Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Does fair value accounting exacerbate the pro-cyclicality of bank lending?
(USC Thesis Other)
Does fair value accounting exacerbate the pro-cyclicality of bank lending?
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
DOES FAIR VALUE ACCOUNTING EXACERBATE THE
PRO-CYCLICALITY OF BANK LENDING?
by
Biqin Xie
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)
August 2012
Copyright 2012 Biqin Xie
ii
Dedication
This dissertation is dedicated to my parents Xuanhua Xie and Hanying Mao, my husband
Bin Gu, my siblings Bishuang Xie, Bifeng Xie, Zhenglong Xie, and Zhengjun Xie, and
my grandparents.
iii
Acknowledgements
I would like to express my deepest gratitude to my dissertation chair, K.R.
Subramanyam, for his continuous guidance and insightful suggestions in the development
of this paper. I am also thankful to my other committee members Mark DeFond, Jeffrey
Nugent and, especially, Jieying Zhang for their comments and support. Special thanks to
Yihong Jiang for his longtime support. This paper benefited from helpful comments from
Randy Beatty, Elizabeth Chuk, Wayne Ferson, Jessica Halenda, Rebecca Hann, Mingyi
Hung, Ayse Imrohoroglu, Yuri Loktionov, Kevin Murphy, Siqi Li, David Maber, Jeff
McMullin, Suresh Nallareddy, Bryce Schonberger, Karen Ton, Robert Tresevant, Selale
Tuzel, Yanhui Wu, Alicia Yancy, and workshop participants at University of Southern
California, Penn State University, Arizona State University, Florida International
University, and George Mason University. I am grateful to the Marshall School of
Business at the University of Southern California for financial support. The Dissertation
Completion Fellowship from USC Graduate School is gratefully acknowledged.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vi
List of Figures vii
Abstract viii
Chapter 1: Introduction 1
Chapter 2: Motivation 11
Chapter 3: Hypothesis Development 15
3.1. Hypothesis on Necessary Condition 1: Whether Unrealized
Gains (Losses) Affect Lending 16
3.2 Hypothesis on Necessary Condition 2: Whether Fair Values of
Bank Assets are Pro-cyclical 18
3.3 Hypothesis Development: Whether Fair Value Accounting
Exacerbates the Pro-cyclicality of Lending 20
Chapter 4: Sample and Data 22
Chapter 5: Empirical Analyses and Results 23
5.1 Does Fair Value Accounting Exacerbate the Pro-cyclicality of
Lending? 23
5.1.1 Research Design 23
5.1.2 Descriptive Statistics 29
5.1.3 Empirical Results 32
5.2 Do Unrealized Gains (Losses) Affect Lending? 40
5.2.1 Research Design 40
5.2.2 Descriptive Statistics 42
5.2.3 Empirical Results 43
5.3 Are Fair Values Pro-cyclical? 49
5.3.1 Research Design 49
5.3.2 Empirical Results: Industry-level Analysis 51
5.3.3 Empirical Results: Firm-level Analysis 55
5.3.4 Why are Fair Values of Bank Assets not Pro-cyclical? 57
5.4 Generalizability of Results 63
v
Chapter 6: Conclusion 64
Bibliography 66
Appendix: Variable definitions 71
vi
List of Tables
Table 1. Summary Statistics 30
Table 2. Pearson Correlations 31
Table 3. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending:
Not Controlling for Firm Fixed Effects 33
Table 4. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending:
After Controlling for Firm Fixed Effects 37
Table 5. Whether Unrealized Gains (Losses) Affect Banking Lending:
Not Controlling for Firm Fixed Effects 44
Table 6. Whether Unrealized Gains (Losses) Affect Banking Lending:
Controlling for Firm Fixed Effects 47
Table 7. Whether Fair Values are Pro-cyclical: Industry-level Analysis 54
Table 8. Whether Fair Values are Pro-cyclical: Firm-level Analysis 56
Table 9. Whether Changes in Fair Values are Related to Interest Rate Changes 61
vii
List of Figures
Figure 1. Two Necessary Conditions for Fair Value Accounting to Exacerbate
Lending Pro-cyclicality 15
Figure 2. Intrinsic Pro-cyclicality of Bank Lending 21
Figure 3. Industry Aggregate Unrealized Gains (Losses) Recognized
during the Year 52
Figure 4. Time Series of Unrealized Gains (Losses) Recognized during the
Quarter for the Banking Industry in Relation to the Risk-free Rate
during 1995-2010 59
Figure 5. Times Series of Term Spread and Default Spread during 1995-2010 59
Figure 6. Time Series of Unrealized Gains (Losses) Recognized during the
Quarter for the Banking Industry in Relation to the Interest Rates
during 1995-2010 60
viii
Abstract
This paper examines whether fair value accounting increases the pro-cyclicality of banks’
lending behavior. Exploiting cross-sectional variation in individual banks’ exposure to
fair value accounting, I find that fair value accounting does not exacerbate the pro-
cyclicality of bank lending over the past two business cycles during 1995-2010. This
result holds despite the fact that every one dollar of unrealized gains (losses) is associated
with at least 12 cents of new lending (cutbacks in lending). The probable cause of this
non-exacerbation finding is that interest rates rise (fall) during some of the expansionary
(recessionary) periods, resulting in movements of bank assets’ fair value that are not pro-
cyclical.
1
Chapter 1: Introduction
“The second problem involves ways of making the system less pro-cyclical so
that the financial system is less susceptible to exuberant booms and disastrous
busts…. Capital rules, accounting policies, and other regulatory standards should
not make this job even more difficult by encouraging excessively pro-cyclical
behavior by financial institutions, that is, behavior that causes financial
institutions to tighten credit in downturns and ease credit in booms more than is
justified by changes in the credit worthiness of borrowers.”
- Ben Bernanke, March 20, 2009
1
Fair value accounting has come under severe attack for its alleged role in
precipitating the financial crisis of 2007-2008 (U.S. Congress, 2008; American Bankers
Association, 2008). Central to the concerns about fair value accounting is whether it
exacerbates the pro-cyclical lending behavior of financial institutions, i.e., the tendency
of banks to expand credit during economic booms and curtail credit during busts.
2,3
Excessive pro-cyclicality in the financial system amplifies economic cycles, increases
financial instability, and hence is of great concern to bank regulators and policymakers
(Financial Stability Forum, 2009; Financial Stability Board, 2009; Bank of Italy, 2009).
Financial regulators are especially concerned that fair value accounting may have
exacerbated the pro-cyclicality of the banking system (European Central Bank, 2004;
Banque de France, 2008). Supporting these concerns, two simulation-based studies show
1
From Ben Bernanke’s speech at the Independent Community Bankers of America's National Convention
and Techworld, Phoenix, Arizona. Emphasis added.
2
The standard textbook definition of pro-cyclicality is the tendency of an economic variable to move in the
same direction as aggregate economic activity over the business cycle (up in expansions, down in
contractions) (see, e.g., Abel and Bernanke, 2005).
3
At the height of the financial crisis, sixty-five members of Congress write a letter to the SEC, stating that
“the ‘mark-to-market’ rule, while well intended, has the unintended consequences of exacerbating
economic downturns by hamstringing the ability of banks to make loans to consumers and businesses” (see
U.S. Congress, 2008).
2
that fair value accounting induces unintended pro-cyclicality in banks’ balance sheets and
can, thereby, exacerbate the pro-cyclicality of their lending (European Central Bank,
2004; International Monetary Fund, 2009). Despite its importance, however, there is little
empirical evidence on this issue. My paper attempts to fill this gap.
Specifically, the objective of my paper is to study whether fair value accounting
exacerbates the intrinsic pro-cyclicality of banks’ lending behavior over the business
cycle, i.e., whether fair value accounting induces banks to expand credit during economic
expansions and curtail credit supply during recessions. Bank lending is intrinsically pro-
cyclical (Fisher, 1933; Kiyotaki and Moore, 1997; Asea and Blomberg, 1998; Lown and
Morgan, 2006), and fair value accounting is alleged to increase lending pro-cyclicality
because fair values of bank assets are conjectured to increase (decrease) during economic
upturns (downturns), resulting in “surplus capital” (“capital shortfall”) that facilitates
(restrains) lending during those periods.
I use a broad sample of 2,956 unique bank holding companies comprising 69,578
bank-quarter observations during 1995-2010 to conduct the main empirical analysis. To
measure fair value accounting, I exploit the cross-sectional variation in the extent of
individual banks’ fair value accounting exposure, measured as the fraction of bank assets
carried at fair value on the balance sheet (Nissim and Penman, 2007; SEC, 2008; Khan,
2010).
4,5
To capture bank lending behavior, I use the quarterly growth in outstanding
4
Not all assets are carried at fair value on the balance sheet. During 1994-2007, only trading assets and
available-for-sale securities are carried at fair value on the balance sheets (SFAS 115); since 2008, assets
that banks elect to use fair value accounting are also carried at fair value on the balance sheet (SFAS 159).
5
The measure used by Nissim and Penman (2007) and Khan (2010), however, includes both assets and
liabilities that are recognized on the balance sheet at fair value or disclosed in the footnote at fair value.
3
loans as in Berger and Udell (2004), Beatty and Liao (2011), and Cornett et al. (2011),
among others. I next partition the sample period into recessions (2001Q2-2001Q4 and
2008Q1-2009Q2) and expansions (the remaining quarters during 1995-2010) in order to
get time-series variation in macroeconomic conditions.
6
My primary analysis examines
differential lending behavior across banks with high versus low fair value accounting
exposure during expansionary and recessionary periods. If fair value accounting has pro-
cyclical effects, I would expect to find that banks with high fair value accounting
exposure increase lending more rapidly during expansions and curtail lending more
substantially during recessions than banks with low fair value accounting exposure.
I begin my empirical analysis by investigating my primary research question –
whether banks’ fair value exposure impacts the pro-cyclicality of their lending. Contrary
to conventional wisdom, I find that banks with greater fair value accounting exposure do
not increase lending more rapidly in expansions, nor do they curtail lending more
substantially in recessions relative to banks with less fair value accounting exposure.
These results are robust to controlling for factors shown in the literature to affect bank
lending.
One potential concern with using the fraction of total assets carried at fair value
(i.e., Fair-Valued Assets /Total Assets) to measure fair value exposure is that this
Since fair values recognized on the balance sheet affect bank examiners’ assessments of a bank’s capital
adequacy (Office of the Comptroller of the Currency, 1994) and hence are more likely to affect a bank’s
lending behavior compared to fair values disclosed in the footnote, the fair value accounting exposure
measure used here only includes assets that are recognized at fair value on the balance sheet.
6
The partition into expansions and recessions is based on National Bureau of Economic Research (NBER)
business cycle dates.
4
measure is inexorably linked to banks’ business models – heavy involvement in securities
business (vs. lending business) automatically results in greater (vs. less) fair value
accounting exposure. Furthermore, banks’ business models obviously affect their lending
behavior. Thus, it is important to control for cross-bank differences in their business
models. I achieve this in two ways: (1) controlling for banks’ asset structures measured as
the fraction of bank assets that is securities (i.e., Book Value of Securities /Total Assets;
higher values indicate more securities-oriented business models), and (2) controlling for
firm fixed effects. After controlling for banks’ business models, I still find that fair value
accounting does not appear to have exacerbated the pro-cyclicality of bank lending over
the past two business cycles.
The above finding begs the question: what caused this counterintuitive no-
exacerbation finding? For fair value accounting to exacerbate the pro-cyclicality of bank
lending, two necessary conditions must be met. First, increases (decreases) in fair values,
i.e., unrealized gains (losses), through their impact on the strength of banks’ balance
sheets, must induce banks to increase (reduce) lending.
7,8
Second, fair values of bank
7
With the five exceptions detailed in footnote 8 below, most unrealized gains (losses) from securities are
included in “Accumulated Other Comprehensive Income” as part of “Shareholders’ Equity” on the balance
sheet, and are excluded from earnings and regulatory capital. According to the rules set by Office of the
Comptroller of the Currency (1994), “examiners will consider both unrealized gains and losses in their
evaluation of the adequacy of a bank's regulatory capital. When unrealized losses could threaten a bank's
financial condition, other regulatory actions that are based on regulatory capital may be initiated”.
8
The five exceptions in which unrealized gains (losses) are included in earnings and regulatory capital are:
(1) Unrealized gains (losses) from trading assets are included in earnings and regulatory capital; trading
assets, however, represents a very small fraction of a typical bank holding company’ assets. (2) Unrealized
losses from equity securities are included in Tier 1 regulatory capital; equity securities, however, constitute
a very small fraction of a typical bank holding company’s investment securities. (3) Up to 45% of pretax
unrealized gains from available-for-sale equity securities can be included in Tier 2 regulatory capital. See
http://www.gpo.gov/fdsys/pkg/FR-1998-09-01/html/98-23379.htm. (4) Unrealized losses from available-
for-sale and held-to-maturity securities that are deemed “other than temporary” (referred to as “other-than-
temporary impairment” or “OTTI”). The part of OTTI representing credit losses must be included in
5
assets must move in a pro-cyclical manner on average, increasing during economic
expansions and decreasing during recessions. In other words, banks must experience
unrealized gains in economic upturns and unrealized losses in downturns. Violations of
either of these two necessary conditions would imply that fair value accounting is
unlikely to increase lending pro-cyclicality. Thus, to understand why fair value
accounting does not increase lending pro-cyclicality, I examine which of these two
necessary conditions are violated.
I first investigate whether the first necessary condition is violated, i.e., whether
unrealized gains and losses recognized on the balance sheet do not affect banks’
subsequent lending behavior. Because the objective of this analysis is simply to examine
whether a link exists between unrealized gains (losses) and subsequent lending behavior
without considering whether such a relationship is pro-cyclical, I perform these tests on a
pooled cross-sectional and time-series basis over the full sample period. Given that bank
regulators consider unrealized gains (losses) recognized on the balance sheet in assessing
banks’ capital adequacy (Office of the Comptroller of the Currency, 1994), I hypothesize
that unrealized gains (losses) affect subsequent bank lending. Consistent with this
hypothesis, I find that unrealized gains (losses) recognized on the balance sheet during
the past two to four quarter(s) increase (decrease) banks’ lending activities in the current
earnings and regulatory capital, and the rest of OTTI are included in Other Comprehensive Income and
excluded from regulatory capital. OTTI mostly occurred during the recent financial crisis and recession.
However, these OTTI could have also been recognized even under historical cost accounting with
impairment and writedowns. (5) For fiscal years ending after November 15, 2007, unrealized gains (losses)
of assets elected to use fair value accounting under FAS 159 are included in earnings and regulatory capital
(the exception being that changes in fair values of liabilities caused by deterioration of the banks’ own
creditworthiness go into earnings, but not into tier 1 capital.) I find that the extent to which my sample bank
holding companies use this “fair value option” is limited.
6
quarter during 1995-2010 – a period in which fair value accounting rules are in effect. In
particular, every one dollar of unrealized gains (losses) recognized over the past two to
four quarter(s) is associated with at least 12 cents of increase (decrease) in lending during
the current quarter. In contrast, holding gains (losses) that occurred over the past one to
four quarter(s) have very small or no impact on the current quarter’s lending during 1988-
1991 – a period in which fair value accounting rules are not in effect.
9 , 10
These
comparative analyses suggest that it is not changes in fair values per se, but rather the use
of fair value accounting to recognize those changes on the balance sheet that affects
subsequent lending. These findings are robust to the inclusion of time fixed effects and
factors shown by prior literature to affect lending, suggesting that the results are less
likely to be driven by correlated omitted variables. Thus, the first necessary condition for
fair value accounting to exacerbate lending pro-cyclicality does not appear to be violated.
I next examine whether the second necessary condition (for fair value accounting
to exacerbate lending pro-cyclicality) is violated, i.e., whether the fair values of bank
assets are not pro-cyclical. Contrary to conventional wisdom, I find that the fair values of
9
During the fair-value-accounting periods, unrealized gains (losses) in my empirical analyses refer to
unrealized gains (losses) from available-for-sale securities that are included in “Accumulated Other
Comprehensive Income (AOCI)” as part of “Shareholders’ Equity” on the Balance Sheet. The reason I
exclude unrealized gains (losses) from trading assets is that the data are not available; such exclusion,
however, has very small impact on my analysis because trading assets represent a very small fraction of
total assets of my sample bank holding companies. During pre-fair-value-accounting periods, unrealized
gains (losses) refer to holding gains (losses) from investment securities; these unrealized gains (losses) are
not recognized on the balance sheet during the pre-fair-value-accounting periods.
10
My empirical analyses focus on fair values of asset side instead of liability side of the balance sheet. The
reason is that only a small fraction of liabilities are fair valued (Prior to 2008, only trading liabilities and
derivative liabilities are fair valued, and trading and derivative liabilities constitute a very small portion of
banks’ total liabilities; although banks can elect to fair value their liabilities for fiscal years ending after
November 15, 2007, I find that the extent to which my sample bank holding companies use this “fair value
option” on their liabilities is limited.)
7
bank assets are not pro-cyclical at either the industry level or the firm level. At the
industry level, I find that the banking industry incurs unrealized losses during almost half
of the expansionary quarters, and these losses offset 68% of the unrealized gains that the
industry experiences during the remaining expansionary quarters. Equally surprisingly,
the industry reaps unrealized gains during one-third of the recessionary quarters, and
these gains almost completely offset the entire unrealized losses that the industry incurs
during the remaining recessionary quarters. At the firm level, I find that banks with
greater fair value accounting exposure actually incur slightly greater unrealized losses
during expansion periods relative to banks with less fair value accounting exposure;
however, I do not observe a similar relationship with respect to unrealized gains or losses
in recessions. Taken together, this evidence suggests that the fair values of banks assets
are not pro-cyclical at either the industry or the firm level. Thus, the second necessary
condition appears to be violated, which helps explain why fair value accounting does not
exacerbate the pro-cyclicality of bank lending.
11
Given the above finding, in my final analysis I investigate why the fair values of
bank assets are not pro-cyclical. I find that the answer probably resides in the pattern of
interest rate movements, as such interest rate movements profoundly impact fair values of
bank assets. Studying how the three components of interest rates – namely, risk-free rate,
term spread, and default spread – move over the business cycle, I find that the risk-free
11
My prior analyses suggest that the relationship between unrealized gains (losses) and bank lending is a
lagged relationship, i.e., unrealized gains (losses) recognized during the past one to four quarters affects
bank lending in the current quarter. Accordingly, I perform a sensitivity test to examine whether the lagged
fair values of bank assets are pro-cyclical. I find that the lagged fair values of bank assets are not pro-
cyclical, either.
8
rate rises (the risk-free rate and the default spread fall) during some of the expansionary
(recessionary) periods, causing the fair values of banks’ fair-valued assets to fall (rise)
during those periods and therefore, making the fair values of bank assets not pro-cyclical.
Overall, the evidence in this paper suggests that fair value accounting does not
exacerbate the pro-cyclicality of bank lending behavior over the past two business cycles
during 1995-2010.
12
This result holds despite the fact that changes in fair values (i.e.,
unrealized gains/losses) have a significant impact on subsequent lending behavior. The
evidence further suggests that this non-exacerbation finding is probably due to the fact
that fair values of bank assets do not move in a pro-cyclical manner – interest rates rise
(fall) during some of the expansionary (recessionary) periods, resulting in movements of
the fair values of bank assets that are not pro-cyclical.
My paper makes the following contributions. First, my study adds to the fair value
accounting and banking literature by providing empirical evidence on the relationship
between fair value accounting and banks’ lending behavior. Despite deep concerns
among financial regulators and policymakers that fair value accounting may have a pro-
cyclical impact on bank lending and, thereby, increase financial instability, little
empirical evidence has been presented to support such concerns. To my knowledge, this
is the first archival study to examine whether fair value accounting exacerbates the pro-
cyclicality of bank lending. In so doing, this paper helps inform the policy debate on fair
value accounting and its impact on the financial system.
12
Importantly, the finding that fair value accounting does not exacerbate the pro-cyclicality of lending
would probably still hold even if we had used full fair value accounting during the past two business cycles,
since fair values of banks’ entire balance sheets would still likely not be pro-cyclical under full fair value
accounting as bank assets are largely fixed-income instruments whose fair values are very sensitive to
interest rate changes.
9
Second, my paper provides empirical evidence that sheds light on whether fair
value accounting contributed to the recent financial crisis. As I find that the banking
sector incurs unrealized losses almost as much as it experiences unrealized gains during
expansionary periods, fair value accounting is less likely to have contributed to banks’
excessive leverage and risk-taking during the recent booms, which contradicts critics’
claims.
13
Likewise, I find that the unrealized gains reaped by the banking industry during
some of the recessionary quarters almost completely offset the unrealized losses incurred
during the remaining recessionary quarters. This result suggests that fair value accounting
is less likely to have contributed to the substantial contraction in credit supply during the
2007-2008 financial crisis, again contradicting critics’ allegations. Rather, my results are
consistent with recent papers that suggest that fair value accounting did not contribute to
the recent financial crisis (e.g., SEC, 2008; Ryan, 2008; Ball, 2008; Laux and Leuz,
2010; Barth and Landsman, 2010; Shaffer, 2010; Badertrcher, Burks, and Easton, 2012).
Finally, this paper should be of interest to accounting standard setters and
policymakers in deciding whether to require firms to carry other financial instruments,
such as bank loans, at fair value on their balance sheets. On the one hand, the finding that
fair value accounting has not exacerbated the pro-cyclicality in bank lending over the past
two business cycles should ease concerns about the potential pro-cyclical effects of fair
value accounting on banks. On the other hand, the finding that unrealized gains (losses)
13
Adrian and Shin (2010, 2011) show that investment banks actively expand (contract) their balance sheets
during economic upturns (downturns), resulting in high (low) leverage during upturns (downturns).
10
do have a significant impact on bank lending suggests that a wider application of fair
value accounting to other bank assets such as bank loans must be taken cautiously.
14
14
FASB (2010) proposed to require fair value of bank loans, but backed off due to strong opposition from
banks and American Bankers Association (see Moore, 2011).
11
Chapter 2: Motivation
My primary interest is whether fair value accounting exacerbates the pro-
cyclicality of bank lending behavior over the business cycle. While financial regulators
and policymakers are deeply concerned that fair value accounting may introduce
unintended pro-cyclicality in bank balance sheets and, thereby, exacerbate the pro-
cyclicality in bank lending over the business cycle (European Central Bank, 2004;
Banque de France, 2008), to my knowledge, no direct archival research addresses this
question. However, several related strands of literature provide background and
motivation for my study.
One related stream of research is the theoretical work on the impact of mark-to-
market accounting in bad times.
15
Allen and Carletti (2008) model the contagion effect of
mark-to-market accounting during financial crises. They find that the interaction between
liquidity pricing and mark-to-market accounting can depress asset prices below
fundamental values, causing contagion from the insurance sector to the banking sector.
Plantin, Sapra, and Shin (2008) model the feedback effect of mark-to-market accounting
in bad times. They show that as liquidity dries up, mark-to-market leads to pro-cyclical
trades that inject excessive, endogenous volatility into prices, which induces suboptimal
real decisions, especially for illiquid, long-lived, and senior claims.
16
However, neither of
15
Mark-to-market accounting can be viewed as the pure form of fair value accounting, as it uses market
prices to determine fair value.
16
Based on premises quite different from those of Plantin, Sapra, and Shin (2008), the model in O’Hara
(1993) predicts that in the presence of information asymmetry, mark-to-market accounting induces banks to
prefer issuing short-term loans over long-term loans.
12
these analytical works addresses whether and how fair value accounting affects banks’
lending behavior.
A second related literature is the burgeoning empirical work and discussion
papers on whether fair value accounting contributed to the recent financial crisis.
17
Overall, the evidence is mixed. On one hand, multiple opinion pieces and empirical
studies come to similar conclusions that fair value accounting played little or no role in
the financial crisis (SEC, 2008; Ryan, 2008; Ball, 2008; Laux and Leuz, 2010; Barth and
Landsman, 2010; Shaffer, 2010; Badertrcher, Burks, and Easton, 2012).
18
On the other
hand, Khan (2010) finds that fair value accounting increases contagion effects among
banks, especially during periods of market illiquidity; additionally, Bhat, Frankel, and
Martin (2011) find evidence of a feedback effect between trading and prices of mortgage-
backed securities during the recent financial crisis. Given the mixed evidence and the
importance of this issue, Laux and Leuz (2009) call for more research to resolve the
debate on fair value accounting. As loan growth is an important driver of the riskiness of
banks (Foos, Norden, and Weber, 2010), and the seeds of the financial crisis could have
been sown during the boom years preceding the financial crisis, my paper helps shed
light on the fair value accounting debate by examining whether fair value accounting
17
Another stream of literature focuses on relevance of fair values for equity investors (and creditors); for
example, Hann, Heflin, and Subramanyam (2007) examine value relevance and credit relevance of fair
value pension accounting. For an overview of early literature on value relevance of fair values, see Wahlen
et al. (2000). Other strands of the recent literature center on reliability of fair value estimates (e.g., Kolev
2009), risk information of fair values (e.g., Riedl and Serafeim, 2011; Blankespoor, Linsmeier, Petroni, and
Shakespeare, 2011), and information content of unrealized cash flow hedge gains/losses (Campbell, 2011).
18
For analyses of broad issues that have contributed to the financial crisis, see Acharya and Richardson
(2009), Diamond and Rajan (2009), and Rajan (2010).
13
induced banks to increase lending more rapidly (and thus take more risk) during the
boom years preceding the crisis.
Additionally, closely related to this paper are two simulation studies by
International Monetary Fund (2009) and European Central Bank (2004), which
investigate whether full fair value accounting (under which all assets are carried at fair
value on the balance sheet) has a pro-cyclical effect on banks’ equity capital over
hypothetical business cycles. Both of these simulation studies conclude that fair value
accounting can indeed have a pro-cyclical impact on lending. However, it is unclear
whether these simulation results are generalizable to the real world, since the assumptions
underlying those hypothetical business cycles may not be valid in the real world. To
overcome such limitations of these simulation exercises, my paper uses real-world
empirical data to investigate whether fair value accounting actually induces banks to
increase lending more rapidly during economic expansions and forces banks to curtail
lending more substantially during recessions. To my knowledge, this is the first study to
provide large-sample archival evidence on whether fair value accounting exacerbates
lending pro-cyclicality over the business cycle.
Finally, an interesting study by Beatty and Liao (2011) examines whether the
current loan loss provisioning rules magnify the pro-cyclicality of bank lending. They
find that banks that delay less in recognizing expected loan losses experience smaller
reduction in lending during the most recent recession. My paper is related to Beatty and
Liao (2011) in that both papers study whether accounting rules increase the pro-
14
cyclicality of bank lending; my focus is very different from theirs, however, as they focus
on delays in loan loss provisioning.
To summarize, although there are related studies, none addresses whether fair
value accounting exacerbates the pro-cyclicality of bank lending. Given the importance
of this issue, my paper contributes to both the literature and the policy debates on fair
value accounting.
15
Chapter 3: Hypothesis Development
As discussed above, for fair value accounting to have a pro-cyclical effect on
bank lending, two necessary conditions must be met: (1) changes in the fair values of
bank assets affect lending; (2) the fair values of bank assets are pro-cyclical (see Figure
1). Whether fair value accounting increases lending pro-cyclicality crucially hinges on
whether both of these necessary conditions are met. Below I discuss each of these two
necessary conditions, as well as the implications for whether fair value accounting
increases lending pro-cyclicality.
Figure 1. Two Necessary Conditions for Fair Value Accounting to Exacerbate
Lending Pro-cyclicality
Fair value accounting
exposure
Expansion
Recession
Fair values of bank
assets increase
Fair values of bank
assets decrease
Lending increases
Lending decreases
Necessary Condition 1 Necessary Condition 2
Necessary Condition 1
Necessary Condition 2
16
3.1. Hypothesis on Necessary Condition 1: Whether Unrealized Gains (Losses)
Affect Lending
Unrealized gains (losses) can affect banks’ lending behavior by impacting their
regulatory costs.
19
Regulatory costs are costs arising from regulatory intervention as
banks’ capital ratios fall toward or below the statutory minimum; the extreme form of
such intervention is seizure of the bank by the Federal Deposit Insurance Corporation
(FDIC).
With a few exceptions noted in footnote 8, most unrealized gains (losses) are
excluded from both earnings and regulatory capital (Office of the Comptroller of the
Currency, 1994).
20
As a result, these unrealized gains may not turn into “surplus capital”
that banks can deploy to make more loans, and unrealized losses may not lead to “capital
shortfalls” that force banks to curtail lending. For this reason, it is often implicitly
assumed that these unrealized gains and losses have little impact on bank behavior, an
assumption that has never been empirically tested. That said, although fluctuations in the
fair values of securities are mostly excluded from regulatory capital, “examiners will
consider both unrealized gains and losses in their evaluation of the adequacy of a bank's
19
I focus on regulatory costs instead of contracting costs for the following reason: unrealized gains and
losses recognized on the balance sheet mostly come from available-for-sale securities; these unrealized
gains and losses from available-for-sale securities are excluded from earnings and, thus, are less likely to
affect contractual constraints tied to earnings, such as managerial compensation contracts and debt
contracts. Thus, unrealized gains and losses are less likely to affect banks’ contracting costs. Note that (1)
unrealized gains and losses of assets elected to use the fair value option afforded by SFAS 159 since
November 15, 2007 are included in earnings; however, I find that the extent to which banks use the fair
value option during 2008-2010 is limited, and (2) under SFAS 159, banks can elect to use the fair value
option on securities that were originally classified as available-for-sale securities under SFAS 115.
Consequently, they must reclassify these available-for-sale securities as trading assets.
17
regulatory capital. When unrealized losses could threaten a bank's financial condition,
other regulatory actions that are based on regulatory capital may be initiated” (Office of
the Comptroller of the Currency, 1994).
Since unrealized gains (losses) affect bank examiners’ overall assessment of a
bank’s capital adequacy and, in turn, affect the likelihood that banks will face regulatory
intervention, they likely impact bank managers’ lending decisions. In particular, as
unrealized gains recognized on the balance sheet are viewed favorably by bank examiners,
these gains put the bank in a better capital position than its regulatory capital ratios
suggest, reduce the probability of regulatory intervention and, thereby, allow bank
managers to take on more risk by issuing more loans. Conversely, since unrealized losses
are viewed negatively by bank examiners, these losses put the bank in a worse capital
position than its regulatory capital ratios suggest, increase the probability of regulatory
intervention and, thereby, force bank managers to reduce risk exposures by curtailing
lending.
21
This motivates the following hypothesis:
Hypothesis 1: Unrealized gains are associated with an increase in subsequent lending,
whereas unrealized losses are associated with a cutback in subsequent lending.
To my knowledge, no empirical papers have investigated whether and how
unrealized gains (losses) excluded from earnings and regulatory capital affect bank
managers’ behavior.
21
Extending more credit is considered risk-taking for two reasons. First, from a regulatory standpoint,
making loans is riskier than the alternative of investing in highly-rated securities such as highly-rated
agency-sponsored mortgage-backed securities, as banks need to hold much more capital against loans than
against investments in highly-rated agency-sponsored mortgage-backed securities. Second, loan growth has
been shown to be an important driver of bank risk (Foos, Norden, and Weber, 2010).
18
3.2 Hypothesis on Necessary Condition 2: Whether Fair Values of Bank Assets
are Pro-cyclical
It is often assumed that fair values of bank assets move in a pro-cyclical manner:
in good times, most asset classes appreciate in value, resulting in fair value gains; in bad
times, most asset classes decline in value, resulting in fair value losses. This relationship
may not hold for banks, however. A unique aspect of banks is that banks’ assets consist
largely of fixed-income instruments such as debt securities and loans whose fair values
are very sensitive to interest rate changes. In fact, interest-rate risk is by far the greatest
risk faced by debt securities investors (Fabozzi, 2005). Importantly, the fair value of debt
securities moves in the opposite direction of changes in interest rates. To illustrate, the
fair value of debt securities (as well as other fixed-income instruments such as bank loans
and medium-term notes) can be calculated as follows:
22
∑
=
+ + +
=
n
t
t
t t t
t
Spread Default Spread Term rf
Flow Cash Expected
Value Fair
1
) 1 (
where
t t t
Spread Default Spread Term rf + + is the time-varying interest rate required by
investors to hold the debt securities,
23
and n is the contractual maturity of the debt
securities. This equation illustrates that, as interest rates increase (decrease), the fair value
of the debt securities decreases (increases).
24
22
The expected cash flow takes into account factors such as probability of default and early prepayment of
principal, factors that are very important in valuing mortgage-backed securities (Schwartz and Torous,
1989; Stanton, 1995).
23
The interest rate required by investors is also referred to as required yield. The relationship between
required yield and fair value of debt securities is convex (Fabozzi, 2005).
24
In theory, only unexpected changes in interest rates will affect banks’ net worth.
19
Since interest-rate risk is by far the greatest risk faced by debt securities investors
(Fabozzi, 2005), whether fair values of debt securities are pro-cyclical crucially depends
on how the three components of interest rates behave over the business cycle. In
recessions, both term spread and default spread tend to increase (Fama and French, 1989;
Chen, 2010), but the risk-fee rate rf often decreases as monetary policy is shifted to
boost the weak economy. Thus, the sum of the three (i.e., interest rates) may actually
decrease, causing an increase in the fair values of debt securities.
25
In addition, default
spread may decrease during the late phases of recessions, as seen in the recent recession,
which ultimately works to push interest rates down and fair values up. Conversely, during
some expansionary periods, although both term spread and default spread tend to
decrease (Fama and French, 1989; Chen, 2010), rf may increase as monetary policy
changes (for example, during 2004Q1-2007Q1); therefore, the sum of the three may
actually increase, causing fair values of debt securities to fall. Altogether, interest rates
may increase (decrease) during some expansionary (recessionary) periods, resulting in a
decrease (increase) in fair values of debt securities in those periods. Therefore, fair values
of debt securities may not move pro-cyclically with the aggregate economic activity. The
same analysis applies to fair values of loans as loans are also fixed-income instruments.
26
25
The impact of a decreasing interest rate on the fair value of mortgage-backed securities is not as
straightforward. Decreasing interest rates lead to an increase in prepayment risk because mortgage
borrowers may be induced to pay off the loan earlier in order to refinance, which would have a negative
impact on fair value of the mortgage-backed securities.
26
Ideally, banks can use derivatives to hedge interest rate risks so that changes in interest rate will not
affect their financial conditions. However, more than 80% of insured commercial banks in the U.S. do not
use derivatives to hedge interest rate risk or credit risks, according to the bank regulator the Office of the
Comptroller of the Currency (1994-2011). This is consistent with findings in Minton, Stulz, and
20
Therefore, theoretically, there is no guarantee that fair values of bank assets move pro-
cyclically with the aggregate economic activity.
Ample anecdotal evidence exists that changes in interest rates have substantial
impact on the fair values of debt securities held by banks. As the Federal Reserve
continuously raised interest rates during 2005, the commercial banking sector incurred
unrealized losses of -$13.4 billion, losses that many banks explicitly attributed to the
increase in interest rates. For instance, Wachovia, which posted more than -$1.4 billion of
unrealized losses on its debt securities that year, stated in its 10-K that “the gross
unrealized losses at December 31, 2005 and 2004, were primarily caused by interest rate
changes.” Likewise, International Bancshares Corporation, which incurred -$69 million
of unrealized losses on its mortgage-backed securities in 2005, stated in its annual report
that “the unrealized losses on investments in mortgage-backed securities [were] caused
by changes in market interest rates. … The decrease in fair value [was] due to market
interest rates and not other factors.”
Given the above discussion, whether the fair values of bank assets are pro-cyclical
is an empirical question.
3.3 Main Hypothesis: Whether Fair Value Accounting Exacerbates the
Pro-cyclicality of Lending
Bank lending is intrinsically pro-cyclical (Fisher, 1933; Kiyotaki and Moore,
1997; Asea and Blomberg, 1998; Lown and Morgan, 2006), a phenomenon confirmed in
Williamson (2009) that banks make very limited use of credit derivatives to hedge credit risk associated
with loans.
21
Figure 2 below. Whether fair value accounting exacerbates this pro-cyclicality is the
central research question in this study. As discussed above, for fair value accounting to
increase the pro-cyclicality of lending, two necessary conditions must be met. I have
hypothesized that the first necessary condition (i.e., changes in fair values affect
subsequent lending ex ante) is met; however, it is not clear ex ante whether the second
necessary condition (i.e., the fair values of bank assets are pro-cyclical) is met. Thus,
whether fair value accounting exacerbates the pro-cyclicality of actual lending is an
empirical question that only data can answer. I move to this empirical analysis next.
Figure 2. Intrinsic Pro-cyclicality of Bank Lending
-10
-5
0
5
10
15
1995 1997 1999 2001 2003 2005 2007 2009
Loan percent change from year ago (%)
Figure 2 shows the percent change of total loans and leases outstanding at all U.S. commercial
banks. Data are from St. Louis Fed.
22
Chapter 4: Sample and Data
The sample consists of all U.S. bank holding companies with total assets of more
than $150 million over the 1995-2005 period, and those with total assets of more than
$500 million over the 2006-2010 period.
27
Financial data are collected from quarterly FR
Y-9C reports, supplemented by Compustat data when necessary.
28
Risk-free interest rate
data are from CRSP. Term spread, default spread, and real GDP growth data are from the
St. Louis Fed. Data are collected quarterly for the sample period 1995-2010. The sample
begins in 1995 for two reasons. First, 1994 is the first year in which fair value accounting
affects banks’ balance sheets, since SFAS 115 mandates that trading assets and available-
for-sale securities be reported at fair value on the balance sheet (with unrealized gains
and losses recorded in accumulated other comprehensive income) for fiscal years
beginning after December 15, 1993; prior to SFAS 115, fair values of assets are not
reported on the balance sheet. Second, the analysis requires that fair value data lead bank
loan data by one year, thus the sample begins in 1995. To mitigate potential effects of
major mergers and acquisitions, I follow the literature to exclude bank-quarters with total
asset growth during the quarter exceeding 10% (see, e.g., Cornett et al., 2011). The final
main sample consists of 69,578 bank-quarter observations comprising 2,956 unique
banks over the 1995-2010 period.
27
According to the Bank Holding Company Act, bank holding companies with total assets exceeding
certain asset-size threshold are required to file quarterly FR Y-9C to the Federal Reserve. The filing
threshold is $150 million of total assets prior to 2006, and $500 million of total assets from 2006 onward.
28
The quarterly FR Y-9C report contains a consolidated balance sheet, a consolidated income statement,
and detailed supporting schedules, including a schedule of off balance-sheet items. The data are available
from Federal Reserve Bank of Chicago website from 1986Q3 onward.
23
Chapter 5: Empirical Analyses and Results
5.1 Does Fair Value Accounting Exacerbate the Pro-cyclicality of Lending?
5.1.1 Research Design: Whether Fair Value Accounting Exacerbates the Pro-
cyclicality of Lending
In this section I examine the central question of this paper: whether fair value
accounting exacerbates the pro-cyclicality of bank lending. To answer this research
question, I exploit the cross-sectional variation in individual banks’ exposure to fair value
accounting, and examine whether banks with greater fair value accounting exposure
exhibit more pro-cyclical lending behavior relative to banks with less fair value
accounting exposure over the business cycle.
29
My measure of individual banks’
exposure to fair value accounting is the fraction of total assets carried at fair value on the
balance sheet (i.e., fair-valued assets divided by total assets). Variants of this measure
have been used to measure banks’ exposure to fair value accounting (e.g., Nissim and
Penman, 2007; SEC, 2008; Khan, 2010). The measure used by Nissim and Penman
(2007) and Khan (2010), however, includes both assets and liabilities recognized at fair
value on the balance sheet and those disclosed at fair value in the footnote. Since fair
values recognized on the balance sheet affect bank examiners’ assessment of banks’
capital adequacy (Office of the Comptroller of the Currency, 1994) and hence likely
29
Ideally one would use a difference-in-difference design that exploits both time-series and cross-sectional
variation in individual banks’ exposure to fair value accounting. However, difference-in-difference design
is not suitable for this study because important confounding events happened during my long sample period
of 1995-2010. In particular, Basal II risk-based capital requirements started to be fully implemented in
1993Q3, which happened around the same time as the introduction of fair value recognition rule SFAS 115
in late 1993. Many policymakers and researchers suspect that the Basal II risk-based capital requirements
may have had pro-cyclicality impact on bank lending. It is therefore difficult to attribute the over-time
increase in bank lending pro-cyclicality, if there is any, to the introduction of fair value accounting.
Accordingly, I focus on examining whether the cross-sectional difference in banks’ fair value accounting
exposure is associated with cross-sectional variation in the pro-cyclicality of their lending behavior.
24
affect banks’ risk taking, while fair values disclosed in the footnote are less likely to have
a similar effect, my fair value accounting exposure measure only includes assets
recognized at fair value on the balance sheet and excludes those disclosed at fair value in
the footnote (which mainly comprise held-to-maturity securities and loans).
I define expansionary and recessionary periods based on National Bureau of
Economic Research (NBER) business cycle dates, a widely used business cycle indicator.
There are two recessions during my sample period (1995-2010), March 2001-November
2001 and December 2007-June 2009, which gives me two recessionary periods, 2001Q2-
2001Q4 and 2008Q1-2009Q2.
30
Quarters other than these recession quarters are
classified as expansion quarters. My measure of lending behavior is quarterly growth in
total outstanding loans.
31
Variants of this measure are frequently used in the literature to
capture banks’ lending behavior (e.g., Beatty and Liao, 2011; Berger and Udell, 2004;
Cornett et al., 2011). The main regression models are panel regressions that controls for
quarter fixed effects. Quarter fixed effects control for time-specific factors that affect
lending but are invariant across banks, making it unnecessary to control for time-varying
macroeconomic factors that affect bank lending, such as unemployment rates and risk-
free interest rate. All standard errors are clustered at the firm level. Except for the dummy
30
I define a quarter as a recessionary quarter when at least two months during the quarter are in a recession.
31
Outstanding loans are the equilibrium outcome of supply of loans and demand for loans. However, the
focus of my paper is on examining whether changes in fair values of bank assets affect banks’ ability to
supply credit. Thus, I need to control for the demand effect. My solution is to control for the time fixed
effects and, thus, hold constant across banks the creditworthiness of borrowers and their demand for loans,
which allows me to focus on how the cross-sectional variation in fair value accounting exposure affects
banks’ ability to supply credit.
25
variable, all other variables are winsorized at the top and bottom 99% within each quarter.
The model is given as follows:
) 1 ( /
Re Re *
, 1 ,
2 1 , 2 1 , 1 0 ,
t i t
k
t i k
t t t i t i t i
Effects Fixed Firm Quarter Variable Control
cession cession FVA FVA Loan
ε α
α α α α
+ +
+ + + + = Δ
−
− −
∑
where i stands for bank, and t for quarter.
t i
Loan
,
Δ is quarterly loan growth measured as
1 , 1 , ,
/ ) (
− −
−
t i t i t i
Assets Loan Loan , where Loan
i,t
is total loans and leases outstanding at the
end of quarter t, Loan
it-1
is total loans and leases outstanding at the beginning of quarter t,
and Assets
i,t-1
is total assets at historical costs. FVA
i,t-1
is bank i’s exposure to fair value
accounting at the beginning of quarter t, measured as Fair-valued assets/Assets
i,t-1
. Fair-
valued assets are assets carried at fair value on the balance sheet, measured as the sum of
fair value of trading assets and fair value of available-for-sale securities during 1995-
2007, and the sum of fair value of trading assets, fair value of available-for-sale securities,
and fair value of other assets elected to use fair value accounting under FAS 159 during
2008-2010.
32
Recession
t
is a business cycle indicator based on NBER business cycle
dates; it takes the value of 1 for recessionary quarters 2001Q2-2001Q4 and 2008Q1-
2009Q2, and 0 for expansionary quarters (remaining quarters over the 1995-2010 period).
Significantly positive
1
α and significantly negative
2 1
α α + would suggest that fair value
32
Under SFAS 159, banks can elect to measure essentially any financial assets and liabilities at fair value
in their financial statements for fiscal years beginning after November 15, 2007. This is the so-called “fair
value option”. The option is available at the inception of a financial instrument on an instrument-by-
instrument basis and is irrevocable. Unrealized gains (losses) for the assets and liabilities elected to use the
fair value option are included in earnings.
26
accounting exacerbates lending pro-cyclicality, whereas insignificant
1
α and
2 1
α α + would suggest that fair value accounting does not increase lending pro-cyclicality.
The panel regression above controls for banks’ business models and other factors
shown by the literature to affect banks’ lending behavior:
• Banks’ business models: banks with securities-oriented business models likely
exhibit different lending behavior compared to banks with lending-oriented
business models. I control for business models in two ways: (1) controlling for
banks’ asset structures measured as the fraction of total assets in investment
securities – Historical Costs of Securities/Assets
i,t-1
, with a larger number
proxying for a more securities-oriented business model, and (2) controlling for
firm fixed effects.
• Quality of existing loan portfolios: Poor quality of existing loan portfolios
predicts high future credit losses, which erodes banks’ capital positions and
directly diminishes banks’ lending capacity (Kwan, 2010). Accordingly, I control
for the ratio of non-performing loans to total loans. Non-performing loans are
loans 90 days or more past due plus loans no longer accruing interest.
• Profitability: Profitability affects banks’ capital and in turn their abilities to
support and fund new lending (Berger and Udell, 2004). Banks typically engage
in two kinds of activities: investing in securities and originating loans. Prior to
2008, unrealized gains (losses) come exclusively from securities business.
33
To
33
For fiscal years starting after November 15, 2007, unrealized gains (losses) also come from other assets
such as loans elected to use fair value accounting under SFAS 159.
27
mitigate concerns that the observed impact of fair value accounting exposure on
lending may be driven by realized gains on securities, I separately control for
profitability of banks’ securities businesses, Securities ROE
i,t-1
,
measured as
realized gains (losses) on available-for-sale and held-to-maturity securities during
the prior quarter scaled by average total equity capital. I also control for the
profitability of loan businesses, Non-securities ROE
i,t-1
, measured as total return
on equity (ROE) minus Securities ROE
i,t-1
.
• Reliance on core deposits as a stable source of financing: Banks that rely more on
core deposits as opposed to short-term financing may exhibit different lending
behavior than banks that rely less on core deposits (Ivashina and Scharfstein,
2010; Cornett et al., 2011).
34
Accordingly, I control for banks’ reliance on core
deposits using Core Deposits/Assets
i,t-1
, measured as core deposits scaled by total
assets at historical cost. I also control for Recesssion
t
* Core Deposits/Assets
i,t-1
,
which is the interaction between the core deposit ratio and the recession dummy.
The interaction term captures the differential impact of core deposits on lending in
recessions relative to expansions, as prior research shows that banks that rely
more on core deposits as a stable source of funding cut lending less during the
recent financial crisis relative to other banks (Ivashina and Scharfstein, 2011;
Cornett et al., 2011).
• Leverage: Due to regulation, leverage directly affects banks’ ability to originate
new loans (Berger and Udell, 2004). Leverage also signals the amount of risk that
34
Core deposits are deposits under $100,000 plus all transactions deposits.
28
the bank has undertaken and, thereby, directly affects how much additional risk
the bank can take in making new loans. Accordingly, I control for Leverage
i,t-1
,
which is measured as total assets divided by total equity capital.
35
• Firm size: since firm size affects lending behavior (see, for e.g., Berger et al.
2005), I control for firm size using the natural log of total assets.
• Quarter fixed effects: Controlling for quarter fixed effects helps control for time-
varying demand for loans. This control is important because my measure of
lending behavior is the growth rate of outstanding loans; outstanding loans,
however, is the equilibrium outcome of demand for loans and supply of loans;
through controlling for quarter fixed effects, I keep the demand for loans constant
across banks, which enables me to focus on examining the impact of fair value
accounting on banks’ ability to supply loan.
I also conduct subsample analysis based on size and leverage partitions. I partition
on size because large banks likely exhibit different lending behavior than small banks.
Large banks have different lending practices compared to small banks (Berger et al.,
2005); large banks are more likely to receive federal assistance during crises than small
banks (the so-called “too big to fail” doctrine).
36
Thus, I partition the sample into large
banks (total assets of more than $1 billion) and small banks (total assets of $1 billion or
less). I also partition on leverage because high-leverage banks that are closer to violating
35
I do not use regulatory capital ratios and risk-based leverage ratios such as Tier 1 risk-based capital ratio
or Tier 1 leverage ratio because data for these measures are not available until 2001.
36
For analyses of the benefits and costs of the large-scale government bailout of the ten largest U.S. banks
during the recent financial crisis, see Veronesi and Zingales (2010).
29
the minimum capital thresholds are likely affected by changes in the fair values of their
assets to a greater extent than banks that maintain a sizable buffer in their capital
structure. High-leverage banks are thus more likely to be affected by fair value
accounting in a pro-cyclical manner relative to low-leverage banks. In the partition, high-
leverage (low-leverage) banks are those banks with leverage in the top (bottom) tercile.
5.1.2 Descriptive Statistics of Loan Growth and Fair Value Accounting Exposure
Table 1 provides summary statistics of the sample banks’ characteristics. Mean
quarterly loan growth indicates that for an average bank, new lending in an average
quarter is equal to 1.3% of total assets. The mean (median) fair value accounting
exposure as captured by FVA indicates that 19% (17%) of an average (median) bank’s
total assets are measured at fair value on the balance sheet. With respect to the
persistence of banks’ fair value accounting exposure, the AR(1), AR(2), AR(3), and
AR(4) are 0.97, 0.94, 0.91, and 0.89, respectively, suggesting that individual banks’ fair
value accounting exposure is very persistent over time.
Table 2 presents Pearson correlations. The correlation between loan growth and
fair value accounting exposure is significantly negative. Such negative correlation likely
reflects the significantly negative correlation between loan growth and banks’ asset
structures (Historical Costs of Securities/Assets). This further suggests that my fair value
accounting exposure measure is closely linked to banks’ asset structures and, therefore, it
is very important to control for banks’ asset structures when examining the relationship
between loan growth and fair value accounting exposure.
30
Table 1. Summary Statistics
VARIABLES N Mean Minimum p10 p25 Median p75 p90 Maximum Stddev
ΔLoan 69,578 0.013 -0.073 -0.014 -0.001 0.012 0.026 0.041 0.085 0.023
FVA 69,578 0.19 0.00 0.06 0.11 0.18 0.26 0.35 0.89 0.12
UGL_P4 (×100) 69,578 0.009 -1.579 -0.315 -0.113 0.003 0.134 0.335 1.696 0.312
UGL_P2 (×100) 69,578 0.009 -0.983 -0.228 -0.087 0.003 0.098 0.245 1.164 0.219
UGL_P1 (×100) 69,578 0.005 -1.021 -0.166 -0.062 0.000 0.076 0.187 0.748 0.168
UGL_CUM (×100) 69,578 0.047 -2.139 -0.256 -0.081 0.018 0.166 0.385 2.052 0.313
Pre_UGL_P4 (×100) 16,043 0.041 -7.625 -0.702 -0.226 0.062 0.340 0.680 13.479 0.878
Pre_UGL_P2 (×100) 16,043 0.079 -5.172 -0.415 -0.160 0.048 0.296 0.578 14.863 0.685
Pre_UGL_P1 (×100) 16,043 0.061 -3.823 -0.303 -0.114 0.034 0.217 0.483 2.886 0.441
Pre_UGL_CUM (×100) 16,043 -0.029 -8.220 -0.605 -0.228 0.008 0.245 0.554 2.128 0.622
UGL_EP4 (×100) 69,578 0.177 -32.362 -3.612 -1.277 0.037 1.528 3.955 40.593 3.959
UGL_EP2 (×100) 69,578 0.141 -21.344 -2.619 -0.988 0.029 1.123 2.878 27.508 2.793
UGL_EP1 (×100) 69,578 0.084 -18.496 -1.923 -0.702 0.005 0.859 2.212 19.112 2.134
UGL_ECUM (×100) 69,578 0.466 -38.342 -3.075 -0.931 0.206 1.888 4.379 27.842 3.828
BVSR 69,578 0.239 0.004 0.096 0.153 0.224 0.310 0.404 0.655 0.120
PLOANR 69,578 0.011 0.000 0.001 0.003 0.007 0.013 0.024 0.183 0.015
ROEO 69,578 0.027 -1.012 0.010 0.021 0.030 0.039 0.049 0.155 0.039
ROES 69,578 0.001 -0.223 0.000 0.000 0.000 0.000 0.003 0.061 0.006
CDR 69,578 0.68 0.15 0.54 0.62 0.69 0.76 0.80 0.87 0.11
Leverage 69,578 12.01 4.63 8.12 9.66 11.41 13.59 16.35 84.45 4.03
Size
69,578
13.30 11.48 12.17 12.42 12.98 13.76 14.88 18.98 1.23
Table 1 provides summary statistics for all firm-level variables based on all bank-quarter observations.
UGL_P4, UGL_P2, UGL_P1, and UGL_CUML are variants of Unrealized Gains (Losses)/Assets
t-1
based
on unrealized gains (losses) recognized during the past four quarters, the past two quarters, the past one
quarter, and cumulative as of the beginning of the quarter, respectively, during the fair-value-accounting
period of 1995-2010. Pre-UGL_P4, Pre-UGL_P2, Pre-UGL_P1, and Pre-UGL_CUML are the counterpart
of UGL_P4, UGL_P2, UGL_P1, and UGL_CUML, respectively, during the pre-fair-value-accounting
period of 1988-1991. UGL_EP4, UGL_EP2, UGL_EP1, and UGL_ECUM are defined in a similar fashion
as UGL_P4, UGL_P2, UGL_P1, and UGL_CUM, except that their deflators are lagged total equity capital.
BVSR refers to Historical Costs of Securities/Assets, PLOANR refers to Nonperforming Loans/Loans,
ROEO refers to Non-securities ROE, ROES refers to Securities ROE, and CDR refers to Core
Deposits/Assets. All variables are calculated as defined in the Appendix.
31
Table 2. Pearson Correlations
VARIABLES FVA UGL_P4 UGL_P2 UGL_P1 UGL_CUM BVSR PLOANR ROEO ROES CDR Leverage Size Recession
ΔLoan -0.074 -0.085 -0.092 -0.053 -0.064 -0.071 -0.315 0.225 -0.021 0.022 0.001 -0.067 -0.047
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.835) (0.001) (0.001)
FVA -0.008 -0.016 -0.008 0.104 0.709 -0.016 0.007 0.064 -0.055 -0.085 -0.005 -0.031
(0.033) (0.001) (0.038) (0.001) (0.001) (0.001) (0.049) (0.001) (0.001) (0.001) (0.225) (0.001)
UGL_P4 0.688 0.535 0.533 -0.019 0.065 -0.034 0.040 0.016 -0.059 0.018 0.275
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
UGL_P2 0.639 0.354 -0.009 0.057 -0.036 -0.013 0.010 -0.037 0.016 0.201
(0.001) (0.001) (0.016) (0.001) (0.001) (0.001) (0.006) (0.001) (0.001) (0.001)
UGL_P1 0.264 0.002 0.035 -0.008 -0.066 0.003 -0.031 0.011 0.105
(0.001) (0.593) (0.001) (0.038) (0.001) (0.377) (0.001) (0.005) (0.001)
UGL_CUM 0.044 0.004 0.006 0.164 0.045 -0.203 -0.032 0.052
(0.001) (0.254) (0.114) (0.001) (0.001) (0.001) (0.001) (0.001)
BVSR -0.098 0.059 0.032 0.067 -0.170 -0.078 -0.084
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
PLOANR -0.525 0.055 -0.091 0.236 0.104 0.055
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
ROEO -0.083 0.062 -0.318 -0.044 -0.061
(0.001) (0.001) (0.001) (0.001) (0.001)
ROES -0.019 0.038 0.008 -0.027
(0.001) (0.001) (0.032) (0.001)
CDR -0.064 -0.412 -0.076
(0.001) (0.001) (0.001)
Leverage 0.066 0.001
(0.001) (0.792)
Size 0.068
(0.001)
Table 2 reports Pearson correlation coefficients for variables used in the analysis. UGL_P4, UGL_P2,
UGL_P1, and UGL_CUML are variants of Unrealized Gains (Losses)/Assets
t-1
based on unrealized gains
(losses) recognized during the past four quarters, the past two quarters, the past one quarter, and cumulative
as of the beginning of the quarter, respectively. BVSR refers to Historical Costs of Securities/Asset,
PLOANR refers to Nonperforming Loans/Loans, ROEO refers to Non-securities ROE, ROES refers to
Securities ROE, and CDR refers to Core Deposits/Assets. All variables are calculated as defined in the
Appendix. Statistical significance is based on two-sided t-tests and is indicated as follows: *** p<0.01, **
p<0.05, and * p<0.1.
32
5.1.3 Empirical Results on Whether Fair Value Accounting Exacerbates Pro-
cyclicality of Lending
Table 3 and 4 reports results on the relationship between fair value accounting
exposure and actual lending – specifically, whether banks with greater fair value
accounting exposure exhibit more pro-cyclical lending behavior relative to banks with
less fair value accounting exposure. The analyses in Table 3 use asset structures to
control for banks’ business models, whereas those in Table 4 use both asset structures
and firm fixed effects to control for business models.
Panel A, Table 3 presents results for the full sample and the size subsamples. For
the full sample, the coefficient on FVA
i,t-1
is 0.002 and insignificant, indicating that fair
value accounting exposure has a positive but insignificant impact on lending during
expansions; the sum of the coefficients on FVA
i,t-1
and FVA
i,t-1
*Recession
t
is -0.001 and
insignificant, suggesting that fair value accounting exposure has a negative but
insignificant impact on lending during recessions. For small banks, fair value accounting
has a significantly positive impact on lending in expansions (the coefficient on FVA
it-1
is
0.003 and statistically significant) but no statistically significant impact on lending during
recessions (the sum of the coefficients on FVA
i,t-1
and FVA
i,t-1
*Recession
t
is 0.004 and
insignificant). For large banks, fair value accounting exposure has no significant impact
on lending during either expansions (the FVA
i,t
coefficient is -0.0003 and insignificant)
or recessions (the sum of the coefficients on FVA
i,t-1
and FVA
i,t-1
*Recession
t
is -0.011
and insignificant). The coefficient on the Recession
t
dummy is -0.02 and significant,
indicating that bank lending is indeed pro-cyclical: banks curtail lending equivalent to 2%
of total assets during a recessionary quarter relative to an expansionary quarter.
33
Table 3. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending:
Not Controlling for Firm Fixed Effects
) 1 (
Re Re *
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t
k
t i k
t t t i t i t i
Effects Fixed Quarter Variable Control
cession cession FVA FVA Loan
ε α
α α α α
+ + +
+ + + = Δ
−
− −
∑
Panel A: Full sample and size subsamples
Small banks Large banks
(Total assets ≤ $1B) (Total assets > $1B)
Intercept 0.04 0.01 0.06
(12.96)*** (2.14)** (10.04)***
FVA i,t-1 0.002 0.003 -0.0002
(1.29) (1.95)** (-0.06)
FVA i,t-1 *Recession t -0.003 0.001 -0.010
(-0.80) (0.25) (-1.66)*
Recession t -0.02 -0.01 -0.02
(-5.60)*** (-3.34)*** (-3.80)***
Historical Costs of Securities/Assets i,t-1 -0.03 -0.03 -0.02
(-14.72)*** (-15.5)*** (-4.36)***
Recession t * Historical Costs of Securities/Assets i,t-1 0.01 0.01 0.02
(3.26)*** (2.05)** (2.37)**
Nonperforming Loan/Loan i,t-1 -0.33 -0.36 -0.26
(-24.03)*** (-23.17)*** (-10.36)***
Non_securities ROE i,t-1 0.05 0.04 0.05
(11.51)*** (7.69)*** (9.48)***
Securities ROE i,t-1 0.01 0.04 -0.02
(0.59) (1.27) (-0.8)
Core Deposits/Assets i,t-1 -0.01 -0.01 -0.01
(-4.96)*** (-3.86)*** (-3.72)***
Recession * Core Deposits/Assets i,t-1 0.010 0.007 0.004
(2.92)*** (1.47) (0.79)
Leverage i,t-1 0.0004 0.0004 0.0003
(10.49)*** (9.84)*** (3.17)***
Size i,t-1 -0.001 0.001 -0.002
(-4.41)*** (4.42)*** (-7.45)***
FVA i,t-1 + FVA i,t-1 *Recession t -0.001 0.004 -0.011
Clustered t-stat of FVA i,t-1 + FVA i,t-1 *Recession t -0.24 0.92 -1.52
Quarter fixed effects Yes Yes Yes
Standard errors clustered by firm Yes Yes Yes
Observations 69,578 52,890 16,688
Adjusted R-squared 0.18 0.16 0.25
Full sample VARIABLES
Panel A, Table 3 presents results of whether fair value accounting exacerbates lending pro-cyclicality for
the full sample, small-bank subsample, and large-bank subsample without controlling for firm fixed effect.
t-statistics are based on standard errors clustered at the firm level. Quarter fixed effects are included in the
model. All variables are winsorized at the 1% and 99% levels, and are calculated as defined in the
34
Appendix. Statistical significance is based on two-sided t-tests and is indicated as follows: *** p<0.01, **
p<0.05, and * p<0.1.
With respect to the control variables in the full sample, Panel A of Table 3 shows
that banks with a greater emphasis on securities businesses grow loans more slowly in
expansions, with a coefficient of -0.03 on Historical Costs of Securities/Assets
i,t-1
;
however, these banks curtail lending by less during recessions, as suggested by the
significant coefficient of 0.01 on Recession
t
*Historical Costs of Securities/Assets
i,t-1
.
These results indicate that asset structures as proxied by Historical Costs of
Securities/Assets
i,t-1
indeed influence banks’ lending behavior. Banks that rely more on
stable financing from core deposits appear to grow their loans more slowly in expansions,
as evidenced by the coefficient of -0.01 on Core Deposit/Assets
i,t-1
; however, during
recessions, these banks curtail lending by less relative to other banks, with a significantly
positive coefficient of 0.010 on Recession
t
*Core Deposit/Assets
i,t-1.
37
This finding is
consistent with the results in Cornett et al. (2011). Not surprisingly, banks with poor loan
quality grow their loans more slowly, as evidenced by the significant coefficient of -0.33
on Nonperformaning Loan/Loan
i,t-1
. Profitability of non-securities businesses (i.e., loan
origination) has a statistically significant and economically large impact on lending, as
indicated by the coefficient of 0.05 on Non_securities ROE
i,t-1
. In contrast, the
profitability of securities businesses (i.e., realized gains and losses on securities scaled by
average total equity) has little impact on lending.
37
However, the coefficient on Recession
t
*Core Deposit/Assets
it-1
is insignificant for both the small-bank
and the large-bank subsamples.
35
Table 3. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending:
Not Controlling for Firm Fixed Effects (Continued)
) 1 (
Re Re *
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t
k
t i k
t t t i t i t i
Effects Fixed Quarter Variable Control
cession cession FVA FVA Loan
ε α
α α α α
+ + +
+ + + = Δ
−
− −
∑
Panel B: Leverage subsamples
Low-leverage Banks High-leverage Banks
(Leverage at the bottom tercile) (Leverage at the top tercile)
Intercept 0.03 0.04
(7.08)*** (8.82)***
FVA i,t-1 0.003 0.001
(1.56) (0.43)
FVA i,t-1 *Recession t 0.003 -0.007
(0.49) (-0.85)
Recession t -0.01 -0.01
(-2.08)** (-2.95)***
Historical Costs of Securities/Assets i,t-1 -0.02 -0.02
(-10.25)*** (-7.41)***
Recession t * Historical Costs of Securities/Assets i,t-1 0.01 0.01
(1.93)** (1.38)
Nonperforming Loan/Loan i,t-1 -0.29 -0.34
(-14.36)*** (-15.79)***
Non_securities ROE i,t-1 0.06 0.03
(4.25)*** (6.09)***
Securities ROE i,t-1 0.07 0.00
(1.90)* (-0.05)
Core Deposits/Assets i,t-1 -0.01 -0.01
(-2.95)*** (-2.47)***
Recession * Core Deposits/Assets i,t-1 0.004 0.012
(0.55) (1.90)
Leverage i,t-1 0.0004 0.0002
(2.46)*** (3.01)***
Size i,t-1 -0.0003 -0.0010
(-1.39) (-4.1)***
FVA i,t-1 + FVA i,t-1 *Recession t 0.006 -0.006
Clustered t-stat of FVA i,t-1 + FVA i,t-1 *Recession t 1.02 -0.64
Quarter fixed effects Yes Yes
Standard errors clustered by firm Yes Yes
Observations 23,172 23,195
Adjusted R-squared 0.14 0.22
VARIABLES
Panel B, Table 3 presents results of whether fair value accounting exacerbates lending pro-cyclicality for
low-leverage banks and high-leverage banks. t-statistics are based on standard errors clustered at the firm
level. Quarter fixed effects are included in the model. All variables are winsorized at the 1% and 99%
levels, and are calculated as defined in the Appendix. Statistical significance is based on two-sided t-tests
and is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
36
Panel B, Table 3 presents results for the low-leverage (leverage in the bottom
tercile) and the high-leverage (leverage in the top tercile) subsamples. For both the low-
and the high-leverage banks, the coefficient on FVA
i,t-1
is positive but insignificant,
indicating that fair value accounting exposure does not have significant impact on lending
in expansions. Furthermore, the sum of the coefficients on FVA
it-1
and FVA
i,t-1
*Recession
t
is insignificant, suggesting that fair value accounting exposure does not have a significant
impact on lending in recessions either.
The analyses in Table 3 only use asset structures to control for banks’ business
models. The analyses in Table 4 add firm fixed effects to further control for business
models. The results in Table 4 confirm the finding that fair value accounting does not
appear to exacerbate lending pro-cyclicality. Panel A, Table 4 presents results for the full
sample and the size subsamples. For the full sample, small-bank subsample, and the
large-bank subsample, the coefficient on FVA
i,t-1
is 0.008, 0.005, 0.014, respectively, and
significant, indicating that fair value accounting exposure has a significantly positive
impact on lending during expansions. However, untabulated results based on finer size
partitions suggest that the positive impact of fair value accounting on lending during
economic expansions is mainly restricted to large banks in the top 89 percentile of size
distribution, while for the rest of the banks (which constitutes 88% of all sample banks),
fair value accounting exposure has no significant positive effect on lending during
economic expansions. For the full sample as well as both size subsamples, the sum of the
coefficients on FVA
i,t-1
and FVA
i,t-1
*Recession
t
is insignificant, suggesting that fair value
accounting exposure has no significant impact on lending during recessions.
37
Table 4. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending
after Controlling for Firm Fixed Effects
) ' 1 (
Re Re *
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t t
k
t i k
t t t i t i t i
Effects Fixed Firm Effects Fixed Quarter Variable Control
cession cession FVA FVA Loan
ε α
α α α α
+ + + +
+ + + = Δ
−
− −
∑
Panel A: Full sample and size subsamples
Small banks Large banks
(Total assets = $1B) (Total assets > $1B)
Intercept -0.01 0.01 0.01
(-11.24)*** (4.64)*** (11.27)***
FVA
i,t-1
0.008 0.005 0.014
(2.39)** (1.72)* (1.96)**
FVA
i,t-1
*Recession
t
-0.006 -0.005 -0.010
(-1.65)* (-1.23) (-2.16)***
Recession
t
-0.01 -0.01 -0.01
(-2.62)*** (-1.68)* (-1.98)**
Historical Costs of Securities/Assets
i,t-1
0.01 0.02 -0.01
(3.09)*** (5.25)*** (-0.96)
Recession
t
* Historical Costs of Securities/Assets
i,t-1
0.02 0.02 0.02
(2.82)*** (2.56)*** (3.13)***
Nonperforming Loan/Loan
i,t-1
-0.38 -0.40 -0.34
(-15.13)*** (-15.53)*** (-11.72)***
Non_securities ROE
i,t-1
0.02 0.01 0.03
(2.48)*** (0.97) (3.94)***
Securities ROE
i,t-1
-0.02 -0.01 -0.03
(-1.17) (-0.28) (-1.71)*
Core Deposits/Assets
i,t-1
0.007 0.008 0.001
(1.74)* (1.80)* (0.26)
Recession * Core Deposits/Assets
i,t-1
0.009 0.004 0.007
(1.64)* (0.77) (1.03)
Leverage
i,t-1
-0.0001 -0.0001 -0.0001
(-1.69)* (-0.8) (-1.81)*
Size
i,t-1
-0.01 -0.01 -0.01
(-9.58)*** (-8.25)*** (-7.60)***
FVA
i,t-1
+ FVA
i,t-1
*Recession
t
0.0024 0.0004 0.0040
Clustered t-stat of FVA
i,t-1
+ FVA
i,t-1
*Recession
t
0.52 0.08 0.56
Firm fixed effects Yes Yes Yes
Quarter fixed effects Yes Yes Yes
Standard errors clustered by firm and quarter Yes Yes Yes
Observations 69,578 52,890 16,688
R-squared 0.19 0.15 0.29
VARIABLES Full sample
Panel A, Table 4 presents results of whether fair value accounting exacerbates lending pro-cyclicality for
the full sample, small-bank subsample, and large-bank subsample after controlling for firm fixed effects. t-
statistics are based on standard errors clustered at the firm level. Quarter fixed effects are also included in
the model. All variables are winsorized at the 1% and 99% levels, and are calculated as defined in the
Appendix. Statistical significance is based on two-sided t-tests and is indicated as follows: *** p<0.01, **
p<0.05, and * p<0.1.
38
Panel B, Table 4 below presents results for the low-leverage (leverage in the
bottom tercile) and the high-leverage (leverage in the top tercile) subsamples. For the
low-leverage banks, the coefficient on FVA
i,t-1
is positive but insignificant, indicating that
fair value accounting exposure does not have significant impact on lending during
expansions. Furthermore, the sum of the coefficients on FVA
it-1
and FVA
i,t-1
*Recession
t
is
also insignificant, suggesting that fair value accounting exposure has no significant effect
on lending during recessions either. For the high-leverage banks, the coefficient on
FVA
i,t-1
is 0.014 and significant, indicating that fair value accounting exposure has
significantly positive effect on lending in expansions; however, the sum of the
coefficients on FVA
it-1
and FVA
i,t-1
*Recession
t
is insignificant, suggesting that fair value
accounting exposure has no significant effect on lending during recessions either.
39
Table 4. Whether Fair Value Accounting Exacerbates Pro-cyclicality in Lending
After Controlling for Firm Fixed Effects (Continued)
) 6 (
Re Re *
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t t
k
t i k
t t t i t i t i
Effects Fixed Firm Effects Fixed Quarter Variable Control
cession cession FVA FVA Loan
ε α
α α α α
+ + + +
+ + + = Δ
−
− −
∑
Panel B: Leverage subsamples (Dependent variable: quarterly loan growth
t i
Loan
,
Δ )
Low-leverage Banks High-leverage Banks
(Leverage at the bottom tercile) (Leverage at the top tercile)
Intercept -0.01 0.01
(-14.95)*** (3.15)***
FVA
i,t-1
0.005 0.014
(1.34) (2.59)***
FVA
i,t-1
*Recession
t
0.002 -0.011
(0.38) (-2.23)**
Recession
t
-0.01 -0.01
(-0.87) (-2.46)***
Historical Costs of Securities/Assets
i,t-1
0.01 0.01
(3.04)*** (1.45)
Recession
t
* Historical Costs of Securities/Assets
i,t-1
0.01 0.02
(2.39)*** (1.76)*
Nonperforming Loan/Loan
i,t-1
-0.35 -0.37
(-12.15)*** (-11.15)***
Non_securities ROE
i,t-1
0.05 0.01
(3.20)*** (1.63)*
Securities ROE
i,t-1
0.02 -0.03
(0.57) (-1.33)
Core Deposits/Assets
i,t-1
0.012 0.003
(2.29)*** (0.61)
Recession * Core Deposits/Assets
i,t-1
0.004 0.006
(0.48) (1.10)
Leverage
i,t-1
-0.00031 -0.00002
(-1.95)** (-0.20)
Size
i,t-1
-0.01 -0.01
(-6.00)*** (-9.00)***
FVA
i,t-1
+ FVA
i,t-1
*Recession
t
0.007 0.003
Clustered t-stat of FVA
i,t-1
+ FVA
i,t-1
*Recession
t
1.29 0.04
Firm fixed effects Yes Yes
Quarter fixed effects Yes Yes
Standard errors clustered by firm and quarter Yes Yes
Observations 23,172 23,195
R-squared 0.16 0.23
VARIABLES
Panel B, Table 4 presents results of whether fair value accounting exacerbates lending pro-cyclicality for
low-leverage banks and high-leverage banks after controlling for firm fixed effects. t-statistics are based on
standard errors clustered at the firm level. Quarter fixed effects are also included in the model. All variables
are winsorized at the 1% and 99% levels, and are calculated as defined in the Appendix. Statistical
significance is based on two-sided t-tests and is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
40
Overall, Table 3 and 4 provide evidence that fair value accounting exposure does
not amplify the pro-cyclicality of bank lending over the past two business cycles during
1995-2010. In other words, banks with greater fair value accounting exposure do not
grow loans more rapidly during economic upturns or curtail lending more substantially
during downturns relative to banks with less fair value accounting exposure.
This counterintuitive finding begs the question: why doesn’t fair value accounting
increase the pro-cyclicality of bank lending? As previously discussed, two necessary
conditions must be met for fair value accounting to increase lending pro-cyclicality: (1)
increases (decreases) in fair values, i.e., unrealized gains (losses), through their impact on
the strength of banks’ balance sheets, induce banks to increase (reduce) lending; (2) fair
values of banks’ assets move, on average, in a pro-cyclical manner (see Figure 2).
Violation of either necessary condition makes it unlikely that fair value accounting will
exacerbate lending pro-cyclicality. In subsections 5.2 and 5.3 below, I investigate
whether, in fact, either of these two necessary conditions is violated.
5.2 Do Unrealized Gains (Losses) Affect Lending?
5.2.1 Research Design: Whether Unrealized Gains (Losses) Affect Lending
To test whether unrealized gains (losses) recognized on the balance sheet affect
bank lending, I examine whether unrealized gains (losses) recognized during the past one
to four quarter(s) affect lending in the current quarter. Note that a positive relationship
between unrealized gains (losses) and subsequent lending could arise simply because an
increase (decrease) in fair values improves (worsens) a bank’s financial position and
thereby induces more (less) lending, even if fair value accounting is not used to recognize
41
these unrealized gains (losses) on the balance sheets. To rule out this possibility, I use a
difference-in-difference design. In this design, I also examine the relationship between
unrealized gains (losses) and subsequent lending in a period when fair value accounting
rules have not come into effect (“pre-fair-value-accounting period”). If the relationship
between unrealized gains (losses) and subsequent lending is not significant in the pre-
fair-value-accounting period but becomes significant after fair value accounting rules
came into effect (“fair-value-accounting period”), then I can conclude that the significant
relationship between unrealized gains (losses) recognized on the balance sheet and bank
lending in the fair-value-accounting period results from the use of fair value accounting.
The panel regression model used in this analysis is given as follows:
) 2 ( /
* / ) ( / ) (
, 1 , 3
1 , 2 1 , 1 0 ,
t i t
k
t i k t
t t i t i t i
Effects Fixed Firm Quarter Variable Control Post
Post Assets Losses Gains Unrealized Assets Losses Gains Unrealized Loan
ε α α
α α α
+ + + +
+ + = Δ
−
− −
∑
where i stands for bank, and t for quarter.
t i
Loan
,
Δ is quarterly loan growth measured as
1 , 1 , ,
/ ) (
− −
−
t i t i t i
Assets Loan Loan .
t i
Assets Losses Gains Unrealized
,
/ ) ( is unrealized gains
(losses) during the past one to four quarter(s) deflated by total assets at historical cost,
with the definition slightly different for the pre-fair-value-accounting period of 1988-
1991 and fair-value-accounting period of 1995-2010.
38
For the fair-value-accounting
period, ) (Losses Gains Unrealized are unrealized gains (losses) recognized on the
38
I exclude 1992-1993 because this period does not qualify as a pure pre-fair-value-accounting period or a
pure fair-value-accounting period. SFAS No. 107, which is effective for fiscal years ending after December
15, 1992, mandates disclosure of fair value of financial instruments; thus, banks with fiscal years ending
after December 15, 1992 already disclose the fair values of their assets in 1992, so 1992-1993 does not
qualify as a pure pre-fair- value-accounting period. Additionally, 1992-1993 does not qualify as a pure fair-
value-accounting period as fair values do not affect balance sheet until fiscal years beginning after
December 15, 1993. I also exclude 1994 as fair value data are available since 1994 and the analysis
requires that fair value data lead bank loan data by one year.
42
balance sheets during the past one to four quarter(s) based on fair value accounting rules.
For the pre-fair-value-accounting period, ) (Losses Gains Unrealized are holding gains
(losses) for investment securities that occurred during the past one to four quarter(s). Post
is a dummy variable equal to 0 for the pre-fair-value-accounting period and 1 for the fair-
value-accounting period. Asset structures (and firm fixed effects) are included in the
model to control for the heterogeneity of business models. All control variables are the
same as those in Model (1), except that interaction terms with Recession
t
are not included.
5.2.2 Descriptive Statistics of Unrealized Gains (Losses)
As shown in Table 1, for unrealized gains (losses) recognized during the past four
quarters during the fair-value-accounting period, its mean value equals 0.009% of total
assets, and its 90
th
percentile (10
th
percentile) represents 0.335% (-0.315%) of total assets,
suggesting that the small mean value is partly driven by unrealized gains and losses from
banks at the extremes offsetting each other. The mean value of unrealized gains (losses)
is small also partly because total assets are used as the deflator. To see this, Table 1 also
reports summary statistics for unrealized gains (losses) using total equity as the deflator:
the mean unrealized gains (losses) recognized during the past four quarters equals 0.2%
of total equity (after gains and losses offset each other), with the 90
th
percentile (10
th
percentile) totaling 4.0% (-3.6%) of total equity. The summary statistics of cumulative
unrealized gains (losses) are similar to those of unrealized gains (losses) recognized
during the past one or two quarters. Further, the summary statistics for unrealized gains
(losses) for the pre-fair-value-accounting period are similar to those for the fair-value-
accounting period.
43
5.2.3 Empirical Results on Whether Unrealized Gains (Losses) Affect Subsequent
Lending
Table 5 below presents results on whether unrealized gains (losses) impact
subsequent lending without controlling for firm fixed effects. To ensure that the observed
impact is driven by accounting rather than by economic fundamentals, I compare the pre-
fair-value-accounting period (1988-1991) and the fair-value-accounting period (1995-
2010). Different columns represent unrealized gains (losses) that occurred at different
time horizons as the main independent variable. Columns 1 and 2 show that during the
pre-fair-value-accounting period, unrealized gains (losses) that occurred during the past
four or two quarters, respectively, have a minimal impact on current-quarter lending. In
contrast, during the fair-value-accounting period, for every one dollar of unrealized gains
(losses) recognized on the balance sheets during the past four or two quarters, banks
increase (reduce) lending by 25 or 32 cents, respectively, in the current quarter. Column 3
reports that every one dollar of unrealized gains (losses) that occurred during the past one
quarter is associated with 8 cents of increase (decrease) in current-quarter lending during
the pre-fair-value-accounting period, whereas the lending impact is three times as large at
25 cents during the fair-value-accounting period. Finally, column 4 shows that during the
pre-fair-value-accounting period, every one dollar of cumulative unrealized gains (losses)
is associated with about 6 cents of increased (decreased) lending in the current quarter,
while the lending impact is insignificant during the fair-value-accounting period. The
very small impact of cumulative unrealized gains (losses) on lending makes intuitive
sense: some of the cumulative unrealized gains (losses) may have occurred long time ago
and, thereby, is less likely to affect banks’ lending decisions during the current quarter.
44
Table 5. Whether Unrealized Gains (Losses) Affect Banking Lending: Not Controlling Firm Fixed Effects
) 2 (
* / ) ( / ) (
, 1 , 3
1 , 2 1 , 1 0 ,
t i t
k
t i k t
t t i t i t i
Effects Fixed Quarter Variable Control Post
Post Assets Losses Gains Unrealized Assets Losses Gains Unrealized Loan
ε α α
α α α
+ + + +
+ + = Δ
−
− −
∑
Table 5 presents results on whether unrealized gains (losses) affect banks’ subsequent lending during the pre-fair-value-accounting period of 1988-1991
versus the fair-value-accounting period of 1995-2010 without controlling for firm fixed effects. In columns 1, 2, 3, and 4, the main independent variable
unrealized gains (losses) are those occurred during the past four quarters, the past two quarters, the past one quarter, and cumulative as of the beginning
of the quarter, respectively. t-statistics based on standard errors clustered at the firm level are in parentheses. Quarter fixed effects are included in the
model. All non-dummy variables are winsorized at the 1% and 99% levels. Post is a dummy variable equal to 1 for the fair-value-accounting period and
0 for the pre-fair-value-accounting period. All other variables are calculated as defined in the Appendix. Statistical significance is based on two-sided t-
tests and is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
45
Table 5 (Continued)
(1) (2) (3) (4)
VARIABLES
Unrealized Gains
(Losses) Occurred during
Past Four Quarters
Unrealized Gains
(Losses) Occurred during
Past Two Quarters
Unrealized Gains
(Losses) Occurred
during Past One Quarter
Cumulative Unrealized
Gains (Losses)t-1 at the
Beginning of the Quarter
Intercept 0.02 0.02 0.02 0.02
(9.39)*** (9.41)*** (9.35)*** (9.34)***
Unrealized Gains (Losses)/Assets i,t-1 0.011 0.032 0.077 0.062
(0.54) (1.34) (2.04)** (1.97)**
Unrealized Gains (Losses)/Assets i,t-1 * Post 0.24 0.29 0.17 -0.11
(5.37)*** (5.03)*** (2.38)*** (-2.09)***
Post -0.01 -0.01 0.00 0.00
(-5.11)*** (-5.25)*** (-4.76)*** (-4.38)***
Historical Costs of Securities/Assets i,t-1 -0.02 -0.02 -0.02 -0.02
(-17.94)*** (-17.98)*** (-17.98)*** (-17.93)***
Nonperforming Loan/Loan i,t-1 -0.32 -0.32 -0.32 -0.32
(-32.93)*** (-32.93)*** (-32.90)*** (-32.87)***
ROE i,t-1 0.05 0.05 0.05 0.05
(16.37)*** (16.41)*** (16.40)*** (16.33)***
Core Deposits/Assets i,t-1 -0.01 -0.01 -0.01 -0.01
(-4.10)*** (-4.10)*** (-4.10)*** (-4.11)***
Leverage i,t-1 0.0002 0.0002 0.0002 0.0002
(5.95)*** (5.89)*** (5.85)*** (5.70)***
Size i,t-1 -0.001 -0.001 -0.001 -0.001
(-5.32)*** (-5.32)*** (-5.32)*** (-5.34)***
Unrealized Gains (Losses)/Assets i,t-1 +
Unrealized Gains (Losses)/Assets i,t-1 * Post 0.25 0.32 0.25 -0.05
(Clustered t-stat) (6.27)*** (6.20)*** (4.06)*** (-1.11)
Quarter fixed effects Yes Yes Yes Yes
Standard errors clustered by firm Yes Yes Yes Yes
Observations 85,621 85,621 85,621 85,621
Adjusted R-squared 0.19 0.19 0.19 0.19
Unrealized Gains (Losses) Occurred at Different Time Horizons as the Main Independent Variable
46
The economic impact of unrealized gains (losses) on lending during the fair-
value-accounting period is significant. On average, an increase in unrealized gains
(recognized during the past four quarters) from the bottom to the top quartile (decile) of
the distribution results in an increase in quarterly lending equivalent to 0.06% (0.16%) of
total assets; these effects are economically large considering that the average quarterly
loan growth for the entire sample is equivalent to 1.3% of total assets. Moreover,
approximately 200 sample banks fall into the bottom or top decile of the unrealized gains
(losses) distribution; thus, the impact of unrealized gains (losses) on lending is
economically significant for a sizable number of banks.
Thus far, Table 5 uses only asset structures to control for heterogeneity in banks’
business models. Table 6 below adds firm fixed effects to further control for business
models. The results in Table 6 are similar to those in Table 5, except that now the lending
impact of unrealized gains (losses) on lending during the fair-value-accounting period is
weaker: every one dollar of unrealized gains (losses) recognized during the past four and
the past two quarters are now associated with 12 cents and 17 cents of increase (decrease)
in current-quarter lending, respectively, after controlling for fixed firm effects; every one
dollar of unrealized gains (losses) recognized during the past one quarters is weakly
associated with 12 cents of increase (decrease) in current-quarter lending (t value of 1.10).
Again, unrealized gains (losses) disclosed in the footnote has no impact on lending during
the pre-fair-value-accounting period after controlling for fixed firm effects.
47
Table 6. Whether Unrealized Gains (Losses) Affect Banking Lending: After Controlling for Firm Fixed Effects
) 2 ( /
* / ) ( / ) (
, 1 , 3
1 , 2 1 , 1 0 ,
t i t
k
t i k t
t t i t i t i
Effects Fixed Frim Quarter Variable Control Post
Post Assets Losses Gains Unrealized Assets Losses Gains Unrealized Loan
ε α α
α α α
+ + + +
+ + = Δ
−
− −
∑
Table 6 presents results on whether unrealized gains (losses) affect banks’ subsequent lending during the pre-fair-value-accounting period of 1988-1991
versus the fair-value-accounting period of 1995-2010 after controlling for firm fixed effects. In columns 1, 2, 3, and 4, the main independent variable
unrealized gains (losses) are those occurred during the past four quarters, the past two quarters, the past one quarter, and cumulative as of the beginning
of the quarter, respectively. t-statistics based on standard errors clustered at the firm level are in parentheses. Quarter fixed effects are included in the
model. All non-dummy variables are winsorized at the 1% and 99% levels. Post is a dummy variable equal to 1 for the fair-value-accounting period and
0 for the pre-fair-value-accounting period. All other variables are calculated as defined in the Appendix. Statistical significance is based on two-sided t-
tests and is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
48
Table 6 (Continued)
(1) (2) (3) (4)
VARIABLES
Unrealized Gains
(Losses) Occurred during
Past Four Quarters
Unrealized Gains
(Losses) Occurred during
Past Two Quarters
Unrealized Gains
(Losses) Occurred
during Past One Quarter
Cumulative Unrealized
Gains (Losses)t-1 at the
Beginning of the Quarter
Intercept 0.002 0.001 0.002 0.002
(6.16)*** (5.24)*** (6.08)*** (6.74)***
Unrealized Gains (Losses)/Assets i,t-1 0.009 0.029 0.041 0.001
(0.93) (1.55) (1.10) (0.05)
Unrealized Gains (Losses)/Assets i,t-1 * Post 0.11 0.15 0.08 0.08
(2.02)** (1.88)* (0.70) (1.19)
Post 0.01 0.01 0.01 0.01
(5.79)*** (5.80)*** (5.81)*** (5.76)***
Historical Costs of Securities/Assets i,t-1 0.02 0.01 0.01 0.02
(6.34)*** (6.31)*** (6.26)*** (6.29)***
Nonperforming Loan/Loan i,t-1 -0.36 -0.36 -0.36 -0.36
(-20.14)*** (-20.14)*** (-20.11)*** (-20.07)***
ROE i,t-1 0.02 0.02 0.02 0.02
(4.95)*** (4.95)*** (4.94)*** (4.95)***
Core Deposits/Assets i,t-1 0.01 0.01 0.01 0.01
(1.81)* (1.81)** (1.82)** (1.83)**
Leverage i,t-1 -0.0001 -0.0001 -0.0001 -0.0001
(-2.46)*** (-2.50)*** (-2.57)*** (-2.44)**
Size i,t-1 -0.005 -0.005 -0.005 -0.005
(-7.26)*** (-7.28)*** (-7.28)*** (-7.29)***
Unrealized Gains (Losses)/Assets i,t-1 +
Unrealized Gains (Losses)/Assets i,t-1 * Post 0.12 0.17 0.12 0.09
(Clustered t-stat) (2.19)** (2.31)** (1.10) (1.35)
Firm fixed effects Yes Yes Yes Yes
Quarter fixed effects Yes Yes Yes Yes
Standard errors clustered by firm and quarter Yes Yes Yes Yes
Observations 85,621 85,621 85,621 85,621
R-squared 0.18 0.18 0.18 0.18
Unrealized Gains (Losses) Occurred at Different Time Horizons as the Main Independent Variable
49
Overall, consistent with Hypothesis 1, I find that unrealized gains (losses)
recognized on the balance sheet under fair value accounting during the past one to four
quarter(s) induce banks to increase (reduce) lending in the current quarter,
39
whereas the
effect is much weaker or even absent during the pre-fair-value-accounting period. Thus,
the first necessary condition for fair value accounting to exacerbate lending pro-
cyclicality does not appear to be violated. This result also suggests that under fair value
accounting, changes in fair values do affect lending behavior; therefore, concerns that fair
value accounting may exacerbate lending pro-cyclicality are legitimate.
Thus far, I have shown that the first necessary condition for fair value accounting
to exacerbate the pro-cyclicality of bank lending is not violated. Thus, the finding that
fair value accounting does not increase the pro-cyclicality of bank lending is not caused
by the violation of the first necessary condition. Then is the finding caused by the second
necessary condition (i.e., the fair values of bank assets are pro-cyclical) being violated?
Section 5.3 below investigates whether fair values of bank assets are, indeed, pro-cyclical.
5.3 Are Fair Values Pro-cyclical?
5.3.1 Research Design: Testing Whether Fair Values are Pro-cyclical
To test whether fair values tend to increase in expansions and decrease in
recessions, I examine whether changes in fair values (i.e., unrealized gains/losses) are
pro-cyclical, both at the industry and at the firm level. At the industry level, I test whether
39
In untabulated analyses, I find that the impact of unrealized gains and unrealized losses on lending is
symmetric. Put differently, unrealized losses reduce subsequent lending to the same extent that unrealized
gains increase subsequent lending.
50
the aggregate unrealized gains (losses) for the banking industry are pro-cyclical using the
following models:
) 3 ( Re ) (
1 0
A cession Losses Gains Unrealized of Sum Industry
t t t
ε α α + + =
) 3 ( ) (
1 0
B GDPgrowth Losses Gains Unrealized of Sum Industry
t t t
ε β β + + =
Models (3A) and (3B) use the recession dummy and real GDP growth as the business
cycle indicator, respectively. Positively significant
1
α and
1
β would indicate that fair
values of bank assets are pro-cyclical. Conversely, insignificant
1
α and
1
β would suggest
that fair values of bank assets are not pro-cyclical.
It is important to note that since I find in Section 5.2 that unrealized gains (losses)
recognized during the past one to four quarter(s) affect bank lending during the current
quarter, I also need to investigate whether the lagged fair values of bank assets are pro-
cyclical. Thus as a sensitivity test, I examine whether the lagged aggregate unrealized
gains (losses) for the banking industry are pro-cyclical using the following models:
) 3 ( Re ) (
1 0
C cession Losses Gains Unrealized of Sum Industry
t t i t
ε α α + + =
−
) 3 ( ) (
1 0
D GDPgrowth Losses Gains Unrealized of Sum Industry
t t i t
ε β β + + =
−
where i =1, 2, or 4 in examining whether the industry aggregate unrealized gains (losses)
recognized during the past one, two, and four quarters, respectively, are pro-cyclical.
At the firm level, I test whether banks with greater fair value accounting exposure
tend to recognize more unrealized gains in expansions and incur more unrealized losses
51
in recessions relative to banks with less fair value accounting exposure.
40
To do so, I use
the following model:
) 4 (
Re Re * / ) (
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t
k
t i k
t t t i t i t i
effect quarter Fixed Variable Control
cession cession FVA FVA Assets Losses Gains Unrealized
ε α
α α α α
+ + +
+ + + =
−
− −
∑
where i stands for bank and t for quarter. The dependent variable
t i
Assets Losses Gains Unrealized
,
/ ) ( is unrealized gains (losses) recognized on the
balance sheet during quarter t divided by total assets at historical cost. FVA
i,t-1
is bank i’s
exposure to fair value accounting at the beginning of quarter t, measured as Fair-valued
assets/Assets
i,t-1
. Recession
t
is the business cycle indicator based on NBER business cycle
dates; it takes the value of 1 for recession quarters 2001Q2-2001Q4 and 2008Q1-2009Q2,
and 0 for expansion quarters (remaining periods during 1995-2010). The control variables
are Securities ROE
i,t-1
, Leverage
i,t-1
, and Size
i,t-1
, as defined in Section 5.1.
5.3.2 Empirical Results: Industry-level Analysis of Whether Fair Values of Bank
Assets are Pro-cyclical
I first examine whether aggregate unrealized gains (losses) for the banking
industry as a whole are pro-cyclical. Figure 3 below shows the time series of industry
aggregate unrealized gains (losses) recognized during the year.
41
40
Changes in fair value equals unrealized plus realized gains (losses) (Barth 1994). However, because
realized gains (losses) do not differ under fair value accounting versus historical cost accounting,
unrealized gains (losses) is what sets fair value accounting apart from historical cost accounting. I therefore
focus on unrealized gains (losses) instead of the sum of unrealized and realized gains (losses). Nonetheless,
the results are very similar when I define changes in fair value as unrealized plus realized gains (losses)
following Barth (1994) and Venkatachalam (1996), among others.
41
For ease of exposition, here I use annual data instead of quarterly data.
52
Figure 3. Industry Aggregate Unrealized Gains (Losses) Recognized during the
Year
5.9
-0.3
2.9
-1.1
-15.5
8.0
0.8
10.9
-5.2
-1.5
-13.4
1.0
9.3
-15.2
45.3
17.4
-20
-10
0
10
20
30
40
50
1995 1997 1999 2001 2003 2005 2007 2009
Industry Aggregate Unrealized Gains (Losses) Recognized
During the Year ($ billions)
Figure 3 illustrates industry aggregate unrealized gains (losses) recognized during the year (in
$billion). The shaded areas denote NBER-dated recessions.
Interestingly, the banking industry incurs unrealized losses in six out of the
thirteen expansionary years: -$0.3B, -$1.1B, -$15.5B, -$5.2B, -$1.5B, and -$13.4B in
1996, 1998, 1999, 2003, 2004, and 2005, respectively. In fact, during 24 out of the 54
expansionary quarters, the banking industry incurs unrealized losses totaling -$122B that
offsets 68% of the $178B total unrealized gains that the industry recognizes during the
remaining expansionary quarters. Equally surprisingly, during three out of the nine
recessionary quarters, the industry reaps unrealized gains of $19B that almost completely
offsets the entire -$21B of unrealized losses that the industry incurs during the remaining
recessionary quarters. Thus, Figure 3 suggests that the banking industry as a whole does
not tend to experience unrealized gains in expansions or incur unrealized losses in
53
recessions; rather, unrealized gains (losses) appear to be equally likely to occur regardless
of good times or bad.
Table 7 below presents results from a more formal test that regresses industry
aggregate unrealized gains (losses) on the recession dummy (in Panel A) and real GDP
growth (in Panel B). In Panel A, the coefficient on Recession
t
, which captures the
incremental impact of recessions relative to expansions on industry aggregate unrealized
gains (losses), is insignificant. Panel B uses a more fine-tuned measure of aggregate
economic activity, namely quarterly real GDP growth, as the independent variable, and
finds results similar to those in Panel A. Overall, Table 7 suggests that the banking
industry as a whole does not tend to enjoy fair value gains in economic upturns or suffer
fair value losses in downturns. Thus, fair values of bank assets for the whole banking
industry do not appear to be pro-cyclical.
54
Table 7. Whether Fair Values are Pro-cyclical: Industry-level Analysis
Panel A: Recession dummy as the business cycle indicator
) 3 ( Re ) (
1 0
A cession Losses Gains Unrealized of Sum Industry
t t t
ε α α + + =
COEFFICIENTS
(t-value)
Intercept 0.96
(0.90)
Recession t -1.21
(-0.43)
Observations 63
Adjusted R-squared 0.003
VARIABLES
Panel B: Real GDP growth as the business cycle indicator
) 3 ( ) (
1 0
B GDPgrowth Losses Gains Unrealized of Sum Industry
t t t
ε β β + + =
COEFFICIENTS
(t-value)
Intercept 1.49
(1.17)
GDPgrowth t -1.17
(-0.86)
Observations 63
Adjusted R-squared 0.012
VARIABLES
Table 7 reports results on whether fair values are pro-cyclical at the industry level, i.e., whether the banking
industry tends to recognize unrealized gains during economic upturns and incur unrealized losses during
downturns. Panel A and B use dummy variable Recession
t
and quarterly real GDP growth rate as the
business cycle indicator, respectively. All variables are calculated as defined in the Appendix.
Additionally, since changes in fair values of bank assets recognized during the
past one to four quarter(s) affect current-quarter lending, I also use model (3C) and (3D)
to examine whether the lagged fair values of assets for the whole banking industry are
pro-cyclical. The untabulated results suggest that the lagged fair values of bank assets for
the banking industry are not pro-cyclical, either.
55
It is important to note that, as the unrealized losses incurred by the banking sector
during about almost half of the expansionary quarters offset approximately 68% of the
unrealized gains recognized during the remaining expansionary quarters, fair value
accounting is less likely to have contributed to excessive leverage and risk taking during
the 2003-2006 boom, which contradicts critics’ allegations. Likewise, as the unrealized
gains experienced by the banking industry during about one third of the recession
quarters almost completely offset the unrealized losses that incurred during the remaining
recession quarters, fair value accounting is less likely to have contributed to excessive
write-downs during the 2007-2008 bust, which again contradicts critics’ allegation.
42
5.3.3 Empirical Results: Firm-level Analysis of Whether Fair Values of Bank Assets
are Pro-cyclical
Table 8 below presents the results of a firm-level analysis on whether (changes in)
fair values of bank assets are pro-cyclical. The coefficient on FVA
i,t-1
is -0.0001,
suggesting that banks with higher fair value accounting exposure actually recognize more
unrealized losses (as opposed to gains) during expansions. The coefficient of -0.0001 on
FVA
i,t-1
, however, is too small to have any meaningful impact on lending.
43
The sum of
coefficients on FVA
i,t-1
and FVA
i,t-1
*Recession
t
is -0.0003 and insignificant, indicating that
42
It is important to note that although banks write down billions of dollars of their debt securities as “other-
than-temporary impairment” (OTTI) during the recent financial crisis, those OTTI could have also arisen
even under historical cost accounting with impairment and writedowns.
43
Consider a bank with 30% of total assets carried at fair value (i.e., FVA
i,t-1
=0.30, which is greater than
the 75th percentile of the fair value accounting distribution as shown in Table 1). This bank will recognize -
0.0001 x 0.30 =-0.00003 of unrealized losses during quarter t, which will translate into a change in lending
equal to -0.00003 x 0.26 = -0.000008, or a 0.0008% reduction in lending during quarter t – too small to be
economically significant.
56
banks with greater fair value accounting exposure do not tend to recognize more
unrealized losses in recessions.
Table 8. Whether Fair Values are Pro-cyclical: Firm-level Analysis
) 4 (
Re Re * / ) (
, 1 ,
3 1 , 2 1 , 1 0 ,
t i t
k
t i k
t t t i t i t i
Effects Fixed Quarter Variable Control
cession cession FVA FVA Assets Losses Gains Unrealized
ε α
α α α α
+ + +
+ + + =
−
− −
∑
Dependent variable: Unrealized Gains (Losses)/Assets
i,t
Small banks Large banks
(Total assets ≤ $1B) (Total assets > $1B)
Intercept 0.00 0.00 0.00
(22.36)*** (12.66)*** (8.86)***
FVA i,t-1 -0.0001 -0.0002 0.0003
(-2.46)*** (-4.09)*** (2.88)***
FVA i,t-1 *Recession t -0.0002 0.0001 -0.0009
(-0.84) (0.26) (-1.76)*
Recession t -0.001 -0.001 -0.001
(-14.02)*** (-13.40)*** (-6.40)***
Securities ROE i,t-1 -0.01 -0.02 -0.01
(-8.10)*** (-6.99)*** (-4.35)***
Leverage i,t-1 0.000006 0.000008 0.000001
(5.36)*** (6.38)*** (-0.28)
Size i,t-1 0.000004 0.000005 0.000001
(1.38) (0.56) (0.21)
FVA i,t-1 + FVA i,t-1 *Recession t -0.0003 -0.0001 -0.0006
Clustered t-stat of FVA i,t-1 + FVA i,t-1 *Recession t -1.39 -0.69 -1.26
Qarterly time dummies Yes Yes Yes
Standard errors clustered at the firm level Yes Yes Yes
N 69578 52890 16688
Adjusted R
2
0.49 0.49 0.49
Full sample VARIABLES
Table 8 presents results on whether fair values are pro-cyclical at the firm level. t-statistics are based on
standard errors clustered at the firm level. Quarter fixed effects are included in the model. All variables are
winsorized at the 1% and 99% levels, and are calculated as defined in the Appendix. Statistical significance
is based on two-sided t-tests and is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
57
Overall, the analysis in Table 8 reveals that banks with greater fair value
accounting exposure do not appear to recognize more unrealized gains during expansions,
nor do they appear to incur more unrealized losses during recessions. Thus, unrealized
gains (losses) do not appear to be pro-cyclical at the firm level, which is consistent with
the results at the industry level. These results suggest that the fair values of bank assets
are not pro-cyclical at the industry or firm level. Hence, the second necessary condition
for fair value accounting to exacerbate lending pro-cyclicality is violated, which has
caused fair value accounting not to have pro-cyclical impact on bank lending. In the
following subsection, I investigate a possible explanation for this phenomenon.
5.3.4 Why are Fair Values of Bank Assets not Pro-cyclical? The Relationship
between Changes in Fair Values and Changes in Interest Rates
Why are fair values of bank assets not pro-cyclical? I posit that the answer resides
in changes in interest rates because (1) banks’ fair-valued assets are mostly debt
securities, and (2) interest-rate risk is by far the biggest risk that debt security investors
face (Fabozzi, 2005). As the valuation model for debt securities in Section 3.2 illustrate,
all three components of interest rates (namely, risk-free rate, term spread, and default
spread) are important determinants of debt securities’ fair values.
44
Accordingly, I
examine whether changes in the fair values of bank assets for the whole industry (i.e., the
industry aggregate unrealized gains/losses) are associated with changes in interest rates
44
In my empirical analysis, risk-free rate is 3-month T-bill rate; term spread is 10-year Treasury bond yield
minus 3-month T-bill rate; default spread is Moody's Seasoned Baa Corporate Bond Yield minus Aaa
Corporate Bond Yield. choose to use 10-year Treasury bond yield for two reasons: 1) 10 year likely
reflects expected life of mortgage-backed securities due to potential prepayment (Boudoukh et al. 1997);
and 2) I find that mortgage-backed securities constitute about 35% (50%) of an average bank’s securities
holdings prior to (since) 2008. These three rates are approximation of the rates required by debt security
investors, since data on the required rates of specific debt securities are not available.
58
and with changes in the three components of interest rates using the following regression
models:
) 5 ( ) (
1 0
A Rates Interest Losses Gains Unrealized of Sum Industry
t t t
ε α α + Δ + =
) 5 (
) (
3
2 1 0
B Spread Default
Spread Term Rate free - Risk Losses Gains Unrealized of Sum Industry
t t
t t t
ε β
β β β
+ Δ +
Δ + Δ + =
Figure 4 below shows the time series of quarterly risk-free rate in relation to
industry aggregate unrealized gains (losses) recognized during the quarter; it illustrates
that the risk-free rate tends to rise during the second half of the expansions and fall
sharply during the recessions. Figure 5 below presents the times series of term spread and
default spread; this figure shows that the term spread and the default spread tend to rise
during the recessions, consistent with Fama and French (1989) and Chen (2010), with the
exception that the default spread falls sharply during the second half of the most recent
recession, resulting in $14B of unrealized gains; the term spread tends to fluctuate during
most of the expansionary periods (although it continuously decreases during the extended
expansion period from 2004 to early 2007), while the default spread is relatively stable
throughout both expansions. Changes in these three components of interest rates have a
significant impact on fair values of bank assets. During 1999Q1-Q3, as the Federal
Reserve continuously increased rates and both the risk-free rate and term spread rose, the
industry experienced unrealized losses of -$13B. Again when the risk-free rate and term
spread increased in 2004Q2, the industry incurred unrealized losses of -$18B during that
quarter.
59
Figure 4. Time Series of Unrealized Gains (Losses) Recognized during the Quarter
for the Banking Industry in Relation to Risk-free Rate during 1995-2010
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
-30
-20
-10
0
10
20
30
40
1995 1997 1999 2001 2003 2005 2007 2009
Risk-free Rate (%)
Industry Sum of Unrealized Gains and Losses
Industry sum of unrealized gains and losses ($billion) Risk-free rate (%)
Figure 4 presents the times series of industry aggregate unrealized gains (losses) recognized during the
quarter in relation to risk-free rate during 1995-2010. The shaded areas denote NBER-dated recessions.
Figure 5. Times Series of Term Spread and Default Spread during 1995-2010
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1995 1997 1999 2001 2003 2005 2007 2009
Term spread (%) Default spread (%)
Figure 5 illustrates the times series of quarterly term spread and default spread during 1995-2010. The
shaded areas denote NBER-dated recessions.
60
Figure 6 below presents the time series of interest rates in relation to industry
aggregate unrealized gains (losses) during 1995-2010. The figure clearly illustrates that
the industry aggregate unrealized gains (losses) move in the opposite direction of changes
in the interest rates. That is, unrealized gains (losses) tend to coincide with decreases
(increases) in the interest rates. Importantly, during both recessions the interest rates
increase in the first half of the recessions and then decrease in the second half, resulting
in unrealized losses in the first half of the recessions but then unrealized gains in the
second half. Similarly, the interest rates fluctuate during both expansions, resulting in
unrealized gains as well as unrealized losses during these periods.
Figure 6. Time Series of Unrealized Gains (Losses) Recognized during the Quarter
for the Banking Industry in Relation to the Interest Rates during 1995-2010
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
-30
-20
-10
0
10
20
30
40
1995 1997 1999 2001 2003 2005 2007 2009
Interest Rate (%)
Industry Sum of Unrealized Gains and Losses
Industry sum of unrealized gains and losses ($billion) Interest rate (%)
Figure 6 illustrates the time series of industry aggregate unrealized gains (losses) recognized during the
quarter in relation to the interest rates (the sum of risk-free rate, term spread, and default spread) during
1995-2010. The shaded areas denote NBER-dated recessions.
61
As a more formal test of the inverse relationship between changes in interest rates
and changes in the fair values of bank assets, I conduct a regression analysis. Table 9
provides the results. Panel A shows that, as the interest rates increase (decrease), the
banking industry experiences unrealized losses (gains). Panel B further illustrates that, as
each of the three components of the interest rates increases (decreases), the banking
industry experiences unrealized losses (gains).
Table 9. Whether Changes in Fair Values are Related to Interest Rate Changes
Panel A: Changes in Interest Rates as the Independent Variable
) 5 ( ) (
1 0
A Rates Interest Losses Gains Unrealized of Sum Industry
t t t
ε α α + Δ + =
COEFFICIENT
(t-value)
Intercept 0.51
(0.55)
∆Interest Rates t -5.49
(-3.13)***
Observations 62
Adjusted R-squared 0.13
VARIABLE
62
Table 9. Whether Chang es in Fair Values are Related to Interest Rate Changes
(Continued)
Panel B: Changes in the Three Components of Interest Rates as the Independent Variable
) 5 (
) (
3
2 1 0
B Spread Default
Spread Term Rate free - Risk Losss Gains Unrealized of Sum Industry
t t
t t t
ε β
β β β
+ Δ +
Δ + Δ + =
COEFFICIENT
(t-value)
Intercept 0.53
(-0.58)
∆Risk Free Rate t -5.66
(-2.34)***
∆Term Spread t -4.62
(-2.55)***
∆ Default Spread t -13.98
(-3.76)***
Observations 62
Adjusted R-squared 0.19
VARIABLE
Table 9 reports results on whether changes in the fair values for the banking industry are associated with
change in interest rates and changes in the three components of interest rates, namely risk-free rate, term
spread, and default spread. Panel A uses quarterly changes in interest rates as the independent variable, and
Panel B uses quarterly changes in the three components of interest rates as the independent variable. All
variables are calculated as defined in the Appendix. Statistical significance is based on two-sided t-tests and
is indicated as follows: *** p<0.01, ** p<0.05, and * p<0.1.
Collectively, this evidence suggests that the reason the fair values of bank assets
are not pro-cyclical is because interest rates fluctuate during both expansions and
recessions. In particular, interest rates rise (fall) during some of the expansionary
(recessionary) periods, resulting in movements of the fair values of bank assets that are
not pro-cyclical.
63
5.4 Generalizability of Results
The overall conclusion of this paper is that fair value accounting does not amplify
the pro-cyclicality of bank lending over the past two business cycles spanning 1995-2010,
despite the fact that unrealized gains (losses) affect subsequent lending. Given the large
sample size of close to 70,000 bank-quarter observations, these results are not likely to be
driven by low power of the tests.
Note that during the sample period 1995-2010 the accounting system does not
require banks to carry all assets at fair value on the balance;
45
in particular, the largest
portion of banks’ assets – namely, loans – are not required to be carried at fair value on
the balance sheet, although banks could elect to use fair value on loans beginning in 2008.
This raises the question of whether the main findings of this study would be generalizable
to full fair value accounting of the balance sheet. The answer is likely yes. Like debt
securities, loans are fixed-income instruments and, thus, their fair values likely depend on
interest rates in a similar way. Specifically, as interest rates rise (fall) during expansions
(recessions), fair values of bank loans are also likely to decrease (increase). Thus, similar
to fair values of debt securities, fair values of bank loans are unlikely to be pro-cyclical
over the past two business cycles. Therefore, fair values of banks’ total assets are less
likely to be pro-cyclical under full fair value accounting since debt securities and loans
represent the bulk of banks’ assets. This argument implies that my main finding that fair
value accounting does not exacerbate lending pro-cyclicality over the past two business
cycles likely holds even under full fair value accounting.
45
Under full fair value accounting, all assets are reported at fair value on the balance sheet.
64
Chapter 6: Conclusion
Using a large sample of 2,956 unique bank holding companies comprising 69,578
bank-quarter observations over the 1995-2010 period, this paper finds that fair value
accounting does not exacerbate the pro-cyclicality of bank lending over the past two
business cycles. Although some banks with greater fair value accounting exposure seem
to increase lending more rapidly during expansions, banks with greater fair value
accounting exposure do not reduce lending more substantially during recessions relative
to banks with less fair value accounting exposure. This non-exacerbation finding holds
after controlling for cross-sectional differences in banks’ business models. The non-
exacerbation finding also remains despite the evidence that unrealized gains (losses)
under fair value accounting affect subsequent lending, even though these holding gains
(losses) are not explicitly reflected in earnings or regulatory capital. The finding that
unrealized gains (losses) affect bank lending challenges the conventional view that
unrealized gains (losses) that are excluded from earnings and regulatory capital do not
affect bank behavior.
The two seemingly contradictory and counterintuitive findings (i.e., fair value
accounting does not increase lending pro-cyclicality; unrealized gains/losses affect bank
lending) can be reconciled by the observation that fair values of bank assets are not pro-
cyclical. Specifically, I find that during almost half (one-third) of the expansionary
(recessionary) periods studied, fair values of bank assets actually decrease (increase), as a
result of an increase (decrease) in interest rates following, for example, shifts in monetary
policy.
65
In summary, this paper shows that over the past two business cycles during 1995-
2010, fair value accounting does not increase the pro-cyclicality of bank lending. This
result could ease concerns that fair value accounting injects excessive pro-cyclicality into
the credit supplied by banks.
This paper also highlights that fair value accounting exposes banks to an
additional layer of interest rate risk. Changes in interest rates following, for example,
shifts in monetary policy, have immediately impact on banks’ balance sheets (and on
banks’ income statements if changes in fair values are included in earnings) under fair
value accounting.
66
Bibliography
Abel, Andrew, and Ben Bernanke, 2005, Macroeconomics, 5
th
edition, Pearson Addison
Wesley.
Acharya, Viral, and Matthew Richardson, 2009, Restoring Financial Stability: How to
Repair a Failed System, New York University Stern School of Business, (c) John
Wiley & Sons.
Adrian, Tobias, and Hyun Song Shin, 2010, Liquidity and Leverage, Journal of Financial
Intermediation 19, 418-437.
Adrian, Tobias, and Hyun Song Shin, 2011, Procyclical Leverage and Value-at-Risk,
Federal Reserve Bank of New York Staff Reports.
Allen, Franklin, and Elena Carletti, 2008, Mark-to-Market Accounting and Liquidity
Pricing, Journal of Accounting and Economics 45, 358-378.
American Bankers Association, 2008, Letter to SEC. September 23, 2008.
Asea, Patrick, and Brock Blomberg, 1998, Lending Cycles, Journal of Econometrics 83,
89-128.
Badertscher, Brad, Jeffrey Burks, and Peter Easton, 2012, A Convenient Scapegoat:
Fair Value Accounting by Commercial Banks during the Financial Crisis, The
Accounting Review 87, 59-90.
Banque de France, 2008, Fair Value Accounting and Financial Stability: Challenges and
Dynamics, Financial Stability Review No. 12 – Valuation and Financial Stability.
Bank of Italy, 2009, Financial Sector Pro-cyclicality: Lessons from the Crisis.
Barth, Mary, 1994, Fair Value Accounting: Evidence from Investment Securities and the
Market Valuation of Banks, The Accounting Review 69, 1-25.
Barth, Mary, Wayne R. Landsman, 2010, How did Financial Reporting Contribute to the
Financial Crisis? European Accounting Review 19, 399-423.
Beatty Anne, Scott Liao, 2011, Do Delays in Expected Loss Recognition Affect Banks’
Willingness to Lend? Journal of Accounting and Economics 52, 1-20.
Berger, Allen, Nathan Miller, Mitchell Petersen, Raghuram Rajan, and Jeremy Stein,
2005, Does Function Follow Organizational Form? Evidence from the Lending
Practices of Large and Small Banks, Journal of Financial Economics 76, 237-269.
67
Berger, Allen, and Gregory Udell, 2004, Institutional Memory Hypothesis and the Pro-
cyclicality of Bank Lending Behavior, Journal of Financial Intermediaries 13, 458-
495.
Bhat, Gauri, Richard Frankel, and Xiumin Martin 2011, Panacea, Pandora’s box, or
Placebo: Feedback in Bank Holdings Mortgage-backed Securities and Fair Value
Accounting, forthcoming, Journal Accounting and Economics 52, 153-173.
Blankespoor, Elizabeth, Thomas Linsmeier, Kathy Petroni, and Catherine Shakespeare,
2011, Bank Leverage and Credit Risk: What do fair values tell us? Working paper.
Boudoukh, Robert Whitelaw, Matthew Richardson, and Richard Stanton, 1997, Pricing
Mortgage-Backed Securities in a Multifactor Interest Rate Environment: a
Multivariate Density Estimation Approach, Review of Financial Studies 10, 405-
446.
Campbell, John, 2011, The Fair Value of Cash Flow Hedges, Future Profitability and
Stock Returns, working paper, University of Georgia.
Cantrell, Brett, John McInnis, and Christopher Yust, 2011, Predicting Credit Losses:
Loan Fair Values versus Historical Costs, working paper, University of Texas at
Austin.
Chen, Hui, 2010, Macroeconomic Conditions and the Puzzles of Credit Spreads and
Capital Structure, Journal of Finance 66, 2171-2212.
Office of Comptroller of the Currency, 1994-2011, OCC’s Quarterly Report on Bank
Trading and Derivatives Activities, http://www.occ.treas.gov/topics/capital-
markets/financial-markets trading/derivatives/derivatives-quarterly-report.html
Cornett, Marcia, Jamie McNutt, Philip Strahan, and Hassan Tehranian, 2011, Liquidity
Risk Management and Credit Supply in the Financial Crisis, Journal of Financial
Economics 101, 297-312.
Diamond, Douglas, Raghuram Rajan, 2009, The Credit Crisis: Conjectures about Causes
and Remedies, American Economic Review 99, Papers and Proceedings, 606-610.
European Central Bank, 2004, Fair Value Accounting and Financial Stability, Occasional
Paper Series, No. 13, April.
Fama, Eugene, Kenneth French, 1989, Business Conditions and Expected Returns on
Stocks and Bonds, Journal of Financial Economics 25, 23-49.
68
The Financial Accounting Standards Board (FASB), 2010, Proposed Accounting
Standards Update, Accounting for Financial Instruments and Revisions to the
Accounting for Derivative Instruments and Hedging Activities.
Financial Stability Board, 2009, The Role of Valuation and Leverage in Pro-cyclicality.
Financial Stability Forum, 2009, Addressing Procyclicality in the Financial System: a
possible framework.
Fisher, Irving, 1933, The Debt-Deflation Theory of Great Depressions, Econometrica 1,
337-357.
Foos, Daniel, Lars Norden, Martin Weber, 2010, Loan Growth and Riskiness of Bank,
Journal of Banking & Finance 34, 2929-2940.
Hann, Rebecca, Frank Heflin, and K.R. Subramanayam, 2007, Fair-value Pension
Accounting, Journal of Accounting and Economics 44, 328-358.
Fabozzi, Frank, 2005, The Handbook of Fixed Income Securities, 7th ed., McGraw Hill
Higher Education.
International Monetary Fund, 2009, Global Financial Stability Report. Chapter 3: Pro-
cyclicality and Fair value accounting.
Ivashina, Victoria, and David Scharfstein, 2010, Bank Lending during the Financial
Crisis of 2008, Journal of Financial Economics 97, 319-338.
Khan, Urooj, 2010, Does Fair Value Accounting Contribute to Systemic Risk in the
Banking Industry? working paper, Columbia University.
Kiyotaki, Nobuhiro, and John Moore, 1997, Credit Cycles, Journal of Political Economy
105, 211-248.
Kolev, Kalin, 2009, Do Investors Perceive Marking-to-Model as Marking-to-Myth? Early
Evidence from FAS 157 Disclosure, working paper, Yale University.
Kwan, Simon, 2010, Financial Crisis and Bank Lending, working paper, Federal Reserve
Bank of San Francisco.
Laux, Christian, and Christian Leuz, 2009, The Crisis of Fair value accounting: Making
Sense of the Recent Debate, Accounting, Organizations and Society 34, 826-834.
Laux, Christian, and Christian Leuz, 2010, Did Fair-Value Accounting Contribute to the
Financial Crisis? Journal of Economic Perspectives 24, 93-118.
69
Lown, Cara, and Donald Morgan, 2006, The Credit Cycle and the Business Cycle: New
Findings Using the Loan Officer Opinion Survey, Journal of Money, Credit, and
Banking 38, 1575-1597.
Minton, Bernadette, René Stulz, and Rohan Williamson, 2009, How Much Do Banks Use
Credit Derivatives to Hedge loans? Journal of Financial Services Research 35, 1-31.
Moore, Michael, 2011, FASB Backs Off Fair-Value of Loans Proposal After Opposition,
http://www.bloomberg.com/news/2011-01-25/financial-panel-scraps-plan-
requiring-banks-to-mark-assets-to-market-value.html#
Nissm, Doron, and Stephan Penman, 2007, Fair Value Accounting in the Banking
Industry, working paper, Columbia University.
Office of the Comptroller of the Currency, 1994, Capital Adequacy: Net Unrealized
Holding Gains and Losses on Available-For-Sale Securities. Final Rule. Federal
Register Volume 59, Number 226. 12 CFR Part 3. http://www.gpo.gov/fdsys/pkg/
FR1994-11-25/html/94- 29110.htm
Office of Comptroller of the Currency, 1994-2011, OCC’s Quarterly Report on Bank
Trading and Derivatives Activities, http://www.occ.treas.gov/topics/capital-
markets/financial-markets trading/derivatives/derivatives-quarterly-report.html
O’Hara, Maureen, 1993. Real Bills Revisited: Market Value Accounting and Loan
Maturity, Journal of Financial Intermediation 3, 51-76.
Peek, Joe, Eric Rosengren, and Geoffrey Tootell, 2003, Identifying the Macroeconomic
Effect of Loan Supply Shocks, Journal of Money, Credit and Banking 35, 931-946.
Plantin, Guillaume, Haresh Sapra, and Hyun Song Shin, 2008, Marking-to-market:
panacea or pandora’s box? Journal of Accounting Research 46, 435-460.
Rajan, Raghuram, 2010, Fault Lines: How Hidden Cracks Still Threaten the World
Economy, Princeton University Press.
Riedl, Edward, and George Serafeim, 2011, Information Risk and Fair Value: An
Examination of Equity Betas, Journal of Accounting Research 49, 1083-1122.
Ryan, Stephan, 2008, Accounting in and for the Subprime Crisis, The Accounting Review
83, 1605-1638.
Schwartz, Eduardo, and Walter Torous, 1989, Prepayment and the Valuation of
Mortgage-Backed Securities, Journal of Finance 44, 375-392.
70
Shaffer, Sanders, 2010, Fair Value Accounting: Villain or Innocent Victim? working
paper, Federal Reserve Bank of Boston QAU Research Paper Series.
Statement of Financial Accounting Standards No. 115 (SFAS 115), Accounting for
Certain Investments in Debt and Equity Securities.
Statement of Financial Accounting Standards No. 115 (SFAS 159), The Fair Value
Option for Financial Assets and Financial Liabilities.
Stanton, Richard, 1995, Rational Prepayment and the Valuation of Mortgage-backed
Securities, Review of Financial Studies 8, 677-708.
United States Congress (U.S. Congress), 2008, Letter to U.S. Securities and Exchange
Chairman Christopher Cox, September 30.
U.S. Securities and Exchange Commission (SEC), 2008, Report and Recommendations
Pursuant to Section 133 of the Emergency Economic Stabilization Act of 2008:
Study on Mark-To-Market Accounting.
Van den Heuvel, Skander, 2011, Banking Conditions and the Effects of Monetary Policy:
Evidence from U.S. States, working paper, Federal Reserve Board.
Venkatachalam, Mohan, 1996, Value-relevance of Banks' Derivatives Disclosures,
Journal of Accounting and Economics 22, 327-355.
Veronesi, Pietro, and Luigi Zingales, 2010, Paulson's gift, Journal of Financial
Economics 97, 339-368.
Wahlen, James, James Boatsman, Robert Herz, Gregory Jonas, Krishna Palepu, Stephen
G. Ryan, Katherine Schipper, Catherine M. Schrand, and Douglas J. Skinner, 2000,
Response to the FASB preliminary views: Reporting financial instruments and
certain related assets and liabilities at fair value, Accounting Horizons 14, 501-508.
71
Appendix: Variable Definitions
Variable Definition
Main variables of
interest
Loan Δ
Quarterly loan growth, measured as
1 1
/ ) (
− −
−
t t t
Assets loan Loan , where Loan
t
is total loans and
leases outstanding (bhck2122) at the end of quarter t, Loan
t-1
is total loans and leases outstanding at the beginning of
quarter t, and Assets
t-1
is total assets at historical cost,
measured as total assets (bhck2170) minus cumulative fair
value unrealized gains and losses (bhck8434).
FVA Individual banks’ exposure to fair value accounting,
measured as fair-valued assets deflated by total assets. Fair-
valued assets are assets carried at fair value on the balance
sheet, measured as the sum of fair value of trading assets
(bhck3545) and fair value of available-for-sale securities
(bhck1773) during 1995-2007, and the sum of fair value of
trading assets, fair value of available-for-sale securities, and
fair value of other assets elected to use fair value accounting
under FAS 159 during 2008-2010.
Unrealized Gains
(Losses)/Assets
Changes in fair values, measured as unrealized gains (losses)
divided by total assets at historical cost. During the fair-value
accounting period of 1995-2010, the unrealized gains (losses)
are those recognized on the balance sheet from available-for-
sale securities during the past four/two/one quarter(s), the
current quarter, or cumulative unrealized gains (losses)
(bhck8434). During the pre-fair-value accounting period of
1988-1991, the unrealized gains (losses) are holding gains
(losses) of investment securities that occurred during the past
four/two/one quarter(s), the current quarter, or cumulative
holding gains (losses) (the difference between market value
of investment securities (bhck0391) and book value of
investment securities (bhck0390)).
Industry Sum of
Unrealized Gains
(Losses)
Industry aggregate unrealized gains (losses) recognized
during quarter t for all bank holding companies in the
sample.
72
Post A dummy variable indicating whether the quarter belongs to
the pre-fair-value-accounting period of 1988-1991 (Post=0)
or the fair-value-accounting period of 1995-2010 (Post=1).
Recession
Business cycle indicator based on NBER business cycle
dates. Recession takes the value of 1 for recession quarters
2001Q2-2001Q4 and 2008Q1-2009Q2, and 0 for expansion
quarters (i.e., remaining quarters during 1995-2010).
Control variables
Historical Costs of
Securities/Assets
The fraction of total assets that is securities, measured as
historical costs of securities divided by total assets at
historical cost. Historical costs of securities is the sum of
trading assets at fair value (bhck3545), available-for-sale
securities at fair value (bhck1773), and held-to-maturity
securities at fair value (bhck1754), minus cumulative
unrealized gains and losses (bhck8434).
Nonperforming
Loan/Loan
Quality of preexisting loan portfolios, measured as
nonperforming loan divided by total loans (bhck2122).
Nonperforming loans are the sum of loans past due for 90
days or more (bhck5525) and loans no longer accruing
interest revenue (bhck5526).
Securities ROE Profitability of securities businesses, measured as the sum of
realized gains (losses) on available-for-sale securities
(bhck3196) and realized gains (losses) on held-to-maturity
securities (bhck3521) during the last quarter, divided by
average total equity capital (bhck3210) during the last
quarter.
Non-securities ROE The profitability of non-securities businesses, measured as
total return on equity (ROE) minus Securities ROE.
Core Deposits/Assets Reliance on core deposits as a stable source of financing,
measured as core deposits divided by total assets at historical
cost. Core deposits are the sum of deposits under $100,000
plus all transactions deposits, and are measured as bhcb2210
+ bhcb 3187 + bhcb 2389 + bhcb 6648 + bhod3189 + bhod
3187 + bhod 2389 + bhod 6648. Total assets at historical cost
are total assets (bhck2170) minus cumulative unrealized
gains and losses (bhck8434).
73
Leverage Total assets (bhck2170) divided by equity capital
(bhck3210).
Size Firm size, measured as the natural log of total assets
(bhck2170).
Macroeconomic
variables
Δ Risk Free Rate Quarterly changes in risk-free rate. Risk-free rate is 3-month
Treasury bill rate from CRSP.
Δ Term Spread Quarterly changes in term spread. Term spread is measured
as yield of 10-year treasury with constant maturity minus
risk-free rate. Data on yield of 10-year treasury come from
St. Louis Fed.
Δ Default Spread Quarterly changes in default spread measured as Moody's
Seasoned Baa Corporate Bond Yield minus Moody's
Seasoned Aaa Corporate Bond Yield. Data come from St.
Louis Fed.
Δ Interest Rates Quarterly changes in interest rates. Interest rates are the sum
of risk-free rate, term spread, and default spread.
GDPgrowth GDP growth rate adjusted for inflation. Data come from St.
Louis Fed.
Abstract (if available)
Abstract
This paper examines whether fair value accounting increases the pro-cyclicality of banks’ lending behavior. Exploiting cross-sectional variation in individual banks’ exposure to fair value accounting, I find that fair value accounting does not exacerbate the pro-cyclicality of bank lending over the past two business cycles during 1995-2010. This result holds despite the fact that every one dollar of unrealized gains (losses) is associated with at least 12 cents of new lending (cutbacks in lending). The probable cause of this non-exacerbation finding is that interest rates rise (fall) during some of the expansionary (recessionary) periods, resulting in movements of bank assets’ fair value that are not pro-cyclical.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Can I borrow your footnotes? Learning and network benefits of footnote similarity
PDF
Who manages the firm matters: the incremental effect of individual managers on accounting quality
PDF
Real asset liquidity and asset impairments
PDF
Does mandatory adoption of International Accounting Standards reduce the cost of equity capital?
PDF
Accrual quality and expected returns: the importance of controlling for cash flow shocks
PDF
Does mandatory adoption of international financial reporting standards decrease the voting premium for dual-class shares: theory and evidence
PDF
Two essays on financial crisis and banking
PDF
Do shared auditors improve audit quality? Evidence from banking relationships
PDF
Does differential sensitivity to aggregate earnings shocks drive post-earnings-announcement drift?
PDF
Gone with the big data: institutional lender demand for private information
PDF
Leveling the playing field: unbiased tests of the relative information content of book income and taxable income
PDF
A comparison of value-added, orginary least squares regression, and the California Star accountability indicators
PDF
Effectiveness of the SEC’s comment letters in initial public offerings
PDF
The role of accounting information in the sentiment-price relation
PDF
Essays in asset pricing
PDF
Share repurchases: how important is market timing?
PDF
The term structure of CAPM alphas and betas
PDF
Two essays in econometrics: large N T properties of IV, GMM, MLE and least square model selection/averaging
PDF
For the love of the game? ownership and control in the NBA
PDF
Comparing dependent groups with missing values: an approach based on a robust method
Asset Metadata
Creator
Xie, Biqin Summer
(author)
Core Title
Does fair value accounting exacerbate the pro-cyclicality of bank lending?
School
Leventhal School of Accounting
Degree
Doctor of Philosophy
Degree Program
Business Administration
Degree Conferral Date
2012-08
Publication Date
07/06/2012
Defense Date
05/30/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
bank lending,fair value accounting,OAI-PMH Harvest,pro-cyclicality
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Subramanyam, K.R. (
committee chair
), DeFond, Mark L. (
committee member
), Nugent, Jeffrey B. (
committee member
), Zhang, Jieying (
committee member
)
Creator Email
biqinxie@usc.edu,bxx5@psu.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-49714
Unique identifier
UC11288382
Identifier
usctheses-c3-49714 (legacy record id)
Legacy Identifier
etd-XieBiqin-915.pdf
Dmrecord
49714
Document Type
Dissertation
Rights
Xie, Biqin
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
bank lending
fair value accounting
pro-cyclicality