Loan Loss Reserves, Regulatory Capital, and Bank Failures: Evidence from the 2008-2009 Economic Crisis

Jeffrey Nga jeffng@mit.edu

Sugata Roychowdhuryb sugata.roychowdhury@bc.edu

Abstract

This paper tests the influence of regulatory capital guidelines on the link between banks’ loan loss reserve decisions in 2007 and the risk of bank failure during the 2008-2009 economic crisis. Our primary finding is that larger loan loss reserve increases in 2007 are associated with an increased risk of bank failure in 2008-2009, but only for banks that are allowed to add back loan loss reserves as Tier 2 capital. The results suggest that the ability to add back loan loss reserves to regulatory capital creates the illusion of non-declining financial health, which allows banks to avoid taking the actions necessary to reduce failure risk. Our results thus challenge the notion that the buffer created by loan loss reserves lowers bank failure risk. Simultaneously they question the soundness of recent regulatory proposals to allow even more loan loss reserves to be added back in the calculation of regulatory capital. Keywords: bank failure, bank risk, regulatory capital, capital adequacy, loan loss reserves, loan loss provisions, loan charge-offs JEL Codes: G21, G28, G32, G38, M41, M48

First draft: January 27, 2010 Current draft: July 16, 2010

_____________________________ a. MIT Sloan School of Management, 50 Memorial Drive E52-325, Cambridge, MA 02142. b. Boston College, 140 Commonwealth Avenue, Fulton 520A, Chestnut Hill, MA 02142.

This paper has benefited from comments we received from Anne Beatty, Michelle Hanlon, S.P. Kothari, Ewa Sletten, Ross L. Watts, and from seminar participants at the University of Missouri and Boston College. We thank the Federal Deposit Insurance Corporation for responding to our queries on regulations regarding bank capital calculations. All remaining errors are ours.

1

1. Introduction

This paper is motivated by recent concerns, prompted by the financial crisis of 2008-2009,

that regulatory capital guidelines on loan loss reserves can generate dysfunctional outcomes.

Under current guidelines, a bank can add back its loan loss reserves, which are cumulative

accrued loan losses, as Tier 2 regulatory capital, up to a maximum of 1.25% of a bank’s gross

risk-weighted assets. The treatment of loan loss reserves as capital has received considerable

attention in the wake of the financial crisis, with regulators actively debating its pros and cons.

For example, in speaking at the American Bankers Association meeting on March 17, 2010,

Comptroller of the Currency John Dugan argued for the relaxation of restrictions on the inclusion

of loan loss reserves as capital, to encourage banks to report adequate and timely reserves. A day

later at that same meeting, Federal Deposit Insurance Corporation (FDIC) Chairperson Sheila

Bair contested this view, opining that “letting more reserves count [towards capital] could

dramatically, in our view, dilute the quality of capital.”1 In light of this debate, we examine

whether regulatory capital guidelines influence the association between loan loss reserve

accounting and the risk of bank failure during the financial crisis.

Various hypotheses exist in the literature regarding the link between loan loss reserve

changes and bank failure risk. First, accounting accruals are expected to reflect contemporaneous

economic events that have implications for future cash flows (Dechow 1994). For banks,

increases in loan loss reserves reflect increases in the accrued expected losses in their loan

portfolios. Thus, banks with larger increases in loan loss reserves are expected to experience

more severe cash flow losses in their loan portfolios in the future and will thus be more likely to

fail, particularly during an economic crisis (the “troubled-bank” hypothesis). Second, a number

of academic studies (Elliott et al. 1991, Wahlen 1994, and Beaver and Engel 1996) indicate the

1 She also referenced the late L. William Seidman, a former chairman of the FDIC, in saying that the rationale for adding back reserves as capital was ambiguous.

2

possibility that in practice, banks report larger loan loss provisions, thus increasing their loan loss

reserves, when they are financially stronger and expect better future performance. Hence, banks

with greater loan loss reserve increases signal financial strength and are more likely to survive

during a financial crisis (the “signal-of-strength” hypothesis).

We propose that the relation between failure risk and loan loss reserve decisions during a

period of worsening credit quality is likely to be more complex than has been suggested by either

the troubled-bank or signal-of-strength hypothesis. This is because we expect existing regulatory

guidelines to have a particularly pronounced and possibly dysfunctional effect on bank

managers’ incentives during times of deteriorating economic conditions. The dysfunctional

effect can arise because, according to current regulatory guidelines, loan loss reserve increases

during periods of worsening credit conditions do not necessarily result in a decline in regulatory

capital, and in fact can increase regulatory capital.2

Regulatory capital adequacy is a key metric that regulators rely on to judge a bank’s

solvency status (see Estrella et al. 2000), and increasing loan loss reserves can create an illusion

of improving solvency if they result in higher regulatory capital. This illusion of improving

financial health can lead to more lenient regulatory monitoring of banks with deteriorating loan

quality.3 The impression of financial health can also allow banks to avoid taking prudent actions

such as restricting risky lending, improving collection efficiency, etc., which would strengthen

their underlying financial position. Lenient regulatory monitoring and inaction in the presence of

underlying loan problems can threaten banks’ financial stability, particularly when an economic

2 The mechanics involved in the regulatory capital adequacy calculation are discussed in detail in Section 2.1. Briefly, Tier 1 capital consists primarily of shareholder equity. Loan loss provisions, which increase loan loss reserves and reduce net income, decrease shareholder equity net of tax. Since the entire loan loss reserve balance counts as Tier 2 capital, every dollar of the loan loss provision increases the sum of Tier 1 and Tier 2 capital by the tax rate times that dollar. This increase happens particularly over the range that Tier 2 capital is below the maximum allowable limit of 1.25% of gross risk-weighted assets. 3There exists a long history of bank regulators relying on low capital adequacy ratios to identify banks that require greater supervisory attention (Estrella et al. 2000).

3

downturn occurs. This generates the following prediction: loan loss reserve increases in 2007 are

positively associated with bank failure risk during 2008-2009, but only among banks that are less

constrained in adding back loan loss reserves as Tier 2 capital. Importantly, we do not expect

loan loss reserve increases to exhibit a positive association with bank failure risk when the

former cannot be added back as Tier 2 capital; in that range, loan loss reserve increases via

provisions unambiguously reduce regulatory capital. We refer to our hypothesis as the

“regulatory-capital-effects” hypothesis.

For our analyses, we obtain bank failure data from Federal Deposit Insurance

Corporation (FDIC) press releases. Data on banks’ accounting information is obtained from the

call reports commercial banks file with the Federal Reserve, Federal Deposit Insurance

Corporation, and/or the Office of the Comptroller of the Currency. We also separately

investigate the two components of loan loss reserve changes, given their distinct balance sheet

and income statement effects: (i) loan loss provisions and (ii) loan charge-offs.4 Loan loss

provisions increase loan loss reserves; they reduce net income and hence, stockholders’ equity.

Loan charge-offs reduce loan loss reserves, and have no income statement effect.

We document that, after controlling for various economic determinants of bank failure,

loan loss reserve changes are positively associated with bank failure risk. More importantly, the

positive association between loan loss reserve changes and bank failure risk is restricted to banks

with loan loss reserves that are below the maximum limit allowable as Tier 2 capital. Above that

limit, higher loan loss reserve changes are not significantly associated with failure risk. The

results are as predicted by the regulatory-capital-effects hypothesis, and cannot be explained by

either the troubled-bank or signal-of-strength hypothesis. The troubled-bank hypothesis in

4 Loan loss provisions are used to adjust the loan loss reserve account to the desired balance. A loan charge-off decision involves identifying specific accounts in default and removing them from the books.

4

particular cannot explain why loan loss reserve increases would be associated with higher bank

failure risk only when they are below the maximum limit allowable as Tier 2 capital.

In examining the components of loan loss reserve changes, we find that bank failure risk

is associated positively with loan loss provisions and negatively with loan charge-offs. In the

cross-sectional analyses, we document that failure risk’s positive (negative) association with loan

loss provisions (loan charge-offs) is more pronounced when banks are less constrained in

including loan loss reserves as capital. Since loan loss reserves increase with loan loss provisions

and decrease with charge-offs, the results are consistent with the regulatory-capital-effects

hypothesis and inconsistent with the signal-of-strength hypothesis. Further, the troubled-bank

hypothesis, which would predict that both loan loss provisions and charge-offs are associated

with increased failure risk, cannot explain the observed results.

Next, we investigate whether the association between failure risk and loan loss reserve

decisions is at least partially a result of managers’ accounting discretion.5 If managers exercise

discretion in loan loss reserve accounting in order to report higher capital, and if the upward-

managed capital allows managers to avoid addressing underlying problems, then bank failure

risk should be positively associated with managers’ discretionary loan loss provisions. Using

models for normal loan loss provisions and charge-offs, we find that bank failure risk is

positively associated with discretionary loan loss provisions and negatively with discretionary

charge-offs. Further, failure risk’s positive association with discretionary loan loss provisions,

and its negative association with discretionary charge-offs, is restricted to banks that are less

constrained in reporting loan loss reserves as Tier 2 capital.

In additional analyses, we first document preliminary evidence that banks’ risky activities

in the first quarter of 2008, including lending, appear to be positively associated with changes in 5 Evidence in the literature indicates that managers can actively use the discretion available to them in recording loan loss provisions to manage their reported regulatory capital upwards (Moyer 1990, Beatty et al. 1995, Ahmed et. al. 1999).

5

Tier 2 capital, but negatively associated with changes in Tier 1 capital. This finding supports the

concern raised by our primary results: the ability of loan loss reserves to increase Tier 2 capital

engenders greater failure risk. Second, we measure bank failure severity using the cost of the

failure to the FDIC. We find that within the sample of failed banks, bank failure severity varies

positively with loan loss reserve changes. Thus, not only are loan loss reserve increases

associated with greater failure risk, they also contribute to more severe bank failures.

Recently, there has been considerable interest in banks’ loan loss accounting practices.

Beatty and Liao (2009) argue that more conservative loan loss provisioning practices appear to

make banks’ lending less sensitive to their capital adequacy during recessionary periods.

Bushman and Williams (2009) provide evidence suggesting that greater discretion in loan loss

provisioning weakens regulators’ ability to monitor and discipline banks. Our paper focuses on a

related but distinct issue: regulations regarding capital can themselves have dysfunctional

consequences on bank failure risk. In this respect, our study is related to a large body of literature

that examines the link between regulatory capital requirements and bank risk (Koehn and

Santomero 1980, Buser et al. 1981, Furlong and Keeley 1989, Saunders et al. 1990, Shrieves and

Dahl 1992, Diamond and Rajan 2000, Rime 2001, Laeven and Levine 2009). Specifically, we

document that regulatory capital guidelines that allow total capital to increase with loan loss

reserves can, during economic crises, contribute to heightened bank failure risk.

The findings in our paper have policy implications. By including loan loss reserves,

regulatory capital departs from the conventional notion of capital as a cushion of funds available

to partially insulate a bank against shocks to the value of deposit liabilities (see, for example,

Diamond and Rajan 2000). As noted earlier, in the wake of the economic downturn in 2008-

2009, bank regulators are considering increasing the limit to which loan loss reserves can count

towards Tier 2 capital. This proposal has received the support of banks and bank organizations

6

such as the Georgia Bankers’ Association.6 Following intense bank lobbying at the time of the

issuance of Statement of Financial Accounting Standards (SFAS) 166 and 167 in late 2009,

regulators allowed “transitional relief” for banks from the 1.25% limit on Tier 2 capital.7 Our

findings, however, question the very rationale for considering loan loss reserves as Tier 2 capital.

At the very least, the results strike a cautionary note against expanding the add-back of loan

reserves to regulatory capital.

The rest of the paper is organized as follows. Section 2 discusses our setting. Section 3

describes our sample construction and data. Our results are presented in Section 4, and section 5

concludes.

2. Setting and Hypotheses

The economic crisis of 2008-2009 provides a rich setting for our paper because in

addition to raising questions about the role of loan loss reserves and their treatment as regulatory

capital, it is characterized by a significant number of bank failures. Bank failures are

unambiguously negative outcomes, and they facilitate the direct examination of the

consequences of increasing loan loss reserves, as well as the impact of regulatory capital

guidelines. Below, we discuss two critical institutional factors – (a) the calculation of regulatory

capital and (b) the process of a bank failure – before discussing our hypotheses.

6 In its comment letter to the Federal Reserve Board on October 15, 2009 relating to bank regulators’ proposed rule-making on risk-based capital guidelines and related issues, Discover Financial Services argues for the increase or elimination of the current cap on loan loss reserve amount eligible to qualify as Tier 2 capital (Federal Reserve Board, 2009). In an oral statement to the Domestic Policy Subcommittee of the U.S. House Oversight and Government Reform Committee on November 2, 2009, Joe Brennan, President and CEO of the Georgia Bankers Association argued for a higher limit on the loan loss reserves to be counted as regulatory capital. He stated that 76% of all Georgia banks were adversely affected by the restriction and that billions in capital among Georgia banks would be freed up to support more lending if the limit is suspended. 7 In the first two quarters after the implementation of SFAS 166 and 167 in November 2009, a banking organization, under certain conditions, is permitted to include without limit in Tier 2 capital the full amount of the loan loss reserves.

7

2.1 Regulatory capital and loan loss reserves

The capital adequacy ratio, or the ratio of regulatory capital to risk-weighted assets, is the

metric most widely relied on by regulators to monitor bank solvency (Estrella et al. 2000). There

are two main sources of regulatory capital: Tier 1 and Tier 2. Tier 1 capital is “core” capital; it

includes shareholders’ equity (the primary component) and disclosed reserves. Tier 2 capital is

“secondary” capital; it includes general loss reserves, undisclosed reserves, and subordinated

term debt. The International Basel Committee requirements specify a minimum limit of 4% for

Tier 1 capital, and 8% for total capital.

In practice, Tier 2 capital consists primarily of loan loss reserves, with the restriction that the

latter is limited to 1.25% of gross-risk-weighted assets (GRWA).8 From an accounting

perspective, loan loss reserves recognize expected losses in the loan portfolio. That is, losses are

accrued when expected, as opposed to expensed when realized. Hence, loan loss reserves provide

a buffer to absorb declines in the value of the bank’s loan portfolio when the losses are realized

in the future. In part, the add-back of loan loss reserves to regulatory capital is meant to serve as

an incentive for the timely recording of loan losses (Wall and Koch 2000).

A simple example illustrates the role of loan loss reserve components in influencing

regulatory capital.9 Assume a bank increases its loan loss reserves by reporting a loan loss

provision of $100, and the statutory tax rate is 40%. This transaction, ceteris paribus, has two

effects on regulatory capital: i) a Tier 1 effect and ii) a Tier 2 effect. The loan loss provision

reduces after-tax income by $100*(1 - tax rate), or $60, which in turn reduces shareholders’

equity, and hence Tier 1 capital, by $60. Since banking capital regulations allow loan loss

8 Gross risk-weighted assets equal risk-weighted assets used in the computation of the capital ratios plus excess allowance for loan and lease losses plus allocated transfer risk reserve. The limit of 1.25% of gross risk-weighted assets on the amount of the loan loss reserves that a banking organization may include in tier 2 capital is a standard included in the first capital accord of the Basel Committee on Banking Supervision (Basel Accord). See Basel Committee on Banking Supervision, International Convergence of Capital Measurement and Capital Standards (1988), paragraph 21. 9 We thank the FDIC for confirming that our example is correct in representing the effect of the regulations.

8

reserves to be considered as Tier 2 capital, Tier 2 capital increases by the provision amount of

$100. Total regulatory capital (the sum of Tier 1 and Tier 2) increases by $40 as a result of the

loan loss provision, that is, the tax rate times the provision amount. If loan loss reserves prior to

the provision were already equal to or greater than 1.25% of GRWA, the $100 provision in the

example would not increase Tier 2 capital. If loan loss reserves were below the 1.25% limit but

significantly close to it, it is possible that only a portion of the $100 loan loss provision would

count towards Tier 2 capital, not the entire amount.

Next, consider the effect of increasing a loan charge-off, assuming no loan loss provision. A

charge-off of $100 would reduce loan loss reserves by $100, ceteris paribus. Since charge-offs

do not affect the shareholders equity account, the sole effect of the charge-off would be to reduce

Tier 2 capital, and hence total regulatory capital, by $100 (to the extent that loan loss reserves

were within the maximum allowable limit).

The example highlights that loan loss provisions and loan charge-offs have significantly

distinct effects on regulatory capital. Furthermore, the effect of loan loss changes on regulatory

capital is dependent on the size of total available Tier 2 capital relative to the maximum limit

allowable under current regulations. Section 2.3 discusses the implications of these institutional

features for the relation between loan loss reserve changes and the likelihood of bank failure. The

process of such a failure is discussed below.

2.2 Bank failure

A bank failure is an extreme event involving the chartering authority or the Federal

Deposit Insurance Corporation (FDIC) “closing” the bank.10 Closing a bank involves shutting

10 The chartering authority for state-chartered banks is usually the state banking department; for national banks, the Office of the Comptroller of the Currency (OCC); and for federal savings institutions, the Office of Thrift Supervision (OTS). While it is much more common for the chartering authoring to close a bank, the FDIC has the authority, under the FDIC Improvement Act of 1991, to close any bank that it considers to be critically undercapitalized and that does not have a plan to restore capital to an adequate level.

9

down its operations, re-distributing its assets and liabilities and, if necessary, paying off insured

depositors.

Generally, a bank is closed when the regulating authority determines that it is “critically

undercapitalized” and is deemed unable to meet its obligations to depositors and other creditors.

The key attribute determining undercapitalization is insolvency, which occurs when the bank’s

assets are worth less than its liabilities according to either book or market values.11 The Federal

Deposit Insurance Corporation Improvement Act (FDICIA) of 1991 requires regulators to close

banks before they reach book-value insolvency, since the market values of bank assets are

uncertain and typically, for troubled banks, below their book values.

In the event of a bank failure, the FDIC acts as a receiver and is in charge of the failure

resolution. FDICIA mandates the use of the least-cost resolution method for bank failures, the

objective of which is to minimize the present value of the net losses incurred by the FDIC. There

are two primary types of failure resolution methods: (1) purchase-and-assumption transactions

and (2) deposit pay-offs.

In a purchase-and-assumption transaction, a healthy bank acquires the failed bank by

purchasing “some or all” of the assets and assuming “some or all” of the liabilities. The FDIC

often provides assistance to the acquiring bank, e.g., in the form of loan-loss sharing agreements,

and then liquidates the remaining assets and liabilities, internalizing the cost of doing so. The

acquiring bank usually compensates the FDIC for the franchise value from the failed bank’s

established customer relationships, which helps reduce the insurer’s resolution cost. In a deposit-

payoff transaction, the FDIC pays the failed bank’s depositors the full amount of their insured

11 Another reason for bank failure is illiquidity, which occurs when a bank is unable to meet its current obligations as they come due. For example, when depositors are concerned that a bank is failing, they may withdraw their deposits and precipitate a liquidity crisis at the bank (bank run). Illiquidity appears to drive bank failures more commonly in the European Union. Because of deposit insurance and the U.S. Federal Reserve’s ability to provide liquidity, banks in the United States typically fail because they are insolvent as opposed to illiquid.

10

deposits. Typically deposit payoffs are observed when no other bank is interested in assuming

the assets and liabilities of the failed bank.12

2.3 Hypotheses

Given the widespread acknowledgement of the role banks’ poor lending decisions played in

the 2008-2009 economic crisis, managers’ decisions regarding loan loss reserves have been

under considerable scrutiny. In particular, a key question has been whether increasing loan loss

reserves in the period leading up to the 2008-2009 economic downturn would have enabled

banks to better weather the crisis. Direct evidence on the effect of loan loss reserves on bank

failures is lacking.

A relation between loan loss reserves and the likelihood of bank failure is expected on

various grounds. Elliott et al. (1991), Whalen (1994) and Beaver and Engel (1996) argue that

banks are likely to make greater loan loss provisions (and increase their loan loss reserves) when

they expect to perform better in the future. In other words, greater loan loss provisions are a

signal of strength. Consistent with this “signal of strength” hypothesis, they document that loan

loss provisions are associated with contemporaneously positive market returns. However, these

findings have been subsequently questioned in the literature (see Ahmed et al. 1999). Further,

loan loss provisions and charge-offs recorded in periods close to an economic recession are

likely to reflect the accumulated effects of the risks assumed in past lending practices. This

“troubled bank” hypothesis suggests that banks report the loan loss provisions and charge-offs

simply because they are compelled to, as they contend with more negative realizations in the

12 Variations in the two primary methods exist. For example, in a deposit transfer transaction, the FDIC transfers the insured deposits to a healthy bank that is willing to be an agent of the FDIC. The depositors can either withdraw their deposits or keep them in the new bank. In a bridge transaction, the FDIC itself temporarily acquires the assets and liabilities and takes over the operations of the failed bank as it decides upon the least-cost resolution method. In a more significant departure, the FDIC can engage in an open-bank transaction, in which it provides financial assistance to the bank while it continues operations. Open-bank transactions are not classified as bank failures in our sample, since they are implemented when banks’ liquidity and/or solvency issues are perceived as temporary.

11

periods immediately preceding an economic downturn. This can potentially generate a positive

association of both loan loss provisions and charge-offs with the likelihood of bank failure.

Our paper proposes an association between loan loss reserve decisions and the risk of

bank failure based on the special circumstances that prevail during economic downturns. As a

bank’s loan quality deteriorates in the period leading up to a crisis, “building up” loan loss

reserves can generate dysfunctional incentives. Since loan loss reserves can be added back as

Tier 2 capital, they generate an illusion of non-declining capital adequacy and even of improving

capital adequacy despite deteriorating loan quality. The illusion of financial health due to higher

regulatory capital can induce complacency among banks, preventing them from responding to

underlying loan problems by restricting lending, improving collection efficiency, etc. Higher

capital adequacy also potentially allows banks to take actions that could exacerbate the

likelihood of failure, such as granting more risky loans than the given deteriorating economic

conditions justify. This possibility is derived from Shrieves and Dahl (1992), who document

evidence that banks with higher capital engage in riskier activities.

From a regulatory monitoring perspective, Buser et al. (1981) point out that regulation

generally permits banks with higher capital to make riskier investments. Further, the appearance

of higher total capital adequacy can lead to more lenient regulatory monitoring of these banks.

During an ensuing economic crisis, these complacent but nevertheless troubled banks then

experience more severe liquidity issues and asset value declines that ultimately drive their

failure.13

13 Note that as loan assets increasingly lose value during the economic downturn, loan loss provisions and hence loan loss reserves are expected to mount. At the time of failure, loan loss reserves are probably much beyond the 1.25% of risk weighted assets allowable as Tier 2 capital. Thus, the sole effect of earnings losses induced by the provisions is likely to be a reduction in total regulatory capital. Collectively, the earnings losses and declining capital adequacy could generate bank failure. See the illustration in Section 2.1.

12

It is conceivable that bank managers actively use their accounting discretion to report

larger loan loss reserves, in order to avoid declines in total capital adequacy ratios.14 A number

of studies document evidence that banks use their discretion in loan loss provisions to manage

their capital ratios (Moyer 1990, Beatty et al. 1995, Ahmed et al. 1999). Generally, declining

capital ratios can be associated with increased financial costs such as higher deposit insurance

premiums, as well the political costs that arise from more severe regulatory scrutiny.15

Discretionary loan loss reserve accounting decisions potentially represent managers’ attempts to

overstate their financial health during deteriorating economic conditions in order to avoid these

costs. However, to the extent that such capital management bears deadweight costs (i.e., it

distracts managers from taking the real actions necessary to address fundamental weaknesses),

such banks are then at an increased failure risk in the event of an economic crisis.

Irrespective of whether large loan loss reserves induce complacency or are a result of

opportunistic accounting, we predict a positive association between loan loss reserve increases

and failure risk. Further, we expect this positive association to be concentrated among banks that

are less constrained in adding back loan loss reserves as Tier 2 capital.

H1: Loan loss reserve increases in 2007 are associated with a higher risk of bank failure

during 2008-2009, particularly when banks are less constrained in adding back loan loss

reserves to regulatory capital.

Finding evidence consistent with H1 would support the regulatory-capital-effects

hypothesis: loan loss reserve increases lead to increased bank failure risk because the regulatory

14 There already exists evidence that banks use their discretion in loan loss provisions to manage their capital ratios (Moyer 1990, Beatty et al. 1995, Ahmed et al. 1999). 15 As Kim and Kross (1998) report, greater regulatory scrutiny can involve more frequent audits by regulatory agencies and/or restrictions on capital distributions, on payments to bank managers and on asset growth.

13

guidelines that allow loan loss reserves to be added back to regulatory capital generate

dysfunctional outcomes.

We further analyze the components of loan loss reserve increases. Ceteris paribus, loan

loss provisions increase loan loss reserves, while loan charge-offs reduce reserves. Consistent

with the regulatory-capital-effects hypothesis, we predict that bank failure risk is positively

associated with loan loss provisions and negatively associated with loan charge-offs. We also

predict that bank failure risk’s positive association with loan loss provisions and its negative one

with loan charge-offs are both more pronounced when banks are less constrained in adding back

loan loss reserves to regulatory capital.

To examine the role of bank managers’ discretion in using loan loss reserves to influence

regulatory capital, we test the association of discretionary components of loan loss provisions

and charge-offs with subsequent bank failure risk. Our predictions for these discretionary

components are similar to those for the total loan loss provisions and charge-offs.

Finally, the FDIC’s least-cost method of failure resolution, discussed in Section 2.2,

implies that the costs borne by the FDIC are likely to increase with the severity of a bank failure.

Banks whose asset values are more severely depressed relative to their liability values and those

that have more depleted franchise values would generate greater closing costs for the FDIC. To

the extent that loan loss reserve increases, by increasing regulatory capital, allow managers to

ignore underlying problems for a longer period, they are likely to exacerbate the severity of bank

failure. Thus, for the sample of failed banks, we examine whether loan loss reserve increases are

associated with the severity of bank failure.

14

3. Sample construction

The timeline in our research design is shown below:

The National Bureau of Economic Research (NBER) identifies December 2007 as the

beginning of the recent recession. Our empirical analysis focuses on the association between loan

loss reserve changes and its components (loan loss provisions and net charge-offs) prior to the

economic crisis (i.e., in 2007) and the bank failures that occurred during the economic crisis (i.e.,

in 2008 and 2009).

3.1 Data on bank failures

We obtain data on bank failures from the Federal Deposit Insurance Corporation (FDIC)

website: http://www.fdic.gov. The FDIC, which is appointed as the receiver in the event of a

bank failure, makes public a press release that provides details about the bank at the time of the

failure, including actions being taken to deal with it. We collect the following information from

the press releases (available on the FDIC website): the name of the failed bank, the bank’s

estimated assets and deposits at the time of the failure, and the cost of the failure to the FDIC. As

an example, the press release for the failure of Corus Bank is provided in Appendix A. Corus

Bank’s failure date was September 11, 2009; its estimated assets and deposits at the time of

failure were both approximately $7 billion, and the cost of the failure to FDIC was assessed at

$1.7 billion.

2008 & 2009

Bank failures

(Data obtained from FDIC press releases)

2007

Measurement of loan loss reserve changes

(Data obtained from call reports)

15

Table 1 provides descriptive information on the failure of commercial banks and thrifts

(which includes savings and loans associations and savings banks) from 2001 to 2009. While, for

the reasons discussed below, the focus of this paper is on the failure of commercial banks, we

also provide descriptive information on the failure of thrifts to provide a broader overview of

bank failures and to highlight the enormity of the problems facing the banking industry in

general. Failures of commercial banks and thrifts were relatively infrequent prior to the

economic recession. A total of 21 commercial banks and 4 thrifts failed between the seven years

from 2001 to 2007, compared to a total of 140 commercial banks and 24 thrifts during the

economic recession in 2008 and 2009.16 Available data indicates that the FDIC bore the large

costs that arose from bank and thrift failures in 2008 and 2009. For example, the total cost to the

FDIC insurance fund on account of failed commercial banks was $4.58 billion in 2008 and $24.1

billion in 2009. In fact, failure costs were significant enough to deplete the FDIC insurance fund

to the point of insolvency as of December 31, 2009.

In this paper, we focus on commercial banks because i) commercial banks and thrifts file

different regulatory reports, ii) detailed regulatory report data for individual commercial banks,

both private and public, are publicly available in a machine-readable form, and iii) the number

of failed commercial banks is significantly larger than the number of thrifts, facilitating wide-

sample empirical analyses.

3.2 Data from call reports

We obtain data on loan loss reserves, loan loss provisions, and net charge-offs, as well as

other accounting variables, from the call reports filed by commercial banks with the Federal

Reserve, the Federal Deposit Insurance Corporation, or the Office of the Comptroller of the

16 While not included in our analyses, we note that there were 86 more failures of commercial banks and thrifts by the end of June 2010. This suggests that many banks are still facing difficulties despite some indications of economic stabilization in 2010.

16

Currency. The data is available in machine-readable form at the Chicago Federal Reserve

website.17 We begin with the 8,098 call reports filed by banks for the fiscal year ending in

December 2007. To be included in our sample, the bank must have positive total assets and total

loans for the fiscal years ending in December 2006 and December 2007.18 To merge the bank

failure data with the call report data, we obtain the RSSD ID of the banks in the bank failure

dataset. The RSSD ID is the unique identifying number assigned by the Federal Reserve for all

financial institutions, main offices, and branches. This final sample consists of 7,463 commercial

banks, including 137 that subsequently failed during the 2008-2009 crisis.19 For brevity, we

henceforth use the term “banks” to refer to the commercial banks in our sample.

Table 2 presents the distribution of the 7,463 banks across the different states and regions

of the United States. The states with the most number of bank failures are Georgia, Illinois, and

California, with 29, 21, and 17 failures, respectively. In terms of the percentage of banks that

failed, the state with the highest percentage is Nevada, with 16.67% of its banks having failed in

2008 and 2009. From a regional perspective, while there were more bank failures in the south,

the percentage of all banks that failed is higher in the west. The uneven distribution of bank

failures across different states and regions is consistent with the fact there was significant

variation in the impact of the economic crisis across the United States.

Table 3 provides some descriptive information on the regulatory capital ratios and loan

loss reserves in 2007 for of the 7,463 banks. The mean total risk-based capital ratio and the Tier

1 risk-based capital ratio in 2007 are 17.28% and 16.18%, respectively. The inclusion of Tier 2

17 http://www.chicagofed.org/webpages/banking/financial_institution_reports/commercial_bank_data.cfm 18 We require data in 2006 and 2007 to construct our variables, in particular, those variables that measure changes from 2006 to 2007. 19 Three of the 140 failed banks were dropped from the sample because they did not file call reports in 2006. The three banks are Waterford Village Bank, Pacific National Bank, and Bank USA.

17

capital adds, on average, 1.097% of risk-weighted assets to the total risk-based capital ratio.20

The median total risk-based capital ratio and Tier 1 risk-based capital are 14.11% and 13.04%,

respectively. Given that the minimum required total risk-based capital ratio and the Tier 1 capital

ratio are, respectively, 8% and 4%, a typical bank in the sample meets the minimum

requirements, as expected. Based on the difference between the minimum requirements and the

means and medians of the actual ratios, one can also observe that the total risk-based capital ratio

requirement is more binding on average than the Tier 1 capital ratio requirement. As reported in

Table 3, Panel B, banks that failed in 2008-2009 had, in 2007, significantly lower Tier 1 but

significantly higher Tier 2 capital relative to banks that survived. 

Finally, Table 3, Panel A demonstrates that in practice, Tier 2 capital is almost entirely

made up of loan loss reserves. Mean loan loss reserves scaled by risk-weighted assets used in

Tier 2 capital is 1.037%, relative to a mean Tier 2 capital ratio of 1.097%. Loan loss reserves

make up about 95% (1.037/1.097) of Tier 2 capital, about 6% (= 1.037 / 17.279) of total risk-

based capital, and 1.365% of total loans.

4 Bank failure and loan loss reserve accounting

4.1 Research design and descriptive statistics

In this paper, we rely on survival analyses that use hazard models to study the relation

between bank failure and loan loss reserve accounting. Hazard models are appropriate in our

context since they incorporate information about the time that elapses before an event (in our

case, a bank failure) occurs. Shumway (2001) demonstrates that, hazard models outperform

static models such as logit models in predicting bankruptcy.

20 Note that the allowable Tier 2 capital (i.e., the capital that can be considered as part of total capital) is the lesser of Tier 1 capital and Tier 2 capital. This constraint does not appear to matter. For example, in 2006 and 2007, the allowable Tier 2 capital is the same as the actual Tier 2 capital for all commercial banks in our sample.

18

Our tests rely on the widely used Cox proportional hazard model (Cox 1972; Cox and

Oakes 1984), which has the following form:

0 i ih( t ) h ( t )exp( X ) (2)

where Xi is a vector of explanatory variables and βi is a vector of coefficients.21 Note that the

hazard rate is the risk of failure at a certain point in time, conditional on survival until that point

in time. h0(t) represents the baseline hazard rate that is a function only of time, while exp(Xiβi)

represents the dependence of the hazard rate on a vector of explanatory variables X. In the Cox

model, the coefficient on the explanatory variable represents the proportional change in the

hazard rate for a one-unit change in the explanatory variable.

The focus of this paper is to examine how bank failure hazard is associated with changes

in loan loss reserves and its components: i) loan loss provisions and ii) net charge-offs. To

examine the above, we use the following hazard models:

h(t) = h0(t) exp (β1 CH_LLR +∑i βi CONTROLi + ε) (3)

h(t) = h0(t) exp (β1 LLP + β2 NCO +∑i βi CONTROLi + ε) (4)

The initiation date for the survival analysis is January 1, 2008, and the final date is December 31,

2009. h(t) is the hazard of bank failure at time t. t is the number of days from January 1, 2008 to

the failure date.

The independent variables of interest are CH_LLR, LLP, and NCO, which are,

respectively, the loan loss reserve changes, loan loss provisions and net charge-offs, expressed as

a percentage of the total loans. CONTROL is a set of control variables added to mitigate omitted

correlated variable bias: REAL_ESTATE_LOAN, NPL, ASSET_RISK, OVERHEAD,

INSIDER_LOAN, ROA, UNINSURED_DEPOSIT, LIQUIDITY, LEVERAGE, TOTAL_RBC,

21 Note that no explicit intercept parameter is estimated for the proportional hazards model. The log-likelihood function and survival function estimators are invariant with respect to translations of any of the independent variables, so no intercept is needed

19

SIZE, and REGION dummies. All the independent variables are measured at the end of 2007,

i.e., before the occurrence of the bank failures in 2008 and 2009.

If increases in loan loss reserves increase the risk of bank failure, we expect the

coefficients on CH_LLR in Eq. (3) to be positive and the coefficients on LLP and NCO in Eq. (4)

to be positive and negative, respectively. REAL_ESTATE_LOAN is loans and leases as a

percentage of average beginning and ending total assets. Exposure to real estate loans was a key

factor behind the financial difficulties that many banks faced during the crisis. We expect banks

with relatively more real estate loans to be at a greater risk of failure. NPL is non-performing

loans as a percentage of total loans; we expect banks with relatively greater NPL to exhibit

greater failure risk. ASSET_RISK is assets risk-weighted at 100% as a percentage of total risk-

weighted assets; we expect failure risk to be higher for banks with relatively riskier assets.

OVERHEAD is the non-interest expense as a percentage of total assets. On one hand,

banks with higher overheads are probably less efficient and therefore more likely to fail. On the

other hand, higher overheads (e.g., higher salaries and employee benefits) might be indicative

that managers are likely to prefer a quiet life and take less risk (Betrand and Mullainathan 2003).

This suggests that banks with higher overheads are less likely to fail. INSIDER_LOAN denotes

loans to executive officers, directors, principal shareholders, and their related interests as a

percentage of total assets. While more insider loans are a possible indicator of agency problems,

they can also induce in managers a greater commitment to a bank’s continued survival, and

hence a decreased preference for risk-taking. Thus, the sign of failure risk’s association with both

OVERHEAD and INSIDER_LOAN is ambiguous.

ROA is net income as a percentage of average beginning and ending total assets. We

expect more profitable banks to be less likely to fail. UNINSURED_DEPOSIT represents the

uninsured deposits as a percentage of the total deposits. The FDIC places a cap on maximum

20

insurable deposits. We expect banks with more uninsured deposits to be at greater failure risk

during times of crisis due to the greater possibility of “deposit runs”. LIQUIDITY denotes the

cash and balances due from depository institutions as a percentage of total deposits. Cash and

balances due from depository institutions provide liquidity during deposit withdrawals, which

tend to be higher during economic crises. Hence, a bank with higher LIQUIDITY is likely to face

fewer difficulties in meeting withdrawal requests, and be less likely to fail.

Higher LEVERAGE, or total liabilities as a percentage of total assets, is expected to be

associated with higher failure risk. TOTAL_RBC is the total risk-based capital ratio; it is

expected to be associated negatively with failure risk. SIZE is the natural logarithm of total assets

(with total assets in millions). From casual observation of the failed banks, it becomes apparent

that both smaller and large banks have failed. However, we control for size because it is an

important consideration when closing a bank, particularly in light of the possibility of

governmental support if the bank is “too big to fail”.

REGION2 is a dummy variable equaling one if a bank is in the Midwest region and zero

otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western

regions respectively. By construction, the Northeast serves as the benchmark region. We include

region dummies to mitigate concerns that the empirical results are driven by heterogeneous

regional characteristics. Examples of such heterogeneity include differences in the expansion of

the property sector, or unemployment differences.22

Table 4, Panel A presents the summary statistics for the above variables used in the

regressions. All the continuous variables are winsorized at the 1st and 99th percentile, to mitigate

the effect of outliers. The mean value of FAIL is 0.018, or 1.8%, reflecting that 137 of the 7,463

commercial banks in our sample failed between January 2008 and December 2009. In 2007, the 22 We are not able to include state dummies because there are many states with no bank failures. Hence, it is not possible to examine how within-state variation in loan loss reserve accounting is associated with within-state variation in the risk of bank failure.

21

mean change in loan loss reserves, loss loan provisions and net charge-offs as a percentage of

total loans are 0.105%, 0.347% and 0.247%, respectively.

Table 4, Panel B presents a preliminary comparison of surviving banks (FAIL=0) and

failed banks (FAIL=1). Compared to the banks that did not fail in 2008 and 2009, the banks that

failed have greater increases in loan loss reserves, higher loan loss provisions and higher net

charge-offs in 2007. Further, 82.7% of failed banks’ total loans in 2007 were to the real estate

sector; the corresponding fraction for surviving banks was significantly lower, at 68.4%. This is

consistent with the fact that real estate exposure was a key driver of banks’ financial difficulties

during the crisis. Banks that failed also had more non-performing loans (NPL), had invested

more in relatively riskier assets (ASSET_RISK), and were less profitable (ROA). They also had

more uninsured deposits (UNINSURED_DEPOSIT), less cash to meet potential deposit

withdrawals (LIQUIDITY), higher leverage (LEVERAGE), and lower total risk-based capital

(TOTAL_RBC). Interestingly, they also tend to be larger in terms of total assets

(TOTAL_ASSETS), which is not consistent with the notion that regulators are more likely to

intervene to prevent big banks from failing.

4.2 Primary results

Table 5 presents the results of the hazard models that examine how the likelihood of bank

failures is associated with the accounting for loan loss reserves. Panel A presents the regression

coefficients along with associated standard errors. Panel B presents the hazard ratios.

In Panel A, the coefficients of interest are those on change in loan loss reserves in

Column I, and those on loan loss provisions and net charge-offs in Column II. Column I

examines the effect of increases in loan loss reserves on the likelihood of bank failure. The

coefficient on CH_LLR is positive and statistically significant at the 5% level, suggesting that

increases in loan loss reserves are associated with higher failure risk, consistent with H1.

22

In Column II, the coefficient on LLP is positive and statistically significant at the 5%

level, suggesting that higher loan loss provisions are associated with a higher likelihood of bank

failure. The coefficient on NCO is negative and statistically significant at the 10% level,

suggesting that higher net charge-offs are associated with a lower likelihood of bank failure.

Given that higher loan loss provisions and lower net charge-offs generate higher

regulatory capital, our results are consistent with the regulatory-capital-effects hypothesis. Loan

loss reserve increases, by increasing capital adequacy and creating an impression of financial

health even as loan quality deteriorates, increase bank failure risk during an economic crisis. Our

results are inconsistent with the signal-of strength hypothesis, which would predict that loan loss

reserve increases are associated with a reduced failure risk. They also cannot be explained by the

troubled-bank hypothesis, which predicts that both loan loss provisions and loan charge-offs are

associated positively with failure risk.

Panel B of Table 5 reports the hazard of bank failure in terms of the hazard ratio, which is

the effect of an explanatory variable on the hazard of bank failure. For each continuous

explanatory variable (i.e., all variables except the region dummies), we use a single-standard

deviation increase in the variable to assess its economic effect on bank failure risk.23 For each

region dummy, we compute the hazard of bank failure if a bank is in a particular region, as

opposed to the Northeast region.

The hazard ratio for CH_LLR in Column I indicates that a single-standard deviation

increase in loan loss reserves increases the risk of bank failure by 18.9% (= 118.9% - 100%). The

results in Column II indicate that a single standard deviation increase in loan loss provisions (as a

percentage of total loans) increases the risk of bank failure by 18.9%. In contrast, a single

23 To compute the hazard ratio for a one standard-deviation increase, we first multiply the coefficient on the explanatory variable from Table 4, Panel A by the standard deviation of the variable from Table 3, Panel A. We then take the exponential of this result. For example, the hazard ratio of CH_LLR is the exponential of coefficient of 0.507 on CH_LLR multiplied by its standard deviation of 0.341.

23

standard deviation increase in net charge-offs (as a percentage of total loans) decreases the risk

of bank failure by 14.6% (= 100% - 85.4%).

Among bank characteristics, the most economically significant variable is asset risk

(ASSET_RISK). A single standard-deviation increase in risk-weighted assets increases failure

risk by 85.9%. This is consistent with the notion that investment in risky assets was a significant

contributor to banking problems during the crisis. Almost as economically significant is the

exposure to real estate. A single standard-deviation increase in REAL_ESTATE_LOAN increases

the risk of bank failure by 83.6%; this result is consistent with the notion that the collapse of the

real estate market was a key driver of banks’ financial problems. Finally, we also observe

economically significant effects for liquidity and risk-based capital. A standard-deviation

increase in LIQUIDITY and TOTAL_RBC reduces the risk of bank failure by 41.7% and 58.5%,

respectively.

The first-order drivers of bank failure appear to be, as expected, the fundamental

economic characteristics of the bank. Importantly, even after controlling for these first-order

factors, loan loss reserve changes have a statistically and economically significant positive

association with bank failure risk. Further analyses of the components of loan loss reserve

changes indicate that failure risk is positively associated with loan loss provisions and negatively

associated with net charge-offs. Collectively, the results so far are more consistent with the

regulatory-capital-effects hypothesis than with the signal-of-strength or troubled-bank-

hypotheses.

Further evidence in support of the regulatory-capital-effects hypothesis is obtained when

we rely on the regulatory constraint that the maximum loan loss reserves allowable as Tier 2

capital are limited to 1.25% of gross risk-weighted assets. This implies cross-sectional variation

in the extent to which banks are constrained in reporting a given amount of loan loss reserves as

24

regulatory capital. We exploit this cross-sectional variation to test whether the positive

association between bank failure risk and loan loss reserve changes is less pronounced when

banks are more constrained in reporting loan loss reserves as regulatory capital. For this test, Eq.

(3) is adapted as follows:

H(t) = h0(t) exp (β1 CH_LLR x CONSTRAINT + β3 CH_LRR

+ β3 CONSTRAINT +∑i βi CONTROLi + ε) (5)

To examine the components of loan loss reserve changes, we adapt Equation (4) as

follows:

H(t) = h0(t) exp (β1 LLP x CONSTRAINT + β2 NCO x CONSTRAINT +

β3 LLP + β4 NCO + β5 CONSTRAINT +∑i βi CONTROLi + ε) (6)

All variables in Eq. (5) and Eq. (6), other than CONSTRAINT, are as defined in Eq. (3)

and Eq. (4). CONSTRAINT represents two proxies for the extent to which a bank is constrained

in adding back loan loss reserves to capital in 2007. Both proxies are based on the bank’s loan

loss reserve balance as a percentage of gross-risk weighted assets at the end of 2006, LLR2006.

Our first proxy, CONSTRAINT1, is a binary zero/one variable set equal to one if the

bank’s LLR2006 is above the maximum allowable limit of 1.25% of GWRA, zero otherwise.

Thus, when CONSTRAINT1 = 1, additional loan loss reserve increases in 2007 arising from

higher loan loss provisions and/or lower net charge-offs do not increase regulatory capital in

2007 (assuming no change in GWRA in 2007).

To construct our second proxy CONSTRAINT2, banks with LLR2006 below 1.25% of

GWRA are sorted into two equal-sized groups based on increasing LLR2006. Banks with

LLR2006 above 1.25% of GWRA constitute the third group. Thus, CONSTRAINT2 classifies

banks into a total of 3 groups based on increasing LLR2006.24 The tercile ranks are scaled such

24 Within our sample of 7,463 banks, 2,631 banks fall into the third group with LLR2006 above 1.25% of GWRA.

25

that the ranks range from zero to one. Thus, CONSTRAINT2 captures cross-sectional variation in

the constraint below the 1.25% cutoff. In particular, the further LLR2006 is below the maximum

limit allowable as capital, the less constrained the bank is in adding back to regulatory capital

any increase in loan loss reserves in 2007.

The coefficient on the interaction term CH_LLR x CONSTRAINT in Eq. (5) captures how

the association of bank failure hazard with CH_LLR varies with constraints on the add-back of

loan loss reserves to regulatory capital. The coefficients on LLP x CONSTRAINT and NCO x

CONSTRAINT in Eq. (6) can be interpreted analogously.

Table 6 presents the results of the cross-sectional analyses. In Columns I and II, the

cross-sectional analyses are performed with CONSTRAINT1. The results indicate that the effect

of an increase in loan loss reserves on bank failure risk is significantly lower when banks are

constrained in adding back loan loss reserves to capital in 2007. In particular, the coefficient on

CH_LLR x CONSTRAINT1 is negative and significant at a 5% level. The coefficients imply that,

among banks whose loan loss reserves are below the 1.25% limit, a single standard-deviation

increase in loan loss reserves increases is associated with an increase in failure hazard of 27.9%.

The association is statistically significant at the 1% level. In contrast, the association between

loan loss reserve increases and failure risk is statistically insignificant among banks whose loan

loss reserves exceed the maximum limit allowable as capital.

Column II reports results with the components of loan loss reserve changes. We find that

the positive association between loan loss provisions and bank failure risk observed in Table 5 is

concentrated among banks whose loan loss reserves are below the maximum limit allowable as

capital. The coefficients also suggest a stronger negative association between loan charge-offs

and the failure risk among banks with loan loss reserves below the 1.25% limit relative to those

above this limit. However, the coefficients on either NCO or NCO x CONSTRAINT1 are not

26

statistically significant at conventional levels. We next turn to the analysis with CONSTRAINT2,

which allows for more variation in the extent to which banks are constrained in reporting loan

loss reserves as capital, and hence is expected to provide more power than CONSTRAINT1.

Columns III and IV presents the cross-sectional analyses with CONSTRAINT2. The

results are largely similar to those in Columns I and II, with one important exception. Unlike

results in Column II, the coefficient on NCO in Column IV is significantly negative at the 1%

level, and that on NCO x CONSTRAINT2 is significantly positive at the 5% level. The

coefficients imply a significant negative association between loan charge-offs and the failure risk

among banks that are the least constrained in adding back loan loss reserves to regulatory capital,

and an insignificant association among banks that are the most constrained.

4.3 Abnormal loan loss provisions and abnormal net charge-offs

The results thus far (Sections 4.3 and 4.4) can be interpreted as evidence of the

potentially dysfunctional consequences of reporting large loan loss provisions without

accompanying charge-offs during times of economic crisis. The contribution of loan loss

reserves to regulatory capital can create ex post incentives for managers to delay taking actions

that would avoid future financial instability. However, it is also possible that the results are a

consequence of managers’ actively exploiting their discretion in loan loss accounting to

artificially inflate regulatory capital when banks are financially troubled. To explore this

possibility, we rely on models for normal loan loss provisions and net charge-offs to identify

their discretionary components (e.g., Wahlen 1994, Ahmed et al. 1999, Beatty et al. 2002).25

The model for “normal” loan loss provisions and net charge-offs is as follows:

25 We do not use these models in our main analyses for a number of reasons. First, there do not appear to be consistent models of loan loss provisions and net charge-offs that are used across different papers and results could be sensitive to the choice of the models. Second, it is possible that banks engage in “real activities” management (Roychowdhury 2006). For example, banks could discretionally classify more of their loans as non-performing and this will, by the construction of existing models in the literature, increase (decrease) the amounts of “normal” (“abnormal”) loan loss provisions and net charge-offs.

27

LLP = β0 + β1 CH_NPL + β2 LLR + β3 LOAN + β4 LOANR + β5 LOANOTH + εLLP (7)

NCO = β0 + β1 CH_NPL + β2 LLR + β3 LOAN + β4 LOANR + β5 LOANOTH + εNCO (8)

where CH_NPL is the change in non-performing loans as a percentage of total loans from 2006 to

2007, LLR is the loan loss reserves as a percentage of total loans at the beginning of 2007, LOAN

is the loans as a percentage of total assets in 2007, LOANR is real estate loans as a percentage of

total loans in 2007, and LOANOTH is the sum of commercial and industrial loans, loans to

depository institutions, agricultural loans, loans to individuals, and loans to foreign governments

as a percentage of total loans in 2007.26 To assess managers’ discretionary choices, we extract

the “abnormal” components of loan loss provisions (ABN_LLP) and of net charge-offs

(ABN_NCO) from the residuals ε of Eq. (7) and Eq. (8), respectively.

Table 7 presents the results of regressions (7) and (8). The results indicate that loan loss

provisions and net charge-offs are higher when there are increases in non-performing loans and

higher beginning loan loss reserves. As expected, banks with more loans as a percentage of total

assets also have higher loan loss provisions and net charge-offs. These results are consistent with

Beatty et al. (2002). Interestingly, we find that more real estate loans as a percentage of total

loans is associated with lower loan loss provisions and net charge-offs. This result appears to

reflect the notion that prevailed prior to the current economic crisis—that real estate loans are

typically “safe loans” due to the perceived low likelihood of a downturn in the real estate market.

Next, we examine the effect of abnormal levels of loan loss provisions and net charge-

offs on bank failure risk by replacing LLP and NCO in Eq. (4) and Eq. (6) with ABN_LLP and

ABN_NCO, respectively. Table 8 reports the regression results with ABN_LLP and ABN_NCO.

Similar to the results in Table 5, we find in Column I that bank failure risk increases with

26 Unlike Beatty et al. (2002), who include each loan type in LOANOTH individually in the decomposition regression, we use the sum of the loan types because there is almost no cross-sectional variation in some of the loan types. For 2007, the average LOANR and LOANOTH is 68.68% and 18.58%, respectively. This means that the loan types in LOANR and LOANOTH make up most of the loan portfolios of the banks in our sample.

28

abnormal loan loss provisions and decreases with net charge-offs. Further, in cross-sectional

analyses we find evidence similar to Table 6: failure risk’s positive association with abnormal

loan loss provisions and its negative association with abnormal net charge-offs is significantly

more pronounced among banks that are less constrained in adding back loan loss reserves to

regulatory capital.

The evidence is consistent with the possibility that bank managers use their accounting

discretion to inflate regulatory capital via loan loss reserve changes in 2007. To the extent that

capital inflation is more likely among banks that are in financial trouble, and that capital inflation

allows managers to avoid taking the actions necessary to address underlying problems, such

banks were more likely to fail during the crisis in 2008-2009.

4.4 Additional analyses

4.4.1 Regulatory capital and banks’ risky activities

In Sections 4.2 and 4.3, we observe that banks with higher loan loss reserves are more

likely to fail only when these reserves can be added back as Tier 2 capital. This is predicted by

our hypothesis that banks are less likely to take actions such as restricting risky activities when

the loan loss reserve add-back keeps their capital adequacy from declining. In this section, we

undertake some preliminary analysis of the association between regulatory capital changes and

changes in the risky activities undertaken by banks.

In the context of the 2008-2009 crisis, banks started failing in the first quarter of 2008.

Bank failures significantly intensified during the second quarter of 2008; consequently, at that

point, the banks most severely affected by the crisis begin to drop from our sample. To obtain a

sense of the influence of regulatory capital on bank actions during the crisis, we measure changes

in the banks’ risky activities in the first quarter of 2008 (from January 1, 2009 to March 31,

29

2009). We use three different measures to capture changes in the banks’ risky activities: (a)

increases in total loans, (b) increases in loans risk-weighted at 50% or above, hereafter risky

loans, and (c) increases in all assets risk-weighted at 50% or above, hereafter risky assets.

Changes in the capital ratios are measured from the fiscal period ending in December 2007 to

that ending in December 2008.

Table 9 reports the correlations between changes in capital ratios, loan loss reserve

changes scaled by risk-weighted assets, and changes in risk-taking in the first quarter of 2008.

Since loan loss reserves are the primary component of Tier 2 capital, changes in both are highly

positively correlated ( = 0.872) Changes in Tier 1 capital exhibit a negative and significant

correlation with both changes in Tier 2 capital and in loan loss reserves. This is expected; an

increase in loan loss reserves via loan loss provisions decreases Tier 1 and increases Tier 2

capital as long as the latter is below 1.25% of GRWA.

In terms of the correlations between changes in capital and changes in risk-taking, we

first observe that increases in Tier 1 capital in 2007 are negatively associated with all three

measures of increased risk-taking in 2008: increases in total loans, increases in risky loans, and

increases in risky assets. This is probably because banks with greater increases in Tier 1 capital

(e.g., through raising additional capital as a buffer) generally followed more prudent policies and

reduced their risky activities in early 2008 due to better assessments of the impending crisis.

More importantly, we find that both Tier 2 capital increases and loan loss reserve increases are

associated positively with increased risk-taking. For example, the correlation coefficient between

the change in Tier 2 capital and the change in risky loans is a statistically significant 0.029.

Overall, the correlations suggest the possibility that the ability to add back loan loss reserve

increases in 2007 as Tier 2 capital made banks less likely to reduce and possibly even increase

their risky activities in 2008, thereby increasing the risk of failure.

30

4.4.2 Severity of bank failure analyses

Finally, for the sample of banks that failed in 2008 and 2009, we examine the association

between the severity of bank failure and changes in the loan loss reserve and its components. To

do so, we make use of the following regression specifications:

FDIC_COST = β0 + β1 CH_LLR + ∑i βi CONTROLi + ε (9)

FDIC_COST = β0 + β1 LLP + β2 NCO + ∑i βi CONTROLi + ε (10)

FDIC_COST = β0 + β1 ABN_LLP + β2 ABN_NCO + ∑i βi CONTROLi + ε (11)

where FDIC_COST is the estimated cost of bank failure to the FDIC, scaled by total assets. Note

that the cost to the FDIC is higher the more severe the bank failure. The bank failure is more

severe when the value of bank’s assets relative to its liabilities is lower, and/or the value of the

bank’s deposit franchise is more depleted.

In addition to the control variables used in earlier specifications for failure hazard, we

include DEPOSITS as a control variable. DEPOSITS represent the total deposits at the time of

the failure scaled by total assets at the time of the failure (according to the FDIC press release).

The cost of bank failure can depend on compensating bank depositors for their insured deposits

(to the extent that the banks’ existing net assets are insufficient to cover these deposits). The

larger the size of the deposit base, the greater the likely compensation to depositors; hence we

control for deposits. All other variables are as defined earlier, with continuous variables

winsorized at the 1st and 99th percentiles to mitigate the effect of outliers.

Table 10 reports the results. The sample used for the analyses is constrained to only those

failed banks for which the FDIC provides the cost estimates. In particular, while there are 137

failed banks in our sample, the FDIC provides cost estimates for 136 banks. In Column I, the

coefficient on CH_LLR is positive and statistically significant, suggesting that failed banks with

31

larger loan loss reserve increases cost the FDIC insurance fund more upon failure. A single

standard-deviation increase in CH_LLR is associated with an increase in the cost of bank failure

of 2.4% (= 1.214 x 0.020 x 100%) of total assets.27 Given that the average total assets for our

sample of 136 failed banks is $887.92 million, a single standard-deviation increase in CH_LLR

costs the FDIC insurance fund an additional $21.31 million (= 0.024 x 887.92).

In Column II, the coefficient on LLP is positive and statistically significant at the 1%

level, while that on NCO is negative and statistically significant at the 10% level. Thus, the cost

of bank failures in 2008 and 2009 are associated positively with loan loss provisions and

negatively with net charge-offs. We present similar results with ABN_LLP and ABN_NCO in

Column III. Overall, the results in the section show that loan loss reserve decisions could be

associated with significant problems for the financial system during an economic crisis, not only

in terms of the incidence of bank failures, but also in their severity.

5. Conclusion

We exploit the 2008-2009 economic crisis to test the effects of adding back loan loss

reserves to regulatory capital on the future financial instability of banks. Our setting offers a

unique opportunity to observe financial instability in the form of bank failures. We document

that banks’ loan loss reserve increases in 2007 are associated positively with their failure risk in

2008-2009. More importantly, this positive association is observed only for banks that are less

constrained in adding back loan loss reserves to regulatory capital, suggesting that the treatment

of loan loss reserves as Tier 2 capital bears some potentially dysfunctional consequences that can

threaten bank survival during an economic downturn.

27 The standard deviation of CH_LLR within the sample of 136 failed banks is 1.214.

32

In examining the components of loan loss reserve changes, we find that bank failure risk

is influenced positively by loan loss provisions and negatively by loan charge-offs. The above

relations are all driven by banks that are less constrained in adding back loan loss reserves to

regulatory capital. Further, managerial discretion in loan loss accounting appears to play a

significant role in influencing the association of failure risk with both loan loss provisions and

loan charge-offs. Our findings are particularly important in the light of the recent debate on

whether to treat loan loss reserves as Tier 2 regulatory capital. They challenge the notion that the

buffer created by loan loss reserves lowers the risk of bank failure. The results also highlight that

opportunistic accounting decisions by bank managers can lead to adverse outcomes during a

crisis. Finally, our results indicate that the more specific analysis involved in reporting loan loss

charge-offs is probably more beneficial for a bank’s continued health than are loan loss

provisions.

On a broader note, our paper identifies a regulatory setting where there is a departure of

regulation from economic and accounting principles, in particular in the treatment of loan loss

reserves as Tier 2 capital. We provide evidence that such a departure could have adverse

consequences for the banking sector, in terms of the risk of bank failure. Current proposals under

consideration by regulators include calls to increase, even eliminate, the limit on the amount of

loan loss reserves that can be added back as regulatory capital. In this context, our findings are

important not just from a policy review perspective, but also from a policy-making perspective.

33

References

Ahmed, A. S., C. Takeda, and S. Thomas, 1999. Bank loan loss provisions: A reexamination of capital management, earnings management and signaling effects. Journal of Accounting and Economics 28: 1-25.

Beatty, A., S. L. Chamberlain, and J. Magliolo, 1995. Managing financial reports of commercial

banks: The influence of taxes, regulatory capital, and earnings. Journal of Accounting Research 33: 231-261.

Beatty, A., B. Ke, and K. Petroni, 2002. Earnings management to avoid earnings declines across

publicly and privately held banks. The Accounting Review 77: 547-570. Beatty, A. and S. Liao, 2009. Regulatory Capital ratios, loan loss provisioning and pro-

cyclicality, working paper. Beaver, W. H., and E. E. Engel, 1996. Discretionary behavior with respect to allowances for loan

losses and the behavior of security prices. Journal of Accounting and Economics 22: 177-206.

Betrand, M. and S. Mullainathan, 2003. Enjoying the quite life? Corporate governance and

managerial preferences. Journal of Political Economy 111: 1043-1075. Buser, S. A., A. H. Chen, and E. J. Kane, 1981. Federal deposit insurance, regulatory policy, and

optimal bank capital. Journal of Finance 35: 51-60. Bushman, R.M., and C.D. Williams, 2009. Accounting discretion, loan loss provisioning, and

discipline of banks’ risk-taking, working paper. Cox, D. R., 1972. Regression models and life-tables. Journal of the Royal Statistics Society 34:

187-220. Cox, D. R., and Oakes, D. 1984. Analysis of Survival Data. New York: Chapman & Hall. Dechow, P. M. 1994. Accounting earnings and cash flows as measures of firm performance: The

role of accounting accruals. Journal of Accounting and Economics 18: 3-42. Diamond, D. W., and R. G. Rajan, 2000. A theory of bank capital. Journal of Finance 55: 2431-

2465. Elliott, J. A., and J. D. Hanna, and W. H. Shaw, 1991. The evaluation by the financial markets of

changes in bank loan loss Reserve Levels. The Accounting Review 66: 847-861.

34

Estrella, A., S. Park, and S. Peristiani, 2000. Capital ratios as predictors of bank failure. FRBNY Economic Policy Review July 2000: 33-52.

Federal Reserve Board, 2009. Invitation to comment: Risk-Based Capital Guidelines; Capital

Adequacy Guidelines; Capital Maintenance: Regulatory Capital; Impact of Modifications to Generally Accepted Accounting Principles; Consolidation of Asset-Backed Commercial Paper Programs (Regulations H and Y) [R-1368]. http://www.federalreserve.gov/generalinfo/foia/index.cfm?doc_id=R%2D1368&doc_ver=1&ShowAll=Yes

Furlong, F. T., and M. C., Keeley, 1989. Capital regulation and bank risk-taking: A note. Journal

of Banking and Finance 13: 883–891. Kim, M., and W. Kross, 1998. The impact of the 1989 change in bank capital standards on loan

loss provisions and loan write-offs. Journal of Accounting and Economics 25: 69-99. Koehn, M., and A. Santomero, 1980. Regulation of bank capital and portfolio risk. Journal of

Finance 35: 1235-1244. Laeven, L., and R. Levine, 2009. Bank governance, regulation and risk taking. Journal of

Financial Economics 93: 259-275. Moyer, S. E., 1990. Capital adequacy ratio regulations and accounting choices in commercial

banks. Journal of Accounting and Economics 13: 123-154. Rime, B., 2001. Capital requirements and bank behavior: Empirical evidence for Switzerland.

Journal of Banking and Finance 25: 789-805. Saunders, A., E. Strock, and N. G. Travlos, 1990. Ownership structure, deregulation, and bank

risk taking. Journal of Finance 45: 643-654. Shrieves, R. E., and D. Dahl, 1992. The relationship between risk and capital in commercial

banks. Journal of Banking and Finance: 16 439–457. Shumway, T., 2001. Forecasting bankruptcy more accurately: A simple hazard model. Journal of

Business 74: 101-124. Wall, D. L., and T. W. Koch, 2000. Bank loan-loss accounting: A review of theoretical and

empirical evidence. Economic Review Q2 2000: 1-20. Whalen, J. M., 1994. The nature of information in commercial bank loan loss disclosures. The

Accounting Review 69: 455-478.

35

Appendix A Example of FDIC press release on bank failure

MB Financial Bank, National Association, Chicago, Illinois, Assumes All of the Deposits of Corus Bank, National Association, Chicago, Illinois

FOR IMMEDIATE RELEASE September 11, 2009

Media Contact: LaJuan Williams-Dickerson Office (202) 898-3876 Email: lwilliams-dickerson@fdic.gov

Corus Bank, National Association, Chicago, Illinois, was closed today by the Office of the Comptroller of the Currency, which appointed the Federal Deposit Insurance Corporation (FDIC) as receiver. To protect the depositors, the FDIC entered into a purchase and assumption agreement with MB Financial Bank, National Association, Chicago, Illinois, to assume all of the deposits of Corus Bank, N.A.

The eleven branches of Corus Bank will reopen on their next normally scheduled business day as branches of MB Financial Bank. Depositors of Corus Bank will automatically become depositors of MB Financial Bank. Deposits will continue to be insured by the FDIC, so there is no need for customers to change their banking relationship to retain their deposit insurance coverage. Customers should continue to use their existing branches until MB Financial Bank can fully integrate the deposit records of Corus Bank.

This evening and over the weekend, depositors of Corus Bank can access their money by writing checks or using ATM or debit cards. Checks drawn on the bank will continue to be processed. Loan customers should continue to make their payments as usual.

As of June 30, 2009, Corus Bank had total assets of $7 billion and total deposits of approximately $7 billion. MB Financial Bank will pay the FDIC a premium of 0.2 percent to assume all of the deposits of Corus Bank. In addition to assuming all of the deposits of the failed bank, MB Financial Bank agreed to purchase approximately $3 billion of the assets, comprised mainly of cash and marketable securities. The FDIC will retain the remaining assets for later disposition. The FDIC plans to sell substantially all of the remaining assets of Corus Bank in the next 30 days in a private placement transaction.

Customers who have questions about today's transaction can call the FDIC toll-free at 1-800-823-5017. The phone number will be operational this evening until 9:00 p.m., Central Daylight Time (CDT); on Saturday from 9:00 a.m. to 6:00 p.m., CDT; on Sunday from noon to 6:00 p.m., CDT; and thereafter from 8:00 a.m. to 8:00 p.m., CDT. Interested parties can also visit the FDIC's Web site at http://www.fdic.gov/bank/individual/failed/corus.html.

The FDIC estimates that the cost to the Deposit Insurance Fund (DIF) will be $1.7 billion. MB Financial Bank's acquisition of all the deposits was the "least costly" resolution for the FDIC's DIF compared to alternatives. Corus Bank is the 90th FDIC-insured institution to fail in the

36

nation this year, and the sixteenth in Illinois. The last FDIC-insured institution closed in the state was Platinum Community Bank, Rolling Meadows, on September 4, 2009.

# # #

Congress created the Federal Deposit Insurance Corporation in 1933 to restore public confidence in the nation's banking system. The FDIC insures deposits at the nation's 8,195 banks and savings associations and it promotes the safety and soundness of these institutions by identifying, monitoring and addressing risks to which they are exposed. The FDIC receives no federal tax dollars – insured financial institutions fund its operations.

FDIC press releases and other information are available on the Internet at www.fdic.gov, by subscription electronically (go to www.fdic.gov/about/subscriptions/index.html) and may also be obtained through the FDIC's Public Information Center (877-275-3342 or 703-562-2200). PR-168-2009

37

Appendix B Call report: Reporting changes in loan loss reserves

The image part with relationship ID rId15 was not found in the file.

38

Appendix C Variable definitions

The image part with relationship ID rId16 was not found in the file.

39

The image part with relationship ID rId17 was not found in the file.

40

Table 1 Bank failures by year

This table provides information on bank and thrift failures by calendar year from 2001 to 2009. Panel A (B) shows the failure of commercial banks (thrifts) from 2001 to 2009.

Panel A: Failure of commercial banks

Year FailuresTotal

Assets ($m)Total

Deposits ($m)Bank failures

with FDIC cost infoTotal

Cost ($m)

2001 3 58.6 51.6 3 4.62002 10 2,656.4 2,291.6 4 361.92003 3 961.2 903.2 2 135.62004 3 150.8 140.1 3 14.12005 0 0.0 0.0 . .2006 0 0.0 0.0 . .2007 2 102.5 89.2 1 3.02008 20 17,963.8 14,898.6 19 4,580.52009 120 119,175.1 97,596.8 120 24,100.9

Panel B: Failure of thrifts

Year FailuresTotal

Assets ($m)Total

Deposits ($m)Bank failures

with FDIC cost infoTotal

Cost ($m)

2001 1 2,300.0 1,600.0 0 .2002 1 52.0 40.0 0 .2003 0 0.0 0.0 . .2004 1 12.3 9.8 0 .2005 0 0.0 0.0 . .2006 0 0.0 0.0 0 0.02007 1 2,500.0 2,300.0 1 110.02008 5 401,694.6 224,332.1 5 12,842.02009 19 51,709.1 39,844.6 18 12,174.8

41

Table 2 Within-sample distribution of commercial bank failures across the United States This table shows the distribution of commercial bank failures in 2008 and 2009 within our sample of 7,493 commercial banks. This sample of 7,493 banks is the number of commercial banks in existence at the end of 2007 that have the data needed to compute the variables used in our analysis of the relation between loan loss reserve accounting and bank failure (see Table 4). Within each parenthesis, the number of bank failures is indicated on the left; the total number of banks is indicated on the right.

New England Middle Atlantic

Connecticut (0 / 46) New Jersey (2 / 86)Maine (0 / 27) New York (0 / 131)Massachusetts (0 / 159) Pennsylvania (0 / 198)New Hampshire (0 / 17)Rhode Island (0 / 8)Vermont (0 / 14)

East North Central

Indiana (1 / 121) Iowa (0 / 371) Nebraska (1 / 236)Illinois (21 / 612) Kansas (4 / 338) North Dakota (0 / 93)Michigan (4 / 148) Minnesota (7 / 417) South Dakota (1 / 84)Ohio (0 / 186) Missouri (4 / 330)Wisconsin (1 / 268)

South Atlantic East South Central West South Central

Delaware (0 / 23) Alabama (2 / 143) Arkansas (1 / 142)District of Columbia (0 / 6) Kentucky (0 / 185) Louisiana (0 / 138)Florida (12 / 257) Mississippi (0 / 92) Oklahoma (1 / 252)Georgia (29 / 319) Tennessee (0 / 181) Texas (6 / 612)Maryland (0 / 55)North Carolina (2 / 92)South Carolina (0 / 68)Virginia (0 / 100)West Virginia (0 / 62)

Mountain Pacific

Arizona (4 / 45) Montana (0 / 74) Alaska (0 / 5)Colorado (3 / 140) Utah (2 / 59) California (17 / 261)Idaho (0 / 15) Nevada (5 / 30) Hawaii (0 / 7)New Mexico (0 / 46) Wyoming (1 / 40) Oregon (3 / 36)

Washington (3 / 88)

Region 1: Northeast (2 / 686 = 0.29% )

Region 2: Midwest (44 / 3204 = 1.37% )

West North Central

Region 3: South (53 / 2727 = 1.94% )

Region 4: West (38 / 846 = 4.49% )

42

Table 3 Capital ratios and loan loss reserves This table provides information on the capital ratios and loan loss reserves of banks in our sample of 7,463 banks. Panel A presents the summary statistics for all banks. In Panel B, tests of differences between non-failed and failed banks are conducted. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Panel A: Summary statistics for all banks Variable Mean Std P25 Median P75

Total risk-based capital ratio -Total capital scaled by risk-weighted assets 17.280 25.094 11.680 14.110 18.610

Components of total risk-based capital ratio:Tier 1 capital scaled by risk-weighted assets 16.184 25.099 10.590 13.040 17.530Tier 2 capital scaled by risk-weighted assets 1.097 0.448 0.912 1.119 1.252Deductions for total risk-based capital scaled by risk-weighted assets 0.001 0.032 0.000 0.000 0.000

Loan loss reserves used in Tier 2 capital scaled by risk-weighted assets 1.037 0.247 0.897 1.103 1.251

Loan loss reserves scaled by gross risk-weighted assets 1.211 0.819 0.894 1.098 1.358

Excess loan loss reserves scaled by gross risk-weighted assets 0.183 0.733 0.000 0.000 0.122

43

Panel B: Univariate comparison of non-failed banks and failed banks

FAIL = 0 FAIL = 1

Number 7,326 137

Total risk-based capital ratio -Total capital scaled by risk-weighted assets 17.384 11.713 -5.671 ***

Components of total risk-based capital ratio:i) Tier 1 capital scaled by risk-weighted assets 16.291 10.474 -5.818 ***ii) Tier 2 capital scaled by risk-weighted assets 1.094 1.240 0.146 ***iii) Deductions for total risk-based capital scaled by risk-weighted assets 0.001 0.000 -0.001

Loan loss reserves used in Tier 2 capital scaled by risk-weighted assets 1.035 1.134 0.099 ***

Loan loss reserves scaled by gross risk-weighted assets 1.202 1.727 0.525 ***

Excess loan loss reserves scaled by gross risk-weighted assets 0.175 0.602 0.427 ***

Difference

44

Table 4 Descriptive statistics

This table provides some descriptive statistics of the variables used in the analysis of bank failure. Panel A provides the summary statistics for the banks in our sample of 7,463 banks. Panel B presents univariate comparisons between non-failed and failed banks. All the various have been winsorized at the 1st and 99th percentile within the year. The definitions of all the variables are found in Appendix C. In Panel B, tests of differences between non-failed and failed banks are conducted. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Panel A: Summary statistics for all banks

Mean Std P25 Median P75Main variables

FAIL 0.018 0.134 0.000 0.000 0.000CH_LLR 0.105 0.341 -0.026 0.053 0.190LLP 0.347 0.570 0.043 0.173 0.397NCO 0.247 0.481 0.011 0.091 0.261

Control variables

REAL_ESTATE_LOAN 68.681 19.597 57.689 72.321 83.048NPL 2.511 2.582 0.769 1.795 3.360ASSET_RISK 57.009 17.225 45.304 58.191 69.595OVERHEAD 3.019 1.079 2.374 2.869 3.449INSIDER_LOAN 1.444 1.650 0.248 0.890 2.055ROA 0.859 0.888 0.514 0.918 1.294UNINSURED_DEPOSIT 0.405 0.156 0.295 0.382 0.490LIQUIDITY 5.317 4.814 2.812 3.935 5.895LEVERAGE 88.722 4.166 87.515 89.898 91.363TOTAL_RBC 16.661 7.656 11.680 14.110 18.610TOTAL_ASSETS ($b) 0.469 1.424 0.063 0.134 0.308

45

Panel B: Univariate comparison of non-failed banks and failed banks

FAIL = 0 FAIL = 1 FAIL = 0 FAIL = 1

Number 7,326 137 7,326 137

Main variables

CH_LLR 0.096 0.560 0.464 *** 0.051 0.437 0.386 ***LLP 0.331 1.161 0.830 *** 0.168 0.873 0.706 ***NCO 0.241 0.567 0.327 *** 0.090 0.268 0.178 ***

Control variables

REAL_ESTATE_LOAN 68.418 82.738 14.320 *** 72.088 84.297 12.208 ***NPL 2.431 6.812 4.381 *** 1.762 5.688 3.926 ***ASSET_RISK 56.731 71.875 15.145 *** 57.906 72.944 15.038 ***OVERHEAD 3.019 3.017 -0.002 2.869 2.804 -0.066INSIDER_LOAN 1.446 1.339 -0.108 0.891 0.783 -0.109ROA 0.874 0.053 -0.821 *** 0.925 0.419 -0.506 ***UNINSURED_DEPOSIT 0.404 0.460 0.056 *** 0.380 0.447 0.067 ***LIQUIDITY 5.359 3.090 -2.269 *** 3.964 2.374 -1.590 ***LEVERAGE 88.694 90.241 1.547 *** 89.875 91.195 1.320 ***TOTAL_RBC 16.749 11.916 -4.833 *** 14.180 10.820 -3.360 ***TOTAL_ASSETS ($b) 0.461 0.878 0.416 *** 0.132 0.273 0.142 ***

Difference DifferenceMeans Medians

46

Table 5 Bank failure and loan loss reserve changes

This table presents Cox proportional hazard regressions that analyze the relation between bank failure and loan loss reserve changes. The sample consists of 7,463 banks. The definitions of all the variables are found in Appendix C. The t-statistic of each coefficient is provided in brackets below the coefficient. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Panel A: Hazard regression results

I II

CH_LLR 0.507**[2.37]

LLP 0.304**[1.98]

NCO -0.328*[-1.90]

REAL_ESTATE_LOAN 0.032*** 0.031***[4.74] [4.44]

NPL 0.209*** 0.215***[9.17] [9.32]

ASSET_RISK 0.036*** 0.036***[4.09] [4.07]

OVERHEAD -0.145 -0.146[-1.58] [-1.57]

INSIDER_LOAN -0.106* -0.117*[-1.74] [-1.89]

ROA -0.349*** -0.395***[-3.44] [-3.37]

UNINSURED_DEPOSIT 0.195 0.197[0.33] [0.33]

LIQUIDITY -0.099** -0.112**[-2.22] [-2.49]

LEVERAGE 0.107** 0.098**[2.55] [2.28]

TOTAL_RBC -0.113** -0.115**[-2.23] [-2.26]

TOTAL_ASSETS ($b) 0.101** 0.104**[2.18] [2.24]

REGION2 1.660** 1.678**[2.27] [2.28]

REGION3 1.555** 1.598**[2.13] [2.18]

REGION4 2.510*** 2.510***[3.37] [3.36]

Dependent variable is H(t), hazard of bank failure at time t

47

Panel B: Hazard ratios

CH_LLR 1.189LLP 1.189NCO 0.854REAL_ESTATE_LOAN 1.872 1.836NPL 1.716 1.742ASSET_RISK 1.859 1.859OVERHEAD 0.855 0.854INSIDER_LOAN 0.840 0.824ROA 0.733 0.704UNINSURED_DEPOSIT 1.031 1.031LIQUIDITY 0.621 0.583LEVERAGE 1.562 1.504TOTAL_RBC 0.421 0.415TOTAL_ASSETS ($b) 1.155 1.160REGION2 5.259 5.355REGION3 4.735 4.943REGION4 12.305 12.305

48

Table 6 The role of regulatory capital constraints This table presents Cox proportional hazard regressions that analyze the role of regulatory capital constraints on the relation between bank failure and loan loss reserve changes. The definitions of all the variables are found in Appendix C. The t-statistic of each coefficient is provided in brackets below the coefficient. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

I II III IV

CH_LLR x CONSTRAINT -0.644** -0.603*[-2.05] [-1.68]

LLP x CONSTRAINT -0.601** -0.996*** [-2.13] [-3.12]NCO x CONSTRAINT 0.449 0.979**

[1.17] [2.28]CH_LLR 0.721*** 0.827***

[2.96] [2.79]LLP 0.392*** 0.797***

[2.59] [3.93]NCO -0.268 -0.758***

[-1.43] [-2.81]CONSTRAINT -0.124 -0.085 -0.101 0.033

[-0.47] [-0.30] [-0.33] [0.10]REAL_ESTATE_LOAN 0.033*** 0.033*** 0.032*** 0.032***

[4.74] [4.67] [4.72] [4.57]NPL 0.216*** 0.223*** 0.216*** 0.227***

[9.58] [9.75] [9.53] [9.93]ASSET_RISK 0.039*** 0.041*** 0.039*** 0.040***

[4.33] [4.46] [4.33] [4.42]OVERHEAD -0.137 -0.155 -0.128 -0.147

[-1.46] [-1.64] [-1.37] [-1.55]INSIDER_LOAN -0.104* -0.103* -0.102* -0.103*

[-1.71] [-1.67] [-1.66] [-1.68]ROA -0.371*** -0.423*** -0.355*** -0.407***

[-3.62] [-3.63] [-3.47] [-3.51]UNINSURED_DEPOSIT 0.333 0.377 0.304 0.371

[0.56] [0.62] [0.51] [0.61]LIQUIDITY -0.098** -0.108** -0.096** -0.103**

[-2.22] [-2.42] [-2.17] [-2.34]LEVERAGE 0.138*** 0.135*** 0.139*** 0.121***

[3.21] [3.04] [3.20] [2.65]TOTAL_RBC -0.090* -0.091* -0.093* -0.105**

[-1.79] [-1.79] [-1.83] [-2.02]TOTAL_ASSETS ($b) 0.104** 0.106** 0.101** 0.096**

[2.25] [2.29] [2.17] [2.04]

REGION2 1.749** 1.747** 1.680** 1.668**

[2.39] [2.37] [2.30] [2.27]

REGION3 1.524** 1.551** 1.494** 1.473**

[2.08] [2.11] [2.04] [2.00]

REGION4 2.513*** 2.504*** 2.496*** 2.461***

[3.36] [3.34] [3.34] [3.29]

Dependent variable is H(t), hazard of bank failure at time tCONSTRAINT1 CONSTRAINT2

49

I II III IV

CH_LLR x CONSTRAINT 0.077 0.224 + CONSTRAINT [0.28] [0.87]

LLP x CONSTRAINT -0.209 -0.199 + CONSTRAINT [-0.74] [-0.87]

NCO x CONSTRAINT 0.181 0.221 + CONSTRAINT [0.52] [0.81]

CONSTRAINT1 CONSTRAINT2

50

Table 7 Decomposition of loan loss provisions and net charge-offs

This table presents the ordinary least squares regressions that decompose loan loss provisions and net charge-offs to obtain abnormal loan loss provisions and abnormal net charge-offs, respectively. The sample of consists of 7,463 banks. The definitions of all the variables are found in Appendix C. The t-statistic of each coefficient is provided in brackets below the coefficient. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

Dependent variable is LLP Dependent variable is NCO

Intercept 0.063 -0.108**[1.12] [-2.28]

CH_NPL 0.079*** 0.035***[25.03] [13.15]

LLR 0.124*** 0.203***[11.97] [23.12]

LOAN 0.006*** 0.003***[13.59] [8.71]

LOANR -0.004*** -0.002***[-8.36] [-5.49]

LOANOTH -0.001* 0.001*[-1.96] [1.80]

Observations 7463 7463Adjusted R-square (%) 12.12 11.17

51

Table 8 Bank failure, abnormal loan loss provisions, and abnormal charge-offs

This table presents Cox proportional hazard regressions that analyze the relation between bank failure, abnormal loan loss provisions, and abnormal net charge-offs. The definitions of all the variables are found in Appendix C. The t-statistic of each coefficient is provided in brackets below the coefficient. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

CONSTRAINT1 CONSTRAINT2I II III

ABN_LLP 0.287* 0.386** 0.801***[1.79] [2.42] [3.69]

ABN_NCO -0.321* -0.273 -0.792***[-1.72] [-1.36] [-2.67]

ABN_LLP x CONSTRAINT -0.666** -1.062***[-2.06] [-2.98]

ABN_NCO x CONSTRAINT 0.478 1.046**[1.11] [2.17]

CONSTRAINT -0.281 -0.213[-1.25] [-0.82]

REAL_ESTATE_LOAN 0.031*** 0.033*** 0.031***[4.50] [4.65] [4.59]

NPL 0.223*** 0.232*** 0.235***[10.44] [10.90] [11.00]

ASSET_RISK 0.037*** 0.041*** 0.041***[4.20] [4.56] [4.51]

OVERHEAD -0.15 -0.155* -0.145[-1.62] [-1.65] [-1.55]

INSIDER_LOAN -0.115* -0.103* -0.104*[-1.85] [-1.68] [-1.69]

ROA -0.399*** -0.432*** -0.417***[-3.41] [-3.74] [-3.66]

UNINSURED_DEPOSIT 0.215 0.358 0.352[0.36] [0.60] [0.59]

LIQUIDITY -0.113** -0.109** -0.104**[-2.53] [-2.44] [-2.36]

LEVERAGE 0.101** 0.136*** 0.122***[2.36] [3.08] [2.71]

TOTAL_RBC -0.114** -0.092* -0.105**[-2.24] [-1.80] [-2.03]

TOTAL_ASSETS ($b) 0.103** 0.103** 0.093*[2.22] [2.23] [1.96]

REGION2 1.675** 1.756** 1.685**[2.28] [2.38] [2.29]

REGION3 1.598** 1.568** 1.495**[2.18] [2.13] [2.03]

REGION4 2.518*** 2.529*** 2.486***[3.37] [3.37] [3.32]

Dependent variable is H(t), hazard of bank failure at time t

52

ABN_LLP x CONSTRAINT -0.280 -0.261

+ CONSTRAINT [-0.89] [-1.05]

ABN_NCO 0.204 0.254 + CONSTRAINT [0.52] [0.83]

53

Table 9 Correlations between changes in capital ratios and changes in risk-taking This table presents the Pearson correlations between changes in capital ratio and changes in risk-taking. CH_TIER1_RW is the change in Tier 1 capital scaled by total risk-weighted assets, CH_TIER2_RW is the change in Tier 2 capital scaled by total risk-weighted assets, CH_LLR_RW is the change in loan loss reserves scaled by total risk-weighted assets, CH_LOANS is the change in loans as a percentage of total assets, CH_LOAN_RISK is the change in loans at 50% and 100% risk weights as a percentage of total loans. CH_ASSET_RISK is the change in assets at 50% and 100% risk weights as a percentage of total assets. This exploratory analysis of the association between changes in capital ratio and changes in risk-taking is limited to changes that occur in the first quarter of 2008, which is the first quarter after the recession began. Significance levels are based on two-tailed tests. The p-value for each correlation is indicated in italics.

CH_TIER2_RW CH_LLR_RW CH_LOANS CH_LOAN_RISK CH_ASSET_RISK

CH_TIER1_RW -0.208 -0.213 -0.108 -0.119 -0.079

<.0001 <.0001 <.0001 <.0001

CH_TIER2_RW 0.872 0.028 0.029 0.027

<.0001 0.015 0.012 0.021

CH_LLR_RW 0.030 0.031 0.031

0.010 0.009 0.007

CH_LOANS 0.959 0.845

<.0001 <.0001

CH_LOAN_RISK 0.854

<.0001

54

Table 10 FDIC cost and loan loss reserve changes This table presents the ordinary least squares regressions that analyze, for the banks that have failed, the relation between the FDIC cost of bank failure, loan loss provisions, and net charge-offs. The sample of consists of 136 banks that failed in 2008 and 2009. The definitions of all variables are found in Appendix C. The t-statistic of each coefficient is provided in brackets below the coefficient. Significance levels are based on two-tailed tests. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

I II III

Intercept -0.386 -0.212 -0.245[-0.74] [-0.41] [-0.48]

CH_LLR 0.020**[2.09]

LLP 0.044***[3.22]

NCO -0.031*[-1.94]

ABN_LLP 0.043***[3.01]

ABN_NCO -0.029*[-1.67]

REAL_ESTATE_LOAN 0.000 0.000 0.000[0.36] [0.22] [0.12]

NPL -0.001 0.000 0.000[-0.39] [-0.32] [0.20]

ASSET_RISK 0.001 0.001 0.001[1.22] [0.62] [0.76]

OVERHEAD -0.004 -0.001 -0.001[-0.49] [-0.13] [-0.18]

INSIDER_LOAN -0.012* -0.011* -0.011*[-1.91] [-1.75] [-1.85]

ROA 0.007 0.011 0.011[0.96] [1.46] [1.43]

UNINSURED_DEPOSIT 0.014 0.006 0.006[0.25] [0.10] [0.11]

LIQUIDITY 0.000 -0.002 -0.002[-0.13] [-0.54] [-0.64]

LEVERAGE 0.002 0.000 0.000[0.36] [-0.02] [0.08]

TOTAL_RBC 0.001 -0.002 -0.001[0.13] [-0.38] [-0.26]

TOTAL_ASSETS ($b) -0.011** -0.011** -0.011**[-2.15] [-2.14] [-2.17]

DEPOSITS 0.449*** 0.438*** 0.450***[3.69] [3.69] [3.77]

REGION2 0.034 0.076 0.061[0.44] [0.99] [0.79]

REGION3 0.053 0.094 0.083[0.69] [1.22] [1.08]

REGION4 0.032 0.072 0.060[0.40] [0.91] [0.75]

Observations 136 136 136Adjusted R-square (%) 26.34 29.92 29.11

Dependent variable is FDIC_COST