How do banks react to increased asset risks?
Evidence from Hurricane Katrina∗
Claudia Lambert†, Felix Noth and Ulrich Schüwer‡
September 2012
Abstract
The instability of banks during the recent financial crisis underlines the importance of
understanding how banks determine their capital ratios. This paper conducts the first
empirical assessment on how banks adjust their capital ratios following an exogenous
shock to their asset risks. The existing literature, which uses non-experimental identification, faces the difficulty that banks typically determine capital ratios and asset risks
simultaneously. Using Hurricane Katrina as a natural experiment, we find that banks in
the disaster areas increase their risk-based capital ratios after the hurricane. This finding
shows that banks act precautious by themselves irrespective of regulatory requirements.
However, when we examine low-capitalized and high-capitalized banks separately, we find
that results are driven by high-capitalized banks. In addition, high-capitalized banks
increase their risk-based capital ratios by decreasing loans and not by increasing capital.
Keywords: financial crisis, bank regulation, capital requirements, natural experiment
JEL Classification: G21, G28
∗
We would like to thank Horst Entorf, Jan P. Krahnen, Gregory Nini, Jörg Rocholl, Reinhard H. Schmidt,
Adi Sunderam, Marcel Tyrell, Greg Udell, Laurent Weill and participants at the 2012 European Finance
Association Conference, the 2012 Financial Intermediation Research Society Conference, the 2012 AEA annual
meeting in Chicago and the 2012 CEPR Winter Conference on Financial Intermediation in Lenzerheide for
valuable comments and suggestions. Any remaining errors are, of course, our own.
†
Goethe University of Frankfurt, Department of Econometrics, Grüneburgplatz 1, 60323 Frankfurt am
Main, Germany, E-mail: [email protected] (corresponding author).
‡
Goethe University Frankfurt, Department of Finance, Grüneburgplatz 1, 60323 Frankfurt am Main, Germany, E-mail: [email protected], [email protected], [email protected].
1
Introduction
The U.S. financial crisis demonstrates that the stability of the banking system is essential
to economic welfare. In particular, the crisis caused unstable banks to cut lending and to
increase loan spreads (e.g., Ivashina and Scharfstein, 2010; Puri et al., 2010; Santos, 2011),
which led to a decline in corporate investments (e.g., Campello et al., 2010; Duchin et al.,
2010). Because the deposit insurance creates incentives for banks to become excessively risky,
banking regulation is generally a necessity. Regulation builds on the idea that minimum
capital ratios mitigate distortions from inefficiently priced deposit insurance and thus reduces
banks’ moral hazard. With sufficient equity ”at stake“, banks should have less incentives to
take excessive risks and thus should be more stable. The traditional view is that capital requirements are binding for banks, because banks try to economize on costly capital. However,
empirical evidence shows that banks hold levels of capital well above the regulatory minimum
(e.g., Flannery and Rangan, 2008). An important question resulting from the financial crisis
is whether ”banks choose the right capital structure, left to themselves, or . . . does the government have to force them to raise more capital” (Gale, 2010). In summary, bank capital
ratios are the cornerstone of banking policy and regulation, but the determination of these
ratios is not well understood.
This paper tests the hypothesis that an exogenous adverse shock to banks’ assets, which
initially increases their asset and bankruptcy risks, causes banks to choose a new optimal
capital ratio. A positive relation suggests that banks wish to mitigate bankruptcy risks. A
negative relation points towards incentives for banks to exploit the deposit insurance. No
significant relation indicates that risk does not affect a bank’s capital-ratio decisions outside
regulatory requirements.
In order to identify causality between a bank’s asset risk and its capital ratio, we use Hurricane
Katrina as a natural experiment. Hurricane Katrina caused estimated damages to property
in excess of $200 billion and ranks among the costliest natural disasters in United States
history (Congleton, 2006). Consequently, Katrina exposed banks in the U.S. Gulf Coast
region to unexpected losses and increased their asset risks in August 2005 because a large
part of the potential damages for borrowers was not insured. Importantly, the hurricane
1
caused uncertainty for banks on how commercial and individual borrowers could cope with the
damages. In particular, asymmetric information between banks and their borrowers increased,
and it was also uncertain how the overall economy in the affected areas would recover from
the shock. Fig. 1(a) illustrates loan loss provisions over loans for independent banks located
in areas that the hurricane affected (affected banks) and for banks located in areas unaffected
(unaffected banks).1 While affected and unaffected banks follow similar trajectories before
the hurricane, it is clear that loan loss provisions over loans increased sharply for affected
banks in Q4 2005. The FDIC (2006) characterized the situation as follows:
Hurricane Katrina had a devastating effect on the U.S. Gulf Coast region that
will continue to affect the business activities of the financial institutions serving
this area for the foreseeable future. Some of these institutions may face significant
loan quality issues caused by business failures, interruptions of borrowers’ income
streams, increases in borrowers’ operating costs, the loss of jobs, and uninsured or
underinsured collateral damage.
Along the same lines, the major rating agencies announced close monitoring of capital adequacy and the risk-management processes of affected banks in the aftermath of Hurricane
Katrina (Moody’s, 2005; Guy Carpenter, 2006). Furthermore, Hurricane Katrina changed the
(perceived) hurricane risks, as reflected in the increases of 30% or more for property insurance
premiums (USA TODAY, 2010).
[Fig. 1]
For the estimation of the effect of an exogenous increase in the riskiness of banks’ asset risks
on the total risk-based capital, we employ a difference-in-difference estimation technique. All
independent affected banks in the U.S. Gulf Coast region comprise our treatment group,
and all independent unaffected banks in the U.S. Gulf Coast region comprise our control
group. The observation that affected banks on average increase their risk-based capital ratios
1
Our classification into affected banks and unaffected banks builds on hurricane data from the Federal
Emergency Management Agency (FEMA). The analysis focuses on independent banks and leaves out banks
that are part of a bank-holding company. Details are provided in Section 2.
2
provides the first indication on how banks reacted to the increased asset risks relative to the
control group after the hurricane, shown in Fig. 2.
[Fig. 2]
The key findings of our analysis are as follows: A shock that increases banks’ asset risks
induces the affected banks to increase their risk-based capital ratios after the event relative
to the banks in the control group (unaffected by the shock). This increase shows that asset
risk is an important determinant of banks’ capital ratios outside of any regulatory requirements. Affected banks presumably reestablish their bank-specific cushion against insolvency
and choose a capital ratio that corresponds to their risk-return profile before the hurricane.
Furthermore, when we examine low-capitalized and high-capitalized banks separately, we find
that the precautionary behavior only holds for high-capitalized banks.2 A potential explanation from the literature is that banks with high franchise values and/or high bankruptcy
costs have incentives to avoid bankruptcy, and are thus characterized by high-capital ratios
(before a hurricane) and precautionary measures (after a hurricane). Further, our analysis
shows that high-capitalized banks react to the higher asset risks by shifting investments from
risky loans to risk-free U.S. government securities, and thereby increase their capital ratios.
We do not find a significant effect for changes in leverage.
We add to the literature by establishing a causal effect of asset risks on capital ratios. This
causal effect is difficult to examine because asset risk and capital ratios are typically determined simultaneously by banks. Moreover, it is difficult to measure asset risk. For example,
risk weighted assets only partially capture the risk a bank carries, and the standard deviation
of asset returns reflects past but not future risk. In our study, the difficulty of establishing a
causal effect and the difficulty of measuring asset risk is circumvented by using an exogenous
shock. Previous studies do not use exogenous variation to identify a relation and to measure
asset risk. Using simultaneous equations, two-stage, or standard OLS estimation techniques,
these studies typically find a positive relation between risk and capital ratios (e.g., Shrieves
and Dahl, 1992; Flannery and Rangan, 2008; Gropp and Heider, 2010).3 Our findings are in
2
We define all banks with an average capital ratio before the hurricane below the median of 15.8% as
“low-capitalized” and all other banks as “high-capitalized”.
3
Further, the main message from Gropp and Heider (2010) is that capital regulations are of second-order
3
line with findings from these studies and add further evidence on the causal relation between
asset risk and capital ratios.
This paper also adds to the growing literature that explores the real effects of substantial
adverse credit shocks. Puri et al. (2010) find that the U.S. crisis led to a contraction in banks’
retail lending in Germany, which they attributed to the supply side. Ivashina and Scharfstein
(2010) find that banks are less likely to cut down on lending if sufficient refinancing from
deposits is available such that they do not need to rely on short-term debt. Using the landprice collapse in Japan in the early 1990s, Gan (2007) reports that firms with greater collateral
losses get less credit and reduce investments. Our paper presents complementary evidence
that high-capitalized banks cut their loans and substitute them with risk-free government
securities. Our results that banks do not significantly adjust their leverage ratios shows the
relative “stickiness” of equity capital as discussed by Adrian and Shin (2011).
More generally, this paper adds to the literature on the determinants of banks’ risk taking
and the related impact of bank regulation. Kim and Santomero (1988) and Furlong and
Keeley (1989) consider how capital requirements affect banks’ risk taking. On the one hand,
higher capital requirements reduce incentives for banks to increase asset risk. On the other
hand, there is a contrary effect for banks to invest in riskier projects as a consequence of the
restricted risk-return profile. Hellmann et al. (2000) study the combined effect of financial
liberalization and capital regulation on banks’ risk taking. Diamond and Rajan (2000) show
that bank capital reduces the probability of financial distress in times of crisis. They analyze
how capital as a buffer against shocks affects banks’ liquidity creation. Marcus (1984) shows
that when charter values are high, exploiting the deposit insurance is not optimal for banks.
Along these lines, Keeley (1990) analyzes how increased competition triggers a decline in
banks’ charter values and causes banks to become more risky. Laeven and Levine (2009)
analyze how bank regulation affects risk taking and show that it depends on bank governance.
Allen et al. (2011) analyze the role of credit-market competition in capital regulation and find
that market discipline can be induced from the asset side.
This paper proceeds as follows. Section 2 contains the data and summary statistics. Section 3
importance and time-invariant bank fixed effects are the most important determinant of banks’ capital structures.
4
presents our identification strategy and empirical model. Section 4 shows the main estimation
results. Section 5 explores the differences between high-capitalized and low-capitalized banks.
Section 6 further investigates mechanisms through which affected banks change their total
risk-based capital ratios. Section 7 concludes. All figures and tables appear in the appendix.
2
Data and summary statistics
We start by describing the data sources and sample selection for our study. Next, we discuss
how we measure bank asset risks, our main explanatory variable, and bank capital ratios,
which is the main dependent variable. Further, we provide summary statistics for our sample.
2.1
Data sources and sample description
Our data come from three public sources. As regards the impact of Hurricane Katrina4 on
the U.S. Gulf Coast region, we use data from the Federal Emergency Management Agency
(FEMA). The FEMA is a U.S. agency established in 1979 with “the mission of helping communities nationwide prepare for, respond to and recover from natural and man-made disasters”
(FEMA, 2008). Further, we use the FDIC database that provides “Reports of Income and
Condition” (Call Reports) that comprises detailed quarterly financial and regulatory bank
data for all commercial and domestic banks in the U.S. The data refer to individual FDICinsured institutions, and the FDIC does not consolidate the data into banking groups. Also,
we use quarterly unemployment data at the county level from the Bureau of Labor Statistics to examine the robustness of our results with regard to time-varying macroeconomic
conditions.
Our sample includes all independent banks (i.e., banks that are not part of a bank holding
company) in the Gulf Coast region and neighbor states, that is, Louisiana, Mississippi, Texas,
Florida, Alabama, Oklahoma, Georgia, Arkansas and Tennessee. We focus on independent
banks and exclude all banks that are part of a bank-holding company (“BHC banks”) for two
4
Note that we also consider effects of Hurricanes Rita and Wilma in our estimation. Hurricane Rita hit the
Gulf Coast region in September 2005 and Hurricane Wilma in October 2005.
5
reasons.5 First, these banks typically operate regionally and are therefore likely to be affected
by an exogenous shock that hits their region. Second, we want to exclude distortive effects
from internal capital markets within bank-holding groups that are due to capital allocations
or implicit and explicit guarantees (Houston et al., 1997; Froot and Stein, 1998). Therefore,
we do not expect a significant impact from the hurricane on banks’ capital ratios that are part
of a banking group such as Bank of America, Citigroup, or Wachovia. Independent banks
share features with community banks discussed by DeYoung et al. (2004) and are a viable part
of the U.S. banking sector. Further, as shown by the summary statistics in Subsection 2.4,
independent banks and BHC banks are not significantly different across several key financial
figures.
The sample covers a period of the ten quarters before the hurricane, Q1 of 2003 to Q2 of
2005; the ten quarters afterwards, Q1 of 2006 to Q2 of 2008: and Q3 and Q4 of 2005 around
the hurricane. Our baseline regressions cover a period of eight quarters around the event
that more clearly avoids measuring effects from the U.S. financial crisis. We also run some
regressions for shorter time periods of four and six quarters around the event.
Previous studies that have also used the FDIC data set find that some of the data is erroneous
or includes banks that are “not viable”. Therefore, we follow Berger and Bouwman (2009)
and exclude banks that (1) have no commercial real estate or commercial and industry loans
outstanding, (2) have zero or negative equity capital, (3) hold assets below $25 million or (4)
hold consumer loans exceeding 50% of gross total assets.6 Further, we leave out banks with
very high total risk-based capital ratios above 40%, which represents five times the regulatory
requirement of 8%. We also exclude biases from newly founded banks. Therefore, we require
that banks be in existence for at least two years before the event took place. Naturally, we
also exclude all banks that were established after the hurricane took place. The final data set
includes 287 banks of which we classify 99 as affected and 178 as unaffected by the hurricane.
5
6
Technically, we require that the FDIC data field that denotes a bank-holding company is left blank.
Some further exclusion criteria used by Berger and Bouwman (2009) are not relevant to our sample.
6
2.2
Bank asset risks
Our main explanatory variable for bank capital ratios is bank asset risk. Measures for bank
asset risks that are frequently used in the literature are the change in risk-weighted assets
(e.g., Avery and Berger, 1991), the standard deviation of the return over assets (e.g., Laeven
and Levine, 2009, calculate a bank z-score) or the standard deviation of (unlevered) stock
price returns (e.g., Gropp and Heider, 2010; Flannery and Rangan, 2008). In our study,
we depart from using these traditional measures because they cause endogeneity concerns.
For example, a total risk-based capital ratio on the left hand side of the model by definition
reflects risk-weighted assets. Further, banks typically determine their asset risk and capital
ratio simultaneously.
The measure we use for bank asset risk is the exogenous variation that Hurricane Katrina
introduces that exposes banks in the U.S. Gulf Coast region to unexpected losses and increased
credit risk in the second half of 2005. We are thus able to identify a causal relation, and to
evaluate whether a shock to the banks’ asset risk induces banks to adapt their total risk-based
capital ratio.
Following Hurricane Katrina in the second half of 2005, the FEMA designated 135 out of 534
counties in the Gulf Coast region (Louisiana, Mississippi, Texas, Florida and Alabama) as
eligible for individual and public disaster assistance. Using this information we classify banks
as affected by a hurricane if their headquarters are located in counties that were eligible for
individual and public disaster assistance (dark-grey shaded region in Fig. 3). Next we classify
banks as unaffected if their headquarters are located in counties not affected by Hurricane
Katrina (light-grey shaded area in Fig. 3). Last, we exclude banks from the sample if their
headquarters are located in counties eligible for public disaster assistance but not eligible
for individual disaster assistance, because this criterion is ambiguous. The sample includes
counties that are both somehow affected and very distant from the wind fields, for example,
counties in the northwest Texas region that possibly were designated for public assistance for
political reasons. To guarantee that we are dealing with banks that were clearly affected or
clearly not affected by the hurricane, we exclude banks from these counties. Consequently,
we are left with a clean identification of banks located in affected counties and banks located
7
in unaffected counties.
[Fig. 3]
2.3
Bank capital ratios
The main dependent variable that we use in our estimations is the total risk-based capital
ratio as reported for each bank at the end of each quarter in the FDIC database. The
sum of a bank’s Tier 1 and Tier 2 capital divided by its risk-weighted assets represents its
regulatory capital ratio.7 Risk-sensitivity is the basis for the definition of risk weights in
accordance with the relevant Basel accord. The banks in our sample operate in a Basel I
regulatory environment. Consequently they can assign risk weights corresponding to five
different categories that range from zero to 100%. For example, U.S. government securities
have a risk weight of zero, uncovered mortgage loans have a risk weight of 50%, and commercial
loans have a risk weight of 100%. Banks are required to hold capital equal to 8% of riskweighted assets.
2.4
Summary statistics and similarity between groups
Table 1 provides summary statistics for all of the variables used in the regressions and some
further bank characteristics for the two years before the hurricane, that is, Q3 2003 to Q2
2005.8 Because our later difference-in-difference estimation uses affected banks as a treatment
group and unaffected banks as a control group, it is important that banks in both groups
have similar characteristics. The table therefore reports mean values and standard deviations
separately for both groups. As suggested by Imbens and Wooldridge (2009), the table also
reports normalized differences to compare the similarity between both distributions for important bank characteristics.9 As a rule of thumb, groups are regarded as sufficiently equal
7
For some banks, the denominator also includes Tier 3 capital allocated for market risk, net of all deductions.
For details, see “Schedule RC-R – Regulatory Capital” of the FDIC.
8
The appendix in Table 12 provides a description of the FDIC data.
9
Normalized differences are calculated as “the difference in averages by treatment status, scaled by the
square root of the sum of the variances” (Imbens and Wooldridge, 2009, p. 24).
8
and adequate for linear regression methods if normalized differences are basically in the range
of ± 0.25.
Overall, the summary statistics confirm that affected and unaffected banks are on average
relatively similar. In particular, banks in both groups hold similar levels of total risk-based
capital ratios of around 17.7% during the two years before the hurricane event. Note that this
level substantially exceeds the regulatory minimum of 8%. This observation is in line with
Flannery and Rangan (2008) who report high ratios for the U.S., and Barth et al. (2005) and
Berger et al. (1995) who examine high cross-country ratios. We observe the same pattern for
the tier capital-to-asset ratio where the mean value for banks’ risk-weighted assets to assets
is higher for the unaffected group but not significantly according to normalized differences of
about ± 0.25. If we consider volumes of total risk-based capital, risk-weighted assets, and
total assets, then we also do not find any significant differences. The same applies for loans
and different loan types as well as for government securities, profitability, and risk measures
for both groups. We also find that the share of commercial banks and banks that are classified
as stock institutions are similar in counties that the hurricane affected and those that it did
not. The same is true for the regional quarterly unemployment rate.
As regards external validity in our sample, we are also interested in the comparison between
affected independent banks, which we include in our regressions, and affected nonindependent
banks, bank-holding company (BHC) banks, which we exclude from our regressions. Therefore
columns 5 and 6 of the table present the summary statistics for BHC banks. The statistics
show that independent and BHC banks on average do not differ across a broad range of bank
characteristics. The BHC banks tend to be larger and comprise only stock institutions, but
the normalized differences are not outside the ± 0.25 range. Thus, our sample of independent
banks is not significantly different from the sample of BHC banks.
[Table 1]
9
3
Identification strategy and empirical model
To assess how the hurricane affects a bank’s asset risk and consequently the bank’s capital
structure decision, this study considers potentially parallel macroeconomic and industry-wide
factors that affect all banks independent of the shock. It would be misleading to simply test
how affected banks adapt their total risk-based capital ratios after the hurricane. Another
concern is that unobservable bank characteristics might influence the analysis. To account for
both aspects, we use a difference-in-difference estimation technique with bank fixed effects.
3.1
Difference-in-difference identification
Ideally, a comparison of our observations on the capital ratios of a bank affected by Hurricane
Katrina could be made with a bank’s hypothetical capital ratio in case the bank had not been
exposed to the hurricane. To do so, let DIS be a binary variable indicating whether bank i
1
is affected (DIS = 1) or is not affected (DIS = 0) at time t. Let yi,t+1
represent the value
for the total risk-based capital ratio of an affected bank i in t + 1 after Hurricane Katrina at
0
time t. The yi,t+1
represents the hypothetical adaption of the total risk-based capital ratio of
bank i in time t + 1 had the bank not gone through a hurricane. The effect of Katrina on the
average total risk-based capital ratio for bank i, classified in the evaluation literature as the
average treatment effect on the treated (see, for example, Angrist and Pischke, 2009), can be
formally stated as:
1
0
AT T = E(yi,t+1
|DIS = 1) − E(yi,t+1
|DIS = 1).
(1)
1
The term E(yi,t+1
|DIS = 1) represents the expected value for the total risk-based capital ratio
of an affected bank i in t + 1 after Hurricane Katrina at time t. This term can be identified
through the observed average effect on total risk-based capital for banks that were affected by
0
Katrina. For this sample of affected banks, the term E(yi,t+1
|DIS = 1) represents the coun-
terfactual expected mean or hypothetical effect on the total risk-based capital ratio assuming
these banks were initially not affected by the hurricane. This effect being unobservable needs
to be approximated and represents the central problem of causal inference (Holland, 1986).
10
In experimental studies, the identification problem is typically solved through random assignment for both groups, the treated and the control. When treatment is randomized across
individuals, self-selection biases are excluded such that the mean independence assumption
holds:
1
1
E(yi,t+1
|DIS = 1) = E(yi,t+1
) and
0
0
E(yi,t+1
|DIS = 1) = E(yi,t+1
).
(2)
In our study we exploit exogenous variation from Hurricane Katrina. This natural disaster
randomly selected banks into a treatment group (affected by Katrina) and a control group
0
(not affected by Katrina), such that E(yi,t+1
|DIS = 1) can be identified by measuring the
observed average effect on total risk-based capital ratios of the control group.
The comparison of both the treatment and control groups after the event could be misleading
because trajectories of total risk-based capital could already point in different directions for
the treatment and control groups before the hurricane. Therefore, we rely on a difference-indifference estimation that compares the ratio of the total risk-based capital after treatment
both to the treatment group before treatment and to the control group.
3.2
Baseline estimation
In applying a difference-in-difference estimation technique, we estimate whether higher capital
ratios are systematic and can be attributed to the hurricane. Formally, we estimate the
following equation:
CAPit = β0 + β1 Eventt + β3 (Eventt ∗ Affectedi ) + τγ + νi + it .
(3)
The dependent variable CAPit is the total risk-based capital ratio of bank i at time t. Our
event window is Q3 and Q4 of 2005. Accordingly, the variable Eventt is a time dummy
with a value of zero for all quarters before the hurricane (t ≤ Q2 2005) and a value of
one for all quarters after the hurricane (t ≥ Q1 2006). The variable Affected i is a dummy
variable of bank i that is one if the bank is located in a county classified by FEMA as eligible
for “public and private disaster assistance” and thus belongs to the treatment group, and
zero otherwise (for the control group). Hence, the interaction term Event t ∗Affected i is one
11
if both the variable Eventt and the variable Affected i amount to one, and zero otherwise.
The corresponding coefficient β3 is the main interest. It captures the average effect of the
hurricane on the total risk-based capital ratios of affected banks. The variable τγ represents
yearly time fixed effects (Time FE). Further, we are concerned that unobserved differences
between banks might influence our results. Thus, we include fixed effects νi for each bank i
(Bank FE). Finally, it is the idiosyncratic error term. To account for heterogeneity among
banks, we use clustered standard errors at the bank level.
For robustness, we reestimate our baseline estimation with two variations. First, we estimate
Eq. (3) without bank fixed effects. The variable Affected i that otherwise interferes with bank
fixed effects then enters the equation. Second, we estimate Eq. (3) with control variables
that are common in the banking literature. In particular, we add bank size represented by
the log of the total number of employees, the ratio of non-performing loans to assets and the
return over assets (RoA).10 For example, bank size is generally an important factor for banks
that operate with less capital (e.g., Demsetz and Strahan, 1997). Note that these control
variables only matter for the estimation to the degree that they are time variant because
they are otherwise already included in the bank fixed effects. As an additional time-varying
control variable that captures differences in local economic developments, we use quarterly
unemployment rates at the county level.
4
The effect of increased asset risk on risk-based capital ratios
We begin by presenting the main estimation results and then describe several further estimation results to expand and corroborate those results.
4.1
Main estimation results
We present our main results in Table 2. Column 1 shows the difference-in-difference estimation
without bank fixed effects. With regard to our main variable of interest, the interaction term
10
Results remain qualitatively the same if we use total assets instead of employees for size effects and RoE
instead of RoA as the performance indicator.
12
Eventt *Affectedi , we observe a positive and significant coefficient that shows that affected
banks increase their capital ratios after the hurricane. This effect is also highly significant
economically. Affected banks have a total risk-based capital ratio that is on average 1.2
percentage points higher than the ratio of unaffected banks after the hurricane, as shown by
the point estimate of the interaction term.
Next, column 2 shows results for our baseline estimation with bank fixed effects [Eq. (3)].
Importantly, the results remain robust and confirm that a change in the risk environment
of a bank is highly relevant. The average effect of the hurricane on affected banks’ total
risk-based capital ratios of 1.1 percentage points is slightly lower but still is in the same range
as before. Note that bank fixed effects explain a lot of the variation in total risk-based capital
ratios (note the adjusted R2 increases from 0.3% to 85%). This suggests that unobserved
time-invariant bank-specific factors are also important.
Last, we add bank characteristics that are regarded as relevant for banks’ capital ratios and
the unemployment rate at the county level. We find that size approximated by employees adds
some explanation to our regression but bank profitability and non-performing loans are not
significant. In addition, we find that banks that operate in counties with higher unemployment
have higher total risk-based capital ratios. But foremost, adding other covariates leaves the
effect of the interaction term intact.11
[Table 2]
This first set of results strongly advocates that independent banks react when confronted with
a shock that increases the economic risk in their business region. They do this by increasing
their (regulatory) risk-based capital ratio relative to banks that do not experience this shock.
In other words, the results suggest that banks strengthen their bank-specific cushion against
insolvency, and they are not reacting purely to regulatory capital requirements. This finding
adds to Flannery and Rangan (2008) who suggest that a change in the banking environment
rather than supervisory pressure leads to higher capital ratios for U.S. banks during the 1990s.
Moreover, we are in line with Gropp and Heider (2010) who argue that banks rely on “own
11
In unreported results we find the same effect when we consider Tier 1 risk-based capital ratios instead of
Total risk-based capital ratios.
13
judgement” to define the appropriate amount of total risk-based capital and that regulatory
requirements are of second-order importance.
4.2
Alternative time horizons
In the previous subsection, we established our main results for the period of ± 2 years around
the event. Next, we investigate whether this effect differs for alternative time horizons around
the event. Therefore, Table 3 shows results for the fixed-effects estimation of Eq. (3) for the
periods of 4, 6, 8 and 10 quarters around the hurricane.
We find that the effect on banks’ total risk-based capital ratios for the shortest period is also
positive but becomes less statistically significant. This finding suggests that banks require
some time to adjust to the new risk environment and to build up their total risk-based capital
ratios. For periods of six quarters or longer, the effect stays significant at the 5% level and is
also economically significant with a value around 1 percentage point. Furthermore, we also
find that the statistical significance maintains for the longest period of ± 10 quarters. This
finding might be interpreted as the uncertainty about borrower quality and credit risks with
the economic prospects of the region might not have diminished more than two years after
the hurricane.
[Table 3]
4.3
Alternative regional samples
This subsection examines whether a smaller or larger sample of the states that we consider for
the composition of the control and the treatment groups might change our main results. Recall
that our previous results are based on a sample with 99 affected banks and 178 unaffected
banks in Alabama, Florida, Louisiana, Mississippi, Texas, Georgia, Tennessee, Arkansas and
Oklahoma. For the robustness check in this section, we make the following changes: First,
we restrict the sample to banks that operate in Florida and Alabama only. The reason
is that only these states comprise both counties affected and counties unaffected by the
14
hurricane. Second, we restrict the sample to counties in the core states affected by the
hurricane (Louisiana, Mississippi, Texas, Florida, and Alabama) and thus exclude banks in
neighbor states (Georgia, Tennessee, Arkansas, and Oklahoma) from the control group relative
to our baseline sample. Third, we extend the sample to banks in neighboring U.S. Southeast
states, that is, South Carolina and North Carolina, and thus add banks from these states to
the control group.
We rerun our main regression and provide results for our baseline sample and the three
alternative regional samples in Table 4. Across all groups we find significant results for the
treatment effect from Hurricane Katrina on the total risk-based capital ratio of affected banks.
We also find that the effect is economically stronger for the core regions. Here, affected banks
increase their total risk-based capital ratio by nearly 2.1 percentage points relative to their
unaffected peers after the event. Considering the largest sample in the last column, we find
very similar results to our baseline regression in Table 2. Overall, Table 4 shows that our
results do not hinge on the choice of a specific control group.
[Table 4]
4.4
Potential credit demand effects
Because we are interested in bank behavior, we have to rule out that the results mainly reflect
a shortfall in credit demand after the hurricane, which technically would also lead to higher
regulatory capital ratios for banks in affected areas.12 To mitigate such concerns, we explore
the development of gross loans and different types of loans in counties affected and unaffected
by Hurricane Katrina. In particular, we consolidate data for all gross loans, commercial &
industrial loans, real estate loans, commercial real estate loans, consumer loans, and total
assets by quarter and by county comprising both data for independent and BHC banks.
Table 5 shows regression results for the collapsed sample for different loan categories (all
standardized by total assets) in columns 1 to 5. The regressions comprise all counties for
12
Remember that a bank’s total risk-based capital ratio is calculated as its total Tier 1 and Tier 2 capital
over its risk-weighted assets. Risk-weighted assets often substantially reflect consumer, commercial & industrial
loans.
15
the baseline sample for the period of ± 8 quarter around the event. Importantly, we do not
observe a significant effect for the ratio of total gross loans or the various loan categories.
Moreover, we investigate the total risk-based capital ratio for BHC banks around Hurricane
Katrina. If demand effects are relevant, we should detect an effect on the total risk-based
capital ratio for BHC banks, too. As shown in the last column of Table 5, we do not find
a significant effect from the hurricane on the total risk-based capital ratio of affected BHC
banks relative to their non-affected BHC peers. Thus, we are confident that our results are
not driven by demand effects.
[Table 5]
The Federal Reserve Bank of Atlanta even expected credit demand to increase in the aftermath
of the hurricane, as stated in their 2005 annual report (Federal Reserve Bank of Atlanta, 2005).
Thus, it is highly unlikely that a shortfall in credit demand is driving the results during the
two-year period that we examine.
4.5
4.5.1
Additional robustness tests
Parallel-trend assumption
To alleviate potential biases we have to guarantee that the parallel-trend assumption prior to
the treatment is satisfied. In other words, the total risk-based capital ratios should follow a
similar trend for the treatment and control groups. Analogous to previous studies, in Fig. 2
we graphically inspect the trend of mean total risk-based capital for both groups and confirm
the parallel-trend assumption. Further, as already discussed in Subsection 2.4 and shown in
Table 1, the groups of affected and unaffected banks largely do not differ significantly with
respect to common bank characteristics.
16
4.5.2
Cross-section estimation
In order to show that the results are robust against problems with difference-in-difference
technique in the presence of serial correlation, Bertrand et al. (2004) suggest ignoring the
time structure of the data. Therefore, we average the data before and after the hurricane and
rerun the estimation for this collapsed sample.
Table 6 presents results for the collapsed baseline sample over the four different time periods.
We find the treatment effect for all different periods intact and in the range of about 1 to
1.5 percentage points. As opposed to the panel results presented so far, the adjusted R2 is
much lower. When we rerun the estimation for alternative regional samples, as discussed in
Subsection 4.3, we again find that the effect of the hurricane on affected banks’ total risk-based
capital ratio is significant and positive (not reported).
[Table 6]
4.5.3
Time-placebo estimation
The possibility that the results are driven by time trends unrelated to Hurricane Katrina needs
to be ruled out. Therefore, we run a “placebo estimation” where the treatment shifts from
the Q3 and Q4 of 2005 to Q3 and Q4 of 2002. We then rerun the estimation for observations
two years before and after this “2002 pseudo hurricane” event. Table 7 shows the results for
this analysis, which can be directly compared to our baseline results in Table 2. We do not
find an effect for the 2002 pseudo hurricane in any of our baseline specifications. This finding
supports our assumption that our results are not driven by factors unrelated to Hurricane
Katrina.
[Table 7]
17
5
Behavior of low-capitalized and high-capitalized banks
Theory suggests that bank characteristics such as bankruptcy costs, franchise value, value of
deposit guarantees, and bank governance are important determinants for banks’ risk taking.
Accordingly, banks can choose to be more risky or more safe that is approximately reflected
in lower and higher capital ratios. Therefore, we examine subsamples of low-capitalized and
high-capitalized banks separately.
5.1
Subsample characteristics
In order to construct these subsamples, we first calculate the average total risk-based capital
ratio for each bank during the eight quarters pre-Katrina (Q3 2003 to Q2 2005). We then
calculate the corresponding median value of all banks, which is a total risk-based capital ratio
of 15.8%. Next, we classify all banks with an average below the median as low-capitalized and
all other banks as high-capitalized. Note that even in the low-capitalized subsample, more
than 95% of the banks held an average total risk-based capital ratio pre-Katrina above 10%,
which is well above the required 8% and considered as “well capitalized” by the FDIC.13
When we compare banks that are high-capitalized and low-capitalized we have to be careful
that within both subsamples affected and unaffected banks have similar characteristics before the hurricane. Therefore, Table 8 presents mean values and standard deviations for the
same bank characteristics and regional control variables explained in Table 1. Again, normalized differences check whether these characteristics differ significantly between affected and
unaffected banks in both subsamples.
Considering high-capitalized banks on the left side of the table there are no significant differences along all characteristics. If we further consider the sample of low capitalized banks,
we exclude the six largest low-capitalized banks from the subsample to guarantee similar
characteristics of affected and unaffected banks. Otherwise, we find that affected banks and
unaffected banks differ across several bank characteristics. Nevertheless, the following regres13
Further requirements to be classified as “well capitalized” by the FDIC are a Tier 1 risk-based capital
ratio equal to or greater than 6% and a Tier 1 leverage capital ratio equal to or greater than 5%.
18
sion results hold whether this adjustment is made or not. In the sample of low-capitalized
banks some differences are more pronounced but are not overall significant according to normalized differences. In particular, low-capitalized unaffected banks hold slightly higher shares
of total loans that are mirrored in the higher ratio of risk-weighted assets also. Overall, normalized differences confirm that the split in high- and low-capitalized banks does not produce
very different samples.
When we consider the immediate impact of Hurricane Katrina, we expect that affected banks
in both subsamples faced an uncertain economic environment and higher asset risks after
2005 as noted by the FDIC and rating agencies. However, accounting figures only reflect
significantly higher loan loss provisions for the subsample of affected high-capitalized banks.
One possible explanation is different accounting practices and use of regulatory forbearance
after the hurricane. Nevertheless, we are cautious about the interpretation of results for the
subsample of low-capitalized banks and focus on the behavior of high-capitalized banks.
[Table 8]
5.2
Estimation results
Along the sample split of low-capitalized and high-capitalized banks, we reestimate our baseline regression for both groups separately. Table 9 shows values for the interaction term
β3 , its standard errors and the overall fit for each regression based on four alternative time
horizons from ±4 to ±10 quarters around the hurricane (equivalent to Subsection 4.2) and
four alternative regional samples (equivalent to Subsection 4.3).14 We observe that banks
in our high-capitalized subsample significantly modify their capital ratios upward which is
in line with the results for the whole sample. This result holds over all regions and time
periods. Considering the baseline regression (± 8 quarters), we find an economic effect of
around 1.75 percentage points triggered by the hurricane event. For the other specifications,
the effect ranges from 1.32 percentage points to 4.10 percentage points. To the contrary and
consistently over all periods and control regions, low-capitalized banks do not significantly
alter their total risk-based capital ratios.
14
Full information tables can be provided upon request.
19
[Table 9]
The analysis yields an interesting result about how banks react to increased asset risks. Banks
that ex ante choose to be relatively conservative (the high-capitalized subsample) take higher
risks into account for their capital ratio adjustments. They seem to be driven by the motive
to avoid bankruptcy under all circumstances. Banks that ex ante choose to be more risky
(the low-capitalized subsample) do not adjust their capital ratios.
5.3
Robustness
We run several robustness checks along the analyses in the previous subsection for the subsample of high-capitalized and low-capitalized banks (not reported). In particular, we find
that if we consider a regression without bank fixed effects and a regression in which we add
further covariates as in Table 2 (column 3), the results in Table 9 remain intact. We also
find that high-capitalized banks increase their total risk-based capital ratios relative to their
unaffected peers after the event in cross-section regressions for all regions and time periods.
6
Mechanisms of capital ratio adjustments
This section examines the mechanisms that are responsible for the capital-ratio adjustments
of banks.
6.1
The role of capital level and risk-weighted assets
Banks can adjust their capital ratios in two ways. First, they can adjust their capital level
(the numerator of the ratio), for example, by raising new capital or by increasing the share of
retained earnings. Second, banks can cut back their risk-weighted assets (the denominator of
the ratio), for example, by restricting the provision of new loans. Therefore, we are interested
in whether banks adjust their total capital, their risk-weighted assets, or both.
20
Building on the results from the previous section that high-capitalized and low-capitalized
banks behave differently, we continue to examine these groups separately in this subsection.
We do not expect that low-capitalized banks significantly adjust their capital or risk-weighted
assets, but we include both groups in our analysis to get a better understanding of the full
picture.
The analysis starts with focusing on the numerator of banks’ total risk-based capital ratio, and
therefore uses total tier capital over assets as the dependent variable. We rerun the baseline
regression for high-capitalized and low-capitalized banks for the four different time horizons
previously reported. We find no significant effects for banks in either the low-capitalized group
or the high-capitalized group, as shown in Table 10. In unreported regressions we also check
whether we find an effect on dividends or retained earnings after the hurricane. Again, we do
not find any significant effect for either high-capitalized or low-capitalized affected banks.
[Table 10]
Next, we use risk-weighted assets over assets as the dependent variable and rerun our baseline
regression. As Table 10 shows, we find that affected banks of the high-capitalized group
decrease their risk-weighted assets significantly in the aftermath of the hurricane. This effect
is significant and economically very strong for the shortest time period, but it loses some
strength for longer observation periods. Ten quarters after the event, the effect for the
high-capitalized affected banks remains significant. These results corroborate our previous
findings from Section 5 that high-capitalized banks significantly increase their total risk-based
capital ratios and show that they do this by decreasing their risk-weighted assets, and not by
increasing their level of capital. These findings are in line with Thakor (1996) who suggests
that banks typically face high costs in raising equity. To meet the capital requirement, which
might well exceed the minimum requirement, a bank can constrain further lending. Instead
of extending lending activity a bank is likely to invest in marketable securities.
Results are also in line with our previous result that low-capitalized banks do not adjust their
capital ratios. In particular, we do not find a significant effect on the risk-weighted assets of
affected low-capitalized banks, as Table 10 also shows.
21
6.2
Bank lending
We further investigate whether a reduction of risk-weighted assets is associated with a reduction in lending. To investigate this issue, we first use banks’ gross loans over assets as the
dependent variable. As Table 10 shows, we find that high-capitalized affected banks decrease
loan business in the aftermath of the hurricane while results for low-capitalized banks are not
significant. The reduction of loans takes place within an equal time horizon and becomes less
significant for the longest period, ± 10 quarters around the event. These results suggest that
high-capitalized affected banks become more stable in terms of their total risk-based capital
ratio after the event by reducing their loan supply. This finding is in line with Garmaise and
Moskowitz (2009) who show that, in the context of earthquake risk, local banks are reluctant to offer loans to high earthquake-risk properties in the period after the 1994 Northridge
earthquake.
Next, we explore several other balance sheet figures that are related to a bank’s risk-weighted
assets. Interestingly, we find that high-capitalized affected banks increase their investments
in risk-free U.S. government securities post Hurricane Katrina, as the last columns of Table
10 indicate. Economically this effect is very similar to the reduction in loans to assets but
is positive, and thereby roughly offsets the loan reduction. A reduction in risk weights is
obviously driven by a decline in loan supply accompanied by investments in government
securities. Again, there is no such effect for low-capitalized banks.
These results are also meaningful as regards potential concerns that the observed reduction
of risk-weighted assets is demand driven, as already discussed in Subsection 4.4. Note that
a shortfall in demand is also present for low-capitalized banks that operate in the hurricane
regions. As shown in this subsection, we find no such significant effect for low-capitalized
affected banks. Thus, this finding further strengthens the interpretation that our results
reflect the behavior of banks and not the economic conditions in the hurricane regions.
22
6.3
Commercial & industrial, real estate and consumer loans
This set of regressions further investigates whether the overall reduction of loans by highcapitalized affected banks is consistent over different loan types. Therefore, Table 11 provides
results for the baseline regression that uses banks’ commercial and industrial loans, real estate
loans, commercial real estate loans and consumer loans as the dependent variable. We find
that the overall loan reduction of high-capitalized affected banks after the event stems from
less engagement in corporate and mortgage lending. Here the effect is negative and significant
for the shorter two periods around the event whereas we do not find an effect for the subsample
of low-capitalized banks. We also do not find an effect for commercial real estate loans and
consumer loans for both groups.
To conclude, high-capitalized banks that were affected by the Hurricane Katrina did not
significantly adjust their capital levels. However, they reacted to the higher asset risks by
shifting investments from risky loans to risk-free U.S. government securities and thereby
increased their capital ratios. We find that these effects are significant and economically
relevant over different periods around the events. Further our results even hold when we
consider different loan types. Here, banks reduced risky assets in the form of mortgage and
corporate lending in the phase after Hurricane Katrina.
[Table 11]
7
Conclusion
In this paper, we conduct the first empirical assessment of the impact of bank asset risks
on capital ratios using a natural experiment. Therefore, we examine the effects of Hurricane
Katrina that exposed banks in the U.S. Gulf Coast region to unexpected losses and increased
credit risks in August 2005. This examination allows us to identify a causal relation between
asset risk and capital ratios that is otherwise difficult because of mutual influences and feedback effects. A better understanding of this relation is important because capital ratios are
a key factor in banking regulation.
23
We find that a shock that increases banks’ credit risk, measured as the impact of Hurricane
Katrina, induces affected banks to increase their capital ratios relative to the banks in the
control group (not affected by the hurricane). Affected banks thereby strengthen their buffer
against future income shocks and mitigate bankruptcy risks. This strengthening shows that
asset risk is an important determinant of bank capital ratios besides any regulatory requirements.
However, when we examine low-capitalized and high-capitalized banks separately, we only find
a significant effect for high-capitalized banks. This effect demonstrates that the risk-taking
behavior of banks cannot be generalized for all banks but depends on bank characteristics.
Based on existing research, relevant bank characteristics could be a bank’s franchise value,
its bankruptcy costs, and/or the value of its deposit guarantees.
Furthermore, we find that the high-capitalized banks that increase their capital ratios do this
by reducing their risk-weighted assets and not by increasing their capital levels. In particular,
these banks shift their investments from risky loans to risk-free U.S. government securities.
This shift has ambiguous effects for the economy: Banks become more stable, which has
positive effects, but firms might suffer from a lower supply of credit.
24
References
Adrian, T. and Shin, H. (2011). Financial intermediary balance sheet management. Annual
Review of Financial Economics, 3.
Allen, F., Carletti, E. and Marquez, R. (2011). Credit market competition and capital
regulation. Review of Financial Studies, 24, 983–1018.
Angrist, J. D. and Pischke, J.-S. (2009). Mostly Harmless Econometrics. Princeton University Press.
Avery, R. B. and Berger, A. B. (1991). Risk-based capital and deposit insurance reform.
Journal of Banking & Finance, 15, 847–874.
Barth, J., Caprio, G. and Levine, R. (2005). Rethinking bank regulation: Till angels
govern. Cambridge and New York: Cambridge University Press.
Berger, A. and Bouwman, C. (2009). Bank liquidity creation. Review of Financial Studies,
22, 3779–3837.
Berger, A. N., Herring, R. J. and Szegö, G. P. (1995). The role of capital in financial
institutions. Journal of Banking & Finance, 19, 393–430.
Bertrand, M., Duflo, E. and Mullainathan, S. (2004). How much should we trust
differences-in-differences estimates? Quarterly Journal of Economics, 119, 249–275.
Campello, M., Graham, J. and Harvey, C. (2010). The real effects of financial constraints: Evidence from a financial crisis. Journal of Financial Economics, 97, 470–487.
Congleton, R. (2006). The story of Katrina: New Orleans and the political economy of
catastrophe. Public Choice, 127, 5–30.
Demsetz, R. S. and Strahan, E. P. (1997). Diversification, size, and risk at bank holding
companies. Journal of Money, Credit and Banking, 29, 300–313.
DeYoung, R., Hunter, W. C. and Udell, G. F. (2004). The past, present, and probable
future for community banks. Journal of Financial Services Research, 25, 85–133.
25
Diamond, D. and Rajan, R. G. (2000). A theory of bank capital. The Journal of Finance,
55, 2431–2465.
Duchin, R., Ozbas, O. and Sensoy, B. (2010). Costly external finance, corporate investment, and the subprime mortgage credit crisis. Journal of Financial Economics, 97,
418–435.
FDIC
(2006).
pervisory
Hurricane
Guidance
Katrina
for
Examiner
Institutions
Guidance.
Affected
by
Interagency
Hurricane
Su-
Katrina.
http://www.fdic.gov/news/news/financial/2006/fil06012.html.
Federal Reserve Bank of Atlanta (2005). Annual report 2005: Resilience.
FEMA
(2008).
Prepared.
responsive.
committed.
URL
http://www.fema.gov/pdf/about/brochure (Download June, 2011).
Flannery, M. J. and Rangan, K. P. (2008). What caused the bank capital build-up of
the 1990s? Review of Finance, 12, 391.
Froot, K. A. and Stein, J. C. (1998). Risk management: capital budgeting and capital
structure policy for financial institutions: an integrated approach. Journal of Financial
Economics, 47, 55–82.
Furlong, F. and Keeley, M. (1989). Bank capital regulation and risk taking: a note.
Journal of Banking & Finance, 13, 883–91.
Gale, D. (2010). Capital regulation and risk sharing. International Journal of Central Banking, 23, 187–204.
Gan, J. (2007). Collateral, debt capacity, and corporate investment: Evidence from a natural
experiment. Journal of Financial Economics, 85, 709–734.
Garmaise, M. and Moskowitz, T. (2009). Catastrophic risk and credit markets. The
Journal of Finance, 64, 657–707.
Gropp, R. and Heider, F. (2010). The determinants of bank capital structure. Review of
Finance, 14, 587–622.
26
Guy Carpenter (2006). Rating agency update:
Stepping up to new criteria. URL
http://www.guycarp.com/portal/extranet/pdf/GCPub/Rating Agency Update 2006.pdf
(Download December, 2011).
Hellmann, T., Murdock, K. and Stiglitz, J. (2000). Liberalization, moral hazard in
banking, and prudential regulation. American Economic Review, 90 (1), 147–165.
Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical
Association, 81, 945–960.
Houston, J., James, D. and Marcus, D. (1997). Capital market frictions and the role of
internal capital markets in banking. Journal of Financial Economics, 46, 135–164.
Imbens, G. W. and Wooldridge, J. M. (2009). Recent developments in the econometrics
of program evaluation. Journal of Economic Literature, 47, 5–86.
Ivashina, V. and Scharfstein, D. (2010). Bank lending during the financial crisis of 2008.
Journal of Financial Economics, 97, 319–338.
Keeley, M. (1990). Deposit insurance, risk and market power in banking. American Economic Review, 80, 1183–1200.
Kim, D. and Santomero, A. (1988). Risk in banking and capital regulation. Journal of
Finance, 43, 1219–1233.
Laeven, L. and Levine, R. (2009). Bank governance, regulation and risk taking. Journal
of Financial Economics, 93, 259–275.
Marcus, A. J. (1984). Deregulation and bank financial policy. Journal of Banking & Finance,
8, 557–565.
Moody’s (2005). Moody’s comments on Hurricane Katrina’s impact on regional bank ratings.
Global Credit Research, 08 Sep 2005.
Puri, M., Rocholl, J. and Steffen, S. (2010). Global retail lending in the aftermath
of the us financial crisis: Distinguishing between supply and demand effects. Journal of
Financial Economics.
27
Santos, J. (2011). Bank corporate loan pricing following the subprime crisis. Review of
Financial Studies, 24, 1916.
Shrieves, R. E. and Dahl, D. (1992). The relationship between risk and capital in commercial banks. Journal of Banking & Finance, 16, 439–457.
Thakor, A. V. (1996). Capital requirements, monetary policy, and aggregate bank lending.
Theory and empirical evidence. The Journal of Finance, 51, 279–324.
USA TODAY (2010). 5 years after Katrina, homeowners insurance costs more (August 26).
URL http://www.usatoday.com/money/industries/insurance/ (Download 18 April 2011).
28
A
Appendix: Figures and Tables
−.005
0
Provisions for loans/Loans
.005 .01 .015 .02 .025
.03
Figure 1: Impact of Hurricane Katrina on the Gulf Coast banking market
2001q2
2002q4
2004q2
2005q4
Affected banks (mean)
2007q2
2008q4
2010q2
Unaffected banks (mean)
This figure shows the development of mean values of annualized loan loss provisions (to loans) for the first
quarter of 2000 to the fourth quarter of 2010. The shaded area illustrates Q3 and Q4 of 2005 when Hurricane
Katrina hit the U.S. Gulf Coast region. The mean value for independent banks located in areas affected by
the hurricane are represented by a solid line. The mean values for independent banks located in the U.S. Gulf
Coast region but not affected by the hurricane are represented by a dotted line.
29
Total risk−based capital ratio
.15 .16 .17 .18 .19 .2 .21 .22 .23 .24 .25
Figure 2: Total risk-based capital ratios
2001q2
2002q4
2004q2
2005q4
Affected banks (mean)
2007q2
2008q4
2010q2
Unaffected banks (mean)
This figure shows the development of the mean values of total risk-based capital ratios for the first quarter of
2000 to the fourth quarter of 2010. The shaded area illustrates Q3 and Q4 of 2005 when Hurricane Katrina hit
the U.S. Gulf Coast region. The mean value for independent banks located in areas affected by the hurricane
are represented by a solid line. The mean values for independent banks located in the U.S. Gulf Coast region
but not affected by the hurricane are represented by a dotted line.
30
Figure 3: 2005 hurricane disaster areas
Oklahoma
Tennessee
Arkansas
Georgia
Louisiana
Alabama
Mississippi
Florida
Texas
Treatment group
Control group
Major city
USA
This figure shows counties in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas,
and Oklahoma that Hurricane Katrina affected. The dark-grey shaded area comprises counties that were
eligible for individual and public disaster assistance. The light-grey shaded area comprises counties that did
not get disaster assistance. The white shaded area includes counties that were eligible only for public disaster
assistance.
31
Table 1: Descriptive statistics
This table reports descriptive statistics for all variables used in later analyses for the period two years before the hurricane
event in 2005. We provide mean values and standard deviations for independent banks (i.e., they do not belong to a BHC)
that belong to counties that were affected by the hurricane (affected), banks operating in counties unaffected by the event
(unaffected), and also for affected banks that belong to a BHC. The sample consists of 99 affected and 178 unaffected
independent banks and 222 affected BHC banks. The sample includes all banks in Alabama, Louisiana, Mississippi,
Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma. The last two columns show the normalized differences
(Norm. Diff.) according to Imbens and Wooldridge (2009) and compare differences between independent banks that were
affected versus banks that were not affected ((1) vs (2)) and affected independent banks with affected BHC banks ((1)
vs (3)). As a rule of thumb, values between ± 0.25 are equivalent to nonsignificant differences between groups. Variables
are defined as follows: Total risk-based capital ratio represents banks’ total risk-based capital to its risk-weighted assets;
Risk-weighted assets/Assets is the ratio of risk-weighted assets to total assets; Tier capital/Assets relates the sum of a
bank’s Tier 1 and Tier 2 capital to total assets; Risk-weighted assets represent the dollar amount of banks’ risk-weighted
assets; Total Tier capital is the dollar amount of the sum of banks’ Tier 1 and Tier 2 capital; Total assets are banks’
total assets; Total loans/Assets is the ratio of banks’ total (gross) loans to total assets; C&I loans/Assets is the ratio
of banks’ corporate and industrial loans to total assets; Real estate loans/Assets is the ratio of banks’ mortgage loans
to total assets; Commercial RE loans/Assets is the ratio of banks’ nonresidential loans secured by real estate to total
assets; Consumer loans/Assets is the ratio of consumer loans to total assets; U.S. Gov. securities/Assets is the ratio of
banks’ U.S. government securities to total assets; RoA is banks’ return over assets; Non-perf. loans/Assets shows the
ratio of total assets past due 90 or more days and still accruing interest to banks’ total assets; Provisions for loans/Loans
is the ratio of loan loss provisions to total loans; No. of employees is the number of full-time employees on the payroll of
the bank and its subsidiaries at the end of the quarter; Share of commercial banks is a dummy variable that equals one
for commercial banks and zero otherwise; Share of stock institutions is a dummy variable that equals one if a bank is a
stock institution or zero otherwise; Regional unemployment rate shows the unemployment rate per county per quarter.
The U.S. $ values are denoted in millions. Furthermore, a very detailed description of all variables with FDIC codes is
given in Table 12.
Total risk-based capital ratio
Risk-weighted assets/Assets
Tier capital/Assets
Risk-weighted assets
Total tier capital
Total assets
Total loans/Assets
C&I loans/Assets
Real estate loans/Assets
Commercial RE loans/Assets
Consumer loans/Assets
U.S. Gov. securities/Assets
RoA
Non-perf. loans/Assets
Provisions for loans /Loans
No. of employess
Share of commerical banks
Share of stock institutions
Regional unemployment rate
Number of banks
Number of observations
Independent
(1) affected
(2) unaffected
Mean
SD
Mean
SD
0.1777
0.0649 0.1772 0.0667
0.6305
0.1302 0.6802 0.1285
0.1071
0.0312 0.1152 0.0326
247.90
647.82
126.88 357.50
36.15
87.01
20.78
57.51
403.52 1098.99 209.48 745.01
0.6332
0.1800 0.6672 0.1631
0.0761
0.0688
0.0876 0.0696
0.4956
0.2117
0.4870 0.1896
0.1618
0.1285 0.1466
0.1127
0.0495
0.0549 0.0698 0.0594
0.1875
0.1562 0.1750 0.1503
0.0094
0.0096 0.0093 0.0111
0.0013
0.0029 0.0015 0.0029
0.0038
0.0085 0.0043 0.0087
96.19
203.61
54.35 111.62
0.6486
0.4777
0.6914 0.4621
0.9419
0.2342
0.9250 0.2635
0.0533
0.0168 0.0543 0.0156
99
178
774
1387
32
BHC
(3) affected
Mean
SD
0.1606
0.0550
0.6711
0.1300
0.1027
0.0224
407.52 1325.93
52.23
147.35
565.19 1720.25
0.6230
0.1570
0.0976
0.0735
0.4341
0.1666
0.1806
0.1021
0.0705
0.0592
0.1938
0.1311
0.0111
0.0075
0.0014
0.0030
0.0033
0.0097
175.62
482.65
0.9074
0.2900
1.0000
0.0000
0.0586
0.0197
222
1760
Norm. Diff.
(1) vs (2) (1) vs (3)
0.01
0.20
-0.27
-0.22
-0.18
0.11
0.16
-0.11
0.15
-0.09
0.15
-0.08
-0.14
0.04
-0.12
-0.21
0.03
0.23
0.089
-0.11
-0.25
-0.26
0.06
-0.03
0.01
-0.14
-0.05
-0.02
-0.04
0.04
0.18
-0.15
-0.06
-0.46
0.05
-0.25
-0.04
-0.20
Table 2: Main results
This table shows results for regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent variable.
Regressions presented in this table are for the period of ± 2 years around the event. The sample includes all banks
in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma that are independent,
i.e., do not belong to a BHC. Event is a dummy variable that is zero for the pre-hurricane period and one after the event.
Affected is a dummy variable that separates banks in counties that were affected by the hurricane and has a value of one
and those that were unaffected have a value of zero. Event*Affected is an interaction term for the variables Event and
Affected. RoA is banks’ return on assets. Non-perf. loans/Assets shows the ratio of total assets past due 90 or more
days and still accruing interest to banks’ total assets, and Log (No. of employees) is the natural logarithm of banks’
number of employees. Regional unemployment rate represents the quarterly unemployment rate for each quarter. A
very detailed description of all variables with FDIC codes is given in Table 12. We also include bank fixed effects (Bank
FE) and year dummies (Time FE) in the regressions. We show clustered standard errors on bank level in parentheses.
The ***, ** and * stand for significant coefficients at the 1%, 5%, and 10% levels respectively.
Total risk-based capital ratio
-0.0034 -0.0103***
-0.0044
(0.0027)
(0.0029)
(0.0029)
0.0005
(0.0078)
0.0122**
0.0108**
0.0105**
(0.0054)
(0.0047)
(0.0043)
-0.1085
(0.2756)
0.1544
(0.2287)
-0.0484***
(0.0101)
0.1409**
(0.0669)
0.1739***
0.1772***
0.3476***
(0.0047)
(0.0011)
(0.0375)
No
Yes
Yes
Yes
Yes
Yes
4138
4138
4137
0.0034
0.8538
0.8654
Event
Affected
Event*Affected
RoA
Non-perf. loans/Assets
Log(No. of employess)
Regional unemployment rate
Constant
Bank FE
Time FE
Number of observations
Adj. R2
33
Table 3: Alternative time horizons
This table shows the results for regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent
variable. The sample includes all banks in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas
and Oklahoma that are independent, i.e., do not belong to a BHC. Event is a dummy variable that is zero for the prehurricane period and one after the event. Affected is a dummy variable that separates banks in counties that were
affected by the hurricane and has a value of one and those that were unaffected have a value of zero. Event*Affected
is an interaction term for the variables Event and Affected. Regressions presented in this table are for the periods of ±
4, 6, 8, and 10 quarters around the event. We include bank fixed effects (Bank FE) and year dummies (Time FE) in
each regression. We show clustered standard errors on bank level in parentheses. The ***, ** and * stand for significant
coefficients at the 1%, 5%, and 10% levels respectively.
Event
Event*Affected
Constant
Bank FE
Time FE
Number of observations
Adj. R2
± 4 quarter
-0.0066***
(0.0022)
0.0071*
(0.0041)
0.1767***
(0.0010)
Yes
Yes
2153
0.9101
Total risk-based capital ratio
± 6 quarter ± 8 quarter ± 10 quarter
-0.0107***
-0.0103***
-0.0156***
(0.0028)
(0.0029)
(0.0036)
0.0091**
0.0108**
0.0108**
(0.0045)
(0.0047)
(0.0049)
0.1807***
0.1772***
0.1831***
(0.0013)
(0.0011)
(0.0017)
Yes
Yes
Yes
Yes
Yes
Yes
3161
4138
5085
0.8789
0.8538
0.8383
Table 4: Alternative regional samples
This table shows results for regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent
variable. Regressions presented in this table are for different samples: AL & FL shows results for banks in Florida
and Alabama only; Core states comprise counties in Alabama, Louisiana, Mississippi and Florida; Baseline includes
counties in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma; Southeast
U.S. comprises all counties in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas, Oklahoma,
Georgia, North Carolina and South Carolina. All banks are independent, i.e., do not belong to a BHC. Regressions
presented in this table are for the period of ± 2 years around the event. Event is a dummy variable that is zero for the
pre-hurricane period and one after the event. Affected is a dummy variable that separates banks in counties that were
affected by the hurricane and has a value of one and those that were unaffected have a value of zero. Event*Affected is
an interaction term for the variables Event and Affected. We include bank fixed effects (Bank FE) and year dummies
(Time FE) in each regression. We show clustered standard errors on bank level in parentheses. The ***, ** and * stand
for significant coefficients at the 1%, 5%, and 10% levels respectively.
Event
Event*Affected
Constant
Bank FE
Time FE
Number of observations
Adj. R2
AL & FL
-0.0241***
(0.0060)
0.0215***
(0.0077)
0.1706***
(0.0029)
Yes
Yes
1536
0.7880
34
Total risk-based capital ratio
Core states
Baseline
Southeast U.S.
-0.0225*** -0.0103***
-0.0108***
(0.0064)
(0.0029)
(0.0023)
0.0209***
0.0108**
0.0109**
(0.0070)
(0.0047)
(0.0044)
0.1806***
0.1772***
0.1777***
(0.0024)
(0.0011)
(0.0010)
Yes
Yes
Yes
Yes
Yes
Yes
2123
4138
4917
0.8415
0.8538
0.8677
35
Table 5: Potential credit demand effects
Bank FE
Time FE
County FE
Number of observations
adj. R2
Constant
Event*Affected
Event
Total loans/Assets
0.0244***
(0.0036)
-0.0128
(0.0083)
0.6384***
(0.0022)
No
Yes
Yes
6303
0.8533
All Banks (Independent + BHC banks) collapsed per county and quarter
C&I loans/Assets
Real estate loans/Assets
Commercial RE loans/Assets
-0.0048**
0.0471***
0.0066***
(0.0019)
(0.0036)
(0.0025)
-0.0014
-0.0100
-0.0029
(0.0044)
(0.0077)
(0.0044)
0.0918***
0.4379***
0.1528***
(0.0011)
(0.0021)
(0.0015)
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
6303
6303
6303
0.7923
0.9009
0.8435
Consumer loans
-0.0144***
(0.0010)
-0.0007
(0.0019)
0.0752***
(0.0007)
No
Yes
Yes
6303
0.891
BHC Banks
Total risk-based capital ratio
-0.0060***
(0.0011)
0.0024
(0.0020)
0.1584***
(0.0007)
Yes
Yes
No
16078
0.8909
This table shows results for regressions of Eq. (3) for two different sample. In the first five columns we use a collapsed sample to investigate the effect of Hurricane Katrina on the sum of total gross
loans (over assets) per county and quarter and four different loan categories: commercial and industrial loans, real estate loans, commercial real estate loans and consumer loans all standardized by
all banks assets per count per quarter. The collapsed sample comprises independent and BHC banks. The last column considers BHC banks only and presents results for a regression of Eq. (3) in
which banks’ total risk-based capital ratio is the dependent variable. The sample includes all banks in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma.
Event is a dummy variable that is zero for the pre-hurricane period and one after the event. Affected is a dummy variable that separates banks in counties that were affected by the hurricane and has
a value of one and those that were unaffected have a value of zero. Event*Affected is a interaction term for the variables Event and affected. We consider the period of ± 8 quarters around the event.
We also include bank fixed effects (Bank FE), county fixed effects (County FE) and year dummies (Time FE) in the regressions. We show clustered standard errors on bank level in parentheses. The
***, ** and * stand for significant coefficients at the 1%, 5%, and 10% levels respectively.
Table 6: Cross section results
This table shows results for cross-section regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent variable. The sample includes all banks in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee,
Arkansas and Oklahoma that are independent, i.e., do not belong to a BHC. Event is a dummy variable that is zero for
the pre-hurricane period and one after the event. Affected is a dummy variable that separates banks in counties that
were affected by the hurricane and has a value of one and those that were unaffected have a value of zero. Event*Affected
is a interaction term for the variables Event and affected. Cross-sections for each time period (i.e., for the period of ±
4, 6, 8, and 10 quarters around the event) comprise collapsed mean values for each bank before and after the event for
each period respectively. We show clustered standard errors on bank level in parentheses. The ***, ** and * stand for
significant coefficients at the 1%, 5%, and 10% levels respectively.
Event
Event*Affected
Affected
Constant
Number of observations
Adj. R2
± 4 quarter
-0.0058**
(0.0024)
0.0097**
(0.0049)
0.0038
(0.0084)
0.1752***
(0.0050)
558
0.0003
36
Total risk-based capital ratio
± 6 quarter ± 8 quarter ± 10 quarter
-0.0092***
-0.0098***
-0.0119***
(0.0026)
(0.0029)
(0.0030)
0.0126**
0.0141***
0.0145***
(0.0051)
(0.0054)
(0.0055)
0.0009
-0.0011
-0.0014
(0.0083)
(0.0081)
(0.0080)
0.1778***
0.1797***
0.1805***
(0.0049)
(0.0050)
(0.0049)
562
566
569
0.0009
0.0008
0.0025
Table 7: Placebo event
This table presents results for regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent
variable. Regressions presented in this table are for the period of ± 2 years around the event. The sample includes
all banks in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma that are
independent, i.e., do not belong to a BHC. Event is a dummy variable that is zero before the third quarter of 2002 and
one after the fourth quarter of 2002 (the placebo event). Affected is a dummy variable that separates banks in counties
that were affected by the hurricanes and equals one and those that were unaffected equal zero. Event*Affected is an
interaction term for the variables Event and Affected. All other variables are defined as in Table 2. We include bank
fixed effects (Bank FE) and year dummies (Time FE) in each regression. We show clustered standard errors on bank
level in parentheses. The ***, ** and * stand for significant coefficients at the 1%, 5%, and 10% levels respectively.
Total risk-based capital ratio
-0.0030 -0.0160***
-0.0074
(0.0034)
(0.0047)
(0.0053)
-0.0004
(0.0090)
-0.0005
0.0018
0.0011
(0.0057)
(0.0055)
(0.0053)
-0.5564**
(0.2619)
-0.2848
(0.2814)
-0.0396**
(0.0156)
0.0856
(0.0577)
0.1836***
0.1909*** 0.3299***
(0.0053)
(0.0026)
(0.0540)
No
Yes
Yes
Yes
Yes
Yes
3852
3852
3852
0.0016
0.8157
0.8297
Event
Affected
Event*Affected
RoA
Non-perf. loans/Assets
Log(No. of employess)
Regional unemployment rate
Constant
Bank FE
Time FE
Number of observations
Adj. R2
37
38
Total risk-based capital ratio
Risk-weighted assets/Assets
Tier capital/Assets
Risk-weighted assets
Total tier capital
Total assets
Total loans/Assets
C&I loans/Assets
Real estate loans/Assets
Commercial RE loans/Assets
Consumer loans/Assets
U.S. Gov. securities/Assets
RoA
Non-perf. loans/Assets
Provisions for loans /Loans
No. of employees
Share of commercial banks
Share of stock institutions
Regional unemployment rate
Number of banks
Number of observations
High-capitalized banks
affected (1)
unaffected (2)
Mean
SD
Mean
SD
0.2258 0.0593
0.2270
0.0608
0.5662 0.1231
0.6042
0.1160
0.1246 0.0330
0.1341
0.0337
108.45 219.02
124.69
476.25
23.06
41.63
25.40
78.60
190.20 376.70
240.50
1019.81
0.5511 0.2012
0.5676
0.1614
0.0754 0.0782
0.0785
0.0674
0.4026 0.2200
0.3878
0.1744
0.1215 0.1056
0.1085
0.0933
0.0575 0.0497
0.0759
0.0684
0.2498 0.1810
0.2518
0.1682
0.0100 0.0090
0.0089
0.0133
0.0015 0.0029
0.0015
0.0027
0.0043 0.0085
0.0040
0.0102
56.22
89.71
49.05
115.62
0.6527 0.4767
0.7235
0.4476
0.8825 0.3224
0.9176
0.2751
0.0575 0.0174
0.0559
0.0165
49
89
383
680
Norm. Diff.
(1) vs (2)
-0.01
-0.22
-0.20
-0.03
-0.03
-0.05
-0.06
-0.03
0.05
0.09
-0.22
-0.01
0.07
0.00
0.02
0.05
-0.11
-0.08
0.07
Low-capitalized banks
affected (3)
unaffected (4)
Mean
SD
Mean
SD
0.1312
0.0201
0.1294
0.0221
0.7045
0.0976
0.7532
0.0926
0.0918
0.0160
0.0970
0.0179
134.95
113.92
128.99
181.00
17.22
13.94
16.34
22.62
195.47
171.08
179.64
295.95
0.7100
0.1039
0.7631
0.0918
0.0831
0.0587
0.0963
0.0707
0.5751
0.1541
0.5824
0.1508
0.2187
0.1339
0.1832
0.1175
0.0428
0.0607
0.0639
0.0485
0.1259
0.0876
0.1011
0.0773
0.0089
0.0108
0.0097
0.0085
0.0013
0.0030
0.0015
0.0031
0.0036
0.0089
0.0045
0.0070
61.63
53.27
59.44
107.47
0.7262
0.4465
0.6605
0.4739
1.0000
0.0000
0.9321
0.2517
0.0497
0.0156
0.0529
0.0145
44
89
347
707
Norm. Diff.
(3) vs (4)
0.06
-0.36
-0.22
0.03
0.03
0.05
-0.38
-0.14
-0.03
0.20
-0.27
0.21
-0.06
-0.05
-0.08
0.02
0.10
0.27
-0.15
This table shows descriptive statistics for low-capitalized and high-capitalized banks. The Total risk-based capital ratio of low-capitalized banks is below the sample median
before the hurricane event. The Total risk-based capital ratio of high-capitalized banks is above the median before the hurricane event. We provide mean values and standard
deviations for independent banks (i.e., they do not belong to a BHC) that belong to counties that were affected by the hurricanes (affected) and banks operating in counties
unaffected by the event (unaffected). ND is a short hand and shows normalized differences according to Imbens and Wooldridge (2009) that compares differences between
affected and unaffected banks for high-capitalized ((1) vs (2)) or low-capitalized ((1) vs (3)). As a rule of thumb, values around ± 0.25 relate to nonsignificant differences
between groups. Variables are defined as in Table 2. The sample consists of 118 affected and 199 unaffected independent banks. 51 are affected and high-capitalized and 57
are affected and low-capitalized. 93 are not affected and high-capitalized and 90 are not affected and low-capitalized The sample includes all banks in Alabama, Louisiana,
Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma, U.S. $ values are denoted in millions. Furthermore, a very detailed description of all variables
with FDIC codes is given in Table 12.
Table 8: Descriptive statistics for high-capitalized and low-capitalized banks
39
4 quarter
6 quarter
8 quarter
10 quarter
4 quarter
6 quarter
8 quarter
10 quarter
±
±
±
±
±
±
±
±
Adj. R2
0.7759
0.6757
0.6484
0.6495
Adj. R2
0.4476
0.4584
0.4532
0.4063
AL & FL
SE
(0.0128)
(0.0167)
(0.0177)
(0.0181)
AL & FL
SE
(0.0075)
(0.0077)
(0.0076)
(0.0080)
β3
0.0299**
0.0351**
0.0396**
0.0410**
β3
0.0035
0.0060
0.0084
0.0079
β3
0.0022
0.0039
0.006
0.006
β3
0.0233***
0.0302***
0.0353***
0.0374***
High-capitalized banks
Total risk-based capital ratio
Core states
Baseline
SE
Adj. R2
β3
SE
(0.0084)
0.865
0.0132**
(0.0060)
(0.0101)
0.7974
0.0156**
(0.0070)
(0.0105)
0.763
0.0175**
(0.0073)
(0.0110)
0.7412
0.0179**
(0.0077)
Low-capitalized banks
Total risk-based capital ratio
Core states
Baseline
SE
Adj. R2
β3
SE
(0.0061)
0.5127
0.0025
(0.0058)
(0.0061)
0.5004
0.0040
(0.0057)
(0.0061)
0.4902
0.0057
(0.0059)
(0.0064)
0.4457
0.006 (0.0062)
Adj. R2
0.5651
0.5097
0.4578
0.4132
Adj. R2
0.8814
0.8233
0.7858
0.7538
Southeast U.S.
β3
SE
Adj. R2
0.0034
(0.0056)
0.5789
0.0048
(0.0056)
0.5235
0.0063
(0.0057)
0.4712
0.0068
(0.0060)
0.415
Southeast U.S.
β3
SE
Adj. R2
0.0132**
(0.0056)
0.9002
0.0151**
(0.0065)
0.8502
0.0165**
(0.0068)
0.8155
0.0172**
(0.0072)
0.7829
This table shows results for regressions of Eq. (3) in which banks’ total risk-based capital ratio is the dependent variable. Regressions presented in this table are for the
periods of ± 4, 6, 8, and 10 quarters around the event for the four different regions explained in Table 4. We only show results for the interaction term (Event*Affected)
β3 for each regression (full information can be obtained upon request). The top panel of this table includes banks that are high-capitalized banks, i.e., their level of total
risk-based capital ratios is above the sample median before the hurricane event. The lower panel comprises all banks that show ratios below the median before the hurricane
event. All variables are defined as in Table 2. We include bank fixed effects (Bank FE) and year dummies (Time FE) in each regression. We show clustered standard errors
on bank level in parentheses. The ***, ** and * stand for significant coefficients at the 1%, 5%, and 10% levels respectively.
Table 9: Results for total risk-based capital ratios (high- vs. low-capitalized banks)
40
4 quarter
6 quarter
8 quarter
10 quarter
4 quarter
6 quarter
8 quarter
10 quarter
±
±
±
±
±
±
±
±
Tier capital/Assets
β3
SE
Adj. R2
0.0016
(0.0037)
0.6891
0.0026
(0.0036)
0.6651
0.0039
(0.0036)
0.6483
0.0041
(0.0036)
0.5999
Tier capital/Assets
β3
SE
Adj. R2
0.0020
(0.0037)
0.8104
0.0037
(0.0042)
0.7564
0.0051
(0.0041)
0.7346
0.0056
(0.0041)
0.7034
High-capitalized banks
Risk-weighted assets/Assets
Total
β3
SE
Adj. R2
β3
-0.0345*** (0.0103)
0.9180
-0.0411***
-0.0338*** (0.0117)
0.8910
-0.0410***
-0.0320** (0.0129)
0.8712
-0.0365**
-0.0283** (0.0140)
0.8402
-0.0308*
Low-capitalized banks
Risk-weighted assets/Assets
Total
β3
SE
Adj. R2
β3
0.0058 (0.0125)
0.8468
-0.0043
0.0048 (0.0131)
0.8185
-0.0015
0.0049 (0.0142)
0.7875
0.0019
0.0045 (0.0147)
0.7693
0.0036
loans/Assets
SE
Adj. R2
(0.0162)
0.7766
(0.0174)
0.7457
(0.0190)
0.7139
(0.0207)
0.6843
loans/Assets
SE
Adj. R2
(0.0120)
0.9479
(0.0136)
0.9314
(0.0149)
0.9176
(0.0160)
0.8994
U.S. Gov.
β3
-0.0001
-0.0008
-0.0006
0.0022
U.S. Gov.
β3
0.0371***
0.0388***
0.0373**
0.0349**
securities/Assets
SE
Adj. R2
(0.0126)
0.7727
(0.0133)
0.7400
(0.0144)
0.7062
(0.0151)
0.6864
securities/Assets
SE
Adj. R2
(0.0122)
0.9316
(0.0138)
0.9149
(0.0148)
0.9012
(0.0157)
0.8835
This table shows results for regressions of Eq. (3) in which the banks’ Tier capital/Asset ratio, the Risk-weighted assets/Assets ratio, the ratio of banks’ total loans to assets,
and the ratio of banks’ U.S. government securities are the dependent variables. Regressions presented in this table are for the periods m 4, 6, 8, and 10 quarters around the
event. We only show results for the interaction term (Event*Affected) β3 for each regression (full information can be obtained upon request). The sample includes all banks
in Alabama, Louisiana, Mississippi, Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma that are independent, i.e., do not belong to a BHC. We include bank fixed
effects (Bank FE) and year dummies (Time FE) in each regression. We show clustered standard errors on bank level in parentheses. The ***, ** and * stand for significant
coefficients at the 1%, 5%, and 10% levels respectively.
Table 10: Results for total risk-based capital ratio components and major bank assets (high- vs. low-capitalized banks)
41
4 quarter
6 quarter
8 quarter
10 quarter
4 quarter
6 quarter
8 quarter
10 quarter
±
±
±
±
±
±
±
±
Adj. R2
0.9113
0.8924
0.8669
0.8443
Adj. R2
0.9026
0.8903
0.8810
0.8708
C&I loans/Assets
Event*Impacted
SE
-0.0141**
(0.0063)
-0.0134*
(0.0070)
-0.0121
(0.0078)
-0.0089
(0.0080)
C&I loans/Assets
Event*Impacted
SE
-0.0002
(0.0051)
-0.0008
(0.0054)
-0.0010
(0.0059)
-0.0006
(0.0063)
Real estate
Event*Impacted
0.0051
0.0088
0.0101
0.01
Real estate
Event*Impacted
-0.0242**
-0.0241**
-0.0206
-0.0183
loans/Assets
SE
Adj. R2
(0.0144)
0.9153
(0.0153)
0.8995
(0.0166)
0.8888
(0.0181)
0.8737
Commercial RE loans/Assets
Event*Impacted
SE
Adj. R2
0.0008
(0.0060)
0.9505
0.0023
(0.0070)
0.9353
0.0019
(0.0077)
0.9208
0.0011
(0.0080)
0.9106
High-capitalized banks
loans/Assets
Commercial RE loans/Assets
SE
Adj. R2
Event*Impacted
SE
Adj. R2
(0.0105)
0.9658
-0.0017
(0.0072)
0.9522
(0.0120)
0.9540
0.0021
(0.0080)
0.9390
(0.0133)
0.9440
0.0045
(0.0088)
0.9330
(0.0142)
0.9338
0.0078
(0.0095)
0.9242
Low-capitalized banks
Consumer
Event*Impacted
-0.0019
-0.0024
-0.0014
0.0006
Consumer
Event*Impacted
-0.0024
-0.0029
-0.0035
-0.0040
loans/Assets
SE
Adj. R2
(0.0029)
0.9573
(0.0033)
0.9440
(0.0036)
0.9301
(0.0041)
0.9155
loans/Assets
SE
Adj. R2
(0.0027)
0.9755
(0.0032)
0.9632
(0.0037)
0.9525
(0.0044)
0.9370
This table shows results for regressions of Eq. (3) in which banks’ ratio of C&I loans to assets, the ratio of banks’ real-estate loans to total assets, the ratio of commercial real estate loans to totla
assets, and the ratio of total consumer loans to total assets are the dependent variables. Regressions presented in this table are for the periods ± 4, 6, 8, and 10 quarters around the event. We
only show results for the interaction term (Event*Affected) β3 for each regression (full information can be obtained upon request). The sample includes all banks in Alabama, Louisiana, Mississippi,
Florida, Texas, Georgia, Tennessee, Arkansas and Oklahoma that are independent, i.e., do not belong to a BHC. We include bank fixed effects (Bank FE) and year dummies (Time FE) in each
regression. We show clustered standard errors on bank level in parentheses. The ***, ** and * stand for significant coefficients at the 1%, 5%, and 10% levels respectively.
Table 11: Results for different loan types (high- vs. low-capitalized banks)
Table 12: Variable description
Notes: The source for all variables as well as their descriptions is the FDIC. For more details, refer to
http://www2.fdic.gov/SDI/main.asp.
Variable name
FDIC code
Description
Total risk-based capital ratio
rbcrwaj
Tier 1 capital and Tier 2 capital divided by the bank’s riskweighted assets
Risk-weighted assets
rwaj
Assets adjusted for risk-based capital definitions that comprise on-balance-sheet as well as off-balance-sheet items multiplied by risk weights that range from 0 to 100% (under
Basel I).
Total assets
asset
The sum of all assets owned by the institution including
cash, loans, securities, bank premises and other assets. This
total does not include off-balance-sheet accounts.
Tier capital
(rbct1j+rbct2)
Tier 1 capital and Tier 2 capital.
RoA
roa
Net income after taxes and extraordinary items (annualized)
to average total assets.
Non-perf. loans
p9asset
Total assets past due 90 or more days and still accruing
interest.
Provisions for loans
elnatr
The amount needed to make the allowance for loan and lease
losses adequate to absorb expected loan and lease losses.
Total loans
lnlsgr
Total loans and lease financing receivables, net of unearned
income.
C&I loans
lnci
Commercial and industrial loans. Excludes all loans secured
by real estate, loans to individuals, loans to depository institutions and foreign governments, loans to states and political
subdivisions and lease financing receivables.
Real-estate loans
lnre
Loans secured primarily by real estate, whether originated
by the bank or purchased.
Commercial RE loans
lnrenres
Nonresidential loans primarily secured by real estate.
Consumer loans
lncon
Loans to individuals for household, family, and other personal expenditures including outstanding credit-card balances and other secured and unsecured consumer loans.
U.S. Gov. securities
scus
Total U.S. Treasury securities plus U.S. Government agency
and corporation obligations.
No. of employees
numemp
The number of full-time employees on the payroll of the bank
and its subsidiaries at the end of the quarter.
Commercial banks
bkclass
A classification code assigned by the FDIC based on the
institution’s charter type (commercial bank or savings institution).
Stock institutions
mutual
Banking institutions fall into one of two ownership types,
stock or non-stock. An institution that sells stock to raise
capital is called a stock institution. It is owned by the shareholders who benefit from profits earned by the institution.
A non-stock institution, or mutual institution, is owned and
controlled solely by its depositors. A mutual does not issue
capital stock.
Independent
namehcr
Dummy variable that we assign a value of one if “namehcr”
is blank and a value of zero otherwise
42