The Short-Term Effect of the Paycheck Protection Program on Unemployment (Barraza, Rossi and Yeager)

Summary

Universidad de San Andrés working paper D.T. No. 144, The Short-Term Effect of the Paycheck Protection Program on Unemployment, by Santiago Barraza, Martín A. Rossi and Timothy J. Yeager, dated August 2020. The abstract states that the authors use the 2019 density of Small Business Administration member bank offices in a county as an instrument for PPP loans originated in that county during April 2020 and find statistically and economically significant effects of the program on unemployment. The introduction states that during the month following the launch of the program on April 3, 2020 the authors estimate it reduced unemployment by 1.4 percentage points, using data covering more than 98% of the labor force. It reviews other studies on the effects of PPP on employment.

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

Issuer: Source documents
Document type: The Short Term Effect Of The Paycheck Protection Program On Unemployment
Case: The Short Term Effect Of The Paycheck Protection Program On Unemployment

Cite this document

“The Short-Term Effect of the Paycheck Protection Program on Unemployment (Barraza, Rossi and Yeager).” Original at repec.udesa.edu.ar. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/the-short-term-effect-of-the-paycheck-protection-program-on-unemployment/, SHA-256 51d5472cd4bd5688f613438831c59a93f9af2bab0638137bab6da5cb4b5815b0.

Full text

The Short-Term Effect of the Paycheck Protection Program on Unemployment
Por Santiago Barraza (Universidad de San Andrés),
Martín A. Rossi (Universidad de San Andrés) y
Timothy J. Yeager (University of Arkansas)

D.T.: Nº 144 Agosto 2020

Vito Dumas 284, B1644BID, Victoria, San Fernando, Buenos Aires, Argentina
Teléfono 4725-7020
Email: economia@udesa.edu.ar
The Short-Term Effect of the Paycheck
Protection Program on Unemployment

Santiago Barraza
Universidad de San Andrés

Martín A. Rossi
Universidad de San Andrés

Timothy J. Yeager
University of Arkansas

This version: August 2020

Abstract

We study the short-term causal effect of the Paycheck Protection Program on unemployment.
Using the 2019 density of Small Business Administration member bank offices in a county as an
instrument for PPP loans originated in that county during April 2020, we find statistically and
economically significant effects from the program on unemployment. Our results highlight the
importance of this relief policy and the financial system infrastructure in preserving jobs during
the COVID-19 crisis.

Keywords: Banks; Financial system; COVID-19; Labor Market.
JEL classification: G18.


Santiago Barraza (sbarraza@udesa.edu.ar), Business School, Universidad de San Andrés; Martín A. Rossi
(mrossi@udesa.edu.ar), Department of Economics, Universidad de San Andrés; Timothy J. Yeager
(tyeager@uark.edu), Walton College of Business, University of Arkansas. Cao Fang provided excellent research
assistance.
I. Introduction

We study the short-term causal effect of the Paycheck Protection Program (PPP) on

unemployment. We estimate that during the month following the launch of the program on April

3, 2020, it reduced unemployment by 1.4 percentage points.

Our main contribution is the identification of a causal effect of PPP on unemployment. Our

identification strategy exploits the heterogeneity in the capacity of the banking sector in each

county to process PPP loan applications as measured by the density of banking offices that were

members of the Small Business Administration (SBA) in 2019.1 This county-level instrument

differs in important ways over alternative instruments used in the literature. First, it permits a

comprehensive estimate of the effect of PPP on U.S. unemployment using data that cover more

than 98% of the labor force. In addition, the instrument is predetermined so it excludes the

endogenous increase in PPP processing capacity as additional banks that were not SBA members

began to process loan applications. Our results show that the unemployment path between counties

with high and low SBA member office densities diverges significantly post-intervention, as

expected.

The rapidly evolving literature thus far shows mixed results regarding the effects of PPP on

the employment. Three studies find strong effects from the program. Bartik et al. (2020) use

Homebase time-clock tracking data on small business hourly employees primarily in retail and

food service. They find that small-business employees in states that received more PPP loans as a

share of payroll were less likely to stop work in April 2020. The authors concede, however, that

their analysis is suggestive rather than causal. Autor et al. (2020) use payroll processing data from

ADP to identify causal effects from PPP through a difference-in-difference design that compares

1
We estimate that two-thirds of the first round of PPP loans were made by 2019 SBA members, which includes all
the largest banks.
1
employment trends of small businesses below and above 500 employees, which is the approximate

size cutoff for PPP loan eligibility. Their preferred estimate is that PPP boosted employment at

eligible firms by 3.25 percent. Humphries et al. (2020) use survey evidence to show that small

businesses that received PPP loans were 9 percentage points less likely to have fewer employees

than in January 2020.

Two studies find modest or no effects on employment from PPP. As with Autor et al. (2020),

Chetty et al. (2020) use a difference-in-difference approach around the 500-employee PPP

eligibility cutoff using payroll information from Earnin. The authors conclude that the PPP had

little effect on small business employment; however, they cannot rule out a positive employment

effect of 3 or 4 percentage points, which they consider small relative to the large number of jobs

lost in the weeks after the COVID-19 shock. Closer to our approach, Granja et al. (2020) exploit

the geographical heterogeneity of banks’ observed origination of PPP loans relative to their typical

small business loan volume. Because the average large bank processed PPP loans slowly in the

first round of the program, small business access to PPP loans was hindered in the large-bank

lending markets. The authors show that counties where bank processing of PPP loan applications

greatly exceeded their typical small business loan volumes had significantly lower unemployment

insurance filings than other counties, consistent with a beneficial effect from PPP. However, the

divergence in unemployment claims between counties with relatively high-volume versus low-

volume processors emerges in the two weeks prior to the start of PPP, which they attribute to

adverse effects from the economic lockdown induced by the pandemic. The authors find no

additional divergence in claims after PPP implementation. Autor et al. (2020) also find some

evidence of a positive employment effect in the weeks just prior to the April 3 implementation

date, but they attribute the results in part to firms’ retention of workers in anticipation of PPP loans,

and they do find an effect from PPP post-intervention.

2
We find that, on average, the PPP improved county employment across the board.

Consistent with Granja et al. (2020) and Liu and Volker (2020), our evidence also attributes small

and regional banks a significant role in this result because counties with high-density offices tend

to have a high proportion of these banks that were particularly efficient at processing loan

applications in the initial weeks of the program.

Our paper proceeds as follows. Section II provides background on the PPP and discusses the

identification strategy. Section III describes the data. Section IV presents the econometric methods

and reports the results. Section V concludes.

II. Paycheck Protection Program

The Paycheck Protection Program was created as part of the Coronavirus Aid, Relief, and

Economic Security (CARES) Act, which was signed into law on March 27, 2020. The program

objective is to provide financial support to small businesses to keep their businesses afloat and

workers employed. Through delegated authority from the SBA, banks lend 100% government-

guaranteed funds to businesses to cover eight weeks of eligible payroll, rent, and utility expenses.2

Although the interest rate is 1%, the loan is forgiven in full or in part after lenders verify that the

funds have been used for eligible expenses. Payments for ineligible expenses can be deferred at

least six months following loan disbursement.

Businesses eligible for a loan include small business concerns with primary residence in the

U.S. that were in operation on or before February 15, 2020 and have 500 or fewer employees (or

otherwise meet employee-based industry size standards). Lenders already certified as SBA 7(a)

lenders prior to passage of the CARES Act are automatically eligible to participate as lenders in

the PPP program.3 Most other federally insured depository institutions (and non-bank institutions)

2
“Business Loan Program Temporary Changes; Paycheck Protection Program,” SBA Interim Final Rule, April 5,
2020.
3
The 7(a) loan program is the main program through which SBA provides financing to small businesses.
3
are automatically eligible to participate upon submission of the Lender Agreement (SBA Form

3506).

The first phase of PPP set aside $349 billion in government-guaranteed forgivable loans.

The SBA began processing new lender applications on April 3, 2020. Demand was strong, and

funding was depleted as of April 16. Congress replenished the fund with an additional $310 billion

in funding, signed into law April 24, and the SBA resumed processing small business loan

applications on April 27.

III. Data

Unemployment data

We collect monthly county-level labor data through April 2020 from the Local Area

Unemployment Statistics (LAUS) program, Bureau of Labor Statistics.4 For each county-month,

we compute the unemployment rate as the ratio of unemployed to total civilian labor force. During

the early stages of the COVID-19 crisis in April 2020, the average unemployment rate in U.S.

counties rose by 7.64 percentage points, from 4.81% at the end of March to 12.45% at the end of

April. We also use changes in weekly unemployment initial claim rates (UICR), generously made

available by Chetty et al. (2020).5

Banking data

The FDIC Summary of Deposit database contains the geographic location and deposits of

every (thrift and commercial) bank office, which includes the headquarters office and branches as

of June 30 each year.6 We retrieve the 2019 data and retain full service, retail, and loan production

offices, and we aggregate by bank the number of offices and deposits at the county level.

4
https://www.bls.gov.
5
https://github.com/Opportunitylab/EconomicTracker/tree/main/data.
6
https://www7.fdic.gov/sod/dynaDownload.asp?barItem=6.

4
We also identify banks that were “SBA members” in 2019.7 We do so by extracting from

the SBA 7(a) loan database distinct banks that made at least one loan in 2019 through the 7(a)

program.8 Unlike the PPP data, this database includes the bank headquarters zip code, city, county,

and state. We match banks in the SBA 7(a) database to the FDIC list of banks in 2019 to obtain

the unique FDIC certificate (cert) number via a two-step process. We first match by a nine-

character string of bank name, state, and zip code. We then match the remaining banks by the same

nine-character name and state, excluding zip code. Because this process is less precise than

matching by zip code, we manually identify incorrect matches based on full name discrepancies

and remove them from the database. In the end, we identify 1,383 SBA member banks that

participated in the SBA 7(a) lending program in 2019. They spread across 2,778 counties and 50

states plus DC, encompassing 98.9 percent of the population and 98.4 of the labor force in the U.S.

We then merge by FDIC cert number the SBA-member database with the county-bank

aggregated summary of deposits database. Finally, we aggregate the county-bank database to the

county level and merge in county-level population data for 2019 from the U.S. Census Bureau.

The resulting database allows us to compute the per-capita number of offices, banks, and deposits

at the county level for all banks and for SBA member banks.

PPP data

Because our analysis is at the county level, we aggregate the PPP loan data by the county of

the borrower. We retrieve the PPP data for the period April 3-June 30, 2020 from the SBA

website.9 The borrower’s county is not in the database so we map zip code to county using the

HUD-USPS Zip Crosswalk Zip-County file.10 Roughly, one-third of zip codes match to multiple

7
Of course, all PPP lenders are also SBA members eligible to participate in that program, but we use this terminology
for clarity to identify SBA 7(a) members prior to the PPP.
8
https://www.sba.gov/about-sba/open-government/foia.
9
https://www.sba.gov/funding-programs/loans/coronavirus-relief-options/paycheck-protection-program.
10
Obtained at https://www.huduser.gov/portal/datasets/usps_crosswalk.html.
5
counties, so we split the PPP loan count across the counties by the percentage of businesses in the

zip code located in each county using the Business Ratio weight provided in the crosswalk.

Other data

Our estimation equations include four control variables. In addition to county-level deposits,

we include the county COVID case rate as of March 31, 2020 provided by Chetty et al. (2020) and

downloaded from Economic Tracker.11 The share of small businesses in a county with 500 or fewer

employees is computed from the 2017 County Business Patterns data.12 Finally, county population

density is computed as 2019 population divided by square miles of land area in 2010 using data

from the 2019 Census Bureau.

Table 1 reports summary statistics of the data.

IV. Econometric methods and results

Our objective is to estimate the causal effect of PPP loans on unemployment. Formally, we

wish to estimate the following equation:

ΔUnemploymentcs = β + α PPP Loanscs + γ Xcs + δs + εcs (1)

where c indexes counties, s indexes states, Xcs is a vector of county-level controls, δs is a state fixed

effect, and εcs is an error term. The coefficient of interest is α.

PPP Loans is potentially endogenous in equation (1). A self-evident reason is that firms in

counties with more acute unemployment problems could be more likely to apply for PPP loans,

thus reversing the order of causality. Hence, we estimate equation (1) by 2SLS, where the variable

SBA Member Offices is an instrument for PPP Loans. The choice of our instrument is motivated

by the frictions introduced by the sheer magnitude of originating the millions of PPP loans in a

11
https://tracktherecovery.org.
12
2017 County Business Patterns and 2017 Economic Census, “Number of Firms, Number of Establishments,
Employment, and Annual Payroll by Enterprise Employment Size for Counties, Totals: 2017,” downloaded 31 June
2020 at https://www.census.gov/data/tables/2017/econ/susb/2017-susb-annual.html.

6
short timeframe.13 Thus, our identification strategy exploits the heterogeneity across counties in

the agility of the banking industry to process the loan applications. The most apparent reason why

counties with a high-density banking sector would have an advantage in processing loan

applications quickly is that more employees are available to do the time-intensive work – 45

minutes per loan application according to one loan officer at a large bank – of inputting borrower

information into E-Tran, the SBA’s electronic system for loan processing. 14 Moreover, banks in

counties with high per-capita densities of SBA member offices may have also had an advantage in

loan processing over counties with lower SBA member office densities because banks involved in

the SBA 7(a) lending program already had relevant administrative experience and compliance

information required by the SBA on their small business clients.

More formally, to be a suitable instrument for PPP Loans, the variable SBA Member Offices

needs to be both relevant and exogenous. We first explore the exclusion restriction, or the

exogeneity of SBA Member Offices in equation (1). Our key identifying assumption is that SBA

member office density affects unemployment only through PPP loans. A natural concern is that

banks can affect unemployment via other functions they perform, such as holding deposits or

extending credit lines that provide liquidity on demand. Because the application process for PPP

loans began April 3, we test the identification assumption by examining the correlation between

the density of SBA member offices and pre-treatment changes in unemployment. We compute the

change in unemployment at the county level for each of the four months prior to the launch of the

program (December 2019 to March 2020). As shown in Table 2, the density of SBA member

offices in a county is uncorrelated with changes in unemployment for all the pre-treatment periods,

13
Lenders originated, and the SBA approved, roughly 2.9 million business loans during April 2020. A ballpark
estimate based on the outstanding 24 million small business loans on bank balance sheets on the Call Reports and a
56-month average loan maturity obtained from the Federal Reserve Bank of Kansas City Small Business Lending
Survey (fixed-rate, term loans), suggests this volume is about 6.8 times the number of small business loans banks
process during a typical month.
14
WSJ, 7 April 2020, “Lawmakers, White House Vow to Quickly Provide More Small-Business Loans.”
7
even during March 2020 when Li et al. (2020) find evidence of significant drawdowns at certain

banks, validating our identification assumption.

Although the density of SBA member offices is orthogonal to pre-treatment changes in

unemployment, a potential concern is that the orthogonality measured at a monthly frequency in

normal times no longer holds in the early days or weeks of a labor market crisis such as the

COVID-19 pandemic just prior to the policy intervention. To address this concern, and to further

validate our identification assumption, we examine changes in weekly unemployment initial claim

rates (UICR) beginning January 11, 2020. These data cover a subset of 1,424 counties. Figure 1

presents average weekly UICR for counties in the fourth quartile of SBA member office density

and the rest of the quartiles. According to the figure, the evolution of UICR before PPP was

launched is similar across counties with different SBA member office density – even when claims

suffered their sharpest jump the week ending Saturday March 28. The paths of UICR diverge only

after PPP is launched as seen in the figure for the week ending Saturday, April 4. Thus, this

evidence eases concern that other banking functions could have affected unemployment differently

across counties with different SBA member office densities in the wake of the onset of the COVID-

19 pandemic and prior to the policy intervention.

We now explore the relevance of our instrument by examining the link between SBA Member

Offices and PPP Loans.15 Columns (1) and (2) in Table 3 report first-stage estimates, without and

with controls. In all cases, first-stage effects are precisely estimated and indicate that those counties

with more SBA member offices are more likely to originate PPP loans compared to counties with

fewer SBA member offices.

15
Figure A1 in the Online Appendix displays the average number of PPP loans originated in April 2020 by counties
in the top quartile of SBA member office density and the remaining counties. Banks in counties in the top office
density quartile originated, on average, nearly twice as many loans as the other counties, consistent with the intuition
that our instrument identifies the most agile banking offices across counties. Figure A2, also in the Online Appendix,
presents heat maps of SBA member office density (Panel A) and PPP loans (Panel B) that reinforce the intuition.
8
Figure 2 provides a preview of our main results. The figure displays the evolution of county-

level unemployment between November 2019 and April 2020 according to county-level density

quartiles of SBA member offices in 2019. Consistent with results displayed in Table 2 and Figure

1, the figure shows that both the level and trend of unemployment are unrelated to the density of

SBA member offices in all the pre-intervention period (November 2019-March 2020). In contrast,

the divergence begins only in April when the PPP was launched. In the post-intervention period,

counties with high SBA member office densities experience far lower increases in unemployment

rates than counties with lower densities.

Columns (3) and (4) in Table 3 report reduced-form estimates of the link between SBA

Member Offices and ΔUnemployment, without and with controls. The coefficient on SBA Member

Offices is always negative and statistically significant, indicating that those counties with higher

SBA member office densities are more likely to have smaller increases in unemployment during

the COVID-19 crisis than those counties with lower SBA member office densities.

Columns (5) and (6) in Table 3 report our main results, the 2SLS estimates of equation (1).

The coefficient of PPP Loans is negative and significant. The magnitude of the coefficient is

meaningful: an increase of one standard deviation in PPP loans per 10,000 people is associated

with approximately a 0.25 standard deviation decrease in the change in unemployment at the early

stages of the COVID-19 crisis.16 The average-sized county in our sample (with approximately

104,000 people) received 1,180 loans, which saved 1.4% of the 48,880 labor force in that county,

or 684 jobs.17

We also conducted additional analysis to explore the beneficial role that small and regional

16
All results are robust to using the change in unemployment between December 2019 and April 2020. All findings
mentioned but not shown are available from the authors upon request.
17
County-level BLS data show a mean labor force participation rate of 47% of the population as of December 2019.

9
banks played in distributing PPP funding. Importantly, all our results hold if we define the measure

of SBA member office density as the number of bank offices with assets less than $100 billion at

year-end 2019, an asset threshold that excludes the 29 largest banks in the U.S.18 In fact, the

economic significance of the effect of PPP on unemployment is larger than when including all

SBA member offices. This result is expected because the share of small and regional offices in the

quartile of counties with the highest office density is 82% compared with a 60% share in the

quartile of counties with the lowest office density.

V. Final remarks

We present causal evidence on the palliative effect of the Paycheck Protection Program on

unemployment during its first month of implementation. To our knowledge, this is the first causal

work that estimates the impact of PPP on nearly the entire labor force in the United States, which

should make a valuable contribution to the evolving analysis of the economic effects of this historic

relief program.

The issue of efficacy of PPP embeds two fundamental questions. First, was the program

allocated where it was most necessary? Second, where allocated, did the program achieve its

desired goals? Although our paper directly addresses the latter question, in so doing, it also hints

at some subtle relations with allocation. Our findings suggest that the pre-existing banking

infrastructure was a significant determinant of loan allocation. To the extent that we can assume

that this infrastructure was less than perfectly related with the pandemic shock, we can also assume

that loans have not necessarily been allocated where they were needed the most. Thus, studies

whose designs are meant to assess the efficacy of the program in terms of allocation efficiency

18
See Table A1 in the Online Appendix.

10
only or combine both questions at once, are likely to find slightly different results. Those results

should be deemed complementary to ours, rather than substitute or contradictory.

11
References

Autor, David, David Cho, Leland D. Crane, Mita Goldar, Byron Lutz, Joshua Montes,

William B. Peterman, David Ratner, Daniel Villar, and Ahu Yildirmaz (2020). “An Evaluation of

the Paycheck Protection Program Using Administrative Payroll Micro-data." Working Paper, MIT

July 2020.

Bartik, Alexander, Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matthew Unrath

(2020). “Measuring the Labor Market at the Onset of the COVID-19 Crisis.” University of

Chicago, Becker Friedman Institute for Economics Working Paper No. 2020-83.

Chetty, Raj, John N. Friedman, Nathaniel Hendren, and Michael Stepner (2020). “The

Opportunity Insights Team. 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 No 27431.

Granja, João, Christos Makridis, Constantine Yannelis, and Eric Zwick (2020). “Did the

Paycheck Protection Program Hit the Target?” NBER Working Paper No. 27095.

Humphries, John, Christopher Nelson, and Gabrile Ulyssea (2020). “Information Frictions

and Access to the Paycheck Protection Program.” IRS Working Papers no. 643.

Li, Lei, Philip E. Strahan, and Song Zhang (2020). “Banks as Lenders of First Resort:

Evidence from the COVID-19 Crisis.” NBER Working Paper No. 27256.

Liu, Haoyang and Desi Volker (2020). “Where Have the Paycheck Protection Loans Gone

So Far?” Liberty Street Economics, Federal Reserve Bank of New York.

12
Table 1. Summary statistics
Mean Standard Minimum Maximum
Deviation

ΔUnemployment 7.64 4.92 -4.38 32.43
SBA Member Offices 2.28 1.72 0.14 21.01
PPP Loans 117.88 82.48 0.00 774.25
Deposits ($000s) 23.22 65.43 0.93 3,454.59
Population Density 269.31 1,781.35 0.04 71,340.39
COVID exposure 19.10 56.76 0.00 1,030.00
Small Business Share 0.90 0.04 0.51 1.00

Notes: ΔUnemployment is the change in the monthly county-level unemployment rate in
April 2020. SBA Member Offices is the county-level number of offices of SBA-affiliated
banks in 2019, per 10,000 people. PPP Loans is the number of loans originated in April
2020 to firms located in a given county, per 10,000 people. Deposits is the county-level,
per capita dollar amount of deposits in June 2019, in thousands of U.S. dollars. Population
Density is the population density of the county, as in 2019. COVID exposure is the county
COVID case rate as of March 31, 2020. Small Business Share is the share of firms with
500 or fewer employees in the county.

13
Table 2. SBA member offices and pre-treatment changes in unemployment
Dependent variable: ΔUnemployment
March 2020 February 2020 January 2020 December 2019
(1) (2) (3) (4)

SBA Member Offices -0.012 0.002 0.002 0.011
(0.012) (0.008) (0.011) (0.009)

State dummies Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Number of observations 2,777 2,777 2,777 2,777

Notes: Standard errors clustered at the state level are shown in parentheses. All models use county-
level data in the U.S. and are estimated using Ordinary Least Squares. County-level control
variables are Deposits, Population Density, COVID exposure, and Small Business Share.
*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.

14
Table 3. Main results
First-stage Reduced-form 2SLS
(1) (2) (3) (4) (5) (6)

SBA Member 13.670*** 13.761*** -0.198*** -0.195***
Offices (2.195) (2.044) (0.063) (0.060)
PPP Loans -0.014** -0.014***
(0.006) (0.005)

F-statistic [CV] 38.78[16.38] 45.34[16.38]
State dummies Yes Yes Yes Yes Yes Yes
Controls No Yes No Yes No Yes
Observations 2,778 2,777 2,778 2,777 2,778 2,777
Estimation method OLS OLS OLS OLS 2SLS 2SLS

Notes: Standard errors clustered at the state level are shown in parentheses. All models use county-
level data in the U.S. In models (1) and (2) the instrumented variable is PPP Loans. In models (3)
to (6) the dependent variable is ΔUnemployment. F-statistic is the Kleibergen-Paap Wald rk F
statistic, which is compared to Stock-Yogo weak ID test critical values (10% level). County-level
control variables are Deposits, Population Density, COVID exposure, and Small Business Share.
In 2SLS models, PPP Loans is instrumented using SBA Member Offices. *Significant at the 10%
level. **Significant at the 5% level. ***Significant at the 1% level.

15
Figure 1. Evolution of weekly unemployment claims by density of SBA member offices

Notes: Weekly unemployment claims at county level, January-April 2020. Qs
indicate quartiles of the county-level density of SBA member offices.

16
Figure 2. Evolution of monthly unemployment rate by density of SBA member offices

Notes: Monthly county unemployment rate, November 2019-April 2020. Qs
indicate quartiles of county-level density of SBA member offices.

17
Online Appendix

Table A1. Results using only small and regional banks
First-stage Reduced-form 2SLS
(1) (2) (3) (4) (5) (6)

SBA Member 13.307*** 13.514*** -0.318*** -0.304***
Offices (2.055) (1.870) (0.046) (0.044)
PPP Loans -0.024*** -0.022***
(0.005) (0.004)

F-statistic [CV] 41.93[16.38] 52.25[16.38]
State dummies Yes Yes Yes Yes Yes Yes
Controls No Yes No Yes No Yes
Observations 2,703 2,702 2,703 2,702 2,703 2,702
Estimation method OLS OLS OLS OLS 2SLS 2SLS

18
Figure A1. Average number of county PPP loans by density of SBA member offices

Notes: Average number of PPP loans approved in April 2020 by counties
in the first three quartiles (Q1-Q3) and top quartile (Q4) of the distribution
of the density of county-level SBA member offices.

19
Figure A2. Panel A. SBA Member Office Density

Notes: Heat map of SBA member office density. Darker shades represent counties with higher
office density in 2019. Map courtesy of Tableau software. We are grateful for their generosity.

Figure A2. Panel B. PPP Loans

Notes: Heat map of PPP Loans. Darker shades represent counties with more PPP loans in April
2020. Map courtesy of Tableau software. We are grateful for their generosity.

File and source

File: the-short-term-effect-of-the-paycheck-protection-program-on-unemployment.pdf
Size: 1,025,477 bytes
SHA-256: 51d5472cd4bd5688f613438831c59a93f9af2bab0638137bab6da5cb4b5815b0
Our copy: the-short-term-effect-of-the-paycheck-protection-program-on-unemployment.pdf
Original: repec.udesa.edu.ar