How bank size relates to the impact of bank stress on the real economy

Journal Pre-proof How Bank size relates to the impact of Bank stress on the real economy

Amy G. Lorenc, Jeffery Y. Zhang PII:

S0929-1199(20)30036-5

DOI:

https://doi.org/10.1016/j.jcorpfin.2020.101592

Reference:

CORFIN 101592

To appear in:

Journal of Corporate Finance

Received date:

24 October 2018

Revised date:

15 October 2019

Accepted date:

9 February 2020

Please cite this article as: A.G. Lorenc and J.Y. Zhang, How Bank size relates to the impact of Bank stress on the real economy, Journal of Corporate Finance(2020), https://doi.org/ 10.1016/j.jcorpfin.2020.101592

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier.

Journal Pre-proof How Bank Size Relates to the Impact of Bank Stress on the Real Economy Amy G. Lorenc and Jeffery Y. Zhang1 September 18, 2019 Abstract We examine whether financial stress at larger banks has a different impact on the real economy

of

than financial stress at smaller banks. Our empirical results show that stress experienced by

ro

banks in the top 1 percent of the size distribution leads to a statistically significant and negative

-p

impact on the real economy. This impact increases with the size of the bank. The negative impact

re

on quarterly real GDP growth caused by stress at banks in the top 0.15 percent of the size

lP

distribution is more than twice as large as the impact caused by stress at banks in the top 0.75 percent, and more than three times as large as the impact caused by stress at banks in the top 1

na

percent. These results are broadly informative as to how the stringency of regulatory standards should vary with bank size, and support the idea that the largest banks should be subject to the

Jo

requirements.

ur

most stringent requirements while smaller banks should be subject to successively less stringent

Keywords: Bank Size, Bank Failures, Systemic Risk, Financial Regulation, Tailoring JEL: G21, G28

1

Amy G. Lorenc was a senior financial analyst and Jeffery Y. Zhang was an economist in the Division of Supervision and Regulation at the Board of Governors of the Federal Reserve System. We thank Brian Chernoff, Michael Gibson, Art Lindo, Elizabeth MacDonald, Dennis Mawhirter, Marco Migueis, Michael Suher, Teng Wang, and other colleagues in the Division of Supervision and Regulation for helpful discussions. We also thank Sean Campbell of the Financial Services Forum for his contributions to this project during his time in the Division of Supervision and Regulation. Please note that the views presented in this article are the authors’ alone and do not necessarily represent the views of the Board of Governors of the Federal Reserve System or the views of the United States government.

Jo

ur

na

lP

re

-p

ro

of

Journal Pre-proof

Journal Pre-proof

I.

Introduction This article contributes to an ongoing effort to understand the impact of stress in the

banking sector on the real economy. The experiences of past financial crises (Bernanke 1983, Calomiris and Mason 2003, Kupiec and Ramirez 2013, Reinhart and Rogoff 2009) and of the most recent one (Bernanke 2015, Chodorow-Reich 2014, Geithner 2015, Gorton 2010) have taught us that stress within the aggregate banking sector can lead to stress in the real economy.

of

This article considers whether stress at large banks has a different impact on the real economy

ro

than stress at small banks.

-p

Using the assets of banks that failed or received assistance from the Federal Deposit

re

Insurance Corporation (FDIC) as a proxy for bank stress, our empirical results show that

lP

financial stress at large banks has a statistically significant and negative impact on the real economy. Moreover, this impact increases with bank size. For instance, after scaling our

na

empirical results to the size distribution of banks in the fourth quarter of 2017, we find that the

ur

negative impact on real quarterly GDP growth associated with stress at banks with greater than $200 billion in total assets is more than twice as large as the impact associated with stress at

Jo

banks with greater than $50 billion in total assets, and more than three times as large as the impact associated with stress at banks with greater than $30 billion in total assets. These results are qualitatively similar when considering the impact on the unemployment rate. Our conclusion that the failure of larger banks has a disproportionate impact on the real economy than the failure of smaller banks should not be surprising. As noted by Bremus et al. (2018), scholars have long theorized a link between bank size and the real economy through the destabilization of aggregate credit—bail-out expectations may invite imprudent risky behavior by large banks, and close linkages between large banks and highly leveraged shadow financial

Journal Pre-proof institutions may destabilize the entire financial system. In addition, large banks are more interconnected (Federal Reserve 2016), and they operate in a much broader array of services and customers. Thus, when a large bank fails, a greater number of connected parties are affected simultaneously, and there is a greater risk of contagion (Scott 2016). On the other end of the size spectrum, smaller banks typically do not enjoy the same level of implicit subsidies; do not

generally have fewer product lines and fewer counterparties.

of

engage in the same frequency or magnitude of transactions with nonbank financial entities; and

ro

Recent literature has shown that systemic measurements of large banks are better at

-p

forecasting future economic downturns than those of small banks (Allen, Bali, and Tang 2012);

re

that large banks take on more leverage relative to small banks in times of stress (Dávila and

lP

Walther 2017); and that large banks have an outsized economic impact even in the absence of contagion, spillover effects, or shared responses to macroeconomic shocks (Bremus et al. 2018).

na

In addition to contributing to the academic literature, our results broadly inform the

ur

policy “threshold” debate that ensues among regulators, public officials, and the private sector.

Jo

There is a growing consensus that financial regulation should be tailored according to a firm’s risk profile, with the goal of limiting the damage done to the real economy in future financial crises while minimizing the costs of increased regulation. It is generally agreed upon that banks posing the greatest risks to financial stability should be subject to the most stringent regulations and that regulatory burden should be reduced for less risky banks. Determining where to draw the line between banks that should and should not be subject to enhanced regulation has proven to be a challenging task. For one, size is not the only predictor of the potential impact that a distressed bank would have on the economy. The complexity of a bank’s operations could also play a role (OFR 2017). Unfortunately, reliable indicators of bank

Journal Pre-proof complexity are not available over the sample period examined in this paper. We therefore focus on size alone. Our baseline regression specification assumes that bank stress affects economic performance—GDP growth or the unemployment rate—within the same quarter. Thus, at each point in time, we calculate the specific percentiles of the bank distribution, denoted by p. We then derive the relevant amount of total assets for a given value of p in each quarter.

As

of

discussed later, we also create a stress proxy for each quarter, which captures the total assets of

ro

failed banks above a certain percentile p of the size distribution. Similarly, in every quarter, we

-p

have size and stress proxies that corresponds to failed banks below a certain percentile p of the

re

distribution. We include the stress measure for the 1-p set of banks in order to control for the

lP

impact of stress at smaller banks. We therefore allow the impact of financial stress to differ across large and small banks.

na

The rest of this article is organized as follows: Section II reviews the relevant literature.

ur

Section III describes the data used in our empirical analysis, including the proxy we use for bank

Jo

stress. Section IV details the regression specifications. Section V contains the baseline results for our GDP and unemployment rate regressions; it also presents the baseline results in the form of a hypothetical scenario in which we assume a particular level of bank stress. Section VI describes a series of robustness checks such as conducting regional analysis and including an array of additional control variables to mitigate concerns of omitted variable bias; using monthly data instead of quarterly data to confirm our contemporaneous modeling assumption; performing a check for reverse causality; and removing the financial crisis from the sample. Section VII concludes with a brief discussion of policy implications and suggestions for future research.

Journal Pre-proof

II.

Literature Review Our analysis is inspired by Bernanke’s seminal 1983 article, which demonstrates the

existence of a statistically and economically significant link between bank failures and real economic activity during the Great Depression. 2 Since Bernanke (1983), scholars have further examined the connection between banks and economic activity, across banks of varying size and

of

with a particular focus on large banks.

ro

For example, Boyd and Gertler (1993) argue that bad real estate loans by large banks was the primary source of the U.S. banking crisis in the 1980s. Dávila and Walther (2017) examine

-p

how large and small banks make funding decisions when the government provides system-wide

re

bailouts to the financial sector, and conclude that bank size is a key determinant of banks’

lP

leverage choices, even when bailout policies treat large and small banks symmetrically. They show that a large bank takes on more leverage than a small bank because, relative to a small

na

bank, a large bank internalizes the fact that its actions directly affect the magnitude of the

ur

government’s optimal bailout, which generates an additional incentive to take on debt. Acharya and Steffen (2015) argue that Eurozone bank risks during 2007-2012 can be understood as a

Jo

“carry trade.” That is, bank equity returns load positively on peripheral—Greece, Ireland, Portugal, Spain and Italy—bond returns and negatively on German government bond returns, a position that generated “carry” until the deteriorating peripheral bond returns inflicted losses on banks. The authors find evidence that large banks are more likely to engage in carry trades to earn higher and riskier returns on their diminished economic capital. Bremus et al. (2018)

2

As discussed further in Sections III and IV, our proxy of bank stress is not an exact replica of Bernanke’s (1983) bank stress proxy. The value of deposits at failed banks during the Great Depression —the three-year period of analysis in Bernanke (1983)—was in the hundreds of thousands of dollars. Measuring deposits in levels (namely, thousands of dollars) makes sense for that time period. Our analysis spans almost six decades, so we have to account for the structural changes in the banking sector.

Journal Pre-proof construct a model of granularity in the financial sector and demonstrate the outsized influence of large banks. They conclude that policies that ultimately increase concentration in the banking sector can increase aggregate volatility in macroeconomic outcomes. The literature also covers large banks and proxies of systemic risk. Allen, Bali, and Tang (2012) construct a measure of systemic risk, designated CATFIN, and show that the CATFIN of both large and small banks can forecast macroeconomic declines, though the CATFIN of large

of

banks can successfully forecast lower economic activity sooner than that of small banks. In their

ro

article, a bank is small if its stock market capitalization is below the top quintile among firms

-p

traded on the NYSE. Adrian and Brunnermeier (2016) construct a measurement of systemic risk,

re

designated CoVar, and show that firms with higher leverage, more maturity mismatch, and larger

lP

size are associated with larger systemic risk contributions. Specifically, the authors find that if a bank is 10 percent larger than another bank, then the size coefficient predicts that the larger

na

bank’s CoVaR per unit of capital is 27 basis points higher than the smaller bank’s CoVaR.

ur

Similarly, research conducted by the Bank for International Settlements arrives at the same qualitative conclusion as the one presented in this article. Tarashev et al. (2010) measure

Jo

individual institutions’ systemic importance by applying the Shapley value methodology to the field of risk attribution. The authors find that, all else equal, the ratio of one institution’s systemic importance to a smaller institution’s systemic importance is larger than the ratio of their respective sizes. It implies that prudential regulations for systemic importance should increase faster than size. Notably, Ashcraft (2005) provides a key contribution that sheds light on the link between bank size, bank stress, and economic performance. Using county-level data over multiple decades, the author demonstrates that the failures of small banks are associated with a greater

Journal Pre-proof decline in county-level income than the failures of large banks. It is important to note that, in the author’s analysis, “small [bank] failures involve a ratio of deposits to income below the ninetieth percentile.” Given the fact that the United States has thousands of banks, banks become very small once we go below the 90th percentile of the distribution, no matter along which dimension. Indeed, at the 99th percentile of the size distribution, the banks are already down to $30 billion in assets. Therefore, we believe that our analysis complements that of Ashcraft (2005), as we focus

of

on the top 1 percent of the bank distribution. This distinction also is crucial because our analysis

ro

can inform the macroprudential regulatory debate following the Economic Growth, Regulatory

-p

Relief, and Consumer Protection Act of 2018. That statute requires the FDIC, the OCC, and the Federal Reserve to tailor prudential regulations for banks between $100 billion and $250 billion

Jo

ur

na

lP

re

in size. These sizes map onto the 99.5th percentile of the distribution and above.

Journal Pre-proof

III.

Data We examine the relationship between stress experienced by banks of different sizes and

real economic activity. To do so, we employ proxies for bank size, bank stress, and real economic activity. We perform our analysis over the period from first quarter of 1960 through the fourth quarter of 2017, which includes the savings and loan crisis of the 1980s and 1990s and the Great Recession. This long sample period is necessary to ensure an appropriate number of

ro

of

observations.

As is standard practice, we use total assets as reported in the Consolidated Reports of

-p

Condition and Income (Call Reports) as a proxy for bank size. 3 This data is reported quarterly by

re

every national bank, state member bank, state nonmember bank, and savings association. 4

lP

We define stress experienced by an individual bank as a failure or assistance transaction

na

included in the FDIC’s Historical Statistics on Banking (HSOB). The HSOB contains transactions for FDIC-insured institutions, including insured commercial banks, insured savings

ur

institutions, and insured branches of foreign banks. A failed or assisted institution is defined as

Jo

an institution that has failure stature or has been provided assistance by the FDIC in merging with another institution. For simplicity, we refer to both failures and assistance transactions as “failures” in the remainder of this article.5 Similarly, “failed assets” (or “failed deposits”) refers to the amount of assets (or deposits) at institutions that failed or received assistance transactions.

3

We use data collected by the Federal Financial Institutions Examination Council 031, 041, and 051 reporting forms. 4

We also performed the analysis described in this article using bank holding company (BHC) data from the Y-9C. The results show a similar relationship between stress experienced by BHCs and impact on the real economy. The greatest economic impact results from stress at the largest BHCs, and the impact declines as the universe of BHCs included in the analysis encompasses smaller and smaller BHCs. 5

The FDIC data used to create our bank stress measure includes both bank failures and assistance. We performed a robustness check using only bank failures, and the results remain qualitatively identical.

Journal Pre-proof The HSOB contains information on the total assets of the failed bank, as well as a number of other fields. We create a time series measure of stress in the broader banking sector equal to the total assets of banks that failed within a particular quarter, based on the effective date provided by the HSOB, and use this as our baseline measure of bank stress. We inflation adjust our stress measure using the Consumer Price Index (CPI) and take the natural logarithm. Our measurement of bank stress differs from that of Bernanke (1983) in a couple of

of

ways, in order to account for structural changes in the banking sector. First, note that the value of

ro

deposits at failed banks from 1929 to 1933—the period of analysis in Bernanke (1983)—was in

-p

the hundreds of thousands of dollars. Thus, Bernanke (1983) used thousands of dollars as the

re

unit of measurement. Our analysis spans almost six decades, during which the banking sector

lP

changed dramatically. We therefore cannot use nominal dollar amounts.6 Instead, we inflation adjust the value of assets and then take the natural logarithm. Using the level—that is, using

na

thousands, millions, billions, or trillions of dollars—would place a disproportionately greater

ur

weight on bank stress experienced during the 2008 financial crisis. Second, we use failed assets instead of failed deposits as our baseline measure of bank stress, because the United States

Jo

established deposit insurance following the episode analyzed by Bernanke (1983). 7 To illustrate our bank stress measure, consider Figures 1 and 2. In Figure 1, we plot our bank stress measure for all failed banks over our full sample period. Figure 2 provides greater detail on the calculation of the series in the first quarter of 2017. As shown in Figure 2, the series took a value of nearly $500 million in that quarter. This value was caused by the failure of three

6

If we were to use, for example, trillions of dollars, the sheer volume of the nominal dollar amount in 2008 would mask the thousands of bank failures in prior decades. 7

To be sure, failed deposits and failed assets are highly correlated. The results are qualitatively identical. See Lorenc and Zhang (2018).

Journal Pre-proof banks, each with a different amount of assets. Therefore, at any point in time, the value of the series depends on the number of banks that have failed and the amount of assets at those banks. Figure 1: Assets of All Failed Banks, 1960-2017 $1,000,000,000,000 $100,000,000,000

$10,000,000,000 $1,000,000,000

of

Logarithmic Scale

$100,000,000 $10,000,000

ro

$1,000,000

-p

$100,000 $10,000

re

$1,000

$10

1975

1990

2005

na

$1 1960

lP

$100

Millions of Dollars

$500 $400

Jo

$600

ur

Figure 2: Assets of All Failed Banks, First Quarter of 2017

$300 $200

$100 $0 Seaway Bank and Trust Company

Harvest Community Bank

Proficio Bank

Journal Pre-proof Since we are interested in examining the impact of financial stress at banks of varying size on economic performance, we construct measures of large bank stress by including only failed banks with total assets above a percentile of the bank size distribution. To illustrate, Figure 3 depicts the large bank stress measure using the 98 th percentile of the bank size distribution as the cutoff. That is, this stress measure includes the assets of failed banks only when the bank that failed was one of the largest 2 percent of banks in that quarter. For example, during the financial

of

crisis—in the third quarter of 2008—the cutoff for the 98th percentile of the bank size

ro

distribution was $4.7 billion in total assets. Of the nine bank failures that occurred in that quarter,

-p

two were above the 98th percentile threshold: Washington Mutual Bank and IndyMac Bank

re

F.S.B. The assets of these banks are included in Figure 3. The assets of the seven other, smaller failed banks are included in the aggregate measure of bank stress shown in Figure 1 but not in

Jo

ur

na

lP

the measure of large bank stress shown in Figure 3. 8

8

Because we use Call Reports data for our analysis, the size percentiles described in this article correspond to the distribution of insured depository institutions (banks). If we want to map the results to the distribution of bank holding companies (BHCs), we must “scale-up” the bank size numbers. To derive the scale-up factors, we analyzed data from the largest BHCs in the first quarter of 2018, specifically the ratio of their insured depository institutions to their total consolidated assets. These ratios range from 0.27 to 0.40. Therefore, the 98th percentile corresponds to a range of $31 billion to $45 billion in BHC size.

Journal Pre-proof Figure 3: Assets of Failed Banks above the 98 th Percentile Size Threshold, 1960-2017 $1,000,000,000,000

$100,000,000,000 $10,000,000,000

$100,000,000

$10,000,000

of

$1,000,000 $100,000 $10,000

ro

Logarithmic Scale

$1,000,000,000

-p

$1,000 $100

$1 1960

1980

1990

2000

2010

lP

1970

re

$10

na

We use two quarterly variables to proxy economic performance in our baseline model— real GDP growth and the change in the civilian unemployment rate. The GDP series is already at

ur

a quarterly frequency. We convert the monthly unemployment rate series to a quarterly

Jo

frequency by taking the unemployment rate observed in the last month of a particular quarter. Figures 4 and 5 show the series from 1960 onward.

Journal Pre-proof Figure 4: Quarterly Real Gross Domestic Product Growth 4 3

2

0

of

Percent

1

ro

-1

-p

-2

re

-3

lP

-4 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

Percentage Point

1.0 0.5

ur

1.5

Jo

2.0

na

Figure 5: Change in Quarterly Unemployment Rate

0.0 -0.5

-1.0 -1.5 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

Journal Pre-proof

IV.

Methodology Our baseline methodology assumes that bank stress affects economic performance within

the same quarter (i.e., contemporaneously). The baseline regression specifications for GDP and the unemployment rate are, respectively: (

)

(

)

(

)

(

)

(

)

(

)

)

ro

(

)

of

(

-p

The quarter in which the contemporaneous effect takes place is denoted by subscript t and

re

the specific percentile of the bank size distribution used to calculate the stress measure is denoted

lP

by subscript p. Note that the complement to the set of banks above percentile p is denoted by 1-p. We update the relevant total asset distribution and percentiles for a given value of p each quarter.

na

The subscripts g and u correspond to regressions using GDP and the unemployment rate, in each regression to account for well-

ur

respectively. We include an AR(1) persistence term

Jo

documented persistence in real economic activity. captures the total assets of failed banks above a certain percentile p of the size

distribution during quarter t. Similarly,

corresponds to the total assets of failed banks

below a certain percentile p of the size distribution during quarter t. We include the stress measure for the 1-p set of banks in order to control for the impact of stress at smaller banks. The first regression we run is at the 99.9 th percentile, and we continue running the regressions in intervals of 0.05 for the top 1 percent of the distribution (i.e., 99.85 th percentile, 99.80th percentile, 99.75th percentile, and so on). After each regression, we record the estimated

Journal Pre-proof values of ,

,

, and

. The parameter of interest is

, which captures the impact of stress at

banks above size percentile p (large banks) on real GDP growth or the change in the unemployment rate. The parameter

, on the other hand, captures the impact of stress at banks

below size percentile p (small banks). We therefore allow the impact of financial stress to differ

Jo

ur

na

lP

re

-p

ro

of

across large and small banks.

Journal Pre-proof

V.

Baseline Results Table 1 presents the estimated parameters of interest at select percentiles within the top 1

percent

of the

bank

size

distribution.

Each

estimated

coefficient represents the

contemporaneous change in quarterly real GDP growth associated with a 1 percent increase in assets among failed banks above a particular percentile. Each estimated

coefficient represents

of

the contemporaneous change in quarterly real GDP growth associated with a 1 percent increase in assets among failed banks below a particular percentile. Therefore,

and

represent the

ro

impact on GDP of stress at large and small banks, respectively.

at the 99.8th percentile is -0.064, which means that a 1

-p

For example, the estimated

re

percent increase in assets among failed banks above the 99.8th percentile is associated with a

lP

contemporaneous 0.064 percentage point drop in quarterly real GDP growth. This result is significant at the 1 percent level.9 The estimated

at the 99.8th percentile is -0.006, which

na

means that a 1 percent increase in assets among failed banks below that percentile is associated

ur

with a contemporaneous 0.006 percentage point decline in quarterly real GDP growth. This

Jo

result, however, is not statistically significant. These empirical results show that financial stress at banks in the top 0.7 percent of the bank size distribution yields a statistically significant and negative impact on the real economy. The estimated

coefficients are significant through the 99.3rd percentile and then lose

significance. Notably, none of the estimated

coefficients are statistically significant, which

supports the claim that it is indeed stress at the large banks affecting the real economy.

9

Our baseline results are calculated using robust standard errors.

Journal Pre-proof Table 1: Baseline GDP Results Estimated

$318

-0.049** (0.025)

-0.008 (0.010)

0.127

99.8

$200

-0.064*** (0.023)

-0.006 (0.010)

0.155

99.7

$138

-0.051*** (0.019)

-0.004 (0.010)

0.145

99.6

$120

-0.039** (0.017)

-0.004 (0.010)

0.132

99.5

$93

-0.033** (0.016)

-0.007 (0.010)

0.130

99.4

$74

-0.025* (0.014)

-0.007 (0.010)

0.120

99.3

$53

-0.025* (0.014)

-0.007 (0.010)

0.121

99.2

$44

-0.018 (0.013)

-0.007 (0.010)

0.113

$35

-0.018 (0.013)

-0.007 (0.010)

0.114

$32

-0.016 (0.012)

-0.007 (0.010)

0.111

99.9

-p re

lP

ur

99.0

na

99.1

ro

Corresponding Asset Size

of

Estimated

Percentile

Jo

Notes: The asterisks placed next to each point estimate represent the corresponding level of statistical significance. In particular, * corresponds to significance at the 10 percent level, ** corresponds to significance at the 5 percent level, and *** corresponds to significance at the 1 percent level. The corresponding asset size is as of the fourth quarter of 2017. Figure 6 plots the estimated

coefficients and corresponding confidence intervals for

the top 1 percent of the bank size distribution. As shown in the chart, the estimated impact on GDP growth falls as the threshold defining large banks declines, though the point estimate remains negative at all thresholds. At the point where the confidence interval first includes zero, and the estimate is no longer statistically significant, the point estimate is -0.018. Although these

Journal Pre-proof point estimates are small, keep in mind that the increase in failed assets during a crisis is substantially higher than one percent. Figure 6: Impact of Failed Assets on Quarterly GDP 0.02

0.00

of

-0.02

ro

-0.04

-p

-0.06

re

-0.08

lP

-0.10

99.9%

99.8%

na

-0.12

99.7%

99.6%

99.5%

99.4%

99.3%

99.2%

99.1%

99.0%

Jo

ur

Notes: This figure shows the estimated coefficients in the top 1 percent of the bank size distribution. The estimated coefficients are depicted by the solid line, and the dashed lines represent the 90th percent confidence interval around the point estimates. Each point on the solid line represents an estimated coefficient. The results are similar for the change in the unemployment rate. In this case, positive values of

represent a negative impact on the real economy. Table 2 presents the estimated

parameters of interest at select percentiles in the top 1 percent of the bank size distribution for our baseline unemployment regression. Similar to the above discussion, each estimated coefficient represents the contemporaneous change in the unemployment rate associated with a 1 percent increase in assets among failed banks above a particular percentile. Each estimated coefficient represents the contemporaneous change in the unemployment rate associated with a 1

Journal Pre-proof percent increase in assets among failed banks below a particular percentile. Therefore,

and

represent the impact on the unemployment rate of stress at large and small banks, respectively. at the 99.8th percentile is 0.033. This means that a 1 percent increase in

The estimated

assets among failed banks above the 99.8 th percentile is associated with a contemporaneous 0.033 percentage point increase in the unemployment rate. This result is significant at the 1 at the 99.8th percentile is -0.001, which is essentially zero and

percent level. The estimated

parameters are statistically

of

not statistically significant. In fact, none of the estimated

ro

significant.

coefficients are significant through the 99 th percentile, and the

-p

The estimated

re

estimated impact is the most severe for stress at the largest banks. Indeed, the estimated impact of stress at banks in the top 0.2 percent of the size distribution is more than three times as large

lP

as the estimated impact of stress at banks in the top 1 percent. Figure 7 plots the estimated

na

coefficients and corresponding confidence intervals for the top 1 percent of the bank size

ur

distribution.

Jo

Table 2: Baseline Unemployment Rate Results Percentile

Corresponding Asset Size

99.9

$318

0.021* (0.011)

0.002 (0.003)

0.289

99.8

$200

0.033*** (0.011)

-0.001 (0.003)

0.343

99.7

$138

0.024** (0.010)

-0.001 (0.003)

0.315

99.6

$120

0.017** (0.008)

-0.001 (0.003)

0.296

99.5

$93

0.015** (0.007)

-0.001 (0.003)

0.294

Estimated

Estimated

Journal Pre-proof

$74

0.012* (0.007)

-0.001 (0.004)

0.286

99.3

$53

0.014** (0.006)

-0.002 (0.006)

0.298

99.2

$44

0.012** (0.006)

-0.002 (0.006)

0.289

99.1

$35

0.012** (0.006)

-0.002 (0.006)

0.290

99.0

$32

0.010* (0.006)

-0.002 (0.006)

0.282

of

99.4

-p

ro

Notes: The asterisks placed next to each point estimate represent the corresponding level of statistical significance. In particular, * corresponds to significance at the 10 percent level, ** corresponds to significance at the 5 percent level, and *** corresponds to significance at the 1 percent level. The corresponding asset size is as of the fourth quarter of 2017.

re

Figure 7: Impact on Quarterly Unemployment Rate

lP

0.06

na

0.05

ur

0.04

Jo

0.03

0.02

0.01

0

99.9%

99.8%

99.7%

99.6%

99.5%

99.4%

99.3%

99.2%

99.1%

99.0%

Notes: This figure shows the estimated coefficients in the top 1 percent of the bank size distribution. The estimated coefficients are depicted by the solid line, and the dashed lines represent the 90th percent confidence interval around the point estimates. Each point on the solid line represents an estimated coefficient.

Journal Pre-proof Last, but not least, we present our baseline results in terms of a hypothetical deviation from trend growth. Performing this exercise requires us to calculate the predicted change in real GDP growth that results from an assumed level of bank stress, then comparing the predicted change in GDP with trend GDP. Showing the results in this way allows us to see that the failure of $

billion assets at a single large bank has greater costs for the real economy than a large

of

number of failures at smaller banks that results in the same $ billion in failed assets. This calculation requires plugging all necessary variables into the right hand side of the

ro

regression specification in order to compute the predicted economic performance on the left hand

̂ ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ( )

re

̂

-p

side. This is shown generally in the following equation for any percentile p: ̂ ̅̅̅̅̅̅̅̅ ( )

lP

Presenting the results in this way requires two additional assumptions. First, we must make an assumption regarding the value of the bank stress measure. As a demonstration, we

na

choose $200 billion, which we then convert into natural logarithm units. This is ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ( )

ur

Second, we make an assumption regarding the average value of lagged GDP growth, which we

0.756.10

Jo

designate ̅̅̅̅̅̅̅̅ ( ). We use the average GDP growth rate of the regression sample, which is

Using these inputs, our results suggest that the failure of a single bank—above the 99.8th size percentile and with an assumed $200 billion in assets—would result in approximately a 150 percent decline in quarterly real GDP growth, while failures of five banks—each above the 99.1th size percentile and with an assumed $40 billion in assets—would result in approximately a 33 percent decline in quarterly real GDP growth. While both scenarios assume $200 billion in 10

The 1-p version of the stress measure is not here because, in this hypothetical scenario, we are assuming that only banks above size threshold p fail. Thus, there are no failed banks in the 1-p set.

Journal Pre-proof total failed assets, the negative impact is greatest when larger banks fail. These results are qualitatively similar for a change in the unemployment rate.

VI.

Robustness Checks The purpose of this section is to subject our baseline results to a battery of tests to check

for robustness. We begin with a couple of tests for omitted variable bias. First, we conduct our

of

analysis at the regional level instead of the aggregate country level. Second, we include

ro

additional control variables into our baseline specification. Third, we examine the assumption of contemporaneous impact in our baseline model by analyzing monthly economic data. Fourth, we

-p

run a simple Granger test to see if bank stress is causing a decline in economic performance, as

re

opposed to a decline in economic performance causing bank stress. Fifth, we re-run our baseline

lP

model with a shorter time series—stopping at the end of 2006 to remove the recent financial

large banks.

na

crisis—because the severity of the recent financial crisis could overstate the impact of stress at

ur

In some of these robustness tests, the estimated point estimates have larger standard

Jo

errors than those in our baseline case; this is because we are dropping a nontrivial number of observations to conduct those robustness checks. Nevertheless, our main results survive—the estimated impact of bank stress on the real economy is negative, and the estimated impact is greater for stress at larger banks than at smaller banks.

A. Regional Analysis To address the possibility of omitted variable bias in our results, we examine our baseline specification at the regional level. To the extent an omitted variable might be correlated with both macroeconomic performance and bank failures, it is unlikely that an omitted variable would

Journal Pre-proof have the same effect on all regions of the country simultaneously. For example, it is plausible that a downturn in the national housing market could simultaneously decrease GDP and increase bank failures. However, from historical episodes like the New England housing bubble in the 1990s, we have seen that downturns in the national housing market are often driven by vulnerabilities in one or two particular regions. For this reason, finding significant results at a regional level would help alleviate the concern regarding omitted variable bias.

of

We run our baseline regression using the U.S. Bureau of Labor Statistics unemployment

ro

rate in each of the four Census Bureau-designated regions: the Northeast, Midwest, South, and

coefficients are significant through the 99 th percentile, which is similar

re

The estimated

-p

West.11 For illustration, we present the Census West results in Table 3 below.

lP

to the results in the baseline case presented earlier. The magnitude of the estimated coefficients is similar to, but smaller than, that of the baseline scenario. It is still true that the estimated impact

na

is the most severe for stress at the largest banks. In this region, the estimated impact of stress at

ur

banks in the top 0.2 percent of the size distribution is nearly twice as large as the estimated

Jo

impact of stress at banks in the top 1 percent. In addition, the estimated essentially zero. Figure 8 plots the estimated

coefficients also are

coefficients and corresponding confidence

intervals for the top 1 percent of the bank size distribution. These results are almost identical between the Census West region and the Census South region. In the Census Midwest region, the results are qualitatively identical but are not as statistically significant. In the Census Northeast region, there were no bank failures above the

11

We use the U.S. Bureau of Labor Statistics unemployment rate for Census Bureau -designated regions because a sufficiently long quarterly time series of GDP at a state or regional level is not available.

Journal Pre-proof 99.3rd percentile, though the estimated coefficients between that and the 99th percentile are highly significant, both economically and statistically.

Table 3: West Census Region Results Percentile

Estimated

Estimated

Baseline 0.021* (0.011)

0.025*** (0.008)

0.000 (0.002)

0.716

99.8

0.033*** (0.011)

0.017** (0.008)

0.000 (0.002)

0.709

99.7

0.024** (0.010)

0.013* (0.008)

0.000 (0.002)

0.704

99.6

0.017** (0.008)

0.011 (0.007)

0.000 (0.002)

0.702

99.5

0.015** (0.007)

0.010* (0.005)

0.000 (0.002)

0.703

99.4

0.012* (0.007)

0.010* (0.005)

0.000 (0.002)

0.703

0.014** (0.006)

0.010* (0.005)

0.000 (0.002)

0.703

0.012** (0.006)

0.010* (0.005)

0.000 (0.002)

0.703

99.1

0.012** (0.006)

0.010* (0.005)

0.000 (0.002)

0.703

99.0

0.010* (0.006)

0.010** (0.004)

0.000 (0.002)

0.705

ro

-p

re

lP

na

Jo

99.2

ur

99.3

of

99.9

Notes: The asterisks placed next to each point estimate represent the corresponding level of statistical significance. In particular, * corresponds to significance at the 10 percent level, ** corresponds to significance at the 5 percent level, and *** corresponds to significance at the 1 percent level.

Journal Pre-proof

Figure 8: Impact of Failed Assets on Change in Unemployment Rate, West Census

0.040

of

0.035

ro

0.030

-p

0.025

re

0.020

lP

0.015 0.010

-0.005

ur

0.000

na

0.005

Jo

99.9% 99.8% 99.7% 99.6% 99.5% 99.4% 99.3% 99.2% 99.1% 99.0% Notes: This figure shows the estimated coefficients in the top 1 percent of the bank size distribution. The estimated coefficients are depicted by the solid line, and the dashed lines represent the 90th percent confidence interval around the point estimates. Each point on the solid line represents an estimated coefficient.

B. Additional Control Variables We next explicitly consider whether our baseline results are robust to the inclusion of additional control variables that also are linked with economic performance. Specifically, we include the percent change in median wages of full-time employed workers (inflation adjusted), the percent change in the S&P 500 index, the corporate BBB option-adjusted spread, the percent

Journal Pre-proof change in the Case-Shiller national home price index (seasonally adjusted), and the first difference in total new privately owned housing starts (seasonally adjusted). The GDP growth results are presented in Table 4, and the unemployment rate results are presented in Table 5. The main results hold for both the GDP and unemployment rate specifications—with the estimated impact remarkably similar to that of the baseline case—though they are slightly less statistically significant in some cases. This is particularly true in four of the five specifications—

of

wages, S&P, BBB, and housing starts. The house price regression, which covers the period 1987

ro

through 2017, appears to be an outlier. Notably, when examining the specifications that cover the

-p

full sample period of 1960 through 2017—S&P and housing starts—we see that the estimated

re

coefficients are very close to the results we obtained under the baseline specification. This gives

lP

us confidence that our results are robust to potential omitted variable bias.

Percentile

Size

99.9

$318

99.8

na

Table 4: GDP Growth Results

Wages

S&P

BBB

House Price

Housing Starts

-0.049** (0.025)

-0.042* (0.023)

-0.044* (0.024)

-0.046*** (0.014)

-0.055*** (0.019)

-0.048** (0.024)

$200

-0.064*** (0.023)

-0.059** (0.025)

-0.061** (0.024)

-0.046*** (0.014)

-0.054*** (0.019)

-0.063*** (0.023)

99.7

$138

-0.051*** (0.019)

-0.045** (0.021)

-0.049** (0.019)

-0.046*** (0.014)

-0.035** (0.016)

-0.050*** (0.019)

99.6

$120

-0.039** (0.017)

-0.032* (0.017)

-0.037** (0.016)

-0.046*** (0.014)

-0.021 (0.013)

-0.036** (0.016)

99.5

$93

-0.033** (0.016)

-0.032** (0.015)

-0.031** (0.015)

-0.046*** (0.014)

-0.020 (0.013)

-0.031** (0.015)

99.4

$74

-0.025* (0.014)

-0.023 (0.014)

-0.024* (0.014)

-0.030* (0.017)

-0.010 (0.012)

-0.023* (0.014)

99.3

$53

-0.025* (0.014)

-0.025* (0.014)

-0.023* (0.014)

-0.030* (0.017)

-0.010 (0.012)

-0.023* (0.013)

Jo

ur

Baseline

Journal Pre-proof

99.2

$44

-0.018 (0.013)

-0.016 (0.013)

-0.017 (0.013)

-0.030* (0.017)

-0.005 (0.011)

-0.016 (0.012)

99.1

$35

-0.018 (0.013)

-0.016 (0.013)

-0.018 (0.012)

-0.030* (0.017)

-0.006 (0.011)

-0.017 (0.012)

99.0

$32

-0.016 (0.012)

-0.014 (0.012)

-0.016 (0.012)

-0.021 (0.019)

-0.006 (0.010)

-0.015 (0.012)

Year

2017

1960-2017 1979-2017 1960-2017 1997-2017 1987-2017 1960-2017

ro

of

Notes: The asterisks placed next to each point estimate represent the corresponding level of statistical significance. In particular, * corresponds to significance at the 10 percent level, ** corresponds to significance at the 5 percent level, and *** corresponds to significance at the 1 percent level.

BBB

House Price

Housing Starts

0.020* (0.011)

0.022*** (0.007)

0.028*** (0.009)

0.021* (0.011)

0.029** (0.012)

0.033*** (0.012)

0.022*** (0.007)

0.028*** (0.009)

0.033*** (0.011)

0.019* (0.011)

na

0.023** (0.010)

0.021*** (0.007)

0.015* (0.008)

0.023** (0.010)

0.013 (0.008)

0.017* (0.008)

0.021*** (0.007)

0.009 (0.006)

0.016* (0.009)

Jo

Size

99.9

0.015** (0.007)

0.013* (0.007)

0.014** (0.007)

0.021*** (0.007)

0.009 (0.006)

0.014** (0.007)

$74

0.012* (0.007)

0.011 (0.007)

0.012* (0.007)

0.016** (0.007)

0.007 (0.005)

0.012* (0.007)

99.3

$53

0.014** (0.006)

0.014** (0.007)

0.014** (0.006)

0.016** (0.007)

0.007 (0.005)

0.014** (0.006)

99.2

$44

0.012** (0.006)

0.011* (0.006)

0.012** (0.006)

0.016** (0.007)

0.007 (0.005)

0.011* (0.006)

99.1

$35

0.012** (0.006)

0.011* (0.006)

0.012** (0.006)

0.016** (0.007)

0.007 (0.005)

0.011** (0.006)

99.0

$32

0.010* (0.006)

0.009 (0.006)

0.010* (0.006)

0.018** (0.009)

0.008 (0.005)

0.009* (0.006)

Wages

$318

0.021* (0.011)

0.021* (0.011)

99.8

$200

0.033*** (0.011)

99.7

$138

0.024** (0.010)

99.6

$120

0.017** (0.008)

99.5

$93

99.4

S&P

ur

lP

Baseline

re

Percentile

-p

Table 5: Unemployment Rate Results

Journal Pre-proof

Year

2017

1979-2017 1979-2017 1960-2017 1997-2017 1987-2017 1960-2017

Notes: The asterisks placed next to each point estimate represent the corresponding level of statistical significance. In particular, * corresponds to significance at the 10 percent level, ** corresponds to significance at the 5 percent level, and *** corresponds to significance at the 1 percent level.

C. Monthly Frequency In our baseline model, we investigate the contemporaneous impact of bank stress on quarterly real GDP growth and the quarterly change in the unemployment rate. To check the

of

robustness of our contemporaneous modeling assumption—that is, to test whether the impact

ro

does indeed materialize within the same quarter—we re-run the GDP and unemployment

-p

regressions using monthly economic performance indicators. Note that the unemployment rate

re

series is available at a monthly frequency, but the GDP series is not. We therefore replace real GDP growth with the Chicago Fed National Activity Index (CFNAI), which closely tracks real

lP

GDP growth.

Jo

ur

na

Figure 9: Monthly Chicago Fed National Activity Index (CFNAI)

Journal Pre-proof 4 3

2 1

Index

0

-1

of

-2 -3

ro

-4

-p

-5

lP

re

-6 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

The regression specifications are modified to include more lags, as there are three months

na

in a quarter. Otherwise, the two specifications are unchanged from the baseline model discussed

(

(

Jo

∑

ur

previously:

)

)

∑

∑

∑

(

(

(

)

)

∑

(

∑

)

)

We begin our estimation at the 99.9th percentile and progress down in intervals of 0.05. The coefficient of interest is still

, except we now have three of them. The first one

Journal Pre-proof corresponds to the contemporaneous effect;

captures the impact with a one-month lag; and

captures the impact with a two-month lag. Figure 10 presents the estimated

coefficients of the CFNAI regressions in the solid

blue line. The magnitude of the monthly point estimates is smaller than that of the baseline GDP results, though it is still true that the largest point estimate is more than three times the point estimate at the top 1 percent threshold. In addition, the shape of the solid blue line is slightly

of

different. The solid blue line begins to flatten out—and lose its statistical significance—below

lP

re

-p

ro

the top 0.5 percent, whereas the baseline green line flattens out later.

na

Figure 10: Impact on Monthly CFNAI

ur

0.02

Jo

0.00 -0.02 -0.04 -0.06 -0.08

-0.10 -0.12 99.9%

99.8%

99.7%

99.6%

99.5%

Robustness Check

99.4%

99.3% Baseline

99.2%

99.1%

99.0%

Journal Pre-proof Notes: This figure shows the estimated coefficients for the monthly CFNAI regressions in the top 1 percent of the bank size distribution. The estimated coefficients are depicted by the solid blue line with circular markers, and the dashed lines represent the 90 th percent confidence interval around the point estimates. The point estimates from the baseline quarterly GDP regressions are depicted by the solid green line with triangular markers. Each point on the two solid lines represents an estimated coefficient. The results using the unemployment rate are similar, but with a couple of notable differences. First, the contemporaneous impact and

. Second, as shown in Figure 11, while the monthly point estimates are smaller

of

the lags,

is not the main driver of the results, but rather

in magnitude than the baseline ones, the monthly point estimates are statistically significant

lP

re

-p

ro

through the 99th percentile.

Figure 11: Impact on Monthly Unemployment Rate

na

0.040

ur

0.035

Jo

0.030

0.025 0.020

0.015 0.010 0.005

0.000 99.9%

99.8%

99.7%

99.6%

99.5%

Robustness Check

99.4%

99.3% Baseline

99.2%

99.1%

99.0%

Journal Pre-proof Notes: This figure shows the estimated coefficients for the monthly unemployment rate regressions in the top 1 percent of the bank size distribution. The estimated coefficients are depicted by the solid blue line with circular markers, and the dashed lines represent the 95 th percent confidence interval around the point estimates. The point estimates from the baseline quarterly unemployment rate results are depicted by the solid green line with triangular markers. Each point on the two solid lines represents an estimated coefficient.

D. Granger Causality Another interpretive issue in this type of analysis is reverse causality. How do we know

of

bank stress is causing a decline in economic performance, as opposed to a decline in economic

ro

performance causing bank stress? We test the robustness of our baseline results by following the time series methods developed in Granger (1969). Specifically, we first run a vector auto-

-p

regression (“VAR”) using proxies of bank stress and economic performance, with a one-quarter

re

lag. We then compare the p-values of the coefficients on the lagged terms. We are essentially

lP

asking which lagged series is a better predictor of the other series—for instance, are lagged failed assets better at predicting GDP growth, or is lagged GDP growth better at predicting failed

na

assets? The results are, on balance, favorable to our model.

ur

Consider columns (1) and (2) of Table 6. Up until the top 0.5 percent, the Granger tests

Jo

suggest that lagged assets at these large failed banks are better at predicting GDP growth than vice versa, because a smaller p-value corresponds to greater significance. Once the size threshold reaches the top 0.5 percent, however, the story changes: lagged values of GDP growth become a stronger predictor of assets at failed banks. That implies reverse causality for stress at banks below the top 0.5 percent size threshold. However, the issue does not appear as starkly when analyzing changes in the unemployment rate. In particular, Columns (3) and (4) show that lagged assets are uniformly better predictors of changes in the unemployment rate than vice versa. 12

12

To be sure, the p-value in Column (4) falls under 0.1 after the top 0.2 percent, which suggests that the unemployment rate and failed assets have a statistically significant impact on each other. While comparing p -values

Journal Pre-proof Thus, while these results provide some evidence that poor economic performance causes bank stress, they provide as much, if not more, evidence that bank stress causes poor economic performance. Table 6: Comparing p-Values of Granger Tests

re

-p

ro

of

Lagged Failed Lagged GDP Lagged Failed Lagged UR Assets Predict Predicts Failed Assets Predicts Failed Percentile GDP Assets Predict UR Assets (1) (2) (3) (4) 99.9 0.044 < 0.668 0.001 < 0.541 99.8 0.011 < 0.281 0.002 < 0.379 99.7 0.048 < 0.123 0.004 < 0.056 99.6 0.057 < 0.137 0.008 < 0.083 99.5 0.131 0.012 0.002 < 0.019 99.0 0.066 0.004 0.002 < 0.005 Notes: This table contains the p-values of Granger tests at select percentiles. A smaller p-value corresponds to greater statistical significance.

lP

E. Sample Period: Financial Crisis

na

The recent financial crisis was a period of extreme bank stress compared to most periods in our full sample period. In addition, the crisis resulted in a large amount of assistance and

ur

acquisitions, such as the Troubled Asset Relief Program (TARP), that were the result of

Jo

government initiatives. To assess the impact of the financial crisis on our baseline results, we run our baseline regressions from 1960 through 2006, removing the financial crisis entirely. Figure 12 compares the results of our GDP regression for the full and non-crisis samples. The shape of the estimated series is similar and the point estimates are uniformly negative, which suggest that the impact is still more severe for stress at the largest banks. The magnitude of the estimated impact is smaller in this no-crisis sample, and the standard errors are larger (as shown by the wider confidence intervals), especially for the point estimates at the top of the bank size of 0.002 and 0.005 (99th percentile) does not tell us much, we can learn something from comparing p-values of 0.001 and 0.541 (99.9th percentile) or from comparing p-values of 0.002 and 0.019 (99.5th percentile).

Journal Pre-proof distribution.13 The results are qualitatively similar for the unemployment rate—the shape of the estimated series is the same, the point estimates are uniformly positive, and the standard errors

re

-p

ro

of

are larger.

lP

Figure 12: Shorter Sample Period and Quarterly GDP 0.06

na

0.04 0.02

ur

0.00 -0.02

Jo

-0.04 -0.06 -0.08 -0.10

-0.12 -0.14 99.9%

99.8%

99.7%

99.6%

99.5%

Robustness Check

13

99.4%

99.3%

99.2%

Baseline

Figure 12 omits the point estimate at the 99.9th percentile because of small sample size.

99.1%

99.0%

Journal Pre-proof

re

-p

ro

of

Notes: This figure shows the estimated coefficients in the top 1 percent of the bank size distribution. The sample period ends in the fourth quarter of 2006, before the financial crisis. The estimated coefficients are depicted by the solid blue line with circular markers, and the dashed lines represent the 90th percent confidence interval around the point estimates. The point estimates from the baseline results are depicted by the solid green line with triangular markers. Each point on the two solid lines represents an estimated coefficient.

lP

VII. Conclusion

na

This article builds upon the important and early work of Bernanke (1983), which uncovers a link between aggregate bank stress and macroeconomic stress. We document

ur

evidence of a link between stress in the banking sector and stress in the economy that varies

Jo

along the bank-size spectrum. Importantly, we find that stress among larger banks has a greater negative consequence on the economy than does stress at smaller banks. To this end, our analysis can broadly inform the size threshold at which enhanced standards take effect. There are, however, trade-offs involved with choosing the appropriate level. For instance, one might consider the level at which the impact of bank stress on the economy becomes statistically insignificant. Alternatively, one could choose to calibrate the threshold to a level at which the economic impact of stress is greater than zero and economically significant rather than statistically significant. Doing so would result in a higher threshold, and

Journal Pre-proof the precise level of economic significance would represent policymakers’ views on the trade-off between the costs of enhanced regulation and the benefits associated with more resilient banks. Our results survive multiple robustness tests, but are nonetheless subject to caveats. First, they are dependent on the particular measures of economic performance and bank stress considered, as well as the sample period employed. Second, an assumption implicit in this approach is that the cost of complying with enhanced standards is the same for banks of varying

of

size. If costs of compliance vary with bank size, then these costs need to be considered in the

ro

analysis.

-p

In future research, we would like to explore an alternative measure of bank stress. All

re

analyses to date have used assets (or deposits) of failed banks as a proxy for bank stress. We

lP

would like to repeat our analysis using the assets of banks with low capital ratios, which would provide a less extreme signal of bank stress than failing or receiving assistance from the FDIC.

na

Assets of thinly capitalized banks can be adjusted to consider varying degrees of stress or

ur

“thinness,” whereas failed assets are binary—a bank has either failed or it has not. Additionally, as mentioned in the introduction, size is not the sole predictor of what impact a bank would have

Jo

on the economy when under stress. The complexity of a bank’s operations also factors into the equation. Future research could explore various ways to measure bank complexity so that size and complexity are analyzed together. In sum, our results are broadly informative about how the stringency of regulatory standards should vary with bank size, and support the idea that the largest banks should be subject to more stringent macroprudential requirements while smaller banks should be subject to successively less stringent macroprudential requirements. We emphasize macroprudential in this context because we argue that the failure of the largest banks have a greater impact on the

Journal Pre-proof macroeconomy than the failure of small banks. We do not claim that small bank failures are not costly. Indeed, our analysis using aggregate macroeconomic variables does not shed light on the

-p

ro

of

impact of small bank lending along other dimensions.

re

References

lP

Acharya, Viral V. and Sascha Steffen. 2015. “The ‘Greatest’ Carry Trade Ever? Understanding

na

Eurozone Bank Risks.” Journal of Financial Economics Vol. 115, pp. 215-236. Adrian, Tobias and Markus K. Brunnermeier, 2016. “CoVaR.” American Economic Review Vol.

ur

106, pp. 1705-1741.

Jo

Allen, Linda, Tura G. Bali, and Yi Tang. 2012. “Does Systemic Risk in the Financial Sector Predict Future Economic Downturns?” The Review of Financial Studies Vol. 25, No. 10, pp. 3000-3036. Ashcraft, Adam B. 2005. “Are Banks Really Special? New Evidence from the FDIC-Induced Failure of Healthy Banks.” The American Economic Review Vol. 95, No. 5, pp. 17121730. Bank Policy Institute. 2019. Response to Proposed Changes to Applicability Thresholds for Regulatory Capital and Liquidity Requirements.

Journal Pre-proof Basel Committee on Banking Supervision. 2014. “The G-SIB Assessment Methodology – Score Calculation.” Bernanke, Ben S. 1983. “Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression.” The American Economic Review Vol. 73, No. 3, pp. 257-276. Bernanke, Ben S. 2015. The Courage to Act: A Memoir of a Crisis and Its Aftermath. Norton.

Outcomes: Theory and

Cross-Country Evidence of Granularity.”

ro

Macroeconomic

of

Bremus, Franziska, Claudia Buch, Katheryn Russ, and Monika Schnitzer. 2018. “Big Banks and

-p

Journal of Money, Credit and Banking Vol. 50, pp. 1785-1825.

re

Calomiris, Charles W. and Joseph R. Mason. 2003. “Consequences of Bank Distress during the

lP

Great Depression.” The American Economic Review Vol. 93, No. 3, pp. 937-947. Chodorow-Reich, Gabriel. 2014. “The Employment Effects of Credit Market Disruptions: Firm-

na

Level Evidence from the 2008-9 Financial Crisis.” The Quarterly Journal of Economics

ur

Vol. 129, No. 1, pp. 1-59.

Jo

Dávila, Eduardo and Ansgar Walther. 2017. “Does Size Matter? Bailouts with Large and Small Banks” NBER Working Paper No. 24132. Federal Reserve. 2016. “Calibrating the Single-Counterparty Credit Limit between Systemically Important Financial Institutions.” White Paper. Geithner, Timothy F. 2015. Stress Test: Reflections on Financial Crises. Broadway Books. Gorton, Gary B. 2010. Slapped by the Invisible Hand: The Panic of 2007. Oxford University Press.

Journal Pre-proof Granger, C. W. J. 1969. “Investigating Causal Relationships by Econometric Models and CrossSpectral Methods.” Econometrica Vol. 37, No. 3, pp. 424-438. Haldane, Andrew G. and Robert M. May. 2011. “Systemic Risk in Banking Ecosystems.” Nature Vol. 469, pp. 351-355. Kashyap, Anil K. and Jeremy C. Stein. 2000. “What Do a Million Observations on Banks Say

of

about the Transmission of Monetary Policy?” The American Economic Review Vol. 90,

ro

No. 3, pp. 407-428.

-p

Kupiec, Paul H. and Carlos D. Ramirez. 2013. “Bank Failures and the Cost of Systemic Risk:

re

Evidence from 1900 to 1930.” Journal of Financial Intermediation Vol. 22, pp. 285-307. Lorenc, Amy L. and Jeffery Y. Zhang. 2018. “The Differential Impact of Bank Size on Systemic

lP

Risk.” FEDS Working Paper 2018-066.

ur

Important Banks.”

na

Office of Financial Research. 2017. “Size Alone Is Not Sufficient to Identify Systemically

Jo

Quarles, Randal K. 2018. “Getting It Right: Factors for Tailoring Supervision and Regulation of Large Financial Institutions.” Speech at American Bankers Association Summer Leadership Meeting, Salt Lake City, Utah. Reinhart, Carmen M. and Kenneth S. Rogoff. 2009. This Time Is Different: Eight Centuries of Financial Folly. Princeton University Press. Scott, Hal S. 2016. Connectedness and Contagion: Protecting the Financial System from Panics. MIT Press.

Journal Pre-proof Tarashev, Nikola, Claudio Borio, and Kostas Tsatsaronis. 2010. “Attributing Systemic Risk to Individual Institutions: Methodology and Policy Applications.” BIS Working Paper No. 308. United States Congress. 2018. Economic Growth, Regulatory Relief, and Consumer Protection

Jo

ur

na

lP

re

-p

ro

of

Act.

Journal Pre-proof Highlights: 

The experiences of past financial crises have taught us that stress within the aggregate banking sector can lead to stress in the real economy.



This study shows that the failure of larger banks has a disproportionate impact on the real economy than the failure of smaller banks.

of

The largest banks should be subject to the most stringent requirements while smaller

ur

na

lP

re

-p

ro

banks should be subject to successively less stringent requirements.

Jo

