Zwick SBA Appendix PPP Hit Target
Cite this document
“Zwick SBA Appendix PPP Hit Target.”
Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/zwick-sba-appendix-ppp-hit-target-b9ed60e709/
, SHA-256 ebb47b804f88ab40e0629589bb44864c56ca0840de22aff10384c7533043d8b9.Full text
For Online Publication
Appendix A. Data Appendix
We draw upon microdata made available through the Small Business Administration (SBA)
and the Department of Treasury containing all PPP loans, allowing us to observe all loans
approved under the program.1 For all loans, the data include loan amount, lender name,
the borrower’s self-reported industry and corporate form, workers covered by the loan, and
some demographic data on firm owners, including borrower name. Our targeting analysis and
bank exposure research design use data for all loans aggregated to either the regional or local
geography level.
We merge this data set with the Reports of Condition and Income (Call Reports) filed by
all active commercial banks as of 2020:Q1. Specifically, we use a bigram string comparator
for the lender name to match the lender names in the PPP data set to commercial and savings
banks in the Call Reports data set. The main challenge in this process is that many lender
names are matched to multiple distinct banks with the same legal name. For instance, there
are fifteen distinct banks whose legal name is “Community State Bank” filing a call report in
the first quarter of 2020. We address this issue by assigning each loan made by these distinct
banks with similar legal names to the similarly-named bank with the branch that is closest
to the zip code where the loan was made.2 We are able to match 4,370 bank participants in
the PPP program to the Call Reports data set. We did not match 795 commercial and savings
banks that filed a Call Report in the first quarter of 2020. We assume that these banks did not
participate in the PPP program and made no PPP loans. Overall, lenders in the PPP sample
that we matched to the Call Report account for 90.5% of all loans disbursed under the PPP.
We classified 926 PPP program participants as credit unions and 45 participants as agricul-
tural credit associations. We also classified the remaining 123 participants as non-bank PPP
lenders. This group is very heterogenous and comprises small community development funds
(e.g. Montana Community Development Corporation), as well as finance companies and Fin-
tech lenders. After careful investigation of companies websites, we classified thirteen non-bank
lenders as Fintech lenders. Interestingly, Fintech lenders account for 4.2% of the total number
of loans in the program and a single Fintech lender, Kabbage Inc., accounts for more than half
1
An earlier version of this paper used data from a Freedom of Information Act request on the number of
approved PPP loans and approved PPP amounts during the first round of the program.
2
Most of these banks with similar legal names are small and operate in different states. Given the proximity
between lenders and PPP borrowers across the entire sample, we are confident that our allocation process assigns
most loans to their correct lender.
1
of the loans made by Fintech lenders.
We obtain financial characteristics of all banks from the Call Reports, which provide detailed
data on the size, capital structure, and asset composition of each commercial and savings bank
operating in the United States. Importantly, we obtain information on the number and amount
of small business loans outstanding of each commercial and savings bank from the “Loans to
Small Business and Small Farms Schedule” of the Call Reports. Using this information, we
benchmark the participation of all commercial and savings banks in the PPP program relative
to their share of the small business lending market prior to the program.
As noted in the main text, we use the matched-PPP-Call-Reports data and Summary of De-
posits data containing the location of all branches and respective deposit amounts for all de-
pository institutions operating in the United States as of June 30th, 2019. A significant number
of depository institutions merged in the second half of 2019, which means that some branches
are assigned to commercial and savings banks that no longer exist as stand-alone institutions.
Notably, SunTrust Banks, Inc. merged with Branch Banking and Trust Company (BB&T) to
create the sixth largest financial institution in the United States. We use the bank mergers file
from the National Information Center to adjust the branch network of merged institutions and
account for these mergers. We use data from the County Business Patterns dataset to approxi-
mate the amount of PPP lending per establishment and the fraction of establishments receiving
PPP loans in the region. It is important to note that the County Business Patterns data include
establishments for all firms, including those too large to qualify for the PPP. We use these data
to examine how the use of the banking system to deploy the PPP funds affected their distri-
bution. The maintained assumption is that the unobservable share of establishments that are
not eligible to receive PPP funds is not systematically associated with the exposure to the PPP
performance of local banks in the region.
To evaluate whether PPP amounts were allocated to areas that were hardest-hit by the
COVID-19 crisis and whether the program improved economic employment and other eco-
nomic outcomes following its passage, we use data from multiple available sources on the
employment, social distancing, and health impact of the crisis. We obtained detailed data on
hours worked among employees of firms that use Homebase to manage their scheduling and
time clock. Homebase processes exact hours worked by the employees of a large number of
businesses in the United States. We use information obtained from Homebase to track em-
ployment indicators at a weekly frequency at the establishment level. The Homebase data set
disproportionately covers small firms in food and beverage service and retail; therefore, it is
not representative of aggregate employment. At the same time, the Homebase data are quite
2
useful for evaluating the employment impacts of the PPP specifically, since many hard-hit firms
are in the industries Homebase covers and much of the early employment losses came from
these firms.
We use the Homebase data in our bank exposure and matched sample analysis to measure
the impact of PPP funding on employment and business shutdowns. To broaden this analy-
sis, we supplement the Homebase data with three additional data sources. First, we obtain
county-by-week initial unemployment insurance claims from state web sites or by contacting
state employment offices for data. We use initial unemployment insurance claims as a mea-
sure of flows into unemployment. Second, we supplement the Homebase data with data from
Womply, a company that aggregates data from credit card processors. The Womply data in-
cludes aggregate card spending at small businesses at the county-industry level, defined by the
location where a transaction occurred. We aggregate the Womply dataset at the county level
to harmonize the level of aggregation we use across different analyses. We find similar results
when we use the Womply dataset at the county-industry level. Small businesses are defined
as businesses with revenues below SBA thresholds. We complement these data sources with
additional county-level employment data from Opportunity Insights, which are described in
detail in Chetty et al. (2020).3 The employment rates are based on employment data that
Opportunity Insights obtained from Paychex, Earnin, Intuit, and Kronos. The data are at the
county/week level and span the period from January 2020 to the end of August.
We obtain counts of COVID-19 cases by county and state from the Center for Disease Con-
trol and use data on the effectiveness of social distancing from Unacast. Unacast provides a
social distancing scoreboard that describes daily changes in average physical mobility. Unacast
measures the change in average distance travelled using individuals’ GPS signals. The data are
available on a daily basis at the county level. We obtain information on the effective dates of
statewide shelter-in-place orders from the New York Times.4
To understand the mechanisms underlying our results, we draw on data from the Census
Bureau’s Small Business Pulse Survey (SBPS), launched within seven weeks of the national
emergency declaration in March (Buffington et al., 2020). To obtain real-time information tai-
lored towards small businesses, the SBPS was run weekly from April 26th to June 27th with
businesses contacted via email based on the Census Bureau’s Business Register, which is popu-
3
We also refer readers to Chetty et al. (2020) who provide comparisons between HomeBase and aggregate
employment, showing that it provides an overall good glimpse of employment dynamics.
4
See https://www.nytimes.com/interactive/2020/us/coronavirus-stay-at-home-order.
html.
3
lated using responses to the Economic Census across the 50 states (and D.C. and Puerto Rico).5
Furthermore, the SBPS focuses on businesses with receipts that are greater than or equal to
$1,000 but retain 500 employees or fewer. This sampling frame closely fits the target popula-
tion for the PPP. Each week, the sample weights are adjusted to maintain representativeness.
Our goal in assembling these diverse data is to conduct a comprehensive assessment of
how business liquidity support affects firm behavior. We observe both intensive and extensive
margin employment and operating responses by targeted firms and in their local labor markets.
We link this behavior to data on the performance and geographic footprint of banks, the agents
used to transmit funds to eligible firms as quickly as possible. With the Census data, we draw
upon responses to questions about small business liquidity, loans, defaults, and applications for
various forms of private and public government assistance, including specific questions about
the PPP and EIDL programs.
5
We focus on the first phase of the survey that spans the period from April 26th to June 27th because the
second phase of the survey only started mid-August and ended in October after the PPP ended.
4
Appendix B. PPPE Statistics and Heterogeneous Demand for PPP Loans
This Appendix provides further details about some statistical properties of our PPPE mea-
sure of local exposure to banks that over- or underperformed in the deployment of PPP funds
and further investigates the impact that local differences in demand for PPP funds might have
on the empirical analyses.
Figure B.1 is a histogram of the distribution of bank PPPE at the end of the first round and
during the second round of the program. Both histograms show a wide dispersion of relative
performance. The second round histogram shows a shift in PPPE, with some banks that barely
participated in the first round considerably improving their performance subsequently.
Figure B.2 provides a description of the spatial distribution of PPPE. Panel A plots the spa-
tial distribution of county-level PPPE across the United States and Panels B and C describe the
spatial distribution of ZIP-level PPPE in Chicago and New York City, respectively. Panel A indi-
cates that the Midwest and New England regions were exposed to banks that performed well
in delivering PPP during the first round whereas Western areas were less exposed to banks that
performed well in deploying PPP. Panels B and C show that, even within cities, there could be
relevant differences in local exposure to banks that performed well. The less affluent areas of
the city of Chicago, such as the South Side of the city, were exposed to banks that performed
well, whereas firms in the downtown area were served by banks that did not do so well in
deploying PPP funds. On the other hand, in New York City, less affluent and less densely pop-
ulated areas were served by banks that performed worse in the first round relative to banks
elsewhere in the city. These differences in exposure between Chicago and New York City prob-
ably reflect the heterogenous role of exposure to bank PPP performance.
Figure B.3 is a survival function of the time to receive PPP funds after partitioning firms in
the sample based on their local exposure to ZIP-PPPE. The figure shows that exposure to un-
derperforming PPPE banks is associated with meaningful differences in when borrowers could
access funds. Only 25% of all PPP borrowers located in ZIP codes whose banks underperformed
obtained PPP approval prior to the end of the first round. By contrast, approximately 42% of
all PPP borrowers in ZIP codes whose banks overperformed had access to funds in the first
round.
Figure B.4 shows that we find a similar relationship between our state exposure to PPPE
and the share of firms that report receiving and requesting PPP funds at the end of the first
round when we measure bank PPP performance based on the share of the total amount of PPP
and total amount of small business loans disbursed by each bank, rather than the share of the
number of PPP and small business loans by each bank.
5
Table B.1 shows summary statistics of the PPPE variable and other outcome variables in the
paper.
We believe the differences in PPPE mostly capture initial variation in banks’ abilities to
process PPP loans. Here, we present some supplementary evidence to support this conjecture.
In Appendix Table B.2, we show that the ability of banks to include information about the
program in their websites and to receive online applications is positively associated with bank
performance.
In Table B.3, we repeat the empirical analysis of Table 4 but include county⇥industry fixed
effects rather than state⇥industry fixed effects. By including county⇥industry fixed effects, we
ensure that our results compare firms that are exposed to banks with weak performance during
the first round with firms in the same county and in the same industry that were close to banks
that were quick to disburse PPP loans and had a strong performance during the first round.
This empirical strategy further assuages concerns that PPPE captures differences in demand,
since it is comparing firms that are exposed to very similar economic and health conditions
(same county and same industry) but that happened to be located in areas close to banks that
were swift in processing PPP applications.
To further allay concerns about the possibility that PPPE captures differences in demand for
PPP loans across regions, we created a predicted PPPE measure that uses variation in supply-
side restrictions that limit banks’ ability to process PPP loans. Panel A of Figure B.5 suggests
that banks that had greater labor intensity ratios performed relatively better in deploying PPP
loans during the first round. Panel B of Figure B.5 shows that banks with no prior relationship
over the past three years performed worse, on average, than banks that originated at least one
SBA guaranteed loan over the past three years. Moreover, the intensity of the SBA relationship
is also associated with improved bank performance. Banks with a greater fraction of their small
business lending in the form of SBA-guaranteed loans performed relatively better than banks
that only sporadically interacted with the SBA.
The generous terms of the PPP and the broad eligibility criteria likely meant that demand for
the program was very high almost everywhere. Our evidence in Figure 3 suggests that, even
at the end of the first round, approximately 75% of establishments surveyed in the Census
Pulse Survey had applied for PPP funds. Our conjecture is that differences in access to banks
that quickly processed PPP were the main factor explaining differences in timing of receipt
of PPP funds. Because of capacity constraints and difficulties in processing the sudden influx
of PPP applications, banks prioritized their existing relationships. Small businesses that had
relationships with banks that were swift in deploying PPP obtained access to PPP funds quickly.
6
Other businesses that were close to or were in relationships with banks that were unable or
unwilling to process their applications either waited until their bank was able to process their
application or found another bank that was willing and able to process their application.
To provide further evidence supporting this conjecture, we obtained access to a proprietary
dataset from a Chicago community bank. This bank performed well in delivering PPP to their
clients, i.e., it had a high bank PPPE. In this proprietary dataset, we are able to observe the
date in which each PPP loan was fully funded and, more importantly, we also learn about the
location of each small business that received a PPP loan from that bank and whether it is a
new or old client of the bank. In Figure B.6, we show that this bank delivered funds to its
existing clients much quicker than it did to small businesses that opened an account with the
bank after the announcement of the program. Consistent with the idea that new clients were
likely in relationships with other banks that were not able to process their applications quickly,
in Table B.4 we find that the new clients that were served by this community banks came from
areas with lower PPPE, lower predicted PPPE, and greater exposure to Wells Fargo. We also
find that these new clients came from farther away and were smaller than existing clients of
the community bank. This pattern suggests that these new clients likely were in relationships
with other banks that were not processing PPP applications quickly. Alternatively, their banks
did not prioritize theses applications, and these firms had to search for other, more distant
banks to obtain access to PPP funds.
7
Figure B.1: Histogram of Bank Paycheck Protection Program Exposure (PPPE)
Figure B.1 plots the distribution of bank PPPE measured at the end of the first round (April 15th, 2020) and
when the flow of second round funds approximately ends (June 30th, 2020). We compute this measure as:
P P P E b,N br = Share Nbr. PPP Share Nbr. SBL
Share Nbr. PPP+Share Nbr. SBL ⇥ 0.5. We weigh each bank observation by its size measured as total assets
as of the end of 2019. Data are from the SBA and commercial bank Call Reports.
3
Density (Asset-Weighted)
1 2
0
-.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5
Bank PPPE
As of 2020-04-15 As of 2020-06-30
8
Figure B.2: Map of Exposure to PPPE
Figure B.2 shows the spatial distribution of our PPPE measure across the U.S. Panel A shows the average exposure of each county to the PPPE.
County exposure to PPPE is computed as the average of the PPPE of each bank with a branch presence in the county. The PPPE of each bank is
weighted by the share of the number of branches of the bank in the county as of June 30th, 2019. Panels B and C show the average exposure
of each zip code in Chicago and New York City, respectively. Zip exposure to PPPE is the weighted average of bank PPPE during the first round
at the ZIP level. The weights are defined by the share of the number of branches of each bank within 10 miles of the center of the respective
ZIP. Data is from the SBA, Call Reports, and FDIC’s Summary of Deposits.
Panel A: County PPPE
−
−
−
−
1RGDWD
Panel B: ZIP PPPE in Chicago Panel C: Zip PPPE in New York City
−
−
−
−
−
− 9
−
−
Figure B.3: Kaplan-Meier Survival Functions
Figure B.3 plots Kaplan-Meier survival functions. The blue line represents the survival function for the group of firms located in zip codes
exposed to banks with low PPPE and the red line plots the survival function for the group of firms located in zip codes exposed to banks with
high PPPE. Data is obtained from the SBA and call reports.
1.00 Low PPPE Zip
High PPPE Zip
Survival Function (Kaplan-Meier)
0.75
0.50
0.25
0.00
20 20 20 02
0
02
0 0
02 20 20 02
0
02
0
02
0
02
0
02
0
02
0 20 20 20 20 20
20 20 20 ,2 ,2 20 20 ,2 20 20 20 20 20
t, 2 ,2 ,2 t, 2 ,2 t, 2
st, th, th, 6th 6th st, st, 6th 6th 5th th, th, th, th, th,
l1 1s
y1 y6 y1 1s 1s
l6 l 11 l1 l2 l2 1th y1 y2 y2 y3 ne
10 15 20 25 30
Ap Ap Ap Ap Ma Ma ne ne ne ne ne
ri ri ri Ap ri Ap Ma Ma Ma Ma Ma Ju
ri ri Ju Ju Ju Ju Ju
Date PPP Received
10
Figure B.4: State Exposure to PPPE and PPP per Establishment
Figure B.4 are scatterplots of state exposure to the volume-based PPPE at the end of Round #1 of the PPP and the
percentage of firms reporting receiving PPP funds at the end of the first round (Panel A) and of state exposure to
the volume-based PPPE at the end of Round #1 and the percentage of firms in each state reporting requesting PPP
funds (Panel B). Data comes from the Census Bureau Small Business Pulse Survey, SBA, Call Reports, Summary
of Deposits, and County Business Patterns.
Panel A: State Exposure to Volume-based PPPE and Share Receiving PPP
100
Share of Small Businesses Requesting PPP (Round 1) AR OK ME
ND
NE MO
MN SD
AL MTWIKS
IA MS
ID KY
LA IN NH
10 20 30 40 50 60 70 80 90
UT
ILWY TN DE
TX AK WV PA
OH VT
VA CO MA RI MI
SC NCCT MD HI
GA
WA
NY
AZ OR FL NM
NJ
NV
CA DC
0
-.2 -.1 0 .1 .2
State Vol. PPPE (Round 1)
Panel B: State Exposure to Volume-Based PPPE and Share Requesting PPP
100
Share of Small Businesses Requesting PPP (Round 1)
GA AR OK ME
MN ILNJ
LACTTXALMO WI MA PA
MSOHRI
CA FLVATN
NY UT NC D MD IN MI
OR SD KS
DE NH
AZ SC
WA IA CO KY HI
NV DC ID
NE NM AK WV
MT VT
WY
10 20 30 40 50 60 70 80 90
0
-.2 -.1 0 .1 .2
State Vol. PPPE (Round 1)
11
Figure B.5: Bank PPPE and Supply Side Factors
Figure B.5 plots the average bank PPPE of each bank decile based on partitioning banks on their ratio of wage expenses to data processing
expenses (Panel A) and on the ratio between the number of SBA loans and small business loans originated in the past three years (Panel B).
Data is obtained from the SBA and call reports
.35
.3
Bank PPPE (Number of Loans)
.05 .1 .15 .2 .25
0
-.05
-.1
1 2 3 4 5 6 7 8 9 10
Labor Intensity Bin (Decile)
.5
.45
Bank PPPE (Number of Loans)
.2 .25 .3 .35 .4
.15
.1
.05
No SBA Loan 1 2 3 4 5 6 7 8 9 10
SBA Relationship Intensity Bin (Decile)
12
Figure B.6: Kaplan-Meier Survival Functions by Bank Relationship
Figure B.6 plots Kaplan-Meier survival functions of the time that took the PPP applicants of a small community bank to receive PPP funds.
The blue line represents the survival function for the group of firms that represented a new relationship for the bank and the red line plots the
survival function for the group of firms that had a banking relationship with the bank that had begun prior to March 2020. Data is obtained
from the SBA, call reports, and from a proprietary dataset obtained from a small community bank.
1.00 New Relationship
Survival Function (Kaplan-Meier)
Existing Relationship
0.75
0.50
0.25
0.00
ApAp
Ap ril
ril 1s t, 2
6th 20
,2 0
Ap
Ap ril
ril
ril 11
16 th,
th, 2
2002
0 2 0
0
ApMaril 21
26
y1 st,
th, 20
20 20
20
20
Ma
Ma y1 y6 st,
st,
1th 2020
20 20
Ma
Ma
Ma y1
y2 ,2
6th 20
1s ,2
t, 2 0
02 0
Ma
Ju y2
y3
ne 6th
1s t,2
, 2 02
02
0 0
0
Ju
Ju ne
ne 105th
th,,220022
200
0
Ju
Ju ne
ne 15
20
25 th,
th, 20
20 20
20
Ju ne
Ju 30
ly 5 th,
th,
th, 20
20 20
20
Ju
Ju
Ju ly 1
ly 1 0th ,2
5th 20
,2200 2 0
Ju
Ju ly 2
ly 2
ly 3 0th 20
,2
5th 20
,2 0
0
0
Au
Au
Au gu
gu st
st 0th 20
,2
4th 20
,2
9th 20 0
0
gu st 14 th,,22002200
Date PPP Received
13
Table B.1: Summary Statistics for Bank Exposure Analysis
Table B.1 reports summary statistics for the analyses in Tables 5 and 6. See the notes to those tables for variable
definitions.
Mean 10th Median 90th N
Outcomes
Bus. Shutdown (ZIP/Week) -0.08 -1.00 0.00 0.00 819834
Hours Worked (ZIP/Week) 0.19 -0.13 0.06 0.73 819834
Nbr. Employees 0.22 -0.14 0.06 0.78 819834
UI Claims (Cnty/Week) -1.78 -4.05 -1.55 0.01 46092
Small Bus. Rev. (Cnty) 0.19 -0.05 0.18 0.47 43930
OI Emp. (Cnty/Week) -0.01 -0.11 -0.00 0.09 17112
Exposure
PPPE in Round 1 (ZIP) 0.02 -0.18 0.01 0.24 35645
PPPE in Round 1 (Cnty) 0.05 -1.28 0.20 1.20 2111
Cross Sectional Characteristics
Pre-PPP Soc. Dist. Index (ZIP) -0.01 -1.03 -0.12 1.21 35645
Pre-PPP Cases per cap ⇥103 (ZIP) 0.05 0.00 0.02 0.09 35645
Pre-PPP Deaths per cap ⇥106 (ZIP) 0.99 0.00 0.00 1.66 35645
Pre-PPP Soc. Dist. Index (Cnty) -0.30 -0.47 -0.34 -0.09 2081
Pre-PPP Cases per cap ⇥103 (Cnty) 0.02 0.00 0.00 0.04 2110
Pre-PPP Deaths per cap ⇥106 (Cnty) 0.37 0.00 0.00 0.00 2110
14
Table B.2: PPP in Bank Websites and PPP Performance
Table B.2 reports results of OLS regressions examining the relation between bank PPPE and the availability of
information about applications to the PPP program in each bank’s internet websites as of April 10th, 2020. The
dependent variables are PPP info, Receiving PPP applications, and Online Application. PPP info is an indicator
variable that takes the value of one if the bank provides any information about the PPP program in its internet
website. Receiving PPP applications is an indicator variable that takes the value of one if the bank states in its
website that is receiving applications to the PPP program as of April 10th. Online Application is an indicator
variable that takes the value of one if the bank receives online applications through its internet website. Bank
PPPE is the bank PPPE measured as of the end of round one of the PPP. Bank Predicted PPPE is the predicted bank
PPPE. The predicted values of bank PPPE are obtained from estimating the empirical specifications of column (8)
of table 2. All specifications include controls for size deciles. Data about the PPP offerings in bank’s websites
was hand-collected from banks’ websites during April 9th and April 10th, 2020. Standard errors are presented
in parentheses, and are clustered at the level of the state. ***, **, and *, represent statistical significance at 1%,
5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
PPP info Receiving PPP applications Online Application
Bank PPPE 0.159⇤⇤⇤ 0.145⇤⇤⇤ 0.154⇤⇤⇤ 0.139⇤⇤⇤ 0.041⇤⇤ 0.032
(0.025) (0.027) (0.024) (0.025) (0.019) (0.020)
Observations 4857 4856 4857 4856 4857 4856
Adjusted R2 0.184 0.196 0.145 0.153 0.055 0.061
State Fixed Effects No Yes No Yes No Yes
Size Deciles Yes Yes Yes Yes Yes Yes
PPP info Receiving PPP applications Online Application
Bank Predicted PPPE 0.832⇤⇤⇤ 0.815⇤⇤⇤ 0.807⇤⇤⇤ 0.788⇤⇤⇤ 0.426⇤⇤⇤ 0.420⇤⇤⇤
(0.094) (0.099) (0.090) (0.094) (0.059) (0.058)
Observations 4852 4851 4852 4851 4852 4851
Adjusted R2 0.194 0.207 0.156 0.164 0.064 0.070
State Fixed Effects No Yes No Yes No Yes
Size Deciles Yes Yes Yes Yes Yes Yes
15
Table B.3: ZIP PPPE in Round 1 and PPP Reallocation (County⇥Year Fixed Effects)
Table B.3 shows the correlation between PPPE and the fraction of establishments receiving PPP loans from dif-
ferent sources in the first and second rounds of the program. The left-hand-side variable in column (1) is the
fraction of establishments within a ZIP and 2-digit NAICS industry that received PPP in the first round in Panel
A and in both rounds in Panel B. Left-hand-side variables in other columns represent a decomposition of the de-
pendent variable in column (1) into the fraction of establishments within a ZIP and 2-digit NAICS industry that
received PPP from local banks, non-local banks, credit unions, FinTech companies, and other nonbanks. ZIP PPPE
(Round 1) is the weighted average of bank PPPE during the first round at the ZIP level. The weights are defined
by the share of the number of branches of each bank within 10 miles of the center of the respective ZIP. ZIP PPPE
is standardized to permit coefficients to be interpreted as the effect of a one-standard-deviation increase in ZIP
PPPE and observations are weighted by the number of eligible establishments in each zip-industry pair. Predicted
PPPE is the weighted average of predicted bank PPPE during the first round at the ZIP level. The predicted values
of bank PPPE are obtained from estimating the empirical specification of column (8) of table 2. The weights are
defined by the share of the number of branches of each bank in the zip code or within 10 miles of the center of
the respective ZIP. Eligible establishment counts equal all establishments in a ZIP less an estimate of the share of
establishments with more than 500 employees (which are not eligible for PPP) plus an estimate of the number of
proprietorships likely to apply for PPP. All regressions include county-by-NAICS fixed effects. Standard errors are
presented in parentheses, and are clustered at the state level. ***, **, and *, represent statistical significance at
1%, 5%, and 10% levels, respectively.
Panel A: Allocation in Round 1
(1) (2) (3) (4) (5) (6)
PPP Loans Relative to All Establishments by Lender Source
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
Zip PPPE (Round #1) 2.237⇤⇤⇤ 2.714⇤⇤⇤ -0.245⇤⇤ -0.090⇤⇤⇤ -0.110⇤⇤⇤ -0.032
(0.229) (0.239) (0.115) (0.026) (0.016) (0.019)
Observations 239411 239411 239411 239411 239411 239411
Adjusted R2 0.477 0.470 0.225 0.315 0.374 0.192
State⇥Industry FE Yes Yes Yes Yes Yes Yes
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
Predicted PPPE 0.775⇤⇤⇤ 0.786⇤⇤⇤ 0.034 -0.008 -0.037⇤⇤ -0.002
(0.185) (0.250) (0.139) (0.034) (0.014) (0.004)
Observations 239411 239411 239411 239411 239411 239411
Adjusted R2 0.475 0.466 0.224 0.315 0.373 0.192
County⇥Industry FE Yes Yes Yes Yes Yes Yes
Panel B: Allocation in Round 1 and 2
(1) (2) (3) (4) (5) (6)
PPP Loans Relative to All Establishments by Lender Source
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
Zip PPPE (Round #1) 0.321 2.075⇤⇤⇤ -0.744⇤⇤ -0.264⇤⇤⇤ -0.612⇤⇤⇤ -0.133⇤⇤⇤
(0.418) (0.468) (0.335) (0.064) (0.127) (0.048)
Observations 239411 239411 239411 239411 239411 239411
Adjusted R2 0.466 0.483 0.274 0.413 0.258 0.215
State⇥Industry FE Yes Yes Yes Yes Yes Yes
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
Predicted PPPE 0.150 0.345 0.104 -0.030 -0.249⇤⇤ -0.021⇤
(0.291) (0.443) (0.323) (0.092) (0.105) (0.012)
Observations 239411 239411 16 239411 239411 239411 239411
Adjusted R2 0.466 0.482 0.273 0.412 0.257 0.215
County⇥Industry FE Yes Yes Yes Yes Yes Yes
Table B.4: Differences in Characteristics of PPP Applicants With and Without Previous Bank Relationship
Table B.4 uses a proprietary dataset obtained from a small community bank to examine the characteristics of firms
that received PPP funds from that bank. We examine the differences in characteristics of PPP applicants with and
without a previous bank relationship. We define a new bank relationship as a firm that did not have any account
with the bank prior to the start of the pandemic. Data is from a proprietary dataset of a small community bank,
SBA, and Call Reports.
New Bank Relationship Existing Bank Relationship
Mean Std. Dev. N Mean Std. Dev. N Diff t-stat
Zip PPPE (Round #1) -0.0116 0.0884 239 0.0280 0.0944 584 -0.0396 -5.568
Predicted PPPE (Round #1) 0.237 0.0245 239 0.248 0.0264 585 -0.0115 -5.776
Zip Exposure to Wells Fargo 0.00750 0.0275 239 0.00390 0.0183 584 0.00360 2.210
Distance to Bank 43.07 210.8 239 15.10 93.18 585 27.97 2.641
PPP Loan Amount 59670 116015 239 192471 418803 585 -132801 -4.825
Number of Employees 7.864 14.17 236 19.58 42.51 569 -11.72 -4.140
17
Appendix C. Wells Fargo Appendix
This appendix repeats the main empirical analyses of the paper using the local exposure
to Wells Fargo as our main variable of interest. Wells Fargo did not accept PPP applications
out of concerns that it might breach the asset cap restriction imposed by the Office of the
Comptroller of the Currency in the aftermath of the 2016 fake accounts scandal. On April
8th, the Federal Reserve issued a press release exempting PPP loans from counting toward the
total assets formula used to determine compliance with the asset cap restriction. Given that
this restriction was externally imposed and Wells Fargo is a very large depository institution in
the United States, we use each local area exposure to Wells Fargo’s branches as an alternative
measure of exposure to supply-side difficulties in accessing the PPP. Table C.1 and Figure C.1
show that local areas that were more exposed to Wells Fargo branches received fewer loans
during the first round. Figure C.2 and Table C.2 use local differences in exposure to Wells
Fargo to examine the effect of the program on employment outcomes using Homebase data.
Figure C.3 and Table C.3 use local differences in exposure to Wells Fargo to examine the effect
of the program on employment outcomes using county unemployment filings, Womply, and
Opportunity Insights Tracker data. Overall, the estimated results when we use this alternative
empirical strategy are quantitively and qualitatively similar to those presented in the paper.
18
Figure C.1: Exposure to Wells Fargo and PPP Coverage over Time
Figure C.1 plots binned scatter plots of the average fraction of small business establishments that received a PPP
loan versus exposure to Wells Fargo branches at the Zip level. Eligible establishment counts equal all establish-
ments in a ZIP less an estimate of the share of establishments with more than 500 employees (which are not
eligible for PPP) plus an estimate of the number of proprietorships likely to apply for PPP. Both variables are de-
meaned at the state level to present the within-state relationship. Data come from SBA, Call Reports, Summary
of Deposits, and County Business Patterns.
PPP Loans Relative to
Eligible Establishments (Residualized)
.1 .2 .3 .4 .5 .6 .7 .8
-.4 -.2 0 .2 .4
Wells Fargo Exposure (Residualized)
04/12-04/18 04/26-05/02 05/10-05/16
05/24-05/30 06/07-06/13 06/21-06/27
19
W W W
ee ee ee
k1 Ratio of Employees Relative to January k1 Ratio of Hours Worked Relative to January k1 Share of Businesses Shutdown
W : Jan W : Jan W : Jan
ee 0 .2 .4 .6 .8 1 ee 0 .2 .4 .6 .8 1 ee 0 .1 .2 .3 .4
k 2 19 - k 2 19 - k 2 19 -
W W W
ee : Ja Jan ee : Ja Jan ee : Ja Jan
k 3 n2 25 k 3 n2 25 k 3 n2 25
W 6 W 6 W 6
ee : Fe -Fe ee : Fe -Fe ee : Fe -Fe
W k 4: b02 b1 W k 4: b02 b1 W k 4: b02 b1
ee - ee - ee -
k 5 Feb Feb k 5 Feb Feb k 5 Feb Feb
W 9 8 W 9 8 W 9 8
ee : Fe -Fe ee : Fe -Fe ee : Fe -Fe
k 6 b1 b k 6 b1 b k 6 b1 b
: F 6-F 15 : F 6-F 15 : F 6-F 15
W eb eb W eb eb W eb eb
e 2 2 e 2 2 e 2 2
W ek 7 3-F 2 W ek 7 3-F 2 W ek 7 3-F 2
ee :M eb ee :M eb ee :M eb
ar 29 ar 29 ar 29
W k 8:
ee 1 W k 8:
ee 1 W k 8:
ee 1
Ma -Ma Ma -Ma Ma -Ma
W k 9:
ee r8 r7 W k 9:
ee r8 r7 W k 9:
ee r8 r7
k 1 Ma -Ma k 1 Ma -Ma k 1 Ma -Ma
r r r r r r
W 0: M 15-M 14 W 0: M 15-M 14 W 0: M 15-M 14
ee ee ee
k 1 ar2 ar2 k 1 ar2 ar2 k 1 ar2 ar2
W 1 2 - 1 W 1 2 - 1 W 1 2 - 1
ee : M Ma ee : M Ma ee : M Ma
W k 12 ar29 r28 Initial Lockdown Orders W k 12 ar29 r28 Initial Lockdown Orders W k 12 ar29 r28 Initial Lockdown Orders
The panels on the left of Figure C.2 shows the ratio of hours worked over time, the percent of businesses shut down,
ee ee ee
k 1 : Ap -Apr k 1 : Ap -Apr k 1 : Ap -Apr
W 3 r 5 4 W 3 r5 4 W 3 r5 4
ee :A - PPP Launch ee :A - PPP Launch ee :A - PPP Launch
k pr Apr k pr Apr k pr Apr
W 14: 12- 11 W 14: 12- 11 W 14: 12- 11
ee Ap Ap ee Ap Ap ee Ap Ap
k r r k r r k r r
W 15: 19-A 18
e A 1st Round Exhausted W 15: 19-A 18
e A 1st Round Exhausted W 15: 19-A 18
e A 1st Round Exhausted
W ek 1 pr2 pr25 W ek 1 pr2 pr25 W ek 1 pr2 pr25
ee 6: 6- ee 6: 6- ee 6: 6-
W k 17 Ma May W k 17 Ma May W k 17 Ma May
ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins
k 1 Ma 3-M k 1 Ma 3-M k 1 Ma 3-M
W y a W y a W y a
Low Wells Fargo Exposure Low Wells Fargo Exposure Low Wells Fargo Exposure
ee 8: M 10-M y9 ee 8: M 10-M y9 ee 8: M 10-M y9
k1 ay a k1 ay a k1 ay a
W 9: M 17- y16 W 9: M 17- y16 W 9: M 17- y16
ee
k2 a y
Ma ee
k2 a y
Ma ee
k2 a y
Ma
W 0 24 y23
W 0 24 y23
W 0 24 y23
ee : Ma -Ma ee : Ma -Ma ee : Ma -Ma
W k 21 y 3 y30 W k 21 y 3 y30 W k 21 y 3 y30
ee 1 ee 1 ee 1
k 2 : Ju -Jun k 2 : Ju -Jun k 2 : Ju -Jun
W n W n W n
Time Series Time Series
ee 2: J 7-Ju 6 ee 2: J 7-Ju 6 ee 2: J 7-Ju 6
k 2 un n1 k 2 un n1 k 2 un n1
1 1 1
W 3: J 4-Ju 3 W 3: J 4-Ju 3 W 3: J 4-Ju 3
ee un ee un ee un
k 21 n20 k 21 n20 k 21 n20
W 24: -J W 24: -J W 24: -J
ee Ju un2 ee Ju un2 ee Ju un2
W k 25 n28 7 W k 25 n28 7 W k 25 n28 7
ee - ee - ee -
k 2 : Ju Jul4 k 2 : Ju Jul4 k 2 : Ju Jul4
W l5- W l5- W l5-
ee 6: J ee 6: J ee 6: J
k u Jul1 k u Jul1 k u Jul1
W 27: l12-J 1 W 27: l12-J 1 W 27: l12-J 1
ee ee ee
k 2 Jul1 ul18 k 2 Jul1 ul18 k 2 Jul1 ul18
W 9 W 9 W 9
and the ratio of number of employees splitting the sample into firms with above- and below-median exposure to
ee 8: Ju -Ju ee 8: Ju -Ju ee 8: Ju -Ju
W k 29 l26 l25 W k 29 l26 l25 W k 29 l26 l25
ee : -A ee : -A ee : -A
W k 3 0 Au g u g 1 W k 30 Aug ug1 W k 30 Aug ug1
ee 2 ee 2 ee 2
k 3 : Au -Au k 3 : Au -Au k 3 : Au -Au
W g g W g g W g g
ee 1: A 9-A 8 ee 1: A 9-A 8 ee 1: A 9-A 8
k 3 ug ug k 3 ug ug k 3 ug ug
2: 1 1 2: 1 1 2: 1 1
Au 6-Au 5 Au 6-Au 5 Au 6-Au 5
g2 g g2 g g2 g
3- 22 3- 22 3- 22
Au Au Au
g2 g2 g2
High Wells Fargo Exposure High Wells Fargo Exposure High Wells Fargo Exposure
9 9 9
Panel C. Change in Nbr. Employees
20
We We We
e k1 Coefficient (Wells Fargo Exposure) e k1 Coefficient (Wells Fargo Exposure) e k1 Coefficient (Wells Fargo Exposure)
0: 0: 0:
We
Ma
r22 -.02 0 .02 .04 We
Ma
r22 -.02 0 .02 .04 We r22
Ma -.02 0 .02 .04
e k1 -M e k1 -M e -M k1
1: ar2 1: ar2 ar2 1:
8 8 8
Panel B. Change in Hours Worked
We Ma We Ma We Ma
e r29 e r29 ek r29
k1 -Ap k1 -Ap 12 -Ap
We
2:
Ap r4 We
2:
Ap r4 We :A r4
e k1 r5-
Ap PPP Launch e k1 r5 -Ap PPP Launch ek pr5
-Ap PPP Launch
13
Wells Fargo branches. The panels on the right of Figure C.2 plot coefficients and standard errors of regressions
3: r11 3: r11 :A r11
We Ap We Ap We pr1
e r12 e r12 ek 2-A
k1 -Ap k1 -Ap 14 pr1
4: r18 4: r18 :A
We Ap We Ap We pr1 8
r19 1st Round Exhausted r19 1st Round Exhausted 1st Round Exhausted
e k1 -Ap e k1 -Ap ek 9-A
5: r25 5: r25 15 pr2
Ap Ap :A 5
We r26 We r2 We pr2
e k1 -M e k1 6-M ek 6-M
ay ay 16 ay
Panel A. Business Shutdowns
6: 2 6: 2 :M 2
We
e Ma 2nd Round Begins We
e Ma 2nd Round Begins We
ek ay 2nd Round Begins
k1 y3 k1 y3
7: -M 7: -M 17
:
3-M
ay ay ay
We Ma 9 We Ma 9 We Ma 9
e k1 y1
0 e k1 y1
0 ek y1
8: -M 8: -M 18 0-M
Ma ay Ma ay :M ay
We 16 We 16 We ay 16
e k1
y1
7-M e k1
y1
7-M ek 17
9: 9: 19 -M
Ma ay Ma ay :M ay
We y2 23 We y2 23 We ay 23
e 4-M e 4-M ek 24
k2 k2 20 -M
0: ay 0: ay :M ay
Ma 30 Ma 30 30
We y3 We y3 We ay
e k2 1-J e k2 1-J ek 31
-Ju
1: un 1: un 21
We Ju 6 We Ju 6 We :J
un
n6
e k2
n7
-Ju e k2
n7
-Ju ek 7-J
2: n1 2: n1 22
:J u n
We Ju
n1 3 We Ju
n1 3 We un 13
e k2 4-J e k2 4-J ek 14
-Ju
3: un 3: un 23 n2
Ju 20 Ju 20 :J 0
n2 n2 un
We We We 21
e 1-J
un e 1-J
un ek -Ju
k2
4: 27
k2
4: 27 24 n 27
:J
investigating the impact of exposure to Wells Fargo on employment and firm outcomes. The regressions are similar
Ju Ju un
We n2 We n2 We 28
e k2 8-J e k2 8-J ek -Ju
5: ul4 5: ul4 25 l4
We Ju We Ju We :J
e l5- e l5- ek ul5
Ju Ju -Ju
Time Series Local Projection Local Projection
Local Projection
k2 k2 26
6: l11 6: l11 :J l11
We Ju We Ju We ul1
l12 l12
e k2 -Ju e k2 -Ju ek 2-J
7: l18 7: l18 27 ul1
Ju Ju :J 8
We l19 We l19 We ul1
e k2 -Ju e k2 -Ju ek 9-J
l25 l25 28 ul2
8:
Ju
8:
Ju :J 5
We l26 We l26 We ul2
e k2 -Au e k2 -Au ek 6-A
Figure C.2: Wells Fargo Exposure and Evolution of Homebase Employment Outcomes
g1 g1 29 ug
We
9:
Au We
9:
Au We :A 1
e k3 g2 e k3 g 2-A ek ug
2-A
-Au ug 30 u
0: g8 0: 8 :A g8
We Au We Au We
e k3 g9 e k3 g 9-A ek ug
9-A
-Au ug 31 ug
1: g1 1: :A
We Au
g1 5 We Au
g1
15 We ug 15
e k3 6-A e k3 6-A ek 16
-Au
2: ug 2: ug 32 g2
Au 22 Au 22 :A 2
g2 g2 ug
3-A 3-A 23
ug ug -Au
29 29 g2
9
to those of Figure 8 in the main paper. Data are from SBA, Homebase, County Business Patterns.
W W W
ee ee ee
k1 Decline Relative to January k1 Decline Relative to January k1 Initial UI Claims/January Employment
W : Jan W : Jan W : Jan
ee -.25 -.2 -.15 -.1 -.05 0 ee -.6 -.4 -.2 0 .2 ee 0 1% 2% 3% 4%
k 2 19 - k 2 19 - k 2 19 -
W W W
ee : Ja Jan ee : Ja Jan ee : Ja Jan
k 3 n2 25 k 3 n2 25 k 3 n2 25
W 6 W 6 W 6
ee : Fe -Fe ee : Fe -Fe ee : Fe -Fe
W k 4: b02 b1 W k 4: b02 b1 W k 4: b02 b1
ee - ee - ee -
k 5 Feb Feb k 5 Feb Feb k 5 Feb Feb
W 9 8 W 9 8 W 9 8
ee : Fe -Fe ee : Fe -Fe ee : Fe -Fe
k 6 b1 b k 6 b1 b k 6 b1 b
: F 6-F 15 : F 6-F 15 : F 6-F 15
W eb eb W eb eb W eb eb
e 2 2 e 2 2 e 2 2
W ek 7 3-F 2 W ek 7 3-F 2 W ek 7 3-F 2
ee :M eb ee :M eb ee :M eb
The panels on the left of Figure C.3 show the evolution of the ratio of weekly initial unemployment filing claims
ar 29 ar 29 ar 29
W k 8:
ee 1 W k 8:
ee 1 W k 8:
ee 1
Ma -Ma Ma -Ma Ma -Ma
W k 9:
ee r8 r7 W k 9:
ee r8 r7 W k 9:
ee r8 r7
k 1 Ma -Ma k 1 Ma -Ma k 1 Ma -Ma
r r r r r r
W 0: M 15-M 14 W 0: M 15-M 14 W 0: M 15-M 14
ee ee ee
k 1 ar2 ar2 k 1 ar2 ar2 k 1 ar2 ar2
W 1 2 - 1 W 1 2 - 1 W 1 2
ee : M Ma ee : M Ma e : M -M 1
a
Initial Lockdown Orders W ek 1 ar2 r28 Initial Lockdown Orders
W k 12 ar29 r28
ee
W k 12 ar29 r28
ee ee 2 9
k 1 : Ap -Apr k 1 : Ap -Apr k 1 : Ap -Apr
W 3 r 5 4 W 3 r 5 4 W 3 : r5 4
ee :A - ee :A - PPP Launch ee - PPP Launch
k pr Apr k pr Apr k 1 Apr1 Apr1
W 14: 12- 11 W 14: 12- 11 W 4 2 1
ee Ap Ap ee Ap Ap ee : Ap -Ap
k r r k r r k r1 r1
W 15: 19-A 18
e A
W 15: 19-A 18
e A 1st Round Exhausted We 15: A 9-A 8 1st Round Exhausted
W ek 1 pr2 pr25 W ek 1 pr2 pr25 W ek 1 pr2 pr25
ee 6: 6- ee 6: 6- ee 6: 6-
W k 17 Ma May W k 17 Ma May W k 17 Ma May
ee : y 2 ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins
k 1 Ma 3-M k 1 Ma 3-M k 1 Ma 3-M
W y a W y a W y a
Low Wells Fargo Exposure Low Wells Fargo Exposure Low Wells Fargo Exposure
ee 8: M 10-M y9 ee 8: M 10-M y9 ee 8: M 10-M y9
k1 ay a k1 ay a k1 ay a
W 9: M 17- y16 W 9: M 17- y16 W 9: M 17- y16
a a a
at the county level and total county employment (Panel A), the change in aggregate small business revenue at
ee Ma ee Ma ee Ma
k2 y k2 y k2 y
W 0 24 y23
W 0 24 y23
W 0 24 y23
ee : Ma -Ma ee : Ma -Ma ee : Ma -Ma
W k 21 y 3 y30 W k 21 y 3 y30 W k 21 y 3 y30
ee 1 ee 1 ee 1
k 2 : Ju -Jun k 2 : Ju -Jun k 2 : Ju -Jun
W n W n W n
Time Series Time Series
ee 2: J 7-Ju 6 ee 2: J 7-Ju 6 ee 2: J 7-Ju 6
k 2 un n1 k 2 un n1 k 2 un n1
1 1 1
W 3: J 4-Ju 3 W 3: J 4-Ju 3 W 3: J 4-Ju 3
ee un ee un ee un
k 21 n20 k 21 n20 k 21 n20
W 24: -J W 24: -J W 24: -J
ee Ju un2 ee Ju un2 ee Ju un2
W k 25 n28 7 W k 25 n28 7 W k 25 n28 7
ee - ee - ee -
k 2 : Ju Jul4 k 2 : Ju Jul4 k 2 : Ju Jul4
W l5- W l5- W l5-
ee 6: J ee 6: J ee 6: J
k u Jul1 k u Jul1 k u Jul1
W 27: l12-J 1 W 27: l12-J 1 W 27: l12-J 1
ee ee ee
k 2 Jul1 ul18 k 2 Jul1 ul18 k 2 Jul1 ul18
W 9 W 9 W 9
ee 8: Ju -Ju ee 8: Ju -Ju ee 8: Ju -Ju
W k 29 l26 l25 W k 29 l26 l25 W k 29 l26 l25
ee : -A ee : -A ee : -A
W k 3 0 Au g u g 1 W k 3 0 Au g u g 1 W k 30 Aug ug1
ee 2 ee 2 ee 2
k 3 : Au -Au k 3 : Au -Au k 3 : Au -Au
the county level relative to January (Panel B), and the ratio of county employment relative to January (Panel C)
W g g W g g W g g
ee 1: A 9-A 8 ee 1: A 9-A 8 ee 1: A 9-A 8
k 3 ug ug k 3 ug ug k 3 ug ug
2: 1 1 2: 1 1 2: 1 1
Au 6-Au 5 Au 6-Au 5 Au 6-Au 5
g2 g g2 g g2 g
3- 22 3- 22 3- 22
Au Au Au
g2 g2 g2
High Wells Fargo Exposure High Wells Fargo Exposure High Wells Fargo Exposure
9 9 9
Panel B. Small Business Revenues
21
We We We
e k1 Coefficient (Wells Fargo Exposure) e k1 Coefficient (Wells Fargo Exposure) e k1 Coefficient (Wells Fargo Exposure)
0: 0: 0:
We
Ma
r22 -.01 0 .01 .02 .03 We
Ma
r22 -.04 -.02 0 .02 .04 We
Ma
r22 -.5 0 .5
e k1 -M e k1 -M e k1 -M
1: ar2 1: ar2 1: ar2
Ma 8 Ma 8 Ma 8
splitting the sample into firms with above- and below-median exposure to Wells Fargo branches. The panels on
We r29 We r29 We r29
e k1 -Ap e k1 -Ap e k1 -Ap
We
2:
Ap r4 We
2:
Ap r4 We
2:
Ap r4
e k1 r5-
Ap PPP Launch PPP Launch e k1 r5-
Ap e k1 r5 -Ap PPP Launch
Panel A. UI Claims
3: r11 3: r11 3: r11
Ap Ap Ap
We
e r12 We
e r12 We
e r1 2-A
k1 -Ap k1 -Ap k1 pr1
4: r18 4: r18 4:
Ap Ap Ap 8
We r19 PPP Funds Exhausted PPP Funds Exhausted
We r19 We r1 PPP Funds Exhausted
e k1 -Ap ek -Ap e k1 9-A
5: r25 15 r25 5: pr2
Ap :A
pr2 Ap 5
We r26 We We r2
e k1 -M e k1 6-M e k1 6-M
6: ay 6: ay 6: ay
2 2 2
Panel C. County Employment
We
e Ma PPP 2nd Round Begins PPP 2nd Round BeginsWee Ma We
e Ma PPP 2nd Round Begins
k1 y3 k1 y3 k1 y3
7: -M 7: -M 7: -M
ay ay ay
We Ma 9 We Ma 9 We Ma 9
e k1 y1
0 e k1 y1
0 e k1 y1
8: -M 8: -M 8:
0-M
Ma ay Ma ay Ma ay
We 16 We 16 We 16
e k1
y1
7-M e k1
y1
7-M e k1
y1
7-M
9: ay 9: ay 9: ay
Ma 23 Ma 23 Ma 23
We y2 We y2 We y2
e e e
the right of Figure C.3 plot coefficients and standard errors of regressions investigating the impact of exposure to
k2 4-M k2 4-M k2 4-M
0: ay 0: ay 0: ay
Ma 30 Ma 30 Ma 30
We y3 We y3 We y3
e k2 1-J e k2 1-J e k2 1-J
1: un 1: un 1: un
We Ju 6 We Ju 6 We Ju 6
e k2
n7
-Ju e k2
n7
-Ju e k2
n7
-Ju
2: n1 2: n1 2: n1
We Ju
n1 3 We Ju
n1 3 We Ju
n1 3
e k2 4-J e k2 4-J e k2 4-J
3: un 3: un 3: un
Ju 20 Ju 20 Ju 20
We n2 We n2 We n2
e k2
1-J
un e k2
1-J
un e k2
1-J
un
4: 27 4: 27 4: 27
Ju Ju Ju
We n2 We n2 We n2
e k2 8-J e k2 8-J e k2 8-J
5: ul4 5: ul4 5: ul4
We Ju We Ju We Ju
e l5-
Ju e l5-
Ju e l5-
Ju
Local Projection
Time Series Local Projection
Local Projection
k2 k2 k2
6: l11 6: l11 6: l11
We Ju We Ju We Ju
l12 l12 l12
e k2 -Ju e k2 -Ju e k2 -Ju
7: l18 7: l18 7: l18
We Ju We Ju We Ju
l19 l19 l19
Wells Fargo on the same outcomes. The regressions are similar to those of Figure 10 in the main paper. Data are
e k2 -Ju e k2 -Ju e k2 -Ju
8: l25 8: l25 8: l25
We Ju We Ju We Ju
l26 l26 l26
e k2 -Au e k2 -Au e k2 -Au
g1 g1 g1
Figure C.3: Wells Fargo Exposure and Other Local Labor Market and Economic Effects
9: 9: 9:
We Au We Au We Au
e k3 g2 e k3 g2 e k3 g 2-A
-Au -Au ug
0: g8 0: g8 0: 8
We Au We Au We Au
e k3 g9 e k3 g9 e k3 g 9-A
-Au -Au ug
1: g1 1: g1 1:
We Au
g1 5 We Au
g1 5 We Au
g
15
e k3 6-A e k3 6-A e k3 16
-Au
2: ug 2: ug 2: g2
Au 22 Au 22 Au 2
g2 g2 g 23
3-A 3-A -Au
ug ug g2
29 29 9
from SBA, Homebase, County Business Patterns.
Table C.1: Wells Exposure in Round 1 and PPP Reallocation across Funding Sources
Table C.1 shows the correlation between local exposure to Wells Fargo branches and the fraction of establishments
receiving PPP loans from different sources in the first and second rounds of the program. The left-hand-side
variable in column (1) is the fraction of establishments within a ZIP and 2-digit NAICS industry that received PPP
in the first round in Panel A and in both rounds in Panel B. Left-hand-side variables in other columns represent
a decomposition of the dependent variable in column (1) into the fraction of establishments within a ZIP and
2-digit NAICS industry that received PPP from local banks, non-local banks, credit unions, FinTech companies,
and other nonbanks. % Wells (Zip 10 miles) is the share of the number of branches in the zip code or within 10
miles of the center of the respective zip code that belong to Wells Fargo. % Wells (Zip 10 miles) is standardized to
permit coefficients to be interpreted as the effect of a one-standard-deviation increase in exposure to Wells Fargo
and observations are weighted by the number of establishment counts in each zip-industry pair. All regressions
include state-by-NAICS fixed effects. Standard errors are presented in parentheses, and are clustered at the state
level. ***, **, and *, represent statistical significance at 1%, 5%, and 10% levels, respectively.
Panel A: Allocation in Round 1
(1) (2) (3) (4) (5) (6)
PPP Loans Relative to All Establishments by Lender Source
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
% Wells (Zip 10 miles) -0.017⇤⇤⇤ -0.014⇤⇤⇤ -0.004⇤⇤⇤ 0.000 0.001⇤⇤⇤ 0.000
(0.004) (0.004) (0.001) (0.001) (0.000) (0.000)
Observations 251511 251511 251511 251511 251511 251511
Adjusted R2 0.356 0.329 0.126 0.166 0.285 0.043
State⇥Industry FE Yes Yes Yes Yes Yes Yes
Panel B: Allocation in Round 1 and 2
(1) (2) (3) (4) (5) (6)
PPP Loans Relative to All Establishments by Lender Source
PPP/Est (%) Local Banks Non-Local Banks Credit Unions FinTech Nonbanks
% Wells (Zip 10 miles) 0.008⇤ 0.002 -0.002 0.002 0.006⇤⇤⇤ 0.001⇤⇤⇤
(0.005) (0.004) (0.003) (0.002) (0.002) (0.000)
Observations 251511 251511 251511 251511 251511 251511
Adjusted R2 0.399 0.392 0.187 0.219 0.194 0.090
State⇥Industry FE Yes Yes Yes Yes Yes Yes
22
Table C.2: Wells Exposure and Homebase Employment Outcomes
Table C.2 reports the results of OLS regressions examining the relation between exposure to Wells Fargo and the difference between a firm’s
average employment outcomes in the two weeks prior to the launch of PPP and the firm’s outcomes in each of the following months. The
left-hand-side variable in Panel A, Bus. Shutdown, is the difference between the firm’s shutdown status in a week and its average shutdown
status in weeks 10 and 11, where shutdown status takes a value of one if the business reported zero hours worked over the entire week. The
left-hand-side variable in Panel B, Hours Worked, is the difference in the ratio of hours worked in each establishment in a week and the
average ratio of hours worked in that establishment in weeks 10 and 11. The ratio of hours worked in each establishment is measured as the
hours worked in that week relative to the hours worked in that same establishment during the last two weeks of January. Nbr. Employees, is
the difference in the ratio of the number of employees in each establishment in a week and the average ratio of number of employees in that
establishment in weeks 10 and 11. The ratio of number of employees in each establishment is measured as the number of distinct employees
that worked in the establishment in that week relative to the number of distinct employees working in that same establishment during the last
two weeks of January. % Wells is the share of the number of branches in the zip code or within 10 miles of the center of the ZIP that belongs
to Wells Fargo. I(Month=‘M’), where M = {Apr il, M a y, June, Jul y, August} are indicator variables for the weeks that span those respective
months. Other control variables include interactions between the median household income, social distance index, COVID cases per capita
and deaths per capita measured as of week 9 interacted with the indicator variables for April, May, June, July, and August and controls for
the average tier 1 capital and core deposit ratios of all banks within the zip code or within a 10 miles radius of the zip code also interacted
with the indicator variables for April, May, June, July, and August. Standard errors are clustered at the state level. ***, **, and *, represent
statistical significance at 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Bus. Shutdown Hours Worked Nbr. Employees
% Wells ⇥ I(Month=April) -0.004⇤ -0.004⇤ 0.002 0.003 0.002 0.001
(0.003) (0.003) (0.002) (0.002) (0.002) (0.002)
% Wells ⇥ I(Month=May) 0.002 0.004 -0.012⇤⇤ -0.009⇤⇤ -0.011⇤⇤ -0.009⇤⇤
(0.003) (0.003) (0.005) (0.004) (0.004) (0.004)
% Wells ⇥ I(Month=June) -0.002 0.003 -0.021⇤⇤⇤ -0.018⇤⇤⇤ -0.017⇤⇤ -0.016⇤⇤
(0.004) (0.004) (0.007) (0.007) (0.007) (0.007)
% Wells ⇥ I(Month=July) -0.005 0.001 -0.015⇤⇤ -0.015⇤⇤⇤ -0.010 -0.013⇤
(0.003) (0.003) (0.006) (0.005) (0.007) (0.007)
% Wells ⇥ I(Month=August) -0.006 0.000 -0.014⇤⇤ -0.013⇤⇤ -0.007 -0.010
(0.004) (0.004) (0.006) (0.005) (0.007) (0.007)
Observations 819834 819834 819834 819834 819834 819834
Adjusted R2 0.058 0.602 0.133 0.629 0.110 0.571
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables No Yes No Yes No Yes
Firm Fixed Effects No Yes No Yes No Yes
23
Table C.3: Wells Fargo Exposure and Local Labor Market and Economic Effects
Table C.3 reports the results of OLS regressions examining the relation between exposure to Wells Fargo and county-level unemployment
filings, small business revenue from Womply, and employment growth from Opportunity Insights. UI Claims is the difference between
the county unemployment filings during a week and the average unemployment filings in the county in weeks 10 and 11. Small Business
Revenue is the difference between the county aggregate change in small business revenue relative to January and the average change in small
business revenue in weeks 10 and 11 relative to January. Aggregate change in small business revenue is from Womply. OI Emp., is the
difference between county employment growth relative to January in a week and the average county employment growth relative to January
in weeks 10 and 11. The county-level employment data come from Opportunity Insights. County % Wells is the share of the number of Wells
Fargo branches in the county. I(Month=‘M’), where M = {Apr il, M a y, June, Jul y, August} are indicator variables for the weeks that span
those respective months. Other control variables include interactions between the median household income, social distance index, COVID
cases per capita and deaths per capita measured as of week 9 interacted with the indicator variables for April, May, June, July, and August
and controls for the average tier 1 capital and core deposit ratios of all banks within the county also interacted with the indicator variables
for April, May, June, July, and August. Standard errors are clustered at the state level. ***, **, and *, represent statistical significance at 1%,
5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
UI claims Small Bus. Rev. OI Emp.
County % Wells ⇥ I(Month=April) -0.020 -0.003 -0.014⇤⇤⇤ -0.002 0.002 0.001
(0.072) (0.053) (0.003) (0.003) (0.004) (0.004)
County % Wells ⇥ I(Month=May) -0.006 0.013 -0.023⇤⇤ -0.005 -0.001 0.002
(0.074) (0.058) (0.009) (0.008) (0.004) (0.004)
County % Wells ⇥ I(Month=June) -0.006 0.025 -0.011 -0.006 -0.004 0.001
(0.071) (0.058) (0.010) (0.010) (0.006) (0.003)
County % Wells ⇥ I(Month=July) -0.094 0.016 -0.003 -0.001 -0.004 0.001
(0.075) (0.059) (0.006) (0.006) (0.007) (0.003)
County % Wells ⇥ I(Month=August) -0.147⇤ -0.047 0.001 -0.001 -0.003 0.003
(0.080) (0.062) (0.007) (0.007) (0.008) (0.004)
Observations 46092 45533 43886 43863 17112 17112
Adjusted R2 0.744 0.949 0.489 0.734 0.682 0.878
State⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables No Yes No Yes No Yes
County Fixed Effects No Yes No Yes No Yes
24
Appendix D. Pre-Targeting Appendix
This Appendix expands the pre-targeting analysis of section 5.5 using the Homebase data
and other data sources. The purpose of this section is to show that the pre-targeting results
presented in the paper generalize to other outcomes and other levels of aggregation. Figure
D.1 shows that the Homebase pre-PPP employment outcomes at the state level are worse in
states that received a smaller PPP allocation in the first round. Figure D.2 shows that similar
patterns hold when we examine other state-level outcomes such as pre-PPP initial unemploy-
ment insurance filings, pre-PPP change in small business revenue from Womply, and pre-PPP
change in employment from the Opportunity Insights tracker. Figure D.3 shows that the pre-
PPP disease spread was greater in states that received a smaller allocation of the PPP during
the first round and Figure D.4 shows that states receiving greater PPP allocation in the first
round had less pre-PPP social distancing and issued shelter-in-place orders later. Figures D.5
and D.6 show similar relations between pre-PPP employment outcomes and PPP allocation
during the first round when we implement the analysis at the county level. Figure D.7 shows
that the pre-PPP evolution of Homebase employment outcomes did not vary with the exposure
of firms to banks with supply-side constraints. When we partition firms based on deciles of
predicted PPPE, we see similar pre-PPP levels of business shutdown, declines in hours worked,
and declines in number of workers.
25
Figure D.1: Pre-PPP Homebase Employment Outcomes and PPP Allocation by State (Round 1)
Figure D.1 presents scatterplots of the share of businesses in each state that shutdown in the week of March 22nd
to March 28th (Panel A), of the decline in hours worked in each state in the week of March 22nd to March 28th
relative to a January Baseline (Panel B), and of the decline in the number of employee counts in each state in
the week of March 22nd to March 28th relative to a January baseline (Panel C). The figures on the left plot the
pre-PPP state-level employment outcomes from Homebase against the number of PPP loans received by small
businesses in each state during the first round of the program divided by the total number of small businesses in
the state. The figures on the right plot the same pre-PPP employment outcomes and the state-level PPPE measure.
Data is from Homebase, SBA, and Call Reports.
Panel A. Business Shutdowns
Fraction Receiving PPP State PPPE
.5 .5
Share of Businesses Shutdown during 03/22 to 03/28 (Pre-PPP) Share of Businesses Shutdown during 03/22 to 03/28 (Pre-PPP)
NY NY
PA PA
.4 ME .4 ME
IL IL
DC
NV NV DC
NJ WV RI NJ RI WV
CA CT CA CT
OR OR MA
MA NH NH
MI KY MI
KY
MD VT .3 MD LA VT
.3 OHLA OH
DE WI
MN WA DE MN WI
WA
IN HI IN HI
FL FL
TX
NC NM GA TX CO IA NC GA CO NM IA
MO SC VA MO
VASC AL AL
.2 AZ
.2 AZ AR
AR MS
TN MS
MT TN MT
KS KS
OK OK
WY WY
ID NE ID NE
AK AK
UT UT
.1 ND .1 ND
SD SD
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel B. Decline in Hours Worked
Fraction Receiving PPP State PPPE
.8 SD .8 SD
Ratio of Hours Worked relative to January (Pre-PPP) Ratio of Hours Worked relative to January (Pre-PPP)
.7 .7
UT UT
NE ND NE ND
.6 OK .6 OK
ID ID
MS MS
KSMT KS
MT
IA IA
AL AK AL MO
AZ AK MO .5 AZ
.5 SC AR SC GA TN KY AR
GA TN KY
NC MN NC TX MN
WY
VA TX WY VA
FL FL CO
CO DE IN
MD DE IN MD
WI
WA NM WI WA
.4 OR NM
.4 OR LA
NV OHLA RI NH NV OHRI NH
CA CA
NJ CT NJ CT
MI WV HI ME MI WV
HI ME
IL MA IL MA VT
VT
.3 NY .3 NY
PA DC PA
DC
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel C. Decline in Number of Employees
Fraction Receiving PPP State PPPE
SD SD
.8 .8
Ratio of Employees relative to January (Pre-PPP) Ratio of Employees relative to January (Pre-PPP)
UT .7 UT
.7
NE ND NE ND
ID ID OK
.6 OK .6
MS MS
KS KS
MT MT
AZ AK AL MO IA AZ AK AL MO IA
GA
SC TN AR SC GA TN AR
.5 KY .5 NC TX KY
NC VA TX MN FL
VA MN
WY
FL WY CO
CO
MD DE MD DE
WA NM IN WI WA LANM WI IN
OR LA OR
CANV OH RI
.4 NV CA OHRI NH
.4 NH
NJ CT NJ CT
IL HI IL MI HI
MI WV MA ME MA WV ME
VT
VT
26
DCNY DC NY
PA .3 PA
.3
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Figure D.2: Other Labor and Economic Outcomes and PPP Allocation by State (Round 1)
Figure D.2 presents scatterplots of the average pre-PPP UI claims as a percentage of employment (Panel A), of the
average decline in small business revenues at the state level from Womply (Panel B), and of the average decline
in employment at the state level from OI (Panel C). The figures on the left plot the pre-PPP outcomes against the
number of PPP loans received by small businesses in each state during the first round of the program divided by
the total number of small businesses in the state. The figures on the right plot the same pre-PPP outcomes and
the state-level PPPE measure. Data is from State Labor Departments, Womply, Opportunity Insights, SBA, and
Call Reports.
Panel A. UI Claims
Fraction Receiving PPP State PPPE
.2 .2
RI RI
Pre-PPP State Unemployment Claims as % of Employment
MI MI
Pre-PPP State Unemployment Claims as % of Employment
PA PA
NV HI NV HI
NH
.15 KY
NH
.15 KY
WA LA WA LA
NJ NJ
OH MA OHMA
CA ME
MT CA MT ME
GA GA
AK MN VT AK MN VT
DE DE IN
IN ID IA ID IA
AL .1 AL
.1
NC MD WI KS NC MD WI KS
NM SC AZ SC NM
AZ MO NY MO
NY VA IL AR OK VA IL AR OK
ND ND
OR TN MS OR TN MS
NE NE
TX WY TX WY
FL CT UT FL UT CT
.05 WV CO .05 CO WV
SD SD
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel B. Small Business Revenues
Fraction Receiving PPP State PPPE
Total Consumer Spending 03/22 to 03/28 relative to January (Pre-PPP) Total Consumer Spending 03/22 to 03/28 relative to January (Pre-PPP)
-.2 -.2
NE NE
AL AL SD
SD
ID MO WY ND ID WYMO AR KSND
NC KY AR KSMT NC TX MNKY MT OK
SC TX IN
GA
MN MS OK SC GA IN MS
AZ TN AZ OR TN
OR NM
NMVA UT VA
-.4 UT
NH -.4 MI NH
MI IA FL IA
FL
CT OHLA WI LA CT OH WI
WA IL WA IL
CA WV CO CA CO WV
MD DE MD DE
NY AK ME
NV NY AK VT ME NV VT
PA PA
RI RI
MA
NJ MA NJ
-.6 -.6
DC DC
HI HI
-.8 -.8
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel C. OI Employment
Fraction Receiving PPP State PPPE
Decline in OI Employment 03/22 to 03/28 relative to January (Pre-PPP) Decline in OI Employment 03/22 to 03/28 relative to January (Pre-PPP)
0 0
IA IA
SD SD
MO MO
AZ OK AZ OK
-.05 DC WY -.05 DC WY
FL SC MN SC
FLWA MN
WA MS NE NEMS
NV NC GA TX INLA UT
VA WI KS NV UT GA VA
NC LA
TX WI IN KS
NM WV CO ID AR ID CO NM ARWV
CA MD IL AL CA MD IL AL
DE TN KY DE TN KY
OR OH OR OH
PA PA
-.1 ME
-.1 AK ME
MI AK NJ MI ND
NJ ND
CT CT
NH NH
HI HI
MA MA
NY NY
-.15 MT -.15 MT
RI RI
VT VT
-.2 -.2
27
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Figure D.3: Pre-PPP Health Outcomes and PPP Allocation by State
Figure D.3 presents scatterplots of the cumulative number of pre-PPP COVID-19 cases and deaths per thousand
in each state as of April, 3rd 2020. The figures on the left plot the pre-PPP health outcomes against the number
of PPP loans received by small businesses in each state during the first round of the program divided by the total
number of small businesses in the state. The figures on the right plot the same pre-PPP health outcomes and the
state-level PPPE measure. Data comes from the Center for Disease Control, SBA, Call Reports, and FDIC Summary
of Deposits.
Panel A. Covid Cases per Capita
Fraction Receiving PPP State PPPE
3 3
Cumulative Covid-19 Cases per Thousand until 04/03 (Pre-PPP) Cumulative Covid-19 Cases per Thousand until 04/03 (Pre-PPP)
NJ NJ
NY NY
LA LA
2 2
MA MA
CT CT
MI MI
1 1
WA WA
IL PA CO IL
CO PA
VT VT
GA IN RI ID NV GAID RI IN
NV FL MD DE MS FL MD DE TN MS
TN UT NH UT NH
CA SC OH AL WI
MO ME CA SC AL WYWI
MO ME
AZ ORNC NMVA TXAK AR HI MT
KSIA
WY OK
SD ND
AZ OR NC NM OH
VA AKTX MT
AR HI OK
KSND
IA
SD
KY MN NE MN KY NE
WV
WV
0 0
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel B. Covid Deaths per Capita
Fraction Receiving PPP State PPPE
.08 .08
Cumulative Covid-19 Deaths per Thousand until 04/03 (Pre-PPP) Cumulative Covid-19 Deaths per Thousand until 04/03 (Pre-PPP)
LA LA
NJ NJ
.06 .06
MI MI
.04 WA
.04 WA
CT CT
MA MA
NY NY
.02 CO .02 GA CO
GA IL
IL IN
NV DE IN NV DE
VT VT
MS OK WI MSOK
WI FL SC PA
AL OH
CA FL MD SC PAOH KY AL KS ME CA
IDMD RIKY KS ME
AZ OR NMVA TN RI ID NH
MO MT AZ OR NM TNMN MO AR IA
VA AKTX MTNH
TX
AK MN AR IA NE ND NC NESDND
NC WV UT HI SD UT HI
WV
0 WY 0 WY
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
28
Figure D.4: Pre-PPP Social Distancing and Public Health Interventions and PPP Allocation by State
Figure D.4 presents scatterplots of measures of pre-PPP social distancing (Panel A) and of the timing of statewide
shelter-in-place orders (Panel B). The figures on the left plot these pre-PPP social distancing and public health
intervention outcomes against the number of PPP loans received by small businesses in each state during the first
round of the program divided by the total number of small businesses in the state. The figures on the right plot
the same outcomes and the state-level PPPE measure. Data comes from the New York Times, SBA, Call Reports,
and FDIC Summary of Deposits.
Panel A. Social Distancing Index
Fraction Receiving PPP State PPPE
WY WY
-.2 -.2
Social Distancing Index during 03/22 to 03/28 (Pre-PPP) Social Distancing Index during 03/22 to 03/28 (Pre-PPP)
-.3 -.3
NM SC AL SC NM
AL
AZ MT OK AZ MT OK
UT ID
WV TN UT ID AR MS TN ARWV
MS
NC GA NC GA
-.4 OR
KY
ME -.4 OR
KY
ME
ND ND
VA MO VA MO
LA SD
NE LA NESD
TX KSIA TX IAKS
CA RI CA RI
AK IN NH -.5 AK IN NH
-.5 FL
WA MD IL OH CO
FLWA MD CO
IL OH
NV DE
CT NV DE CT WI
PA WI PA
MA MA
VT VT
NJ NJ
NY
NY -.6
-.6 MI MI
MN MN
HI HI
DC DC
-.7 -.7
.1 .2 .3 .4 .5 -.2 0 .2 .4
Share of Small Businesses Receiving PPP (Round 1) State PPPE (Round 1)
Panel B. Shelter in Place Orders
Fraction Receiving PPP State PPPE
.5 .4
VT ME
Share of Small Businesses Receiving PPP (Round 1)
ND
NE NH OK
ND
SD MT KS MS SD
WV HI IA
NE
IN AR
OK .2 WI KY MO
ME WY
State PPPE (Round 1)
MI
MA WY
MT
KS MS IA
OH MN RI PA
HI TN
VT AR CT NM AL
LA TX
WI MN
NH MO IL DE CO AK
ID
.2 .3 .4
AL VA
MD
RI ID
MA KY UT GA
LA IN CO TN NY NJ NC SC UT
OH 0 WA
AK PA TX FL
IL WV SC
CT DE VA GA OR
MI
NM MD
NC
WA
OR FL DC
NY NJ
DC
AZ NV CA AZ
CA
.1 -.2 NV
20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 r 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 r
Or Or
r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 de r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 r20 de
ma ma ma ma ma ma ma ma ma ma ma ma ma ma ap ap ap ap ap ap ap te ma ma ma ma ma ma ma ma ma ma ma ma ma ma ap ap ap ap ap ap ap te
18 19 20 21 22 23 24 25 26 27 28 29 30 31 01 02 03 04 05 06 07 Sta 18 19 20 21 22 23 24 25 26 27 28 29 30 31 01 02 03 04 05 06 07 Sta
No No
Date Shelter Order Date Shelter Order
29
Figure D.5: Pre-PPP Homebase Employment Outcomes and PPP Allocation by County (Round 1)
Figure D.5 presents scatterplots of the share of businesses in each county that shutdown in the week of March
22nd to March 28th (Panel A), of the decline in hours worked in each county in the week of March 22nd to March
28th relative to a January Baseline (Panel B), and of the decline in the number of employee counts in each county
in the week of March 22nd to March 28th relative to a January baseline (Panel C). The figures on the left plot the
pre-PPP county-level employment outcomes from Homebase against the average number of PPP loans received
by small businesses in each county during the first round of the program divided by the total number of small
businesses in the county. The figures on the right plot the same pre-PPP employment outcomes and the average
county-level PPPE measure. Data is from Homebase, SBA, and Call Reports.
Panel A. Business Shutdowns
Fraction Receiving PPP County PPPE
.8 .8
Share of Business Shutdowns during 03/22 to 03/28 (Pre-PPP) Share of Business Shutdowns during 03/22 to 03/28 (Pre-PPP)
.6 .6
.4 .4
.2 .2
0 0
-2 -1 0 1 2 3 -.5 -.3 -.1 .1 .3 .5
Share of Businesses receiving PPP (Round #1) County PPPE (Round #1)
Panel B. Decline in Hours Worked
Fraction Receiving PPP County PPPE
Ratio of Hours Worked 03/22 to 03/28 relative to January (Pre-PPP) Ratio of Hours Worked 03/22 to 03/28 relative to January (Pre-PPP)
1 1
.75 .75
.5 .5
.25 .25
0 0
-2 -1 0 1 2 3 -.5 -.3 -.1 .1 .3 .5
Share of Businesses receiving PPP (Round #1) County PPPE (Round #1)
Panel C. Decline in Number of Employees
Fraction Receiving PPP County PPPE
Ratio of Nbr. Employees 03/22 to 03/28 relative to January (Pre-PPP) Ratio of Nbr. Employees 03/22 to 03/28 relative to January (Pre-PPP)
1 1
.75 .75
.5 .5
.25 .25
0
0 1 2 3
30 0
-.5 -.3 -.1 .1 .3 .5
-2 -1
Share of Businesses receiving PPP (Round #1) County PPPE (Round #1)
Figure D.6: Pre-PPP Other Employment and Economic Outcomes and PPP Allocation by County (Round 1)
Figure D.6 presents scatterplots of the average county pre-PPP UI claims as a percentage of employment (Panel
A), of the average pre-PPP decline in small business revenues at the county level from Womply (Panel B), and of
the pre-PPP decline in employment at the county level from OI (Panel C). The figures on the left plot the pre-PPP
county-level employment outcomes against the average number of PPP loans received by small businesses in each
county during the first round of the program divided by the total number of small businesses in the county. The
figures on the right plot the same pre-PPP employment outcomes and the average county-level PPPE measure.
Data is from State Labor Departments, Womply, Opportunity Insights, SBA, and Call Reports.
Panel A. UI Claims
Fraction Receiving PPP County PPPE
15 15
Pre-PPP County Unemployment Claims as a % of Employment Pre-PPP County Unemployment Claims as a % of Employment
10 10
5 5
0 0
0 .2 .4 .6 .8 1 -.5 -.3 -.1 .1 .3 .5
Share of County Businesses receiving PPP (Round #1) County PPPE (Round #1)
Panel B. Small Business Revenues
Fraction Receiving PPP County PPPE
1 1
Decline in County Small Buz Rev 03/22 to 03/28 relative to Jan Decline in County Small Buz Rev 03/22 to 03/28 relative to Jan
.5 .5
0 0
-.5 -.5
-1 -1
0 .2 .4 .6 .8 1 -.5 -.3 -.1 .1 .3 .5
Share of County Businesses receiving PPP (Round #1) County PPPE (Round #1)
Panel C. OI Employment
Fraction Receiving PPP County PPPE
Decline in County OI Employment 03/22 to 03/28 relative to January Decline in County OI Employment 03/22 to 03/28 relative to January
.1 .1
0 0
-.1 -.1
-.2 -.2
-.3 -.3
31
0 .2 .4 .6 .8 1 -.5 -.3 -.1 .1 .3 .5
Share of Businesses receiving PPP (Round #1) County PPPE (Round #1)
Figure D.7: Targeting of PPP Allocation across Quantiles of PPPE and Predicted PPPE (First Round)
Figure D.7 stratifies all businesses in Homebase in 10 bins based on the PPPE (left panels) and Predicted PPPE
(right panels) of the their ZIP codes during the first round. Panel A plots for each bin the share of Homebase
businesses that shut down in the week of March 22nd–March 28th. Panel B plots for each bin the average decline
in hours worked in the week of March 22nd–March 28th relative to a baseline of the average weekly hours worked
in the last two weeks of January. Panel C plots for each bin the average decline in the number of employees in
the week of March 22nd–March 28th relative to a baseline of the average number of employees in the last two
weeks of January. Data are from SBA, Call Reports, Homebase, and County Business Patterns.
Panel A. Business Shutdowns and Fraction of Establishments Receiving PPP
PPPE (Round 1) Predicted PPPE (Round 1)
.4 .4
Share of Businesses Shutdown Share of Businesses Shutdown
during 03/22 to 03/28 (Pre-PPP) during 03/22 to 03/28 (Pre-PPP)
.25 .3 .35 .25 .3 .35
.2 .2
t 2 3 4 5 6 7 8 9 t t 2 3 4 5 6 7 8 9 t
/Es /Es /Es /Es
PP PP PP PP
st P st P st P st P
Lea Mo Lea Mo
PPPE Quantile (Round 1) Predicted PPPE Quantile (Round 1)
Panel B. Decline in Hours Worked and Fraction of Establishments Receiving PPP
PPPE (Round 1) Predicted PPPE (Round 1)
Ratio of Hours Worked 03/22 to 03/28 Ratio of Hours Worked 03/22 to 03/28
Relative to January (Pre-PPP) Relative to January (Pre-PPP)
.35 .4 .45 .5 .35 .4 .45 .5
.3 .3
/Es 2 3 4 5 6 7 8 9 st /Es 2 3 4 5 6 7 8 9 st
t P/E t P/E
PP PP
st P st P st P st P
P P
Lea Mo Lea Mo
PPPE Quantile (Round 1) Predicted PPPE Quantile (Round 1)
Panel C. Decline in Number of Employees and Fraction of Establishments Receiving PPP
PPPE (Round 1) Predicted PPPE (Round 1)
Ratio of Nbr. Employees from 03/22 to 03/28 Ratio of Nbr. Employees from 03/22 to 03/28
Relative to January (Pre-PPP) Relative to January (Pre-PPP)
.35 .4 .45 .5 .55 .35 .4 .45 .5 .55
/Es 2 3 4 5 6 7 8 9 st /Es 2 3 4 5 6 7 8 9 st
t P/E t P/E
PP PP
st P st P st P st P
P P
Lea Mo Lea Mo
PPPE Quantile (Round 1) 32 Predicted PPPE Quantile (Round 1)
Appendix E. Crowd-out Appendix
This appendix presents evidence that suggests that the PPP crowded out some private loans.
We find some evidence of modest crowd-out, but the results suggest that magnitudes are small
and private lending would not have fully offset PPP lending.
Appendix Figure E.1 shows suggestive evidence of crowd-out from California Uniform Com-
mercial Code (UCC) filings.1 The figure shows a significant spike in UCC filings in May, follow-
ing the exhaustion of PPP funds, which is consistent with the program crowding-out private
lending. On the other hand, the time series could simply reflect bureaucratic delays in filing or
the recovery. Appendix Figure E.2 shows scatterplots of the ratio of UCC filings per establish-
ment and county-level PPPE under different time horizons. The relationship is relatively flat,
suggesting little relationship between the availability of PPP loans and commercial lending.2
While suggestive of some crowd-out, we use the Call Reports to explore this pattern more
broadly and formally. The Call Reports are quarterly at the bank level, and the first two annual
quarters (January-March, April-Jun) almost perfectly align with the disbursement of PPP loans,
which began on April 3. The number of loans in the second quarter is given by C&I LoansQ2 =
C&I LoansQ1 +P P P +N L P, where C&I LoansQi are commercial and industrial loans in quarter
i, P P P refers to PPP loans, N L are other new non-PPP commercial loans, and P are loans that
are paid or charged-off. We rearrange the equation in terms of quarterly loan growth and write:
C&I LoansQ2 P P P Loans
=1+ + ⇣. (E.1)
C&I LoansQ1 C&I LoansQ1
The coefficient captures crowd-out. If there is no crowd-out, an additional PPP loan leads
to one additional total loan, and = 1. Under full crowd-out, an additional PPP loan is offset
by a reduction in another commercial loan, and = 0. Appendix Figure E.3 plots the ratio of
Commercial and Industrial Loans in Q2 2020 and Commercial and Industrial Loans in Q1 2020
and the ratio of PPP loans and Commercial and Industrial Loans in Q1 2020. Appendix Table
E.1 reports OLS and IV regressions examining the relation between the ratio of Commercial and
C&I Loans
Industrial Loans in Q2 2020 and Commercial and Industrial Loans in Q1 2020 ( C&I LoansQ2
Q1
) and
P P P Loans
the ratio of PPP loans and Commercial and Industrial Loans in Q1 2020 ( C&I LoansQ1 ). Column
1
We obtained data from California on all UCC filings, which are required for all secured business loans to protect
creditor claims. These UCC laws are set at the state level, although the National Conference of Commissioners
has sought to make them fairly uniform across states. We are able to observe the names and addresses of the
debtor, which we use to subsequently match with the SBA data. We refer readers to Edgerton (2012) for further
details about the UCC data and its features.
2
Note that PPP loans are unsecured and hence are not included in UCC filings.
33
(2) instruments using lender PPPE. In column (1), the coefficient is 0.558 and statistically
significant at the one-percent level. This is suggestive of some crowd-out, but not full crowd-
P P P Loans
out. The OLS estimates may be biased if C&I LoansQ1 and ⇣ are correlated, so in column (2) we
instrument using lender PPPE. The IV estimate in column (2) provides an estimate of = 0.991,
and we cannot reject no crowd out. The confidence interval of the IV estimate allows us to
reject full crowd-out.
Though we find limited evidence of private sector crowd-out, the PPP may have crowded
out other federal loan programs, namely, the Economic Injury Disaster Loan (EIDL) program.
If firms could obtain other federally guaranteed loans in the absence of PPP, this would also
lead to the program having muted effects. The COVID-19 EIDL program is a SBA program
that provides economic relief to small businesses that experience a temporary loss of revenue
due to the coronavirus. The program offers advantageous terms for regular businesses with
interest rates set at 3.75% and maturity of 30 years with no prepayment penalties. Given these
terms, it is possible that the second-best option of firms that did not obtain access to PPP was
to apply for a loan under the EIDL program.
Appendix Figure E.4 shows cumulative PPP and EIDL lending over time. Similar to UCC
filings, we see an uptick in EIDL loans after PPP funds level off in May. The SBA was slower
to open up the expanded EIDL provisions of the CARES Act, which may also account for this
lagged increase. The fact that EIDL loans only start rising in late May implies that crowd-out
of EIDL is unlikely to explain the modest effects we estimate for April and May.
Appendix Figure E.5 shows scatterplots of the average fraction of small business establish-
ments that received an EIDL loan in each percentile bin based on state PPPE in the weeks of
May 3-9 and June 28-July 4. The figure shows a weak relationship in the early period with
very few firms receiving EIDL loans. In the later period, there is a strong negative relationship,
consistent with crowd-out and indicating that higher PPP exposure is associated with fewer
EIDL loans. If firms that were unable to access PPP were more likely to apply for and receive
an EIDL loan, and if EIDL loans were sufficiently good substitutes for PPP loans, this fact could
help account for modest estimated effects of PPP in June.
34
We
ek1 Number of UCC filings (California)
:J
We an 0 20000 40000 60000
19
ek -J
2: an
We Ja 25
ek n2
3: 6-F
We F eb e b1
ek 02
4: -Fe
We
ek
Fe
b9 b 8
5: -Fe
We Fe b1
ek b 16 5
6: -Fe
Fe
b2 b 2
We 3-F 2
ek eb
7:
We
ek M ar1
29
8 - M
We : M a r7
ek ar8
We 9: -M
Ma ar1
ek
10 r15 4
: M -M
We ar2 ar21
ek 2-
11
: M Mar2
We
ek ar2 8
12 9-A
We :A pr4
ek pr5
13 -Ap
We :A r11
ek pr1
14 2-A
: p
Figure E.1 shows the number of UCC filings between January and July 2020. Data comes from California UCC
We Ap r18
ek r19
35 1 -Ap
We 5: Ap r25
ek r26
We 16 -M
:M ay
Week ek
17 ay 2
We :M 3-M
ek ay a y9
18 1 0-M
We :M ay
ek ay 16
19 17
:M -M
We ay ay
Figure E.1: UCC Filings Over Time
ek 24 23
20 -M
:M ay
We ay 30
ek 31
21 -Ju
We
ek
:J
un n 6
22 7-J
We :J un
ek un 13
23 14
:J -Ju
un n2
We
ek 21 0
24 -Ju
:J n2
We un 7
ek 28
25 -Ju
We
ek
:J
ul5 l4
26 -Ju
We :J
ul1 l 11
ek 2-J
27 ul1
:J 8
We
ek u l19
28 -Ju
:J
ul2 l 2 5
6-J
ul3
1
filings.
Figure E.2: UCC Filings and PPPE
Figure E.2 are scatterplots of the the ratio of UCC filings per establishment in each county and the county exposure
to the number-based PPPE during the first round of PPP. Data comes from SBA, Call Reports, Summary of Deposits,
County Business Patterns and California UCC filings.
.4
Ratio UCC filings/Establishments
.3
.2
.1
0
-.5 0 .5
PPPE Round 1
03/22-03/28 04/19-04/25 May17-May23 Jun14-Jun 20 Jul12-Jul18
36
Figure E.3: PPP Lending and Commercial & Industrial Loans
Figure E.3 are scatterplots of the ratio of Commercial and Industrial Loans in Q2 2020 and Commercial and
C&I Loans
Industrial Loans in Q1 2020 ( C&I LoansQ2
Q1
) and the ratio of PPP loans and Commercial and Industrial Loans in Q1
P P P Loans
2020 C&I LoansQ1 . Data comes from Federal Reserve Call Reports and SBA.
1.08
1.06
C&I Loans Q2 / C&I Loans Q1
1.04
1.02
1
0 .02 .04 .06 .08 .1
PPP Loans / C&I Loans Q1
37
Table E.1: PPP Lending and Crowd-Out
Table E.1 reports the results of ordinary least squares (OLS) and instrumental variables (IV) regressions examining
the relation between the ratio of Commercial and Industrial Loans in Q2 2020 and Commercial and Industrial
C&I Loans
Loans in Q1 2020 ( C&I LoansQ2 Q1
) and the ratio of PPP loans and Commercial and Industrial Loans in Q1 2020
P P P Loans
C&I LoansQ1 . Column (2) instruments using lender PPPE. Data comes from Federal Reserve Call Reports and SBA.
Robust standard errors are presented in parentheses. ***, **, and *, represent statistical significance at 1%, 5%,
and 10% levels, respectively.
(1) (2)
OLS IV
C&I LoansQ2 C&I LoansQ2
Dep. Variable: C&I LoansQ1 C&I LoansQ1
P P P Loans
C&I LoansQ1 0.558⇤⇤⇤ 0.991⇤⇤⇤
(0.162) (0.278)
Constant 1.017⇤⇤⇤ 1.011⇤⇤⇤
(0.00255) (0.00457)
Observations 4845 4845
Adjusted R2 0.010 0.004
38
&XPPXODWLYH/RDQ&RXQW
39 333/RDQV
Figure E.4: EIDL and PPP
(,'//RDQV
Figure E.4 shows cumulative PPP and EIDL lending between April and July, 2020. Data comes from the SBA.
Figure E.5: EIDL and PPPE
Figure E.5 are scatterplots of the fraction of small business establishment that received an EIDL loan in each state
and the respective State PPPE during the first round. Data comes from SBA, Call Reports, Summary of Deposits,
and County Business Patterns.
19
/$
)UDFWLRQRI6PDOO%XVLQHVVHV
&$ &7
5,
'& :$ *$ 1+
UHFHLYLQJ(,'/DVRI:HHN
)/
0(
1- &2
1& 0'
9$ 2+0$ 1'
$= 6&
87 97
,/ 7;$/ 01 0, +,
25 1< 10 3$ 6' 2.
,$
.6
.< ,1
'( 71 :, 1(07
:< 06
$5
,' 02
:9
$.
6WDWH333( 5RXQG
+,
)/
&$ *$
/$
19
)UDFWLRQRI6PDOO%XVLQHVVHV 1-
1< 7;
UHFHLYLQJ(,'/DVRI:HHN
06
$= $.
6& 0' &7
,/ $/
'& 25 1&
:$ 9$ &2 10 71 5,
0,
0$ 2. 97
'( 1+
87 6'
3$ $5 07
,' 2+
01 0(
:<
.<:, 1(
02 .61'
,1
:9
,$
6WDWH333( 5RXQG
40
Appendix F. Supplemental Analysis
This appendix presents additional robustness checks and additional empirical analyses ex-
ploring possible mechanisms that can account for the effect of the program on employment
outcomes.
Figure F.1 examines the possibility that the program allowed firms to maintain their em-
ployees at home without pay. The plot suggests that conditional on households reporting not
working in each week of the survey, more than 85% reported receiving no pay. Moreover, the
share that report receiving no pay for staying at home does not have a strong relationship with
State PPPE. This suggests that the weak relationship between access to the program and num-
ber of hours worked that we documented in Table 5 is likely not explained by the possibility
that PPP recipients were keeping their workers at home with pay.
In Table F.1, we explore whether the expansion of unemployment insurance could have led
to difficulties for firms in recalling workers and, in turn, attenuated the employment effects of
the PPP. Following the work of Ganong et al. (2020), we partition the sample between state
with above- and below-median state-level UI replacement rates. The results of Table F.1 do
not suggest that the PPP was more effective in states with low UI replacement rates. In Table
F.2 we examine the impact of PPP on changes in initial unemployment insurance claims, small
business revenues, and Opportunity Insights employment after partitioning the sample into
two subsamples based on the median state UI replacement rates. Again, we do not see striking
differences between the measured impact of the program in both subsamples. In fact, with the
exception of the empirical specifications using OI employment data as the outcome variable,
greater access to PPP is more effective in states with relatively more generous employment
benefits.
Figure F.2 uses data from Homebase to plot binned scatter plots of the share of small busi-
ness establishments that “permanently” closed versus ZIP-level PPPE. Both variables are resid-
ualized with respect to the controls in Table 5, an additional control for the average number
of shutdowns in a ZIP code in the two weeks prior to PPP, and demeaned at the state level to
present the within-state relationship. Permanent shutdowns are defined as the establishment
being closed for all weeks from the beginning of the PPP through the end of August. The
graph reports regression coefficients with and without the Table 5 controls and standard errors
clustered at the state level.
The graph suggests a non-trivial impact of PPP on firms over the medium run, as a one-
standard-deviation change in PPPE is associated with a 0.97 percentage point reduction in
permanent shutdowns. This effect is reasonably large compared to the overall permanent
41
shutdown rate of 12%. The result is consistent with our interpretation that, by helping firms
bolster their balance sheets during the crisis, receiving PPP funds enabled these firms to reopen
at higher rates than if they had not received funds. Such impacts are also consistent with
modest short-term employment impacts providing an incomplete overall picture of the impact
of the program.
Figure F.3 uses the Homebase-PPP matched sample and shows the coefficients of week-by-
week regressions that repeat the OLS and IV specifications of columns (1) and (3) of Table 8
for every week in the sample. We obtain results that are qualitatively similar to those of our
regional analyses. Both the OLS and IV specifications show a small impact of receiving PPP
earlier on business shutdowns. This impact is small and fades away over time. The coefficients
of Panels B and C indicate firms receiving PPP earlier have better employment outcomes in
terms of number of hours and number of employees employed and that these gaps in the
number of employees and hours worked persist until August. Again, we caution that the OLS
estimator may be biased if firms that obtained loans earlier are fundamentally different from
firms that received PPP loans later. We note, however, that the IV coefficients show larger
standard errors likely due to a smaller sample size relative to our regional analyses. These
larger standard errors limit our ability to draw strong conclusions from this analysis.
Tables F.3 and F.4 partition the sample into two subsamples based on the median share of
establishments in the county that are eligible for participation in the PPP. The goal is to inves-
tigate whether eligible firms might expand at the expense of local competitors. Such business
stealing spillovers could account for low employment effects at the labor market level. Alter-
natively, the program might have positive local demand effects, for instance on the suppliers
of treated firms. Given the scale and severity of the labor market disruption due to the pan-
demic, traditional measures of labor market tightness are unlikely to be useful. However, we
can ask whether regions with a larger share of employment in PPP-eligible establishments ex-
hibit different effects relative to those with fewer eligible establishments. Employment effects
are generally similar or greater in regions where a larger share of establishments are eligible
for funds, inconsistent with a business stealing effect and possibly consistent with the presence
of some local demand effects. The results are also useful as a robustness check, since omitted
variables would have to account for stronger effects in regions with a larger share of firms that
are eligible for the program.
In Figure F.4 we repeat the empirical exercise in Figure 9 with and without the state of
California. In Figure 9 of the main draft, we exclude California counties from the time series of
Panel A due to a large outlier in UI claims that is likely due to a backlog in UI claims processing
42
in that state. In this figure, we show sensitivity of this analysis to this sample selection criterion.
In Panel C of this figure, we can see that this backlog likely created an abnormal jump in the
number of UI claims in the week of April 26 to May 2nd. While, this jump is significant for the
state of California, it would not materially change our interpretation of the results in the main
draft.
Finally, in Table F.5 we evaluate heterogeneity in effects of the program across industries by
repeating our empirical specification of column (6) of Table 5 for each industry in the Homebase
sample. The estimated effects of the PPP are more pronounced for the Food & Drink and retail
industry, which is consistent with the idea that the program was relatively more effective in
contact-intensive industries whose businesses were more disrupted by the pandemic.
43
Figure F.1: Share of Households reporting receiving no pay and Exposure to State PPPE
Figure F.1 are scatterplots of state exposure to the state-level PPPE in Round 1 and the percentage of households and the share of households
in the Census Household Pulse Survey that report not receiving any payment for time not working in the previous week. The plots represent
the evolution of the relation between Round 1 State PPPE and the share of respondent reporting receiving no pay over the first six weeks of
the survey. Data comes from the Census Bureau and SBA.
Week 15: Apr23-May5 Week 16: May7-May12
100 100
VT
Share of Households reporting receiving no pay for time not working Share of Households reporting receiving no pay for time not working
MO MT
ME
NC
NY IL AR SD
OR PA WI NM WV
AR HI OR FL AK
DE TN
NC ID IL WY WV
ND VT AZ CO TX
ID PA
90 UT CO 90 WAUT AL OHMIKY HI
CA FLWA
SC IA NH ME ND
NY AK AL OHRI VA MN WI
AZ TN MI MO CA DC OK
MD TX SD
MS
MT MD IN
LA CT
MN NEKS SC DE CT
NV KY WY KS
NJ VA MA NJ GA IA NH
IN RI
DC NM NV LA
GA MA NE
OK
80 80 MS
70 70
-.2 0 .2 .4 -.2 0 .2 .4
State PPPE (Round 1) State PPPE (Round 1)
Week 17: May14-May19 Week 18: May 21-May26
100 100
Share of Households reporting receiving no pay for time not working Share of Households reporting receiving no pay for time not working
NH
OR AK OR WY
CT ME ID
AZ AZ VT
FL VT MD
NC MN ME
NJ OH WV NH NE
ID NM
TX MI MT
ND FL NC IAKS
UT GA MD CO TN WY INARNEIA
KS
SD
SC
NY IL
90 SC IL 90 WA
UT CO MT
WI VA AK NM CTTN MA
AL
KYWI
MO
RI MO NJ RI
MN
DC VA PA OK CA SD OK
NV GA TX
PA IN WV ND
CA WA KY LA MI
NY AL NV AR MS
MA DC HI
HI
MS
LA DE
80 80
OH
DE
70 70
-.2 0 .2 .4 -.2 0 .2 .4
State PPPE (Round 1) State PPPE (Round 1)
Week 19: May 28-Jun2 Week 20: Jun4-Jun9
100 100
Share of Households reporting receiving no pay for time not working Share of Households reporting receiving no pay for time not working
ID
OR
ID AK FL ME
WA MTNH VT
ME
CO TNMN IA ND
ARNE AK PA NESD
MT
AZ OR FL DELA WY HI CA NC
SC ND
OK VT
MO KYWI
90 DC NY MD IL
NM 90 NV CO NM TN AR KS
UT VA NY
WA
NJ CT OHMI
AL WY IN IA NH
NJ
NC KYWI MS AZ DE
ILLA MA
NV CA MI
PA KS OK
SD
WV MD MN HIMS
SC AL
CT IN TX WV
MA
RI RI MO
GA
OH UT GA
TX VA
DC
80 80
70 70
-.2 0 .2 .4 -.2 0 .2 .4
State PPPE (Round 1) State PPPE (Round 1)
44
Figure F.2: PPP Exposure and Permanent Shutdowns
Figure F.2 plots binned scatter plots of the share of small business establishments that permanently closed versus ZIP-level PPPE. Both variables
are residualized with respect to the controls in Table 5, an additional control for the average number of shutdowns in a ZIP code in the
two weeks prior to PPP, and demeaned at the state level to present the within-state relationship. Permanent shutdowns are defined as the
establishment being closed for all weeks from the beginning of the PPP through the end of August. The graph reports regression coefficients
with and without the Table 5 controls and standard errors clustered at the state level. Outcome data come from Homebase.
16
No Controls: = 0.72(0.32)
Controls: = 0.97(0.37)
Permanent Shutdown (Residualized, %)
10 12 14
8
-2 -1 0 1 2
Exposure (Residualized and Standardized)
45
46
We We
e k1 Coefficient ek Coefficient
0: 10
: M
We
Ma
r22 -.02 0 .02 .04 We a r22 -.02 0 .02 .04
ek -M ek -M
11 ar2 11 ar2
: M 8 : M 8
We a r29 We ar2
e k1 -Ap ek 9-A
2: r4 12 pr4
Ap W :A
We
e r5 PPP Launch e ek p r5-A PPP Launch
k1 -Ap 13
3: r11 :A pr1
We Ap We pr1 1
e r12 ek 2-A
k1 -Ap 14 pr1
4: r18 :A
We Ap W pr1 8
e
r19
-A 1st Round Exhausted e ek 9 -Ap 1st Round Exhausted
k1 pr2 15
5: 5 : r25
We A pr2 We Ap
r26
e k1 6-M ek -M
6: ay 16 ay
We Ma 2 2nd Round Begins We : M 2 2nd Round Begins
e k1 y3 ek ay
7: -M 17 3-M
We M ay
ay
9 We :M
ay
ay
9
e k1 1 0-M e k1 1
0-M
8: ay 8: ay
Figure F.3 investigates business shutdowns, changes in the ratio of hours worked, and changes in the number of employees for firms in the Homebase sample
We Ma 16 We Ma 16
ek y1 ek y1
19 7 -M 19 7 -M
:M ay :M ay
We ay 23 We ay 23
24 24
ek
20 -M ek
20 -M
:M ay :M ay
30 30
We ay We ay
e k2
31
-Ju e k2
31
-Ju
1: n6 1: n6
We J un We J un
e k2 7-J e k2 7-J
2: un 2: un
We Ju 13 We Ju 13
n1 n1
e k2 4-J e k2 4-J
3: un 3: un
J un 20 J un 20
We 2 1 We 2 1-J
e -Ju e
k2
4 n 27
k2
4
un
27
:J :J
We un We un
e 28 e 28
k2 -Ju k2 -Ju
5: l4 5: l4
We Ju We Ju
e k l5-J e k l5-J
26 ul1 26 ul1
: Ju 1 : Ju 1
We l12 We l12
e k2 -Ju e k2 -Ju
7: l18 7: l18
We J ul1 We J ul1
e k2 9-J e k2 9-J
ul2 ul2
8:
Ju 5 8:
Ju 5
We l26 We l26
e k2 -Au e k2 -Au
9: g1 9: g 1
We Au We Au
e k3 g 2-A e k3 g 2-A
0: ug 0: ug
A 8 A 8
that are name-matched to the PPP data set from SBA. In the upper panels, we plot OLS estimates from week-by-week regressions examining the impact of
We ug We ug
ek 9 ek 9
Week of PPP Projections (IV) Panel A. Business Shutdowns
31 -Au 31 -Au
:A g1 :A g1
5 5
Week of PPP Projections (OLS)
We ug We ug
ek 16 ek 16
3 -A ug 3 -A ug
2: 22 2: 22
Au Au
g2 g2
3-A 3-A
ug ug
29 29
We We
e k1 Coefficient e k1 Coefficient
0: 0 :
W e
Ma
r2 -.15 -.1 -.05 0 .05 We
Ma -.02 0 .02 .04
ek 2 -M ek
r22
-M
11 ar2 11 ar2
: M 8 : M 8
We ar2 We ar2
e k1 9-A ek 9-A
2: pr4 12
: p r4
We A pr5 We Ap
e k1 -Ap PPP Launch ek r5-A PPP Launch
3: 13 pr1
Ap r11 :A 1
We r12 We pr1
e k1 -Ap ek 2-A
4: 14 pr1
Ap r18 :A 8
We r1 We p r19
e k1 9-A 1st Round Exhausted ek -Ap 1st Round Exhausted
delays in receiving PPP on differences in employment and firm outcomes. The OLS regressions repeat the specifications of column (1) of Table 8 for every
5: pr2 15
A pr2 5 W
:A
p
r25
We
6 e r2
e k1 -M ek 6-M
6: ay 16 ay
We M 2 2nd Round Begins We :M 2 2nd Round Begins
ek ay e k a y
1 7 3 -M 17 3 -M
:M ay : M ay
We a 9 We ay 9
e y1
0-M ek 10
k1
8: 18 -M
ay :M ay
We Ma
y 16 We a y 16
e 17 ek 17
k1
9: -M 19 -M
Ma ay :M ay
We y2 23 We a y 23
e 4-M 24
k2 ek
20 -M
0: ay ay
M ay
30 :M
ay
30
We We
e 3 1 e 31
k2 -Ju k2 -Ju
1 :J n 6 1 :J n6
We un We un
e k2 7-J e k2 7-J
2: un 2: un
We Ju 13 We Ju 13
n1 n1
e k2 4-J e k2 4-J
3 :J u n2 3 :J u n2
u n2 0 u n2 0
We 1-J We 1-J
e k2 un e k2 un
4: 27 4: 27
We J un We J un
e 28 e 28
k2 -Ju k2 -Ju
5: l4 5: l4
We J ul5 We J ul5
e k2 -Ju e k2 -Ju
6: l11 6: l11
We Ju We Ju
l12 l12
e k2 -Ju e k2 -Ju
7: l18 7: l18
week in the sample. In the bottom panels, we plot IV estimates from week-by-week regressions examining the impact of delays in receiving PPP on differences
We J ul1 We J ul1
e k2 9 -Ju e k2 9 -Ju
8: l25 8: l25
We Ju We Ju
l26 l26
e k -Au e k -Au
29
: g1 29
: g 1
We Au We Au
e k3 g2
-Au e k3 g2
-Au
0 :A g8 0 :A g8
We
e u g9
We
e u g9
Week of PPP Projections (IV)
k3 -Au k3 -Au
1: g1 1: g1
A 5 A 5
Week of PPP Projections (OLS)
We ug We ug
e k3 1 6-A e k3 1 6-A
2: ug 2: ug
Au 22 Au 22
Panel B. Change in Hours Worked
g2 g2
3-A 3-A
ug ug
29 29
We We
ek
10
Coefficient ek
10
Coefficient
: M : M
We a r22 -.15 -.1 -.05 0 .05 We a r22 -.02 0 .02 .04
ek -M ek -M
11 ar2 11 ar2
: M 8 : M 8
We ar2 We ar2
9-A 9-A
in employment and firm outcomes. The IV regressions repeat the specifications of column (3) of Table 8 for every week in the sample. Outcome variables
ek ek
12 pr4 12 pr4
W :A W :A
e ek p r5-A PPP Launch e ek p r5-A PPP Launch
13 pr1 13 pr1
:A 1 :A 1
We pr1 We pr1
ek 2-A ek 2-A
14 pr1 14 pr1
: Ap 8 : Ap 8
We r19 We r19
ek -Ap 1st Round Exhausted ek -Ap 1st Round Exhausted
15 15
: A
r25 : A
r25
We pr2 We pr2
e k 6-M ek 6-M
16 ay 16 ay
We :M 2 2nd Round Begins We : M 2 2nd Round Begins
ek ay ek ay
17 3-M 17 3-M
Figure F.3: PPPE and Post-PPP Outcomes in the Matched Sample (Dynamic Analysis)
:M ay :M ay
We ay 9 We ay 9
ek 10 ek 10
18 -M 18 -M
:M ay :M ay
We a 1 6 We ay 16
e y1
7-M e 17
k1
9:
k1
9: -M
Ma ay Ma ay
We y2 23 We y 23
e 4-M e 24
k2 k2 -M
0 :M ay
30 0 :M ay
30
We a y3 We ay
e k2 1-J e k2
31
1 :J u n6 1 :J
-Ju
n6
We u We u
e k2
n7
-Ju e k2
n7
-Ju
2: n1 2: n1
We J un 3 We J un 3
e 1 4-J e 1 4-J
k2 un k2 un
3: 20 3: 20
Ju Ju
We n2 We n2
e 1-J e 1-J
k2 un k2 un
are defined as the difference between measured outcomes in each week and their average in the two weeks preceding program launch. The right-hand-side
4: 27 4: 27
J un J un
We
e 2 8
We
e 2 8
k2 -Ju k2 -Ju
5 :J l4 5 :J l4
We ul5 We ul5
e k2 -Ju e k2 -Ju
6: l11 6: l11
We Ju We Ju
l12 l12
e k2 -Ju e k2 -Ju
7 :J l18 7 :J l18
We u l19 We u l19
e k2 -Ju e k2 -Ju
8: l25 8: l25
Ju Ju
We l2 We l2
e k2 6-A e k2 6-A
ug ug
We
9:
Au 1 We
9:
Au 1
e k3 g2
-Au e k3 g2
-Au
0: g8 0: g8
We Au We Au
ek g9 ek g9
Week of PPP Projections (IV)
31 -A u 31 -A u
:A g1 :A g1
u 5 u 5
Week of PPP Projections (OLS)
We g1 We g1
e k3 6-A e k3 6-A
2: ug 2: ug
Au 22 Au 22
g2 g2
3-A 3-A
Panel C. Change in Nbr. Employees
ug ug
29 29
variable is the week of PPP receipt. Data is from Call Reports, SBA, and Homebase.
47
W W
ee ee
k 1 Initial UI Claims/January Employment k 1 Initial UI Claims/January Employment
W : Jan W : Jan
ee 0 1% 2% 3% 4% ee 0 1% 2% 3% 4%
k 2 19 - k 2 19 -
W W
ee : Ja Jan ee : Ja Jan
k 3 n2 25 k 3 n2 25
W 6 W 6
ee : Fe -Fe ee : Fe -Fe
W k b0 b1 W k b0 b1
ee 4: F 2-F ee 4: F 2-F
k 5 e e k 5 e e
W : F b9-F b8 W : F b9-F b8
e e e e
k 6 eb1 eb1 k 6 eb1 eb1
: F 6-F 5 : F 6-F 5
W eb eb W eb eb
e 2 2 e 2 2
W ek 7 3-F 2 W ek 7 3-F 2
ee :M e ee :M e
ar b29 ar b29
W k 8:
e e M 1 -
W k 8:
e e M 1 -
ar Mar ar Mar
W k 9:
ee 8 7 W k 9:
ee 8 7
k 1 Ma -Ma k 1 Ma -Ma
r r r r
W 0: M 15-M 14 W 0: M 15-M 14
e e a e e a
k r2 ar2 k r2 ar2
W 11: 2 - 1 W 11: 2 - 1
ee M Ma ee M Ma
W k 12 ar29 r28 Initial Lockdown Orders W k 12 ar29 r28 Initial Lockdown Orders
ee ee
k 1 : Ap -Ap k 1 : Ap -Ap
W r r W r r
ee 3: A 5-A 4
k p pr PPP Launch ee 3: A 5-A 4
k p pr PPP Launch
W 14: r12- 11 W 14: r12- 11
ee Ap Ap ee Ap Ap
k r 19 r18 k r 19 r18
W 1 1st Round Exhausted W 1 1st Round Exhausted
e 5: A -A e 5: A -A
W ek 1 pr2 pr25 W ek 1 pr2 pr25
e e 6 6- e e 6 6-
: :
W k 17 Ma May W k 17 Ma May
ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins
Low Predicted PPPE
Low PPPE
k 1 Ma 3-M k 1 Ma 3-M
W y W y
ee 8: M 10- ay9 ee 8: M 10- ay9
k1 ay Ma k1 ay Ma
Figure F.4 repeats the analysis of Panel A of Figure 9 excluding the state of California (Panel A), with all states including the state of California (Panel B) and
W 9: M 17- y16 W 9: M 17- y16
ee ay Ma ee ay Ma
k 2 2 k 2 2
W 0 4 y23 W 0 4 y23
ee : Ma -Ma ee : Ma -Ma
W k 21 y 3 y30 W k 21 y 3 y30
ee 1 ee 1
k 2 : Ju -Jun k 2 : Ju -Jun
W n W n
ee 2: J 7-J 6 ee 2: J 7-J 6
k u k u
2 n1 un1 2 n1 un1
W 3: J 4-J 3 W 3: J 4-J 3
ee un u ee un u
k 21 n20 k 21 n20
W 24: -J W 24: -J
ee Ju un2 ee Ju un2
W k n2 7 W k n2 7
ee 25: 8 - ee 25: 8 -
k 2 Ju Jul4 k 2 Ju Jul4
W l5- W l5-
ee 6: J ee 6: J
k u Jul1
1 k u Jul1
1
W 27: l12- W 27: l12-
High PPPE
ee Ju Jul1 ee Ju Jul1
k l1 8 k l1 8
W 28: 9 W 28: 9
ee J -Ju ee J -Ju
Panel A. UI Claims (No CA)
W k 29 ul26 l25 W k 29 ul26 l25
ee : -A ee : -A
W k 30 Aug ug1 W k 30 Aug ug1
ee 2 ee 2
k 3 : Au -Au k 3 : Au -Au
W g g W g g
ee 1: A 9-A 8 ee 1: A 9-A 8
k 3 u u k 3 u u
High Predicted PPPE
2: g16- g15 2: g16- g15
Au A Au A
g2 ug2 g2 ug2
3- 3-
Au 2 Au 2
g2 g2
9 9
W W
ee ee
k1 Initial UI Claims/January Employment k1 Initial UI Claims/January Employment
W : Jan W : Jan
ee 0 1% 2% 3% 4% ee 0 1% 2% 3% 4%
k 2 19 - k 2 19 -
W W
with only the state of California. In Figure 9 of the main draft, we exclude California counties from the time series of Panel A due to a large outlier in UI
ee : Ja Jan ee : Ja Jan
k k
W 3: n26- 25
F W 3: n26- 25
F
e e e Fe e e e Fe
W k 4: b02 b1 W k 4: b02 b1
ee - ee -
k 5 Feb Feb k 5 Feb Feb
W 9 8 W 9 8
ee : Fe -Fe ee : Fe -Fe
k 6 b1 b k 6 b1 b
: F 6-F 15 : F 6-F 15
W e W e
e b2 eb2 e b2 eb2
W ek 7 3-F 2 W ek 7 3-F 2
ee :M e ee :M e
W k 8: ar b29 W k 8: ar b29
ee Ma 1-Ma ee Ma 1-Ma
W k 9:
e r8 r7 W k 9:
e r8 r7
e k 1 Ma -Ma e k 1 Ma -Ma
r r r r
W 0: M 15-M 14 W 0: M 15-M 14
ee ee
k ar
22 ar21 k ar
22 ar21
W 11: - W 11: -
ee Ma Mar ee Ma Mar
W k r 2 Initial Lockdown Orders W k r 2 Initial Lockdown Orders
ee 12: 29-A 8 ee 12: 29-A 8
k 1 Ap pr k 1 Ap pr
W r W r
ee 3: A 5-A 4
k p pr PPP Launch ee 3: A 5-A 4
k p pr PPP Launch
W 14: r12- 11 W 14: r12- 11
ee Ap Ap ee Ap Ap
k r r k r r
W 15: 19-A 18
e A 1st Round Exhausted W 15: 19-A 18
e A 1st Round Exhausted
W ek 1 pr2 pr25 W ek 1 pr2 pr25
ee 6: 6- ee 6: 6-
W k 17 Ma May W k 17 Ma May
ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins
Low Predicted PPPE
Low PPPE
k 1 Ma 3-M k 1 Ma 3-M
W y W y
ee 8: M 10- ay9 ee 8: M 10- ay9
k1 ay Ma k1 ay Ma
W 9: M 17- y16 W 9: M 17- y16
ee M ee M
k 2 ay 2 ay2 k 2 ay 2 ay2
W 0: 4 3 W 0: 4 3
e e M -M
a a e e M -M
a a
W k y 3 y30 W k y 3 y30
ee 21: 1 ee 21: 1
k 2 Ju -Jun k 2 Ju -Jun
W n W n
ee 2: J 7-Ju 6 ee 2: J 7-Ju 6
k 2
un n1 k 2
un n1
1 1
W 3: J 4-J 3 W 3: J 4-J 3
ee un u ee un u
k 21 n20 k 21 n20
W 24: -J W 24: -J
claims that is likely due to a backlog in UI claims processing in that state. In Panel C of this figure, we can see that this backlog likely created an abnormal
ee Ju un2 ee Ju un2
W k 25 n28 7 W k 25 n28 7
e e - e e -
k 2 : Ju Jul4 k 2 : Ju Jul4
W l5- W l5-
ee 6: J ee 6: J
k u Jul1
1 k u Jul1
1
W 27: l12- W 27: l12-
High PPPE
ee Ju Jul1 ee Ju Jul1
k l19 8 k l19 8
W 2 -J W 2 -J
ee 8: J u ee 8: J u
W k 29 ul26 l25 W k 29 ul26 l25
Panel B. UI Claims (with CA)
ee : -A ee : -A
W k 30 Aug ug1 W k 30 Aug ug1
ee 2 ee 2
k 3 : Au -Au k 3 : Au -Au
W g g W g g
ee 1: A 9-A 8 ee 1: A 9-A 8
k 3 ug ug k 3 ug ug
High Predicted PPPE
2 : A 16-A 15 2 : A 16-A 15
ug u ug u
23 g22 23 g22
-A -A
ug ug
29 29
Figure F.4: PPP and UI Claims (Including the state of California)
W W
ee ee
k1 Initial UI Claims/January Employment k1 Initial UI Claims/January Employment
W : Jan W : Jan
ee 0 1% 2% 3% 4% ee 0 1% 2% 3% 4%
k 2 19 - k 2 19 -
W W
ee : Ja Jan ee : Ja Jan
k 3 n 2 k 3 n 2
W : F 26-F 5 W : F 26-F 5
e e e e e e e e
W k 4: b02 b1 W k 4: b02 b1
ee - ee -
k 5 Feb Feb k 5 Feb Feb
W 9 8 W 9 8
ee : Fe -Fe ee : Fe -Fe
k 6 b1 b k 6 b1 b
: F 6-F 15 : F 6-F 15
W e b e W e b e
e 23 b22 e 23 b22
W ek W ek
ee 7: M -Feb ee 7: M -Feb
W k 8: a r 29 W k 8: a r 29
ee Ma 1-Ma ee Ma 1-Ma
W k 9:
ee r8 r7 W k 9:
ee r8 r7
k 1 Ma -Ma k 1 Ma -Ma
r r r r
W 0: M 15-M 14 W 0: M 15-M 14
ee a ee a
k ar
2 k ar
2
W 11 2- r21 W 11 2- r21
ee : M Ma ee : M Ma
W k 12 ar29 r28Initial Lockdown Orders W k 12 ar29 r28Initial Lockdown Orders
ee ee
k 1 : Ap -Ap k 1 : Ap -Ap
W r r W r r
ee 3: A 5-A 4
k p pr PPP Launch ee 3: A 5-A 4
k p pr PPP Launch
W 14 r12 11 W 14 r12 11
ee : Ap -Ap ee : Ap -Ap
k r r k r r
W 15: 19-A 18 1st Round Exhausted W 15: 19-A 18 1st Round Exhausted
e A e A
W ek 1 pr2 pr25 W ek 1 pr2 pr25
ee 6: 6- ee 6: 6-
W k 17 Ma May W k 17 Ma May
ee : y 2 2nd Round Begins ee : y 2 2nd Round Begins
Low Predicted PPPE
Low PPPE
k 1 Ma 3-M k 1 Ma 3-M
W y W y
ee 8: M 10- ay9 ee 8: M 10- ay9
k1 ay Ma k1 ay Ma
W 9: M 17- y16 W 9: M 17- y16
ee M ee M
k 2 ay 2 ay2 k 2 ay 2 ay2
W 0 4 3 W 0 4 3
ee : Ma -Ma ee : Ma -Ma
W k 21 y 3 y30 W k 21 y 3 y30
ee 1 ee 1
k 2 : Ju -Jun k 2 : Ju -Jun
W n W n
ee 2: J 7-Ju 6 ee 2: J 7-Ju 6
k 2
un n1 k 2
un n1
1 1
W 3: J 4-J 3 W 3: J 4-J 3
ee un u ee un u
k 21 n20 k 21 n20
W 24: -J W 24: -J
ee Ju un2 ee Ju un2
W k 25 n28 7 W k 25 n28 7
ee
k : -J ee
k : -J
W 2 Jul5 ul4 W 2 Jul5 ul4
ee 6: J - ee 6: J -
k u Jul1
1 k u Jul1
1
jump in the number of UI claims in the week of April 26 to May 2nd. The results of the other analysis include state ⇥ week fixed effects and therefore are
W 27: l12- W 27: l12-
High PPPE
ee Ju Jul1 ee Ju Jul1
k l19 8 k l19 8
W 2 -J W 2 -J
ee 8: J u ee 8: J u
W k 29 ul26 l25 W k 29 ul26 l25
Panel C. UI Claims (CA only)
ee : -A ee : -A
W k 30 Aug ug1 W k 30 Aug ug1
ee 2 ee 2
not affected by these statewide reporting changes.Data is from Call Reports, SBA, and state labor deparments.
k 3 : Au -Au k 3 : Au -Au
W 1: g9 g8 W 1: g9 g8
e e - e e -
k 3 Aug Aug k 3 Aug Aug
High Predicted PPPE
2: 1 1 2: 1 1
Au 6-Au 5 Au 6-Au 5
g2 g g2 g
3- 22 3- 22
Au Au
g2 g2
9 9
Table F.1: PPP Exposure and Homebase Employment Outcomes: Partition by State UI Replacement Rates
Table F.1 reports the results of OLS regressions repeating the analysis of Table 5 after partitioning the sample into two subsamples based on
the median state UI replacement rates following the work of Ganong et al. (2020). Standard errors are clustered at the state level. ***, **,
and *, represent statistical significance at 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Bus. Shutdown Hours Worked Nbr. Employees
Zip PPPE (Round #1) ⇥ I(Month=April) 0.007 -0.003 0.001 0.006⇤⇤ 0.001 0.005⇤⇤
(0.004) (0.005) (0.003) (0.002) (0.003) (0.002)
Zip PPPE (Round #1) ⇥ I(Month=May) -0.000 -0.007 0.021⇤⇤⇤ 0.020⇤⇤⇤ 0.022⇤⇤⇤ 0.019⇤⇤
(0.007) (0.004) (0.005) (0.007) (0.006) (0.007)
Zip PPPE (Round #1) ⇥ I(Month=June) 0.003 -0.009⇤⇤ 0.034⇤⇤⇤ 0.035⇤⇤⇤ 0.035⇤⇤⇤ 0.031⇤⇤⇤
(0.007) (0.003) (0.009) (0.008) (0.011) (0.007)
Zip PPPE (Round #1) ⇥ I(Month=July) 0.007 -0.007 0.030⇤⇤ 0.035⇤⇤⇤ 0.033⇤⇤ 0.033⇤⇤⇤
(0.008) (0.006) (0.012) (0.009) (0.014) (0.009)
Zip PPPE (Round #1) ⇥ I(Month=August) 0.007 -0.008⇤ 0.029⇤⇤ 0.035⇤⇤⇤ 0.028⇤⇤ 0.033⇤⇤⇤
(0.009) (0.004) (0.012) (0.010) (0.014) (0.009)
Observations 446476 373358 446476 373358 446476 373358
Adjusted R2 0.606 0.597 0.635 0.619 0.556 0.595
Sample High Rep Low Rep High Rep Low Rep High Rep Low Rep
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
Bus. Shutdown Hours Worked Nbr. Employees
Predicted PPPE ⇥ I(Month=April) 0.005 -0.006 -0.003 0.002 -0.002 0.003
(0.003) (0.004) (0.002) (0.004) (0.002) (0.004)
Predicted PPPE ⇥ I(Month=May) -0.004 -0.009 0.009⇤⇤ 0.008 0.009⇤ 0.007
(0.004) (0.005) (0.004) (0.006) (0.004) (0.006)
Predicted PPPE ⇥ I(Month=June) -0.007 -0.005 0.022⇤⇤⇤ 0.014⇤ 0.020⇤⇤ 0.013⇤
(0.005) (0.005) (0.007) (0.007) (0.009) (0.008)
Predicted PPPE ⇥ I(Month=July) -0.003 -0.007 0.016⇤⇤⇤ 0.019⇤⇤ 0.011 0.019⇤
(0.004) (0.006) (0.005) (0.008) (0.009) (0.010)
Predicted PPPE ⇥ I(Month=August) -0.002 -0.006 0.014⇤⇤⇤ 0.018⇤⇤ 0.007 0.018⇤
(0.005) (0.006) (0.004) (0.009) (0.009) (0.010)
Observations 446476 373358 446476 373358 446476 373358
Adjusted R2 0.606 0.597 0.635 0.619 0.556 0.595
Sample High Rep Low Rep High Rep Low Rep High Rep Low Rep
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
48
Table F.2: PPP Exposure and Local Labor Market and Economic Effects: Partition by State UI Replacement Rates
Table F.2 reports the results of OLS regressions repeating the analysis of Table 6 after partitioning the sample into two subsamples based on
the median state UI replacement rates following the work of Ganong et al. (2020). Standard errors are clustered at the state level. ***, **,
and *, represent statistical significance at 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
UI claims Small Bus. Rev. OI Emp.
County PPPE ⇥ I(Month=April) -0.210⇤⇤ -0.101 0.001 -0.001 -0.000 -0.001
(0.084) (0.062) (0.006) (0.007) (0.007) (0.002)
County PPPE ⇥ I(Month=May) -0.282⇤⇤ -0.143 0.010 0.014 0.006 0.009⇤⇤
(0.104) (0.086) (0.011) (0.010) (0.005) (0.004)
County PPPE ⇥ I(Month=June) -0.266⇤⇤ -0.092 -0.006 -0.007 0.012 0.014⇤⇤⇤
(0.115) (0.105) (0.015) (0.012) (0.011) (0.004)
County PPPE ⇥ I(Month=July) -0.258⇤⇤ -0.093 -0.017 -0.027⇤ 0.008 0.018⇤⇤⇤
(0.108) (0.102) (0.013) (0.014) (0.009) (0.006)
County PPPE ⇥ I(Month=August) -0.256⇤⇤ -0.051 -0.020 -0.013 0.013 0.019⇤⇤
(0.112) (0.086) (0.013) (0.009) (0.010) (0.007)
Observations 23261 22272 13419 14974 4422 5801
Adjusted R2 0.948 0.945 0.701 0.725 0.868 0.865
Sample High Rep Low Rep High Rep Low Rep High Rep Low Rep
State⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
County Fixed Effects Yes Yes Yes Yes Yes Yes
UI claims Small Bus. Rev. OI Emp.
County Predicted PPPE ⇥ I(Month=April) -0.081 -0.068 0.001 0.001 -0.004⇤ 0.000
(0.061) (0.080) (0.005) (0.007) (0.002) (0.002)
County Predicted PPPE ⇥ I(Month=May) -0.081 -0.155 0.021 0.003 -0.007 0.010⇤⇤
(0.073) (0.130) (0.015) (0.007) (0.004) (0.004)
County Predicted PPPE ⇥ I(Month=June) -0.067 -0.160 0.023 -0.003 0.003 0.012
(0.072) (0.148) (0.020) (0.010) (0.008) (0.008)
County Predicted PPPE ⇥ I(Month=July) -0.064 -0.165 0.003 -0.013 -0.009 0.018
(0.080) (0.141) (0.013) (0.010) (0.017) (0.012)
County Predicted PPPE ⇥ I(Month=August) -0.076 -0.089 0.007 0.002 -0.009 0.023
(0.082) (0.126) (0.015) (0.010) (0.017) (0.014)
Observations 23261 22272 13419 14974 4422 5801
Adjusted R2 0.947 0.945 0.701 0.723 0.868 0.864
Sample High Rep Low Rep High Rep Low Rep High Rep Low Rep
State⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
County Fixed Effects Yes Yes Yes Yes Yes Yes
49
Table F.3: PPP Exposure and Homebase Employment Outcomes: Partition by Share Eligible for PPP in County
Table F.3 reports the results of OLS regressions repeating the analysis of Table 5 after partitioning the sample into two subsamples based on
the median share of establishments in the county that are eligible for participation in the PPP. Standard errors are clustered at the state level.
***, **, and *, represent statistical significance at 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
Bus. Shutdown Hours Worked Nbr. Employees
Zip PPPE (Round #1) ⇥ I(Month=April) 0.004 0.003 0.003 0.001 0.003 0.001
(0.004) (0.006) (0.002) (0.004) (0.002) (0.004)
Zip PPPE (Round #1) ⇥ I(Month=May) -0.005 0.001 0.017⇤⇤⇤ 0.020⇤⇤⇤ 0.016⇤⇤⇤ 0.021⇤⇤⇤
(0.005) (0.008) (0.005) (0.007) (0.005) (0.007)
Zip PPPE (Round #1) ⇥ I(Month=June) -0.008 0.003 0.033⇤⇤⇤ 0.030⇤⇤⇤ 0.029⇤⇤⇤ 0.030⇤⇤
(0.006) (0.006) (0.006) (0.010) (0.008) (0.011)
Zip PPPE (Round #1) ⇥ I(Month=July) -0.010 0.012 0.038⇤⇤⇤ 0.019 0.039⇤⇤⇤ 0.020
(0.008) (0.007) (0.008) (0.013) (0.011) (0.015)
Zip PPPE (Round #1) ⇥ I(Month=August) -0.014⇤ 0.018⇤⇤ 0.038⇤⇤⇤ 0.018 0.039⇤⇤⇤ 0.016
(0.007) (0.008) (0.008) (0.013) (0.011) (0.016)
Observations 409565 409078 409565 409078 409565 409078
Adjusted R2 0.601 0.603 0.634 0.628 0.606 0.551
Sample Hi Elig Low Elig Hi Elig Low Elig Hi Elig Low Elig
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
Bus. Shutdown Hours Worked Nbr. Employees
Predicted PPPE ⇥ I(Month=April) 0.003 -0.003 -0.001 -0.001 -0.000 -0.001
(0.003) (0.005) (0.002) (0.004) (0.002) (0.004)
Predicted PPPE ⇥ I(Month=May) -0.002 -0.011⇤ 0.002 0.014⇤⇤ 0.001 0.013⇤⇤
(0.003) (0.006) (0.004) (0.006) (0.005) (0.006)
Predicted PPPE ⇥ I(Month=June) -0.003 -0.012⇤ 0.006 0.027⇤⇤⇤ 0.002 0.025⇤⇤
(0.004) (0.007) (0.005) (0.009) (0.009) (0.010)
Predicted PPPE ⇥ I(Month=July) -0.004 -0.006 0.006 0.022⇤⇤⇤ 0.006 0.013
(0.005) (0.007) (0.006) (0.006) (0.012) (0.010)
Predicted PPPE ⇥ I(Month=August) -0.004 -0.004 0.006 0.022⇤⇤⇤ 0.005 0.013
(0.005) (0.007) (0.006) (0.006) (0.013) (0.010)
Observations 409565 409078 409565 409078 409565 409078
Adjusted R2 0.601 0.603 0.633 0.628 0.606 0.551
Sample Hi Elig Low Elig Hi Elig Low Elig Hi Elig Low Elig
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes Yes Yes
50
Table F.4: PPP Exposure and Local Labor Market and Economic Effects: Partition by Share Eligible for PPP in
County
Table F.4 reports the results of OLS regressions repeating the analysis of Table 6 after partitioning the sample into two subsamples based on
the median share of establishments in the county that are eligible for participation in the PPP. Standard errors are clustered at the state level.
***, **, and *, represent statistical significance at 1%, 5%, and 10% levels, respectively.
(1) (2) (3) (4) (5) (6)
UI claims Small Bus. Rev. OI Emp.
County PPPE ⇥ I(Month=April) -0.113⇤⇤⇤ -0.057 -0.002 0.004 0.005 -0.004
(0.035) (0.074) (0.006) (0.003) (0.004) (0.006)
County PPPE ⇥ I(Month=May) -0.141⇤⇤ -0.119 -0.001 0.017⇤ 0.009⇤⇤ 0.008
(0.057) (0.094) (0.012) (0.008) (0.004) (0.005)
County PPPE ⇥ I(Month=June) -0.103 -0.111 -0.027⇤⇤ 0.004 0.013⇤⇤⇤ 0.010⇤
(0.066) (0.091) (0.011) (0.009) (0.005) (0.005)
County PPPE ⇥ I(Month=July) -0.107 -0.098 -0.037⇤⇤ -0.011 0.017⇤⇤⇤ 0.009⇤
(0.065) (0.095) (0.016) (0.007) (0.006) (0.005)
County PPPE ⇥ I(Month=August) -0.089 -0.061 -0.031⇤⇤⇤ -0.012 0.015⇤⇤ 0.009
(0.067) (0.096) (0.011) (0.008) (0.007) (0.006)
Observations 22426 22774 20973 21487 7728 8050
Adjusted R2 0.942 0.952 0.694 0.753 0.894 0.862
Sample Hi Elig Low Elig Hi Elig Low Elig Hi Elig Low Elig
State⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
County Fixed Effects Yes Yes Yes Yes Yes Yes
UI claims Small Bus. Rev. OI Emp.
County Predicted PPPE ⇥ I(Month=April) -0.061 -0.028 -0.005 0.006 -0.004 0.000
(0.053) (0.091) (0.006) (0.004) (0.005) (0.003)
County Predicted PPPE ⇥ I(Month=May) -0.082 -0.064 -0.004 0.014 0.008 -0.003
(0.070) (0.101) (0.009) (0.009) (0.005) (0.005)
County Predicted PPPE ⇥ I(Month=June) -0.070 -0.069 -0.006 0.011 0.013⇤⇤ 0.001
(0.065) (0.108) (0.008) (0.012) (0.006) (0.007)
County Predicted PPPE ⇥ I(Month=July) -0.072 -0.071 -0.018⇤⇤ 0.001 0.012 0.002
(0.067) (0.122) (0.009) (0.009) (0.009) (0.007)
County Predicted PPPE ⇥ I(Month=August) -0.063 0.002 -0.016 -0.001 0.013 0.002
(0.069) (0.115) (0.010) (0.009) (0.010) (0.007)
Observations 22426 22774 20973 21487 7728 8050
Adjusted R2 0.942 0.952 0.692 0.752 0.894 0.862
Sample Hi Elig Low Elig Hi Elig Low Elig Hi Elig Low Elig
State⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes
County Fixed Effects Yes Yes Yes Yes Yes Yes
51
Table F.5: PPP Exposure and Homebase Employment Outcomes by Industry
This Table repeats the analysis of column (6) in Table 5 of the main draft for each Homebase Industry category
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Bus. Shutdown
Industry Food & Drink Retail Health & Fit. Prof. Serv Education Home & Repair Leisure Transp. Beauty
Zip PPPE (Round #1) ⇥ I(Month=April) 0.002 0.009 0.009 -0.006 0.019 0.013 0.013 -0.046 0.042
(0.006) (0.006) (0.015) (0.011) (0.016) (0.023) (0.032) (0.028) (0.048)
Zip PPPE (Round #1) ⇥ I(Month=May) -0.005 -0.000 0.026⇤⇤ 0.003 -0.013 0.017 -0.035 -0.073⇤ 0.108
(0.006) (0.010) (0.012) (0.016) (0.021) (0.028) (0.038) (0.036) (0.130)
Zip PPPE (Round #1) ⇥ I(Month=June) -0.008 0.003 0.032⇤ 0.014 -0.018 0.032 -0.055 -0.043 0.095
(0.007) (0.011) (0.018) (0.019) (0.021) (0.030) (0.043) (0.040) (0.143)
Zip PPPE (Round #1) ⇥ I(Month=July) -0.006 0.006 0.035⇤⇤ 0.009 -0.001 0.052 -0.074 -0.036 0.106
(0.008) (0.010) (0.014) (0.021) (0.025) (0.033) (0.045) (0.043) (0.137)
Zip PPPE (Round #1) ⇥ I(Month=August) -0.007 0.004 0.039⇤⇤⇤ 0.022 -0.007 0.056 -0.046 -0.078 0.121
(0.009) (0.010) (0.013) (0.024) (0.027) (0.036) (0.045) (0.053) (0.146)
Observations 394857 109434 52639 24001 19586 14714 11337 6026 4659
Adjusted R2 0.610 0.597 0.576 0.581 0.557 0.616 0.591 0.594 0.671
Hours Worked
Industry Food & Drink Retail Health & Fit. Prof. Serv Education Home & Repair Leisure Transp. Beauty
Zip PPPE (Round #1) ⇥ I(Month=April) 0.005 ⇤⇤
-0.001 -0.005 0.010 0.012 -0.056 ⇤⇤⇤
-0.013 0.025 -0.011
(0.002) (0.005) (0.005) (0.017) (0.011) (0.018) (0.017) (0.030) (0.047)
Zip PPPE (Round #1) ⇥ I(Month=May) 0.022⇤⇤⇤ 0.024⇤ 0.001 0.007 0.024 -0.053⇤ -0.017 0.041 -0.080
(0.006) (0.014) (0.014) (0.028) (0.020) (0.029) (0.035) (0.052) (0.104)
52 Zip PPPE (Round #1) ⇥ I(Month=June) 0.041⇤⇤⇤ 0.031⇤ 0.011 0.006 0.045 -0.053 -0.018 0.042 -0.046
(0.007) (0.016) (0.021) (0.036) (0.032) (0.034) (0.061) (0.065) (0.143)
Zip PPPE (Round #1) ⇥ I(Month=July) 0.043⇤⇤⇤ 0.023 -0.007 -0.011 0.027 -0.038 0.013 0.021 -0.046
(0.009) (0.015) (0.023) (0.032) (0.035) (0.040) (0.058) (0.068) (0.140)
Zip PPPE (Round #1) ⇥ I(Month=August) 0.043⇤⇤⇤ 0.030⇤⇤ -0.017 -0.039 0.029 -0.050 0.001 0.023 -0.020
(0.009) (0.014) (0.025) (0.034) (0.030) (0.037) (0.051) (0.081) (0.147)
Observations 394857 109434 52639 24001 19586 14714 11337 6026 4659
Adjusted R2 0.636 0.630 0.610 0.583 0.584 0.616 0.652 0.616 0.661
Nbr. Employees
Industry Food & Drink Retail Health & Fit. Prof. Serv Education Home & Repair Leisure Transp. Beauty
Zip PPPE (Round #1) ⇥ I(Month=April) 0.005 ⇤⇤
-0.002 -0.005 0.015 0.000 -0.048 ⇤⇤⇤
-0.006 0.047 -0.036
(0.002) (0.005) (0.007) (0.019) (0.012) (0.014) (0.017) (0.030) (0.059)
Zip PPPE (Round #1) ⇥ I(Month=May) 0.022⇤⇤⇤ 0.014 -0.000 0.007 0.024 -0.038 -0.004 0.077 -0.111
(0.006) (0.017) (0.011) (0.029) (0.021) (0.031) (0.030) (0.059) (0.115)
Zip PPPE (Round #1) ⇥ I(Month=June) 0.040⇤⇤⇤ 0.026 0.003 0.005 0.052⇤ -0.033 -0.021 0.044 -0.101
(0.008) (0.017) (0.017) (0.041) (0.028) (0.036) (0.069) (0.084) (0.153)
Zip PPPE (Round #1) ⇥ I(Month=July) 0.046⇤⇤⇤ 0.022 -0.025 -0.020 0.033 -0.020 0.004 -0.170 -0.129
(0.011) (0.017) (0.021) (0.037) (0.033) (0.038) (0.070) (0.241) (0.145)
Zip PPPE (Round #1) ⇥ I(Month=August) 0.044⇤⇤⇤ 0.018 -0.028 -0.053 0.035 -0.022 -0.049 -0.167 -0.104
(0.010) (0.016) (0.024) (0.039) (0.031) (0.034) (0.055) (0.264) (0.151)
Observations 394857 109434 52639 24001 19586 14714 11337 6026 4659
Adjusted R2 0.583 0.620 0.508 0.570 0.568 0.750 0.663 0.319 0.666
State⇥Industry⇥Week Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Other Control Variables Yes Yes Yes Yes Yes Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Appendix G. Bartik Diagnostics Appendix
This appendix presents diagnostic analysis of our PPP Exposure research design following
the tests suggested in Goldsmith-Pinkham et al. (2020) for “unpacking the black box” of Bartik-
style research designs. In Goldsmith-Pinkham et al. (2020), the Bartik estimator is shown to be
a weighted average of just-identified, group-specific IV estimates k , where k denotes a group.
The weights ↵k are called Rotemberg weights. They combine the group-specific shock g k with
the covariance of group-region shares zlk and the endogenous variable X to be instrumented.
The estimand of interest is the effect of X on some outcome Y. (See Proposition 3, p.2600.)
In our setting, k refers to a bank, g k is bank-level PPPE, and zl k are bank-ZIP-level shares
of the number of branches. In the main text, we focus on the reduced form relation between
our Bartik instrument and various outcomes. Here, we use a two-stage setup where Y is the
mean ZIP-level change in hours worked between the base period and week 18 (May 17–May
23) and X is the ZIP-level number of PPP loans received at the end of round one relative to the
number of eligible establishments.
For reference, in this setup, the Bartik estimator Bar t ik estimated by 2SLS with standard
errors clustered at the state-level is 0.41 (s.e.=0.099) with an F statistic of 1150. The regression
in this setup conditions on the controls in our main regression in Table 5.
Table G.1 lists the ten banks with the most positive and most negative Rotemberg weights
in the Bartik IV implementation of our PPPE research design. These weights reflect the banks’
respective influence in the Bartik instrument, which combines bank-ZIP-level shares of the
number of branches with bank-level PPPE. First, the table suggests that influential banks tend
to be either large or mid-sized banks and those banks with PPPE that points to substantial
overperformance or underperformance. For example, JP Morgan, Citibank, and Wells Fargo
all have large Rotemberg weights. Second, Wells Fargo’s weight is the only negative weight
with absolute magnitude over 0.01. This fact indicates that more of the identifying variation
comes from banks with positive weights, which enables the Bartik estimator to be interpreted
as a LATE. Finally, the table suggests that the Bartik estimator reflects contributions from a
large number of banks, as the top ten banks in terms of Rotemberg weight only contribute
approximately 20% of total weight to the Bartik estimator. This result suggests our approach
is not driven just by one or two banks, or even by the top-4 banks alone.
Table G.2 reports statistics about the Rotemberg weights in the Bartik IV implementation
of our PPPE research design (following Table 1 in Goldsmith-Pinkham et al. (2020)). These
weights reflect the banks’ respective influence in the Bartik instrument, which combines bank-
ZIP-level shares of the number of branches with bank-level PPPE. Panel A shows that banks with
53
positive Rotemberg weights account for two-thirds of observations while banks with negative
weights account for one-third of observations. However, Panel C shows that negative weights
contribute relatively little to the overall estimator.
We also considered several of the Bartik diagnostics advocated by Goldsmith-Pinkham et al.
(2020). Table G.3 presents bivariate regressions of bank branch shares (zl k ) on ZIP-level observ-
ables. Both branch shares and observables are residualized with respect to state dummies. Vari-
ables have been normalized, so the coefficients can be interpreted as a one-standard deviation
change in x produces a -standard deviation change in branch shares, where is the reported
coefficient. While bank-branch shares are modestly correlated with local observables—for ex-
ample, branch concentration is higher in low income and less dense areas—all correlations are
below 0.1 in absolute value, supporting our identification assumption of approximate indepen-
dence of bank branch shares from the outcomes of interest.
Goldsmith-Pinkham et al. (2020) also advocate plotting pre-trends for research designs
where regions are sorted based on the level of the Bartik instrument. We show in Figures 7
and 9 that pre-COVID, pre-PPP trends are very similar for high and low PPPE regions for a
variety of outcomes. Our local projections in Figures 8 and 10 shows parallel trends in the
short window after COVID lockdowns and prior to the PPP, once we condition on differences
in targeting.
54
Table G.1: Unpacking PPP Exposure: Influential Banks
Table G.1 lists the ten banks with the most positive and most negative Rotemberg weights in the Bartik IV im-
plementation of our PPPE research design. These weights reflect the banks’ respective influence in the Bartik
instrument, which combines bank-ZIP-level shares of the number of branches with bank-level PPPE. ↵k refers to
Rotemberg weight for bank k. g k is bank-level PPPE. k is the just-identified coefficient. See Goldsmith-Pinkham
et al. (2020) for additional discussion of these diagnostics.
Financial Institution Name ↵k gk k
Top 10 Positive ↵k
KEYBANK NATIONAL ASSOCIATION 0.0667 0.3868 -0.1196
MANUFACTURERS AND TRADERS TRUST COMPANY 0.0585 0.3964 -0.2267
JPMORGAN CHASE BANK, NATIONAL ASSOCIATION 0.0264 -0.3596 -0.5714
CITIZENS BANK, NATIONAL ASSOCIATION 0.0244 0.2529 -0.2744
HUNTINGTON NATIONAL BANK, THE 0.0146 0.4204 1.4877
NBT BANK, NATIONAL ASSOCIATION 0.0143 0.3351 -0.2642
BANCORPSOUTH BANK 0.0123 0.3998 0.0079
CITIBANK, N.A. 0.0115 -0.4552 0.6701
CAPITAL ONE, NATIONAL ASSOCIATION 0.0113 -0.4988 0.2671
UNITED COMMUNITY BANK 0.0111 0.4436 0.6872
Bottom 10 Negative ↵k
WELLS FARGO BANK, NATIONAL ASSOCIATION -0.0353 -0.4849 0.8005
TIMBERWOOD BANK -0.0055 -0.5000 -0.8560
BANK OZK -0.0053 0.4067 -2.0285
COMERICA BANK -0.0048 0.2157 0.3192
BANK OF HAWAII -0.0045 0.4407 0.5173
SOUTH STATE BANK -0.0042 -0.5000 -1.7268
PEDESTAL BANK -0.0041 -0.5000 -1.7775
BOKF, NATIONAL ASSOCIATION -0.0041 0.4705 1.0714
TRI CITY NATIONAL BANK -0.0040 0.4678 0.2811
WESBANCO BANK, INC. -0.0040 0.3507 -1.9260
55
Table G.2: Unpacking PPP Exposure: Summary of Rotemberg Weights
Table G.2 reports statistics about the Rotemberg weights in the Bartik IV implementation of our PPPE research
design (following Table 1 in Goldsmith-Pinkham et al. (2020)). These weights reflect the banks’ respective influ-
ence in the Bartik instrument, which combines bank-ZIP-level shares of the number of branches with bank-level
PPPE. ↵k refers to Rotemberg weight for bank k. g k is bank-level PPPE. k is the just-identified coefficient. V(zk )
refers to the variation in bank-ZIP-level shares within a bank across ZIPs. See Goldsmith-Pinkham et al. (2020)
for additional discussion of these diagnostics.
Sum Mean Share (N)
Panel A. Negative and positive weights
Negative -0.6202 -0.0004 0.3348 (1,716)
Positive 1.6202 0.0005 0.6652 (3,409)
↵k gk k V(zk )
Panel B. Correlations
↵k 1.0000
gk 0.0405 1.0000
k 0.0013 -0.0175 1.0000
V(zk ) 0.0475 -0.0296 -0.0026 1.0000
↵-weighted Share of
sum overall Mean
Panel C. Estimates of k for positive and negative weights
Negative -0.0034 -0.0096 3.7104
Positive 0.3578 1.0096 -5.7637
56
Table G.3: Correlates of Bank Branch Shares
Table G.3 presents bivariate regressions of bank branch shares (zlk ) on ZIP-level observables. Both branch shares
and observables are residualized with respect to state dummies. Variables have been normalized, so the coef-
ficients can be interpreted as a one-standard deviation change in x produces a -standard deviation change in
branch shares, where is the reported coefficient. ***, **, and *, represent statistical significance at 1%, 5%, and
10% levels, respectively, with standard errors clustered at the state level. See Goldsmith-Pinkham et al. (2020)
for additional discussion of these diagnostics.
LHS is Bank Deposit Shares
Coefficient R2 N
Log(Population) -0.032*** 0.0009 668583
(0.004)
Log(Population Density) -0.068*** 0.0034 668583
(0.007)
Social Distancing 0.068*** 0.0023 669985
(0.010)
Covid Cases per Capita -0.071*** 0.0013 670219
(0.008)
Deaths per Capita -0.030*** 0.0005 670219
(0.003)
Unemployment Filing Ratios 0.023 0.0001 444375
(0.016)
Revenue Change of Small Business -0.048*** 0.0012 670219
(0.012)
Median Household Income -0.055*** 0.0019 670204
(0.009)
57
File and source
- File
- zwick-sba-appendix-ppp-hit-target.pdf
- Size
- 4,878,189 bytes
- SHA-256
- ebb47b804f88ab40e0629589bb44864c56ca0840de22aff10384c7533043d8b9
- Original
- No public link identified.