolerance. We also in-
clude year ?xed effects to capture changes in the risk modeling
measures, particularly CRM, over time. This model is:
XRM = ? +?
1
CRO +?
2
SIZE +?
3
TRADD +?
4
INTSEN
+?
5
COMMLOAN +?
6
FVISTD +?
7
TIER1
+year ?xed effects +? (1)
We suppress time subscripts except when necessary for clarity.
We use the portions of the risk modeling measures explained by
the seven proxies, denoted with hats, i.e.,
XRM, as measures of
risk modeling activities. We use the remaining portions of the risk
modeling measures, i.e., XRM ?
XRM, as measures of risk modeling
disclosures.
14
Models of the impact of risk modeling on the
returns-relevance of FVGL and hypotheses restated as
coe?cient restrictions
The returns models using the unadjusted risk modeling mea-
sures (i.e., in the one-stage approach) regress share returns for the
12 months ending four months after the ?scal year end (R)
15
on
net income before FVGL recorded in net income (NIBNIGL), the
three types of FVGL (NIGL, OCIGL, and DISCGL), the risk modeling
measure under consideration (either MRM or CRM), interactions
between that risk modeling measure and each of NIBNIGL and the
13
Using data available on banks’ top ?ve o?cers from SNL Financial, we endeav-
ored to determine whether CROs were one of these o?cers, but this determination
cannot be made from these data.
14
In untabulated robustness analyses, we use an alternative ?rst stage approach
that does not alter any inference in the second stage. Speci?cally, we include ?rm
attributes, that the disclosure literature surveyed by Healy and Palepu (2001) ?nds
to be associated with disclosure, as additional explanatory variables in Eq. (1): an
indicator for issuance of debt or equity, i.e., capital market transactions that mo-
tivate disclosure; an indicator for negative net income, a proxy for performance;
the book-to-market ratio, a proxy for growth; number of analysts forecasting earn-
ings and percentage institutional ownership, proxies for information environment.
We use the portions of the risk modeling measures explained by these proxies as
measures of risk modeling disclosures.
15
We use this return window because some sample banks are non-accelerated
?lers that ?le their Form 10-K ?lings on the last day of the third month following
the end of the ?scal year.
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 87
three types of FVGL, and year ?xed effects.
16
These models are:
R = a + b
1
NIBNIGL + b
2
NIGL + b
3
OCIGL + b
4
DISCGL
+b
5
(NIBNIGL ? XRM) + b
6
(NIGL ? XRM) + b
7
(OCIGL ? XRM)
+b
8
(DISCGL ? XRM) + b
9
XRM +year ?xed effects + e (2A)
The returns models using the risk modeling activity and disclosure
measures from the ?rst-stage model Eq. (1) (i.e., in the two-stage
approach) follow directly from substituting
XRM and XRM ?
XRM
for XRM in Eq. (2A). These models are:
R = a + b
1
NIBNIGL + b
2
NIGL + b
3
OCIGL + b
4
DISCGL
+b
5
(NIBNIGL ?
XRM) + b
6
(NIGL ?
XRM) + b
7
(OCIGL ?
XRM)
+b
8
(DISCGL ?
XRM) + b
9
XRM + b
10
(NIBNIGL ? {XRM ?
XRM})
+b
11
(NIGL ? {XRM ?
XRM}) + b
12
(OCIGL ? {XRM ?
XRM})
+b
13
(DISCGL ? {XRM ?
XRM}) + b
14
{XRM ?
XRM}
+year ?xed effects + e (2B)
We restate Hypotheses H1–H5 as coe?cient restrictions in Eqs.
(2A) and (2B). Hypothesis H1 predicts that b
8
is positive and
larger than b
6
and b
7
for both
MRM and
CRM. Hypothesis H2 pre-
dicts that b
7
is positive for
MRM, especially when OCIGL are for
mortgage-backed and asset-backed securities. Hypothesis H3 pre-
dicts that the primary effect of
CRM is manifested in b
8
. Hy-
pothesis H4 predicts that Hypotheses H1 and H2 pertaining to
MRM hold more strongly in years with above-median interest rate
volatility. Hypothesis H5 predicts that Hypotheses H1 and H3 per-
taining to
CRM hold more strongly during the ?nancial crisis.
Sample selection and data
Sample selection
We study banks for the same reasons as the prior literature
discussed in section ‘Financial report information about estimated
fair values of ?nancial instruments, literature review, and hypoth-
esis development’. In addition, the cost of hand collecting our risk
modeling and CRO variables limits the sample size.
As summarized in Table 1, Panel A, the sample comprises all
public U.S. bank holding companies with data available from 2002
to 2013 from the following ?ve sources: most ?nancial data from
Y-9C regulatory ?lings on the Federal Reserve Bank of Chicago
website, disclosed fair values and carrying values of ?nancial in-
struments from SNL Financial, risk modeling disclosures from Form
10-K ?lings on 10-K Wizard, and stock information from CRSP.
17
The total assets threshold at which bank holding companies are
16
In Eqs. (2A) and (2B), we interact NIBNIGL with the unadjusted risk modeling
measures because certain components of NIBNIGL should be affected by risk mod-
eling, e.g., provisions for loan losses should be affected by CRM. We expect FVGL
to be more strongly affected than NIBNIGL by risk modeling, however; to this ex-
tent, the interaction of NIBIGL with the unadjusted risk modeling measures can be
viewed as a placebo test.
17
We start the sample in 2002 because FAS 133 became effective in 2001 and
two consecutive years of disclosed fair and carrying values are necessary to calcu-
late DISCGL. FAS 133 requires all derivatives to be recognized at fair value on the
balance sheet, with unrealized gains and losses recorded in net income for non-
accounting hedge and fair-value-hedge derivatives and in other comprehensive in-
come for cash-?ow-hedge derivatives. Compared to prior accounting principles and
accounting practices, FAS 133 signi?cantly increased the amount of FVGL recorded
in net income and decreased the amount of FVGL disclosed; the standard also af-
fected the FVGL recorded in other comprehensive income in a less obviously direc-
tional fashion.
required to ?le Y-9Cs increased from $150 million to $500 million
in 2006; we impose the latter size restriction for all bank-year ob-
servations to yield a more comparable sample through time. Some
bank holding companies that ?le Y-9Cs differ slightly from the cor-
responding public companies that ?le Form 10-Ks.
Table 1, Panel A reports the construction of the ?nal sample of
238 unique banks and 2413 bank-year observations for the years
2002–2013. To mitigate the effects of outliers, we winsorize all
continuous model variables at the 0.5% and 99.5% levels of their
distributions.
Descriptive statistics
Table 1, Panel B presents descriptive statistics for the main vari-
ables. As is typical in banking studies, the size-related variables
exhibit considerable variation and right skewness. Despite the ?-
nancial crisis, the vast majority of bank-year observations are prof-
itable and well-capitalized, with the 25th percentile of NIBNIGL
equaling 4.9% of beginning-of-year market value of equity and of
the Tier 1 risk-based capital ratio equaling 10.3%. The distribution
of R is quite spread, however, re?ecting that fact that banks expe-
rienced good stock performance both early and late in the sample
period and poor performance during the crisis.
All three types of FVGL have means and medians of approxi-
mately zero. The spread of DISCGL is much higher than the spread
of NIGL and, to a lesser extent, OCIGL, however, because fair values
for banks’ primary ?nancial asset (loans) and ?nancial liability (de-
posits) are disclosed and only 21.8% of banks (mean TRADD) hold
any trading positions. The median bank-year discloses two out of
?ve MRM activities and one out of four CRM activities. 13.6% of
bank-years employ a chief risk o?cer.
Table 1, Panel C reports the number of observations, the means
of MRM, CRM, and CRO, and the frequency with which each of
the indicators for the risk modeling activities underlying MRM and
CRM take a value of one in each year from 2002 to 2013. At least
170 banks appear in the sample each year. The number of banks
is highest in the middle sample years, peaking at 221 in 2009. The
lesser number of banks in the early (later) years is attributable to
our requirement that observations have at least $500 million total
assets (bank failures and mergers and acquisitions resulting from
the crisis and passage of time). The frequency of banks’ disclosures
of MRM activities is ?at over the sample period due to the 1998
effective date of the main market risk disclosure requirement, FRR
48. In contrast, the frequency of banks’ disclosures of CRM activi-
ties and CRO increases strongly over the sample years.
Table 2 reports the Pearson (Spearman) correlations for the
main variables in the regression; we discuss only the Pearson
correlations. Three insights emerge from this table. First, MRM
and CRM are both signi?cantly positively associated with most of
the variables capturing size and sophistication, such as CRO, SIZE,
NIGL, and TRADD. Second, there are interpretable differences be-
tween the correlations involving MRM and CRM. For example, NIGL
is more positively correlated with MRM than with CRM, consis-
tent with trading-oriented banks primarily trading ?nancial instru-
ments with more market risk than credit risk. In contrast, TRADD
is more positively correlated with CRM than MRM, consistent with
large and sophisticated banks being relatively more likely to con-
duct CRM. Consistent with banks’ chief risk o?cers primarily being
tasked with evaluating credit risk, CRO is more positively corre-
lated with CRM than MRM. Third, unlike NIGL, neither OCIGL nor
DISCGL is correlated with any of the risk modeling measures or
measures of bank size or sophistication.
88 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Table 1
Sample selection and descriptive statistics.
#Bank-years #Banks
Panel A: Sample selection
Bank-years and banks with Y-9C data and total assets >$500 million in 2002–2013 3664 443
Above bank-years and banks with fair value data from SNL Financial 3282 405
Above bank-years and banks with CRSP data 3244 405
Above bank-years and banks with Form 10-K ?lings on 10-K Wizard 2437 239
Final sample bank-years and banks in 2002–2013 with available variables in at least two consecutive years in 2001–2013 (to allow one lag) 2413 238
Variable Mean Stddev Q1 Median Q3
Panel B: Summary statistics for 238 banks from 2002 to 2013 (2413 bank-years)
R 0.060 0.326 ?0.097 0.058 0.224
NIBNIGL 0.004 0.315 0.049 0.067 0.081
NIGL 0.001 0.007 0.000 0.000 0.000
OCIGL ?0.001 0.036 ?0.011 ?0.001 0.009
DISCGL ?0.001 0.224 ?0.034 0.000 0.034
MRM 0.314 0.145 0.200 0.400 0.400
CRM 0.211 0.209 0.000 0.250 0.250
CRO 0.136 0.342 0.000 0.000 0.000
SIZE 14.730 1.531 13.655 14.342 15.388
TRADD 0.218 0.413 0.000 0.000 0.000
INTSEN 1.699 1.987 0.528 1.174 2.075
COMMLOAN 0.121 0.074 0.066 0.105 0.160
FVISTD 0.269 0.338 0.079 0.139 0.312
TIER1 12.328 2.923 10.310 11.870 13.880
Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Panel C: Number of observations, risk modeling measures, and CRO by year
N 170 189 198 215 219 217 216 221 207 194 188 179
Means of risk modeling measures and CRO
MRM 0.327 0.324 0.320 0.316 0.315 0.313 0.311 0.309 0.308 0.304 0.312 0.316
CRM 0.113 0.135 0.145 0.155 0.156 0.162 0.175 0.215 0.301 0.323 0.334 0.332
CRO 0.020 0.042 0.061 0.070 0.087 0.115 0.139 0.163 0.193 0.232 0.234 0.279
MRM components
INTGAP 135 149 147 155 154 148 143 142 131 113 110 104
IRSIMUL 131 146 159 177 176 176 176 182 170 157 156 150
VAR 8 7 8 8 9 8 7 7 7 8 10 10
MRSTRESS 1 1 1 3 3 5 5 5 5 9 9 9
MRBACKTEST 3 3 2 2 3 3 4 5 6 8 8 10
CRM components
MODEL 9 13 16 22 23 22 24 35 48 48 45 45
SCORING 13 17 18 17 19 21 18 22 23 24 27 25
RR 53 71 79 91 91 93 97 113 158 147 144 136
CRSTRESS 2 1 2 3 4 5 12 20 20 32 35 32
Notes: Panel A reports the sample selection process. Panel B reports summary statistics for the primary model variables, including net income before fair value gains
and losses recorded in net income (NIBNIGL), fair value gains and losses recorded in net income (NIGL), fair value gains and losses recorded in other comprehensive
income (OCIGL), disclosed fair value gains and losses (DISCGL), the risk modeling measures (MRM, and CRM), chief risk o?cer (CRO), and other bank characteristics.
Panel C reports the means of the risk modeling measures and CRO by year, as well as and the number of banks that disclose of the components of each risk modeling
measure by year. All variables are de?ned in the Appendix c?De?nitions of variables and other acronymsd? except for the risk modeling components which are de?ned
in the Appendix c?Risk modeling measures and chief risk o?cer indicatord?.
Empirical analysis
First-stage regression results
Table 3 reports the estimation of Eq. (1), the ?rst-stage models
used to explain the unadjusted risk modeling measures in terms
of risk-related explanatory variables and ?xed year effects (untab-
ulated). The ?rst (second) column presents the results for the un-
adjusted market (credit) risk modeling measure MRM (CRM). The
?t is considerably better for the CRM model, primarily because the
year ?xed effects are more signi?cant in this model due to CRM’s
strong upward trend over time reported in Table 1, Panel C.
The coe?cient on CRO is signi?cantly positive at the 5% level
in the MRM model and at the 10% level in the CRM model. The
coe?cient on SIZE is signi?cantly positive at the 1% level in both
models. These results indicate that banks’ discipline over fair value
estimation and technical sophistication explain their risk modeling.
In the MRM model, the coe?cient on INTSEN is signi?cantly neg-
ative at the 5% level and the coe?cient on FVISTD is signi?cantly
negative at the 10% level, suggesting that banks with higher inter-
est rate risk tend to seek out or tolerate rather than manage that
risk. In contrast, the coe?cient on FVISTD is positive at the 10%
level in the CRM model. No other explanatory variable is signi?-
cant in any model.
Primary regression results
Table 4, Panel A reports the estimations of Eqs. (2A) and (2B),
which we use to test hypotheses H1–H5 about whether and how
banks’ risk modeling impacts the returns-relevance of their three
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 89
Table 2
Correlations.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
R (1) 0.485 0.069 ?0.085 ?0.026 0.009 0.071 0.012 ?0.016 0.018 0.113 0.016 ?0.104 0.195
0.000 0.001 0.000 0.194 0.675 0.001 0.572 0.435 0.372 0.000 0.440 0.000 0.000
NIBNIGL (2) 0.201 ?0.076 ?0.049 0.018 ?0.050 ?0.011 ?0.019 ?0.100 ?0.052 ?0.018 ?0.048 ?0.132 0.214
0.000 0.000 0.016 0.385 0.014 0.575 0.354 0.000 0.011 0.375 0.019 0.000 0.000
NIGL (3) 0.106 ?0.068 ?0.004 0.020 0.077 0.209 0.202 0.400 0.669 0.177 0.182 ?0.098 ?0.097
0.000 0.001 0.836 0.316 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OCIGL (4) 0.044 ?0.079 0.067 0.067 ?0.003 ?0.001 ?0.020 ?0.016 ?0.005 0.038 0.004 0.016 ?0.003
0.031 0.000 0.001 0.001 0.894 0.968 0.317 0.445 0.815 0.064 0.846 0.445 0.900
DISCGL (5) 0.017 ?0.096 0.039 0.073 ?0.001 0.033 0.018 0.013 0.031 0.037 0.018 0.000 0.003
0.416 0.000 0.054 0.000 0.963 0.104 0.383 0.512 0.128 0.066 0.382 0.984 0.881
MRM (6) 0.009 0.024 0.303 0.002 0.005 0.173 0.074 0.094 0.076 ?0.011 0.051 ?0.010 ?0.045
0.648 0.239 0.000 0.941 0.793 0.000 0.000 0.000 0.000 0.603 0.012 0.611 0.028
CRM (7) 0.066 ?0.029 0.254 0.003 0.038 0.209 0.229 0.330 0.240 0.144 0.030 ?0.009 0.085
0.001 0.161 0.000 0.888 0.062 0.000 0.000 0.000 0.000 0.000 0.147 0.651 0.000
CRO (8) 0.013 ?0.021 0.227 ?0.009 0.020 0.131 0.261 0.364 0.237 0.141 0.043 0.011 0.004
0.510 0.302 0.000 0.666 0.320 0.000 0.000 0.000 0.000 0.000 0.034 0.592 0.839
SIZE (9) ?0.004 0.011 0.459 0.007 0.002 0.231 0.409 0.452 0.499 0.184 0.201 ?0.128 ?0.102
0.826 0.597 0.000 0.727 0.921 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
TRADD (10) 0.020 0.012 0.388 0.001 0.021 0.136 0.263 0.237 0.593 0.177 0.199 ?0.095 ?0.137
0.315 0.566 0.000 0.953 0.304 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
INTSEN (11) 0.170 ?0.360 0.176 0.084 0.066 ?0.044 0.122 0.123 0.140 0.134 0.156 0.171 ?0.114
0.000 0.000 0.000 0.000 0.001 0.031 0.000 0.000 0.000 0.000 0.000 0.000 0.000
COMMLOAN (12) 0.030 0.072 0.028 ?0.007 ?0.002 0.005 0.009 0.025 0.118 0.183 0.064 ?0.053 ?0.225
0.141 0.000 0.176 0.744 0.905 0.811 0.649 0.215 0.000 0.000 0.002 0.009 0.000
FVISTD (13) ?0.150 ?0.352 ?0.057 0.026 ?0.034 ?0.067 ?0.011 0.021 ?0.070 ?0.056 0.209 ?0.098 ?0.209
0.000 0.000 0.005 0.200 0.099 0.001 0.596 0.292 0.001 0.006 0.000 0.000 0.000
TIER1 (14) 0.192 0.252 ?0.047 ?0.020 0.022 ?0.010 0.061 ?0.011 ?0.134 ?0.142 ?0.148 ?0.188 ?0.219
0.000 0.000 0.020 0.315 0.280 0.616 0.003 0.577 0.000 0.000 0.000 0.000 0.000
Notes: The table reports the Pearson (below the diagonal) and Spearman (above the diagonal) correlations of the primary model variables for 2413 bank-year obser-
vations for 238 banks from 2002 to 2013. Variables are de?ned in the Appendix c?De?nitions of variables and other acronymsd?. p values are in italics below the
corresponding correlation.
Table 3
First-stage regressions of risk modeling measures on risk management variables.
XRM MRM CRM
Variables (1) (2)
CRO 0.022
??
0.022
?
(0.020) (0.066)
SIZE 0.021
???
0.046
???
(0.000) (0.000)
TRADD 0.004 0.012
(0.652) (0.270)
INTSEN ?0.004
??
?0.003
(0.021) (0.196)
COMMLOAN ?0.054 ?0.016
(0.182) (0.757)
FVISTD ?0.017
?
0.022
?
(0.053) (0.059)
TIER1 0.001 ?0.001
(0.347) (0.696)
Constant ?0.012 ?0.357
???
(0.781) (0.000)
Year effects Included Included
Observations 2413 2413
R-squared 0.068 0.273
Notes: This table reports the estimation of Eq. (1), which regresses of the risk mod-
eling measures (MRM and CRM, collectively denoted XRM) on risk management
variables and year effects. Variables are de?ned in the Appendix c?De?nitions of
variables and other acronymsd?. All variables are winsorized at the 0.5% and 99.5%
levels of their distribution. Standard errors are calculated clustering observations by
bank. P-values are in parentheses below the corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
FVGL components. To link to prior research and provide a bench-
mark for the main results, column (1) of the panel reports the
results for the model including only NIBNIGL, the three FVGL
components, and the ?xed year effects (untabulated). Column (2)
[(4)] presents the results for Eq. (2A) using the unadjusted mar-
ket [credit] risk modeling measure MRM [CRM]. Column (3) [(5)]
presents the results for Eq. (2B) using the market [credit] risk mod-
eling activity and disclosure measures
MRM and MRM ?
MRM
[
CRM and CRM ?
CRM].
In the benchmark model reported in column (1), the model ?t
is good with an R
2
of 41.8%. Such a high R
2
for a returns model
obtains in large part from the substantial variation in bank indus-
try performance across the sample years discussed earlier, which
is captured by the ?xed year effects. The coe?cient on NIBNIGL is
signi?cantly positive at the 1% level but well below one at 0.249,
suggesting the presence of considerable noise in this variable. The
coe?cient on NIGL is signi?cantly positive at the 5% level and well
above one at 3.297, indicating that it includes a permanent com-
ponent. This re?ects the inclusion of some fee revenue in trading
revenue on the Y-9C ?lings and the persistence of FVGL from most
dealer and proprietary trading activities due to the incorporation
of trading spreads (i.e., day-one pro?ts). The coe?cient on OCIGL
is signi?cantly positive at the 5% level and somewhat below one
at 0.742, consistent with it primarily capturing transitory income
items. The coe?cient on DISCGL is insigni?cant and close to zero,
consistent with prior research documenting little or no returns rel-
evance for FVGL on less liquid ?nancial instruments discussed in
section ‘Financial report information about estimated fair values
of ?nancial instruments, literature review, and hypothesis devel-
opment’.
The coe?cients on the test variables reported in columns (2)–
(5) of the panel are with one explainable exception consistent with
our main hypotheses H1–H3. The coe?cients on the interactions
of MRM and
MRM with OCIGL in columns (2) and (3), respec-
tively, are both signi?cantly positive at the 10% level, consistent
with Hypothesis H2. The coe?cients on the interactions of MRM
and
MRM with DISCGL in these columns are both insigni?cant,
however, inconsistent with Hypothesis H1. These insigni?cant coef-
?cients can be explained by low power resulting from much of the
90 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
overall sample period exhibiting stable interest rates and the large
set of ?nancial instruments (e.g., loans and deposits) re?ected in
DISCGL likely exhibiting considerable cross-hedging of interest rate
risk. This explanation is supported by results reported in Table 5
discussed below, which show that the coe?cients on these interac-
tions are signi?cantly positive during periods of high interest rate
volatility.
The coe?cients on the interactions of CRM and
CRM with DIS-
CGL in columns (4) and (5), respectively, are both signi?cantly pos-
itive at the 5% level, consistent with Hypotheses H1 and H3. The
coe?cients on the interactions of CRM and
CRM with OCIGL are
insigni?cant, as suggested by Hypothesis H3. These results sug-
gest that DISCGL re?ects banks’ primary assumption of credit risk
through funded loans and unfunded loan commitments.
Table 4, Panel A generally reports insigni?cant coe?cients on
the interactions of the risk modeling measures with the FVGL vari-
ables about which no hypotheses are made. The coe?cients on the
interactions of MRM,
MRM, CRM, and
CRM with NIGL are all in-
signi?cant. The coe?cients on the interactions of the risk modeling
disclosure measures MRM ?
MRM and CRM ?
CRM with all three
types of FVGL are all insigni?cant. Hence, the act of disclosing risk
Table 4
Regression of returns on FVGL, risk modeling measures, and interactions.
XRM MRM MRM CRM CRM
Variables (1) (2) (3) (4) (5)
Panel A: Regression of returns on FVGL, risk modeling measures, and interactions
NIBNIGL 0.249
???
0.255
???
?0.206 0.235
???
0.583
(0.000) (0.000) (0.456) (0.000) (0.229)
NIGL 3.297
??
10.695
???
32.875
???
7.692
??
27.269
??
(0.024) (0.001) (0.006) (0.043) (0.022)
OCIGL 0.742
??
1.649
???
?0.311 0.501 ?1.733
(0.021) (0.002) (0.884) (0.170) (0.427)
DISCGL 0.043 0.075 ?0.327 ?0.025 ?1.165
??
(0.274) (0.453) (0.404) (0.654) (0.016)
NIBNIGL ? XRM ?0.008 0.095
(0.968) (0.680)
NIGL ? XRM ?11.562 ?8.176
(0.103) (0.146)
OCIGL ? XRM 3.194
?
1.044
(0.059) (0.305)
DISCGL ? XRM ?0.125 0.308??
(0.678) (0.027)
XRM ?0.026 0.011
(0.431) (0.657)
NIBNIGL ?
XRM 1.671
?
?0.503
(0.081) (0.492)
NIGL ?
XRM ?64.909 ?25.949
(0.122) (0.143)
OCIGL ?
XRM 3.850
?
3.686
(0.064) (0.265)
DISCGL ?
XRM 1.258 1.791
??
(0.342) (0.014)
XRM ?0.697
???
?0.253
???
(0.000) (0.003)
NIBNIGL ? (XRM ?
XRM) ?0.139 ?0.018
(0.629) (0.950)
NIGL ? (XRM ?
XRM) ?1.715 5.709
(0.731) (0.341)
OCIGL ? (XRM ?
XRM) ?3.647 1.569
(0.125) (0.190)
DISCGL ? (XRM ?
XRM) ?0.650 0.201
(0.147) (0.421)
(XRM ?
XRM) ?0.012 0.036
(0.735) (0.218)
Constant 0.297
???
0.300
???
0.522
???
0.294
???
0.467
???
(0.000) (0.000) (0.000) (0.000) (0.000)
Year effects Included Included Included Included Included
Observations 2413 2413 2413 2413 2413
R-squared 0.418 0.424 0.437 0.421 0.431
(Continued on next column)
Table 4 (continued)
Variables (1) (2)
Panel B: Expanded regression of returns on FVGL, risk modeling measures, and
interactions including M/ABSDUM interactions
OCIGL 2.358
???
0.215
(0.002) (0.933)
OCIGL ? M/ABSDUM ?1.678
?
?2.073
(0.084) (0.588)
OCIGL ? MRM 6.221
(0.113)
OCIGL ? MRM ? M/ABSDUM 6.927
??
(0.031)
OCIGL ?
MRM 1.842
(0.814)
OCIGL ?
MRM ? M/ABSDUM 7.220
?
(0.054)
OCIGL ? (MRM ?
MRM) ? M/ABSDUM ?5.873
(0.113)
OCIGL ? (MRM ?
MRM) ? M/ABSDUM 3.924
(0.260)
R-squared 0.427 0.439
Notes: Column 1 of Panel A reports the estimation of a restricted version of Eq. (2A)
that regresses returns for the 12 months ending 4 months after the ?scal year end
(R) on net income before FVGL recorded in net income (NIBNIGL) and the three
types of FVGL (NIGL, OCIGL, and DISCGL) and includes ?xed year effects. Column 2
(4) reports the estimation of Eq. (2A), which includes one of the unadjusted MRM
and CRM measures described in the Appendix c?Risk modeling measures and chief
risk o?cer indicatord? and interactions involving that measure. Column 3 (5) re-
ports the estimation of Eq. (2B), which includes one of the MRM and CRM risk
modeling activity measures (
MRM or
CRM, collectively denoted
XRM), the corre-
sponding risk modeling disclosure measure (MRM ?
MRM or CRM ?
CRM), and in-
teractions involving these measures. The MRM and CRM activity and disclosure
measures are the predicted values and residuals, respectively, from the estimation
of the corresponding ?rst-stage model reported in Table 3. Panel B reports the re-
sults of expanded versions of the regression reported in columns (2) and (3) of
Panel A that include interactions of M/ABSDUM, an indicator for an above-median
proportion of available-for-sale securities that are mortgage-backed or asset-backed
securities, with the variables involving OCIGL. Variables are de?ned in the Appendix
c?De?nitions of variables and other acronymsd?. All variables are winsorized at the
0.5% and 99.5% levels of their distribution. Standard errors are calculated clustering
observations by bank. P-values are in parentheses under the corresponding coe?-
cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
modeling activities does not appear to affect the returns-relevance
of FVGL.
In summary, the ?ndings reported in Table 4, Panel A indi-
cate that MRM activities enhance the returns-relevance of OCIGL,
which primarily re?ect interest rate risk, consistent with Hypothe-
sis H2 but not Hypothesis H1. CRM activities enhance the returns-
relevance of DISCGL, which re?ect banks’ primary assumption of
credit risk through funded loans and unfunded loan commitments,
consistent with Hypotheses H1 and H3.
Regression results with interactions for potentially illiquid and
di?cult-to-value available-for-sale securities
Hypothesis H2 is based on the assumption that banks do
not determine OCIGL directly from market prices, obviating their
need to conduct MRM to make this determination. As discussed
in section ‘Financial report information about estimated fair val-
ues of ?nancial instruments, literature review, and hypothesis
development’, mortgage-backed and asset-backed securities typi-
cally are illiquid and include embedded interest-rate options and
other interest-rate-related structuring that require MRM to accu-
rately estimate the fair value of the securities. Re?ecting this fact,
the hypothesis states it should hold more strongly for mortgage-
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 91
backed and asset-backed securities than for banks’ other primary
types available-for-sale securities, such as U.S. Treasuries. To test
this part of the hypothesis, we estimate expanded versions of
Eqs. (2A) and (2B) that include interactions of an indicator vari-
able for bank-year observations with above median percentage
of mortgage-backed and asset-backed securities in their available-
for-sale securities portfolios in that year, denoted M/ABSDUM,
with all variables involving OCIGL. We expect the coe?cient on
the interaction of OCIGL with MRM to be more positive and
signi?cant for observations for which M/ABSDUM takes a value
of one.
To conserve space, Table 4, Panel B reports the estimated co-
e?cients only for the critical variables in these expanded models
and also the R
2
s; all omitted coe?cients are essentially identi-
cal to those reported in Panel A. Column (1) [(2)] of the panel re-
ports the results for the expanded versions of Eq. (2A) [(2B)]. In
column (1), the coe?cient on the interaction of OCIGL and MRM is
insigni?cantly positive, consistent with MRM not enhancing the es-
timation of the fair values of available-for-sale securities other than
mortgage-backed and asset-backed securities. The coe?cient on
the interaction of OCIGL and M/ABSDUM is signi?cantly negative
at the 10% level, consistent with fair valuation estimation posing
greater di?culties for mortgage-backed and asset-backed securities
than for other available-for-sale securities. Consistent with Hypoth-
esis H2 that MRM enhances the fair value estimation of mortgage-
backed and asset-backed securities, the coe?cient on the interac-
tion of OCIGL, MRM, and M/ABSDUM is signi?cantly positive at the
5% level.
Column (2) yields similar inferences. In particular, the co-
e?cient on the interaction of OCIGL,
MRM, and M/ABSDUM
is signi?cantly positive at the 10% level and the coe?cient
on the interaction of OCIGL, MRM ?
MRM, and M/ABSDUM is
insigni?cant.
Regression results with interactions for macroeconomic conditions
Hypothesis H4 predicts that MRM activities are more impor-
tant in periods with more volatile interest rates. To test this hy-
pothesis, we expand the models reported in columns (2) and (3)
of Table 4, Panel A to include an indicator variable for years with
above-median interest rate volatility that is interacted with the in-
teractions of MRM,
MRM, and MRM ?
MRM with NIBNIGL and the
three types of FVGL. Similarly, Hypothesis H5 predicts that CRM
activities are more important in periods with higher and more un-
certain credit losses. To test this hypothesis, we expand the mod-
els reported in columns (4) and (5) of Table 4, Panel A to include
an indicator variable for the ?nancial crisis years that is interacted
with the interactions of CRM,
CRM, and CRM ?
CRM with NIBNIGL
and the three types of FVGL.
We measure interest rate volatility during a year as the stan-
dard deviation of the daily yield on seven-year U.S. Treasuries. UN-
STABLE takes a value of one for a year with above-median interest
rate volatility and zero otherwise. CRISIS takes a value of one for
the years 2007–2009, and zero otherwise, because the subprime
crisis began in February 2007 (Ryan, 2008) and banks’ loan loss
provisions peaked in the third quarter of 2009.
Table 5 presents the results of the estimations of the four ex-
panded models; columns (1) and (2) reporting the results for the
models including the indicator variable UNSTABLE and columns (3)
and (4) reporting the results for the models including the indicator
CRISIS. In the table, these indicator variables are generically de-
noted DUMMY and the column headings specify which indicator
DUMMY represents. To conserve space, we discuss only the coef-
?cients on the interactive variables involving these indicators. We
omit ?xed year effects in these models because both indicator vari-
ables partition on time.
Column (1) of Table 5 presents the results for the model in
which UNSTABLE is interacted with the interactive variables in-
volving MRM. Consistent with Hypothesis H4, the coe?cients on
OCIGL ? MRM ? UNSTABLE and DISCGL ? MRM ? UNSTABLE are sig-
ni?cantly positive at the 5% and 1% levels, respectively. The latter
result is related to our explanation for the unexpectedly insignif-
icant coe?cients on the interactions of DISCGL with MRM and
MRM in Table 4, Panel A discussed above. In contrast, the coe?-
cients on all other interactions involving UNSTABLE are insigni?-
cant.
Column (2) of the table presents similar results for the model
where UNSTABLE is interacted with the interactive variables in-
volving
MRM and MRM-
MRM. Again consistent with Hypoth-
esis H4, the coe?cients on OCIGL ?
MRM ? UNSTABLE and
DISCGL ?
MRM ? UNSTABLE are both signi?cantly positive at the
1% levels, while the coe?cients on OCIGL ? (MRM ?
MRM) ? UN-
STABLE and DISCGL ? (MRM ?
MRM) ? UNSTABLE are both in-
signi?cant. In contrast, the coe?cients on all other interactions in-
volving UNSTABLE are insigni?cant except for the signi?cantly neg-
ative coe?cient on NIBNIGL ?
MRM ? UNSTABLE.
Column (3) presents the results for the model where CRISIS
is interacted with the interactive variables involving CRM. Consis-
tent with Hypothesis H5, the coe?cient on DISCGL ? CRM ? CRI-
SIS is signi?cantly positive at the 10% level. In addition, the coef-
?cients on NIGL ? CRM ? CRISIS and OCIGL ? CRM ? CRISIS are sig-
ni?cantly positive, likely re?ecting the pervasive effect of credit
risk on banks’ various sources of net income during the ?nancial
crisis.
Column (4) presents similar results for the model where CRI-
SIS is interacted with the interactive variables involving
CRM
and CRM ?
CRM. Consistent with Hypothesis H5, the coe?cient
on DISCGL ?
CRM ? CRISIS is signi?cantly positive at the 10%
level, while the coe?cients on the interactions of CRISIS with
all variables involving CRM ?
CRM are insigni?cant. In contrast,
the coe?cients on all other interactions involving UNSTABLE are
insigni?cant except for the signi?cantly negative coe?cient on
NIBNIGL ?
CRM ? UNSTABLE and signi?cantly positive coe?cient
on OCIGL ?
CRM ? UNSTABLE.
In summary, the unadjusted risk modeling measures and risk
modeling activity measures generally are signi?cant for economic
conditions for which they are expected to be important and usually
are insigni?cant otherwise. The risk modeling disclosure measures
generally are insigni?cant regardless of the economic conditions.
Changes speci?cation
To help ensure that our primary second-stage results reported
in Table 4, Panel A are robust to omitted variables, and also as an
alternative to the inclusion of year ?xed effects as a means to cap-
ture the trend in CRM over time reported in Table 1, Panel C, we
estimate changes models. These models are identical to the mod-
els reported in columns (2)–(5) of Table 4, Panel A except that
MRM,
MRM, MRM ?
MRM, CRM,
CRM, and CRM ?
CRM are re-
placed with the changes in the respective variables and the year
?xed effects are dropped.
Table 6 reports the estimation of these changes models, with
columns (1) and (2) corresponding to columns (2) and (3), respec-
tively, of Table 4, Panel A. The results in column (1) of Table 6
are similar with those reported in column (2) of Table 4, Panel A.
The coe?cient on the interaction of MRM with OCIGL remains
signi?cantly positive at the 10% level, consistent with Hypothesis
H2. The results in column (2) of Table 6 are similar to those re-
92 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Table 5
Regression of returns on risk modeling measures, FVGL, and interactions impact of
macroeconomic conditions.
XRM MRM CRM
DUMMY Unstable Crisis
Variables (1) (2) (3) (4)
NIBNIGL 0.179
??
?0.808
??
0.099 ?0.686
(0.015) (0.016) (0.140) (0.195)
NIGL 17.018
???
43.237
???
6.563
?
25.126
??
(0.000) (0.001) (0.073) (0.050)
OCIGL 0.995 ?1.320 0.526 0.523
(0.126) (0.627) (0.114) (0.838)
DISCGL 0.133 ?0.321 ?0.086 ?1.166
??
(0.269) (0.560) (0.187) (0.039)
NIBNIGL ? XRM 0.965
?
0.983
???
(0.067) (0.000)
NIGL ? XRM ?16.638 ?10.543
(0.130) (0.158)
OCIGL ? XRM ?6.765 ?2.383
(0.117) (0.234)
DISCGL ? XRM 1.187
?
0.158
(0.094) (0.522)
XRM ?0.098
?
?0.010
(0.077) (0.690)
NIBNIGL ? XRM ? DUMMY ?0.879 ?1.454
(0.199) (0.109)
NIGL ? XRM ? DUMMY ?2.251 26.385
???
(0.492) (0.001)
OCIGL ? XRM ? DUMMY 5.582
??
9.189
???
(0.025) (0.003)
DISCGL ? XRM ? DUMMY 1.080
???
0.850
?
(0.003) (0.064)
XRM ? DUMMY 0.079 0.011
(0.361) (0.889)
NIBNIGL ?
XRM 5.207
???
1.734
??
(0.000) (0.034)
NIGL ?
XRM ?8.767 ?2.431
(0.108) (0.079)
OCIGL ?
XRM ?2.030 ?1.239
(0.816) (0.752)
DISCGL ?
XRM 0.215 1.670
?
(0.907) (0.050)
XRM ?1.187
???
?0.477
???
(0.000) (0.000)
NIBNIGL ?
XRM ? DUMMY ?1.587
??
?0.689
???
(0.040) (0.000)
NIGL ?
XRM ? DUMMY 9.819 12.825
(0.352) (0.108)
OCIGL ?
XRM ? DUMMY 8.692
???
4.100
???
(0.002) (0.000)
DISCGL ?
XRM ? DUMMY 1.238
???
0.271
?
(0.005) (0.074)
XRM ? DUMMY 0.584 0.323
(0.109) (0.162)
NIBNIGL ? (XRM ?
XRM) ?0.855 0.308
(0.590) (0.298)
NIGL ? (XRM ?
XRM) 8.155 0.823
(0.251) (0.888)
OCIGL ? (XRM ?
XRM) 7.239 ?1.184
(0.268) (0.340)
DISCGL ? (XRM ?
XRM) ?0.349 0.395
(0.559) (0.191)
(XRM ?
XRM) 0.047 0.033
(0.717) (0.324)
NIBNIGL ? (XRM ?
XRM) ? DUMMY 0.734 ?0.273
(0.651) (0.634)
NIGL ? (XRM ?
XRM) ? DUMMY ?31.947 8.593
(0.128) (0.833)
OCIGL ? (XRM ?
XRM) ? DUMMY ?8.172 4.763
(0.229) (0.266)
DISCGL ? (XRM ?
XRM) ? DUMMY ?0.741 ?0.661
(0.401) (0.259)
(continued on next page)
Table 5 (continued)
XRM MRM CRM
DUMMY Unstable Crisis
Variables (1) (2) (3) (4)
(XRM ?
XRM) ? DUMMY ?0.042 ?0.102
(0.782) (0.254)
DUMMY ?0.022 ?0.158 ?0.244
???
?0.453
???
(0.402) (0.176) (0.000) (0.003)
Constant 0.067
???
0.379
???
0.110
???
0.413
???
(0.000) (0.000) (0.000) (0.000)
Year effects Excluded Excluded Excluded Excluded
Observations 2413 2413 2413 2413
R-squared 0.086 0.117 0.205 0.232
Notes: Columns (1) and (2) [(3) and (4)] of this table reports the results of expand-
ing the regression reported in the columns (2) and (3) [(4) and (5)] of Table 4,
Panel A to include interactions of a dummy variable that takes a value of 1 for
years with above-median unstable interest rates [the ?nancial crisis years 2007–
2009] with the interactions of NIBNIGL and the three types of FVGL with MRM,
MRM and (MRM-
MRM), [CRM,
CRM and (CRM-
CRM)] respectively. Variables are de-
?ned in the Appendix c?De?nitions of variables and other acronymsd?. All variables
are winsorized at the 0.5% and 99.5% levels of their distribution. Standard errors are
calculating clustering observations by bank. P-values are in parentheses under the
corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
ported in column (3) of Table 4, Panel A. The coe?cient on the
interaction of
MRM with OCIGL becomes more signi?cantly pos-
itive at the 5% level, consistent with Hypothesis H2, and the coe?-
cient on the interaction of (MRM-
MRM) with OCIGL remains in-
signi?cant. Columns (3) and (4) of Table 6 correspond to columns
(4) and (5), respectively, of Table 4, Panel A. The results in col-
umn (3) of Table 6 are similar with those reported in column (4)
of Table 4, Panel A. The coe?cient on the interaction of CRM
with DISCGL remains signi?cantly positive at the 5% level, con-
sistent with Hypotheses H1 and H3. The results in column (4) of
Table 6 are similar with those reported in column (5) of Table 4,
Panel A. The coe?cient on the interaction of
CRM with DISCGL
remains signi?cantly positive, albeit at the lower 10% level, consis-
tent with Hypotheses H1 and H3. The coe?cient on the interaction
of (CRM ?
CRM) with DISCGL remains insigni?cant.
Summary and conclusion
We provide evidence that banks’ market and credit risk model-
ing (MRM and CRM, respectively) enhance the returns-relevance of
banks’ estimated unrealized fair value gains and losses (FVGL) for
?nancial instruments. We employ both a one-stage approach that
uses unadjusted MRM and CRM measures developed from banks’
?nancial report disclosures and a two-stage approach that parti-
tions these unadjusted measures into measures of banks’ risk mod-
eling activities and disclosures of those activities. We predict and
?nd that banks’ MRM and CRM activities, but not disclosures of
these activities, enhance the returns-relevance of their FVGL, more
so for less liquid instruments for which FVGL typically are calcula-
ble from disclosures rather than recorded in net income or other
comprehensive income. We further predict and ?nd that banks’
MRM activities enhance the returns-relevance of FVGL recorded in
other comprehensive income, which primarily result from interest
rate movements, especially when these FVGL are for potentially
illiquid and di?cult-to-value mortgage-backed and asset-backed
available-for-sale securities. We predict and ?nd that banks’ CRM
activities enhance the returns-relevance of their disclosed FVGL,
because banks assume credit risk primarily through their funded
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 93
Table 6
Regression of Returns on Risk Modeling Measures, FVGL, and Interactions
Changes Speci?cation.
XRM MRM CRM
Variables (1) (2) (3) (4)
NIBNIGL 0.148
???
0.189
???
0.139
???
0.174
???
(0.000) (0.000) (0.000) (0.000)
NIGL 7.636
???
6.936
???
7.405
???
7.106
???
(0.000) (0.001) (0.001) (0.001)
OCIGL 0.099 0.073 0.032 0.086
(0.623) (0.713) (0.873) (0.662)
DISCGL 0.033 0.028 0.033 0.024
(0.291) (0.334) (0.286) (0.416)
NIBNIGL ? XRM ?2.810 1.357
???
(0.628) (0.006)
NIGL ? XRM ?168.167 1.402
(0.291) (0.935)
OCIGL ? XRM 10.703
?
4.315
?
(0.052) (0.074)
DISCGL ? XRM ?1.035 0.982
??
(0.666) (0.049)
XRM 0.297 ?0.102
?
(0.410) (0.086)
NIBNIGL ?
XRM 3.748 4.760
??
(0.288) (0.027)
NIGL ?
XRM ?21.199 ?9.753
(0.243) (0.428)
OCIGL ?
XRM 5.515
??
?2.080
(0.033) (0.930)
DISCGL ?
XRM ?0.140 1.136
?
(0.975) (0.060)
XRM ?1.789
??
?1.090
??
(0.033) (0.048)
NIBNIGL ? (XRM ?
XRM) 1.056 0.003
(0.721) (0.995)
NIGL ? (XRM ?
XRM) 37.275 ?25.090
(0.498) (0.141)
OCIGL ? (XRM ?
XRM) 1.693 1.363
(0.879) (0.555)
DISCGL ? (XRM ?
XRM) ?0.748 0.295
(0.630) (0.616)
(XRM ?
XRM) ?0.129 0.003
(0.588) (0.957)
Constant 0.056
???
0.058
???
0.057
???
0.057
???
(0.000) (0.000) (0.000) (0.000)
Observations 2214 2214 2214 2214
R-squared 0.041 0.050 0.050 0.051
Notes: Columns (1) and (2) [(3) and (4)] of this table report the results
of a modi?cation of the regression reported in columns (2) and (3) [(4)
and (5)] of Table 4, Panel A that includes the changes in MRM,
MRM and
(MRM ?
MRM) [CRM,
CRM and (CRM ?
CRM)] rather than the levels of these
variables. Variables are de?ned in the Appendix c?De?nitions of variables and
other acronymsd?. All variables are winsorized at the 0.5% and 99.5% levels
of their distribution. Standard errors are calculated clustering observations by
bank. P-values are in parentheses under the corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
loans and unfunded loan commitments for which the fair values
are disclosed rather than recognized. Finally, we show that the im-
pact of MRM (CRM) activities is stronger during periods of higher
interest rate volatility (the ?nancial crisis).
Our evidence is subject to the following caveats. First, we show
that banks’ MRM and CRM are associated with increased returns-
relevance for speci?c types of FVGL, not that they cause this in-
creased returns relevance. Correlated bank attributes, such as tech-
nical sophistication, may be the actual causes. Second, we do not
control for alternative ?nancial report disclosures related to the
quality of their FVGL estimates, because these alternatives are sig-
ni?cantly limited in terms of one or more of their information
quality, coverage across types of ?nancial instruments, or coverage
across our sample period. Finally, our results based on a U.S. bank
sample may not generalize to other industries or countries.
Despite these caveats, we believe our results have relevance for
accounting standard setters, auditors, and other parties involved
in requiring or evaluating estimates of the fair values of ?nan-
cial instruments and related disclosures in banks’ ?nancial reports.
Re?ecting strong political pressure arose during the ?nancial cri-
sis, a number of the FASB’s and IASB’s recent proposals indicate
retrenchment in what they deem feasible or desirable regarding
the extent of fair value accounting for ?nancial instruments.
18
Our
?ndings suggest that the returns-relevance of FVGL for ?nancial in-
struments is enhanced by banks’ risk modeling.
Risk modeling measures and chief risk o?cer indicator
Risk modeling measures
Our risk modeling measures are simple averages of indicators
for whether banks engage in each of the following risk model-
ing activities (including related risk management processes) men-
tioned a survey developed by the Basel Committee on Banking Su-
pervision (2001):
Market risk modeling:
(1) INTGAP – Does the bank use maturity or repricing analysis?
(2) IRSIMUL – Does the bank use interest rate simulation?
(3) VAR – Does the bank use Value at Risk for market risk?
(4) MRSTRESS – Does the bank stress test its market risk models?
(5) MRBACKTEST – Does the bank back test its market risk models?
Credit risk modeling:
(6) MODEL – Does the bank use statistical credit risk modeling?
(7) SCORING – Does the bank use credit scoring models?
(8) RR – Does the bank have an internal credit risk rating process?
(9) CRSTRESS – Does the bank use stress testing its credit risk
models?
Our market risk modeling measure, MRM, is the simple average
of the ?rst ?ve indicators. Our credit risk modeling measure, CRM,
is the simple average of the next four indicators. As averages of
indicator variables, both measures take values from zero to one.
We identify whether banks engage in these activities from
hand-collected disclosures in their Form 10-K ?ling for the years
2001–2013. We ?rst used search engines to locate words or
phrases such as c?risk management,d? c?risk model,d? c?scoring,d?
and c?risk rating,d? and read each search result to determine
whether one (or more) of the nine risk modeling activities is indi-
cated. Each item is scored one if disclosed and zero otherwise. The
following table provides representative examples of disclosures for
each of the nine activities.
18
For example, see the FASB’s April 12, 2013 proposed Accounting Standards Up-
date, Financial Instruments—Overall (Subtopic 825-10): Recognition and Measure-
ment of Financial Assets and Financial Liabilities, which would allow ?rms consid-
erable ability to classify ?nancial instruments to avoid fair valuation.
94 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Market risk modeling
INTGAP Central Paci?c Financial Corp Form 10-K ?ling 2009
Interest rate risk can be analyzed by monitoring an institution’s interest rate sensitivity gap and changes in the gap over time. An asset or
liability is said to be interest rate sensitive within a speci?c time period if it will mature or reprice within that time period. The interest
rate sensitivity gap is de?ned as the difference between the amount of interest-earning assets and the amount of interest-bearing
liabilities maturing or repricing within a speci?ed time period. A gap is considered positive when the amount of interest rate sensitive
assets exceeds the amount of interest rate sensitive liabilities. A gap is considered negative when the amount of interest rate sensitive
liabilities exceeds the amount of interest rate sensitive assets. During a period of rising interest rates, the earnings of an institution with a
positive gap theoretically may be positively affected due to its interest-earning assets repricing to a greater extent than its interest-bearing
liabilities. An adverse impact would be expected for an institution with a negative gap
IRSIMUL Zions Bancorporation Form 10-K ?ling 2003
Interest rate risk is the most signi?cant market risk to which the Company is regularly exposed. We monitor this risk through the use of
two complementary measurement methods: duration of equity and income simulation. In the duration of equity method, we measure the
changes in the market values of equity in response to changes in interest rates. In the income simulation method, we analyze the changes
in income in response to changes in interest rates. The tool that we use to apply both of these methods is an asset-liability management
system marketed by a third party company, QRM. In general, our goal in managing interest is to have net interest income increase in a
rising interest rate environment, which tends to mitigate any declines in the market value of equity due to higher discount rates. This
approach is based on our belief that in a rising interest rate environment, the market cost of equity, or implied rate at which future
earnings are discounted, would also tend to rise. We refer to this position as c?asset sensitived?
VAR Key Corp Form 10-K ?ling 2004
Management uses a value at risk (c?VARd?) simulation model to measure the potential adverse effect of changes in interest rates, foreign
exchange rates, equity prices and credit spreads on the fair value of Key’s trading portfolio. Using historical information, the model
estimates the maximum potential one-day loss with 95% probability by comparing the relative change in rates to the current day’s rates
and prices
MRSTRESS US Bancorp Form 10-K ?ling 2006
The Company mitigates these uncertainties through regular monitoring of trading activities by management and other risk management
practices, including stop-loss and position limits related to its trading activities. Stress-test models are used to provide management with
perspectives on market events that VaR models do not capture
MRBACKTEST FleetBoston Financial Corp Form 10-K ?ling 2003
Our independent Market Risk Management function routinely validates our measurement framework by conducting backtests, which
compare the actual daily trading-related results against the estimated VAR with a one-day holding period. This measure contrasts with
our standard VAR estimate by excluding the impact of reduced trading liquidity
Credit risk modeling
MODEL Bank of America Corporation Form 10-K ?ling 2009
We use proprietary models to measure the capital requirements for credit, country, market, operational and strategic risks
Statistical models are built using detailed behavioral information from external sources such as credit bureaus and/or internal historical
experience. These models are a component of our consumer credit risk management process and are used, in part, to help determine both
new and existing credit decisions, portfolio management strategies including authorizations and line management, collection practices and
strategies, determination of the allowance for loan and lease losses, and economic capital allocations for credit risk
SCORING BOK Financial Corp Form 10-K ?ling 2009
Credit scoring is assessed based on signi?cant credit characteristics including credit history, residential and employment stability
RR Santander Bancorp Form 10-K ?ling 2007
The corporation has also established an internal risk rating system and internal classi?cations which serve as timely identi?cation of the
loan quality issues affecting the loan portfolio.
CRSTRESS Bank of Hawaii Form 10-K ?ling 2009
In addition, the Company uses a variety of other tools to estimate probable credit losses including, but not limited to, a rolling quarterly
forecast of asset quality metrics; stress testing; and performance indicators based on the Company’s own experience, peers, or other
industry sources
Chief risk o?cer indicator
The chief risk o?cer indicator, CRO, takes the value one if the
bank discloses in its Form 10-K ?ling that it employs a chief risk
o?cer and zero otherwise. We ?rst search ?lings to determine
whether the words c?chiefd? and c?riskd? appear within ?ve words
of each other. We then search the ?lings that meet the ?rst ?lter
to determine whether the words c?riskd? and either c?o?cerd? or
c?directord? appear within ?ve words of each other. We read the
relevant passages to con?rm the existence of a chief risk o?cer for
the bank-years that meet both ?lters.
De?nitions of variables and other acronyms
Variable Description Source
ASC Accounting Standards Codi?cation Acronym
ASU Accounting Standards Update Acronym
COMMLOAN Commercial and industrial (including agricultural) loans divided by lagged average total assets Y-9C
CRISIS Indicator equal to one for the years 2007 to 2009 and zero otherwise
CRM Credit risk modeling measure (see the Appendix c?Risk modeling measures and chief risk o?cer indicatord? for
details)
10-K
CRO Indicator equal to one if the bank employs a chief risk o?cer and zero otherwise (see the Appendix c?Risk modeling
measures and chief risk o?cer indicatord? for details)
10-K
DISCGL Fair value gains and losses calculated from FAS 107 disclosures divided by beginning-of-year market value of equity SNL
(continued on next page)
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 95
Variable Description Source
FAS Statement of Financial Accounting Standards Acronym
FSP FASB Staff Position Acronym
FVGL Fair value gains and losses Acronym
FVISTD Standard deviation of de?ated fair value income—the sum of NIBNIGL, NIGL, OCIGL and DISCGL—over the last ten
years, requiring at least three annual observations
SNL
INTSEN Interest-sensitive assets minus interest-sensitive liabilities repricing in one year divided by beginning-of-year market
value of equity
Y-9C
MRM Market risk modeling measure (see the Appendix c?Risk modeling measures and chief risk o?cer indicatord? for
details)
10-K
M/ABSDUM Indicator variable equal to one if mortgage-backed and asset-backed available-for-sale securities divided by total
available-for-sale securities is above the sample median for the year and zero otherwise
Y-9C
NIBNIGL Net income before fair value gains and losses recorded in net income divided by beginning-of-year market value of
equity
Y-9C
NIGL Fair value gains and losses recorded in net income divided by beginning-of-year market value of equity Y-9C
OCIGL Fair value gains and losses recorded in other comprehensive income divided by beginning-of-year market value of
equity
Y-9C
R Twelve-month stock return ending four months after the ?scal year end CRSP
SIZE Log of total assets Y-9C
TIER1 Tier one risk-based capital ratio Y-9C
TRADD Indicator variable equals to one if the bank engages in trading activities and zero otherwise Y-9C
UNSTABLE Indicator equal to one if the standard deviation of the daily yield on seven-year U.S. Treasuries is above the sample
median for the year and zero otherwise
Federal Reserve
Notes: Regarding bank data sources.
Y-9C refers to banks’ regulatory Y-9C ?lings. 10-K refers to
banks’ Form 10-K ?lings. SNL refers to SNL Financial.
References
Aebi, V., Sabato, G., & Schmid, M. (2012). Risk management, corporate governance,
and bank performance in the ?nancial crisis. Journal of Banking and Finance,
36(2012), 3213–3226.
Ahmed, A., Kilic, E., & Lobo, G. (2006). Does recognition versus disclosure matter?
Evidence from value-relevance of banks recognized and disclosed derivative ?-
nancial instruments. The Accounting Review, 81(3), 567–588.
Ahmed, A., & Takeda, C. (1995). Stock market valuation of gains and losses on com-
mercial banks’ investment securities: An empirical analysis. Journal of Account-
ing & Economics, 20, 207–222.
Badertscher, B., Burks, J., & Easton, P. (2014). The market pricing of other than tem-
porary impairments. The Accounting Review, 89(3), 811–838.
Barth, M. (1994). Fair value accounting: Evidence from investment securities and
the market valuation of banks. The Accounting Review, 69, 1–25.
Barth, M., Beaver, W., & Landsman, W. (1996). Value-relevance of banks’ fair value
disclosure under SFAS 107. The Accounting Review, 71(4), 513–537.
Barth, M., Beaver, W., & Landsman, W. (2001). The relevance of the value-relevance
literature for ?nancial accounting standard setting: Another view. Journal of Ac-
counting & Economics, 71(4), 77–104.
Basel Committee on Banking Supervision (2003). Public Disclosures by Banks: Results
of the 2001 Disclosure Survey, May.
Beaver, W., & Venkatchalam, S. (2003). Differential pricing of components of bank
loan fair values. Journal of Accounting, Auditing and Finance, 18(1), 41–67.
Bhat, G., Ryan, S., & Vyas, D. (2014). The implications of credit risk modeling for banks’
loan loss provision timeliness and loan origination procyclicality. Working paper,
New York University.
Carroll, T., Linsmeier, T., & Petroni, K. (2003). The reliability of fair value versus his-
torical cost information: Evidence from closed-end mutual funds. Journal of Ac-
counting, Auditing, and Finance, 18(1), 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, 557–593.
Chernobai, A., Jorion, P., & Yu, F. (2011). The determinants of operational risk in U.S.
?nancial institutions. Journal of Financial and Quantitative Analysis, 46(6), 1683–
1725.
Chung, S., Goh, B., Ng, J., & Yong, K. (2014). Voluntary fair value disclosures beyond
SFAS 157’s three-level estimates. Working paper, Singapore Management Univer-
sity.
Danbolt, J., & Rees, W. (2008). An experiment in fair value accounting: UK invest-
ment vehicles. European Accounting Review, 17(2), 271–303.
Dong, M., Ryan, S., & Zhang, X. (2014). Preserving amortized costs within a fair-
value-accounting framework: Reclassi?cation of gains and losses on available-
for-sale securities upon realization. Review of Accounting Studies, 19(1), 242–280.
Eccher, E., Ramesh, K., & Thiagarajan, S. (1996). Fair value disclosures by bank hold-
ing companies. Journal of Accounting & Economics, 22, 79–117.
Ellul, A., & Yerramilli, V. (2013). Stronger risk controls, lower risk: Evidence from
U.S. bank holding companies. Journal of Finance, 68(5), 1757–1803.
Goh, B., Li, D., Ng, J., & Yong, K. (2015). Market pricing of banks’ fair value assets
reported under SFAS 157 during the 2008 Financial Crisis. Journal of Accounting
and Public Policy (forthcoming).
Healy, P., & Palepu, K. (2001). Information asymmetry, corporate disclosure, and the
capital markets: A review of the empirical disclosure literature. Journal of Ac-
counting & Economics, 31, 405–440.
Hirst, D., Hopkins, P., & Wahlen, J. (2004). Fair values, income measurement, and
bank analysts’ risk and valuation judgments. The Accounting Review, 79(2), 453–
472.
Holthausen, R., & Verrecchia, R. (1988). The effect of sequential information releases
on the variance of price changes in an intertemporal multi-asset market. Journal
of Accounting Research, 3, 82–106.
Holthausen, R., & Watts, R. (2001). The relevance of the value-relevance literature
for ?nancial accounting standard setting. Journal of Accounting & Economics,
31(1–3), 3–75.
Kaplan, R., & Mikes, A. (2012). Managing risks: A new framework. Harvard Business
Review June: 48–60.
Kolev, Kalin S. (2009). Do investors perceive marking-to-model as marking-to-myth?
Early evidence from FAS 157 Disclosure. Working paper. Yale School of Manage-
ment.
Lang, M., & Lundholm, R. (1993). Cross-sectional determinants of analysts ratings of
corporate disclosure. Journal of Accounting Research, 31(2), 246–271.
Larcker, D., Richardson, S., & Tuna, A. (2007). Corporate governance, accounting out-
comes, and organizational performance. The Accounting Review, 82(4), 963–1008.
Mikes, A. (2011). From counting risk to making risk count: Boundary-work in risk
management. Accounting, Organizations and Society, 36, 226–245.
Nelson, K. (1996). Fair value accounting for commercial banks: An empirical analysis
of SFAS No. 107. The Accounting Review, 71(2), 161–182.
Nissim, D. (2003). Reliability of banks’ fair value disclosure for loans. Review of
Quantitative Finance and Accounting, 20(4), 355–384.
Ryan, S. (2008). Accounting in and for the subprime crisis. The Accounting Review,
83, 1605–1638.
Ryan, S. (2011). Financial reporting for ?nancial instruments. Foundations and Trends
in Accounting, 6(3–4), 187–354.
Song, C., Thomas, W., & Yi, H. (2010). Value-relevance of FAS 157 fair value hierarchy
information and the impact of corporate governance mechanisms. The Account-
ing Review, 85, 1375–1410.
Venkatachalam, M. (1996). Value-relevance of banks’ derivatives disclosures. Journal
of Accounting & Economics, 22, 327–355.
Vickrey, J., & Wright, J. (2013). TBA trading and liquidity in the agency MBS market.
FRBNY Economic Policy Review, 19(1), 1–18.
doc_839302394.pdf
Weexaminewhetherandhowmeasuresofmarketandcreditriskmodelingidentifiedfrombanks’financialreportsenhancethereturns-relevanceoftheirestimatedannualunrealizedfairvaluegainsandlossesforfinancialinstruments.Tocapturedifferencesinmarketliquidityandfairvaluationdifficultiesacrosstypesoffinancialinstruments,wedistinguishunrealizedgainsandlossesthatarerecordedinnetincomeversusrecordedinothercomprehensiveincomeversuscalculableusingfinncialstatementnotedisclosures.Wepredictandgenerallyfindthatbanks’market(credit)riskmodelingenhancesthereturns-relevanceoftheirunrealizedfairvaluegainsandlosses,moresoforlessliquidinstrumentssubjecttogreatermarket-risk-related(credit-risk-related)valuationdifficultiesandduringperiodsforwhichmarket(credit)riskishigher.
Accounting, Organizations and Society 46 (2015) 81–95
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
The impact of risk modeling on the market perception of banks’
estimated fair value gains and losses for ?nancial instruments
R
Gauri Bhat
a,1
, Stephen G. Ryan
b,?
a
Cox School of Business, Southern Methodist University, 6212 Bishop Blvd., Dallas, TX 75275-0333, United States
b
Leonard N. Stern School of Business, New York University, 44 West 4th Street, Suite 10-73, New York, NY 10012-1118, United States
a r t i c l e i n f o
Article history:
Received 5 June 2014
Revised 15 April 2015
Accepted 16 April 2015
Available online 7 May 2015
a b s t r a c t
We examine whether and how measures of market and credit risk modeling identi?ed from banks’
?nancial reports enhance the returns-relevance of their estimated annual unrealized fair value gains and
losses for ?nancial instruments. To capture differences in market liquidity and fair valuation di?culties
across types of ?nancial instruments, we distinguish unrealized gains and losses that are recorded in net
income versus recorded in other comprehensive income versus calculable using ?nancial statement note
disclosures. We predict and generally ?nd that banks’ market (credit) risk modeling enhances the returns-
relevance of their unrealized fair value gains and losses, more so for less liquid instruments subject to
greater market-risk-related (credit-risk-related) valuation di?culties and during periods for which market
(credit) risk is higher. We obtain these ?ndings both for banks’ unadjusted risk modeling measures and
for the portions of these measures that we model as attributable to banks’ risk modeling activities, but
not for the portions we model as attributable to banks’ disclosure of these activities.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
We examine whether and how banks’ risk modeling enhances
the returns-relevance of their estimated annual unrealized fair
value gains and losses (FVGL) on ?nancial instruments. FVGL are
changes in fair value during periods that are not yet realized
through cash received or paid. When the markets for banks’ ?-
nancial instruments are su?ciently illiquid that observable market
inputs do not su?ce to determine the fair values of those instru-
ments, banks must estimate FVGL by developing valuation mod-
els and identifying the inputs necessary to implement those mod-
els. These activities require risk modeling both to predict uncertain
future cash ?ows and to determine appropriate rates to discount
those cash ?ows. To conduct risk modeling effectively, banks must
invest in adequate personnel and information systems and apply
managerial judgment appropriately and with discipline, with inad-
R
This paper was presented at the Accounting, Organizations & Society Confer-
ence on Accounting Estimates on October 23–25, 2014 sponsored by the Deloitte &
Touche LLP. We are grateful for helpful comments from the editor, Hun Tong Tan,
two anonymous reviewers, and conference participants, as well as accounting work-
shop participants at Rice University.
?
Corresponding author. Tel.: +1 (212) 998 0020.
E-mail addresses: [email protected] (G. Bhat), [email protected] (S.G. Ryan).
1
Tel.: +1 (214) 768 2964.
equate investment (self-interested application of judgment) intro-
ducing unintentional (intentional) noise and bias in FVGL. Banks’
investment in risk modeling and other risk management activities
that discipline fair value estimation appears to vary considerably
across banks and time.
2
Banks and their ?nancial instruments exhibit two primary types
of risk, market risk and credit risk. Banks engage in two corre-
sponding types of risk modeling activities, market risk modeling
(MRM) and credit risk modeling (CRM). Market risk is variability
in the value of a position attributable to changes in market prices.
Interest rate risk is the primary market risk for most banks. This
risk manifests through: (1) discounting effects, which are larger for
longer duration positions; (2) prepayment of ?xed-rate mortgage-
related assets (both securities and loans); and (3) the exercise of
other interest rate options, which may be standalone derivatives
or embedded in traditional ?nancial instruments. MRM involves
analyzing the durations of banks’ ?nancial instruments and the
resulting sensitivity of their net interest income and value of eq-
2
For example, Mikes (2011) discusses detailed case studies of two banks and ex-
tensive interviews at ?ve additional banks indicating that the quality of risk man-
agement varies considerably across banks and time. Only 28% (2%) of our sample
banks disclose in their Form 10-K ?lings that they employed a chief risk o?cer in
2013 (2002).http://dx.doi.org/10.1016/j.aos.2015.04.004
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
82 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
uity to interest rate movements. It also involves simulating the ef-
fects of interest rate movements on the prepayment of ?xed-rate
mortgages and exercise of other interest-rate options. Credit risk
is variability of the value of funded assets attributable to uncer-
tainty about default probabilities, losses given default, and timing
of default, as well as variability of the value of unfunded loan com-
mitments due to uncertainty about draws on those commitments,
which are more likely to occur during worse economic times. CRM
involves analysis of these parameters based on attributes of the
borrowers, borrowing contracts, borrowers’ performance to date on
the contracts, and relevant economic conditions.
We argue that banks’ MRM and CRM activities enhance the
quality of their estimates of FVGLs on ?nancial instruments when
two conditions hold: (1) the relevant markets for those instru-
ments are su?ciently illiquid that prices or other information from
these markets do not substantially determine the instruments’ fair
values; and (2) the instruments exhibit features, such as embed-
ded options or complex structuring, that increase the di?culty of
estimating the instruments’ fair values. As a ?rst cut to capture
the applicability of these conditions, we distinguish FVGL that are
recorded in net income versus recorded in other comprehensive in-
come versus calculable using ?nancial statement note disclosures
(c?disclosedd?). Fig. 1 summarizes current relevant U.S. generally
accepted accounting principles, under which FVGL are recorded in
net income for most trading and risk management instruments
and in other comprehensive income for available-for-sale securi-
ties and cash-?ow-hedge derivatives. FVGL are disclosed for most
of banks’ other primary types of ?nancial instruments, including
their largest asset, loans, and largest liability, deposits. We propose
three main hypotheses below that we test by examining whether
and how MRM and CRM enhance the returns-relevance of these
three types of FVGL from 2002 to 2013.
Our ?rst and most general hypothesis is that banks’ MRM and
CRM enhance the returns-relevance of their FVGL, more so for
FVGL on less liquid and more di?cult-to-fair-value ?nancial instru-
ments. In testing this hypothesis, we exploit the fact that banks’
?nancial instruments for which FVGL are disclosed, such as loans
and deposits, usually are less liquid and more di?cult to fair
value than their other instruments. Our second hypothesis is that
banks’ MRM also enhances the returns-relevance of their FVGL
recorded in other comprehensive income. Available-for-sale securi-
ties and cash-?ow-hedge derivatives typically are near credit risk-
less. Moreover, to the limited extent that banks experience credit
losses on these instruments, banks typically record these losses
in net income under impairment accounting rules. Hence, interest
rate risk is the primary risk re?ected in FVGL recorded in other
comprehensive income. We expect this hypothesis to hold only for
available-for-sale securities and cash-?ow-hedge derivatives that
are both less than highly liquid and exhibit fair valuation di?cul-
ties, such as mortgage-backed and asset-backed securities, so that
MRM is essential to estimate the fair values of the instruments
accurately. Our third hypothesis is that banks’ CRM primarily im-
pacts the returns-relevance of their disclosed FVGL, because banks
assume credit risk primarily through their funded loans and un-
funded loan commitments.
To test these hypotheses, we identify banks’ risk modeling ac-
tivities from disclosures in their Form 10-K ?lings. As described in
the Appendix c?Risk modeling measures and chief risk o?cer indi-
catord?, we hand collect disclosures of ?ve MRM activities (inter-
est rate gap analysis, interest rate sensitivity analysis, Value-at-Risk
analysis, stress testing, and backtesting) and four CRM activities
(statistical credit risk measurement, credit scoring, internal credit
risk rating, and stress testing). We equally weight these activities
to construct indices of banks’ MRM and CRM. This approach raises
the issue that many bank-year ?nancial reports include little about
risk modeling activities, particularly CRM early in our sample pe-
riod. Since all banks must engage in at least minimal levels of
MRM and CRM to make investment and ?nancing decisions and to
estimate the fair values of ?nancial instruments for which market
data do not su?ce for the task, it appears that some banks do not
disclose these activities. Hence, non-disclosure of a risk modeling
activity does not imply absence of the activity. We assume, how-
ever, that our MRM and CRM measures capture meaningful varia-
tion in risk modeling intensity across banks and time.
We test all hypotheses using both one-stage and two-stage ap-
proaches. The one-stage approach regresses returns for the twelve
months ending four months after the ?scal year end on net in-
come before FVGL recorded in net income
3
and the three types
of FVGL (recorded in net income, recorded in other comprehen-
sive income, and disclosed), separately and interacted with the un-
adjusted MRM and CRM measures, as well as control variables.
We frame and test our hypotheses as restrictions on the one or
more coe?cients on the interactions of banks’ unadjusted MRM
and CRM measures with speci?c types of FVGL. Empirical results
using this approach support our main hypotheses with one ex-
plainable exception.
We use the two-stage approach to help ensure that the one-
stage approach results are attributable to banks’ risk modeling ac-
tivities rather than to their choice to disclose these activities. In
this approach, we ?rst regress banks’ unadjusted MRM and CRM
measures on proxies for their discipline over risk modeling, techni-
cal sophistication, risk exposures, and risk tolerance, which we ex-
pect primarily indicate banks’ risk modeling activities rather than
their disclosure of those activities. We use the explained (unex-
plained) portions of banks’ unadjusted MRM and CRM measures
from these ?rst-stage models as measures of banks’ risk model-
ing activities (disclosure of these activities) in second-stage re-
turns models. The estimated coe?cients on the MRM and CRM
activity measures in the two-stage approach yield the same in-
ferences as the estimated coe?cients on the unadjusted measures
in the one-stage approach, whereas the estimated coe?cients on
the MRM and CRM disclosure measures generally are insigni?cant.
These results are consistent with the one-stage approach results
being driven by banks’ risk modeling rather than their disclosure
of that modeling.
We further hypothesize that MRM more strongly impacts the
returns-relevance of FVGL that are recorded in other comprehen-
sive income or disclosed in years with high interest rate volatility,
and that CRM more strongly impacts the returns-relevance of dis-
closed FVGL during the ?nancial crisis. To test these predictions,
we interact the primary test variables with indicator variables for
years with above-median interest rate volatility or the crisis period
2007–2009. Empirical results for the unadjusted MRM and CRM
measures and the MRM and CRM activity measures generally sup-
port these further hypotheses.
Our study contributes to the extensive literature beginning
with Barth (1994) that empirically examines the extent and de-
terminants of the value-relevance of fair values and the returns-
relevance of FVGL for ?nancial instruments. Our study is most re-
lated to recent papers examining disclosures of fair valuation in-
puts and other measures of the reliability of recognized fair value
estimates under Statement of Financial Accounting Standards (FAS)
157 (2006, Accounting Standards Codi?cation (ASC) 820), which
became effective in 2008. In particular, Chung, Goh, Ng, and Yong
3
Net income before FVGL recorded in net income includes realized gains and
losses that are distinct from FVGL except for two types of impairment write-down
that are included in net income and thus are accounted for in the same fashion
as realized losses. First, net income includes all or the credit loss portion of other-
than-temporary impairment write-downs of available-for-sale and held-to-maturity
securities. Second, net income includes losses on loans held for sale recognized at
fair value under the lower-of-cost-or-fair-value measurement basis.
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 83
Fig. 1. Summary of three ?nancial reporting treatments for fair value gains and losses on banks’ various types of ?nancial instruments.
(2014) develop measures of the reliability of recognized fair value
estimates from textual analysis of banks’ and insurers’ ?nancial
statement notes. Chung et al. provide evidence that these measures
are associated with enhanced market pricing and lower informa-
tion risk for recognized level 3 fair values. Our study complements
Chung et al. by examining an alternative measure of the reliability
of fair value estimates, a longer and more diverse sample period,
disclosed as well as recognized FVGL, and returns-relevance rather
than value-relevance.
Financial report information about estimated fair values of
?nancial instruments, literature review, and hypothesis
development
In this section, we ?rst describe the various bases in account-
ing standards, accounting research, and other sources for our main
hypotheses about the effects of MRM and CRM on the returns-
relevance of FVGL. We then support and state our further hypothe-
ses about the distinct conditions under which we expect MRM and
CRM to be particularly useful.
Required information about estimated fair values and FVGL
Under current U.S. generally accepted accounting principles,
?rms recognize certain types of ?nancial instruments at fair value
on the balance sheet. They record FVGL in net income for some of
these instruments and in other comprehensive income otherwise.
Banks’ primary types of ?nancial instruments currently recognized
at fair value with unrealized gains and losses recorded in net in-
come are: trading securities under FAS 115 (1993, ASC 320); non-
accounting-hedge and fair value hedge derivatives as well as fair
value hedged items under FAS 133 (1998, ASC 815); most other
trading instruments under industry accounting principles or prac-
tices; and ?nancial instruments for which the fair value option
is selected under FAS 155 (2006, ASC 815.15) and FAS 159 (2007,
ASC 825.10). The primary types of ?nancial instruments that cur-
rently are recognized at fair value with unrealized gains and losses
recorded in other comprehensive income are available-for-sale se-
curities under FAS 115 and cash-?ow-hedge derivatives under FAS
133.
4
We denote FVGL that are recorded in net income (other com-
prehensive income) by NIGL (OCIGL).
FAS 107 (1991, ASC 825.10.50) requires ?rms to disclose the fair
and carrying values of most types of ?nancial instruments in the
notes to ?nancial statements. FAS 107 disclosures are the only in-
formation in banks’ ?nancial reports about the fair values of their
on-balance sheet loans, deposits (excluding core deposit intangi-
bles), and debt as well as off-balance sheet instruments such as
loan commitments; FAS 115 requires these and other disclosures
for held-to-maturity securities. We denote FVGL that can only be
calculated from disclosures by DISCGL. FAS 107 also requires ?rms
to disclose the method(s) and signi?cant assumptions used to es-
timate the fair value of ?nancial instruments. Banks ful?lled this
requirement in boilerplate and minimally informative fashions un-
til the effective date of FAS 157.
FAS 157 de?nes fair value as c?the price that would be received
to sell an asset or paid to transfer a liability in an orderly trans-
action between market participants at the measurement date,d?
and it creates a three-level hierarchy of fair value inputs.
5
Level
1 (highest quality) inputs are quoted prices in active markets for
the identical item. Level 2 inputs are quoted prices in markets that
are not active for the identical item or in active markets for similar
items and most other observable information. Level 3 (lowest qual-
ity) inputs are unobservable reporting-?rm-supplied inputs. The
level of a fair value estimate is determined by the level of its low-
est quality signi?cant input. FAS 157, as amended by FASB Staff Po-
sition (FSP) FAS 157-4 (2009, ASC 820.10.50) and Accounting Stan-
dards Update (ASU) 2010-06 (2010, ASC 820.10.50), requires quar-
terly disclosures of the amounts of each major category of assets
and liabilities that is recognized at fair value on the balance sheet
4
Other standards require certain types of ?nancial instruments to be recognized
at fair value at either: (1) inception, e.g., FAS 166 (2009, ASC 860) for retained inter-
ests in securitizations; or (2) the time of (other than temporary) impairment write-
downs, e.g., FAS 65 (1982, ASC 948.310.35) and SOP 01-6 (2001, ASC 310.10.35) for
held-for-sale loans and FAS 115, as amended by FSP FAS 115-2 & FAS 124-2 (2009,
ASC 320.10.35), for available-for-sale and held-to-maturity securities.
5
Paragraphs C21–C22 of FAS 157 indicates that generally accepted accounting
principles include various practicability exceptions to fair value measurement that
FAS 157 did not change. While banks do not often invoke these exceptions in their
?nancial reports, the exceptions may have some effect on our FVGL measures or be
correlated with our risk modeling measures; e.g., a bank with better risk modeling
may have less need to invoke a practicability exception.
84 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
in each of the three levels, separately for items fair valued each
reporting period versus on a non-recurring basis. For level 2 and
level 3 fair values, ?rms must disclose their valuation techniques
and inputs. For level 3 fair values, ?rms must disclose rollforwards
of the fair values from the beginning to the end of the period, dis-
tinguishing total gains and losses, purchases, sales, issuances, set-
tlements, and transfers in and out of level 3.
6
Prior empirical research on the differential value-relevance of fair
values and returns-relevance of FVGL by type of ?nancial instrument
and by recognition versus disclosure
An extensive literature beginning with Barth (1994) empirically
examines the value-relevance of estimated fair values and returns-
relevance of FVGL on ?nancial instruments. Most of these studies
limit their samples to banks because ?nancial instruments dom-
inate banks’ balance sheets, banks’ ?nancial instruments exhibit
varying levels of liquidity and other factors associated with the
need or ability to exercise discretion over fair value estimation, and
analysis of a single industry mitigates concerns about unmodeled
heterogeneity. Given that several recent and extensive surveys of
this literature exist (e.g., Ryan, 2011, Section 4.5), we brie?y sum-
marize the four main ?ndings in these studies that bear on the
hypotheses developed below.
First, for investment securities and other similarly liquid assets,
fair values are value-relevant and FVGL are returns-relevant, al-
though the returns-relevance results are more sensitive to model
speci?cation and sample (Ahmed & Takeda, 1995; Barth, 1994; Car-
roll, Linsmeier, & Petroni, 2003; Danbolt & Rees, 2008). Second,
inconsistent (essentially no) evidence has been generated as to
whether the fair values of loans and derivatives (other ?nancial
instruments) are value-relevant (Barth, Beaver, & Landsman, 1996;
Eccher, Ramesh, & Thiagarajan, 1996; Nelson, 1996; Venkatacha-
lam, 1996). Third, banks’ disclosed fair values of ?nancial instru-
ments appear to be both noisy (i.e., unexplainable based on con-
temporaneous economic events) and managed to make less sol-
vent banks appear more so. Surprisingly, the value-relevance of
disclosed fair values does not appear to be affected by noise, but
it is reduced by discretionary behavior (Beaver & Venkatchalam,
2003; Nissim, 2003). Fourth, fair values that are recognized (i.e.,
more prominent) are more value-relevant than those that are dis-
closed (Ahmed, Kilic, & Lobo, 2006; Badertscher, Burks, & Eas-
ton, 2014; Chambers, Linsmeier, Shakespeare, & Sougiannis, 2007;
Dong, Ryan, & Zhang, 2014; Hirst, Hopkins, & Wahlen, 2004). Col-
lectively, these ?ndings suggest that estimated fair values and
FVGL for which the signal-to-noise ratio is (perceived to be) higher
have higher value-relevance and returns-relevance, respectively.
We examine returns-relevance rather than value-relevance be-
cause the latter is both more extensively studied in the prior lit-
erature and more subject to omitted variables (but less subject
to measurement error). Barth, Beaver, and Landsman (2001) and
Holthausen and Watts (2001) discuss these methodological trade-
offs.
6
While a massive improvement over FAS 107 disclosures, FAS 157 disclosures ex-
hibit three limitations that make them unsuitable for use in our study. First, these
disclosures are available for the bulk of our sample beginning in 2008, less than half
our sample period and a period dominated by the ?nancial crisis and its aftermath.
Hence, controlling for these disclosures would essentially eliminate our ability to
compare the crisis and non-crisis periods. Second, the most important FAS 157 dis-
closures pertain to level 3 recognized fair values for ?nancial instruments, which
constitute only about 2% of banks’ recognized ?nancial instruments. Third, these
disclosures do not encompass disclosed fair values for the bulk of banks’ ?nancial
instruments.
Prior empirical research on the differential value-relevance of fair
values of ?nancial instruments by FAS 157 level and other reliability
disclosures
Using FAS 157-required disclosures, several studies provide ev-
idence that banks’ recognized fair values of ?nancial instruments
estimated using higher quality inputs are more value-relevant
(Goh, Li, Ng, & Yong, 2015; Kolev, 2009; Song, Thomas, & Yi, 2010).
These studies all ?nd that level 1 and 2 fair values are more value-
relevant than level 3 fair values. Goh et al. ?nd that level 1 fair
values are more value-relevant than level 2 fair values. Song et al.
?nd that level 3 fair values are more value-relevant for banks with
better corporate governance.
Chung et al. (2014) develop binary (disclosure made versus not)
and continuous (number of words in the disclosure divided by to-
tal number of words in the Form 10-K ?ling) measures of FAS 157-
encouraged disclosures in banks’ and insurers’ ?nancial statement
notes about the controls, processes, and procedures used to assure
the reliability of their recognized fair value estimates. Chung et al.
(2014) provide evidence that these measures are associated with
enhanced market pricing (higher share price and lower priced risk)
and lower information risk (higher analyst consensus) for recog-
nized fair values estimated using signi?cant level 3 inputs.
Collectively, these ?ndings suggest that disclosures indicating
that banks more reliably estimate the fair value of their ?nancial
instruments are associated with enhanced value-relevance of these
estimates.
Hypotheses
We discuss banks’ MRM and CRM activities in the introduc-
tion and repeat this discussion here only insofar as it pertains di-
rectly to speci?c hypotheses. All of our hypotheses re?ect the view
that risk modeling typically improves fair value estimation. We ac-
knowledge that risk modeling may instead deteriorate fair value
estimation if it crowds out the appropriate use of judgment or de-
volves into a compliance exercise, as Mikes (2011) and Kaplan and
Mikes (2012) discuss occurred at speci?c banks.
Our ?rst and most general hypothesis is that banks’ MRM and
CRM enhance the returns-relevance of their FVGL, particularly for
less liquid ?nancial instruments that exhibit lower quality mar-
ket information (i.e., less c?price transparencyd?) and thus greater
need for risk modeling to estimate FVGL reliably. This hypothe-
sis is supported by four prior theoretical and empirical literatures:
(1) the theoretical literature indicating that market prices respond
more strongly to more precise information (e.g., Holthausen & Ver-
recchia, 1988); (2) the empirical literature on the value-relevance
of fair values and returns-relevance of FVGL by type of ?nancial
instruments and by recognition versus disclosure discussed above;
(3) the empirical literature that demonstrates that FAS 157 fair
value input level and voluntary fair value reliability disclosures af-
fect this value-relevance discussed above; and (4) Bhat, Ryan, and
Vyas’s (2014) ?ndings that banks’ CRM is positively associated with
the timeliness of their loan loss provisions. We formally state this
hypothesis and all subsequent hypotheses in alternative form.
H1. MRM and CRM enhance the sensitivity of returns to FVGL,
more so for FVGL on less liquid and more di?cult-to-fair-value ?-
nancial instruments.
In testing Hypothesis H1, we exploit the fact that banks’ ?nan-
cial instruments generating DISCGL, such as loans and deposits,
usually are less liquid and more di?cult to fair value than their
other ?nancial instruments. As discussed below, the ?nancial in-
struments that generate OCIGL are a mixed bag of highly liquid
and thus easy-to-fair-value instruments and less liquid and more
di?cult-to-fair-value instruments. The instruments that generate
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 85
NIGL usually are highly liquid and easy to fair value. Hence, we ex-
pect Hypothesis H1 to hold most strongly for DISCGL, less strongly
for OCIGL, and not at all for NIGL.
Our second hypothesis is that banks’ MRM enhances the
returns-relevance of their OCIGL, particularly for mortgage-backed
and asset-backed securities. This hypothesis has three interrelated
bases. First, interest rate risk is the primary risk re?ected in OCIGL,
which are generated by available-for-sale securities and cash-?ow-
hedge derivatives. Banks’ most common types of available-for-sale
securities are U.S. Treasuries, other U.S. federal government se-
curities, and agency-guaranteed mortgage-backed securities. Their
most common type of cash-?ow-hedge derivatives are interest rate
derivatives with highly creditworthy counterparties. These instru-
ments typically are near credit riskless. Moreover, to the limited
extent that banks experience credit losses on these instruments,
they typically record these losses in net income under impairment
accounting rules.
Second, interest rate risk manifests through: (1) discounting,
which more strongly affects longer duration positions; (2) pre-
payment of ?xed-rate mortgages underlying mortgage-backed and
other asset-backed securities; and (3) the exercise of other interest
rate options, which may be standalone cash-?ow-hedge derivatives
or embedded in available-for-sale securities (Ryan, 2011, Section
2.4). Banks must use considerably more sophisticated MRM (e.g.,
interest rate simulation) to determine the effects of prepayment
and other interest rate options than to determine discounting ef-
fects. Relatedly, it is considerably more di?cult for banks to hedge
options than discounting effects.
Third, compared to banks’ other types of available-for-sale secu-
rities, mortgage-backed securities and other types of asset-backed
securities typically are both less liquid and subject to greater
fair valuation di?culties, due to their embedded prepayment
options and complex structuring (e.g., waterfalls). Even agency-
guaranteed mortgage-backed securities generally exhibit low price
transparency unless and until speci?c securities are identi?ed in
to-be-announced trades (Vickrey & Wright, 2013).
7
H2. MRM enhances the sensitivity of returns to OCIGL, especially
OCIGL on less liquid and more di?cult-to-fair-value mortgage-
backed and asset-backed available-for-sale securities.
Our third hypothesis is that banks’ CRM primarily impacts the
returns-relevance of their DISCGL. This hypothesis is motivated
by the fact that banks assume credit risk primarily through their
funded loans and unfunded loan commitments for which esti-
mated fair values are disclosed (Ryan, 2011, Section 2.5).
H3. CRM primarily enhances the sensitivity of returns to DISCGL.
Our fourth and ?fth hypotheses are motivated by the intuition
that banks’ risk modeling should be more important during peri-
ods when the modeled risk is higher. In particular, MRM (CRM)
should be more important during years when interest rates are
more variable (credit losses are unexpectedly higher).
H4. The effect of MRM is stronger during years with above-median
interest rate volatility than in other years.
7
Vickrey and Wright (2013) provide evidence that to-be-announced agency
mortgage-backed securities trades (which are accounted for as derivatives, not in-
vestment securities, until settled) exhibit high volume and price transparency. They
also explain, however, that these trades exhibit various features (e.g., settlement
at a single date each month and the cheapest-to-deliver option) that reduce the
price transparency these trades provide for existing pools of agency-guaranteed
mortgage-backed securities.
H5. The effect of CRM is stronger during the ?nancial crisis than
in other years.
Measurement of variables and methodology
Fair value gains and losses
We measure NIGL as trading revenue reported on line 5.c of
Schedule HI and OCIGL as other comprehensive income reported
on line 4 of Schedule HC-R of Y-9C ?lings. Unfortunately, both NIGL
and OCIGL include items other than FVGL; NIGL includes fee in-
come and realized gains and losses and OCIGL includes other com-
prehensive income from pensions and foreign currency.
8
For ?nan-
cial assets (liabilities) for which fair values are disclosed, we cal-
culate DISCGL as (minus) the change during the year of the differ-
ence between the fair value and carrying value of the instrument.
9
We obtain disclosed fair and carrying values from SNL Financial.
All FVGL components are after taxes calculated using a statutory
tax rate of 35% and divided by beginning-of-year market value of
equity.
Unadjusted risk modeling measures
To construct our MRM and CRM measures, we had to decide
which risk modeling activities to include in the measures and also
the disclosure medium to examine for these activities. To make
the ?rst decision, we considered all of the market and credit risk
modeling activities (including related risk management processes)
mentioned in a 2001 public disclosure survey of large interna-
tional banks conducted by the Basel Committee on Banking Su-
pervision.
10
To accommodate the U.S. banks in our sample, we ex-
cluded risk modeling activities for which disclosures are required
for U.S. banks or that are not provided by any sample bank in any
sample year. This yields ?ve market risk modeling activities (inter-
est rate gap analysis, interest rate sensitivity analysis, Value-at-Risk
analysis, stress testing, and backtesting) and four credit risk mod-
eling activities (statistical credit risk measurement, credit scoring,
internal credit risk rating, and stress testing). We describe these
activities in detail in the Appendix c?Risk modeling measures and
chief risk o?cer indicatord?.
We choose banks’ annual Form 10-K ?lings as the sole disclo-
sure medium for three reasons. First, banks provide the most ex-
tensive disclosures of their risk modeling activities in these ?lings,
primarily due to SEC requirements. For example, SEC FRR 48 re-
quires that banks make extensive disclosures of their market risk
in these ?lings. Second, the extents of disclosure in Form 10-K ?l-
ings and other media are positively correlated (Lang & Lundholm,
1993). Third, our analysis of samples of banks’ ?nancial analyst re-
ports and conference call transcripts yielded no mention of risk
modeling.
We hand collected banks’ risk modeling disclosures from their
Form 10-K ?lings from 2001 to 2013. We electronically searched
8
In untabulated analysis, we tried to remove fee income from NIGL by subtract-
ing its average over the prior three or ?ve years. This transformation of NIGL should
mostly remove fee income, which is much more persistent than fair value gains and
losses. No inferences are affected by this alternative speci?cation of NIGL.
9
Because provisions for credit losses on loans and loan commitments are
recorded in net income, these recognized accruals are not included in DISCGL. This
likely reduces the power of our tests of hypotheses H1, H3, and H5, because prior
research shows that CRM is associated with these accruals (Bhat et al., 2014).
10
While these documents also consider operational risk modeling activities, we
do not deem this type of activity to be directly related to the estimation of fair
values and FVGL for ?nancial instruments. We acknowledge that Chernobai, Jorion,
and Yu (2011) provide evidence that operational risk is positively correlated with
credit risk.
86 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
those ?lings for a large number of relevant words or phrases, such
as c?risk management,d? c?risk model,d? c?scoringd? and c?risk
ratingd?. We read each identi?ed disclosure to verify that the risk
modeling activity was disclosed. We construct an indicator for each
activity that takes a value of one if it is disclosed and zero other-
wise. Our unadjusted market risk modeling measure, MRM (credit
risk modeling measure, CRM), is the simple average of the indica-
tors for the ?ve market risk activities (four credit risk activities).
As averages of indicator variables, MRM and CRM take values from
zero to one.
Several caveats are in order regarding our risk modeling mea-
sures. First, we observe banks’ disclosures of risk modeling activ-
ities, not the activities themselves. Presumably the disclosure of
risk-modeling activities implies banks engage in the activities, be-
cause there does not appear to be any reason for banks to lie in
these disclosures, which generally are too terse to divulge mean-
ingful proprietary information.
11
However, the non-disclosure of an
activity does not imply that the bank does not engage in the activ-
ity. As discussed in the introduction and below, we employ a two-
stage approach to help ensure that our results are attributable to
banks’ risk modeling activities rather than their disclosure of those
activities. Second, our risk modeling measures capture the exis-
tence of particular activities more than the quality of those activi-
ties. Third, and relatedly, despite robustness checks to incorporate
various different weighting approaches, simple averaging of the in-
dicators for risk modeling activities might not re?ect the weights
investors assign to these activities. Lastly, despite careful reading
of each disclosure, our coding of risk modeling activities might be
subjective. The large number of disclosures involved makes coding
by multiple individuals costly, however.
Chief risk o?cer indicator
We use an indicator variable for whether banks employ a chief
risk o?cer, CRO, as our measure of banks’ discipline over risk mod-
eling. We considered using standard proxies for corporate gover-
nance for this purpose, but these proxies do not appear to cor-
respond strongly, if at all, with the estimation of fair values and
FVGL.
12
Consistent with this view, Aebi, Sabato, and Schmid (2012)
?nd that these standard proxies c?are mostly insigni?cantly or
even negatively related to the banks’ performance during the cri-
sis,d? whereas banks that employ chief risk o?cers that report di-
rectly to the board of directors exhibit superior performance dur-
ing the crisis. Ellul and Yerramilli (2013) ?nd similar results for
their risk management index, which depends strongly on the em-
ployment and status of the chief risk o?cer.
CRO takes a value of one if the bank discloses that it employs a
chief risk o?cer in its Form 10-K ?ling and zero otherwise. We
searched ?lings to determine whether the words c?chiefd? and
c?riskd? appear within ?ve words of each other and, if so, whether
the words c?riskd? and either c?o?cerd? or c?directord? appear
within ?ve words of each other. For ?lings that met both ?lters,
we read the relevant passages to con?rm the existence of a chief
11
In addition, disclosures of risk modeling are subject to internal controls, to CEO
and CFO certi?cation of ?nancial reports under Section 302 of the Sarbanes–Oxley
Act, and to audit (reading for consistency) by banks’ auditors if the disclosures
appear in the ?nancial statement notes (MD&A) sections of ?nancial reports. The
observed frequency of disclosures is fairly low, inconsistent with easily imitated
c?cheap talk.d?
12
More generally, Larcker, Richardson, and Tuna (2007) argue that construct valid-
ity issues with corporate governance measures cause mixed and hard-to-interpret
results in many studies, in part due to the absence of theory indicating which cor-
porate governance variables matter in what contexts.
risk o?cer. CRO takes a value of one for these bank-year observa-
tions and zero otherwise.
13
First-stage models explaining risk modeling
In the two-stage approach, the ?rst-stage models regress each
of the unadjusted MRM and CRM measures (collectively denoted
XRM) on seven proxies for banks’ discipline over risk modeling,
technical sophistication, risk exposures, and risk tolerance, all of
which we expect primarily to affect the extent of banks’ risk mod-
eling activities rather than their disclosure of those activities. These
proxies are: CRO, a proxy for banks’ discipline over risk mod-
eling; log of total assets (SIZE) and a trading portfolio indica-
tor (TRADD), proxies for banks’ technical sophistication; 0–1 year
repricing gap (INTSEN), the proportion of assets that are commer-
cial loans (COMMLOAN), and the standard deviation of fair value
income de?ated by beginning market value of equity over the past
ten years (FVISTD), proxies for banks’ interest rate risk, credit risk,
and overall risk, respectively; and tier 1 risk-based capital ratio
(TIER 1), a proxy for banks’ solvency or risk tolerance. We also in-
clude year ?xed effects to capture changes in the risk modeling
measures, particularly CRM, over time. This model is:
XRM = ? +?
1
CRO +?
2
SIZE +?
3
TRADD +?
4
INTSEN
+?
5
COMMLOAN +?
6
FVISTD +?
7
TIER1
+year ?xed effects +? (1)
We suppress time subscripts except when necessary for clarity.
We use the portions of the risk modeling measures explained by
the seven proxies, denoted with hats, i.e.,
XRM, as measures of
risk modeling activities. We use the remaining portions of the risk
modeling measures, i.e., XRM ?
XRM, as measures of risk modeling
disclosures.
14
Models of the impact of risk modeling on the
returns-relevance of FVGL and hypotheses restated as
coe?cient restrictions
The returns models using the unadjusted risk modeling mea-
sures (i.e., in the one-stage approach) regress share returns for the
12 months ending four months after the ?scal year end (R)
15
on
net income before FVGL recorded in net income (NIBNIGL), the
three types of FVGL (NIGL, OCIGL, and DISCGL), the risk modeling
measure under consideration (either MRM or CRM), interactions
between that risk modeling measure and each of NIBNIGL and the
13
Using data available on banks’ top ?ve o?cers from SNL Financial, we endeav-
ored to determine whether CROs were one of these o?cers, but this determination
cannot be made from these data.
14
In untabulated robustness analyses, we use an alternative ?rst stage approach
that does not alter any inference in the second stage. Speci?cally, we include ?rm
attributes, that the disclosure literature surveyed by Healy and Palepu (2001) ?nds
to be associated with disclosure, as additional explanatory variables in Eq. (1): an
indicator for issuance of debt or equity, i.e., capital market transactions that mo-
tivate disclosure; an indicator for negative net income, a proxy for performance;
the book-to-market ratio, a proxy for growth; number of analysts forecasting earn-
ings and percentage institutional ownership, proxies for information environment.
We use the portions of the risk modeling measures explained by these proxies as
measures of risk modeling disclosures.
15
We use this return window because some sample banks are non-accelerated
?lers that ?le their Form 10-K ?lings on the last day of the third month following
the end of the ?scal year.
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 87
three types of FVGL, and year ?xed effects.
16
These models are:
R = a + b
1
NIBNIGL + b
2
NIGL + b
3
OCIGL + b
4
DISCGL
+b
5
(NIBNIGL ? XRM) + b
6
(NIGL ? XRM) + b
7
(OCIGL ? XRM)
+b
8
(DISCGL ? XRM) + b
9
XRM +year ?xed effects + e (2A)
The returns models using the risk modeling activity and disclosure
measures from the ?rst-stage model Eq. (1) (i.e., in the two-stage
approach) follow directly from substituting
XRM and XRM ?
XRM
for XRM in Eq. (2A). These models are:
R = a + b
1
NIBNIGL + b
2
NIGL + b
3
OCIGL + b
4
DISCGL
+b
5
(NIBNIGL ?
XRM) + b
6
(NIGL ?
XRM) + b
7
(OCIGL ?
XRM)
+b
8
(DISCGL ?
XRM) + b
9
XRM + b
10
(NIBNIGL ? {XRM ?
XRM})
+b
11
(NIGL ? {XRM ?
XRM}) + b
12
(OCIGL ? {XRM ?
XRM})
+b
13
(DISCGL ? {XRM ?
XRM}) + b
14
{XRM ?
XRM}
+year ?xed effects + e (2B)
We restate Hypotheses H1–H5 as coe?cient restrictions in Eqs.
(2A) and (2B). Hypothesis H1 predicts that b
8
is positive and
larger than b
6
and b
7
for both
MRM and
CRM. Hypothesis H2 pre-
dicts that b
7
is positive for
MRM, especially when OCIGL are for
mortgage-backed and asset-backed securities. Hypothesis H3 pre-
dicts that the primary effect of
CRM is manifested in b
8
. Hy-
pothesis H4 predicts that Hypotheses H1 and H2 pertaining to
MRM hold more strongly in years with above-median interest rate
volatility. Hypothesis H5 predicts that Hypotheses H1 and H3 per-
taining to
CRM hold more strongly during the ?nancial crisis.
Sample selection and data
Sample selection
We study banks for the same reasons as the prior literature
discussed in section ‘Financial report information about estimated
fair values of ?nancial instruments, literature review, and hypoth-
esis development’. In addition, the cost of hand collecting our risk
modeling and CRO variables limits the sample size.
As summarized in Table 1, Panel A, the sample comprises all
public U.S. bank holding companies with data available from 2002
to 2013 from the following ?ve sources: most ?nancial data from
Y-9C regulatory ?lings on the Federal Reserve Bank of Chicago
website, disclosed fair values and carrying values of ?nancial in-
struments from SNL Financial, risk modeling disclosures from Form
10-K ?lings on 10-K Wizard, and stock information from CRSP.
17
The total assets threshold at which bank holding companies are
16
In Eqs. (2A) and (2B), we interact NIBNIGL with the unadjusted risk modeling
measures because certain components of NIBNIGL should be affected by risk mod-
eling, e.g., provisions for loan losses should be affected by CRM. We expect FVGL
to be more strongly affected than NIBNIGL by risk modeling, however; to this ex-
tent, the interaction of NIBIGL with the unadjusted risk modeling measures can be
viewed as a placebo test.
17
We start the sample in 2002 because FAS 133 became effective in 2001 and
two consecutive years of disclosed fair and carrying values are necessary to calcu-
late DISCGL. FAS 133 requires all derivatives to be recognized at fair value on the
balance sheet, with unrealized gains and losses recorded in net income for non-
accounting hedge and fair-value-hedge derivatives and in other comprehensive in-
come for cash-?ow-hedge derivatives. Compared to prior accounting principles and
accounting practices, FAS 133 signi?cantly increased the amount of FVGL recorded
in net income and decreased the amount of FVGL disclosed; the standard also af-
fected the FVGL recorded in other comprehensive income in a less obviously direc-
tional fashion.
required to ?le Y-9Cs increased from $150 million to $500 million
in 2006; we impose the latter size restriction for all bank-year ob-
servations to yield a more comparable sample through time. Some
bank holding companies that ?le Y-9Cs differ slightly from the cor-
responding public companies that ?le Form 10-Ks.
Table 1, Panel A reports the construction of the ?nal sample of
238 unique banks and 2413 bank-year observations for the years
2002–2013. To mitigate the effects of outliers, we winsorize all
continuous model variables at the 0.5% and 99.5% levels of their
distributions.
Descriptive statistics
Table 1, Panel B presents descriptive statistics for the main vari-
ables. As is typical in banking studies, the size-related variables
exhibit considerable variation and right skewness. Despite the ?-
nancial crisis, the vast majority of bank-year observations are prof-
itable and well-capitalized, with the 25th percentile of NIBNIGL
equaling 4.9% of beginning-of-year market value of equity and of
the Tier 1 risk-based capital ratio equaling 10.3%. The distribution
of R is quite spread, however, re?ecting that fact that banks expe-
rienced good stock performance both early and late in the sample
period and poor performance during the crisis.
All three types of FVGL have means and medians of approxi-
mately zero. The spread of DISCGL is much higher than the spread
of NIGL and, to a lesser extent, OCIGL, however, because fair values
for banks’ primary ?nancial asset (loans) and ?nancial liability (de-
posits) are disclosed and only 21.8% of banks (mean TRADD) hold
any trading positions. The median bank-year discloses two out of
?ve MRM activities and one out of four CRM activities. 13.6% of
bank-years employ a chief risk o?cer.
Table 1, Panel C reports the number of observations, the means
of MRM, CRM, and CRO, and the frequency with which each of
the indicators for the risk modeling activities underlying MRM and
CRM take a value of one in each year from 2002 to 2013. At least
170 banks appear in the sample each year. The number of banks
is highest in the middle sample years, peaking at 221 in 2009. The
lesser number of banks in the early (later) years is attributable to
our requirement that observations have at least $500 million total
assets (bank failures and mergers and acquisitions resulting from
the crisis and passage of time). The frequency of banks’ disclosures
of MRM activities is ?at over the sample period due to the 1998
effective date of the main market risk disclosure requirement, FRR
48. In contrast, the frequency of banks’ disclosures of CRM activi-
ties and CRO increases strongly over the sample years.
Table 2 reports the Pearson (Spearman) correlations for the
main variables in the regression; we discuss only the Pearson
correlations. Three insights emerge from this table. First, MRM
and CRM are both signi?cantly positively associated with most of
the variables capturing size and sophistication, such as CRO, SIZE,
NIGL, and TRADD. Second, there are interpretable differences be-
tween the correlations involving MRM and CRM. For example, NIGL
is more positively correlated with MRM than with CRM, consis-
tent with trading-oriented banks primarily trading ?nancial instru-
ments with more market risk than credit risk. In contrast, TRADD
is more positively correlated with CRM than MRM, consistent with
large and sophisticated banks being relatively more likely to con-
duct CRM. Consistent with banks’ chief risk o?cers primarily being
tasked with evaluating credit risk, CRO is more positively corre-
lated with CRM than MRM. Third, unlike NIGL, neither OCIGL nor
DISCGL is correlated with any of the risk modeling measures or
measures of bank size or sophistication.
88 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Table 1
Sample selection and descriptive statistics.
#Bank-years #Banks
Panel A: Sample selection
Bank-years and banks with Y-9C data and total assets >$500 million in 2002–2013 3664 443
Above bank-years and banks with fair value data from SNL Financial 3282 405
Above bank-years and banks with CRSP data 3244 405
Above bank-years and banks with Form 10-K ?lings on 10-K Wizard 2437 239
Final sample bank-years and banks in 2002–2013 with available variables in at least two consecutive years in 2001–2013 (to allow one lag) 2413 238
Variable Mean Stddev Q1 Median Q3
Panel B: Summary statistics for 238 banks from 2002 to 2013 (2413 bank-years)
R 0.060 0.326 ?0.097 0.058 0.224
NIBNIGL 0.004 0.315 0.049 0.067 0.081
NIGL 0.001 0.007 0.000 0.000 0.000
OCIGL ?0.001 0.036 ?0.011 ?0.001 0.009
DISCGL ?0.001 0.224 ?0.034 0.000 0.034
MRM 0.314 0.145 0.200 0.400 0.400
CRM 0.211 0.209 0.000 0.250 0.250
CRO 0.136 0.342 0.000 0.000 0.000
SIZE 14.730 1.531 13.655 14.342 15.388
TRADD 0.218 0.413 0.000 0.000 0.000
INTSEN 1.699 1.987 0.528 1.174 2.075
COMMLOAN 0.121 0.074 0.066 0.105 0.160
FVISTD 0.269 0.338 0.079 0.139 0.312
TIER1 12.328 2.923 10.310 11.870 13.880
Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Panel C: Number of observations, risk modeling measures, and CRO by year
N 170 189 198 215 219 217 216 221 207 194 188 179
Means of risk modeling measures and CRO
MRM 0.327 0.324 0.320 0.316 0.315 0.313 0.311 0.309 0.308 0.304 0.312 0.316
CRM 0.113 0.135 0.145 0.155 0.156 0.162 0.175 0.215 0.301 0.323 0.334 0.332
CRO 0.020 0.042 0.061 0.070 0.087 0.115 0.139 0.163 0.193 0.232 0.234 0.279
MRM components
INTGAP 135 149 147 155 154 148 143 142 131 113 110 104
IRSIMUL 131 146 159 177 176 176 176 182 170 157 156 150
VAR 8 7 8 8 9 8 7 7 7 8 10 10
MRSTRESS 1 1 1 3 3 5 5 5 5 9 9 9
MRBACKTEST 3 3 2 2 3 3 4 5 6 8 8 10
CRM components
MODEL 9 13 16 22 23 22 24 35 48 48 45 45
SCORING 13 17 18 17 19 21 18 22 23 24 27 25
RR 53 71 79 91 91 93 97 113 158 147 144 136
CRSTRESS 2 1 2 3 4 5 12 20 20 32 35 32
Notes: Panel A reports the sample selection process. Panel B reports summary statistics for the primary model variables, including net income before fair value gains
and losses recorded in net income (NIBNIGL), fair value gains and losses recorded in net income (NIGL), fair value gains and losses recorded in other comprehensive
income (OCIGL), disclosed fair value gains and losses (DISCGL), the risk modeling measures (MRM, and CRM), chief risk o?cer (CRO), and other bank characteristics.
Panel C reports the means of the risk modeling measures and CRO by year, as well as and the number of banks that disclose of the components of each risk modeling
measure by year. All variables are de?ned in the Appendix c?De?nitions of variables and other acronymsd? except for the risk modeling components which are de?ned
in the Appendix c?Risk modeling measures and chief risk o?cer indicatord?.
Empirical analysis
First-stage regression results
Table 3 reports the estimation of Eq. (1), the ?rst-stage models
used to explain the unadjusted risk modeling measures in terms
of risk-related explanatory variables and ?xed year effects (untab-
ulated). The ?rst (second) column presents the results for the un-
adjusted market (credit) risk modeling measure MRM (CRM). The
?t is considerably better for the CRM model, primarily because the
year ?xed effects are more signi?cant in this model due to CRM’s
strong upward trend over time reported in Table 1, Panel C.
The coe?cient on CRO is signi?cantly positive at the 5% level
in the MRM model and at the 10% level in the CRM model. The
coe?cient on SIZE is signi?cantly positive at the 1% level in both
models. These results indicate that banks’ discipline over fair value
estimation and technical sophistication explain their risk modeling.
In the MRM model, the coe?cient on INTSEN is signi?cantly neg-
ative at the 5% level and the coe?cient on FVISTD is signi?cantly
negative at the 10% level, suggesting that banks with higher inter-
est rate risk tend to seek out or tolerate rather than manage that
risk. In contrast, the coe?cient on FVISTD is positive at the 10%
level in the CRM model. No other explanatory variable is signi?-
cant in any model.
Primary regression results
Table 4, Panel A reports the estimations of Eqs. (2A) and (2B),
which we use to test hypotheses H1–H5 about whether and how
banks’ risk modeling impacts the returns-relevance of their three
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 89
Table 2
Correlations.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
R (1) 0.485 0.069 ?0.085 ?0.026 0.009 0.071 0.012 ?0.016 0.018 0.113 0.016 ?0.104 0.195
0.000 0.001 0.000 0.194 0.675 0.001 0.572 0.435 0.372 0.000 0.440 0.000 0.000
NIBNIGL (2) 0.201 ?0.076 ?0.049 0.018 ?0.050 ?0.011 ?0.019 ?0.100 ?0.052 ?0.018 ?0.048 ?0.132 0.214
0.000 0.000 0.016 0.385 0.014 0.575 0.354 0.000 0.011 0.375 0.019 0.000 0.000
NIGL (3) 0.106 ?0.068 ?0.004 0.020 0.077 0.209 0.202 0.400 0.669 0.177 0.182 ?0.098 ?0.097
0.000 0.001 0.836 0.316 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
OCIGL (4) 0.044 ?0.079 0.067 0.067 ?0.003 ?0.001 ?0.020 ?0.016 ?0.005 0.038 0.004 0.016 ?0.003
0.031 0.000 0.001 0.001 0.894 0.968 0.317 0.445 0.815 0.064 0.846 0.445 0.900
DISCGL (5) 0.017 ?0.096 0.039 0.073 ?0.001 0.033 0.018 0.013 0.031 0.037 0.018 0.000 0.003
0.416 0.000 0.054 0.000 0.963 0.104 0.383 0.512 0.128 0.066 0.382 0.984 0.881
MRM (6) 0.009 0.024 0.303 0.002 0.005 0.173 0.074 0.094 0.076 ?0.011 0.051 ?0.010 ?0.045
0.648 0.239 0.000 0.941 0.793 0.000 0.000 0.000 0.000 0.603 0.012 0.611 0.028
CRM (7) 0.066 ?0.029 0.254 0.003 0.038 0.209 0.229 0.330 0.240 0.144 0.030 ?0.009 0.085
0.001 0.161 0.000 0.888 0.062 0.000 0.000 0.000 0.000 0.000 0.147 0.651 0.000
CRO (8) 0.013 ?0.021 0.227 ?0.009 0.020 0.131 0.261 0.364 0.237 0.141 0.043 0.011 0.004
0.510 0.302 0.000 0.666 0.320 0.000 0.000 0.000 0.000 0.000 0.034 0.592 0.839
SIZE (9) ?0.004 0.011 0.459 0.007 0.002 0.231 0.409 0.452 0.499 0.184 0.201 ?0.128 ?0.102
0.826 0.597 0.000 0.727 0.921 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
TRADD (10) 0.020 0.012 0.388 0.001 0.021 0.136 0.263 0.237 0.593 0.177 0.199 ?0.095 ?0.137
0.315 0.566 0.000 0.953 0.304 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
INTSEN (11) 0.170 ?0.360 0.176 0.084 0.066 ?0.044 0.122 0.123 0.140 0.134 0.156 0.171 ?0.114
0.000 0.000 0.000 0.000 0.001 0.031 0.000 0.000 0.000 0.000 0.000 0.000 0.000
COMMLOAN (12) 0.030 0.072 0.028 ?0.007 ?0.002 0.005 0.009 0.025 0.118 0.183 0.064 ?0.053 ?0.225
0.141 0.000 0.176 0.744 0.905 0.811 0.649 0.215 0.000 0.000 0.002 0.009 0.000
FVISTD (13) ?0.150 ?0.352 ?0.057 0.026 ?0.034 ?0.067 ?0.011 0.021 ?0.070 ?0.056 0.209 ?0.098 ?0.209
0.000 0.000 0.005 0.200 0.099 0.001 0.596 0.292 0.001 0.006 0.000 0.000 0.000
TIER1 (14) 0.192 0.252 ?0.047 ?0.020 0.022 ?0.010 0.061 ?0.011 ?0.134 ?0.142 ?0.148 ?0.188 ?0.219
0.000 0.000 0.020 0.315 0.280 0.616 0.003 0.577 0.000 0.000 0.000 0.000 0.000
Notes: The table reports the Pearson (below the diagonal) and Spearman (above the diagonal) correlations of the primary model variables for 2413 bank-year obser-
vations for 238 banks from 2002 to 2013. Variables are de?ned in the Appendix c?De?nitions of variables and other acronymsd?. p values are in italics below the
corresponding correlation.
Table 3
First-stage regressions of risk modeling measures on risk management variables.
XRM MRM CRM
Variables (1) (2)
CRO 0.022
??
0.022
?
(0.020) (0.066)
SIZE 0.021
???
0.046
???
(0.000) (0.000)
TRADD 0.004 0.012
(0.652) (0.270)
INTSEN ?0.004
??
?0.003
(0.021) (0.196)
COMMLOAN ?0.054 ?0.016
(0.182) (0.757)
FVISTD ?0.017
?
0.022
?
(0.053) (0.059)
TIER1 0.001 ?0.001
(0.347) (0.696)
Constant ?0.012 ?0.357
???
(0.781) (0.000)
Year effects Included Included
Observations 2413 2413
R-squared 0.068 0.273
Notes: This table reports the estimation of Eq. (1), which regresses of the risk mod-
eling measures (MRM and CRM, collectively denoted XRM) on risk management
variables and year effects. Variables are de?ned in the Appendix c?De?nitions of
variables and other acronymsd?. All variables are winsorized at the 0.5% and 99.5%
levels of their distribution. Standard errors are calculated clustering observations by
bank. P-values are in parentheses below the corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
FVGL components. To link to prior research and provide a bench-
mark for the main results, column (1) of the panel reports the
results for the model including only NIBNIGL, the three FVGL
components, and the ?xed year effects (untabulated). Column (2)
[(4)] presents the results for Eq. (2A) using the unadjusted mar-
ket [credit] risk modeling measure MRM [CRM]. Column (3) [(5)]
presents the results for Eq. (2B) using the market [credit] risk mod-
eling activity and disclosure measures
MRM and MRM ?
MRM
[
CRM and CRM ?
CRM].
In the benchmark model reported in column (1), the model ?t
is good with an R
2
of 41.8%. Such a high R
2
for a returns model
obtains in large part from the substantial variation in bank indus-
try performance across the sample years discussed earlier, which
is captured by the ?xed year effects. The coe?cient on NIBNIGL is
signi?cantly positive at the 1% level but well below one at 0.249,
suggesting the presence of considerable noise in this variable. The
coe?cient on NIGL is signi?cantly positive at the 5% level and well
above one at 3.297, indicating that it includes a permanent com-
ponent. This re?ects the inclusion of some fee revenue in trading
revenue on the Y-9C ?lings and the persistence of FVGL from most
dealer and proprietary trading activities due to the incorporation
of trading spreads (i.e., day-one pro?ts). The coe?cient on OCIGL
is signi?cantly positive at the 5% level and somewhat below one
at 0.742, consistent with it primarily capturing transitory income
items. The coe?cient on DISCGL is insigni?cant and close to zero,
consistent with prior research documenting little or no returns rel-
evance for FVGL on less liquid ?nancial instruments discussed in
section ‘Financial report information about estimated fair values
of ?nancial instruments, literature review, and hypothesis devel-
opment’.
The coe?cients on the test variables reported in columns (2)–
(5) of the panel are with one explainable exception consistent with
our main hypotheses H1–H3. The coe?cients on the interactions
of MRM and
MRM with OCIGL in columns (2) and (3), respec-
tively, are both signi?cantly positive at the 10% level, consistent
with Hypothesis H2. The coe?cients on the interactions of MRM
and
MRM with DISCGL in these columns are both insigni?cant,
however, inconsistent with Hypothesis H1. These insigni?cant coef-
?cients can be explained by low power resulting from much of the
90 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
overall sample period exhibiting stable interest rates and the large
set of ?nancial instruments (e.g., loans and deposits) re?ected in
DISCGL likely exhibiting considerable cross-hedging of interest rate
risk. This explanation is supported by results reported in Table 5
discussed below, which show that the coe?cients on these interac-
tions are signi?cantly positive during periods of high interest rate
volatility.
The coe?cients on the interactions of CRM and
CRM with DIS-
CGL in columns (4) and (5), respectively, are both signi?cantly pos-
itive at the 5% level, consistent with Hypotheses H1 and H3. The
coe?cients on the interactions of CRM and
CRM with OCIGL are
insigni?cant, as suggested by Hypothesis H3. These results sug-
gest that DISCGL re?ects banks’ primary assumption of credit risk
through funded loans and unfunded loan commitments.
Table 4, Panel A generally reports insigni?cant coe?cients on
the interactions of the risk modeling measures with the FVGL vari-
ables about which no hypotheses are made. The coe?cients on the
interactions of MRM,
MRM, CRM, and
CRM with NIGL are all in-
signi?cant. The coe?cients on the interactions of the risk modeling
disclosure measures MRM ?
MRM and CRM ?
CRM with all three
types of FVGL are all insigni?cant. Hence, the act of disclosing risk
Table 4
Regression of returns on FVGL, risk modeling measures, and interactions.
XRM MRM MRM CRM CRM
Variables (1) (2) (3) (4) (5)
Panel A: Regression of returns on FVGL, risk modeling measures, and interactions
NIBNIGL 0.249
???
0.255
???
?0.206 0.235
???
0.583
(0.000) (0.000) (0.456) (0.000) (0.229)
NIGL 3.297
??
10.695
???
32.875
???
7.692
??
27.269
??
(0.024) (0.001) (0.006) (0.043) (0.022)
OCIGL 0.742
??
1.649
???
?0.311 0.501 ?1.733
(0.021) (0.002) (0.884) (0.170) (0.427)
DISCGL 0.043 0.075 ?0.327 ?0.025 ?1.165
??
(0.274) (0.453) (0.404) (0.654) (0.016)
NIBNIGL ? XRM ?0.008 0.095
(0.968) (0.680)
NIGL ? XRM ?11.562 ?8.176
(0.103) (0.146)
OCIGL ? XRM 3.194
?
1.044
(0.059) (0.305)
DISCGL ? XRM ?0.125 0.308??
(0.678) (0.027)
XRM ?0.026 0.011
(0.431) (0.657)
NIBNIGL ?
XRM 1.671
?
?0.503
(0.081) (0.492)
NIGL ?
XRM ?64.909 ?25.949
(0.122) (0.143)
OCIGL ?
XRM 3.850
?
3.686
(0.064) (0.265)
DISCGL ?
XRM 1.258 1.791
??
(0.342) (0.014)
XRM ?0.697
???
?0.253
???
(0.000) (0.003)
NIBNIGL ? (XRM ?
XRM) ?0.139 ?0.018
(0.629) (0.950)
NIGL ? (XRM ?
XRM) ?1.715 5.709
(0.731) (0.341)
OCIGL ? (XRM ?
XRM) ?3.647 1.569
(0.125) (0.190)
DISCGL ? (XRM ?
XRM) ?0.650 0.201
(0.147) (0.421)
(XRM ?
XRM) ?0.012 0.036
(0.735) (0.218)
Constant 0.297
???
0.300
???
0.522
???
0.294
???
0.467
???
(0.000) (0.000) (0.000) (0.000) (0.000)
Year effects Included Included Included Included Included
Observations 2413 2413 2413 2413 2413
R-squared 0.418 0.424 0.437 0.421 0.431
(Continued on next column)
Table 4 (continued)
Variables (1) (2)
Panel B: Expanded regression of returns on FVGL, risk modeling measures, and
interactions including M/ABSDUM interactions
OCIGL 2.358
???
0.215
(0.002) (0.933)
OCIGL ? M/ABSDUM ?1.678
?
?2.073
(0.084) (0.588)
OCIGL ? MRM 6.221
(0.113)
OCIGL ? MRM ? M/ABSDUM 6.927
??
(0.031)
OCIGL ?
MRM 1.842
(0.814)
OCIGL ?
MRM ? M/ABSDUM 7.220
?
(0.054)
OCIGL ? (MRM ?
MRM) ? M/ABSDUM ?5.873
(0.113)
OCIGL ? (MRM ?
MRM) ? M/ABSDUM 3.924
(0.260)
R-squared 0.427 0.439
Notes: Column 1 of Panel A reports the estimation of a restricted version of Eq. (2A)
that regresses returns for the 12 months ending 4 months after the ?scal year end
(R) on net income before FVGL recorded in net income (NIBNIGL) and the three
types of FVGL (NIGL, OCIGL, and DISCGL) and includes ?xed year effects. Column 2
(4) reports the estimation of Eq. (2A), which includes one of the unadjusted MRM
and CRM measures described in the Appendix c?Risk modeling measures and chief
risk o?cer indicatord? and interactions involving that measure. Column 3 (5) re-
ports the estimation of Eq. (2B), which includes one of the MRM and CRM risk
modeling activity measures (
MRM or
CRM, collectively denoted
XRM), the corre-
sponding risk modeling disclosure measure (MRM ?
MRM or CRM ?
CRM), and in-
teractions involving these measures. The MRM and CRM activity and disclosure
measures are the predicted values and residuals, respectively, from the estimation
of the corresponding ?rst-stage model reported in Table 3. Panel B reports the re-
sults of expanded versions of the regression reported in columns (2) and (3) of
Panel A that include interactions of M/ABSDUM, an indicator for an above-median
proportion of available-for-sale securities that are mortgage-backed or asset-backed
securities, with the variables involving OCIGL. Variables are de?ned in the Appendix
c?De?nitions of variables and other acronymsd?. All variables are winsorized at the
0.5% and 99.5% levels of their distribution. Standard errors are calculated clustering
observations by bank. P-values are in parentheses under the corresponding coe?-
cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
modeling activities does not appear to affect the returns-relevance
of FVGL.
In summary, the ?ndings reported in Table 4, Panel A indi-
cate that MRM activities enhance the returns-relevance of OCIGL,
which primarily re?ect interest rate risk, consistent with Hypothe-
sis H2 but not Hypothesis H1. CRM activities enhance the returns-
relevance of DISCGL, which re?ect banks’ primary assumption of
credit risk through funded loans and unfunded loan commitments,
consistent with Hypotheses H1 and H3.
Regression results with interactions for potentially illiquid and
di?cult-to-value available-for-sale securities
Hypothesis H2 is based on the assumption that banks do
not determine OCIGL directly from market prices, obviating their
need to conduct MRM to make this determination. As discussed
in section ‘Financial report information about estimated fair val-
ues of ?nancial instruments, literature review, and hypothesis
development’, mortgage-backed and asset-backed securities typi-
cally are illiquid and include embedded interest-rate options and
other interest-rate-related structuring that require MRM to accu-
rately estimate the fair value of the securities. Re?ecting this fact,
the hypothesis states it should hold more strongly for mortgage-
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 91
backed and asset-backed securities than for banks’ other primary
types available-for-sale securities, such as U.S. Treasuries. To test
this part of the hypothesis, we estimate expanded versions of
Eqs. (2A) and (2B) that include interactions of an indicator vari-
able for bank-year observations with above median percentage
of mortgage-backed and asset-backed securities in their available-
for-sale securities portfolios in that year, denoted M/ABSDUM,
with all variables involving OCIGL. We expect the coe?cient on
the interaction of OCIGL with MRM to be more positive and
signi?cant for observations for which M/ABSDUM takes a value
of one.
To conserve space, Table 4, Panel B reports the estimated co-
e?cients only for the critical variables in these expanded models
and also the R
2
s; all omitted coe?cients are essentially identi-
cal to those reported in Panel A. Column (1) [(2)] of the panel re-
ports the results for the expanded versions of Eq. (2A) [(2B)]. In
column (1), the coe?cient on the interaction of OCIGL and MRM is
insigni?cantly positive, consistent with MRM not enhancing the es-
timation of the fair values of available-for-sale securities other than
mortgage-backed and asset-backed securities. The coe?cient on
the interaction of OCIGL and M/ABSDUM is signi?cantly negative
at the 10% level, consistent with fair valuation estimation posing
greater di?culties for mortgage-backed and asset-backed securities
than for other available-for-sale securities. Consistent with Hypoth-
esis H2 that MRM enhances the fair value estimation of mortgage-
backed and asset-backed securities, the coe?cient on the interac-
tion of OCIGL, MRM, and M/ABSDUM is signi?cantly positive at the
5% level.
Column (2) yields similar inferences. In particular, the co-
e?cient on the interaction of OCIGL,
MRM, and M/ABSDUM
is signi?cantly positive at the 10% level and the coe?cient
on the interaction of OCIGL, MRM ?
MRM, and M/ABSDUM is
insigni?cant.
Regression results with interactions for macroeconomic conditions
Hypothesis H4 predicts that MRM activities are more impor-
tant in periods with more volatile interest rates. To test this hy-
pothesis, we expand the models reported in columns (2) and (3)
of Table 4, Panel A to include an indicator variable for years with
above-median interest rate volatility that is interacted with the in-
teractions of MRM,
MRM, and MRM ?
MRM with NIBNIGL and the
three types of FVGL. Similarly, Hypothesis H5 predicts that CRM
activities are more important in periods with higher and more un-
certain credit losses. To test this hypothesis, we expand the mod-
els reported in columns (4) and (5) of Table 4, Panel A to include
an indicator variable for the ?nancial crisis years that is interacted
with the interactions of CRM,
CRM, and CRM ?
CRM with NIBNIGL
and the three types of FVGL.
We measure interest rate volatility during a year as the stan-
dard deviation of the daily yield on seven-year U.S. Treasuries. UN-
STABLE takes a value of one for a year with above-median interest
rate volatility and zero otherwise. CRISIS takes a value of one for
the years 2007–2009, and zero otherwise, because the subprime
crisis began in February 2007 (Ryan, 2008) and banks’ loan loss
provisions peaked in the third quarter of 2009.
Table 5 presents the results of the estimations of the four ex-
panded models; columns (1) and (2) reporting the results for the
models including the indicator variable UNSTABLE and columns (3)
and (4) reporting the results for the models including the indicator
CRISIS. In the table, these indicator variables are generically de-
noted DUMMY and the column headings specify which indicator
DUMMY represents. To conserve space, we discuss only the coef-
?cients on the interactive variables involving these indicators. We
omit ?xed year effects in these models because both indicator vari-
ables partition on time.
Column (1) of Table 5 presents the results for the model in
which UNSTABLE is interacted with the interactive variables in-
volving MRM. Consistent with Hypothesis H4, the coe?cients on
OCIGL ? MRM ? UNSTABLE and DISCGL ? MRM ? UNSTABLE are sig-
ni?cantly positive at the 5% and 1% levels, respectively. The latter
result is related to our explanation for the unexpectedly insignif-
icant coe?cients on the interactions of DISCGL with MRM and
MRM in Table 4, Panel A discussed above. In contrast, the coe?-
cients on all other interactions involving UNSTABLE are insigni?-
cant.
Column (2) of the table presents similar results for the model
where UNSTABLE is interacted with the interactive variables in-
volving
MRM and MRM-
MRM. Again consistent with Hypoth-
esis H4, the coe?cients on OCIGL ?
MRM ? UNSTABLE and
DISCGL ?
MRM ? UNSTABLE are both signi?cantly positive at the
1% levels, while the coe?cients on OCIGL ? (MRM ?
MRM) ? UN-
STABLE and DISCGL ? (MRM ?
MRM) ? UNSTABLE are both in-
signi?cant. In contrast, the coe?cients on all other interactions in-
volving UNSTABLE are insigni?cant except for the signi?cantly neg-
ative coe?cient on NIBNIGL ?
MRM ? UNSTABLE.
Column (3) presents the results for the model where CRISIS
is interacted with the interactive variables involving CRM. Consis-
tent with Hypothesis H5, the coe?cient on DISCGL ? CRM ? CRI-
SIS is signi?cantly positive at the 10% level. In addition, the coef-
?cients on NIGL ? CRM ? CRISIS and OCIGL ? CRM ? CRISIS are sig-
ni?cantly positive, likely re?ecting the pervasive effect of credit
risk on banks’ various sources of net income during the ?nancial
crisis.
Column (4) presents similar results for the model where CRI-
SIS is interacted with the interactive variables involving
CRM
and CRM ?
CRM. Consistent with Hypothesis H5, the coe?cient
on DISCGL ?
CRM ? CRISIS is signi?cantly positive at the 10%
level, while the coe?cients on the interactions of CRISIS with
all variables involving CRM ?
CRM are insigni?cant. In contrast,
the coe?cients on all other interactions involving UNSTABLE are
insigni?cant except for the signi?cantly negative coe?cient on
NIBNIGL ?
CRM ? UNSTABLE and signi?cantly positive coe?cient
on OCIGL ?
CRM ? UNSTABLE.
In summary, the unadjusted risk modeling measures and risk
modeling activity measures generally are signi?cant for economic
conditions for which they are expected to be important and usually
are insigni?cant otherwise. The risk modeling disclosure measures
generally are insigni?cant regardless of the economic conditions.
Changes speci?cation
To help ensure that our primary second-stage results reported
in Table 4, Panel A are robust to omitted variables, and also as an
alternative to the inclusion of year ?xed effects as a means to cap-
ture the trend in CRM over time reported in Table 1, Panel C, we
estimate changes models. These models are identical to the mod-
els reported in columns (2)–(5) of Table 4, Panel A except that
MRM,
MRM, MRM ?
MRM, CRM,
CRM, and CRM ?
CRM are re-
placed with the changes in the respective variables and the year
?xed effects are dropped.
Table 6 reports the estimation of these changes models, with
columns (1) and (2) corresponding to columns (2) and (3), respec-
tively, of Table 4, Panel A. The results in column (1) of Table 6
are similar with those reported in column (2) of Table 4, Panel A.
The coe?cient on the interaction of MRM with OCIGL remains
signi?cantly positive at the 10% level, consistent with Hypothesis
H2. The results in column (2) of Table 6 are similar to those re-
92 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Table 5
Regression of returns on risk modeling measures, FVGL, and interactions impact of
macroeconomic conditions.
XRM MRM CRM
DUMMY Unstable Crisis
Variables (1) (2) (3) (4)
NIBNIGL 0.179
??
?0.808
??
0.099 ?0.686
(0.015) (0.016) (0.140) (0.195)
NIGL 17.018
???
43.237
???
6.563
?
25.126
??
(0.000) (0.001) (0.073) (0.050)
OCIGL 0.995 ?1.320 0.526 0.523
(0.126) (0.627) (0.114) (0.838)
DISCGL 0.133 ?0.321 ?0.086 ?1.166
??
(0.269) (0.560) (0.187) (0.039)
NIBNIGL ? XRM 0.965
?
0.983
???
(0.067) (0.000)
NIGL ? XRM ?16.638 ?10.543
(0.130) (0.158)
OCIGL ? XRM ?6.765 ?2.383
(0.117) (0.234)
DISCGL ? XRM 1.187
?
0.158
(0.094) (0.522)
XRM ?0.098
?
?0.010
(0.077) (0.690)
NIBNIGL ? XRM ? DUMMY ?0.879 ?1.454
(0.199) (0.109)
NIGL ? XRM ? DUMMY ?2.251 26.385
???
(0.492) (0.001)
OCIGL ? XRM ? DUMMY 5.582
??
9.189
???
(0.025) (0.003)
DISCGL ? XRM ? DUMMY 1.080
???
0.850
?
(0.003) (0.064)
XRM ? DUMMY 0.079 0.011
(0.361) (0.889)
NIBNIGL ?
XRM 5.207
???
1.734
??
(0.000) (0.034)
NIGL ?
XRM ?8.767 ?2.431
(0.108) (0.079)
OCIGL ?
XRM ?2.030 ?1.239
(0.816) (0.752)
DISCGL ?
XRM 0.215 1.670
?
(0.907) (0.050)
XRM ?1.187
???
?0.477
???
(0.000) (0.000)
NIBNIGL ?
XRM ? DUMMY ?1.587
??
?0.689
???
(0.040) (0.000)
NIGL ?
XRM ? DUMMY 9.819 12.825
(0.352) (0.108)
OCIGL ?
XRM ? DUMMY 8.692
???
4.100
???
(0.002) (0.000)
DISCGL ?
XRM ? DUMMY 1.238
???
0.271
?
(0.005) (0.074)
XRM ? DUMMY 0.584 0.323
(0.109) (0.162)
NIBNIGL ? (XRM ?
XRM) ?0.855 0.308
(0.590) (0.298)
NIGL ? (XRM ?
XRM) 8.155 0.823
(0.251) (0.888)
OCIGL ? (XRM ?
XRM) 7.239 ?1.184
(0.268) (0.340)
DISCGL ? (XRM ?
XRM) ?0.349 0.395
(0.559) (0.191)
(XRM ?
XRM) 0.047 0.033
(0.717) (0.324)
NIBNIGL ? (XRM ?
XRM) ? DUMMY 0.734 ?0.273
(0.651) (0.634)
NIGL ? (XRM ?
XRM) ? DUMMY ?31.947 8.593
(0.128) (0.833)
OCIGL ? (XRM ?
XRM) ? DUMMY ?8.172 4.763
(0.229) (0.266)
DISCGL ? (XRM ?
XRM) ? DUMMY ?0.741 ?0.661
(0.401) (0.259)
(continued on next page)
Table 5 (continued)
XRM MRM CRM
DUMMY Unstable Crisis
Variables (1) (2) (3) (4)
(XRM ?
XRM) ? DUMMY ?0.042 ?0.102
(0.782) (0.254)
DUMMY ?0.022 ?0.158 ?0.244
???
?0.453
???
(0.402) (0.176) (0.000) (0.003)
Constant 0.067
???
0.379
???
0.110
???
0.413
???
(0.000) (0.000) (0.000) (0.000)
Year effects Excluded Excluded Excluded Excluded
Observations 2413 2413 2413 2413
R-squared 0.086 0.117 0.205 0.232
Notes: Columns (1) and (2) [(3) and (4)] of this table reports the results of expand-
ing the regression reported in the columns (2) and (3) [(4) and (5)] of Table 4,
Panel A to include interactions of a dummy variable that takes a value of 1 for
years with above-median unstable interest rates [the ?nancial crisis years 2007–
2009] with the interactions of NIBNIGL and the three types of FVGL with MRM,
MRM and (MRM-
MRM), [CRM,
CRM and (CRM-
CRM)] respectively. Variables are de-
?ned in the Appendix c?De?nitions of variables and other acronymsd?. All variables
are winsorized at the 0.5% and 99.5% levels of their distribution. Standard errors are
calculating clustering observations by bank. P-values are in parentheses under the
corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
ported in column (3) of Table 4, Panel A. The coe?cient on the
interaction of
MRM with OCIGL becomes more signi?cantly pos-
itive at the 5% level, consistent with Hypothesis H2, and the coe?-
cient on the interaction of (MRM-
MRM) with OCIGL remains in-
signi?cant. Columns (3) and (4) of Table 6 correspond to columns
(4) and (5), respectively, of Table 4, Panel A. The results in col-
umn (3) of Table 6 are similar with those reported in column (4)
of Table 4, Panel A. The coe?cient on the interaction of CRM
with DISCGL remains signi?cantly positive at the 5% level, con-
sistent with Hypotheses H1 and H3. The results in column (4) of
Table 6 are similar with those reported in column (5) of Table 4,
Panel A. The coe?cient on the interaction of
CRM with DISCGL
remains signi?cantly positive, albeit at the lower 10% level, consis-
tent with Hypotheses H1 and H3. The coe?cient on the interaction
of (CRM ?
CRM) with DISCGL remains insigni?cant.
Summary and conclusion
We provide evidence that banks’ market and credit risk model-
ing (MRM and CRM, respectively) enhance the returns-relevance of
banks’ estimated unrealized fair value gains and losses (FVGL) for
?nancial instruments. We employ both a one-stage approach that
uses unadjusted MRM and CRM measures developed from banks’
?nancial report disclosures and a two-stage approach that parti-
tions these unadjusted measures into measures of banks’ risk mod-
eling activities and disclosures of those activities. We predict and
?nd that banks’ MRM and CRM activities, but not disclosures of
these activities, enhance the returns-relevance of their FVGL, more
so for less liquid instruments for which FVGL typically are calcula-
ble from disclosures rather than recorded in net income or other
comprehensive income. We further predict and ?nd that banks’
MRM activities enhance the returns-relevance of FVGL recorded in
other comprehensive income, which primarily result from interest
rate movements, especially when these FVGL are for potentially
illiquid and di?cult-to-value mortgage-backed and asset-backed
available-for-sale securities. We predict and ?nd that banks’ CRM
activities enhance the returns-relevance of their disclosed FVGL,
because banks assume credit risk primarily through their funded
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 93
Table 6
Regression of Returns on Risk Modeling Measures, FVGL, and Interactions
Changes Speci?cation.
XRM MRM CRM
Variables (1) (2) (3) (4)
NIBNIGL 0.148
???
0.189
???
0.139
???
0.174
???
(0.000) (0.000) (0.000) (0.000)
NIGL 7.636
???
6.936
???
7.405
???
7.106
???
(0.000) (0.001) (0.001) (0.001)
OCIGL 0.099 0.073 0.032 0.086
(0.623) (0.713) (0.873) (0.662)
DISCGL 0.033 0.028 0.033 0.024
(0.291) (0.334) (0.286) (0.416)
NIBNIGL ? XRM ?2.810 1.357
???
(0.628) (0.006)
NIGL ? XRM ?168.167 1.402
(0.291) (0.935)
OCIGL ? XRM 10.703
?
4.315
?
(0.052) (0.074)
DISCGL ? XRM ?1.035 0.982
??
(0.666) (0.049)
XRM 0.297 ?0.102
?
(0.410) (0.086)
NIBNIGL ?
XRM 3.748 4.760
??
(0.288) (0.027)
NIGL ?
XRM ?21.199 ?9.753
(0.243) (0.428)
OCIGL ?
XRM 5.515
??
?2.080
(0.033) (0.930)
DISCGL ?
XRM ?0.140 1.136
?
(0.975) (0.060)
XRM ?1.789
??
?1.090
??
(0.033) (0.048)
NIBNIGL ? (XRM ?
XRM) 1.056 0.003
(0.721) (0.995)
NIGL ? (XRM ?
XRM) 37.275 ?25.090
(0.498) (0.141)
OCIGL ? (XRM ?
XRM) 1.693 1.363
(0.879) (0.555)
DISCGL ? (XRM ?
XRM) ?0.748 0.295
(0.630) (0.616)
(XRM ?
XRM) ?0.129 0.003
(0.588) (0.957)
Constant 0.056
???
0.058
???
0.057
???
0.057
???
(0.000) (0.000) (0.000) (0.000)
Observations 2214 2214 2214 2214
R-squared 0.041 0.050 0.050 0.051
Notes: Columns (1) and (2) [(3) and (4)] of this table report the results
of a modi?cation of the regression reported in columns (2) and (3) [(4)
and (5)] of Table 4, Panel A that includes the changes in MRM,
MRM and
(MRM ?
MRM) [CRM,
CRM and (CRM ?
CRM)] rather than the levels of these
variables. Variables are de?ned in the Appendix c?De?nitions of variables and
other acronymsd?. All variables are winsorized at the 0.5% and 99.5% levels
of their distribution. Standard errors are calculated clustering observations by
bank. P-values are in parentheses under the corresponding coe?cients.
?
Denotes signi?cance at the 10% level.
??
Denotes signi?cance at the 5% level.
???
Denotes signi?cance at the 1% level.
loans and unfunded loan commitments for which the fair values
are disclosed rather than recognized. Finally, we show that the im-
pact of MRM (CRM) activities is stronger during periods of higher
interest rate volatility (the ?nancial crisis).
Our evidence is subject to the following caveats. First, we show
that banks’ MRM and CRM are associated with increased returns-
relevance for speci?c types of FVGL, not that they cause this in-
creased returns relevance. Correlated bank attributes, such as tech-
nical sophistication, may be the actual causes. Second, we do not
control for alternative ?nancial report disclosures related to the
quality of their FVGL estimates, because these alternatives are sig-
ni?cantly limited in terms of one or more of their information
quality, coverage across types of ?nancial instruments, or coverage
across our sample period. Finally, our results based on a U.S. bank
sample may not generalize to other industries or countries.
Despite these caveats, we believe our results have relevance for
accounting standard setters, auditors, and other parties involved
in requiring or evaluating estimates of the fair values of ?nan-
cial instruments and related disclosures in banks’ ?nancial reports.
Re?ecting strong political pressure arose during the ?nancial cri-
sis, a number of the FASB’s and IASB’s recent proposals indicate
retrenchment in what they deem feasible or desirable regarding
the extent of fair value accounting for ?nancial instruments.
18
Our
?ndings suggest that the returns-relevance of FVGL for ?nancial in-
struments is enhanced by banks’ risk modeling.
Risk modeling measures and chief risk o?cer indicator
Risk modeling measures
Our risk modeling measures are simple averages of indicators
for whether banks engage in each of the following risk model-
ing activities (including related risk management processes) men-
tioned a survey developed by the Basel Committee on Banking Su-
pervision (2001):
Market risk modeling:
(1) INTGAP – Does the bank use maturity or repricing analysis?
(2) IRSIMUL – Does the bank use interest rate simulation?
(3) VAR – Does the bank use Value at Risk for market risk?
(4) MRSTRESS – Does the bank stress test its market risk models?
(5) MRBACKTEST – Does the bank back test its market risk models?
Credit risk modeling:
(6) MODEL – Does the bank use statistical credit risk modeling?
(7) SCORING – Does the bank use credit scoring models?
(8) RR – Does the bank have an internal credit risk rating process?
(9) CRSTRESS – Does the bank use stress testing its credit risk
models?
Our market risk modeling measure, MRM, is the simple average
of the ?rst ?ve indicators. Our credit risk modeling measure, CRM,
is the simple average of the next four indicators. As averages of
indicator variables, both measures take values from zero to one.
We identify whether banks engage in these activities from
hand-collected disclosures in their Form 10-K ?ling for the years
2001–2013. We ?rst used search engines to locate words or
phrases such as c?risk management,d? c?risk model,d? c?scoring,d?
and c?risk rating,d? and read each search result to determine
whether one (or more) of the nine risk modeling activities is indi-
cated. Each item is scored one if disclosed and zero otherwise. The
following table provides representative examples of disclosures for
each of the nine activities.
18
For example, see the FASB’s April 12, 2013 proposed Accounting Standards Up-
date, Financial Instruments—Overall (Subtopic 825-10): Recognition and Measure-
ment of Financial Assets and Financial Liabilities, which would allow ?rms consid-
erable ability to classify ?nancial instruments to avoid fair valuation.
94 G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95
Market risk modeling
INTGAP Central Paci?c Financial Corp Form 10-K ?ling 2009
Interest rate risk can be analyzed by monitoring an institution’s interest rate sensitivity gap and changes in the gap over time. An asset or
liability is said to be interest rate sensitive within a speci?c time period if it will mature or reprice within that time period. The interest
rate sensitivity gap is de?ned as the difference between the amount of interest-earning assets and the amount of interest-bearing
liabilities maturing or repricing within a speci?ed time period. A gap is considered positive when the amount of interest rate sensitive
assets exceeds the amount of interest rate sensitive liabilities. A gap is considered negative when the amount of interest rate sensitive
liabilities exceeds the amount of interest rate sensitive assets. During a period of rising interest rates, the earnings of an institution with a
positive gap theoretically may be positively affected due to its interest-earning assets repricing to a greater extent than its interest-bearing
liabilities. An adverse impact would be expected for an institution with a negative gap
IRSIMUL Zions Bancorporation Form 10-K ?ling 2003
Interest rate risk is the most signi?cant market risk to which the Company is regularly exposed. We monitor this risk through the use of
two complementary measurement methods: duration of equity and income simulation. In the duration of equity method, we measure the
changes in the market values of equity in response to changes in interest rates. In the income simulation method, we analyze the changes
in income in response to changes in interest rates. The tool that we use to apply both of these methods is an asset-liability management
system marketed by a third party company, QRM. In general, our goal in managing interest is to have net interest income increase in a
rising interest rate environment, which tends to mitigate any declines in the market value of equity due to higher discount rates. This
approach is based on our belief that in a rising interest rate environment, the market cost of equity, or implied rate at which future
earnings are discounted, would also tend to rise. We refer to this position as c?asset sensitived?
VAR Key Corp Form 10-K ?ling 2004
Management uses a value at risk (c?VARd?) simulation model to measure the potential adverse effect of changes in interest rates, foreign
exchange rates, equity prices and credit spreads on the fair value of Key’s trading portfolio. Using historical information, the model
estimates the maximum potential one-day loss with 95% probability by comparing the relative change in rates to the current day’s rates
and prices
MRSTRESS US Bancorp Form 10-K ?ling 2006
The Company mitigates these uncertainties through regular monitoring of trading activities by management and other risk management
practices, including stop-loss and position limits related to its trading activities. Stress-test models are used to provide management with
perspectives on market events that VaR models do not capture
MRBACKTEST FleetBoston Financial Corp Form 10-K ?ling 2003
Our independent Market Risk Management function routinely validates our measurement framework by conducting backtests, which
compare the actual daily trading-related results against the estimated VAR with a one-day holding period. This measure contrasts with
our standard VAR estimate by excluding the impact of reduced trading liquidity
Credit risk modeling
MODEL Bank of America Corporation Form 10-K ?ling 2009
We use proprietary models to measure the capital requirements for credit, country, market, operational and strategic risks
Statistical models are built using detailed behavioral information from external sources such as credit bureaus and/or internal historical
experience. These models are a component of our consumer credit risk management process and are used, in part, to help determine both
new and existing credit decisions, portfolio management strategies including authorizations and line management, collection practices and
strategies, determination of the allowance for loan and lease losses, and economic capital allocations for credit risk
SCORING BOK Financial Corp Form 10-K ?ling 2009
Credit scoring is assessed based on signi?cant credit characteristics including credit history, residential and employment stability
RR Santander Bancorp Form 10-K ?ling 2007
The corporation has also established an internal risk rating system and internal classi?cations which serve as timely identi?cation of the
loan quality issues affecting the loan portfolio.
CRSTRESS Bank of Hawaii Form 10-K ?ling 2009
In addition, the Company uses a variety of other tools to estimate probable credit losses including, but not limited to, a rolling quarterly
forecast of asset quality metrics; stress testing; and performance indicators based on the Company’s own experience, peers, or other
industry sources
Chief risk o?cer indicator
The chief risk o?cer indicator, CRO, takes the value one if the
bank discloses in its Form 10-K ?ling that it employs a chief risk
o?cer and zero otherwise. We ?rst search ?lings to determine
whether the words c?chiefd? and c?riskd? appear within ?ve words
of each other. We then search the ?lings that meet the ?rst ?lter
to determine whether the words c?riskd? and either c?o?cerd? or
c?directord? appear within ?ve words of each other. We read the
relevant passages to con?rm the existence of a chief risk o?cer for
the bank-years that meet both ?lters.
De?nitions of variables and other acronyms
Variable Description Source
ASC Accounting Standards Codi?cation Acronym
ASU Accounting Standards Update Acronym
COMMLOAN Commercial and industrial (including agricultural) loans divided by lagged average total assets Y-9C
CRISIS Indicator equal to one for the years 2007 to 2009 and zero otherwise
CRM Credit risk modeling measure (see the Appendix c?Risk modeling measures and chief risk o?cer indicatord? for
details)
10-K
CRO Indicator equal to one if the bank employs a chief risk o?cer and zero otherwise (see the Appendix c?Risk modeling
measures and chief risk o?cer indicatord? for details)
10-K
DISCGL Fair value gains and losses calculated from FAS 107 disclosures divided by beginning-of-year market value of equity SNL
(continued on next page)
G. Bhat, S.G. Ryan/ Accounting, Organizations and Society 46 (2015) 81–95 95
Variable Description Source
FAS Statement of Financial Accounting Standards Acronym
FSP FASB Staff Position Acronym
FVGL Fair value gains and losses Acronym
FVISTD Standard deviation of de?ated fair value income—the sum of NIBNIGL, NIGL, OCIGL and DISCGL—over the last ten
years, requiring at least three annual observations
SNL
INTSEN Interest-sensitive assets minus interest-sensitive liabilities repricing in one year divided by beginning-of-year market
value of equity
Y-9C
MRM Market risk modeling measure (see the Appendix c?Risk modeling measures and chief risk o?cer indicatord? for
details)
10-K
M/ABSDUM Indicator variable equal to one if mortgage-backed and asset-backed available-for-sale securities divided by total
available-for-sale securities is above the sample median for the year and zero otherwise
Y-9C
NIBNIGL Net income before fair value gains and losses recorded in net income divided by beginning-of-year market value of
equity
Y-9C
NIGL Fair value gains and losses recorded in net income divided by beginning-of-year market value of equity Y-9C
OCIGL Fair value gains and losses recorded in other comprehensive income divided by beginning-of-year market value of
equity
Y-9C
R Twelve-month stock return ending four months after the ?scal year end CRSP
SIZE Log of total assets Y-9C
TIER1 Tier one risk-based capital ratio Y-9C
TRADD Indicator variable equals to one if the bank engages in trading activities and zero otherwise Y-9C
UNSTABLE Indicator equal to one if the standard deviation of the daily yield on seven-year U.S. Treasuries is above the sample
median for the year and zero otherwise
Federal Reserve
Notes: Regarding bank data sources.
Y-9C refers to banks’ regulatory Y-9C ?lings. 10-K refers to
banks’ Form 10-K ?lings. SNL refers to SNL Financial.
References
Aebi, V., Sabato, G., & Schmid, M. (2012). Risk management, corporate governance,
and bank performance in the ?nancial crisis. Journal of Banking and Finance,
36(2012), 3213–3226.
Ahmed, A., Kilic, E., & Lobo, G. (2006). Does recognition versus disclosure matter?
Evidence from value-relevance of banks recognized and disclosed derivative ?-
nancial instruments. The Accounting Review, 81(3), 567–588.
Ahmed, A., & Takeda, C. (1995). Stock market valuation of gains and losses on com-
mercial banks’ investment securities: An empirical analysis. Journal of Account-
ing & Economics, 20, 207–222.
Badertscher, B., Burks, J., & Easton, P. (2014). The market pricing of other than tem-
porary impairments. The Accounting Review, 89(3), 811–838.
Barth, M. (1994). Fair value accounting: Evidence from investment securities and
the market valuation of banks. The Accounting Review, 69, 1–25.
Barth, M., Beaver, W., & Landsman, W. (1996). Value-relevance of banks’ fair value
disclosure under SFAS 107. The Accounting Review, 71(4), 513–537.
Barth, M., Beaver, W., & Landsman, W. (2001). The relevance of the value-relevance
literature for ?nancial accounting standard setting: Another view. Journal of Ac-
counting & Economics, 71(4), 77–104.
Basel Committee on Banking Supervision (2003). Public Disclosures by Banks: Results
of the 2001 Disclosure Survey, May.
Beaver, W., & Venkatchalam, S. (2003). Differential pricing of components of bank
loan fair values. Journal of Accounting, Auditing and Finance, 18(1), 41–67.
Bhat, G., Ryan, S., & Vyas, D. (2014). The implications of credit risk modeling for banks’
loan loss provision timeliness and loan origination procyclicality. Working paper,
New York University.
Carroll, T., Linsmeier, T., & Petroni, K. (2003). The reliability of fair value versus his-
torical cost information: Evidence from closed-end mutual funds. Journal of Ac-
counting, Auditing, and Finance, 18(1), 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, 557–593.
Chernobai, A., Jorion, P., & Yu, F. (2011). The determinants of operational risk in U.S.
?nancial institutions. Journal of Financial and Quantitative Analysis, 46(6), 1683–
1725.
Chung, S., Goh, B., Ng, J., & Yong, K. (2014). Voluntary fair value disclosures beyond
SFAS 157’s three-level estimates. Working paper, Singapore Management Univer-
sity.
Danbolt, J., & Rees, W. (2008). An experiment in fair value accounting: UK invest-
ment vehicles. European Accounting Review, 17(2), 271–303.
Dong, M., Ryan, S., & Zhang, X. (2014). Preserving amortized costs within a fair-
value-accounting framework: Reclassi?cation of gains and losses on available-
for-sale securities upon realization. Review of Accounting Studies, 19(1), 242–280.
Eccher, E., Ramesh, K., & Thiagarajan, S. (1996). Fair value disclosures by bank hold-
ing companies. Journal of Accounting & Economics, 22, 79–117.
Ellul, A., & Yerramilli, V. (2013). Stronger risk controls, lower risk: Evidence from
U.S. bank holding companies. Journal of Finance, 68(5), 1757–1803.
Goh, B., Li, D., Ng, J., & Yong, K. (2015). Market pricing of banks’ fair value assets
reported under SFAS 157 during the 2008 Financial Crisis. Journal of Accounting
and Public Policy (forthcoming).
Healy, P., & Palepu, K. (2001). Information asymmetry, corporate disclosure, and the
capital markets: A review of the empirical disclosure literature. Journal of Ac-
counting & Economics, 31, 405–440.
Hirst, D., Hopkins, P., & Wahlen, J. (2004). Fair values, income measurement, and
bank analysts’ risk and valuation judgments. The Accounting Review, 79(2), 453–
472.
Holthausen, R., & Verrecchia, R. (1988). The effect of sequential information releases
on the variance of price changes in an intertemporal multi-asset market. Journal
of Accounting Research, 3, 82–106.
Holthausen, R., & Watts, R. (2001). The relevance of the value-relevance literature
for ?nancial accounting standard setting. Journal of Accounting & Economics,
31(1–3), 3–75.
Kaplan, R., & Mikes, A. (2012). Managing risks: A new framework. Harvard Business
Review June: 48–60.
Kolev, Kalin S. (2009). Do investors perceive marking-to-model as marking-to-myth?
Early evidence from FAS 157 Disclosure. Working paper. Yale School of Manage-
ment.
Lang, M., & Lundholm, R. (1993). Cross-sectional determinants of analysts ratings of
corporate disclosure. Journal of Accounting Research, 31(2), 246–271.
Larcker, D., Richardson, S., & Tuna, A. (2007). Corporate governance, accounting out-
comes, and organizational performance. The Accounting Review, 82(4), 963–1008.
Mikes, A. (2011). From counting risk to making risk count: Boundary-work in risk
management. Accounting, Organizations and Society, 36, 226–245.
Nelson, K. (1996). Fair value accounting for commercial banks: An empirical analysis
of SFAS No. 107. The Accounting Review, 71(2), 161–182.
Nissim, D. (2003). Reliability of banks’ fair value disclosure for loans. Review of
Quantitative Finance and Accounting, 20(4), 355–384.
Ryan, S. (2008). Accounting in and for the subprime crisis. The Accounting Review,
83, 1605–1638.
Ryan, S. (2011). Financial reporting for ?nancial instruments. Foundations and Trends
in Accounting, 6(3–4), 187–354.
Song, C., Thomas, W., & Yi, H. (2010). Value-relevance of FAS 157 fair value hierarchy
information and the impact of corporate governance mechanisms. The Account-
ing Review, 85, 1375–1410.
Venkatachalam, M. (1996). Value-relevance of banks’ derivatives disclosures. Journal
of Accounting & Economics, 22, 327–355.
Vickrey, J., & Wright, J. (2013). TBA trading and liquidity in the agency MBS market.
FRBNY Economic Policy Review, 19(1), 1–18.
doc_839302394.pdf