Levine Lin Xie 2021 06 Local Financial Structure Economic Resilience Worldbank
Summary
In this June 2021 paper, Ross Levine, Chen Lin and Wensi Xie use high-frequency U.S. county-level data on employment, small business revenue and COVID-19 cases to ask whether the structure of the financial system affects resilience to shocks. They find that employment, especially of low-income workers, and the revenues of small firms fall by less in response to local COVID-19 cases in counties with a larger proportion of small banks. They also find that small banks increased lending to small businesses more than large banks did in response to the pandemic, which they read as small banks funding small firms after an adverse shock.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Levine Lin Xie 2021 06 Local Financial Structure Economic Resilience Worldbank.”
Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/levine-lin-xie-2021-06-local-financial-structure-economic-resilience-w-2ca841eccb/
, SHA-256 fe86b49bb74edc13fd0737443ab6a53e797e6e4ca97cefd488b8f1be1b7db0ca.Full text
Local Financial Structure and Economic Resilience
Ross Levine, Chen Lin, and Wensi Xie†
June 2021
Abstract
Does the structure of the financial system influence the economy’s resilience to adverse shocks?
Using high-frequency, U.S. county-level data on employment, small business revenue, and
COVID-19 cases, we discover that employment, especially the employment of low-income
workers, and the revenues of small firms fall by less in response to local COVID-19 cases in
counties with a larger proportion of small banks. Furthermore, small banks increase lending to
small businesses more than large banks in response to the pandemic. Using QWI data,
employment of small firms falls less by less in response to COVID-19 cases in counties with a
larger proportion of small banks. Evidence suggests that small banks provide funding to small
firms following an adverse shock, with positive repercussions on employment.
JEL Classification Codes: G21, E24, E32, J21, H12
Keywords: Financial Structure, Employment, Inequality, Business Fluctuations, Crisis
Management; COVID-19
†
Corresponding author, Levine: Haas School of Business, University of California, Berkeley. Email:
rosslevine@berkeley.edu.
Lin: Faculty of Business and Economics, University of Hong Kong. Email: chenlin1@hku.hk.
Xie: CUHK Business School, Chinese University of Hong Kong. Email: wensixie@cuhk.edu.hk.
1
1. Introduction
A defining feature of financial systems is the extent to which they provide funding to
firms to make positive net present value (NPV) investments (Myers and Majluf 1984; Fazzari,
Hubbard, and Petersen 1988; Levine 1997). The ability of financial systems to fund such
investments and provide immediate liquidity is especially important when adverse shocks
temporarily reduce the profitability of firms (Bernanke, Gertler, and Gilchrist 1999; Acharya and
Steffen 2020). Without access to finance, more firms will reduce employment, magnifying the
impact and transmission of the shock.
Small firms and banks may play a key role in this nexus between financial constraints,
employment, and the propagation of shocks. Besides employing almost half of all private-sector
workers and more than half of low-income workers (Acs and Nichols 2007; Small Business
Administration 2020), research suggests that small firms are often more opaque and hence
subject to tighter financing constraints than larger firms (Petersen and Rajan 1994; Berger and
Udell 1998; Beck, Demirgüç-Kunt, and Maksimovic 2008). As a result, it might be especially
difficult for small firms to obtain countercyclical financing, increasing the severity of recessions.
Research further suggests that small banks have a comparative advantage over large banks in
augmenting the use of “hard” information that is contained in financial and accounting
statements with “soft” information that is acquired through the establishment of long-term
relationships with businesses, where this soft information is often crucial for evaluating and
financing small businesses (Berger and Udell 1995, 2002; Canales and Nanda 2012; Berger et al
2005, 2020; Berger, Goulding, and Rice 2014; Berger, Bouwman, and Kim 2017). These
findings raise a key question: Do small banks foster economic resilience through countercyclical
lending to small firms that limits reductions in employment, especially among low-income
workers?
In this paper, we provide new evidence on whether and how the proportion of small
banks operating in a local economy influences the economy’s resilience to adverse shocks. As an
adverse shock, we use cross-county differences in weekly COVID-19 infection rates across the
2
United States. To gauge economic resilience, we focus on how total employment and the
employment of low-, medium, and high-income workers in each county respond to local
COVID-19 infection rates. We focus on employment both because of the direct welfare
implications of employment reductions and the role of employment reductions in magnifying and
propagating adverse shocks. We focus on how the proportion of small banks operating in a
county differentially shapes employment reductions across low-, medium-, and high-income
workers because understanding the connections between financial structure and income
distributional effects of adverse shocks is valuable for formulating sound policies. Thus, we use
the pre-pandemic period as a benchmark and evaluate how local employment responds to the
pandemic as a function of the proportion of small banks operating in the county. Besides
employment, we examine an additional implication of the view that small banks have a
comparative advantage in lending to small local businesses during the COVID-19 crisis: small
local firms will be better positioned to maintain their operations and revenues in response to the
pandemic in communities with a higher proportion of small banks. Finally, we examine a direct
mechanism linking small banks to local employment and small firm revenues. We assess how
small bank lending to small firms responds to local COVID-19 cases.
In particular, we address three questions concerning the relation between small banks and
the resilience of local economies to COVID-19. First, in response to COVID-19, does
employment—especially the employment of low-income workers—fall by less in counties with a
higher proportion of small banks? To the extent that small banks have a comparative advantage
in ameliorating information asymmetries and financing small firms following adverse shocks, the
pandemic should have a more muted impact on employment in counties with a higher proportion
of small banks. Furthermore, since small firms disproportionately employ low-income workers,
we assess whether the employment of low-income workers falls by less in counties with a larger
share of small banks. That is, we test whether small banks shape the income distributional effects
of an economic crisis. Although past work explores the connection between overall financial
development and the amplitude of business cycle fluctuations (Bernanke, Gertler, and Gilchrist
3
1999; Smets and Wouters 2007; Gilchrist and Zakrajšek 2012; Brunnermeier and Sannikov
2014), we examine how small banks influence changes in both aggregate employment and the
employment of low-, medium-, and high-income workers following the economic disruptions
triggered by COVID-19. Furthermore, as emphasized above, past work suggests that small
banks have a comparative advantage in forming lending relations with and funding small firms
and research also indicates that these lending relationships may be especially important
following adverse shocks because economic downturns often intensify informational frictions
(Bolton, Freixas, and Gambacorta 2016; Beck et al. 2018). We contribute by examining the
degree to which small banks cushion the impact of an adverse shock on total employment and
the distribution of employment across low-, middle-, and high-income workers.
Second, do the revenues of small firms fall by less in response to COVID-19 in counties
with higher proportions of small banks? If small banks have a comparative advantage in
financing small firms during economic downturns, then small firms should perform better in
economies with a higher proportion of small banks than in counties dominated by larger banks.
While past work examines how securities markets, banks, and access to trade credit shape the
financing and performance of large, publicly-traded firms (Demirgüç-Kunt and Maksimovic
1998, 2002; Levine, Lin, and Xie 2016, 2018), we assess the extent to which small banks ease
the impact of economic crises on the performance of small firms.
Third, how do small and large banks change their lending to small firms in response to
the pandemic? As emphasized above, existing research examines the comparative advantages of
small banks in funding small firms and Berger et al (2020) explore the connection between
lending relationships and loan contract terms during the COVID-19 pandemic. We contribute to
this research by examining how a bank’s lending to small firms changes in response to the
bank’s exposure to COVID-19 while differentiating banks by size. In this way, we assess the
differential degrees to which smaller and larger banks provide financing to small local firms
following an adverse shock.
4
To assess the relation between small banks and how employment and the distribution of
employment respond to COVID-19 infection rates, we begin by using ordinary least squares
(OLS) and then employ an instrumental variable (IV) strategy. In these analyses, we use weekly
observations on U.S. counties from February through June 2020. We examine total employment,
unemployment insurance claims, and the employment of low-, middle-, and high-income
workers. We use two measures of the proportion of small banks in a county: the share of total
branches in a county that are branches of small banks and the share of total deposits in a county
held in the branches of small banks. A bank is defined as “small” if its total assets are below $1
billion at the close of 2019. To measure the county’s exposure to the pandemic, we use ln(cases
per capita), which equals the logarithm of one plus the cumulative number of confirmed
COVID-19 cases per 100,000 people.
In these county-week analyses, we regress the five employment measures on ln(cases per
capita), the proportion of small banks, and the interaction between ln(cases per capita) and the
proportion of small banks. In this way, we assess how the relationship between employment and
the adverse shock of COVID-19 varies across local economies with different proportions of
small banks. To condition out all time-invariant county traits and all time-varying state
influences, including state-specific social distancing orders, we include county and state-week
fixed effects. Furthermore, to control for the possibility that other county-level traits besides the
share of small banks influence the relation between employment and the intensity of infection
rates in the county, we control for the interaction between ln(cases per capita) and (a) the total
number of bank branches per capita, (b) income per capita, (c) the percentage of people over the
age of 65, (d) population density, and (e) the pre-pandemic unemployment rate. In addition, we
show that the results are robust to controlling for national policies that may influence lending to
small firms and local employment, e.g., the CARES Act’s Paycheck Protection Program (PPP). 1
Specifically, we condition on the cumulative amount of PPP loans per capita in each county over
1
Compared with large corporations that have access to alternative sources of external finance (e.g., Acharya and
Steffen 2020; Ellul, Erel and Rajan 2020), small firms rely more heavily on bank lending (Lin 2020).
5
time. Thus, in these initial analyses, we use OLS and control for a broad array of factors to limit
concerns that some omitted county-specific factor besides the share of small banks shapes the
local economy’s resilience to COVID-19 cases.
We discover that employment—especially the employment of low-income workers—
falls by less in counties with a greater share of small banks in response to COVID-19 cases. In
particular, we find that aggregate employment falls by less and unemployment insurance claims
increase by less in response to COVID-19 among counties with a greater share of small banks.
We also uncover strong results regarding the income distributional implications of small banks.
In response to COVID-19 infections, we find much smaller declines in the employment of low-
income workers—but not smaller declines in the employment of middle- or high-income
workers—in counties with a higher proportion of small banks. This finding is consistent with the
view that small banks shape the income distributional effects of the economic disruptions
triggered by COVID-19.
Next, we use the same methodology to examine whether—and show that—the revenues
of small firms fall by less in response to COVID-19 in counties with a larger proportion of small
banks. The dependent variable is Small business revenue, which equals the percent change in
seasonally adjusted net revenues for small businesses in each county relative to its pre-pandemic
level. The intensity of a county’s COVID-19 infection rates, ln(cases per capita), tends to reduce
small business revenues, but Small business revenue falls by less in response to the pandemic in
counties with higher small bank shares. These findings are consistent with the view that small
banks have a comparative advantage in financing small firms during crises and this funding
allows those small businesses to operate more effectively, generating greater revenues, during
these periods of economic duress.
To enhance identification, we reassess the findings on employment and revenues using an
IV strategy. Despite the extensive controls used in OLS analyses, there could be omitted county-
specific features that are correlated with the local share of small banks, and these omitted
features, not the share of small banks, could influence cross-county changes in employment and
6
revenues in response to COVID-19 cases. Thus, we use an instrumental variable to identify the
impact of the share of small banks on cross-county responses to the pandemic. The instrument
equals the number of mergers by out-of-state banks that affect branch structure in a county
during the decade from 2010 through 2019, where both the acquiring and target banks have at
least one branch located in the county prior to the merger. Specifically, for a county c located in
state s, the merger involves two banks that are not headquartered in state s, where each bank has
at least one branch in county c. Such mergers will tend to give the consolidated bank a
competitive advantage, reducing the role of small banks in the county. This is what we find: the
merger instrument is negatively associated with the share of small banks. The identifying
assumption is that the merger decision of two banks headquartered in states different from
county c during the 2010-2019 period does not reflect local conditions that drive the county’s
resilience to the 2020 pandemic.
The IV results confirm all of the OLS findings: a larger share of small banks in a county
mitigates the adverse shock of COVID-19 on employment, especially the employment of low-
income workers, and on the revenues of small businesses within the county. In terms of
instrument validity, the data reject the null hypothesis that the excluded instruments do not
account for cross-county variations in the share of small banks.
Next, we use cross-sectional analyses to evaluate the connection between changes in a
bank’s lending to small firms and the exposure of its business clients to COVID-19. To measure
the exposure of a bank’s business clients to the economic duress triggered by COVID-19, we
build on research showing that banks tend to make loans to small businesses that are
geographically very close to their branches (e.g., Nguyen 2019). In particular, we first measure a
county’s exposure to COVID-19, ln(cases per capita), as the logarithm of one plus the number
of confirmed COVID-19 cases per 100,000 people. Second, to measure the exposure of a bank’s
business clients to COVID-19, we compute the weighted average of ln(cases per capita) across
counties in which the bank has branches, where the weights are the shares of deposits held by the
bank in those counties. We call this measure Bank exposure. We define small business loans as
7
loans with origination values of $1 million or less following the classification of Community
Reinvestment Act (CRA) and measure changes in a bank’s lending to small firms using the
growth in the dollar amount of loans to small businesses. We consider two other measures of
credit supply, overall C&I loan growth and PPP loans. C&I loan growth equals the log
difference between commercial & industrial loans between Q2-2020 and Q4-2019 for each bank.
PPP loans equals the amount of PPP loans outstanding as of Q2-2020 scaled by total assets as of
Q4-2019. According to Li and Strahan (2021), bank credit supply through the CARES Act’s
Paycheck Protection Program (PPP) accounts for a major of bank loans during first few months
of 2020. We then regress the three measures of small business loan growth on Bank exposure,
while controlling for each bank’s size, liquid assets, equity ratio, tier 1 ratio, NPL, ROA, and the
U.S. state in which it is headquartered.
We find a positive relation between a bank’s growth in lending to small firms and Bank
exposure among smaller banks but not among larger banks. We measure the differences between
smaller and large banks in two ways: using a discrete cutoff, where banks with less than $1
billion of assets are defined as “small” and all other banks are defined as “large,” and using a
continuous measure of bank size. The findings from both approaches indicate that smaller banks
increase lending to small firms more than larger banks during the pandemic. These cross-
sectional analyses are consistent with the view that small banks have a comparative advantage in
funding small firms during periods of economic duress. The results hold when examining C&I
loan growth and PPP loans. This is consistent with the findings in Li and Strahan (2021), which
stresses a new benefit of bank relationships: small banks prioritize serving their relationship
borrowers during the Covid crisis, facilitating small firms to access government-subsidized
lending.
Finally, we evaluate whether employment at small firms falls by less in response to local
COVID-19 cases in areas with a larger share of small banks. The LEHD Quarterly Workforce
Indicators (QWI) database provides granular employment data at the county, two-digit NAICS
sector, and firm size groups, allowing us to distinguish employment by firm size. The frequency
8
of the QWI data is quarterly, not weekly. The effects of the share of small bank on employment
are more pronounced among small firms. This is consistent with the notion that that small banks
provide liquidity to small firms during periods of economic duress, mitigating the adverse effects
of the Covid-19 pandemic on employment.
Our work relates to a growing body of research on the linkages between financial and
markets. Beck, Levine, and Levkov (2010) find that regulatory changes that improved bank
lending boosted the demand for labor, especially the demand for low-income workers, reducing
income inequality. Focusing on the 2008-9 financial crisis and the relationship between
syndicated lenders and large firms, Chodorow-Reich (2014) shows that firms that had pre-2008
relationships with syndicated lenders that were more severely hurt by the 2008-9 financial crisis
reduced employment by more following the onset of the crisis than otherwise similar firms with
pre-crisis connections to syndicated lenders that were less adversely affected by the crisis.
Giroud and Mueller (2016) show that establishments of more highly levered firms experienced
larger employment losses in response to drops in local consumer demand during the great
recession. Exploiting stock market fluctuations, Lin (2020) shows that stock market booms
reduce household demand for demand deposits, and the deposit outflows induce banks to cut
lending, leading to contractions in local employment. Focusing on the bank lending channel,
Brown and Earle (2017) find that relaxing credit constraints on small firms tend to increase
employment, but Greenstone, Mas, and Nguyen (2020) challenge these results, finding that the
bank lending channel did not have material effects on employment following the 2008-9
financial crisis tightened credit constraints. Our research and contributions are distinct. Rather
than focusing on the ramifications of shocks to bank lending in general or the syndicated lenders
in particular, we examine how the structure of local banking markets shapes the resilience of
local employment, the structure of local employment across low-, medium-, and high-income
workers, and the revenues of small local firms to the COVID-19 pandemic. We discover that a
larger share of small banks in a county cushions the impact of COVID-19 cases on employment,
especially the employment of low-income workers, and on the revenues of small local firms.
9
Our research also relates to studies on bank financing and resilience to shocks. For
example, Chen, Hanson, and Stein (2017) show that large banks, especially the top-four banks,
cut lending to small businesses more than other banks during the 2008 financial crisis and were
slow to resume small business lending during the post-crisis recovery. Related studies investigate
the role of bank lending in how economies recover from natural disasters, such as hurricanes
(Schuwer, Lambert, and Noth 2019) and floods (Koetter, Noth, and Rehbein 2020). Different
from these studies, we examine the ramifications of the COVID-19 pandemic that many
influential economists describe as very different from the past crises (e.g., Bernanke 2020;
Reinhart 2020). For example, the 2020 pandemic was at its core a public health emergency that
triggered an abrupt constriction in economic activity and an especially severe decline in
employment, whereas the GFC arose from concerns about overall financial system stability and
the safety of individual banks. In light of these differences, our research offers the first
investigation of the role of local bank structure in shaping economic resilience to the COVID-19
pandemic by exploiting high-frequency, real-time data on employment and other economic
indicators.
2. Data, Summary Statistics, and Graphical Analyses
2.1 The Economic Tracker database
We use a new public database, the Economic Tracker (“the Tracker” henceforth), that
tracks daily economic activity at the county level. The data are compiled by Chetty, Friedman,
Hendren and Stepner (2020) and the Opportunity Insights Team. The Tracker reports daily
statistics on employment rates, small business revenues, and COVID-19 infection rates. Chetty et
al. (2020) (1) smooth high-frequency fluctuations by reporting 7-day moving averages, (2)
smooth lower-frequency seasonal fluctuations by normalizing each week’s value in 2020 relative
to the corresponding values for the same week in 2019, and (3) report statistics in the form of
percentage changes relative to January 2020, as opposed to raw values, to protect the privacy of
businesses and their clients. Finally, to construct representative series and mitigate selection
10
biases, they compare each time series to national benchmarks and exclude data sources that do
not track corresponding benchmarks closely. We now describe each data series used in our study,
where each series is defined as a percentage deviation from the level in January 2020.
2.2 Employment and unemployment claims
Employment data are compiled by combining employment and earnings information
from four complementary data sources: Paychex, Intuit, Earnin, and Kronos. Paychex collects
paycheck data through offering payroll services to 670,000 small- and medium-sized businesses
across the U.S., accounting for 10% of U.S. private-sector workers. Chetty et al. (2020) convert
these paycheck data into employment estimates (i.e., the number of paid employees). Similarly,
Intuit provides payroll services to one million businesses, through which they estimate
employment data. Both Paychex and Intuit provide a firm-based payroll dataset. As a
complement, Earnin provides a worker-level sample that records individual workers’ earning of
income. Chetty et al. (2020) construct the employment series by converting these paycheck data
into employment measures. Kronos offers workforce management service to 30,000 firms which
employed a total of 3.2 million workers before the pandemic. The employment data are derived
from electronic timesheets on which employees clock into work.
Our sample of employment comprises 17,536 county-week observations over the period
from February through June of 2020, involving 800 counties. Employment is the employment
level for all workers relative to January 2020 at the county-level. As shown in Table 1, the
average Employment equals -10%, suggesting that the employment rate fell by an average
weekly rate of 10% over the period from February through June of 2020.
The Tracker also provides disaggregated employment series by income quartile, allowing
us to evaluate the differential effects on workers in different income groups. In particular,
Employment, low income workers represents the employment level for workers in the bottom
quartile of the income distribution (incomes approximately under $27,000). Employment, middle
income workers represents the employment level for workers in the middle two quartiles of the
11
income distribution (incomes approximately $27,000 to $60,000). Employment, high income
workers represents the employment level for workers in the top quartile of the income
distribution (incomes approximately over $60,000).
We also examine unemployment insurance claims as a robustness check. Data on
unemployment benefits are constructed using public data sources including the Department of
Labor and state government agencies. Unemployment insurance claims equals the total number
of initial claims divided by the 2019 labor force (in 100 people).
2.3 Small business revenue
The Tracker provides daily data on small business revenues at the county level. In
particular, Small business revenue equals the seasonally adjusted net revenue of small businesses
in a county relative to January 4-31 2020. Daily values are reported as a seven-day moving
average of the current day and the previous six days. Raw data on small business transactions
and revenues come from Womply, a company that collects information from credit card
processors. Womply provides a firm-based panel on total revenues of small businesses. As
shown in Table 1, the average Small business revenue equals -20%, suggesting that small
business revenue in an average county falls by 20% over February through June of 2020.
2.4 COVID-19 incidence
The Economic Tracker also includes county-level data on COVID-19 cases. In particular,
Ln(Cases per capita) equals the logarithm of one plus the number of confirmed cases per
100,000 people in a county on each day. The daily number of cases on a particular date is the
cumulative number of cases since January 21, 2020. For a particular week, we use the value of
Ln(Cases per capita) on Sunday of that week. As reported in Table 1, the average number of
cumulative cases per 100,000 people in a county equals 154 across all counties and weeks.
Figure 1 plots the cross-county distribution of the cumulative number of confirmed cases per
100,000 people as of the end of June 2020.
12
2.5 Local bank structure and other country traits
To assess whether small banks have a comparative advantage over big banks in funding
local businesses during the pandemic, we construct two measures of the degree of small bank
presence in a county. Using data from the Summary of Deposits (SoD) in 2019, Share_small
bank branches equals the proportion of bank branches that belong to small U.S. banks.
Following existing studies (e.g., Berger, Bouwman, and Kim 2017), we define a small bank as
having total assets of $1 billion or less. The second measure, Share_small bank deposits, equals
the proportion of deposits held by small U.S. banks in a county. Figure 2 plots the cross-county
distribution of Share_small bank branches. As shown, there exists large variations across
counties, including across counties within the same state. Table 1 shows that the average value of
Share_small bank branches equals 0.4, with a standard deviation equal to 0.3. The other
measure, Share_small bank deposits, exhibits a similar statistical pattern. Furthermore, we
consider the overall availability of bank branches in a county. Specifically, #Branches per capita
equals the total number of bank branches in a county divided by the population (in 100,000
people).
To isolate the relation between economic resilience to COVID-19 and local financial
structure, we consider an array of other county characteristics that might influence local
economic responses to COVID-19. First, to control for possibility that income shapes responses
to the pandemic, as suggested by Chetty et al. (2020), we condition on Income per capita, which
is from the Bureau of Economic Analysis and equals the total amount of personal income
received by residents in a county divided by population in that county. Second, we use two
measures of local demographics based on U.S. congressional data: 2 %65+ pop equals the
proportion of local population aged above 65, and Population density equals the number of
residents per square mile. We control for these demographics indicators because older
2
https://www.jec.senate.gov/public/index.cfm/republicans/2018/4/the-geography-of-social-capital-in-america.
13
individuals and individuals in more dense locales are more susceptible to COVID-19. As a result,
demographics may shape a community’s economic resilience to the pandemic. Third, to capture
a county’s pre-pandemic economic conditions, we account for the unemployment rate in 2019
using data from the Bureau of Labor Statistics. Unemployment rate equals the proportion of
unemployed labor force to total labor force in each county. In this regard, our analyses account
for differences in local communities’ sensitivity to COVID-19 arising from income, population
age, population density, and initial economic conditions. 3
2.6 Graphical analyses
To illustrate the connection between local economic resilience to the pandemic and the
share of small banks in a county, we compare the evolution of employment in counties with
“high” and “low” shares of small banks. We define a county as having a “high” share of small
banks if Share_small bank branches is in the upper tercile of the cross-county sample. We
categorize a county as having a “low” share of small banks if Share_small bank branches is in
the lower tercile. We then plot the average value of Employment in each week from February
through June for counties in the high- and low-small bank share counties respectively. As shown
in Figure 3, employment falls by less in counties with a high share of small banks than it does in
counties with a low share of small banks.
We next focus on the evolution of the structure of employment across low-, middle-, and
high-income workers. We separately trace the deviation of the level of employment of low-,
middle-, and high-workers relative to their pre-pandemic employment levels over the period
from February through June 2020. As shown in Figure 4, the fall in employment is especially
severe and enduring among low-income workers, as Employment, low income workers falls more
sharply in March and April and recovers much more slowly after April than Employment, middle
income workers and Employment, high income workers.
3
For exposition purposes, we multiply the coefficient estimates on Income per capita and Population density by
1000 in the regression tables reported below.
14
We also find that the severity of the pandemic on low-income workers is less pronounced
in counties with higher shares of small banks. Specifically, we focus on the evolution of
Employment, low income workers across counties with different shares of small firms. We again
differentiate between counties with a “high” share of small banks (Share_small bank branches in
the upper tercile of counties) and counties with a “low” share of small banks (Share_small bank
branches is in the lower tercile). We then plot the average value of Employment, low income
workers for high- and low-small bank share counties respectively from February through June.
As shown in Figure 5, the employment of low-income workers falls by less in counties with a
high share of small banks than it does in counties with a low share of small banks.
These graphical analyses are consistent with the view that small banks enhance local
economic resilience—especially among low-income workers. We next implement standard
empirical methods that control for other local factors shaping economic responses to the
pandemic to further assess the independent connection between local economic resilience to
COVID-19 and the structure of local banking markets.
3. Empirical Strategy
To examine the heterogeneous effects of the COVID-19 pandemic on economic activities
while differentiating by local banking structure, we employ the following regression model.
𝑌𝑌𝑐𝑐,𝑡𝑡 = 𝛼𝛼0 + 𝛽𝛽𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 ∗ 𝑆𝑆ℎ𝑎𝑎𝑎𝑎𝑎𝑎 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐵𝐵𝑎𝑎𝑎𝑎𝑎𝑎𝑐𝑐 + 𝛾𝛾𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡
𝑿𝑿𝒄𝒄 ∗ 𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 + 𝛼𝛼𝑐𝑐 + 𝛼𝛼𝑠𝑠,𝑡𝑡 + 𝜀𝜀𝑐𝑐,𝑡𝑡 , (1)
where c and t index county and week, respectively. The dependent variable, Yc,t, represents one
of the indicators of local economic conditions in county c and week t (e.g., Employment). Share
Small Bankc represents one of the two measures of the share of small banks in county c before
the pandemic (i.e., Share_small bank branches or Share_small bank deposits). Ln(Cases per
capita)c,t denotes the logarithm of one plus the cumulative number of confirmed cases per capita
15
in county c on Sunday of week t, which captures county-specific exposure to the pandemic. The
coefficient of interest, β, captures how the relation between changes in county employment rates
and COVID-19 infection rates differs by the share of small banks in counties. We estimate the
model using OLS, with the standard errors clustered at the county level.
To isolate the effects of small banks on local economic resilience to COVID-19, we
include an array of fixed effects and a set of interaction terms between Ln(Cases per capita)c,t
and other county traits. Regarding fixed effects, we include county fixed effects, αc, that
condition out all time-invariant factors at the county level, including differences in the level and
structure of economic activities. The analyses also include state-week fixed effects, αs,t, to
control for all time-varying factors at the state level, including differences in policy responses to
the pandemic. By including state-time effects, we are essentially estimating the differential
sensitivity of local economies to COVID-19 exposures between two counties within the same
state and week. In addition, we condition on the interaction between Ln(Cases per capita)c,t and a
vector of pre-pandemic country traits, Xc , that includes #Branches per capita, Income per
capita, %65+ pop, Population density, and Unemployment rate.
4. Results
4.1 Small banks and employment during COVID-19, baseline results
Results in Panel A of Table 2 show that as COVID-19 cases grew, employment fell, but it
fell by less in counties with a larger proportion of small banks. Specifically, we first note in
column 1 of Panel A that Ln(Cases per capita) enters negatively and significantly, suggesting
that employment rates drop more in counties exposed to a larger number of COVID-19 cases. To
interpret the economic magnitude, the coefficients from column 1 indicate that an average
COVID-19 shock (Ln(Cases per capita) =2.8) is associated with a 2 percentage point
(=2.8*0.72) drop of employment rates per week, which is equivalent to 19% of the standard
deviation of Employment. Second, note in column 2 that the interaction term, Share_small bank
branches*Ln(Cases per capita), enters positively and significantly. This indicates that the
16
adverse effect of COVID-19 on employment is milder in counties with a larger share of small
banks in local banking markets. Furthermore, as shown in column 4, these results hold when
measuring the share of small banks in a county using deposits rather than the number of
branches.
These results hold when accounting for other county traits that might shape how
employment in a county responds to local COVID-19 infection rates. Specifically, we include
the interaction between Ln(Cases per capita) and each of the following county traits: Income per
capita, %65+ pop, Population density, and Unemployment rate. As shown in columns 3 and 5,
we continue to find that employment falls by less in response to COVID-19 in counties with a
larger share of small banks. That is, the interaction term, Share_small bank branches*Ln(Cases
per capita) in column 3 and Share_small bank deposits*Ln(Cases per capita) in column 5,
continues to enter positively and significantly, suggesting that the proportion of small banks in
local banking markets exerts an independent effect on employment resilience to COVID-19.
To illustrate the degree to which small banks in a county mitigate the adverse effects of
local COVID-19 cases on local employment, consider the coefficient estimates from column 3 of
Panel A. Furthermore, consider an average weekly COVID-19 shock of 2.8, which is the sample
mean of Ln(Cases per capita). Next compare two otherwise similar counties, except that the
“high” small bank share county has one standard deviation larger small bank share than the
“low” small bank share county, so that the difference in Share_small bank branches between the
two counties is 0.33. The coefficient estimates from column 3 indicate that the negative impact
of an average weekly COVID-19 shock on the weekly decline in employment would be 0.68
(=0.733*0.33*2.8) percentage points smaller in the county with a “high” share of small banks
than in the county with a “low” share of small banks. This is equivalent to 6% of the standard
deviation of Employment (=10.5).
When examining unemployment insurance claims in Panel B of Table 2, we find that
claims increase with COVID-19 infection rates, but they rise by less in counties with a larger
share of small banks. As shown in column 1, the linear term Ln(cases per capita) enters
17
positively and significantly, indicating that the rate of unemployment insurance claim is higher in
counties with higher COVID-19 infection rates. The interaction between Ln(cases per capita)
and the proportion of small banks in a county, i.e., Share_small bank branches or Share_small
bank deposits, enters negatively and significantly in all columns, suggesting that unemployment
insurance claim rates increase by less in response to COVID-19 in counties with a greater share
of small banks. Again, the effects of small banks on unemployment insurance claims hold when
including (a) a full set of county and state-week fixed effects, and (b) the interaction between
Ln(cases per capita) and each county’s total number of bank branches per capita in the county,
income per capita, percentage of people over the age of 65, population density, and pre-
pandemic unemployment rate. Taken together, the baseline results reported in Table 2 are
consistent with the view that small banks have a comparative advantage in cushioning the
adverse effects of the duress triggered by COVID-19 on labor markets.
4.2 Income distributional implications of small banks
We next examine the income distributional implications of small banks by separately
examining the cushioning effects of small banks in response to COVID-19 infection rates on the
employment rates of low-, middle-, and high-income workers. More specifically, we re-estimate
equation (1) in which the dependent variable is either (1) Employment, low income workers,
which equals the percentage change in the employment rate low-income workers relative to
January 2020, where low-income workers are defined as those in the bottom quartile of the
income distribution, (2) Employment, middle income workers, which equals the percentage
change in the employment rate of middle-income workers relative to January 2020, where
middle-income workers are defined as those in the middle two quartiles of the income
distribution, and (3) Employment, high income workers, which equals the percentage change in
the employment rate of high-income workers relative to January 2020, where high-income
workers are defined as those in the top quartile of the income distribution. Each of these
18
employment rate measures is computed at the county-week level and is measured at the
percentage change from the county’s employment level in January of 2020.
Results reported in Table 3 suggest that small banks have a particularly strong cushioning
effect on the employment of low-income workers during the pandemic. The dependent variable
in columns 1 – 4 is Employment, low-income workers. As shown in columns 1 – 4, where the
dependent variable is Employment, low-income workers, the coefficient estimates on
Share_small bank branches*Ln(cases per capita) and Share_small bank deposits*Ln(cases per
capita) are positive and statistically significant. These findings suggest that the employment of
low-income workers falls by less in response to local COVID-19 infections in counties with a
higher proportion of small banks. With respect to the size of the estimated effect, consider again
an average weekly COVID-19 shock (=2.8), and compare two otherwise similar counties, except
that the “high” small bank share county has one standard deviation (=0.33) larger small bank
share than the “low” small bank share county. The coefficient estimates from column 2 indicate
that the negative impact of an average weekly COVID-19 shocks on the weekly decline in
employment of low income workers would be 1.3 (=1.427*0.33*2.8) percentage points smaller
in the “high” small bank share county than in the “low” small bank share county. This is
equivalent to 8% of the standard deviation of Employment, low-income workers. In contrast,
when examining employment rates of middle-income and high-income workers, the coefficient
estimates on these interaction terms are economically small and not statistically robust (columns
5 – 12). The Table 3 results indicate that the share of small banks in a county mitigated the
adverse impact of local COVID-19 infection rates on the local employment of low-income
workers.
There might be concerns that the connection between the share of small banks in a county
and the reaction of local employment to COVID-19 infection rates is driven by bank traits other
than the local presence of small banks. To mitigate the concern, we control for an additional set
of county traits, including banking market concentration (HHI) and county-aggregated bank
characteristics (including Liquid assets, Equity ratio, Tier 1 ratio, NPL, and ROA averaged at the
19
county level). Banking market concentration equals the HHI of each bank’s market share in a
county as of 2019. Regarding county-aggregate bank characteristics, take Liquid assets as an
illustrative example, we calculate the county-specific Liquid assets as the weighted average of
bank-specific Liquid assets across banks in each county, weighted each bank by its share of
deposits in that particular county. Using the same approach, we construct county-specific bank
characteristics for each of the bank traits.
As shown in Table 4, conditioning on these additional bank market traits and the
interaction with Ln(cases per capita) does not alter the results on the connection between the
proportion of small banks and the reaction of employment to local COVID-19 infection rates
across income distribution. The interaction terms between Ln(cases per capita) and the
proportion of small banks in a county, i.e., Share_small bank branches or Share_small bank
deposits, continue to enter positively and significantly in employment for low-income workers.
In addition, the estimated coefficients on the interaction terms are similar to those in Table 3.
4.3 Small banks and small business revenue during COVID-19
One implication of the view that small banks have a comparative advantage in lending to
local businesses during the economic turmoil triggered by local COVID-19 infection rates is that
local firms in counties with a higher proportion of small banks will be better able to maintain
their operations. To assess this implication, we examine whether the revenues of small local
firms fall by less in response to COVID-19 in counties in which small banks compose a greater
share of the local banking market. We employ the same empirical model as in the employment
analyses, but the dependent variable now becomes Small business revenue, which equals the
percent change in seasonally adjusted net revenues for small businesses in each county relative to
its pre-pandemic level.
Results reported in Table 5 show that in response to a county’s COVID-19 infections,
small business revenue drops by less in counties with higher share of small bank. Column 1
shows that the county-specific COVID-19 infection rate, Ln(cases per capita), enters negatively.
20
Note, however, that the interaction between Ln(cases per capita) and the measure of the share of
small banks in the local banking market (Share_small bank branches or Share_small bank
deposits) enters positively and significantly in columns 2 – 5. This suggests that Small business
revenue falls by less in response to the pandemic in counties with a higher share of small banks.
These findings are consistent with the view that small banks have a comparative advantage in
financing small firms during the pandemic and this funding allows those small businesses to
operate more effectively and generate greater revenues.
5. Instrumental Variables and the Small Bank-Small Firm Mechanism
In assessing the degree to which the share of small banks in a county shapes its resilience
to local COVID-19 infection rates, we have conditioned on the potential influences of many
other local characteristics. Besides controlling for county and state-week fixed effects, the
analyses have conditioned on the interaction between county infection rates and county-level
information on (a) income per capita, (b) the proportion of the population above 65 years old, (c)
population density, (d) the pre-pandemic unemployment rates, and (e) the number of bank
branches per capita. Thus, we have so far used a control function approach to address omitted
variable concerns and draw inferences about how small banks shape a local economy’s resilience
to the COVID-19 infections during the first half of 2020.
Despite the extensive controls included in the analyses above, there could be omitted
county-specific features that are highly correlated with the local share of small banks and it
might be these omitted features, and not financial structure, that shape cross-county responses to
the pandemic. To further address this concern, we use two strategies. First, we use an
instrumental variable to identify the impact of the share of small banks on cross-county
responses to local COVID-19 infection rates. Second, we examine the direct link between small
banks and their lending to small firms in response to the pandemic. Third, we investigate
whether the impact of small banks manifests primarily among small firms. Since the theoretical
mechanism linking small banks to economic resilience operates through small banks having a
21
comparative advantage over larger banks in lending to small firms during periods of economic
turmoil, evidence of this link enhances identification and hence the interpretation of the results.
5.1 IV tests
We use an instrumental variable (IV) for cross-county differences in the share of small
banks. In particular, we use a measure of the consolidation of the local banking market triggered
by the merger of out-of-state banks. For each county, we compute #Mergers by out-of-state
banks as the number of mergers by two out-of-state banks during the ten years, 2010 – 2019,
before the COVID-19 pandemic, where both acquiring and target banks have at least one branch
located in the county prior to the merger. Thus, for a county c located in state s, the merger
involves two banks that are not headquartered in state s, where each bank has at least one branch
in county c. This merger-induced consolidation will tend to give the larger bank a competitive
advantage in county c, reducing the role of small banks in the county. Consistent with this view,
we find that a county’s exposure to #Mergers by out-of-state banks over the past decade is
negatively associated with the market share of small banks at the end of 2019. The identifying
assumption is that the decision of two banks headquartered in states different from county c to
merge during the 2010-2019 period does not reflect local conditions in county c that drive the
county’s resilience to the 2020 pandemic.
We then reassess the degree to which the share of small banks in a county shapes its
resilience to local COVID-19 infection rates using two-stage least squares. In particular, as the
excluded the instrument for Share_small bank branches*Ln(cases per capita), we use #Mergers
by out-of-state banks*Ln(cases per capita). In the second stage, we use the predicted values of
Share_small bank branches*Ln(cases per capita) to reevaluate the baseline regression
specifications using this IV approach.
The first-stage estimation results show that the instruments are strongly correlated with
Share_small bank branches*Ln(cases per capita). As shown in column 2 of Table 6, the F-
22
statistic of the null hypothesis that the excluded instruments do not account for cross-county
variations in the share of small banks prior to the pandemic equals 51, suggesting that the
instruments are not weak. With respect to the exclusion restriction, we note that the instruments
are constructed from a plausibly exogenous source of variation: out-of-state bank mergers.
As shown in Table 6, the second stage of the IV analyses confirms the results reported
above: a larger share of small banks in a county mitigates the adverse shock of COVID-19 on
employment, especially the employment of low-income workers. As reported in column 1 of
Table 6, the instrumented Share_small bank branches*Ln(cases per capita) enters the second-
stage regressions positively and significantly. The results hold when controlling for county and
state-week fixed effects. The estimated “cushioning” effect of small banks on employment
during the pandemic is economically large. To illustrate the economic magnitude, consider an
average weekly COVID-19 shock (Ln(Cases per capita) =2.8), and two otherwise similar
counties, except that the share of small banks in a county (“high” small bank share county) is one
standard deviation (0.33) larger than that of the other county (“low” small bank share county).
The coefficient estimates in column 1 indicate that employment rates in the “high” small bank
share county would fall by 1.8 (=1.92*0.33*2.8) percentage points less than that of the “low”
small bank share county. This is large, as it corresponds to 17% of the standard deviation of
Employment. Furthermore, as shown in column (3), the resilience-enhancing effects of small
banks are especially pronounced for low-income workers.
The IV analyses also confirm that in response to COVID-19 infections, small business
revenues drop by less in counties with higher small bank shares. As stressed above, the argument
that small banks have a comparative advantage in funding small firms implies that small firms in
counties with a higher proportion of small banks will perform better during the pandemic than
otherwise similar firms in counties with a lower proportion of small banks. One implication of
the view that small banks have a comparative advantage in lending to local businesses during the
economic turmoil triggered by local COVID-19 infection rates is that local firms in counties with
a higher proportion of small banks will be better able to maintain their operations. As shown in
23
column 9 of Table 6, the IV regression results are consistent with the view that small local banks
help cushion the adverse effects of the pandemic on small businesses to a greater degree than
larger banks, bolstering small firm operations and revenues.
5.2 Bank lending to small business during COVID-19, small vs. large banks
The analyses so far indicate that employment—especially employment of low-income
workers—and small-firm revenues are more resilient to the COVID-19 pandemic in counties
with a higher proportion of small banks. One explanation for these results is that small banks
have a comparative advantage in lending to small firms during period of economic turmoil,
cushioning the impact of the pandemic in counties with larger small bank shares. The findings on
low-income workers are consistent with this explanation since small firms disproportionately
employ low-income workers. Similarly, the findings on small business revenues support this
explanation as they suggest that high small-bank counties support small firms, allowing them to
better maintain operations and revenues. We now examine a specific theoretical mechanism of
how small banks boost economic resilience: the direct link between small banks and their
lending to small firms during the pandemic.
To examine lending by small banks to small firms, we move from the county-week
analyses of economic resilience to a cross-sectional analysis of the relation between changes in a
bank’s lending to small firms and the bank’s exposure to business clients affected by local
COVID-19 cases. Since we do not have data on the addresses of each client, we measure the
exposure of a bank’s business clients to COVID-19 by building on the findings in Nguyen
(2019), who shows that banks tend to make loans to businesses that are geographically very close
to their branches. Thus, we compute Bank exposure as the weighted average of ln(cases per
capita) across counties in which the bank has branches, where the weights are the shares of
deposits held by the bank’s branches in those counties. To match Bank exposure to the quarterly
Call Reports data, we compute Bank Exposure using Ln(cases per capita) as of June 30, 2020.
24
To measure each bank’s lending to small firms, we follow the classification of
Community Reinvestment Act (CRA) and define small business loans as loans with origination
values of $1 million. We measure changes in a bank’s lending to small firms, Small business
loan growth, as the growth in the dollar amount of loans to small businessesMore specifically,
Small business loan growth equals the log difference between small business loans (in the dollar
amount) between Q2-2020 and Q4-2019 for each bank. We also consider overall C&I loan
growth and PPP loans. C&I loan growth equals the log difference between commercial &
industrial loans between Q2-2020 and Q4-2019 for each bank. PPP loans equals the amount of
PPP loans outstanding as of Q2-2020 scaled by total assets as of Q4-2019. According to Li and
Strahan (2021), bank credit supply through the CARES Act’s Paycheck Protection Program
(PPP) accounts for a major of bank loans during first few months of 2020.
We consider two measures of bank size in our analyses of the connection between bank
size and lending to small firms. First, as a continuous measure of bank size, we use Ln(Assets),
which equals the natural log of total assets of each bank at the close of 2019. Second, as a
discrete measure, we use Small bank, which equals one if a bank’s total assets at the close of are
below $1 billion and zero otherwise.
Given these measures, we estimate the following regression model:
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔ℎ𝑏𝑏 = 𝛼𝛼0 + 𝛽𝛽𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑏𝑏 ∗ 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑏𝑏 + 𝛾𝛾𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑏𝑏 +
𝑋𝑋𝑏𝑏 + 𝛼𝛼𝑠𝑠 + 𝜀𝜀𝑏𝑏 , (2)
where there is one observation per bank, b, and where 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑏𝑏 is either Ln(Assets) or Small bank.
The regression also conditions on, 𝑋𝑋𝑏𝑏 , which is a vector of bank characteristics including
Ln(Assets) (the log of book value of total assets), the Equity ratio (the ratio of total equity to total
assets), ROA (net income divided by total assets), Single market (an indicator of whether a bank
operates in only one market or not), Tier 1 capital (the ratio of tier 1 capital to risk-weighted
assets), Liquid assets (the ratio of liquid assets to total assets), and NPL (non-performing loans to
25
total assets). The term, 𝛼𝛼𝑠𝑠 , denotes a set of state fixed effects for each bank’s headquarters. We
estimate equation (2) using OLS, and report heteroskedasticity robust standard errors.
As reported in Table 7, there is a larger increase in small business lending by small banks
in response to COVID-19 than by large banks. First note that there is a strong, positive relation
between Bank exposure and Small business loan growth, measured either as Small business loan
growth as shown in column 1. This finding suggests that banks increase lending to small firms
during the first few months of the pandemic. Second, note that when distinguishing banks by
size, we find that the increase in small business lending during COVID-19 is stronger among
smaller banks. As shown in columns 2 and 3, Ln(Assets)*Bank exposure enters negatively and
significantly, and Small bank*Bank exposure enters positively and significantly, suggesting that
smaller banks boost lending to small firms more in reaction to the pandemic than larger banks.
When using the discrete measure of bank size (i.e., Small bank), we find that the linear term,
Bank exposure, enters insignificantly in column 3. These findings indicate that small banks with
total assets up to $1 billion increased lending to small firms, whereas large banks with more than
$1 billion total assets did not significantly change lending to small firms. The results hold when
examining C&I loan growth as reported in columns 4 – 6. Furthermore, columns 7 – 9 show that
small banks provide more PPP loans in response to the Covid-19 exposure. This is consistent
with the findings in Li and Strahan (2021), which stresses a new benefit of bank relationships:
small banks prioritize serving their relationship borrowers, facilitating small firms to access
government-subsidized lending. Overall, these cross-bank analyses are consistent with the view
that small banks have a comparative advantage in providing liquidity to small firms during
periods of economic duress.
5.3 Employment resilience to COVID-19, small vs. large firms
We next evaluate whether employment at small firms falls less in areas with a larger
share of small banks. To do this, we use the LEHD Quarterly Workforce Indicators (QWI)
database, which provides disaggregate employment data at the county, two-digit NAICS sector,
26
and firm size groups. The frequency of the QWI data is at the quarterly level. We estimate the
following model specification.
𝐸𝐸𝐸𝐸𝐸𝐸𝑐𝑐,𝑗𝑗,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡,𝑡𝑡 = 𝛽𝛽𝑆𝑆ℎ𝑎𝑎𝑎𝑎𝑎𝑎 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑐𝑐 ∗ 𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 ∗ 𝒕𝒕𝒕𝒕𝒕𝒕𝒕𝒕
+ 𝛾𝛾𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 + 𝜌𝜌𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 ∗ 𝒕𝒕𝒕𝒕𝒕𝒕𝒕𝒕
+ 𝜃𝜃𝑆𝑆ℎ𝑎𝑎𝑎𝑎𝑎𝑎 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑐𝑐 ∗ 𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 + 𝑿𝑿𝒄𝒄
∗ 𝐿𝐿𝐿𝐿(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐)𝑐𝑐,𝑡𝑡 + 𝛼𝛼𝑐𝑐,𝑗𝑗,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 + 𝛼𝛼𝑗𝑗,𝑡𝑡 + 𝛼𝛼𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡,𝑡𝑡 + 𝛼𝛼𝑠𝑠,𝑡𝑡 + 𝜀𝜀𝑐𝑐,𝑗𝑗,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡,𝑡𝑡
Where c, j, type, and t index county, sector, firm size groups, and quarter, respectively. We
consider three firm size groups based 1–49, 50–499, and 500+ employees.
The estimation results reported in Table 8 are consistent with our hypothesis. As shown,
the effects of the share of small bank on employment are more pronounced among small firms.
This is consistent with the notion that that small banks provide liquidity to small firms during
periods of economic duress, mitigating the adverse effects of the Covid-19 pandemic on
employment.
6. Conclusions
In this paper, we examine how the proportion of small banks operating in a local
economy influences the resilience of that economy with respect to employment, the distribution
of employment across low-, middle-, and high-income workers, the revenues of small local
firms, and the countercyclical funding of small businesses to the COVID-19 pandemic. We
exploit high-frequency, county-level data on employment, small business revenue, and COVID-
19 cases.
We discover that employment, especially the employment of low-income workers, falls
by less in response to local COVID-19 infection rates in counties with a larger proportion of
small banks. Furthermore, the pandemic-induced drop in small business revenue was milder
among counties in which small banks compose a larger share of local banking markets. Finally,
we find that in response to the pandemic, small banks increase their small business lending more
27
than larger banks. The evidence suggests that small banks have a comparative advantage in
ameliorating information asymmetries and financing small firms following the adverse shock of
COVID-19.
28
References
Acharya, V. V., and Steffen, S., 2020. The risk of being a fallen angel and the corporate dash for
cash in the midst of COVID, Review of Corporate Finance Studies, 9(3), pp. 430-471.
Acs, G. and Nichols, A., 2007. Low-income workers and their employers. The Urban Institute.
Beck, T., Degryse, H., De Haas, R. and Van Horen, N., 2018. When arm’s length is too far:
Relationship banking over the credit cycle. Journal of Financial Economics, 127(1),
pp.174-196.
Beck, T., Demirgüç-Kunt, A. and Maksimovic, V., 2008. Financing patterns around the world:
Are small firms different?. Journal of Financial Economics, 89(3), pp.467-487.
Beck, T., Levine, R., Levkov, A., 2010. Big bad banks? The winners and losers from bank
deregulation in the United States. Journal of Finance 65, 1637–1667.
Berger, A.N., Bouwman, C.H. and Kim, D., 2017. Small bank comparative advantages in
alleviating financial constraints and providing liquidity insurance over time. The Review
of Financial Studies, 30(10), pp.3416-3454.
Berger, A.N., Bouwman, C.H., Norden, L., Roman, R.A., Udell, G.F., Wang, T., 2020. Is a
Friend in Need a Friend Indeed? How Relationship Borrowers Fare during the COVID-
19 Crisis. Available at SSRN.
Berger, A.N., Cerqueiro, G. and Penas, M.F., 2015. Market size structure and small business
lending: Are crisis times different from normal times?. Review of Finance, 19(5),
pp.1965-1995.
Berger, A.N., Goulding, W. and Rice, T., 2014. Do small businesses still prefer community
banks?. Journal of Banking & Finance, 44, pp.264-278.
Berger, A.N., Miller, N.H., Petersen, M.A., Rajan, R.G. and Stein, J.C., 2005. Does function
follow organizational form? Evidence from the lending practices of large and small
banks. Journal of Financial Economics, 76(2), pp.237-269.
Berger, A.N. and Udell, G.F., 1998. The economics of small business finance: The roles of
private equity and debt markets in the financial growth cycle. Journal of banking &
finance, 22(6-8), pp.613-673.
Berger, A.N. and Udell, G.F., 1995. Relationship lending and lines of credit in small firm
finance. Journal of business, pp.351-381.
Berger, A.N. and Udell, G.F., 2002. Small business credit availability and relationship lending:
The importance of bank organisational structure. The Economic Journal, 112(477),
pp.F32-F53.
29
Bernanke, B.S., 2020. Ben Bernanke on COVID-19 downturn. https://www.marketplace.org
/2020/03/23/former-fed-chair-bernanke-covid19-downturn/.
Bernanke, B.S., Gertler, M. and Gilchrist, S., 1999. The financial accelerator in a quantitative
business cycle framework. Handbook of macroeconomics, 1, pp.1341-1393.
Bolton, P., Freixas, X., Gambacorta, L. and Mistrulli, P.E., 2016. Relationship and transaction
lending in a crisis. The Review of Financial Studies, 29(10), pp.2643-2676.
Brown, J.D. and Earle, J.S., 2017. Finance and growth at the firm level: evidence from SBA
loans. The Journal of Finance, 72(3), pp.1039-1080.
Brunnermeier, M.K. and Sannikov, Y., 2014. A macroeconomic model with a financial
sector. American Economic Review, 104(2), pp.379-421.
Canales, R. and Nanda, R., 2012. A darker side to decentralized banks: Market power and credit
rationing in SME lending. Journal of Financial Economics, 105(2), pp.353-366.
Chen, B. S., S. G. Hanson, and J. C. Stein, 2017. The decline of big-bank lending to small
business: dynamic impacts on local credit and labor markets. Working paper.
Chetty, R., Friedman, J. N., Hendren N., and Stepner M., 2020. How did COVID-19 and
Stabilization Policies Affect Spending and Employment? A New Real-time Economic
Tracker Based on Private Sector Data. NBER working paper.
Chodorow-Reich, G., 2014. The employment effects of credit market disruptions: firm-level
evidence from the 2008–9 financial crisis. Quarterly Journal of Economics 129, 1–59.
Demirgüç-Kunt, A. and Maksimovic, V., 1998. Law, finance, and firm growth. The Journal of
Finance, 53(6), pp.2107-2137.
Demirgüç-Kunt, A. and Maksimovic, V., 2002. Funding growth in bank-based and market-based
financial systems: evidence from firm-level data. Journal of Financial Economics, 65(3),
pp.337-363.
Ellul, A., I. Erel, and Rajan, U., 2020. The Covid-19 pandemic crisis and corporate finance.
Review of Corporate Finance Studies, 9(3), pp. 421-429.
Fazzari, S.M., Hubbard, R.G. and Petersen, B.C., 1988. Financing Constraints and Corporate
Investment. Brookings Papers on Economic Activity, 1988(1), pp.141-195.
Gilchrist, S. and Zakrajšek, E., 2012. Credit spreads and business cycle fluctuations. American
Economic Review, 102(4), pp.1692-1720.
Giroud, X. and Mueller, H. M., 2016. Firm leverage, consumer demand, and employment losses
during the Great Recession. Quarterly Journal of Economics 132(1), pp. 271-316.
30
Greenstone, M., Mas, A. and Nguyen, H.L., 2020. Do Credit Market Shocks Affect the Real
Economy? Quasi-experimental Evidence from the Great Recession and" Normal"
Economic Times. American Economic Journal: Economic Policy, 12(1), pp.200-225.
Koetter, M., F. Noth, and O. Rehbein. 2020. Borrowers under water! Rare disasters, regional
banks, and recovery lending. Journal of Financial Intermediation 43: 100811.
Kroszner, R. S., Laeven, L., Klingebiel, D., 2007. Banking crises, financial dependence, and
growth. Journal of Financial Economics 84, 187–228.
Levine, R., 1997. Financial Development and Economic Growth: Views and Agenda. Journal of
Economic Literature, 35(2), pp.688-726.
Levine, R., Lin, C. and Xie, W., 2016. Spare tire? Stock markets, banking crises, and economic
recoveries. Journal of Financial Economics, 120(1), pp.81-101.
Levine, R., Lin, C. and Xie, W., 2018. Corporate resilience to banking crises: the roles of trust
and trade credit. Journal of Financial and Quantitative Analysis, 53(4), pp.1441-1477.
Li, L., and P. E. Strahan. 2021. Who Supplies PPP Loans (and Does It Matter)? Banks,
Relationships, and the COVID Crisis. Journal of Financial and Quantitative Analysis 56:
2411-38.
Lin, L., 2020. Bank deposits and the stock market. Review of Financial Studies, 33(6), pp. 2622-
2658.
Myers, S.C. and Majluf, N.S., 1984. Corporate financing and investment decisions when firms
have information that investors do not have. Journal of Financial Economics, 13(2),
pp.187-221.
Nguyen, H.L.Q., 2019. Are credit markets still local? Evidence from bank branch
closings. American Economic Journal: Applied Economics, 11(1), pp.1-32
Reinhart, C. M., 2020. This time truly is different. https://www.project-
syndicate.org/commentary/covid19-crisis-has-no-economic-precedent-by-carmen-
reinhart-2020-03.
Schüwer, U., C. Lambert, and F. Noth. 2018. How do banks react to catastrophic events?
Evidence from hurricane Katrina. Review of Finance 23(1), pp.75-116.
Small Business Administration, 2020. Frequently Asked Questions About Small Businesses,
Office of Advocacy, October.
Smets, F. and Wouters, R., 2007. Shocks and frictions in US business cycles: A Bayesian DSGE
approach. American Economic Review, 97(3), pp.586-606.
31
Figure 1. # COVID-19 cases by county, as of June 2020
This figure plots the cumulative number of confirmed COVID-19 cases per 100,000 population for all U.S. counties
as of June 30, 2020. Darker colors indicate a larger number of confirmed cases per capita.
Figure 2. Share of small bank branches by county
This figure plots the proportion of small bank branches for all U.S. counties as of June 30, 2019. Darker colors
indicate a larger share of small bank branches.
32
Figure 3. Employment in counties with high vs. low small bank share
This figure plots the employment rates over time while dividing counties into groups with high- or low-groups of
small bank shares. High (low) small bank share counties refer to those with the proportion of small bank branches is
above (below) the top (bottom) tercile of the sample value of Share_small bank branches.
33
Figure 4. Employment of low-, middle-, and high-income workers
This figure plots the trends of employment rates of low-, middle-, and high-income workers over time. Low-income
workers are those in the bottom quartile of the income distribution, middle-income workers are those in the middle
two quartiles of the income distribution, and high-income workers are those in the top quartile of the income
distribution.
Figure 5. Employment of low-income workers, in counties with high vs. low small bank
share
This figure plots the employment rates of low-income workers over time while dividing counties into groups with
high- or low-groups of small bank shares. High (low) small bank share counties refer to those with the proportion of
small bank branches is above (below) the top (bottom) tercile of the sample value of Share_small bank branches.
34
Table 1 Summary Statistics
variable N mean sd min p10 p25 p50 p75 p90 max
ln(cases per capita) 33930 2.83 2.51 0 0 0 3.24 4.99 6.08 9.08
cases per capita 33930 164 409 0 0 0 24.5 146 436 8744
Employment 17580 -9.611 10.5 -35.6 -25.1 -17.1 -7.59 -0.231 1.67 8.2
Unemployment Insurance Claims 23454 1.163 1.625 0 0.084 0.192 0.619 1.56 2.95 45.5
Employment, low income workers 10914 -18.26 15.92 -52.2 -40.5 -30.9 -17.7 -2.46 0.818 5.56
Employment, middle income workers 12784 -8.603 10.19 -34 -23.8 -15.6 -6.41 0.174 2.18 10.3
Employment, high income workers 7552 -3.651 6.776 -22.7 -13.7 -7.74 -1.74 1.16 3.19 10.7
Small business revenue 9856 -20.78 21.47 -85.1 -49.6 -37.5 -20.1 -3.54 5.92 41.1
share_small bank branches 1797 0.4 0.325 0 0 0.1 0.333 0.667 0.9 1
share_small bank deposits 1797 0.376 0.344 0 0 0.052 0.267 0.671 0.936 1
#branches per capita 1801 37.6 27.7 0 15.5 21.3 29.5 44.4 67.8 255
Income per capita 1814 45735 13755 19141 33736 38013 43296 50107 58878 251728
%65+ pop 1814 16.9 4.29 6.16 11.9 14.2 16.5 19 22 57.3
Population density 1814 381 2303 0.298 10.5 27.6 69.6 210 595 71616
Unemployment rate 1814 3.86 1.29 1.6 2.6 3 3.6 4.5 5.4 18.3
35
Table 2 local financial structure and economic resilience to Covid
The table reports regression results of how employment rates (Panel A) and unemployment insurance claims (Panel
B) respond to COVID-19 as functions of pre-pandemic share of small banks in local banking markets. The
dependent variable is the employment level relative to January 2020 of each county in a week (Panel A) and the
number of initial claims per 100 people in the 2019 labor force (Panel B). COVID19 is the number of confirmed
COVID-19 cases per 100,000 people in a county by the end of each week. To measure the proportion of small banks
in a county, we use Share_small bank branches or Share_small bank deposits. The analyses cover the period from
February 2020 through June 2020. We include county and state by week fixed effects. Heteroskedasticity-robust
standard errors clustered at the county level are reported in parentheses. ***, **, * denote significance levels at 1%,
5% and 10% respectively.
Panel A: Employment
Employment
(1) (2) (3) (4) (5)
ln(cases per capita) -0.721*** -0.730*** -0.871** -0.717*** -0.879**
(0.184) (0.234) (0.407) (0.235) (0.408)
share_small bank branches*ln(cases per capita) 1.095*** 0.733**
(0.260) (0.292)
share_small bank deposits*ln(cases per capita) 1.057*** 0.729***
(0.235) (0.262)
#branches per capita*ln(cases per capita) -0.008 -0.006 -0.007 -0.006
(0.007) (0.007) (0.007) (0.007)
Income per capita*ln(cases per capita) -0.006* -0.006*
(0.003) (0.003)
%65+ pop*ln(cases per capita) 0.033** 0.033**
(0.014) (0.014)
population density*ln(cases per capita) -0.008 -0.008
(0.008) (0.008)
Unemployment rate*ln(cases per capita) 0.026 0.029
(0.053) (0.052)
County FE Y Y Y Y Y
State-time FE Y Y Y Y Y
Observations 17,558 17,536 17,536 17,536 17,536
Adjusted R-squared 0.864 0.866 0.867 0.866 0.867
# of clusters 799 798 798 798 798
36
Panel B: Unemployment Insurance Claims
Unemployment Insurance Claims
(1) (2) (3) (4) (5)
ln(cases per capita) 0.059*** 0.079*** 0.099** 0.073*** 0.091**
(0.014) (0.017) (0.039) (0.017) (0.038)
share_small bank branches*ln(cases per capita) -0.068*** -0.076***
(0.022) (0.023)
share_small bank deposits*ln(cases per capita) -0.047** -0.053***
(0.019) (0.020)
#branches per capita*ln(cases per capita) 0.0002 0.001** 0.0002 0.0005*
(0.000) (0.000) (0.000) (0.000)
Income per capita*ln(cases per capita) -0.001** -0.001**
(0.000) (0.000)
%65+ pop*ln(cases per capita) -0.002 -0.002
(0.001) (0.001)
population density*ln(cases per capita) 0.003*** 0.003***
(0.001) (0.001)
unemployment rate*ln(cases per capita) 0.009* 0.009*
(0.005) (0.005)
County FE Y Y Y Y Y
State-time FE Y Y Y Y Y
Observations 23,435 23,172 23,172 23,172 23,172
Adjusted R-squared 0.749 0.751 0.751 0.751 0.751
# of clusters 1422 1405 1405 1405 1405
37
Table 3 local financial structure and economic resilience to Covid, employment by income level
The table reports regression results of how employment rates respond to COVID-19 as functions of pre-pandemic share of small banks in local banking markets,
while examining employment rates across different income groups. Employment, low income workers represents the employment level for workers in the bottom
quartile of the income distribution (incomes approximately under $27,000). Employment, middle income workers represents the employment level for workers in
the middle two quartiles of the income distribution (incomes approximately $27,000 to $60,000). Employment, high income workers represents the employment
level for workers in the top quartile of the income distribution (incomes approximately over $60,000). Other variables are the same as before. County traits
include Income per capita, %65+ pop, Population density, and Unemployment rate. We include county and state by week fixed effects. Heteroskedasticity-robust
standard errors clustered at the county level are reported in parentheses. ***, **, * denote significance levels at 1%, 5% and 10% respectively.
Employment, Employment, Employment,
low income workers middle income workers high income workers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
ln(cases per capita) -1.354*** -2.402*** -1.419*** -2.429*** -0.788*** -0.383 -0.762*** -0.388 -0.438 -0.067 -0.436 -0.064
(0.403) (0.608) (0.409) (0.609) (0.252) (0.504) (0.250) (0.504) (0.285) (0.488) (0.284) (0.489)
share_small bank branches*
2.022*** 1.427** 0.816** 0.029 0.403 -0.269
ln(cases per capita)
(0.630) (0.693) (0.375) (0.404) (0.453) (0.523)
share_small bank deposits*
1.924*** 1.329** 1.022*** 0.289 0.584 -0.111
ln(cases per capita)
(0.598) (0.662) (0.394) (0.421) (0.440) (0.500)
#branches per capita*
0.010 0.019 0.014 0.023* -0.005 0.001 -0.006 -0.001 -0.019* -0.014 -0.019* -0.016
ln(cases per capita)
(0.012) (0.015) (0.012) (0.014) (0.008) (0.008) (0.007) (0.008) (0.010) (0.012) (0.010) (0.011)
County FE Y Y Y Y Y Y Y Y Y Y Y Y
State-time FE Y Y Y Y Y Y Y Y Y Y Y Y
County traits*ln(cases per capita) N Y N Y N Y N Y N Y N Y
Observations 10,826 10,826 10,826 10,826 12,674 12,674 12,674 12,674 7,442 7,442 7,442 7,442
Adjusted R-squared 0.905 0.906 0.904 0.906 0.877 0.882 0.877 0.882 0.854 0.862 0.854 0.862
# of clusters 493 493 493 493 577 577 577 577 339 339 339 339
38
Table 4 Robustness tests
The table reports regression results of how employment rates respond to COVID-19 as functions of pre-pandemic share of small banks in local banking markets,
while controlling for additional country traits. Additional county traits include bank concentration (HHI) and county-aggregated bank characteristics (liquid
assets, equity ratio, tier 1 ratio, non-performing loans, and ROA). County traits include Income per capita, %65+ pop, Population density, and Unemployment
rate. Other variables are defined the same as before. We include county and state by week fixed effects. Heteroskedasticity-robust standard errors clustered at the
county level are reported in parentheses. ***, **, * denote significance levels at 1%, 5% and 10% respectively.
Employment, Employment, Employment,
low income workers middle income workers high income workers
(1) (2) (3) (4) (5) (6)
ln(cases per capita) -3.290*** -3.406*** -2.062*** -2.076*** -0.100 -0.080
(1.073) (1.070) (0.797) (0.797) (0.794) (0.795)
share_small bank branches*ln(cases per capita) 1.620** -0.141 -0.528
(0.709) (0.408) (0.546)
share_small bank deposits*ln(cases per capita) 1.599** 0.202 -0.461
(0.688) (0.429) (0.541)
County traits*ln(cases per capita) Y Y Y Y Y Y
Additional County traits*ln(cases per capita) Y Y Y Y Y Y
County FE Y Y Y Y Y Y
State-time FE Y Y Y Y Y Y
Observations 10,826 10,826 12,674 12,674 7,442 7,442
Adjusted R-squared 0.907 0.907 0.883 0.883 0.862 0.862
# of clusters 493 493 577 577 339 339
39
Table 5 local financial structure and small business revenue during Covid-19
The table reports regression results of how small business revenues respond to COVID-19 as functions of pre-
pandemic share of small banks in local banking markets. Small business revenue is the percent change in seasonally-
adjusted net revenue for small businesses relative to January 4-31 2020. Other variables are the same as before.
County traits include Income per capita, %65+ pop, Population density, and Unemployment rate. We include
county and state by week fixed effects. Heteroskedasticity-robust standard errors clustered at the county level are
reported in parentheses. ***, **, * denote significance levels at 1%, 5% and 10% respectively.
Small business revenue
(1) (2) (3) (4) (5)
ln(cases per capita) -2.150*** -1.775*** -1.277* -1.660*** -1.271*
(0.493) (0.564) (0.768) (0.569) (0.765)
share_small bank branches*ln(cases per capita) 4.267*** 1.990**
(0.772) (0.814)
share_small bank deposits*ln(cases per capita) 4.091*** 1.845**
(0.874) (0.861)
#branches per capita*ln(cases per capita) -0.038** 0.003 -0.034** 0.006
(0.016) (0.018) (0.016) (0.018)
County FE Y Y Y Y Y
State-time FE Y Y Y Y Y
County traits*ln(cases per capita) - N Y N Y
Observations 9,724 9,724 9,724 9,724 9,724
Adjusted R-squared 0.847 0.851 0.856 0.850 0.855
# of clusters 443 443 443 443 443
40
Table 6 local financial structure and economic resilience to Covid-19, IV
This table reports the 2SLS regression results of how employment rates and small business revenues respond to COVID-19 as functions of pre-pandemic share of
small banks in local banking markets. Columns with the odd number report the second-stage estimation results. The dependent variables in column 1, 3, 5, 7, and
9 are Employment, Employment, low-income workers, Employment, middle-income workers, Employment, high-income workers, and Small business revenue,
respectively. Columns with the even number report the first-stage result, so the dependent variable is the endogenous variable, Share_small bank
branches*Ln(cases per capita). The excluded instrument is #Mergers by out-of-state banks*Ln(cases per capita). We include county and state by week fixed
effects. Heteroskedasticity-robust standard errors clustered at the county level are reported in parentheses. ***, **, * denote significance levels at 1%, 5% and
10% respectively.
Employment,
Employment, Employment, Small
First- First- middle First- First- First-
Employment low income high income business
stage stage income stage stage stage
workers workers revenue
workers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
ln(cases per capita) -0.726*** 0.024 -1.047** -0.099*** -0.652** -0.019 -0.399 -0.038 -1.659*** -0.019
(0.235) (0.029) (0.461) (0.027) (0.285) (0.033) (0.287) (0.028) (0.576) (0.032)
share_small bank branches*
1.926* 4.308** 4.031** 1.074 7.181**
ln(cases per capita)
(1.008) (1.852) (1.818) (2.884) (3.201)
#branches per capita*
-0.014 0.007*** -0.010 0.009*** -0.027* 0.007*** -0.022 0.005*** -0.056** 0.006***
ln(cases per capita)
(0.009) (0.001) (0.019) (0.001) (0.014) (0.001) (0.014) (0.001) (0.024) (0.001)
#Mergers by out-of-state
-0.077*** -0.061*** -0.057*** -0.036*** -0.051***
banks*ln(cases per capita)
(0.011) (0.010) (0.010) (0.010) (0.010)
County FE Y Y Y Y Y Y Y Y Y Y
State-time FE Y Y Y Y Y Y Y Y Y Y
Observations 17,536 17,536 10,826 10,826 12,674 12,674 7,442 7,442 9,724 9,724
Weak_ID_FTest 51.52 40.45 32.31 11.82 27.51
# of clusters 798 493 577 339 443
41
Table 7 Small business lending during Covid-19, Large vs. Small bank
This table reports the regression results analyzing how small business lending responds to COVID-19 as functions of bank size before the pandemic. The
dependent variable is the growth of small business loans, the growth of C&I loans, and the amount of PPP loans scaled by total assets between December 2019
and June 2020 for each bank. Bank exposure is the weighted average of Ln(Cases per Capita) across counties in which a bank has branches, the weight is the
share of deposits held by the bank in a given county. We measure bank size using either the log amount of total assets (ln(Assets)), or a dummy variable equal to
one if a bank’s total assets is below $1 billion and zero otherwise (Small bank). Bank Char. include ln(Assets), single-market indicator, liquid assets, equity ratio,
tier 1 ratio, NPL, and ROA. Heteroskedasticity-robust standard errors are reported in the parentheses. “*” indicates statistical significance at 10% level, “**” at
5% level, and “***” at 1% level.
Small business loan growth C&I loan growth PPP loans
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Bank exposure 0.040*** 0.251*** -0.018 0.024*** 0.332*** -0.045** 0.005*** 0.037*** -0.000
(0.008) (0.070) (0.023) (0.007) (0.056) (0.018) (0.001) (0.006) (0.002)
ln(Assets) 0.048*** 0.155*** 0.063*** -0.018*** 0.138*** 0.013 0.002* 0.018*** 0.004***
(0.009) (0.034) (0.011) (0.007) (0.027) (0.009) (0.001) (0.003) (0.001)
ln(Assets)*Bank exposure -0.017*** -0.025*** -0.003***
(0.006) (0.004) (0.001)
Small bank -0.316** -0.309*** -0.024*
(0.143) (0.112) (0.014)
Small bank*Bank exposure 0.063*** 0.076*** 0.005**
(0.024) (0.019) (0.002)
Single market -0.010 -0.003 -0.006 -0.040** -0.030 -0.031 -0.001 -0.000 -0.001
(0.022) (0.022) (0.022) (0.019) (0.019) (0.019) (0.002) (0.002) (0.002)
Equity -0.826* -0.829* -0.798 -0.931** -0.937** -0.895** -0.014 -0.014 -0.011
(0.490) (0.492) (0.492) (0.421) (0.425) (0.419) (0.057) (0.057) (0.057)
ROA -1.032 -1.262 -1.281 0.661 0.403 0.208 -0.192 -0.219 -0.220
(2.098) (2.110) (2.100) (1.755) (1.763) (1.745) (0.215) (0.215) (0.215)
Tier 1 capital 0.214 0.211 0.217 0.036 0.033 0.052 -0.026 -0.026 -0.025
(0.242) (0.243) (0.242) (0.163) (0.164) (0.162) (0.021) (0.021) (0.021)
Liquid assets -0.152* -0.147 -0.149* -0.136** -0.128* -0.133* -0.079*** -0.078*** -0.079***
(0.090) (0.090) (0.090) (0.069) (0.069) (0.069) (0.007) (0.007) (0.007)
NPL -6.135*** -6.121*** -6.082*** -7.085*** -7.062*** -6.995*** -0.760*** -0.757*** -0.754***
(1.168) (1.169) (1.165) (0.988) (0.986) (0.980) (0.120) (0.120) (0.120)
Headquarter State FE Y Y Y Y Y Y Y Y Y
Observations 4,968 4,968 4,968 5,039 5,039 5,039 5,039 5,039 5,039
Adjusted R-squared 0.113 0.114 0.114 0.087 0.092 0.094 0.142 0.145 0.143
42
Table 8 Local financial structure and employment resilience to Covid, differentiate by firm
size
The table reports regression results of how employment respond to COVID-19 as functions of pre-pandemic share
of small banks in local banking markets. The unit of analysis is at the county-industry-quarter-type level, there type
represents whether firm size falls into [0,49], [50,499], or 500 and above. The analyses cover the period from 2019
Q1 through 2020 Q2. The dependent variable is the employment level relative to 2018 of each county. Ln(cases per
capita) is the log of the number of confirmed COVID-19 cases per 100,000 people in a county by the end of each
quarter. To measure the proportion of small banks in a county, we use Share_small bank branches or Share_small
bank deposits. Basic county control includes the interaction of Ln(cases per capita) and Income per capita, %65+
pop, Population density, and Unemployment rate. Additional control includes the interaction of Ln(cases per capita)
and bank concentration and county-aggregated bank characteristics (liquid assets, equity ratio, tier 1 ratio, non-
performing loans, and ROA). Heteroskedasticity-robust standard errors clustered at the county level are reported in
parentheses. ***, **, * denote significance levels at 1%, 5% and 10% respectively.
Employment
(1) (2) (3) (4) (5)
ln(cases per capita) -0.000814 -0.00240 6.06e-05 -0.00240 5.11e-06
(0.00108) (0.00177) (0.00252) (0.00176) (0.00250)
ln(cases per capita)*Firm size<50 -0.00609*** -0.00809*** -0.00812*** -0.00785*** -0.00788***
(0.00126) (0.00133) (0.00134) (0.00132) (0.00132)
ln(cases per capita)*Firm size∈[50,499] -0.00238 -0.00116 -0.00116 -0.00125 -0.00125
(0.00178) (0.00186) (0.00186) (0.00183) (0.00183)
share_small bank branches*ln(cases per capita) 0.000851 0.00104
(0.00125) (0.00131)
share_small bank branches*ln(cases per capita)*Firm size<50 0.00388*** 0.00385***
(0.00120) (0.00120)
share_small bank branches*ln(cases per capita)*Firm
size∈[50,499] -0.00204 -0.00207
(0.00184) (0.00184)
share_small bank deposits*ln(cases per capita) 0.00108 0.00141
(0.00119) (0.00126)
share_small bank deposits*ln(cases per capita)*Firm size<50 0.00346*** 0.00342***
(0.00116) (0.00116)
share_small bank deposits*ln(cases per capita)*Firm
size∈[50,499] -0.00199 -0.00202
(0.00178) (0.00178)
Basic county control N Y Y Y Y
Additional control N N Y N Y
County-Industry-Type FE Y Y Y Y Y
State-Time FE Y Y Y Y Y
Industry-Time FE Y Y Y Y Y
Type-Time FE Y Y Y Y Y
Observations 502,117 490,710 490,424 490,710 490,424
Adjusted R-squared 0.702 0.703 0.703 0.703 0.703
# of clusters 2950 2860 2856 2860 2856
File and source
- File
- levine-lin-xie-2021-06-local-financial-structure-economic-resilience-worldbank.pdf
- Size
- 755,524 bytes
- SHA-256
- fe86b49bb74edc13fd0737443ab6a53e797e6e4ca97cefd488b8f1be1b7db0ca
- Original
- No public link identified.