Banking infrastructure and the Paycheck Protection Program during the Covid-19 pandemic

ABSTRACT In response to the Covid-19 pandemic, the US federal government distributed US$800 billion in Paycheck Protection Program (PPP) loans to small businesses to preserve employment. Since PPP funding was transmitted through private banks, the characteristics of the regional banking market may have unevenly affected the programme’s reach. This paper examines how variations in market concentration and the presence of community banks contributed to PPP disbursement in US counties. It finds that greater regional banking market concentration correlates with fewer PPP loans, but this negative relationship is mitigated by a greater presence of community banks in highly concentrated markets.


INTRODUCTION
This paper examines how regional banking sector characteristics affected the disbursement of federally guaranteed loans to small businesses, the Paycheck Protection Program (PPP), during the Covid-19 pandemic. The programme is 'the most ambitious and creative fiscal policy response to the Pandemic Recession' (Hubbard & Strain, 2020, p. 1), designed to preserve employment in response to stay-in-shelter orders in the United States. One of its most creative features is that the federal government used the existing private banking infrastructure to distribute over US$800 billion to millions of small businesses in a short period.
Researchers consider the financial subsystem as one of the important determinants of how well regions withstand and recover from an economic shock (Martin & Sunley, 2020). They conjecture that a financial sector promotes regional economic resilience by providing necessary credit for households and businesses to ease a cash crunch. Nonetheless, empirical evidence for the role of the financial sector in economic resilience has been rare in regional studies. The finance literature, however, has discovered that characteristics in regional banking markets contribute to small businesses' loan access during economic recessions. This paper uses findings from the finance literature and tests how PPP's distributional outcomes were determined by multidimensional characteristics of the regional banking sector. This paper takes advantage of the unique opportunity offered by the Covid-19-driven economic shock to study the role of financial sectors. Unlike endogenous economic crises, the pandemic drove unexpected and sudden disruptions to small business operations, while structural changes to the banking sector were kept minimal. Furthermore, while exogenous shocks driven by natural disasters are geographically limited, the impacts of the pandemic were global, which allows one to observe the impacts of the shock in all regions.
Using 3100 US counties, significant impacts of banking market concentration and the presence of community banks in disbursing PPP loans are found. Specifically, higher market concentration correlated with fewer PPP loans per business, but a strong presence of community banks mitigated the negative association. In highly concentrated markets, in particular, a greater presence of community banks even increased the number of PPP loans.
This paper contributes to the literature as follows. First, it contributes to the emerging, but limited, literature on the role of the regional financial subsystem in building economic resilience during and after economic crises. Second, it contributes to the recent finance literature that considers both market concentration and the types of players in the market. Finally, the paper contributes to a growing literature on PPP where the structural environment of a regional banking sector has been overlooked. This paper directly deals with the structural factors using traditional measures of pre-pandemic characteristics of a regional banking sector to make it comparable with previous studies.

Community banks and relationship lending
The discussion on banks' behaviour in small business lending centres on types of lenders: community banks and national banks. In the literature, community banks are often used interchangeably with small banks or local banks. Asset size often defines community banks, but the asset threshold varies in the literature from US$1 billion to US$10 billion at the charter level. Local banks are commonly defined as banks owned or operating within a limited geographical area or by the share of deposits that stay in the local market (Cortés, 2014). To avoid varying definitions and identify community banks more systematically, the Federal Deposit Insurance Corporation (FDIC) provides several standards to defines community banks using indexed asset thresholds, geographical footprint, business plan and number of branches (FDIC, 2020). At the minimum, community banks are 'small' in asset size and 'local' in the geographical scope of business operation. 1 Community banks are deemed to play a vital role in regional economies. They locally acquire deposits and devote a large share of their resources to local businesses (Rogers, 2012;Strahan, 2008). The theoretical underpinning between community banks and small business loans is relationship-lending. The conventional theory posits that big national banks use quantifiable, verifiable and comparable 'hard information' to assess loan risks. Thus, these 'transaction lenders' tend to make loans to already-established firms with externally audited financial statements. Small businesses, however, are often 'opaque' in that they lack an audited financial statement. This is where community banks come into play. They exploit 'soft' information such as the bank's intimate knowledge about business owners' reputations as well as local economic conditions. Evidence shows that those 'relationship lenders' tend to lend to geographically close small businesses, suggesting that spatial proximity facilitates the transmission of soft information (Cotugno et al., 2013;Hakenes et al., 2015;Granja et al., 2018).
Nevertheless, it is the big national banks (assets of more than US$10 billion) that made 58% of the US$645 billion small business loans in 2019 (US Small Business Administration (SBA), 2020b). They actively pursue small business clients using various lending technologies and risk management systems to compensate for lacking hard information (Berger & Black, 2011;Berger et al., 2014;De la Torre et al., 2010). Thus, banks' characteristics alone do not fully explain their lending behaviour because the finance literature shows that the overall banking market environment also shapes their behaviour. For instance, the same community bank may behave differently when they face competition and when they have market power.

Market concentration
The concentration of market power has received considerable attention because the US banking market has been concentrated in a smaller number of big national banks over the years. The FDIC reported a 68% decline in the number of commercial banks between 1986 and 2019 (Brown, 2019). Assets have become concentrated as well. In 2020, the 12 largest banks held 60% of all domestic assets. Regional markets follow the same pattern: 78% of all regional banking markets were highly concentrated in 2017 (Meyer, 2018).
Policymakers are concerned with how the structural changes impact small businesses' loan access. The traditional view posits that market concentration creates an unfavourable environment for small businesses where banks with fewer competitors charge higher interest rates and make fewer loans to small businesses (Berger & Hannan, 1989;Berger et al., 2004;Cetorelli & Strahan, 2004;Hannan & Berger, 1991). This is particularly concerning because, unlike their large counterparts, small businesses heavily rely on commercial loans with no access to the capital market.
Challenging this traditional view, the 'investment theory', however, argues that market concentration benefits small businesses. It suggests that market power enables banks to invest their resources in long-term relationships with small businesses, whereas slim margins from a competitive environment disincentivize banks to develop such relationships (Francis et al., 2008). Furthermore, the investment theory postulates that banks with market power can use economies of scale to diversify risks that may arise from small business loans.
Both views assume a linear relationship between market power and small business loan supply. Recent studies, however, show that the market structure does not necessarily determine banks' lending behaviour. For instance, banks that came to hold bigger market power by acquiring another bank do not always reduce small business loans. When an acquirer bank specializes in small business lending before a merger, it increased, rather than decreased, small business loans even after the merger (Avery & Samolyk, 2004;Peek & Rosengren, 1998;Strahan & Weston, 1998). A similar pattern has been found in Italy (Presbitero & Zazzaro, 2011), Germany (Elsas, 2005) and Mexico (Canales & Nanda, 2012).

Credit access in times of crises
Martin and Sunley (2020) suggest regional subsystems that determine economic resilience: business, finance, governance and the labour market. During economic crises, the impact of economic shocks on businesses depends on the regional financial subsystem. Nonetheless, studies have not directly addressed the role of financial subsystems in times of crisis, but two strands of research may guide one to understand the link. Banks' ability to supply small business loans diminishes during economic downturns because economic contraction not only stresses the financial conditions of businesses but also banks' financial health. According to this view, 'the impairment of the bank-credit channel', compared with the unaffected, banks financially damaged by economic shocks significantly reduced small business loan supply and increased loan interest rates in the post-crisis period (Adams & Amel, 2005;Chava & Purnanandam, 2011).
Nonetheless, the loan accessibility of businesses depends on a couple of other factors. First, banks favour their relationship borrowers. Bolton et al. (2016) propose a theoretical model to predict the small business loan supply during an economic downturn based on the behavioural differences between relationship-and transactionlenders. They suggest that firms with low cash are prepared to take higher interest rates in normal times, expecting a continued relationship with their bank during a credit crunch. In a recession, relationship lenders offer better rates for their long-time clients to prevent them from defaulting for their own interest (Cotugno et al., 2013;Sette & Gobbi, 2015).
In addition to a bank-firm relationship, the financial literature also highlights the impact of a larger structural market environment on crisis lending. Evidence shows that greater market concentration leads to diminished credit flow with a higher interest rate to small businesses after the 2008 economic recession (Chen et al., 2017). However, Cubillas and Suárez (2018) counterargue that banks with market power may increase, not decrease, credit supply with a higher interest rate (Degryse et al., 2018;Hasan et al., 2019;Zhao & Jones-Evans, 2017). Cash-stripped borrowers that had a relationship with a failed bank need to find an alternative lender quickly during an economic downturn. Thus, survived banks with increased market power supply cash with a higher interest rate for desperate non-relationship borrowers. The mixed evidence in the literature may stem from different behavioural assumptions and institutional settings in different countries. To this date, no clear consensus has emerged.
Another line of research on small business crisis lending comes from the natural disaster literature. Unlike endogenous economic crises, natural disasters are exogenous external shocks that less likely affect banks' financial health and more likely damage local small businesses. Most studies find that geographically close local banks increase loans to local businesses after natural disasters, which in turn leads to better economic outcomes in the region (Cortés, 2014;Ivanov et al., 2020;Koetter et al., 2020). In addition, prior relationships between lenders and borrowers help ease the lending restrictions after a natural disaster (Berg & Schrader, 2012).
In summary, previous studies help one to understand the behaviours of lenders and borrowers during endogenous economic downturns and after local disasters. Nevertheless, how those studies can help predict outcomes of PPP is unclear because of the unique characteristics of government-backed PPP 'loans' and the nature of the Covid-19 pandemic that is different from previous shocks. The details of PPP design are important to understand in predicting how regional banking infrastructure contributed to the disbursement of PPP funds.

THE PAYCHECK PROTECTION PROGRAM (PPP) DURING THE COVID-19 PANDEMIC
On 13 March 2020, a national emergency was declared concerning the spread of Covid-19. Aggressive policy actions ensued. On 19 March, the State of California issued a shelter-in-place order to preserve public health and safety from Covid-19. By 6 April, 42 states and Washington, DC ordered non-essential businesses to close temporarily. Within a week after the national emergency declaration, the number of initial unemployment insurance benefits claims jumped from 251,416 to 2.9 million. It further increased to 3 million and 6 million claims in the subsequent weeks (US Department of Labour). 2 Congress responded to the negative economic impacts by enacting the Coronavirus Aid, Relief, and Economic Security (CARES) Act. The Act included PPP designed to prevent large-scale unemployment by aiding small businesses with payroll, rent or other direct operating costs. 3 The first draw of the programme began on 3 April and ended on 8 August 2020. 4 The second draw was implemented in 2021 between January and May. In total, PPP made 12 million loans and distributed US $800 billion to small businesses through 5467 financial institutions (SBA, 2021).
Important design features of PPP make PPP 'loans' distinctive from commercial loans. PPP 'loans' subsidize operating revenue losses of small businesses during the pandemic. The loan amount is based on pre-pandemic payroll costs. The interest rate is fixed at 1% with no collateral, personal guarantee or credit score requirements. The loans are forgivable if employees and wages are maintained, loans are spent on payroll costs and eligible expenses, and at least 60% of the amount is spent on payroll costs. These features of PPP eliminate firm-level heterogeneity in capital structure and financial health that determine the underwriting conditions of commercial loans. The PPP 'loans' can be characterized as conditional federal grants.
One of the innovative features of PPP is the use of existing private banking infrastructure in distributing PPP funds. All existing SBA-certified lenders were eligible to delegate authority 'to speedily process PPP loans', according to the US Department of Treasury. 5 Lenders charge fees from 1% to 5% of the loan amount depending on the loan size. Since all loans are guaranteed by the SBA, lenders are incentivized to participate in PPP to make fee revenues with minimal loan risks.
A growing number of microlevel studies finds that an existing bank-firm relationship significantly increased the likelihood of PPP loan approvals because banks helped alleviate their clients' insufficient information about the 86 Soomi Lee programme (Amiram & Rabetti, 2020;Bartik et al., 2020;Granja et al., 2020;Humphries et al., 2020;James et al., 2020;Li & Strahan, 2020). The evidence is consistent with Bolton et al. (2016) in which banks prioritize their relationship clients during an economic downturn because they have an economic interest in the long-term survival of their borrowers. A few studies in the early stage of PPP implementation have suggested that PPP was ineffective in preserving employment (Autor et al., 2020;Bartik et al., 2020;Chetty et al., 2020;Granja et al., 2020). Yet these studies use surveys or administrative data from private payroll-processing firms with questionable sample representativeness. More recent evidence shows that PPP helped small businesses (Bartik et al., 2020;Cororaton & Rosen, 2021;Hubbard & Strain, 2020) and regional economies (Barrios et al., 2020;Doniger & Kay, 2021;Faulkender et al., 2020;James et al., 2020;Li & Strahan, 2020;Mitchell, 2020).
Nevertheless, the impact of a larger structural environment of the regional banking sector on the distribution of PPP loans has not been well understood. Existing studies mainly focus on microlevel analyses using loan-level data and overlook the structural factor in region-level analyses. The finance literature demonstrates that the market structure is regionally heterogeneous, and it shapes banks' lending behaviour to small businesses. Furthermore, the recent literature suggests that it is not only the market structure but also the types of players in the market that determines small business credit access. Thus, how federal emergency grants channel through the regional banking infrastructure to ease small businesses' cash flow would depend on multidimensional characteristics of the regional banking sector. No prior studies have directly this important aspect examined. To fill the gap, the paper provides county-level analyses to examine how the regional banking market's structural characteristics contribute to the programme's reach.
Although PPP 'loans' differ from commercial loans in extraordinary circumstances, we may predict how market characteristics affect PPP loan disbursement from the previous literature. First, market concentration may be unfavourable to small businesses. When banks enjoy market power, they might prefer clients with a bigger loan to extract bigger fees for themselves since the PPP loan amount is predetermined by pre-Covid operating expenses. On the contrary, in a competitive market, banks pursue clients to earn more fee revenues with minimal loan risks, which would result in a larger number of PPP loans.
Second, community banks have a stronger incentive to actively participate in PPP. For big banks, PPP fees are only a small fraction of their overall revenues, whereas for smaller banks, the fees can be a substantial fraction of their overall revenues. In the early stage of the implementation of PPP, National Public Radio reported a Maryland-based small community bank making one year's worth of loans in just 10 days by participating in PPP. Studies on crisis-lending to small businesses point out that during an economic crisis or a natural disaster, small and local community banks redirect resources to help local and small businesses (Degryse et al., 2018;Hasan et al., 2019;Zhao & Jones-Evans, 2017). Thus, we would expect that a larger presence of community banks would increase the number of PPP loans in the region.

DATA AND METHOD
Cross-sectional data consisting of 3100 US counties in all 50 states and the District of Columbia are used. The data do not include 149 counties with no full-service bank office in 2020. The data were culled from several sources. One of the major sources of the data is the PPP Loan Level Data published by the SBA. The 5.2 million loan information from the first draw of the programme from April and August 2020 is used. The second draw distributed in 2021 was excluded because the main interest of the paper is the immediate PPP's reach to small businesses in response to the pandemic. The number of approved loans by county using businesses' zip code information provided by the SBA is aggregated.
The key variables that capture dimensions of regional banking market characteristics were constructed by using the Summary of Deposit (30 March 2020) and the Community Banking Study Reference Data (2020), both of which came from the FDIC. Counties' economic and industry characteristics were collected from the Bureau of Economic Analysis (BEA) and the US Bureau of Labour Statistics (US Census). 6 Demographic variables were from the US Census; Covid-19-related statistics were from The New York Times (2020). For detailed data descriptions, sources and summary statistics, see Appendix A in the supplemental data online.

Dependent variable
The dependent variable is the county's number of approved PPP loans per 100 businesses between April and August 2020. The number of loans instead of the amount of loan was used for the following reason. The amount of PPP loan is based on the pre-Covid 12month average payroll costs with the 1% fixed interest rate. In the PPP context, where speedy programme reach is the goal, loan count is more important than the predetermined loan amount.
The PPP loan data were obtained from SBA's PPP Loan Level Data. The released files include 5.2 million loan information, but do not provide borrowers' county location. It was identified by using the reported zip code. The crosswalk file from the Housing and Urban Development was used to match zip code and county location. Unmatched observations due to entry errors were manually searched and entered in the data using other identifying information of the business address. The number of loans was then aggregated by county and divided by the number of business establishments. The business establishment data were collected from the Quarterly Census of Employment and Wages (QCED) in the first quarter Banking infrastructure and the Paycheck Protection Program during the  in 2020, the closest period to the national emergency declaration in March 2020.
Key independent variables: the structure of the regional banking sector The first key independent variable is banking market concentration, which is measured by the Herfindahl-Hirschman index (HHI). The formula of the market concentration for county i is defined as follows: where j indicates the bank; k is the number of banks in county i; and S is the market share of j's deposit in county i's total deposit. The index was computed using the statement of deposit (SOD) in June 2020 from the FDIC. The maximum value is 100 if one bank controls 100% of the deposit in i. The second key independent variable is the presence of community banks (CBratio). The traditional HHI measure treats all banks equally (Berger et al., 2004), but the nature of market competition may differ depending on the types of players in the market, for example, big national banks versus community banks. Thus, the level of community bank presence is included to capture different dynamics by taking the ratio between the number of community bank branches to all branches.
When defining community banks, the FDIC's community bank designation made by asset size, geographical footprint, business plan and the number of branches was followed (FDIC, 2020). The FDIC excludes banks with no loans or no core deposits, foreign assets more than 10% of total assets and more than 50% of assets in specialty banks (e.g., credit card specialists). The assent threshold in 2019 was US$1.65 billion. Banks with total assets greater than the threshold also can be designated as a community bank depending on financial ratios, the number of branches, the geographical scope of business operation and other criteria. In 2019, the FDIC identified 4750 community banks that account for 91.8% of the total bank organizations.
As a next step, the total number of all bank branches in each county for the first quarter of 2020 was counted using SOD. Community bank branches were then identified and the ratio of the number of FDIC-designated community banks' office branches to the number of all bank branches in the market was taken. The ratio captures the fraction of branches operated by community banks in each county. The values range from 0 to 1. Higher values indicate a greater presence of community banks.
Note that the FDIC measures the presence of community banks in two ways. One is based on office branches and the other on deposits (FDIC, 2020). The former was used because banks' available deposits do not determine PPP loan amounts since PPP is SBA backed. Further, it took only a few weeks to make 5.7 million loans during the first draw of the programme in 2020 (SBA, 2020a), indicating the importance of the available branch offices channelling the federal government funding to local businesses. Thus, the branch-based measure is more appropriate to capture the presence of community banks in the context of PPP.
The third characteristic of a regional banking sector is the number of full-service bank branches per business. Market concentration (HHI) and community bank presence (CBratio) characterize the structure of the county's banking market. Yet, those indicators do not capture the density of bank brancheshow many bank branches are accessible for small businesses. The total number of fullservice bank branches per 1000 businesses was used in the empirical model, regardless of community bank status.

Control variables
Several control variables are included such as per capita income and total population. Since the economic damage was induced by the Covid-19 pandemic, the number of cumulative confirmed Covid-19 cases by the end of March 2020 was controlled for. Since the economic impacts differed by type of business, it is necessary to control for counties' industry structure. Industries are categorized by the North American Industry Classification System (NAICS). The percentages of jobs in the goodsproducing industry (manufacturing, mining and construction), leisure and hospitality, and trade (retail and wholesale) were controlled for. All continuous variables are logtransformed. An indicator variable for metropolitan areas and state dummy variables are included in all models.

Empirical model
The main estimator is the least absolute deviation regression model. While the standard ordinary least squares (OLS) regression model minimizes the sum of squared errors and estimate conditional mean functions, the quantile regression (QR) model asymmetrically weights absolute residuals to estimate conditional median functions. It also estimates a full range of other conditional quantile functions without strict parametric assumptions and allows more robust, efficient and accurate estimates than OLS (Koenker & Bassett, 1978). The quantile regression model is as follows: where y i is the number of PPP loans in county i; a q is an intercept; x ′ i l q is the vectors of control variables and their coefficients at the q th quantile; 51 j=1 S j is state fixed effects; and e i is the error term. The key variables of interest are (1) the degrees of banking market concentration, (HHI i ); (2) the degrees of community bank's presence (CBratio i ); and (3) their interaction term (HHI i × CBratio i ). The current literature is ambiguous that market concentration could either decrease or increase small business lending, and therefore b 1 , 0 or b 1 . 0.
Community banks are considered to increase small business lending especially in disaster cases, b 2 > 0. The impact of HHI and CBratio depends on the value of the other variable. Therefore, the impact of HHI i is b 1 + b 3 (CBratio i ) and the impact of CBratio i is b 2 + b 3 (HHI i ). Table 1 presents the results. The first two columns show OLS estimates. The QR estimates are presented in columns (3) to (7) at 0.1, 0.25, 0.5, 0.75 and 0.9 percentiles. QR at the median (q ¼ 0.5) serves as the baseline, which is referred to here as an LAD model.

RESULTS
Among the three characteristics of the regional banking market, the density of bank branch offices (Branch Density) is statistically significant with similar magnitudes in all specifications and at all quantiles. The coefficients hover around 1, which indicates that a 1% increase in branch density leads to a one PPP loan increase per 100 businesses.
Column (1) does not include the interaction term between HHI and CBratio. Consistent with the traditional view of market concentration, the negative and statistically significant coefficient of the HHI index indicates that greater banking market concentration is correlated with fewer PPP loans per 100 businesses. The coefficient of CBratio is positive, as expected, but is not statistically significant.
When the interaction term between HHI and CBratio is specified in column (2), the coefficient of CBratio becomes negative and statistically significant, while the coefficient of HHI remains negative and statistically significant with doubled magnitude. The negative coefficient of CBratio is the opposite of the theoretical expectation, but the interpretation should account for the positive and significant interaction term between HHI and CBratio. This pattern, significant effects of HHI, CBratio and their interaction term, remain in almost all quantiles. The only exception is the coefficient of CBratio at q ¼ 0.1, which is significant at the 10% significance level.
Considerable differences in the size of coefficients are observed between OLS estimates (column 2) and the LAD estimates (column 5). The OLS estimates seem inflated compared with the LAD estimates. The coefficient is 1.5 times larger for HHI than the LAD estimate, more than two times larger for CBratio and the interaction term.
In QR models, the magnitude of the coefficient for HHI is similar to all quantiles (around 2.0-2.5), although at the highest percentile (q ¼ 0.9), the effect of HHI is substantially larger (around 3.5). For CBratio, the coefficients vary across quantiles. Its magnitude is smaller at lower quantiles and larger at higher percentiles. It ranges from −5.953 at q ¼ 0.1 to −21.878 at q ¼ 0.9. The interaction effect between HHI and CBratio becomes substantially larger at higher quantiles.
The marginal effects are computed based on the LAD model (q ¼ 0.5). The marginal effect of HHI: It ranges from −2.038 to −0.251 since the CBratio runs from 0 to 0.693. It suggests that the negative effect of market concentration deteriorates with a greater presence of community banks. The interaction effect is presented in Figure 1. The x-axis indicates the degrees of banking market concentration; the y-axis is the predicted number of PPP loans per 100 businesses. The three lines show the effect of HHI on the number of PPP loans at three different levels of CBratio: smaller (CBratio q ¼ 0.25), median (CBratio q ¼ 0.5) and larger (CBratio q ¼ 0.75) share of community bank branches.
The three downward slopes indicate that greater market concentration is associated with fewer PPP loans, supporting the traditional view. Nevertheless, the negative effect of HHI is more pronounced at the lower CBratio (Q1, long-dashed line). At the median level of the community bank ratio (Q2, solid line), the slope is less steep than the long-dashed line. At the third quartile of the community bank ratio (Q3, short-dashed line), the slope is flatter than other slopes. In sum, the findings support the traditional view that greater market concentration decreases small business loans, but the negative effect is diminished by a greater presence of community banks in the county. Figure 2 shows how the effect of market concentration (HHI) changes at different values of community bank presence (CBratio). The y-axis indicates the marginal effect of HHI on the number of PPP loans. At the minimum value of CBratio (no community bank presence), a 1% increase in HHI decreases about two loans per 100 businesses. As the community bank ratios increase, the negative effect becomes smaller, as the line moves upward.
At a CBratio > 0.55, the coefficient of HHI is no longer statistically significant. It translates that when community bank branches make up more than 73% of all bank branches in county i, the degrees of market concentration have no impact on the number of PPP loans. Those counties consist of slightly more than half of all counties in the sample.
The range runs from −3.483 to 3.941 when the interaction with HHI is accounted for. Figure 3 demonstrates the average marginal effects of CBratio at different levels of HHI.
The effects of CBratio depend on the degrees of market concentration. In competitive markets, the average effect of CBratio is negative and statistically significant until the HHI value reaches approximately 2.15. It implies that in a highly competitive market, a greater presence of Banking infrastructure and the Paycheck Protection Program during the Covid-19 pandemic 89 Table 1. Effects of banking structure and Paycheck Protection Program (PPP) loans in US counties. Ordinary least squares (OLS) estimates Quantile regression estimates   Banking infrastructure and the Paycheck Protection Program during the Covid-19 pandemic 91 community banks significantly reduces PPP loans. These markets make up less than 0.35% of all counties. CBratio has no statistically significant impact on the number of PPP loans around the middle level of market concentration, which includes approximately 68% of the counties. In highly concentrated markets where HHI > 3.6, the CBratio significantly increases the number of PPP loans. These markets consist of 32% of the counties. In sum, the evidence suggests that structural characteristics of a regional banking sector unevenly affect the number of PPP loans transmitted from the federal government to small businesses. Market concentration reduces the number of PPP loans, but the negative impact is suppressed by a greater presence of community banks in the region. This negative effect of market concentration is expected and consistent with the traditional view (Berger et al., 2004;Cetorelli & Strahan, 2004) and the prediction in Bolton et al. (2016). In a competitive market, banks compete for PPP clients and processing fees, whereas in a concentrated market, banks with market power can prioritize bigger loans for bigger fee revenues.
However, in this concentrated market, a greater presence of community banks mitigates the negative impact of market concentration on the number of PPP loans because even with the market power, community banks are more likely to lend geographically closer local businesses than national banks, consistent with the previous literature showing spatial proximity between banks and firms increases small business lending in times of economic crisis and natural disasters (Cortés, 2014;Degryse et al., 2018;Hasan et al., 2019;Ivanov et al., 2020;Koetter et al., 2020;Zhao & Jones-Evans, 2017). That is why we see stronger mitigating impacts of community banks' role only in a highly competitive market but not in a competitive market where all banks competitively go after clients.
The insignificant coefficient of the metropolitan indicator suggests no discernible effect. A larger population is associated with fewer PPP loans. The magnitude tends to increase with the quantile. Covid-19 infection rates have no statistically significant impact, consistent with previous studies (Granja et al., 2020;James et al., 2020). James et al. (2020) point out that the PPP loan application rate was the highest, while the approval rates were the lowest in the areas highly impacted by Covid-19. It implies that banks perceive PPP loans in heavily impacted areas as risky. Thus, more confirmed Covid-19 cases are not necessarily associated with more PPP loans.
The last three variables are the percentage of jobs in three NAICS categories to account for the county's industry mix. The share of jobs in goods-producing industries, leisure and hospitality, and trade were included. No robust impact is found in all quantiles. For lower quantiles, a bigger share of jobs in trade is associated with more PPP loans. For higher quantiles, a bigger share of jobs in hospitality and the goods-producing industry is associated with fewer PPP loans. In the LAD model, a bigger Note: The effects of community bank presence depend on the degrees of banking market concentration. The effect is negative and statistically significant in competitive markets until the Herfindahl-Hirschman index (HHI) reaches approximately 2.15. The effect has no statistically significant impact on the number of PPP loans around the middle level of market concentration. In highly concentrated markets where HHI > 3.6, the CBratio significantly increases the number of PPP loans.  Banking infrastructure and the Paycheck Protection Program during the Covid-19 pandemic

REGIONAL STUDIES
share of hospitality jobs leads to fewer PPP loans, while a bigger share of trade jobs leads to more PPP jobs. Table 2 presents sensitivity analyses conducted in two different ways. First, an alternative measure for the market concentration index was used. For the baseline estimates, the market concentration index based on the local deposit share was used. Instead, an alternative HHI based on the share of bank branches in column (2) in Table 2 was used.

Results with alternative variables and specifications
The baseline estimates (LAD estimates in Table 1) are reported in column (1) for comparison purposes. Using the alternative market concentration measure statistically and substantially does not change the baseline results.
Second, additional demographic variables were controlled for in the model because these variables may affect the county's business profiles and the degrees of impact of the Covid-19. The added variables are educational attainment (the percentage of the population aged 25 and older with a bachelor's degree), size of the older population (percentage of the population aged 65 and over), the percentage of population under 18, the percentage of the non-Hispanic white population, the share of the non-Hispanic black population, and finally the share of the Hispanic population. The results are shown in column (3). The coefficients of key variables became slightly smaller, but the general pattern stays the same. Third, instead of the number of PPP loans per 100 businesses, the number of PPP loans per 1000 population was used. The results remain similar to the baseline estimates.

CONCLUSIONS
This paper examines how the existing market characteristics of the regional financial subsystem contribute to emergency credit access for small businesses. In particular, it examines how market concentration and the presence of community banks in the regional market determine the disbursement of PPP loans designed to ease the negative economic impacts of the Covid-19 pandemic. The analysis shows an interplay between different dimensions of financial market characteristics. Currently, 75% of the counties have a highly concentrated market by the Department of Justice's standards. 7 The findings show that the negative impact of HHI on small business loans exists, but community banks play a critical role in suppressing the negative effect. Community banks' moderating effect is especially pronounced in a highly concentrated market where a greater community bank presence significantly increases the number of PPP loans. The paper provides nuanced evidence that goes beyond the simple understanding of the role that community banks play in regional economies.
This paper contributes to the literature on regional economic resilience by providing empirical evidence on how multidimensional regional financial market characteristics influence credit access for small businesses in economic crisis. More than three-fourths of the county-level financial markets in the US are highly concentrated and mostly 'stuck' (Meyer, 2018). The findings suggests that in those concentrated markets, the presence of community banks is particularly important for small businesses, although recent development of financial technologies may substitute credit access for small businesses in markets with less competition (Erel & Liebersohn, 2020;Hannan, 2003).
Finally, the paper demonstrates the importance of channels through which federal policies are implemented. The federal government's fiscal and monetary policies significantly improve regional economic resilience after an economic shock while the impacts of state and local government policies are limited (Wolman et al., 2017). Studies have shown that PPP helped mitigate negative impacts of shelter-in-place orders due to Covid-19, which suggests that regional economic outcomes may depend on heterogeneous regional banking market characteristics. Thus, it is imperative to be aware of not only what federal assistance is implemented but also how it is transmitted to the target. This paper has limitations and raises future research questions. First, the interpretations of the empirical results are limited to correlations because the cross-sectional analysis makes it difficult to identify causality, although we can reasonably rule out a reverse causation and a third factor affecting both sides of the equation. It could be useful to examine whether similar correlations can be found in other parts of the world. Second, existing studies report small, short-run effects of PPP that dampened unemployment (Autor et al., 2020;Chetty et al., 2020;Faulkender et al., 2020;Granja et al., 2020;Hubbard & Strain, 2020;Li & Strahan, 2020). It remains to be seen how the financial market structure and community banks ultimately affected regional economies through PPP loans such as jobs and business survival after the pandemic.