Olga Gorbachev et al., The Tail That Wagged the Dog: What Explains the Persistent Employment Effect of the 10-Day PPP Funding Delay? (2023)
Summary
Federal Reserve Bank of Boston working paper No. 23-6, version of July 2023, by Olga Gorbachev, María J. Luengo-Prado and J. Christina Wang. It studies why the 10-day funding delay in the 2020 Paycheck Protection Program had a large, persistent effect on employment, as estimated by Doniger and Kay (2021). The authors find that the top 1 percent of urban counties by population fully account for the effect, which they attribute to those areas' persistent vulnerability to infectious disease. They also find that receiving more PPP funds and more transfers from other pandemic programs did more for local recovery than receiving PPP funds earlier.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Olga Gorbachev et al., The Tail That Wagged the Dog: What Explains the Persistent Employment Effect of the 10-Day PPP Funding Delay? (2023).”
Original at www.bostonfed.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/olga-gorbachev-et-al-the-tail-that-wagged-the-dog-what-explains-the-pe-98606fc681/
, SHA-256 8381377752c60c9a4ec01fdd932028e5d83669eb88f6117b2067c70b2c06e116.Full text
No. 23-6
The Tail That Wagged the Dog: What Explains the
Persistent Employment Effect of the
10-Day PPP Funding Delay?
Olga Gorbachev, María J. Luengo-Prado and J. Christina Wang
Abstract:
This study explores the mechanisms explaining the large, persistent effect of the 10-day funding
delay in the 2020 Paycheck Protection Program (PPP) on employment recovery during the
COVID-19 pandemic, as estimated by Doniger and Kay (2021). We find that the top 1 percent of
urban counties by population fully account for the significant effect of the delay on county-level
employment. The strong correlation between worse loan delay and slower employment growth
in these counties is due to a factor commonly omitted from analyses: The nature of business and
the high rate of human interactions in major urban centers render these areas exceptionally and
persistently vulnerable to infectious diseases. Moreover, we find that receiving more PPP
funding and more transfers from other pandemic-related assistance programs contributed
significantly more to local economic recovery compared with receiving PPP funds earlier.
JEL Classifications: H81, G28, J21, E24
Keywords: COVID-19, Paycheck Protection Program, small business credit, remote work,
employment
Olga Gorbachev is an associate professor of economics at the Lerner College of Business and Economics at the
University of Delaware. María J. Luengo-Prado and J. Christina Wang (christina.wang@bos.frb.org) are senior
economists and policy advisors in the Federal Reserve Bank of Boston Research Department.
The authors thank Rachel Cummings for research assistance and Daniel Cooper for helpful comments.
This paper presents preliminary analysis and results intended to stimulate discussion and critical comment.
The views expressed herein are those of the authors and do not indicate concurrence by the Federal Reserve Bank of
Boston, the principals of the Board of Governors, or the Federal Reserve System.
This paper, which may be revised, is available on the website of the Federal Reserve Bank of Boston at
https://www.bostonfed.org/publications/research-department-working-paper.aspx.
This version: July 2023 https://doi.org/10.29412/res.wp.2023.06
1 Introduction
As the COVID-19 pandemic swept through US coastal cities in March 2020, Congress
passed the Coronavirus Aid, Relief, and Economic Security (CARES) Act, dispensing
broad-based fiscal assistance on an unprecedented scale. A key component of the fiscal
package was the Paycheck Protection Program (PPP), which provided loans that were
essentially grants to most small businesses whose operations were disrupted by the
pandemic.1 The stated goal of the PPP was to enable small businesses to retain workers
despite having to curtail operations or shut down entirely.
This study focuses on understanding the likely mechanisms through which the PPP
affected the dynamics of the recovery after the initial acute phase of the pandemic in
conjuction with other components of the CARES Act. We find that the persistently
adverse effect of a brief delay in PPP funding found by Doniger and Kay (2021) (who
will be refer to as DK throughout the rest of the paper) was driven by a very small
number of large urban counties, counties that were intrinsically more vulnerable to the
lasting damage inflicted by COVID-19. Accordingly, we show, the receipt of PPP funds
was more important for business recovery and survival than an earlier timing of the
receipt. Furthermore, we quantify the relative importance of the multiple CARES Act
assistance programs (namely, the PPP, expanded unemployment insurance payments,
and tax rebates) in shaping the employment recovery and find that these other programs
were relatively more consequential for the recovery.
In a influential paper, DK estimate the PPP’s treatment effect on employment using
the share of PPP funds delayed due to a 10-day legislative funding pause, which we refer
to as the “share delayed” in all subsequent analysis.2 They document that cities with a
higher share delayed suffered a persistently slower (total) employment recovery through
October 2020. DK attribute the persistent deleterious impact of the share delayed on
the local economy mostly to the crucial importance of liquidity for small businesses.
There is, however, a likely more plausible explanation for DK’s findings. Given the
exceedingly contagious nature of COVID-19 and the public health measures adopted to
contain its spread, it is reasonable to expect, a priori, that economic activity in highly
populated major metropolitan areas would be more adversely affected both at the start
of the pandemic and for a longer period. The more serious disruption of commercial
activity in these areas at the outset likely also resulted in slower underwriting of PPP
loans coupled with greater excess loan demand, and hence a higher share of loans
delayed. These urban areas then suffered more lasting damage after the COVID-19
outbreak, in large part because they experienced a more pronounced shift to remote
work. One important consequence of this structural change is that it impeded the
recovery of local employment that supported office workers. In short, the share delayed
loads on the common omitted risk factor of being a major urban center.
Our analysis indeed confirms this mechanism. First, investigation of the determi-
nants of the share delayed reveals that the share delayed was significantly larger in the
top 1 percent of urban counties in terms of population than in all the other counties,
even after we control for a rich set of covariates. Next, using county-level Quarterly
1
With only a few exceptions, “small” refers to businesses with no more than 500 employees.
2
Lenders had to pause for 10 days after the first round of PPP funding was exhausted mid-day on
April 16, 2020. Lending resumed on April 27 after Congress appropriated additional funds.
2
Census of Employment and Wages (QCEW) data on private employment,3 we find that
the importance of the share delayed in explaining the slow employment recovery is en-
tirely accounted for by these most populous counties, as the share delayed has neither
a significant nor a persistent impact on the employment evolution in the remaining 99
percent of urban counties and no explanatory power for the recovery in rural counties.
Furthermore, we find a greater persistent shift to remote work in the largest urban
counties, on average, which is consistent with the conjecture that remote work took a
lasting toll on local employment in these areas.
Our results based on county-level employment also highlight another important dis-
tinction: The effect of PPP timing most likely differs from the effect of PPP receipt.
Therefore, if we think the policy-relevant treatment effect (as defined in Heckman and
Vytlacil 2001) is PPP loan receipt, then the estimated effect of loan delay is not the
relevant measure, even if the share delayed were conditionally random. To more pre-
cisely isolate the effect of loan receipt from that of loan timing, we turn to firm-level
data and compare business foot traffic between PPP recipients and narrowly matched
(based on geography and line of business) peer firms that did not receive PPP funds.
To the extent that the 10-day delay restrained employment growth for many months
because it deprived some firms of a timely injection of liquidity, we would expect late
borrowers (defined as those that received loans just after the 10-day pause) to per-
form worse than early borrowers (defined as those that received loans just before the
10-day pause) and to do so for the lengths of time documented by DK. Applying ei-
ther a difference-in-differences matching estimator or a staggered treatment estimator
à la Sun and Abraham (2021) to firm-level data from SafeGraph (SG), we find that
early and late borrowers both benefited substantially from receiving PPP loans.4 Both
groups of borrowers saw significantly more visits starting in the second half of 2020 and
significantly fewer closures (defined as showing zero visits) relative to their respective
matched non-recipient peers. In contrast, the treatment-effect differential between early
and late PPP borrowers is rather small and statistically insignificant. Our estimates
show that receiving a PPP loan was much more important than receiving it a little
earlier. This finding provides further support for the interpretation that the significant
effect of the share delayed on city employment found by DK is likely due to confounding
factors.
Finally, we assess the relative importance of different components of the CARES
Act on the employment recovery. The fiscal policy response to COVID-19 was unprece-
dented in both scale and scope. The CARES Act and subsequent legislation authorized
multiple types of spending and transfer payments to support businesses and households.
In addition to the PPP, two other important transfer programs were (1) expanded and
enhanced unemployment insurance benefits and (2) tax rebates (popularly known as
“stimulus checks”). All these payments likely cushioned the pandemic’s impact on in-
come and hence influenced the trajectory of the subsequent employment recovery. To
3
We use QCEW data on private employment instead of individual-level CPS data, which DK use,
because the former are more pertinent to the stated goals of the PPP program—to preserve employee-
employer matches—and the QCEW provides almost universal coverage of private establishments and
geographies. Nevertheless, we essentially replicate DK’s analysis using the CPS data (see section 5.2)
and document the factors (including different data coverage) that contribute to differences between
our findings and theirs.
4
SG records the number of visits and visitors to business locations using cell phone tracking data.
3
the extent that the amount of assistance a community received through each stimulus
is somewhat correlated across programs, not adequately accounting for the influence of
other payments would bias the estimate of a specific program’s impact. Thus, it is use-
ful to gauge the effect of these programs jointly to arrive at a more holistic assessment
of the relative efficacy of the menu of pandemic programs, which can help inform policy
choices should similar situations arise in the future. Comparing the relative contri-
bution of several pandemic fiscal programs, we find that extra unemployment benefits
(with a negative sign) and tax rebates (with a positive sign) account for a larger fraction
of the cross-county variation in employment recoveries than the amount of total PPP
funds received and orders of magnitude more than the slowdown due to the PPP delay.
The remainder of the paper is organized as follows. Section 2 briefly reviews selected
existing studies of the PPP. Section 3 describes our methodology and the data sources
used in our analysis. Section 4 studies the determinants of the share delayed. Section 5
presents our main empirical findings on how the timing and volume of PPP loans, along
with other pandemic-related transfer programs, affected the local employment recovery.
Section 6 illustrates the importance of PPP receipt relative to the timing of its receipt
using SG data. Section 7 concludes.
2 Related Literature
There is a rapidly growing literature studying the effect of the PPP on small businesses’
ability to continue operations after the COVID-19 outbreak and, in turn, on the pace
of the local economic recovery after the initial lockdown phase.5 The rough consensus
has been that the program cushioned the pandemic-related loss of employment (with
most estimates pointing to a marginal impact) but possibly at a high price tag to
taxpayers due to its general lack of targeting (for instance, not limited to the most
affected industries or the hardest hit regions).6 Granja et al. (2022), Balyuk, Prabhala,
and Puri (2020), and Li and Strahan (2021) show that firms with strong bank ties
received better access to PPP loans, especially in the first phase of the program (before
April 17, 2020).
Doniger and Kay (2021) make original use of the 10-day pause of the PPP (the
first round of authorized funding ran out on April 16, 2020) as a plausibly exogenous
variation in loan timing to identify the causal effect of a delay in PPP funding on
local employment.7 DK argue that where the funding pause interrupted the queue of
applicants was random because the borrower composition just before versus just after
the 10-day delay was essentially the same in every CBSA, but the share of loans delayed
differed substantially across CBSAs. Nontrivial harm could have resulted from even a
brief delay at the critical juncture of the COVID-19-induced lockdowns, which forced
many small businesses to shutter at least temporarily and caused revenues to plummet
5
In addition to the studies cited explicitly in this section, a partial list of other studies includes
Autor et al. (2022), Faulkender, Jackman, and Miran (2023), Chetty et al. (2020), and Hubbard and
Strain (2020).
6
For example, Granja et al. (2022) estimate that the PPP saved jobs at a cost per job year of at
least $175,000.
7
DK use the Current Population Survey (CPS) data and conduct their analysis at the individual
level. They match CBSA-level measures of the share of PPP loans delayed to each CPS respondent
based on location identifiers available in the CPS interviews.
4
or even cease entirely, because small businesses generally have limited cash reserves. In
attributing the persistent deleterious impact of the share delayed on the local economy
mostly to the crucial importance of liquidity for small businesses, DK cite the estimate
of the employment effect of an exogenous liquidity shock by Barrot and Nanda (2020).
However, the measure of funding timing in Barrot and Nanda (2020) is a permanent
flow shock that frees up a portion of the stock of working capital (which firms can
deploy to other uses). In contrast, the 10-day PPP funding delay was a brief, one-time
flow shock (albeit rather large).8
Several studies have since used this share delayed to examine the causal impact of the
PPP. For example, Kurmann, Lalé, and Ta (2022) find that, in terms of employment,
small businesses in four of the hardest hit service sectors contracted more severely at
the beginning of the pandemic compared with larger businesses in those sectors, but
they also recovered more strongly afterward.9 Closings and reopenings account for the
bulk of the initial contraction and the subsequent rebound, respectively. Cole (2022)
uses administrative data on private payrolls for very small firms (with a median of
five employees) and documents that, as we find using Safegraph data, it is the receipt
of PPP funding that was important for employee retention and growth rather than
the timing of the funding. PPP borrowers showed greater employment growth than
their peers in the five months following receipt of the loan. Joaquim and Wang (2022)
also make the point that the effect of the funding delay is conceptually, and likely
empirically, different from the effect of the funding receipt, although they provide no
direct empirical evidence.
We contribute to this literature by conducting a more in-depth analysis of the mech-
anisms through which the 10-day delay in 2020 PPP funding was associated with con-
siderable delay in the employment recovery after spring 2020. Our finding that most
of the explanatory power of the funding delay stems from the most populous counties
means that care is needed when interpreting the magnitude of the estimates of studies
that exploit the 10-day delay discontinuity as representing the treatment effect of the
PPP. Several intuitive reasons, none related to credit, can explain why a highly infec-
tious disease such as COVID-19 would be expected to exert more severe and lasting
harm on major metropolises. In short, the pandemic-induced fundamental damage to
the economic advantage of densely populated central business districts (including, for
example, the shift toward remote work)—not the delay of credit—likely explains their
slower and shallower recovery.
8
Even the argument that establishments would have kept their employees and stayed in business
had they received PPP funding sooner receives at best limited support: Doniger and Kay (2021) find
that the number of firms that closed during those 10 days was minimal, even though the funding delay
led to significantly higher nonemployment rates among self-employed individuals and independent
contractors.
9
This study uses Homebase data. Homebase is a scheduling and time-tracking tool that many small
businesses use to track their hourly employees.
5
3 Empirical Design and Data
3.1 Empirical Design
To investigate the impact of PPP loans on the local economy, we start with an em-
pirical specification analogous to that used by DK. Instead of using individual-level
data on employment status from the CPS, our analysis uses county-level employment
at private employer businesses from the QCEW. Two advantages of our data are worth
noting. First, employee count at private businesses (as opposed to total employment) is
more pertinent for evaluating the PPP, a chief goal of which was to preserve employer-
employee matches. Second, relative to that of the QCEW, the geographical coverage
of the CPS is more limited, as will be discussed in Section 5.2.
Specifically, we estimate the following equation:
Yc,t = µc + τs,t + βt Share-Delayedc + γt Xc,t + ηYc,t−12 + c,t , (1)
where Yc,t is the measure of local employment of interest in county c and month t.10 µc
and τs,t are county and state-by-month fixed effects, respectively.11 Xc,t denotes a set of
county-level controls described in Section 3.2 that are allowed to influence the economic
outcome Yc,t differently in each month. Yc,t−12 is the 12-month lag of the dependent
variable, which directly controls for its pre-condition. Standard errors are clustered at
the county level.
The regressor of interest is Share-Delayedc , denoting the share of PPP funds delayed
due to the 10-day pause in lending that occurred after the first round of PPP funding
was exhausted mid-day on April 16, 2020. Lending resumed on April 27 after Congress
appropriated additional funds. Let Lc denote funds received late (on April 27 and
28, 2020, just after funding resumed) and Ec denote funds received early (on April 14
through 16, just before funding ran out). We follow DK and define the share delayed
as follows:
Lc
Share-Delayedc = .
Lc + Ec
To be consistent with DK, we measure the share delayed using the volume of PPP
loans.12 The share delayed is interacted with a full set of monthly indicators to flexibly
model the effect of the funding delay over time. DK argue this measure is as good
as being randomly assigned, conditional on the local characteristics controlled for in
their regressions. To the extent that the share delayed represents exogenous variation
in the timing of PPP lending, βt measures the monthly impact of a delay in access to
credit on county employment relative to March 2020 but not necessarily the impact of
the credit receipt itself. Additional controls in the regressions are discussed next and
10
We measure employment in levels, rather than logs or growth rates, so that our specifications
are comparable to DK’s. Results using logs or growth rates are reported in Appendix Tables A.6-
A.9. These show that the estimated impact of Share-Delayed is quite sensitive to small specification
changes.
11
Note that all regressions control for state-by-month fixed effects, removing any state-specific time-
varying differences in policy response (such as nonpharmaceutical interventions) to the COVID-19
crisis.
12
Note that share delayed could also be computed in terms of the number of loans. Results using
loan counts instead of loan amounts are very similar and thus omitted for brevity.
6
include a wide range of preexisting conditions, the initial severity of COVID-19 spread
and public health measures in response, expanded unemployment insurance benefits,
and tax rebates. Their effects will be measured by γt .
3.2 Data
This section provides a brief description of the array of data used in our analysis.
Summary statistics are reported in Table 1, and more details are supplied in Appendix
A.
Employment and Establishment Counts We use the most timely high-frequency
(monthly) employment data available at the county level, the Quarterly Census of Em-
ployment and Wages (QCEW). The QCEW employee counts cover more than 95 percent
of U.S. jobs, which is pertinent for evaluating the PPP, a chief goal of which was to
preserve employer-employee matches. Our sample spans 2019:Q1 through 2021:Q3. We
also explore the impact of the share delayed on the number of operating establishments.
Such counts are available in the QCEW at a quarterly frequency.
For meaningful cross-county comparisons in terms of cumulative PPP receipts, we
normalize total PPP amounts by 2019 employment levels at small establishments, which
are available in the Quarterly Workforce Indicators (QWI), but not in the QCEW.
We also use data from the 2019 County Business Patterns (CBP) on the number of
establishments by size (range of employee counts) at the county-by-NAICS-industry
level to normalize pre-pandemic county-level small-business lending (SBL) amounts (to
be described later). These detailed establishment counts also are not available in the
QCEW.
PPP Loans and Borrowers For PPP lending, we use official data released by the
Small Business Administration (SBA).13 We use the business name and full address
of each borrower to obtain a unique Placekey identifier. These identifiers then map
the borrowers to their US Census county (or CBSA), which allows us to compute loan
statistics at the desired level of geography, including the share of loans delayed and the
total volume of funds received.14
Preexisting Local Conditions Since bank relationships were an important aspect
of the first phase of the program (see Li and Strahan 2021), we control for the fol-
lowing indicators of local banking market conditions as of 2019: bank branch density
(number of bank branches normalized by population), community banks’ and largest
four banks’ shares of deposits, and the 2019 volume of SBL in each county (from
data reported pursuant to the Community Reinvestment Act). Using 2019:Q1 CBP
data, we normalize SBL by the number of small establishments. We further control
for the following pre-pandemic local demographic and economic conditions using data
13
We use PPP data released as of August 2021 and available at https://www.sba.gov/
funding-programs/loans/COVID-19-relief-options/paycheck-protection-program/ppp-data.
14
Placekey is a free, universal standard identifier for any physical place. For more details, see
https://docs.placekey.io/ and the corresponding white paper. Placekeys in fact allow us to narrow
down borrowers’ locations to census block groups (CBGs), which we employ in later analysis to match
PPP recipients to non-PPP counterparts using SafeGraph data.
7
from the American Community Survey (ACS): population, median family income, and
commuter-to-resident population ratios. We also classify counties as urban or rural
using the 2013 National Center for Health Statistics classification scheme.
COVID-19–Related Factors Counts of COVID-19 cases and deaths are provided
by Johns Hopkins University. The extent of county-level lockdowns is measured as the
share of days in lockdown for early (before April 17, 2020) and late (April 17 through
30, 2020) periods using data from the Keystone-Strategy’s COVID-19 Intervention data
set. To account for potential heterogeneous effects of the COVID-19 shock due to pre-
pandemic industry composition, we control for the share of employees working in the
most adversely affected industries (with two-digit NAICS codes 44/45, 61, 62, 71, 72,
and 81) and the share of employees working in essential industries as defined by the US
Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency
(DHS-CISA). To measure the pandemic’s impact on work patterns, we also use Google
Mobility data provided by Opportunity Insights. In particular, we focus on total time
spent at workplace locations.15
Additional Public Support Programs We also control for industry-weighted,
county-specific unemployment insurance (UI) benefit replacement rates (relative to pre-
pandemic levels). For each industry in a given county, we compute a UI replacement
rate using weekly wages earned by private employees in those industries and the UI
weekly benefits determined by each state’s UI laws supplemented by the CARES Act’s
Pandemic Unemployment Compensation (PUC) payment of $600 a week through July
2020 and a $300 weekly supplemental payment extension.16 Our estimates also account
for the differential end dates in UI extensions, which occurred in some states before
the federal September 6, 2021, deadline. We then compute a county-level measure as
a weighted average of the industry-specific UI replacement rates using each industry’s
2019 employment share in a given county as the weight. We also collect IRS data on
county-level 2020 rebates (that is, stimulus checks) to households. The rebates are
normalized by population to derive a per capita amount.17 Finally, using SBA data on
PPP loans, we compute, for each month, the cumulative sum of PPP receipts for each
county up to this point and normalize it by total employment in small (fewer than 500
employees) establishments as reported in the 2019:Q1 QWI.
4 What Explains the Share of PPP Loans Delayed?
The central assumption underlying the validity of causal inferences using the share of
PPP loans delayed is that the assignment of the share delayed across localities occurs
in a random-like fashion and is conditionally uncorrelated with any unobservables that
could have affected the employment recovery. DK’s argument is that the share delayed
satisfies this requirement because the point where the queue of PPP applicants was
15
For details on the data, see Chetty et al. (2020). Data can be downloaded at https://github.
com/OpportunityInsights/EconomicTracker.
16
We estimate industry-specific UI replacement rates using UI formulae as coded in Ganong, Noel,
and Vavra (2020).
17
County-level data on rebates for 2021 had not been released at the time of our analysis.
8
interrupted when the initial funding ran out—which determines the share of loans
delayed—was as good as random. We investigate this assumption at the county level,
modeling the first-order (that is, linear) relationship between the share delayed and
conditions before the start of the delay window (on April 16, 2020) that can plausibly
affect both the volume of loans delayed and the subsequent economic recovery.
The coefficients reported in Table 2 show that the share of loan volume delayed is
closely related to several county attributes, especially small businesses’ existing lend-
ing relationships.18 Areas with greater volumes of 2019 SBL per small establishment
experienced smaller volumes of delayed loans: A one-standard-deviation (SD) higher
level of SBL per small establishment is associated with a 0.12 SD lower share of vol-
ume delayed.19 Moreover, urban counties with a larger share of community banks saw
more delay, likely because smaller banks were less able to meet the high demand due
to capacity constraints.20
Moreover, differences in the determinants of the share delayed between urban and
rural counties likely relate to how the virus transmission depended on the size of the
potentially affected population and the people’s modes of interaction. On average,
urban counties had a lower share of lending delay by loan volume than rural counties
(0.44 versus 0.50). However, urban counties with higher COVID-19 case rates early on
had a significantly higher share of delay. This correlation for urban counties supports
the conjecture that areas hit harder by COVID-19 suffered a greater disruption to
commercial activity at the outset of the pandemic, which also impeded PPP lending.
The greater demand combined with impaired supply resulted in worse delays in PPP
lending. In fact, the most populous counties (the top 1 percent in terms of population),
arguably the counties most disproportionately impacted by COVID-19, experienced a
significantly larger delay in the volume of funding even when all the covariates are
taken into account.21 Overall, preexisting county attributes explain a higher fraction
of the cross-county variation in the share delayed for urban counties than for rural
counties—about 7 percentage points higher in terms of the adjusted R-squared.
Given that the share delayed can be explained by several preexisting local attributes,
we include these covariates in our employment regressions and allow for their effects to
influence employment dynamically over the sample months (that is, differently in each
month), as we do with the share delayed.
5 Effects of the PPP on Employment
We now turn to the effects of the PPP on county-level employment. We first focus on
the share delayed and later extend our analysis to include the impact of total funding
18
Results based on the number of loans delayed are qualitatively similar and reported in Appendix
Table A.1.
19
Alternatively, counties at the bottom 5th percentile of SBL per small establishment experienced
a 0.39 SD higher delay in funding than counties at the 95th percentile.
20
Before the pandemic, the share of small firms that borrowed from community banks in urban
counties was lower than the share that borrowed from community banks in rural counties, on average.
This hampered small urban firms’ chances of obtaining a PPP loan early in the pandemic. As Balyuk,
Prabhala, and Puri (2020) show, early in the PPP lending program, small firms were better able to
obtain funding from community banks with which they had a preexisting relationship.
21
Their unconditional mean of share delayed is 0.50, as high as that of rural counties.
9
receipts from the PPP and other federal programs.
5.1 Effects of the PPP Delay on QCEW County Employment
Table 3 presents the coefficient estimates from Equation (1). Columns (1) through (4)
show results pooling all counties together, columns (5) and (6) show results for urban
counties only, columns (7) and (8) for urban counties excluding the top 1 percent most
populous ones, and (9) and (10) for rural counties. All specifications include state-by-
month and county fixed effects, and the additional controls included are listed below
the coefficients. In a nutshell, we find that the impact of the funding delay on the local
recovery was driven by the top 1 percent most populous urban counties.22
First, when all the counties are pooled together (column 1), the share delayed is
found to have a lasting impact on county-level employment, even after we control
for pre-COVID-19 county attributes such as median family income, the commuter-
to-residential-population ratio, urban-rural designation, banking and small-business-
lending preexisting conditions, and the 12-month lag of the dependent variable.
More controls are added in column (2) to account for the pandemic’s impact on
public health (cumulative death and case rates per million population), the contain-
ment measures adopted early on, and the likely differential vulnerability of counties
to COVID-19 (the pre-pandemic share of essential employees in a given county and
the pre-pandemic employment share of industries most adversely affected by the crisis,
specifically NAICS 44/45, 61, 62, 71, 72, and 81). Including these additional controls
reduces the impact of the share delayed by about 20 percent.
These controls likely exert influence through several channels. Counties that im-
posed lockdowns early (before April 16, 2020) likely experienced worse disruptions to
commercial activity, which would have raised the demand for PPP loans while ham-
pering banks’ ability to underwrite the loans, resulting in a higher share delayed, on
average. These counties also tended to have higher population density and saw more
infections and deaths early in the pandemic. Their inherent vulnerability to infec-
tious diseases also would make the recovery process slower. Likewise, a higher share
of employment in the most adversely affected industries could have impacted counties’
recovery, but the effect could have been positive or negative; recovery could have been
more rapid if the initial loss of employment was sufficiently greater or more gradual
due to slower recovery of demand for in-person services. Conversely, a higher share of
essential employment likely meant relatively more business activity following the out-
break (as essential businesses could remain open or reopen faster), which would show
up as a more positive trajectory of employment over time.
Next, we control for the plausibly exogenous amounts of federal pandemic assistance
received by each county: tax rebates per capita and industry-weighted UI replacement
rates (column 3). UI replacement rates are fully determined by 2019 wages and thus
do not depend on the actual payout of unemployment benefits and are not driven by
local demand for unemployment benefits. Similarly, the tax rebates are predetermined
by local characteristics of taxpayers in 2019. The cross-county variation in funding
via these programs is also exogenous with respect to PPP receipt (or PPP delay), as
22
We choose a 1 percent cutoff to demonstrate that even a minimal change in sample, in accordance
with the mechanism we propose, is sufficient to eliminate the effect of the share delayed. Expanding
the cutoff to a larger percent (top 5 or 10 percent) only makes the point stronger, see Table A.5.
10
UI payments and rebates did not take into account the volume of PPP funds already
allocated to each county.23 We find that these stimulus payments played a much more
important role in the employment recovery (as will be discussed later), but the inclusion
of these measures only marginally reduces the magnitude of the effect of the delay.
Finally, in column (4), we control for the total county-level amount of PPP funding
per employee in small establishments received through the preceding month. Adding
this last control does not change the size or the statistical significance of the coefficient
on the share delayed in any meaningful way.
The Decisive Role of Large Urban Counties
COVID-19 would a priori be expected to inflict greater and longer-lasting economic
damage to places with a larger population and greater density—major urban centers.
A comparison of urban, smaller urban (excluding the top 1 percent of counties by
population), and rural counties confirms this conjecture. Even when controlling for
only pre-pandemic characteristics, we find that the estimated persistent effect of the
share delayed is mostly driven by the largest urban counties (compare columns 5, 7,
and 9 of Table 3).24 The difference in the magnitude of the estimated coefficients across
subgroups of counties is almost mechanical: The massive difference in scale means that
the top 1 percent most populated urban counties dominate the size of the coefficients in
the pooled sample (as the LHS in these regressions is in levels). In addition, the share
delayed has no significant effect on employment for rural counties.
Once we also account for COVID-19 severity and pandemic-related transfer pay-
ments, the drag on the recovery stemming from the funding delay becomes only marginally
significant through July 2020 for urban counties considered together (shrinking in mag-
nitude by close to 40 percent) and disappears entirely for urban counties excluding
the top 1 percent (see column 6 versus 8). Most estimated coefficients on the share
delayed are positive for the rural counties (see column 10), likely pointing to a spurious
correlation.
For the urban sample, we also estimate an alternative specification that allows for
a differential effect of the share delayed for the top 1 percent most populous counties
versus the rest. Figure 1 plots the estimated coefficients. The lines trace out the
estimated effect of a one-standard-deviation increase in the shared delayed in each
subset, each normalized by its respective average employment level in January 2020 for
easier comparison. The figure clearly shows that the effect of the share delayed is driven
by the most populous counties, at a sizeable 2 percent of pre-pandemic employment in
April 2020 and shrinking gradually over time. For the other urban counties, the share
delayed plays no significant role in the employment recovery.
Population Size versus Population Density and the Influence of Remote
Work
23
Additionally, we find replacement rates to be conditionally uncorrelated with the share delayed
(see Table 2).
24
The top 1 percent most populous counties are Maricopa County, Arizona; Los Angeles County,
California; Orange County, California; Riverside County, California; San Diego County, California;
Miami-Dade County, Florida; Cook County, Illinois; Kings County, New York; Queens County, New
York; Dallas County, Texas; Harris County, Texas; and King County, Washington.
11
Our previous specifications zoomed in on the top 1 percent of counties by popula-
tion because we believe these counties were likely most relevant when accounting for
the effect of the share delayed on employment. Not only do these counties exhibit
significantly higher shares of PPP loans delayed (see Table 2), but they also dominate
the magnitude of the key coefficients in employment regressions that are specified in
levels. A primary candidate for the mechanism through which the pandemic had an
outsized lasting toll on employment in large urban counties is a more pronounced shift
to remote work in these localities. As shown in Althoff et al. (2021) and Ramani and
Bloom (2021), the decline in time spent in urban office centers substantially reduced
the demand for associated services and, in turn, employment in those local businesses
(for example, restaurants, hotels, and dry cleaning). To the extent that the shift to
remote work was a contributing factor to the slower recovery of some urban counties,
these studies would suggest that urban counties with high population (and/or high
population density) would be disproportionally affected.
Using Google mobility data, we confirm that population size and density are in-
dependently associated with a reduction in time spent at workplaces relative to pre-
pandemic averages (see Appendix Figure A.1). Urban counties in the top 1 percent in
terms of population density suffered the largest and most persistent declines in time
spent at workplaces (about 40 percent on average as of September 2021), while coun-
ties in the top 1 percent in terms of overall population also suffered significantly larger
declines than other urban counties (on average, 33 percent as of September 2021).25
If we remove counties in the top 1 percent by population density instead of those
in the top 1 percent by population size from our regressions, the explanatory power of
the share delayed in the employment regressions also disappears, and in this case, even
before we add other controls (see Appendix Table A.3). Thus, a correlation between
the share of loans delayed and remote-work patterns could account for the correlation
between employment and the share of loans delayed.26 Taken together, our findings so
far are consistent with the intuition discussed earlier that the pandemic likely inflicted
disproportionately worse and longer-lasting economic damage to business centers (in
areas that either have large populations or are densely populated).
Effects on the Most Adversely Affected Industries
Next we investigate whether PPP funding delay had a stronger and longer-lasting
impact on contact-intensive industries, such as leisure and hospitality, because of their
intrinsic vulnerability to an infectious disease. We compute the share of the volume of
lending delay specifically for arts, entertainment, and recreation (NAICS 71); accommo-
dation and food services (NAICS 72); and other services except public administration
(NAICS 81). The degree of delay was substantially higher for these industries, almost
25
The top 1 percent of counties by population density are San Francisco County, California; Suf-
folk County, Massachusetts; Hudson County, New Jersey; Bronx County, New York; Kings County,
New York; New York County, New York; Queens County, New York; Richmond County, New York;
Philadelphia County, Pennsylvania; Arlington County, Virginia; and Alexandria County, Virginia.
Only two counties (Queens County and Kings County) appear in both lists. The separate coefficients
for these two counties are not significantly different from those of the most densely populated counties.
26
Appendix Table A.2 shows that that share delayed is also greater in denser counties, conditional on
preexisting conditions and the public health situation early on in the pandemic. Since this difference
is not quite statistically significant, we focus on the top 1 percent of counties by population size in the
paper.
12
54 percent versus 47 percent for all industries. However, as shown in Table 4, the
industry-specific delay had no significant impact on the employment recovery of these
industries. All coefficients are small and statistically insignificant across all specifica-
tions and sub-samples (all, urban, smaller urban, and rural counties). This non-result is
inconsistent with the liquidity-based mechanism advocated in Doniger and Kay (2021),
which would imply that industry-specific funding delay should matter more for those
industries’ employment.27 It is more consistent with our proposed mechanism for the
correlation between the share delayed and the subsequent employment recovery.
Effects on the Number of Establishments
We explore the conjecture that the delay in funding had a lasting impact through
its effect on the number of small establishments that survived the pandemic (see Ap-
pendix Table A.11). In regressions involving the number of establishments of different
employment sizes (fewer than 5, fewer than 50, and fewer than 500), we find that the
share delayed had a significant but short-lived impact, until 2020:Q2, across all specifi-
cations. Similar to our results for the impact on employment, the effects on the number
of establishments are driven by the largest urban counties.
Comparing the Relative Effects and Contributions of Covariates
To illustrate the effects of other controls relative to the effect of share delayed on
the recovery of county employment, we plot their respective estimated coefficients in
Figures 2 and 3. We focus on tax rebates, industry-weighted UI replacement rates, and
the cumulative volume of PPP receipts per employee in small establishments. For ease of
comparison, we standardize these covariates (to a mean of zero and a standard deviation
of one) during the sample period, so that each coefficient measures the impact of a one-
standard-deviation increase. We also normalize the coefficient estimates by county-level
employment in January 2020 to more easily compare their magnitude across regressions.
Even though some of these covariates (particularly PPP receipt) are not necessarily
exogenous, it is nonetheless useful to compare their relative contribution to explaining
the heterogeneity in the employment recovery across counties.
Overall, larger shares of loans delayed and higher UI replacement rates curtail em-
ployment, while higher volumes of PPP receipts and rebates foster employment (see
Figure 2). The effects of stimulus programs are larger and more precisely estimated for
industries hit particularly hard by the pandemic (NAICS 71, 72 and 81; see top versus
bottom panels of Figure 2). The effects of the PPP also differ across urban and rural
counties, corroborating earlier findings that urban counties drive the pooled-sample es-
timates for both the share delayed and cumulative loan receipts (see Figure 3). In fact,
the share delayed does not seem to play much of a role in the employment recovery
of affected industries or rural counties, as the estimated coefficients are all insignifi-
cant and, for rural counties, sometimes even positive. While the share delayed matters
marginally for employment in urban counties, PPP receipts are nonetheless relatively
more important. Both UI benefits and rebates were associated with a more positive
27
Notably, if we use the share delayed for all PPP borrowers in a county, the coefficients on the
share delayed become significantly negative and persistent in regressions that pool all counties and in
those that include just rural counties (see Appendix Table A.10). These less-than-intuitive differences
in coefficients indicate that the share delay likely captures mechanisms other than the importance of
timely access to credit, as proposed by DK.
13
employment recovery in rural counties, acting as demand boosters. Conversely, rela-
tively more generous UI benefits during this period slowed employment growth in urban
counties. This finding suggests that in an urban setting, UI benefits discouraged labor
supply more than they boosted demand, as perhaps both labor supply and demand
for certain goods and services were more restrained in these areas by people’s fear of
COVID-19 infection.
In addition, we quantify the contribution of each covariate to explaining the variance
of employment (the dependant variable) using a Shorrocks-Shapley decomposition of
the R-squared.28 Following the structure of the regression results presented in Tables
3-4, we group the covariates into categories (preexisting conditions, COVID-19 controls,
and CARES Act transfers) to simplify the graphical exposition. We report the contri-
butions of the share delayed, cumulative PPP receipts, rebates to households, and UI
benefits replacement rates separately to assess the relative importance of the different
government interventions.
As Figure 4 shows, all the covariates combined account for close to 30 percent of
the variance in employment, with lag employment levels accounting for the rest.29 Pre-
existing conditions and COVID-19–related controls are relatively more important than
the various government stimulus payments for explaining the evolution of employment.
Among the different programs deployed during the pandemic, the county-specific UI
replacement rates are the most important in explaining the dynamics of employment.30
By comparison, the contribution of the share delayed is rather small. On average, from
March 2020 to September 2021, preexisting conditions, COVID-19 factors, UI replace-
ment rates, rebates, cumulative PPP funds, and share delayed explain 18.3, 3.91, 3.6,
1.01, 0.9, and 0.08 percent of the variation in county employment levels, respectively.
When the Shorrocks-Shapley decomposition is carried out separately for urban and ru-
ral counties, preexisting conditions are relatively more important for urban counties,
while COVID-19–related factors are particularly relevant for rural areas.
Analysis at the CBSA Level
QCEW employment data, as well as many other data sources we use, are compiled
at the county level, which is therefore the geographic level of our analysis. DK argue,
however, that many individuals residing in a metropolitan area travel out of county for
jobs, so core-based statistical areas (CBSAs) better correspond to local labor markets
and thus constitute a more natural level for analysis. We replicate our analysis at
the CBSA level, the data for which are compiled using crosswalks provided by the US
Department of Housing and Urban Development (HUD). The main message remains the
same (see Appendix Table A.13 and Figures A.2 and A.3). In particular, the estimated
effect of the share delayed is again driven by the most populous CBSAs. Curiously, at
the CBSA level, the share delayed has a less persistent effect on employment in every
28
This is implemented with the Stata command shapley2, written by Juarez (2012) and based on
Shorrocks (1982). We run period-by-period regressions after partialling out state-by-time fixed effects
from time-varying variables and state fixed effects from time-invariant controls.
29
Total R-squared in these regressions is high, about 0.99.
30
The uniform supplemental Pandemic Unemployment Compensation (PUC) payments (such as
$600 per week shortly after the COVID-19 outbreak) to all claimants during this period means that
the UI replacement rates vary across counties in inverse relationship to their pre-pandemic average
wage.
14
specification.
5.2 QCEW versus CPS Data on Employment
Doniger and Kay (2021) perform their analysis using individual-level data from the
Current Population Survey (CPS). They merge the share of loans delayed by locality
into the CPS data using geographical identifiers for CPS respondents. To preserve
anonymity, the county identifiers for close to 60 percent of CPS respondents are sup-
pressed in the public-use data, so that only 280 counties (less than 10 percent of all
counties) are covered during our sample period. Information on respondents’ CBSAs—
the level at which DK merge the share delayed into the CPS—is more available and
enables coverage of about 74 percent of respondents and 257 CBSAs (or about 14 per-
cent of all CBSAs). While the QCEW data, covering the population of employees, are
obviously more comprehensive than any survey data, the CPS data are more timely
and include information on self-employed individuals and business owners, who are not
captured in the QCEW. However, to evaluate the efficacy of the PPP program in pre-
serving (private) employment, in particular the valuable matches between employers
and employees, the focus should be mostly on employees at employer businesses.
To fully reconcile our findings with those by DK, it is also important to understand
how CPS employment compares with QCEW employment in the areas observed in both
sources, particularly during the early phase of the pandemic, and whether any observed
differences correlate with the share delayed.31 To this end, we compare employment
for the 280 counties and 257 CBSAs identified in the public CPS with employment
in the QCEW.32 Specifically, we regress the log difference between QCEW private
employment and CPS employment on monthly indicators (the base period is January
2018). As Figure 5 shows, CPS employment appears to be undercounted relative to
QCEW employment during 2020, but the degree of the undercounting is not correlated
with the share delayed (not shown in the figure). This confirms that DK’s estimates
are likely not biased by a possible pandemic-induced distortion to the CPS data.33
Nevertheless, the limited coverage of geographical areas in the CPS should be kept
in mind when comparing results that use all counties versus those that use the CPS
sample.
What happens in the CPS data if we control for the additional covariates in our
county-level regressions using QCEW data (that is, SBL per small establishment, UI
replacement rates, rebates per population, cumulative PPP receipts per employment
in small establishments, etc.)? Does the effect of the share delayed decline? The
short answer is yes, to a great extent, but more so in CBSA-based regressions than in
county-based ones (to be explained below). Note, however, our goal is not to reproduce
DK’s results exactly but to further quantify the influence of our earlier covariates in
individual-level regressions using CPS data, as DK does. We define the left hand side as
31
The CBSAs not covered in the CPS can also affect the coefficients estimated using QCEW data.
32
County-level QCEW data are aggregated to the CBSA level using the HUD crosswalk.
33
The CPS suspended in-person interviewing due to the COVID-19 outbreak. Response rates plum-
meted in March 2020 and remained low through the summer. In-person interviewing resumed nation-
ally in September 2020 and earlier in some areas. According to DK, correcting for the pandemic-induced
distortions strengthens coefficients on the shared delayed. See their Appendix B for their approach to
reweighting based on individuals’ industry and occupation rather than geography.
15
equal to 1 if an individual is employed and 0 otherwise (the opposite of DK’s definition)
to match the LHS of our county-level regressions. In addition, all our specifications
control for state-by-month fixed effects to account for multiple time-varying factors
(such as state-specific containment policies in a specific period), some of which DK
control for directly.34 Finally, our CPS regressions are weighted using the individual
weights included in the public-use CPS because DK weight their regressions. We do
not attempt to correct the CPS data for non-response due to the pandemic, since, as
noted above, this correction did not meaningfully affect DK’s results.
Using the individual-level CPS data, we estimate an analog of Equation (1) for two
samples of individuals: (1) those with information on their county of residence (county
sample) and (2) a larger sample of individuals with CBSA identifiers (CBSA sample, as
in DK’s analysis). The county sample forms our baseline because it enables the use of
more precise location-based controls (including the share delayed) at the county level,
while the CBSA sample with CBSA-level controls matches DK’s specifications and
thus facilitates comparisons to their results. The aggregate controls (such as the share
delayed) match the level of geography considered (that is, county-level controls for the
county sample and CBSA-level for the CBSA sample) in most specifications. We also
explore the difference in estimates for particular subsamples that may be differentially
affected by PPP funding delay (and for better comparison to our QCEW results):
employees of private firms (that is, excluding the self-employed and public employees)
and samples that exclude the top 1 percent most populous counties/CBSAs.
Table 5 presents our findings. Columns (1) and (3) report coefficients for the county
sample, including all individuals (column 1) versus private employees only (column 3),
while column (9) is the counterpart to column (1) for the CBSA sample. These three
regressions control for only individual and state-by-month fixed effects. The share de-
layed exhibits a persistently negative effect on employment, especially for the county
sample. Estimates reported in all the other columns in Table 5 further control for pre-
existing conditions, COVID-19–related factors, rebates, unemployment benefit replace-
ment rates, and cumulative PPP funds received, as in our regressions using county-level
employment. With these additional controls, the effects of the share delayed tend to de-
cline and become much less persistent (specifically, compare columns 1 and 2, columns
5 and 6, and columns 9 and 10).35 Furthermore, the estimated effects of the share
delayed shrink noticeably and lose much of their significance when we focus on private
employees (compare columns 2 and 3) and when we exclude the top 1 percent most
populous counties (compare columns 2 and 4). These findings confirm that a funding
delay was more detrimental to non-employer businesses than to employer businesses (in
the CPS sample), as highlighted by DK, and in large metropolitan areas, as we found
using the QCEW employment data.36
34
DK present some specifications with state-by-month fixed effects in addition to those in their
preferred specification, which include just individual and monthly fixed effects, two-digit NAICS, and
occupation exposure to COVID-19 interacted with monthly FE. In some other specifications, they also
include state-level non-pharmacological interventions and UI-related measures. All state-level controls
are captured by our state-by-month FEs, included in all specifications.
35
In columns (5) through (7), the county sample is paired with CBSA-level controls to quantify
the contribution of more precise local controls to the variation in the estimated coefficients between
county-based and CBSA-based regressions.
36
As already noted, it is not obvious which employer-employee matches would be preserved for
non-employer businesses.
16
As a final comparison, we reestimate Equation (1) using county-level QCEW em-
ployment data for just the sample of counties that appear in the CPS to further un-
derstand how the different geographies covered in the two data sources (CPS versus
QCEW) influence DK’s findings. Figure 6 presents “normalized” coefficients, which are
divided by average county population to make them more comparable to the individual-
level regression coefficients produced using the CPS data.37 We report estimates for
unweighted regressions (our baseline) and weighted regressions (DK’s setup). Only the
weighted regressions produce statistically significant coefficients, which are also larger
in magnitude, indicating that the effects of the share delayed are larger for more pop-
ulous areas. Reassuringly, the coefficients from the weighted county-level regressions
estimated using only the CPS counties are of similar magnitude to those obtained using
the individual-level CPS data. Most important, however, is the fact that the persistent
effect of the share delayed in the CPS sample, whether estimated using individual or
county data, does not carry through to the larger full sample of counties regardless of
weighting.38
The findings so far collectively indicate that the slower trajectory of employment
recovery as a result of PPP delay that DK find was driven by the most populous
counties or cities. More importantly, however, this effect is not necessarily causal in
that those larger metropolitan areas were more vulnerable to a highly infectious disease
such as COVID-19 in ways that are not fully captured by linear functions of observable
preexisting conditions.
6 Funding Receipt More Important than Timeli-
ness: Firm-Level Evidence
To further understand how the impact of PPP funding delay compares with that of
PPP funding receipt, we use firm-level activity indicators from SafeGraph (SG) data.
These data measure foot traffic derived from mobile devices utilizing GPS location
to track movements to and from points of interest (POIs).39 Each POI has its own
unique Placekey, which we use to identify PPP recipients and non-PPP recipients. For
each PPP recipient that we can identify in the SG data, we search for closely matched
competitors, defined as businesses operating in the same Census Block Group (CBG)
and the same six-digit NAICS industry that did not receive a PPP loan. When multiple
non-PPP businesses can be matched to a given PPP recipient, we choose the one
with the most similar number of visits before the pandemic (average weekly visits over
December 2019 to February 2020).40 With this procedure, we are able to match 165,660
37
Recall that the dependent variable in the CPS regressions is a binary variable equal to one if an
individual is employed and zero otherwise. The coefficients thus roughly have the scale of an effect on
the fraction employed.
38
See Appendix Table A.15 for the results using QCEW data on only the CPS counties and Table
A.16 for those on counties not present in the CPS sample. It is clear that the impact of funding delay
was driven by counties represented in the CPS sample.
39
SG does not cover the universe of POIs, but the coverage is extensive. We use visits to POIs as a
proxy for economic activity. See https://docs.safegraph.com/docs/places-data-evaluation for
coverage rates by industry.
40
Businesses might have not received PPP loans because they were not eligible or because they did
not need funding and chose not to apply. Even anecdotes of applicants ultimately being denied are
17
businesses from a list of 830,877 (20 percent of PPP-recipient POIs) with a Placekey.
Our matching criteria (based on geography and line of business) are rather strict because
our goal is to have a sample of businesses that would have faced similar conditions
over time in terms of demand, COVID-19–related factors, and imposed restrictions on
activity.41 Lacking information on pre-pandemic firm characteristics, we rely on firm
fixed effects to control for unobserved heterogeneity. For simplicity, our sample is also
restricted to firms that received just one PPP loan.
To study the effects of PPP receipt and timing, we compare the evolution in the
number of total visits to PPP establishments with the evolution in visits to non-PPP
establishments over time.42 SG data are available at a daily frequency, but we aggregate
visits to a monthly frequency. If a firm is not observed in a given month, its value is
“filled in” as zero visits. The LHS in our regressions is either Log Visits (defined as the
log of visits count plus one) or Zero Visits (an indicator equal to one when visits are
recorded as zero in the SG data, or when we fill in the data). Zero Visits is our proxy
for business closure, which might be temporary or permanent. With our matched-pair
sample, we estimate regressions of the form:
Yijt = αi + δjt + βt PPPi + ijt , (2)
where Yijt is a measure of visits to establishment i of pair j in month t, αi denotes firm
fixed effects, δjt denotes pair-by-time fixed effects, and PPPi is a dummy variable equal
to one if establishment i received PPP funding and zero otherwise. Standard errors are
clustered at the pair-ID level.
We also estimate regressions of the form:
−2
X T
X
l l
Yijt = αi + δjt + µl Dijt + µl Dijt + ijt , (3)
l=−T +1 l=0
l
where Dijt = I{t − Ki = l}, and Ki is the month when firm i0 s PPP loan is received.
These regressions are estimated using the Sun and Abraham (2021) estimator, which
is consistent under staggered heterogeneous treatment effects. Standard errors are
clustered at the pair-ID level.
The left panels of Figure 7 plot the estimated βt coefficients (PPP estimates over
calendar time) from regression (2), while the right panels plot the estimated µl coeffi-
cients (PPP estimates relative to funding approval dates, which are just two or three
days before actual receipt in most cases) from regression (3). The results clearly indi-
cate that businesses that received PPP funding saw significantly more visits (and fewer
closures) than their non-PPP peers, starting in the second half of 2020. The relative
difference between PPP and non-PPP matched firms increased over time, reaching close
rare. It is also possible that some businesses that we labeled as non-PPP recipients might have in fact
received PPP loans. So our estimates are likely a lower bound of the differences between PPP and
non-PPP recipients in terms of the measured outcomes discussed later.
41
CBGs are the smallest statistical area for which the US Census provides information, and NAICS
six-digit codes are narrow classifications of activity. We can obviously produce additional matched
pairs if we move to fewer-digit NAICS classifications or consider wider geographical areas, but the
results would be qualitatively similar.
42
The number of unique visitors, as opposed to visits, is also available in SG data. Results are very
similar when using visitors instead of visits and, for brevity, are not reported.
18
to a 15 percent difference by the end of the sample. About one-fifth of the Log Visits
effect is accounted for by excess closures (Zero Visits) by the end of the sample, and
they account for a larger fraction earlier on (for example, about one-fourth in December
2020).
To gauge the importance of the timing of PPP funding relative to funding receipt, we
restrict our estimations to matched pairs with PPP loans approved either early (April 14
through 16) or late (April 27 and 28), as defined by DK. Results are depicted in Figure 8
and show a small and insignificant average difference in foot traffic between early and
late PPP loan recipients (relative to their pairs). No significant difference in Zero Visits
is observed. The small and insignificant average difference in foot traffic between early
and late PPP loan recipients (relative to their pairs) seems difficult to reconcile with
DK’s report of share delayed having such a significant effect on employment. Apart
from the possibility of a large indirect effect on other firms’ employment, a possible
explanation for this discrepancy might be that the number of visits is only weakly
correlated with employment recovery at the firm level. Specifically, relative to early
PPP recipients, late PPP firms might have had to operate with reduced staff despite the
recovery in customer traffic. This hypothesis, however, cannot be tested using available
data, and it is not obvious why there would be such a systematic difference in the
relative recovery of visits versus employment between early and late PPP borrowers.43
Overall, our findings are consistent with those of Cole (2022), who uses administrative
payroll data for very small firms to argue that it is the receipt of PPP funding that was
important for employee retention and growth rather than the timing of the funding.44
7 Summary and Concluding Remarks
This study documents that the persistent effect of the 10-day delay in PPP funding
estimated by Doniger and Kay (2021) is entirely driven by the top 1 percent most
populous urban counties. These locales suffered higher rates of infections and deaths
in the first wave of the COVID-19 pandemic and thus imposed stricter containment
measures earlier on. The resulting worse disruption to commercial activity raised de-
mand for PPP loans while at the same time hampering their supply, resulting in a
larger backlog of PPP loan applications when funding was halted for 10 days in April
2020. Subsequently, the highly contagious nature of the virus and the shift toward
remote work in response made it more difficult for economic activity in these densely
populated places to recover, beyond what the linear functions of control variables of
pre-pandemic local conditions used in most previous studies could explain. This unac-
counted for heterogeneity, rather than the slightly earlier access to liquidity, explains
the correlation between the share of PPP loans delayed and the subsequent slower em-
ployment recovery. Consistent with this interpretation, employment recovery of the
most adversely affected, and arguably more liquidity-constrained industries, such as
43
In terms of the (log) levels of visits and employment at the county level, the correlation coefficient
is quite high (0.62) even with county and state-by-month fixed effects for all industries or industries
where foot traffic and employment are more directly related (that is, those with 71 or 72 as their
two-digit NAICS code).
44
In a companion paper, Gorbachev, Luengo-Prado, and Wang (2023) show that PPP receipt was
more important for businesses’ survival. They also find that PPP recipients’ credit risk profiles im-
proved relative to the profiles of their non-recipient peers.
19
leisure and hospitality, was not significantly associated with the share of loans delayed
to firms in those industries. Our findings suggest that the economic fundamentals of
these urban centers were more vulnerable to an infectious disease such as COVID-19
and that they would have needed extra government support if the goal was to restore
employment to pre-pandemic levels.
Moreover, consistent with the PPP’s stated goal of speeding up the employment
recovery, we find that receiving (more) PPP funds was indeed much more important
than receiving those funds a little earlier. Put differently, having access to liquidity
is crucial for business operations in general and during the pandemic, but having that
access a little earlier is much less so. In addition, two other important pandemic
support programs—expanded UI payments and rebates—explain a greater share of the
variation in local employment recoveries over time than does the volume of PPP funds
received or, especially, the share of PPP loans delayed. Preexisting local characteristics
(including population density, relationship banking, and 2019 banking conditions) and
the severity of the initial pandemic shock, combined with its heterogeneous impact
across localities due to heterogeneous pre-pandemic industrial compositions, were also
quantitatively important determinants of the recovery. In sum, our findings suggest
that in a pandemic, the availability of public transfer payments and credit assistance
matter more than a slight difference in the timing of funding receipt in shaping the
post-shock recovery.
It is beyond the scope of this paper to address the normative question of what should
be the optimal or most efficient policy response in terms of “bang for the buck” toward
the goal of preserving matches between firms and workers. In fact, a more difficult, but
just as important, question is to what extent it is even optimal to restore employment
to pre-pandemic levels or composition given the structural shift toward remote work
enabled by the latest technology. Many economists have pointed to policies common in
European countries (such as the Kurzarbeit program in Germany) that involved gov-
ernments sharing labor compensation costs with firms so that workers could stay on
the job. It has not been widely recognized, however, that the United States has compa-
rable programs. In particular, short-time compensation (STC) programs allow firms to
avoid layoffs by subsidizing a reduction in working hours through temporary prorated
payments. Workers are able to keep their fringe benefits while also receiving a partial
UI payment to supplement their lower wages. In the United States, 27 states include
STCs as part of their overall UI programs.45 The STC programs might not have helped
with the widely reported severe congestion-induced delays of the UI system early in the
pandemic. Nevertheless, going forward, more studies of the design, effects, and imple-
mentation of STCs and partial UI programs should be conducted. One goal would be
to understand why STC programs were used less during the pandemic than during the
Great Recession—perhaps the PPP, given its scale, made it unnecessary for employ-
ers to explore alternative options. Apart from implementing other worthy options for
government support programs, arriving at the correct understanding of the mechanism
through which the PPP facilitated the economic recovery, in particular whether specific
aspects of the program rules contributed to the ultimate goal of preserving or expanding
45
Houseman et al. (2017) explore the role of STCs during the Great Recession. Rodriguez, Segal,
and von Wachter (2023) find that STC programs generally reduce layoffs, enable workers to receive, on
average, higher UI benefits along with employer benefits, leading to higher earnings for workers over
the long haul.
20
employment, is important for refining the design of such public credit support programs
should the need arise again in the future.
References
Althoff, Lukas, Fabin Eckert, Sharat Ganapati, and Conor Walsh. 2021. “The Geog-
raphy of Remote Work.” National Bureau of Economic Research Working Paper
No. 29181.
Autor, David, David Cho, Leland D. Crane, Mita Goldar, Byron Lutz Byron,
Joshua K. Montes, William B. Peterman, David D. Ratner, Daniel Villar Val-
lenas, and Ahu Yildirmaz. 2022. “The $800 Billion Paycheck Protection Program:
Where Did the Money Go and Why Did It Go There?” National Bureau of Eco-
nomic Research Working Paper No. 29669.
Balyuk, Tetyana, Nagpurnanand R. Prabhala, and Manju Puri. 2020. “Indirect Costs
of Government Aid and Intermediary Supply Effects: Lessons from the Paycheck
Protection Program.” National Bureau of Economic Research Working Paper No.
28114.
Barrot, Jean-Noël, and Ramana Nanda. 2020. “The Employment Effects of Faster
Payment: Evidence from the Federal Quickpay Reform.” Journal of Finance 75
(6): 3139–3173.
Chetty, Raj, John N. Friedman, Michael Stepner, and The Opportunity Insights
Team. 2020. “The Economic Impacts of COVID-19: Evidence from a New Public
Database Built Using Private Sector Data.” National Bureau of Economic Research
Working Paper No. 27431.
Cole, Allison. 2022. “The Impact of the Paycheck Protection Program on (Really)
Small Businesses.” Available at SSRN: https://ssrn.com/abstract=3730268.
Doniger, Cynthia L., and Benjamin Kay. 2021. “Long Lived Employment Effects of
Delays in Emergency Financing for Small Businesses.” Finance and Economics
Discussion Series 2021-003, Board of Governors of the Federal Reserve System.
Faulkender, Michael W., Robert Jackman, and Stephen Miran. 2023. “The Job
Preservation Effects of Paycheck Protection Program Loans.” Available at SSRN:
https://ssrn.com/abstract=3767509 or http://dx.doi.org/10.2139/ssrn.3767509.
Ganong, Peter, Pascal J. Noel, and Joseph S. Vavra. 2020. “US Unemployment
Insurance Replacement Rates During the Pandemic.” National Bureau of Economic
Research Working Paper No. 27216.
Gorbachev, Olga, Maria Luengo-Prado, and J. Christina Wang. 2023. “PPP Receipt
as an Important Stimulant of Small Business Activity.” Working Paper, University
of Delaware.
Granja, João, Christos Makridis, Constantine Yannelis, and Eric Zwick. 2022. “Did the
Paycheck Protection Program Hit The Target?” Journal of Financial Economics
145 (3): 725–761.
Heckman, James J., and Edward Vytlacil. 2001. “Policy-Relevant Treatment Effects.”
American Economic Review 91 (2): 107–111.
21
Houseman, Susan N., Frank Bennici, Katharine Abraham, Susan Labin, Christopher J.
O’Leary, and Richard Sigman. 2017. “Demonstration and Evaluation of the Short-
Time Compensation Program in Iowa and Oregon: Final Report.” Technical re-
port, Washington, D.C., Department of Labor, Chief Evaluation Office.
Hubbard, R. Glenn, and Michael R. Strain. 2020. “Has the Paycheck Protection
Program Succeeded?” National Bureau of Economic Research Working Paper No.
28032.
Joaquim, Gustavo, and J. Christina Wang. 2022. “What Do 25 Million Records of
Small Businesses Say about the Effects of the PPP?” Federal Reserve Bank of
Boston Working Paper No. 22-23.
Juarez, Florian Chavez. 2012. SHAPLEY2: STATA Module to Compute Additive
Decomposition of Estimation Statistics by Regressors or Groups of Regressors.
Statistical Software Components, Boston College Department of Economics.
Kurmann, Andre, Etienne Lalé, and Lien Ta. 2022. “Measuring Small Business
Dynamics and Employment with Private-Sector Real-Time Data.” Institute of
Labor Economics (IZA) Discussion Paper No. 15515.
Li, Lei, and Philip E. Strahan. 2021. “Who Supplies PPP Loans (and Does It Mat-
ter)? Banks, Relationships, and the COVID Crisis.” Journal of Financial and
Quantitative Analysis 56 (7): 2411–2438.
Ramani, Arjun, and Nicholas Bloom. 2021. “The Donut Effect of Covid-19 on Cities.”
National Bureau of Economic Research Working Paper No. 28876.
Rodriguez, Gustavo, Kara Segal, and Till von Wachter. 2023. “Short-Time Compen-
sation in the U.S. and California from 2000 to 2022: A Descriptive Analysis of
Program Incidence and Worker Outcomes.” Working paper.
Shorrocks, A. F. 1982. “Inequality Decomposition by Factor Components.” Econo-
metrica 50 (1): 193–211.
Sun, Liyang, and Sarah Abraham. 2021. “Estimating Dynamic Treatment Effects in
Event Studies with Heterogeneous Treatment Effects.” Journal of Econometrics
225 (2): 175–199. Themed Issue: Treatment Effect 1.
22
Table 1: Summary Statistics for County-Level Preexisting Conditions
All Urban Smaller Urban Rural
Mean SD Mean SD Mean SD Mean SD
PPP Receipt
No. of Early PPP Loans (4/14–4/16/2020) 204 586 453 890 397 617 50 54
No. of Late PPP Loans (4/27–4/28/2020) 235 640 501 974 436 651 70 67
Volume of Early PPP Loans (4/14-16/2020) 28,272 106,663 67,570 164,810 57,511 118,993 3,867 5,296
Volume of Late PPP Loans (4/27-28/2020) 21,321 90,274 50,390 141,023 40,952 92,112 3,269 3,938
Early Jobs Saved 3,209 10,829 7,509 16,598 6,458 11,594 539 685
Late Jobs Saved 2,613 9,708 6,033 15,056 4,984 9,409 488 542
Cum. No. of PPP Loans 958 2,634 2,076 3,998 1,814 2,659 264 266
Cum. Volume of PPP Loans (Million 2016$) 132 467 309 720 261 479 22 26
Avg. Size of PPP Loans (1,000 2016$) 91 42 112 42 112 42 78 36
Share of PPP Loans Delayed (By Count) 0.58 0.12 0.55 0.10 0.55 0.10 0.59 0.12
Share of PPP Loans Delayed (By Vol.) 0.48 0.18 0.44 0.13 0.44 0.13 0.50 0.20
Share of Jobs Delayed 0.49 0.16 0.46 0.12 0.46 0.13 0.51 0.18
COVID-19 Impacts
Cum. COVID-19 Cases per Million Pop. 150 315 210 367 207 360 112 272
Cum. COVID-19 Deaths per Million Pop. 67 189 107 241 103 226 42 142
Covid-10 Stringency Index (Oxford University) 70 8 70 8 70 8 69 9
Share of days in lockdown (pre-4/17/2020) 0.50 0.11 0.51 0.10 0.51 0.10 0.50 0.11
Share of days in lockdown (4/17–4/30/2020) 0.99 0.05 1.00 0.02 1.00 0.02 0.99 0.06
Share of Emp. in Essential Industries 0.88 0.02 0.87 0.02 0.87 0.02 0.88 0.02
Share of Emp. in Impacted Industries 0.32 0.08 0.32 0.07 0.32 0.07 0.31 0.09
Share of Wages in Impacted Industries 0.19 0.07 0.19 0.06 0.19 0.06 0.19 0.08
Share of Emp. in NAICS 71, 72 & 81 0.16 0.06 0.17 0.05 0.17 0.05 0.16 0.06
Share of Wages in NAICS 71, 72 & 81 0.08 0.05 0.08 0.04 0.08 0.04 0.08 0.05
UI Benefits Replacement Rate (Industry-Wtd.) 1.39 0.19 1.30 0.18 1.30 0.18 1.44 0.17
Preexisting Conditions
Rural County Dummy 0.62 0.49
Total Residential Population 108,885 342,828 245,668 525,144 207,098 302,323 23,937 22,533
Commuter to Residential Population Ratio 1.15 0.11 1.13 0.13 1.13 0.12 1.17 0.10
Median Family Income 67,238 16,223 76,058 18,370 76,011 18,391 61,761 11,784
Community Bank Share of Branches 0.45 0.32 0.33 0.27 0.33 0.27 0.52 0.33
Community Bank Share of Deposits 0.43 0.34 0.30 0.28 0.30 0.28 0.52 0.35
Big4 Bank Share of Deposits 0.03 0.07 0.03 0.06 0.03 0.06 0.02 0.08
No. of Branches 28 71 59 108 52 72 9 8
Bank Branch Density (Population per Branch) 3,208 1,927 4,109 1,980 4,095 1,982 2,649 1,665
No. of Small Business Loans in 2019 2,429 10,540 5,784 16,479 4,578 8,619 345 428
Vol. of SBL in 2019 78,721 301,083 186,470 466,393 152,878 262,618 11,805 15,571
Avg SBL Loan in thousands of 2016$ 33 13 34 11 34 11 32 14
SBL Vol. per Small Estabs. (< 500 Emp.) (CBP 2019Q1) 22 11 27 9 27 9 19 11
Private Employment
Private Emp., 2020 34,504 125,274 80,744 193,528 67,430 123,496 5,786 6,301
Private Employment in NAICS 71, 72, and 81 3,841 14,334 9,019 22,179 7,515 13,790 625 872
No. of Private Estabs., 2020 3,140 13,016 7,229 20,364 5,884 10,289 601 639
Share of Employment in Estabs (under 500), QWI 2019 0.54 0.14 0.48 0.11 0.48 0.11 0.57 0.14
Share of Estabs (under 5), CBP 2019Q1 0.56 0.07 0.54 0.06 0.54 0.06 0.57 0.07
Share of Estabs (under 50), CBP 2019Q1 0.96 0.02 0.95 0.02 0.95 0.02 0.96 0.02
Share of Estabs (under 500), CBP 2019Q1 0.99 0.01 1.00 0.01 1.00 0.01 0.99 0.02
Notes: “Smaller Urban” refers to urban counties excluding those in the top 1 percent by population. The values for each variable pertain
to April 2020 unless specified otherwise.
Source: Multiple data sources described in Section 3.2.
23
Table 2: Determinants of Share of PPP Loan Volume Delayed, April 16–26, 2020
All Urban Smaller Rural
∗∗∗ ∗∗∗
Cum. COVID-19 Cases per bil. up to 4/15/2020 0.017 0.064 0.073 –0.006
(0.020) (0.022) (0.022) (0.036)
Cum. COVID-19 Deaths per bil. up to 4/15/2020 0.066 0.030 0.000 0.080
(0.051) (0.065) (0.068) (0.097)
Share of days in lockdown (pre-4/17/2020) 0.027 –0.024 –0.025 –0.036
(0.053) (0.066) (0.069) (0.068)
Share of days in lockdown (4/17–4/30/2020) 0.124 –0.043 –0.045 0.289∗∗∗
(0.144) (0.136) (0.136) (0.072)
Share of Emp. in Essential Industries –0.237 –0.407 –0.440 0.046
(0.216) (0.345) (0.347) (0.258)
Share of Emp. in Impacted Industries –0.031 –0.115 –0.116 0.080
(0.060) (0.116) (0.116) (0.082)
Rural County Dummy 0.005
(0.010)
Most Populous County (Top 1%) 0.068∗∗∗ 0.066∗∗∗
(0.022) (0.024)
Ln Residential Population –0.023∗∗∗ –0.021∗∗∗ –0.021∗∗∗ –0.030∗∗∗
(0.005) (0.007) (0.007) (0.010)
Commuter to Residential Population Ratio –0.027 –0.018 –0.015 –0.030
(0.026) (0.029) (0.029) (0.062)
Ln Median Family Income 0.016 0.014 0.014 0.017
(0.027) (0.035) (0.036) (0.053)
Community Bank Share of Deposits 0.003 0.039∗∗ 0.039∗∗ –0.008
(0.014) (0.015) (0.015) (0.019)
Big4 Bank Share of Deposits 0.072 0.116 0.112 0.052
(0.071) (0.094) (0.092) (0.074)
Ln Bank Branch Density –0.005 –0.006 –0.007 –0.001
(0.009) (0.013) (0.013) (0.015)
SBL Vol. per Small Estabs. (< 500 Emp.) (CBP 2019Q1) –0.002∗∗∗ –0.002∗∗∗ –0.002∗∗∗ –0.002∗∗∗
(0.000) (0.001) (0.001) (0.000)
Proportion of Small Employment in 2020Q1 to 2019Q1, QWI 0.014 –0.160 –0.159 0.092
(0.067) (0.097) (0.097) (0.079)
UI Benefits Replacement Rate (Industry-Wtd.) –0.028 –0.012 –0.016 –0.033
(0.041) (0.058) (0.059) (0.056)
Constant 0.746 1.219∗ 1.265∗ 0.307
(0.497) (0.680) (0.683) (0.717)
Adjusted R-squared 0.13 0.17 0.17 0.10
Observations 2644 1108 1096 1536
State FE Yes Yes Yes Yes
Notes: “Smaller” refers to urban counties excluding those in the top 1 percent by population.
Source: Multiple data sources described in Section 3.2.
24
Table 3: Effects of Share of PPP Loans Delayed on QCEW County Private Employment
All Counties Urban Smaller Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Jan 2020 × Share Delayed 53 175 179 179 270∗ 451∗ 100 201 –20∗ –18
(35) (119) (120) (120) (153) (270) (106) (190) (12) (12)
Feb 2020 × Share Delayed 143∗∗∗ 271∗∗ 283∗∗ 283∗∗ 325∗ 546∗∗ 239∗∗ 370∗ –18∗∗ –16∗
(51) (119) (121) (121) (169) (258) (107) (204) (9) (10)
Apr 2020 × Share Delayed –731∗∗ –553∗ –489∗ –492∗ –3,037∗∗ –2,063 –1,688∗ –1,068 –36 –17
(329) (295) (297) (297) (1,428) (1,311) (931) (887) (57) (57)
May 2020 × Share Delayed –891∗∗∗ –679∗∗ –614∗∗ –609∗∗ –3,400∗∗ –2,182∗ –2,446∗∗ –1,616 –2 20
(336) (285) (284) (282) (1,462) (1,233) (1,213) (1,011) (48) (49)
Jun 2020 × Share Delayed –974∗∗∗ –741∗∗∗ –675∗∗ –667∗∗ –3,739∗∗ –2,260∗ –2,873∗ –1,792 31 50
(366) (280) (273) (273) (1,675) (1,253) (1,564) (1,112) (40) (40)
Jul 2020 × Share Delayed –1,023∗∗∗ –834∗∗∗ –766∗∗∗ –754∗∗∗ –3,383∗∗ –1,944∗ –2,672∗ –1,593 40 57
(354) (275) (268) (267) (1,607) (1,175) (1,596) (1,086) (38) (39)
Aug 2020 × Share Delayed –942∗∗∗ –766∗∗∗ –730∗∗∗ –715∗∗∗ –2,921∗ –1,569 –2,400 –1,356 34 54
(342) (267) (263) (262) (1,537) (1,131) (1,574) (1,078) (36) (36)
Sept 2020 × Share Delayed –818∗∗ –670∗∗∗ –627∗∗ –613∗∗ –2,449∗ –1,210 –2,146 –1,173 30 50
(326) (252) (247) (246) (1,471) (1,059) (1,536) (1,048) (35) (35)
Oct 2020 × Share Delayed –603∗ –450∗ –403∗ –391∗ –1,887 –744 –1,755 –819 34 56
(312) (230) (225) (225) (1,419) (1,003) (1,493) (1,020) (35) (35)
Nov 2020 × Share Delayed –492 –361 –313 –303 –1,827 –668 –1,674 –699 53 75∗∗
(326) (239) (229) (230) (1,477) (1,024) (1,540) (1,045) (34) (35)
Dec 2020 × Share Delayed –464 –341 –298 –281 –1,485 –375 –1,461 –546 56∗ 74∗∗
(331) (249) (241) (241) (1,481) (1,068) (1,548) (1,071) (33) (34)
Jan 2021 × Share Delayed –604∗ –457∗ –413 –392 –1,938 –721 –1,772 –844 58∗ 75∗∗
(359) (276) (269) (268) (1,584) (1,164) (1,662) (1,157) (34) (35)
Feb 2021 × Share Delayed –427 –268 –224 –241 –1,875 –841 –1,676 –878 61∗ 79∗∗
(357) (255) (246) (245) (1,593) (1,072) (1,637) (1,142) (35) (35)
Mar 2021 × Share Delayed –449 –290 –246 –274 –1,809 –766 –1,719 –904 53 68∗
(344) (243) (235) (234) (1,556) (1,038) (1,604) (1,104) (34) (35)
Apr 2021 × Share Delayed –374 –205 –155 –199 –1,264 –282 –1,333 –548 36 43
(306) (219) (215) (214) (1,413) (963) (1,506) (1,032) (40) (41)
May 2021 × Share Delayed –329 –150 –103 –148 –1,194 –217 –1,193 –424 39 44
(297) (211) (209) (209) (1,388) (927) (1,449) (987) (39) (41)
Jun 2021 × Share Delayed –336 –178 –129 –183 –880 19 –860 –143 32 29
(273) (207) (206) (207) (1,256) (893) (1,300) (905) (41) (43)
Jul 2021 × Share Delayed –391 –190 –149 –194 –1,479 –441 –1,134 –349 15 11
(278) (208) (209) (209) (1,296) (900) (1,315) (934) (46) (47)
Aug 2021 × Share Delayed –359 –153 –110 –146 –1,617 –507 –1,182 –360 14 12
(282) (206) (208) (209) (1,315) (907) (1,291) (914) (43) (44)
Sept 2021 × Share Delayed –167 17 43 5 –932 –19 –702 –79 17 18
(248) (188) (188) (189) (1,141) (809) (1,099) (831) (38) (39)
Average Private Employment 38,956 91,195 76,404 6,489
St. Dev. of Private Employment 139,524 215,187 138,117 7,193
Within R-squared 0.86 0.86 0.86 0.86 0.85 0.86 0.78 0.79 0.48 0.51
Observations 61,173 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes Yes No Yes No Yes No Yes
CARES Act Controls No No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top 1 per-
cent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preexisting
Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation,
population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and 2019 small-business-loan vol-
ume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths per population, share of days in April 2020 in
early lockdown, share of employment in essential industries, and share of employment in most impacted industries; CARES Act Controls:
industry-employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
25
Table 4: Effects of Share of PPP Loans Delayed on QCEW County Private Employment
in NAICS 71, 72, and 81
All Counties Urban Smaller Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Jan 2020 × Share Delayed 25 24 24 24 159∗∗ 147 115 101 3 6∗
(17) (31) (31) (30) (80) (127) (71) (128) (3) (4)
Feb 2020 × Share Delayed 27∗ 28 28 28 153∗∗ 145 115∗ 98 1 4
(14) (30) (30) (30) (67) (123) (62) (125) (2) (3)
Apr 2020 × Share Delayed –4 1 –25 –25 –207 –180 –63 –30 15 13
(87) (76) (80) (80) (433) (348) (311) (276) (16) (16)
May 2020 × Share Delayed –16 –19 –40 –45 –164 –8 –114 69 14 9
(113) (106) (107) (106) (527) (438) (418) (363) (17) (16)
Jun 2020 × Share Delayed –117 –102 –129 –138 –692 –456 –577 –304 18 15
(150) (131) (133) (132) (770) (602) (635) (506) (16) (15)
Jul 2020 × Share Delayed –160 –136 –153 –170 –696 –401 –656 –359 8 7
(149) (129) (131) (130) (764) (590) (671) (515) (13) (13)
Aug 2020 × Share Delayed –150 –121 –127 –150 –603 –358 –620 –383 6 6
(142) (123) (124) (123) (718) (574) (645) (511) (13) (12)
Sept 2020 × Share Delayed –116 –96 –102 –122 –512 –299 –522 –303 6 7
(131) (113) (113) (113) (661) (521) (588) (464) (12) (11)
Oct 2020 × Share Delayed –91 –77 –83 –100 –442 –291 –452 –282 10 10
(115) (99) (99) (99) (579) (462) (512) (412) (11) (10)
Nov 2020 × Share Delayed –91 –87 –97 –113 –473 –364 –455 –311 6 5
(113) (102) (103) (103) (578) (463) (512) (408) (11) (10)
Dec 2020 × Share Delayed –38 –37 –49 –76 –68 20 –173 –87 2 0
(111) (106) (106) (106) (552) (460) (505) (411) (9) (10)
Jan 2021 × Share Delayed –50 –57 –70 –105 –193 –35 –368 –243 6 4
(133) (126) (127) (127) (650) (536) (598) (474) (11) (11)
Feb 2021 × Share Delayed –91 –104 –120 –146 –469 –358 –546 –420 9 8
(133) (127) (129) (129) (677) (535) (593) (462) (11) (11)
Mar 2021 × Share Delayed –96 –109 –127 –148 –518 –440 –574 –474 11 11
(133) (127) (128) (127) (682) (547) (593) (468) (11) (10)
Apr 2021 × Share Delayed –97 –99 –119 –136 –477 –380 –508 –370 –6 –0
(126) (118) (119) (118) (653) (516) (550) (435) (17) (18)
May 2021 × Share Delayed –98 –91 –114 –129 –487 –385 –486 –356 0 1
(121) (112) (113) (112) (631) (496) (529) (413) (13) (14)
Jun 2021 × Share Delayed –82 –70 –90 –104 –334 –225 –307 –186 3 4
(108) (101) (101) (100) (559) (448) (467) (376) (12) (13)
Jul 2021 × Share Delayed –54 –40 –57 –65 –327 –186 –260 –86 16 16
(101) (95) (97) (96) (530) (421) (433) (346) (11) (11)
Aug 2021 × Share Delayed –53 –41 –58 –64 –319 –128 –236 –33 14 13
(95) (91) (93) (92) (497) (402) (396) (329) (11) (11)
Sept 2021 × Share Delayed –19 –10 –25 –29 –190 –207 –127 –133 16 16
(74) (71) (71) (71) (385) (332) (288) (259) (10) (10)
Average Private Employment 5,666 13,272 11,221 939
St. Dev. of Private Employment 20,173 31,065 20,276 1,451
Within R-squared 0.89 0.89 0.89 0.89 0.89 0.90 0.85 0.86 0.75 0.76
Observations 61,173 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes Yes No Yes No Yes No Yes
CARES Act Controls No No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top
1 percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed specific to NAICS 71, 72, and
81 industries. Preexisting Conditions Controls: median family income, commuter-to-residential-population ratio, indicators
for 2013 NCHS urban-rural designation, population, community-bank share of deposits, largest four banks’ share of deposits, bank
branch density, and 2019 small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases
and deaths per population, share of days in April 2020 in early lockdown, share of employment in essential industries, and share of
employment in most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits replacement
rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
26
Table 5: Effects of Share of PPP Loans Delayed on Employment, CPS Data
(1) (2) (3) (4) (5) (6) (7) (9) (10)
Respondents with County Resp. with CBSA
Employment All All Private All All All All All All
Employees
Jan20 × Shared Delayed –0.039 –0.070∗∗ –0.064∗ –0.055∗ –0.033 –0.049 –0.043 0.010 –0.004
(0.033) (0.030) (0.036) (0.031) (0.030) (0.031) (0.033) (0.023) (0.027)
Feb20 × Shared Delayed 0.002 –0.024 –0.025 –0.003 0.013 0.004 0.013 0.033∗ 0.033
(0.034) (0.033) (0.034) (0.033) (0.036) (0.040) (0.040) (0.018) (0.024)
Apr20 × Shared Delayed –0.091∗∗ –0.062 –0.059 –0.028 –0.048 –0.029 –0.003 –0.013 0.018
(0.038) (0.041) (0.038) (0.045) (0.043) (0.045) (0.047) (0.036) (0.045)
May20 × Shared Delayed –0.172∗∗∗ –0.131∗∗∗ –0.091∗∗ –0.115∗∗ –0.116∗∗∗ –0.069 –0.060 –0.108∗∗∗ –0.037
(0.042) (0.045) (0.042) (0.048) (0.044) (0.044) (0.046) (0.032) (0.035)
Jun20 × Shared Delayed –0.086∗ –0.047 0.008 0.012 –0.069 –0.027 0.010 –0.021 0.021
(0.050) (0.051) (0.054) (0.054) (0.043) (0.052) (0.056) (0.030) (0.043)
Jul20 × Shared Delayed –0.133∗∗ –0.116∗∗ –0.072 –0.063 –0.099∗∗ –0.070 –0.037 –0.032 –0.003
(0.056) (0.050) (0.057) (0.052) (0.041) (0.047) (0.054) (0.028) (0.040)
Aug20 × Shared Delayed –0.122∗∗∗ –0.095∗ –0.068 –0.058 –0.109∗∗ –0.067 –0.050 –0.061∗∗ –0.032
(0.046) (0.049) (0.057) (0.051) (0.044) (0.049) (0.053) (0.029) (0.039)
Sep20 × Shared Delayed –0.121∗∗∗ –0.090∗∗ –0.028 –0.058 –0.101∗∗ –0.057 –0.038 –0.068∗∗ 0.027
(0.039) (0.044) (0.055) (0.049) (0.048) (0.054) (0.055) (0.033) (0.041)
Oct20 × Shared Delayed –0.086∗∗ –0.061 –0.017 –0.037 –0.072 –0.031 –0.019 –0.031 0.037
(0.036) (0.041) (0.052) (0.045) (0.047) (0.049) (0.052) (0.030) (0.039)
Nov20 × Shared Delayed –0.057 –0.050 –0.030 –0.023 –0.021 0.007 0.019 0.018 0.103∗∗
(0.040) (0.044) (0.051) (0.047) (0.050) (0.054) (0.055) (0.034) (0.042)
Dec20 × Shared Delayed –0.082∗∗ –0.081∗ –0.064 –0.059 –0.091∗∗ –0.063 –0.054 –0.018 0.030
(0.041) (0.044) (0.049) (0.047) (0.043) (0.048) (0.050) (0.031) (0.040)
Jan21 × Shared Delayed –0.081∗ –0.078 –0.072 –0.058 –0.113∗∗ –0.090∗ –0.071 –0.025 0.016
(0.044) (0.047) (0.045) (0.053) (0.051) (0.054) (0.054) (0.034) (0.042)
Feb21 × Shared Delayed –0.093∗∗ –0.071 –0.057 –0.068 –0.060 –0.028 –0.020 –0.012 0.017
(0.042) (0.047) (0.047) (0.051) (0.050) (0.054) (0.054) (0.035) (0.040)
Mar21 × Shared Delayed –0.054 –0.053 –0.054 –0.058 –0.065 –0.045 –0.041 0.008 0.002
(0.044) (0.047) (0.050) (0.049) (0.054) (0.052) (0.053) (0.036) (0.040)
Apr21 × Shared Delayed –0.069∗ –0.042 –0.059 –0.038 –0.069∗ –0.039 –0.031 –0.018 0.012
(0.037) (0.041) (0.053) (0.044) (0.041) (0.045) (0.045) (0.034) (0.044)
May21 × Shared Delayed –0.068∗ –0.045 –0.039 –0.048 –0.057 –0.032 –0.023 –0.005 0.006
(0.037) (0.043) (0.053) (0.048) (0.044) (0.051) (0.055) (0.032) (0.040)
Jun21 × Shared Delayed –0.070 –0.058 –0.047 –0.053 –0.086∗ –0.078 –0.060 –0.031 –0.013
(0.044) (0.047) (0.054) (0.052) (0.048) (0.056) (0.060) (0.033) (0.045)
Jul21 × Shared Delayed –0.079 –0.078 –0.012 –0.069 –0.080 –0.063 –0.057 –0.026 –0.030
(0.052) (0.056) (0.062) (0.061) (0.074) (0.077) (0.079) (0.042) (0.044)
Aug21 × Shared Delayed –0.068∗ –0.059 –0.004 –0.068 –0.073 –0.062 –0.060 –0.017 –0.001
(0.039) (0.046) (0.058) (0.051) (0.048) (0.057) (0.060) (0.035) (0.043)
Sep21 × Shared Delayed –0.053 –0.030 0.003 –0.022 –0.021 –0.005 0.005 –0.008 –0.009
(0.040) (0.046) (0.059) (0.052) (0.050) (0.056) (0.059) (0.032) (0.040)
Adj. R-squared 0.82 0.82 0.80 0.82 0.82 0.82 0.82 0.82 0.82
Within R-squared 0.0001 0.0008 0.0008 0.0009 0.0000 0.0008 0.0009 0.0000 0.0005
Observations 1,612,958 1,612,958 1,612,958 1,352,878 1,612,958 1,612,958 1,352,878 2,934,532 2,934,532
No. Clusters 280 280 280 272 157 157 155 257 257
Avg. Emp 0.60 0.60 0.45 0.60 0.60 0.60 0.60 0.61 0.61
Agg. Controls Level County County County County CBSA CBSA CBSA CBSA CBSA
Additional Agg. Controls NO YES YES YES NO YES YES NO YES
Excludes Top 1% Top 1%
Population Population
Notes: The left-hand side in these regressions is a binary variable for whether the individual reports being employed (any kind of
employment, including self-employed and public sector workers) or being an employee in the private sector (column 3). All regressions
include individual fixed effects and state-by-month fixed effects. Standard errors are clustered at the geography level corresponding
to the aggregate controls as indicated in the table (that is, either county or CBSA level). Locality-level controls, when included, are
the same as those included in our baseline county-level regressions in Table 3. Top 1% refers to counties in the top 1 percent by
population, which are excluded in some regressions as noted.
Source: Multiple data sources described in Section 3.2.
27
Figure 1: Effects of Share of PPP Loans Delayed on QCEW County Private Employment:
Urban Counties by Population
.01
0
-.01
-.02
-.03
-.04
Jan20 May20 Sep20 Jan21 May21 Sep21
Other Urban Top 1% Urban
Notes: Estimated effects of a change of one standard deviation in the share delayed, normalized by
the average employment level in January 2020 of the corresponding sample of counties.
Source: Multiple data sources described in Section 3.2.
28
Figure 2: Effects of Selected Controls on QCEW County Private Employment: All Counties
All Counties, All Industries
.05
.02
.015
0
.01
.005 -.05
0
-.005
-.1
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
All Counties, NAICS 71,72,81
.1 .2
0
.05
-.2
0
-.4
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
Notes: Estimated effects of a change of one standard deviation in a given control, normalized by the
average corresponding county-level employment in January 2020 (all industries in the top panel and
employment in NAICS 71, 72, and 81 in the bottom panel).
Source: Multiple data sources described in Section 3.2.
29
Figure 3: Effects of Selected Controls on QCEW County Private Employment: Urban
vs. Rural Counties
Urban Counties, All Industries
.03 .1
.05
.02
0
.01
-.05
0
-.1
-.01 -.15
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
Rural Counties, All Industries
.03 .1
.05
.02
0
.01
-.05
0
-.1
-.01 -.15
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
Notes: Estimated effects, based on separate regressions for urban versus rural counties, of a change
of one standard deviation in a given control, normalized by the average urban (top panel) or rural
(bottom panel) county level of employment in January 2020. The top 1 percent most populous counties
are included in the urban sample.
Source: Multiple data sources described in Section 3.2.
30
Figure 4: R-Squared Decomposition, QCEW County Private Employment Regressions
All Counties All Counties
100 30
% R sq. Explained (cumulated) % R sq. Explained (cumulated)
80
20
60
40
10
20
0 0
Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21 Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21
Cumulative PPP Share Delayed UI repl. Rebates COVID-19 Preexisting Conditions Lag Empl.
Urban Rural
30 10
% R sq. Explained (cumulated) % R sq. Explained (cumulated)
8
20 6
4
10
2
0 0
Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21 Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21
Cumulative PPP Share Delayed UI repl. Rebates COVID-19 Preexisting Conditions
Notes: Contributions of different variables to explaining the variance in employment over time. The effect
of lagged employment is omitted in some graphs to more easily depict the contribution of other variables.
The top 1 percent most populous counties are included in the urban sample.
Source: Multiple data sources described in Section 3.2.
31
Figure 5: Difference in Private Employment over Time: QCEW versus CPS Data
CPS Counties CPS CBSAs
.15
.1
.1
log difference log difference
.05 .05
0
0
-.05 -.05
2018m1 2018m7 2019m1 2019m7 2020m1 2020m7 2021m1 2021m7 2018m1 2018m7 2019m1 2019m7 2020m1 2020m7 2021m1 2021m7
Notes: The figure depicts dummies from regressions of the log difference between county- or CBSA-level
QCEW employment and CPS employment during the period depicted. The base period is January
2018. County or CBSA fixed effects are included in the regressions, and standard errors are adjusted for
heteroskedasticity and autocorrelation. The county-level comparisons include 280 counties identified
in the CPS, while the CBSA comparison covers 257 CBSAs.
32
Figure 6: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment: All Counties vs. Counties Present in the CPS
No Weights Population Weights
.2
.02
.1
0
-.02
0
-.1
-.04
-.2
-.06
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
All Urban CPS
Notes: Regressions were run separately for each sample (all, urban including the top 1 percent most
populous counties, and only counties that are identified in the CPS). Estimated coefficients were
divided by average county population in each sample in January 2020. The regressions in the right
panel are weighted by county population, while the regressions in the left panel are unweighted.
33
Figure 7: Effect of PPP Loans over Time: CBG-NAICS6 Firm Pairs in Safegraph
Log Visits over Calendar Time Log Visits over Event Time
.15 .15
April 2020 January 2021
average effect average effect
.05 .1 .05 .1
0 0
Dec18 Jun19 Dec19 Jun20 Dec20 Jun21 Dec21 -16 -12 -8 -4 0 4 8 12 16 20
month months relative to loan approval
Zero Visits over Calendar Time Zero Visits over Event Time
.01 .01
0 0
average effect average effect
-.02 -.01 -.02 -.01
-.03 -.03
April 2020 January 2021
-.04 -.04
Dec18 Jun19 Dec19 Jun20 Dec20 Jun21 Dec21 -16 -12 -8 -4 0 4 8 12 16 20
month months relative to loan approval
Notes: The left panels plot the βt coefficients from regressions of the form Yijt = αi + δjt + βt PPPi + ijt ,
where Yijt is a measure of visits/visitors to establishment i of pair j in month t, αi denote firm fixed
effects, δjt are pair-by-time fixed effects, and PPPi is a dummy variable equal to one if establishment i
received PPP funding and P−2zero otherwise. PThe right panels plot the µl coefficients of regressions of the
l T l l
form Yijt = αi + δjt + l=−T +1 µl Dijt + l=0 µl Dijt + ijt , where Dijt = I{t − Ki = l}, and Ki is the
0
month when firm i s first PPP loan is received. The shaded areas represent 95 percent confidence intervals.
The regressions are estimated using the estimator in Sun and Abraham (2021), which is consistent under
heterogeneous treatment effects. Standard errors are clustered at the pair-ID level. The sample is constructed
matching each 2020 PPP (recipient) firm to a non-PPP firm in the same census block group (CBG) and
NAICS-6 sector. If multiple non-PPP recipients were initially matched to a PPP recipient, we kept the
match with the closest average number of visits to the PPP recipient during the months of January and
February 2020 (the omitted time dummy in these regressions). The sample is also restricted to firms that
received just one loan. The data are filled in in the sense that once a firm disappeared from the data, we
assigned them zero visits. Log visits/visitors are defined as the log of visits/visitors counts plus one. Zero
visits is a dummy equal to one when visits are recorded as zero in Safegraph or when we fill in the data, and
it is our proxy for closure, which might be temporary or permanent. Safegraph data and PPP data from the
Small Business Administration were initially matched via Placekeys.
34
Figure 8: Effect of PPP Loans on Visits over Time: Early versus Late PPP Recipients
in Safegraph
Log Visits over Calendar Time Log Visits over Event Time
.2 .2
average effect average effect
April 2020 January 2021
.05 .1 .15 .05 .1 .15
0 0
-.05 -.05
Dec18 Jun19 Dec19 Jun20 Dec20 Jun21 Dec21 -14 -10 -6 -2 2 6 10 14 18 22
month months since loan approved
Zero Visits over Calendar Time Zero Visits over Event Time
.01 .01
average effect average effect
-.04 -.03 -.02 -.01 0 -.04 -.03 -.02 -.01 0
April 2020 January 2021
Dec18 Jun19 Dec19 Jun20 Dec20 Jun21 Dec21 -14 -10 -6 -2 2 6 10 14 18 22
month months since loan approved
early DK PPP late DK PPP
Notes: In these graphs, we compare PPP firms to their non-PPP pairs allowing for a differential effect based
on the date of their loan approval. These regressions include only firms that received loans on April 14, 15,
16, 27, and 28, 2020 (around the Doniger and Kay (2021) discontinuity). Early firms received loans between
April 14 and April 16, and late firms between April 27 and April 28. See notes to Figure 7 for details on the
estimation. The shaded areas represent 95 percent confidence intervals.
35
Online Appendix – Not for Publication
A Data: Additional Details
Pre-pandemic Local Conditions
The Quarterly Census of Employment and Wages (QCEW) reports employment
data at a monthly frequency and the total number of establishments and payroll (wages)
at a quarterly frequency. According to the QCEW, right before the COVID-19 outbreak
(that is, 2020:Q1), there were, on average, about 33,000 employees working in 3,000
establishments (Table 1) in an average county.
On average, each county had about 27.5 bank branches serving about 104,500 indi-
viduals, whose median family income was $66,500 (Table 1). In 2019, an average county
received 2,317 small business loans (SBL) with a total volume of $75.5 million (2016$)
and an average amount of $32,900 (2016$). On average, there were 104,500 residents
living in a county, with a commuter-adjusted daytime population of 115,000. In our
sample, 63 percent of the counties are classified as rural according to the National Cen-
ter for Health Statistics (NCHS) urban-rural 2013 classification scheme. We classify a
county as rural if its urban-rural 2013 classification scheme is greater than 4 (the scale
of population density ranges from 1 to 6, from most to least populated). On average,
a rural county has a population of 24,000 people vs. 240,500 living in an urban county.
Rural population accounts for 14 percent of the total population in our sample.
Public Health Measures and Relative Size of the PPP
In mid-March 2020, in response to the pandemic, the federal, state, and local govern-
ments instituted non-pharmaceutical interventions to curb the spread of the COVID-19
virus, which led to a significant drop in employment, especially for small businesses.
The Coronavirus Aid, Relief, and Economic Security (CARES) Act was introduced to
reduce the economic impact of mandatory shutdowns. As of April 2020, an average
county had received 926 PPP loans for a total of $126 million (2016$) and an average
loan of $90,450 (2016$). However, 58 percent of PPP loans were delayed (48 percent
if we use the total volume, not the number, of loans) because PPP funding ran out
on April 16 and was reinstated on April 27 (see Table 1). By August 2020, the av-
erage county had received 1,637 loans for a total volume of $156 million (2016$) (see
Table 1). Moreover, 635 out of 1,000 small establishments (fewer than 500 employees)
had received PPP funding, with an average loan size of $42,000 (2016$). Importantly,
the volume of 2020 PPP loans substantially exceeded each county’s SBL volume in
2019. In fact, by the end of April 2020, each county already had received, on average,
twice the 2019 SBL volume in PPP funding, and this multiple rose to three by the end
of the 2020 PPP.
B Evolution of the PPP Provisions
The 2020 CARES Act, signed into law on March 27, 2020, appropriated $349 billion in
PPP loans in response to the widespread shutdowns caused by the COVID-19 pandemic.
The PPP funds were provided to businesses that employed fewer than 500 workers and
36
had the resources to maintain or hire back employees that had been laid off and to
cover overhead costs incurred as a result of the pandemic.
Several key provisions in the CARES Act for the PPP were later modified in the
Paycheck Protection Program Flexibility Act (PPPFA).1 Four of these amended pro-
visions had the greatest potential to slow down the recovery of employment after the
initial acute phase of the pandemic. This effect carries the same sign as that of the
delay in funding or lack of funding for small businesses. More importantly, the likely
amendments to the original CARES Act provisions became known before April 27,
2020, when bank lending under the PPP resumed with the additional funding appro-
priation, and thus it could have led to differential behavior of firms that received loans
just before the 10-day window (which we will refer to as the early recipients) versus
those that received loans just after the window (the late recipients). In other words,
the cross-sectional disparity of these amendments’ impact could be correlated with the
degree of funding delay.
First, the PPPFA extended the period in which borrowers could spend their PPP
funds in order to be considered for loan forgiveness from eight weeks following the
date of the loan (that is, disbursement of loan proceeds) to the earlier of 24 weeks
following the date of the loan or December 31, 2020. Businesses that obtained PPP
loans before the effective date of the PPPFA, however, could elect to use the original
eight-week period, thereby allowing them to apply for forgiveness sooner. The proposal
to extend the covered period was first raised by the Main Street Alliance on April 22,
2020, and was reported by the Adhesives & Sealants Industry Magazine on April 23.
On April 29, 2020, it was reported by all journals (including the Portland Business
Journal ) under the umbrella of American City Business Journals. The Small Business
& Entrepreneurship Council more specifically proposed the 24-week expansion on April
30, which was then reported by the Wall Street Journal on May 3, 2020.2
The PPPFA also changed the loan proceeds use formula from 75 percent on payroll
and 25 percent on eligible fixed expenses (such as rent, interest on debt, and utilities)
to 60 percent on payroll and 40 percent on other eligible expenses. The formula change
was first reported by a major news outlet, USA Today, on April 20, 2020. That article
noted that the trade publication Nation’s Restaurant News reported a similar proposed
change to the formula on April 9, 2020. This provision permitted borrowers to keep or
hire back fewer employees because the firms were required to spend less of a given PPP
loan on payroll. The amount of a loan was still capped at 2.5 months of pre-pandemic
payroll.
Perhaps more importantly, the PPPFA relaxed the requirement that borrowers must
rehire employees by June 30, 2020: The amount that could potentially be forgiven would
not be reduced due to a decrease in the borrower’s full-time equivalent (FTE) work-
force count if the firm could document that it (1) attempted, but was unable, to rehire
previous employees as of February 15, 2020; and (2) was unable to hire “similarly qual-
ified employees” before December 31, 2020. This change was implemented to recognize
that some businesses may not have been allowed to reopen by June 30, 2020, and even
if they were allowed to reopen by that date, they may have had to reopen in stages,
thereby allowing them to hire back employees only over a longer period of time.
1
The PPPFA passed the House on May 28, 2020, and the Senate on June 3 of that year. It was
signed into law on June 4, 2020.
2
See articles here and here.
37
Additionally, the amount potentially forgiven would not be reduced due to a de-
creased in FTE workforce count if the borrower, in good faith, could document an
inability to return to the “same level of business activity” with which it operated prior
to February 15, 2020, due to sanitation, social distancing, or worker- or customer-safety
requirements. This provision recognized that businesses may not have been able to lo-
cate and hire qualified employees, because in many industries, the workforce could have
relocated due to the pandemic and become unavailable to employers. These last two
provisions made it less or not at all necessary for businesses to operate at close to their
pre-pandemic levels by June 30, and businesses could cut back permanently yet still
have their PPP loans forgiven. This change, among all the changes described so far,
can probably best help explain the persistent effect of the 10-day delay on employment
at the county level.
News about the broad contour of potential changes to the PPP program (resembling
the actual changes just described) that might be incorporated into a new bill to enhance
the program’s flexibility started circulating widely as early as April 11, 2020, when it
was reported by the New York Times and the Washington Post.3 Also, a bipartisan
group of House representatives sent a letter to House leadership on April 16 requesting
greater flexibility in the program in order to better assist small businesses.4 These
dates suggest that many small businesses, if they had already closed early on due
to the COVID-19 outbreak and the containment measures could reasonably start to
reconsider over the 10 days (when additional funding appropriation for the PPP was
going through the legislative process) their reopening plan. To the extent that a higher
fraction of the late borrowers had closed by April 27, 2020, when PPP lending resumed,
the additional “options” offered by the new provisions would have greater impact on
the late borrowers than the early borrowers.
In sum, several important changes became widely anticipated by small businesses
over the 10-day period of funding delay that could lead to reoptimization, in particular
to delaying reopening, since businesses could reasonably expect a much longer period
over which to spend the funding without having to worry about a reduced amount
of loan forgiveness. This incentive to delay reopening should be particularly strong
for May and June 2020, when demand was far from fully recovered. With the program
changes, businesses were able to preserve PPP funds for later use (to pay employees and
fixed operating expenses) when they could expect a higher volume of demand and/or
when the public health situation improved sufficiently.
The incentive to wait was likely stronger for businesses that closed prior to April
27, since they would need to spend a fixed cost (such as restocking the kitchen or
store shelves) to reopen. Thus, it likely applies especially to essential businesses that
had chosen to close, along with all nonessential businesses, which had to shut down in
most states during the lockdown phase in response to the initial COVID-19 outbreak.
In addition, it should be more relevant for localities that experienced less recovery of
demand in May and June because they were subject to greater mandatory containment
measures or voluntary cutbacks in mobility (due to greater perceived risk of infection,
which could depend on many factors, including higher density or media communication,
3
Similar reporting appeared as early as April 8, 2020, in Maine’s Bangor Daily News and in Restau-
rant Hospitality, a trade publication.
4
This effort was led by representatives Abigail Spanberger, Brian Fitzpatrick, and Josh Gottheimer;
see this House webpage.
38
not just actual infection rates).
More importantly, we argue that the option value of waiting was higher for small
firms that had not received PPP loans before the first round of funding ran out (on
April 16) because they were more likely to have closed due to the lack of liquidity.
In addition, they had not used any of the funds to pay and retain their employees
and thus faced no countervailing incentive to adhere to the original eight-week covered
period so that they could apply for loan forgiveness earlier. Moreover, it is possible
that, even among the recipients within the set considered for the natural experiment
(that is, those that received loans over April 15 and 16 versus April 27 and 28—just
before versus just after the 10-day window of delay), the firms that received PPP loans
earlier had, on average, better prospects even without the PPP, which at the margin
made it more likely for them to have stayed open or to be better equipped to reopen
sooner when public health conditions improved sufficiently. To the extent that such
correlation is present, the difference in reopening dynamics observed cannot be solely
attributed to the lack of funding. Given that this is unobserved heterogeneity, the only
indication we can test is that such firms were more likely to have stayed open prior to
receiving any funding, if they were allowed to, under the assumption that their better
prospect made it more valuable for them to stay open (to gain market share from their
rivals, for example). At the level of the locality, the implication is that those places
with relatively higher shares of PPP loans delayed may have had higher shares of such
firms with poorer prospect, after we control for observables.
In conclusion, the general idea is that the effect of the PPP on employment is not
just due to the liquidity it provided. In particular, the differential effect does not stem
just from the timing delay. Instead, it is the consequence of other design features of
the PPP, as well as its interactive effect with other legislative measures to mitigate the
impact of the pandemic.
39
Figure A.1: Time Spent at Workplaces since the COVID-19 Outbreak: The Roles of
Population and Density
0
-.2
-.4
-.6
-.8
Feb20 Apr20 Jun20 Aug20 Oct20 Dec20 Feb21 Apr21 Jun21 Aug21
Top 1% Pop Top 1% Density, No Top 1% Pop Other Urban
Notes: Dependent variable: fractional decline in time spent at workplaces in urban counties relative to
pre-pandemic averages. County-level fixed effects are partialed out. There are only two urban counties
in both the top 1% by population and the top 1% by density, and their coefficients are included in the
top 1% by population.
Source: Google Mobility data provided by Opportunity Insights.
40
Figure A.2: Effects of Selected Controls on QCEW Private Employment:
CBSA Regressions
All CBSAs, All Industries
.06 .05
.04
0
.02
-.05
0
-.1
-.02
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
CBSAs excluding Top 1%, All Industries
.1
.03
.02
.05
.01
0
0
-.05
-.01
Jan20 May20 Sep20 Jan21 May21 Sep21 Jan20 May20 Sep20 Jan21 May21 Sep21
Cumulative PPP Share Delayed Rebates UI replacement
Notes: Estimated effects for a change of one standard deviation in a given control, normalized by the
average corresponding CBSA-level employment in January 2020. Top 1% refers to population. The
bottom graphs correspond to regressions that exclude CBSAs in the top 1 percent of the population
distribution.
Source: Multiple data sources described in Section 3.2.
41
Figure A.3: R-Squared Decomposition, QCEW CBSA-Level Employment Regressions
All CBSAs CBSAs Excluding Top 1%
30 30
% R sq. Explained (cumulated) % R sq. Explained (cumulated)
20 20
10 10
0 0
Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21 Mar20 Jun20 Sep20 Dec20 Mar21 Jun21 Sep21
Cumulative PPP Share Delayed UI repl. Rebates COVID-19 Preexisting Conditions
Notes: Contributions of different variables to explaining the variance in private employment over time.
The effect of lagged employment is omitted in some graphs to more easily depict the contribution of other
variables. Top 1% refers to population.
Source: Multiple data sources described in Section 3.2.
42
Figure A.4: Map of Share of PPP Volume Delayed by County
Share Delayed
(.62,1]
(.5,.62]
(.42,.5]
(.33,.42]
[0,.33]
Notes: This map depicts, by quintile, the share of PPP volume delayed across all US counties.
Source: Small Business Administration.
Figure A.5: Top Counties by Total Population and by Population Density
County Groups
Top 1 percent by population
Top 1 percent by density
Other Counties
Notes: This map depicts the top 1 percent counties by population and by population density. The top
1 percent most populous counties are Maricopa County, AZ; Lo Angeles County, CA; Orange County,
CA;Riverside County, CA; San Diego County, CA; Miami-Dade County, FL; Cook County, IL; Kings County,
NY; Queen County, NY; Dallas County, TX, Harris County, TX; and King County, WA. The top 1 percent
of counties by population density are San Francisco County, CA; Suffolk County, MA; Hudson County, NJ;
Bronx County, NY; Kings County, NY; New York County, NY; Queens County, NY; Richmond County, NY;
Philadelphia County, PA; Arlington County; and Alexandria, VA.
Source: Census.
43
Table A.1: Determinants of Number of PPP Loans Delayed, April 16–26, 2020
All Urban Smaller Urban Rural
∗ ∗∗
Cum. COVID-19 Cases per bil. up to 4/15/2020 –0.017 0.035 0.044 –0.031
(0.023) (0.018) (0.018) (0.026)
Cum. COVID-19 Deaths per bil. up to 4/15/2020 0.108∗∗∗ 0.041 0.014 0.200∗∗∗
(0.031) (0.039) (0.038) (0.038)
Share of days in lockdown (pre-4/17/2020) 0.004 –0.034 –0.043 0.018
(0.030) (0.032) (0.032) (0.053)
Share of days in lockdown (4/17–4/30/2020) 0.066 0.013 0.020 0.102∗∗
(0.042) (0.134) (0.135) (0.040)
Share of Emp. in Essential Industries 0.021 –0.183 –0.199 0.164
(0.156) (0.261) (0.260) (0.190)
Share of Emp. in Impacted Industries –0.090∗ –0.110 –0.106 –0.065
(0.050) (0.077) (0.077) (0.063)
Rural County Dummy 0.009
(0.006)
Most Populous County (Top 1%) 0.023
(0.018)
Ln Residential Population –0.015∗∗∗ –0.013∗∗∗ –0.014∗∗∗ –0.017∗∗∗
(0.004) (0.004) (0.004) (0.005)
Commuter to Residential Population Ratio –0.053∗∗ –0.030 –0.026 –0.078∗∗
(0.020) (0.022) (0.023) (0.038)
Ln Median Family Income 0.028∗ 0.014 0.016 0.047
(0.016) (0.019) (0.019) (0.028)
Community Bank Share of Deposits 0.018∗ 0.001 0.001 0.021∗
(0.010) (0.017) (0.017) (0.012)
Big4 Bank Share of Deposits 0.070 0.059 0.056 0.071
(0.046) (0.079) (0.078) (0.048)
Ln Bank Branch Density 0.001 –0.014 –0.014 0.010
(0.009) (0.011) (0.011) (0.012)
SBL Vol. per Small Estabs. (< 500 Emp.) (CBP 2019Q1) –0.001∗∗∗ –0.001∗∗∗ –0.001∗∗∗ –0.001∗∗
(0.000) (0.000) (0.000) (0.000)
Proportion of Small Employment in 2020Q1 to 2019Q1, QWI 0.012 –0.102 –0.101 0.038
(0.045) (0.081) (0.080) (0.051)
UI Benefits Replacement Rate (Industry-Wtd.) 0.012 0.061∗ 0.059 –0.016
(0.029) (0.036) (0.036) (0.040)
Constant 0.385 0.926∗ 0.923∗ 0.010
(0.334) (0.502) (0.507) (0.397)
Adjusted R-squared 0.28 0.32 0.32 0.25
Observations 2644 1108 1096 1536
State FE Yes Yes Yes Yes
Notes: “Smaller Urban” refers to urban counties excluding those in the top 1 percent by population.
Source: Multiple data sources described in Section 3.2.
44
Table A.2: Determinants of Share of PPP Loan Volume Delayed (with Population
Density Indicator), April 16–26, 2020
All Urban Smaller Rural
∗∗∗ ∗∗∗
Cum. COVID-19 Cases per bil up to 4/15/2020 0.016 0.064 0.118 –0.006
(0.021) (0.020) (0.023) (0.036)
Cum. COVID-19 Deaths per bil up to 4/15/2020 0.062 0.022 –0.127∗∗ 0.080
(0.056) (0.058) (0.047) (0.097)
Share of days in lockdown (pre-4/17/2020) 0.033 –0.018 –0.014 –0.036
(0.056) (0.070) (0.069) (0.068)
Share of days in lockdown (4/17–4/30/2020) 0.119 –0.054 –0.063 0.289∗∗∗
(0.144) (0.133) (0.135) (0.072)
Share of Emp. in Essential Industries –0.249 –0.433 –0.503 0.046
(0.215) (0.340) (0.342) (0.258)
Share of Emp. in Impacted Industries –0.035 –0.126 –0.119 0.080
(0.060) (0.117) (0.116) (0.082)
Rural County Dummy 0.006
(0.010)
Most Dense Urban County 0.038 0.044
(0.049) (0.053)
Ln Residential Population –0.022∗∗∗ –0.020∗∗∗ –0.022∗∗∗ –0.030∗∗∗
(0.005) (0.007) (0.007) (0.010)
Commuter to Residential Population Ratio –0.028 –0.020 –0.013 –0.030
(0.027) (0.030) (0.031) (0.062)
Ln Median Family Income 0.014 0.012 0.013 0.017
(0.026) (0.035) (0.037) (0.053)
Community Bank Share of Deposits 0.004 0.040∗∗ 0.041∗∗ –0.008
(0.014) (0.016) (0.015) (0.019)
Big4 Bank Share of Deposits 0.070 0.116 0.102 0.052
(0.071) (0.093) (0.088) (0.074)
Ln Bank Branch Density –0.005 –0.007 –0.009 –0.001
(0.009) (0.013) (0.013) (0.015)
SBL Vol. per Small Estabs. (< 500 Emp.) (CBP 2019Q1) –0.002∗∗∗ –0.002∗∗∗ –0.002∗∗∗ –0.002∗∗∗
(0.000) (0.001) (0.001) (0.000)
Proportion of Small Employment in 2020Q1 to 2019Q1, QWI 0.000 –0.002 –0.002∗ 0.001
(0.001) (0.001) (0.001) (0.001)
UI Benefits Replacement Rate (Industry-Wtd.) –0.028 –0.007 –0.026 –0.033
(0.041) (0.059) (0.058) (0.056)
Constant 0.774 1.260∗ 1.390∗∗ 0.307
(0.493) (0.672) (0.685) (0.717)
Adjusted R-squared 0.13 0.17 0.18 0.10
Observations 2644 1108 1097 1536
State FE Yes Yes Yes Yes
Notes: “Smaller” refers to urban counties excluding those in the top 1 percent by population density.
Source: Multiple data sources described in Section 3.2.
45
Table A.3: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment in Less Densely Populated Urban Counties
Total NAICS
Employment 71, 72 and 82
(1) (2) (3) (4)
Jan 2020 × Share Delayed 145 334∗ 113∗ 32
(118) (184) (67) (82)
Feb 2020 × Share Delayed 227 529∗∗ 112∗∗ 67
(176) (218) (49) (76)
Apr 2020 × Share Delayed –779 120 38 112
(920) (917) (221) (232)
May 2020 × Share Delayed –1,243 –422 –182 –33
(897) (953) (330) (339)
Jun 2020 × Share Delayed –1,274 –858 –469 –300
(834) (857) (386) (379)
Jul 2020 × Share Delayed –1,143 –868 –394 –265
(827) (835) (409) (396)
Aug 2020 × Share Delayed –950 –799 –343 –265
(825) (845) (441) (422)
Sept 2020 × Share Delayed –778 –676 –336 –262
(783) (800) (381) (358)
Oct 2020 × Share Delayed –543 –401 –174 –101
(697) (723) (306) (288)
Nov 2020 × Share Delayed –520 –330 –212 –102
(691) (727) (296) (285)
Dec 2020 × Share Delayed –315 –94 –59 110
(770) (827) (377) (376)
Jan 2021 × Share Delayed –446 –221 –255 –75
(851) (916) (435) (434)
Feb 2021 × Share Delayed –319 –257 –261 –230
(715) (749) (339) (328)
Mar 2021 × Share Delayed –399 –341 –264 –255
(702) (723) (330) (316)
Apr 2021 × Share Delayed –180 –173 –91 –120
(681) (689) (290) (275)
May 2021 × Share Delayed –108 –110 –73 –126
(672) (671) (295) (276)
Jun 2021 × Share Delayed 121 90 11 –66
(720) (712) (326) (306)
Jul 2021 × Share Delayed 6 40 105 6
(724) (713) (317) (302)
Aug 2021 × Share Delayed 13 128 113 68
(694) (676) (284) (275)
Sept 2021 × Share Delayed 313 463 149 66
(656) (643) (216) (207)
Average Private Employment 87,100 12,776
St. Dev. of Private Employment 206,184 30,046
Within R-squared 0.87 0.88 0.90 0.91
Observations 23,205 23,205 23,205 23,205
County and State-by-Month FE Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes
COVID-19 Controls No Yes No Yes
CARES Act Controls No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller”
refers to urban counties excluding those in the top 1 percent by population den-
sity. Share Delayed is the share (by volume) of county-level PPP loans delayed
as defined in Equation (2). Preexisting Conditions Controls: median fam-
ily income, commuter-to-residential-population ratio, indicators for 2013 NCHS
urban-rural designation, population, community-bank share of deposits, largest
four banks’ share of deposits, bank branch density, and 2019 small-business-loan
volume per small establishment; COVID-19 Controls: cumulative COVID-
19 cases and deaths per population, share of days in April 2020 in early lock-
down, share of employment in essential industries, and share of employment
in most impacted industries; CARES Act Controls: industry-employment-
share-weighted UI benefits replacement rate, and rebates (“stimulus checks”)
per capita. For employment in NAICS 71, 72, and 82, Cum PPP per Emp
in Small Estab (t-1) is specific to these impacted industries.
Source: Multiple data sources described in Section 3.2.
46
Table A.4: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment: Industry Controls by Share of Wages instead of Employment
All Counties Urban Smaller Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Jan 2020 × Share Delayed 53 179 183 183 270∗ 473∗ 100 209 –20∗ –18
(35) (121) (121) (121) (153) (279) (106) (194) (12) (12)
Feb 2020 × Share Delayed 143∗∗∗ 272∗∗ 284∗∗ 284∗∗ 325∗ 560∗∗ 239∗∗ 372∗ –18∗∗ –16
(51) (120) (122) (122) (169) (261) (107) (210) (9) (10)
Apr 2020 × Share Delayed –731∗∗ –534∗ –473 –476 –3,037∗∗ –2,006 –1,688∗ –1,063 –36 –15
(329) (294) (296) (296) (1,428) (1,320) (931) (887) (57) (57)
May 2020 × Share Delayed –891∗∗∗ –662∗∗ –602∗∗ –597∗∗ –3,400∗∗ –2,113∗ –2,446∗∗ –1,621 –2 21
(336) (286) (283) (282) (1,462) (1,243) (1,213) (1,024) (48) (49)
Jun 2020 × Share Delayed –974∗∗∗ –734∗∗∗ –673∗∗ –665∗∗ –3,739∗∗ –2,265∗ –2,873∗ –1,854 31 51
(366) (281) (274) (274) (1,675) (1,266) (1,564) (1,137) (40) (40)
Jul 2020 × Share Delayed –1,023∗∗∗ –831∗∗∗ –769∗∗∗ –758∗∗∗ –3,383∗∗ –1,995∗ –2,672∗ –1,670 40 57
(354) (276) (269) (269) (1,607) (1,189) (1,596) (1,115) (38) (39)
Aug 2020 × Share Delayed –942∗∗∗ –760∗∗∗ –731∗∗∗ –717∗∗∗ –2,921∗ –1,591 –2,400 –1,411 34 55
(342) (268) (264) (263) (1,537) (1,139) (1,574) (1,103) (36) (36)
Sept 2020 × Share Delayed –818∗∗ –662∗∗∗ –625∗∗ –612∗∗ –2,449∗ –1,195 –2,146 –1,199 30 51
(326) (252) (247) (247) (1,471) (1,062) (1,536) (1,071) (35) (35)
Oct 2020 × Share Delayed –603∗ –443∗ –400∗ –388∗ –1,887 –714 –1,755 –838 34 57
(312) (231) (225) (225) (1,419) (1,006) (1,493) (1,042) (35) (35)
Nov 2020 × Share Delayed –492 –356 –310 –300 –1,827 –641 –1,674 –725 53 76∗∗
(326) (240) (231) (231) (1,477) (1,025) (1,540) (1,068) (34) (35)
Dec 2020 × Share Delayed –464 –332 –294 –277 –1,485 –326 –1,461 –561 56∗ 75∗∗
(331) (249) (242) (242) (1,481) (1,063) (1,548) (1,091) (33) (34)
Jan 2021 × Share Delayed –604∗ –449 –413 –392 –1,938 –703 –1,772 –892 58∗ 76∗∗
(359) (276) (269) (268) (1,584) (1,155) (1,662) (1,177) (34) (35)
Feb 2021 × Share Delayed –427 –264 –224 –242 –1,875 –836 –1,676 –940 61∗ 79∗∗
(357) (256) (247) (247) (1,593) (1,070) (1,637) (1,166) (35) (35)
Mar 2021 × Share Delayed –449 –289 –249 –278 –1,809 –780 –1,719 –984 53 69∗∗
(344) (244) (236) (235) (1,556) (1,038) (1,604) (1,131) (34) (35)
Apr 2021 × Share Delayed –374 –211 –163 –208 –1,264 –333 –1,333 –651 36 43
(306) (220) (216) (215) (1,413) (963) (1,506) (1,059) (40) (41)
May 2021 × Share Delayed –329 –160 –114 –159 –1,194 –293 –1,193 –542 39 45
(297) (213) (210) (209) (1,388) (930) (1,449) (1,015) (39) (40)
Jun 2021 × Share Delayed –336 –190 –143 –196 –880 –73 –860 –262 32 29
(273) (208) (206) (207) (1,256) (896) (1,300) (929) (41) (42)
Jul 2021 × Share Delayed –391 –208 –167 –209 –1,479 –567 –1,134 –490 15 11
(278) (209) (209) (210) (1,296) (907) (1,315) (957) (46) (46)
Aug 2021 × Share Delayed –359 –172 –127 –159 –1,617 –606 –1,182 –478 14 10
(282) (208) (209) (209) (1,315) (913) (1,291) (933) (43) (43)
Sept 2021 × Share Delayed –167 2 27 –9 –932 –103 –702 –192 17 18
(248) (189) (189) (189) (1,141) (811) (1,099) (849) (38) (38)
Average Private Employment 38,956 91,195 76,404 6,489
St. Dev. of Private Employment 139,524 215,187 138,117 7,193
Within R-squared 0.86 0.86 0.86 0.86 0.85 0.86 0.78 0.79 0.48 0.51
Observations 61,173 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes Yes No Yes No Yes No Yes
CARES Act Controls No No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top 1 per-
cent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preexisting
Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation,
population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and 2019 small-business-loan vol-
ume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths per population, share of days in April 2020 in
early lockdown, share of wages in essential industries, and share of wages in most impacted industries; CARES Act Controls: industry-
employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
47
Table A.5: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment at Different Cutoffs - Urban Counties
All Top 99 Top 95 Top 90
(1) (2) (3) (4) (5) (6) (7) (8)
Jan 2020 × Share Delayed 270∗ 451∗ 100 201 25 –11 5 12
(153) (270) (106) (190) (89) (99) (77) (74)
Feb 2020 × Share Delayed 325∗ 546∗∗ 239∗∗ 370∗ 77 74 59 74
(169) (258) (107) (204) (72) (76) (65) (66)
Apr 2020 × Share Delayed –3,037∗∗ –2,063 –1,688∗ –1,068 128 209 184 224
(1,428) (1,311) (931) (887) (548) (548) (427) (424)
May 2020 × Share Delayed –3,400∗∗ –2,182∗ –2,446∗∗ –1,616 –348 –299 –187 –186
(1,462) (1,233) (1,213) (1,011) (514) (514) (385) (387)
Jun 2020 × Share Delayed –3,739∗∗ –2,260∗ –2,873∗ –1,792 –472 –429 –299 –288
(1,675) (1,253) (1,564) (1,112) (479) (483) (375) (377)
Jul 2020 × Share Delayed –3,383∗∗ –1,944∗ –2,672∗ –1,593 –308 –273 –277 –268
(1,607) (1,175) (1,596) (1,086) (446) (455) (350) (353)
Aug 2020 × Share Delayed –2,921∗ –1,569 –2,400 –1,356 –176 –141 –189 –166
(1,537) (1,131) (1,574) (1,078) (442) (453) (337) (341)
Sept 2020 × Share Delayed –2,449∗ –1,210 –2,146 –1,173 –153 –148 –235 –241
(1,471) (1,059) (1,536) (1,048) (431) (440) (331) (335)
Oct 2020 × Share Delayed –1,887 –744 –1,755 –819 –8 –29 –183 –212
(1,419) (1,003) (1,493) (1,020) (422) (431) (306) (308)
Nov 2020 × Share Delayed –1,827 –668 –1,674 –699 55 28 –136 –172
(1,477) (1,024) (1,540) (1,045) (429) (438) (300) (300)
Dec 2020 × Share Delayed –1,485 –375 –1,461 –546 141 90 –84 –141
(1,481) (1,068) (1,548) (1,071) (458) (465) (314) (314)
Jan 2021 × Share Delayed –1,938 –721 –1,772 –844 110 44 –127 –192
(1,584) (1,164) (1,662) (1,157) (469) (486) (316) (323)
Feb 2021 × Share Delayed –1,875 –841 –1,676 –878 122 31 –124 –171
(1,593) (1,072) (1,637) (1,142) (441) (441) (308) (312)
Mar 2021 × Share Delayed –1,809 –766 –1,719 –904 50 –36 –137 –170
(1,556) (1,038) (1,604) (1,104) (427) (419) (301) (302)
Apr 2021 × Share Delayed –1,264 –282 –1,333 –548 129 46 –113 –113
(1,413) (963) (1,506) (1,032) (423) (404) (323) (327)
May 2021 × Share Delayed –1,194 –217 –1,193 –424 322 260 61 78
(1,388) (927) (1,449) (987) (404) (385) (292) (298)
Jun 2021 × Share Delayed –880 19 –860 –143 508 471 265 287
(1,256) (893) (1,300) (905) (422) (412) (324) (334)
Jul 2021 × Share Delayed –1,479 –441 –1,134 –349 398 446 223 315
(1,296) (900) (1,315) (934) (438) (437) (359) (368)
Aug 2021 × Share Delayed –1,617 –507 –1,182 –360 350 424 161 250
(1,315) (907) (1,291) (914) (428) (429) (349) (356)
Sept 2021 × Share Delayed –932 –19 –702 –79 520 563 307 380
(1,141) (809) (1,099) (831) (387) (376) (305) (301)
Average Private Employment 91,195 76,404 55,804 42,245
St. Dev. of Private Employment 215,187 138,117 78,752 50,533
Within R-squared 0.85 0.86 0.78 0.79 0.77 0.78 0.73 0.75
Observations 23,436 23,436 23,184 23,184 22,239 22,239 21,063 21,063
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes No Yes No Yes No Yes
CARES Act Controls No Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Top 99” refers to urban counties excluding those
in the top 1 percent by population. “Top 95” refers to urban counties excluding those in the top 5 percent by popu-
lation. “Top 90” refers to urban counties excluding those in the top 10 percent by population. Share Delayed is the
share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preexisting Conditions Controls:
median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation,
population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and 2019
small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths
per population, share of days in April 2020 in early lockdown, share of wages in essential industries, and share of wages
in most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits replacement
rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
48
Table A.6: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment: Log Employment as Dependent Variable
All Counties Urban Smaller Urban Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Jan 2020 × Share Delayed –0.007∗∗ –0.008∗∗ –0.008∗∗ 0.004 0.003 0.004 0.002 –0.011∗∗∗ –0.011∗∗∗
(0.003) (0.003) (0.003) (0.005) (0.005) (0.005) (0.005) (0.004) (0.004)
Feb 2020 × Share Delayed 0.001 0.000 0.000 0.005 0.004 0.005 0.004 –0.001 –0.001
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) (0.004)
Apr 2020 × Share Delayed 0.014 0.013 0.012 0.048∗∗ 0.041∗∗ 0.050∗∗∗ 0.043∗∗ 0.003 0.005
(0.009) (0.009) (0.009) (0.019) (0.018) (0.019) (0.018) (0.010) (0.010)
May 2020 × Share Delayed 0.015∗ 0.014∗ 0.015∗ 0.027 0.026 0.028 0.027 0.010 0.013
(0.008) (0.008) (0.008) (0.018) (0.018) (0.018) (0.018) (0.010) (0.009)
Jun 2020 × Share Delayed 0.007 0.007 0.008 0.009 0.012 0.010 0.013 0.006 0.008
(0.007) (0.007) (0.007) (0.014) (0.014) (0.014) (0.014) (0.009) (0.009)
Jul 2020 × Share Delayed 0.014∗ 0.014∗ 0.016∗∗ 0.002 0.005 0.003 0.006 0.017∗ 0.020∗∗
(0.008) (0.008) (0.008) (0.014) (0.014) (0.014) (0.014) (0.010) (0.010)
Aug 2020 × Share Delayed 0.010 0.010 0.012 0.002 0.006 0.003 0.007 0.012 0.014
(0.008) (0.008) (0.008) (0.013) (0.013) (0.013) (0.013) (0.010) (0.010)
Sept 2020 × Share Delayed 0.009 0.009 0.011 0.003 0.007 0.004 0.008 0.010 0.012
(0.008) (0.008) (0.008) (0.013) (0.014) (0.013) (0.014) (0.009) (0.009)
Oct 2020 × Share Delayed 0.006 0.006 0.008 0.010 0.014 0.011 0.014 0.004 0.007
(0.008) (0.008) (0.008) (0.014) (0.014) (0.014) (0.014) (0.010) (0.010)
Nov 2020 × Share Delayed 0.008 0.008 0.010 0.012 0.014 0.013 0.015 0.006 0.009
(0.008) (0.008) (0.008) (0.014) (0.014) (0.014) (0.014) (0.010) (0.009)
Dec 2020 × Share Delayed 0.011 0.009 0.012 0.012 0.012 0.012 0.012 0.009 0.012
(0.008) (0.008) (0.008) (0.013) (0.013) (0.013) (0.013) (0.009) (0.009)
Jan 2021 × Share Delayed 0.021∗∗ 0.021∗∗ 0.023∗∗ 0.007 0.009 0.007 0.010 0.023∗∗ 0.027∗∗
(0.009) (0.009) (0.009) (0.016) (0.015) (0.016) (0.015) (0.011) (0.011)
Feb 2021 × Share Delayed 0.013 0.013 0.016∗ 0.007 0.009 0.007 0.009 0.014 0.018
(0.010) (0.009) (0.009) (0.015) (0.014) (0.015) (0.014) (0.012) (0.011)
Mar 2021 × Share Delayed 0.012 0.012 0.014 0.000 0.003 0.000 0.003 0.014 0.018
(0.010) (0.010) (0.009) (0.015) (0.015) (0.015) (0.015) (0.012) (0.011)
Apr 2021 × Share Delayed 0.003 0.004 0.005 –0.009 –0.004 –0.009 –0.004 0.007 0.009
(0.011) (0.011) (0.011) (0.014) (0.014) (0.015) (0.014) (0.014) (0.013)
May 2021 × Share Delayed 0.003 0.004 0.005 –0.003 0.002 –0.004 0.002 0.005 0.007
(0.012) (0.012) (0.012) (0.015) (0.014) (0.015) (0.014) (0.015) (0.014)
Jun 2021 × Share Delayed 0.009 0.011 0.010 0.012 0.018 0.012 0.017 0.010 0.011
(0.013) (0.013) (0.013) (0.016) (0.015) (0.016) (0.015) (0.016) (0.016)
Jul 2021 × Share Delayed 0.004 0.006 0.006 0.012 0.019 0.012 0.018 0.002 0.003
(0.015) (0.014) (0.014) (0.019) (0.018) (0.019) (0.018) (0.018) (0.017)
Aug 2021 × Share Delayed 0.006 0.008 0.008 0.017 0.023 0.017 0.023 0.003 0.004
(0.015) (0.015) (0.014) (0.021) (0.020) (0.021) (0.020) (0.018) (0.017)
Sept 2021 × Share Delayed 0.008 0.010 0.011 0.015 0.021 0.015 0.021 0.006 0.008
(0.014) (0.014) (0.014) (0.019) (0.018) (0.019) (0.018) (0.017) (0.017)
Average Private Employment 8.93 10.14 10.10 8.18
St. Dev. of Private Employment 1.70 1.69 1.65 1.20
Within R-squared 0.16 0.21 0.22 0.16 0.26 0.16 0.26 0.15 0.20
Observations 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes No Yes
CARES Act Controls No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top 1
percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preex-
isting Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural
designation, population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and 2019 small-
business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths per population, share of
days in April 2020 in early lockdown, share of employment in essential industries, and share of employment in most impacted industries;
CARES Act Controls: industry-employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
49
Table A.7: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment in NAICS 71, 72, and 81: Log Employment as Dependent Variable
All Counties Urban Smaller Urban Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Jan 2020 × Share Delayed 0.009 0.007 0.007 0.026∗ 0.017 0.025∗ 0.017 0.006 0.006
(0.010) (0.010) (0.010) (0.014) (0.013) (0.014) (0.013) (0.012) (0.012)
Feb 2020 × Share Delayed 0.004 0.003 0.004 0.010 0.004 0.009 0.004 0.003 0.005
(0.008) (0.008) (0.008) (0.012) (0.012) (0.012) (0.012) (0.010) (0.010)
Apr 2020 × Share Delayed 0.044∗ 0.045∗ 0.045∗ 0.041 0.036 0.042 0.038 0.043 0.045
(0.023) (0.024) (0.024) (0.039) (0.040) (0.039) (0.040) (0.028) (0.028)
May 2020 × Share Delayed 0.022 0.023 0.026 0.026 0.033 0.026 0.032 0.021 0.026
(0.018) (0.018) (0.018) (0.036) (0.037) (0.037) (0.037) (0.020) (0.020)
Jun 2020 × Share Delayed 0.012 0.015 0.017 0.020 0.031 0.021 0.032 0.009 0.015
(0.016) (0.016) (0.016) (0.033) (0.034) (0.034) (0.034) (0.019) (0.019)
Jul 2020 × Share Delayed –0.011 –0.009 –0.006 –0.028 –0.018 –0.027 –0.016 –0.009 –0.005
(0.017) (0.017) (0.017) (0.032) (0.032) (0.032) (0.032) (0.020) (0.020)
Aug 2020 × Share Delayed –0.028 –0.027 –0.024 –0.021 –0.015 –0.020 –0.014 –0.032 –0.028
(0.017) (0.017) (0.017) (0.029) (0.029) (0.029) (0.029) (0.021) (0.020)
Sept 2020 × Share Delayed –0.017 –0.015 –0.013 –0.019 –0.017 –0.018 –0.016 –0.019 –0.015
(0.018) (0.018) (0.018) (0.031) (0.030) (0.031) (0.030) (0.022) (0.022)
Oct 2020 × Share Delayed –0.025 –0.024 –0.021 –0.023 –0.021 –0.023 –0.020 –0.025 –0.020
(0.017) (0.017) (0.017) (0.030) (0.031) (0.030) (0.031) (0.021) (0.021)
Nov 2020 × Share Delayed –0.013 –0.013 –0.011 –0.007 –0.006 –0.006 –0.006 –0.015 –0.012
(0.018) (0.018) (0.018) (0.034) (0.035) (0.034) (0.035) (0.020) (0.020)
Dec 2020 × Share Delayed –0.008 –0.008 –0.006 0.004 0.005 0.004 0.005 –0.012 –0.009
(0.018) (0.018) (0.018) (0.033) (0.035) (0.033) (0.035) (0.021) (0.021)
Jan 2021 × Share Delayed –0.016 –0.015 –0.012 –0.033 –0.028 –0.032 –0.027 –0.012 –0.008
(0.022) (0.022) (0.022) (0.042) (0.043) (0.043) (0.043) (0.026) (0.026)
Feb 2021 × Share Delayed –0.017 –0.016 –0.015 –0.012 –0.013 –0.012 –0.013 –0.016 –0.013
(0.022) (0.022) (0.022) (0.039) (0.041) (0.039) (0.042) (0.026) (0.026)
Mar 2021 × Share Delayed –0.015 –0.014 –0.012 0.004 0.008 0.004 0.008 –0.020 –0.017
(0.023) (0.023) (0.023) (0.040) (0.041) (0.040) (0.041) (0.027) (0.027)
Apr 2021 × Share Delayed –0.025 –0.022 –0.020 –0.043 –0.034 –0.045 –0.036 –0.018 –0.014
(0.029) (0.029) (0.029) (0.048) (0.049) (0.049) (0.049) (0.034) (0.034)
May 2021 × Share Delayed –0.007 –0.004 –0.003 –0.013 –0.003 –0.012 –0.002 –0.002 –0.001
(0.025) (0.025) (0.025) (0.049) (0.048) (0.049) (0.048) (0.030) (0.030)
Jun 2021 × Share Delayed 0.007 0.009 0.009 0.012 0.016 0.012 0.015 0.011 0.011
(0.023) (0.023) (0.023) (0.040) (0.039) (0.040) (0.039) (0.028) (0.028)
Jul 2021 × Share Delayed 0.002 0.002 0.003 0.021 0.022 0.019 0.021 –0.000 –0.000
(0.023) (0.023) (0.023) (0.039) (0.038) (0.039) (0.038) (0.028) (0.028)
Aug 2021 × Share Delayed 0.016 0.017 0.017 0.021 0.026 0.021 0.025 0.016 0.016
(0.023) (0.023) (0.023) (0.042) (0.042) (0.043) (0.042) (0.027) (0.027)
Sept 2021 × Share Delayed 0.030 0.030 0.031 0.038 0.041 0.037 0.040 0.031 0.030
(0.025) (0.025) (0.025) (0.043) (0.043) (0.044) (0.044) (0.030) (0.030)
Average Private Employment 6.42 7.99 7.95 5.44
St. Dev. of Private Employment 2.58 2.24 2.21 2.27
Within R-squared 0.96 0.96 0.96 0.95 0.96 0.95 0.96 0.97 0.97
Observations 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes No Yes
CARES Act Controls No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the
top 1 percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equa-
tion (2). Preexisting Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for
2013 NCHS urban-rural designation, population, community-bank share of deposits, largest four banks’ share of deposits, bank
branch density, and 2019 small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19
cases and deaths per population, share of days in April 2020 in early lockdown, share of employment in essential industries, and
share of employment in most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits
replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
50
Table A.8: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment Growth Rate
All Counties Urban Smaller Urban Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Jan 2020 × Share Delayed –0.010∗∗ –0.010∗∗ –0.010∗∗ 0.009 0.009 0.009 0.008 –0.014∗∗∗ –0.015∗∗∗
(0.004) (0.004) (0.004) (0.007) (0.007) (0.007) (0.007) (0.005) (0.005)
Feb 2020 × Share Delayed 0.001 0.000 0.000 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗ –0.002 –0.003
(0.004) (0.004) (0.004) (0.005) (0.005) (0.005) (0.005) (0.004) (0.004)
Apr 2020 × Share Delayed 0.017∗ 0.016∗ 0.015∗ 0.054∗∗∗ 0.048∗∗∗ 0.056∗∗∗ 0.051∗∗∗ 0.005 0.006
(0.009) (0.009) (0.009) (0.019) (0.018) (0.019) (0.018) (0.010) (0.010)
May 2020 × Share Delayed 0.019∗∗ 0.018∗∗ 0.018∗∗ 0.036∗∗ 0.035∗∗ 0.037∗∗ 0.036∗∗ 0.014 0.015
(0.009) (0.009) (0.009) (0.018) (0.018) (0.018) (0.018) (0.010) (0.010)
Jun 2020 × Share Delayed 0.011 0.011 0.011 0.018 0.019 0.019 0.021 0.009 0.010
(0.008) (0.008) (0.008) (0.014) (0.014) (0.014) (0.014) (0.009) (0.009)
Jul 2020 × Share Delayed 0.021∗∗ 0.021∗∗ 0.022∗∗∗ 0.004 0.004 0.006 0.005 0.024∗∗ 0.026∗∗∗
(0.008) (0.008) (0.008) (0.015) (0.015) (0.015) (0.015) (0.010) (0.010)
Aug 2020 × Share Delayed 0.021∗∗ 0.020∗∗ 0.022∗∗ 0.015 0.015 0.016 0.016 0.021∗∗ 0.024∗∗
(0.009) (0.008) (0.008) (0.013) (0.013) (0.013) (0.013) (0.010) (0.010)
Sept 2020 × Share Delayed 0.018∗∗ 0.018∗∗ 0.019∗∗ 0.011 0.011 0.012 0.013 0.019∗ 0.021∗∗
(0.009) (0.009) (0.009) (0.014) (0.014) (0.014) (0.014) (0.011) (0.010)
Oct 2020 × Share Delayed 0.011 0.011 0.012 0.009 0.011 0.010 0.012 0.011 0.013
(0.009) (0.009) (0.009) (0.016) (0.016) (0.016) (0.016) (0.011) (0.011)
Nov 2020 × Share Delayed 0.011 0.010 0.011 0.012 0.013 0.013 0.014 0.010 0.012
(0.010) (0.010) (0.010) (0.015) (0.015) (0.016) (0.016) (0.012) (0.011)
Dec 2020 × Share Delayed 0.013 0.012 0.014 0.014 0.013 0.014 0.013 0.012 0.015
(0.009) (0.009) (0.009) (0.015) (0.015) (0.015) (0.015) (0.011) (0.011)
Jan 2021 × Share Delayed 0.031∗∗∗ 0.031∗∗∗ 0.033∗∗∗ 0.015 0.019 0.016 0.020 0.034∗∗ 0.036∗∗∗
(0.011) (0.011) (0.011) (0.017) (0.016) (0.017) (0.017) (0.014) (0.014)
Feb 2021 × Share Delayed 0.019 0.019 0.020∗ 0.015 0.018 0.016 0.018 0.019 0.022
(0.012) (0.012) (0.012) (0.017) (0.017) (0.017) (0.017) (0.015) (0.015)
Mar 2021 × Share Delayed 0.017 0.018 0.019 0.011 0.013 0.011 0.014 0.019 0.022
(0.012) (0.012) (0.012) (0.017) (0.017) (0.017) (0.017) (0.015) (0.015)
Apr 2021 × Share Delayed –0.000 0.001 0.001 –0.033 –0.024 –0.035 –0.026 0.010 0.010
(0.015) (0.015) (0.015) (0.023) (0.022) (0.023) (0.022) (0.018) (0.018)
May 2021 × Share Delayed –0.001 0.000 –0.001 –0.012 –0.006 –0.014 –0.007 0.003 0.002
(0.015) (0.015) (0.015) (0.021) (0.021) (0.021) (0.021) (0.018) (0.018)
Jun 2021 × Share Delayed 0.010 0.011 0.010 0.016 0.019 0.015 0.018 0.011 0.010
(0.015) (0.015) (0.015) (0.018) (0.017) (0.018) (0.017) (0.018) (0.018)
Jul 2021 × Share Delayed 0.001 0.003 0.001 0.019 0.023 0.019 0.022 –0.004 –0.004
(0.016) (0.015) (0.015) (0.020) (0.019) (0.020) (0.019) (0.019) (0.019)
Aug 2021 × Share Delayed 0.008 0.009 0.008 0.027 0.030 0.027 0.029 0.003 0.002
(0.016) (0.016) (0.016) (0.022) (0.021) (0.022) (0.021) (0.019) (0.019)
Sept 2021 × Share Delayed 0.010 0.011 0.011 0.023 0.027 0.023 0.026 0.007 0.007
(0.015) (0.015) (0.015) (0.020) (0.019) (0.020) (0.019) (0.019) (0.019)
Average Private Employment –0.02 –0.02 –0.02 –0.02
St. Dev. of Private Employment 0.09 0.09 0.09 0.09
Within R-squared 0.03 0.08 0.09 0.06 0.17 0.06 0.17 0.03 0.07
Observations 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes No Yes
CARES Act Controls No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top 1 percent
by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preexisting Con-
ditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation,
population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and 2019 small-business-loan
volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths per population, share of days in April
2020 in early lockdown, share of employment in essential industries, and share of employment in most impacted industries; CARES Act
Controls: industry-employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
51
Table A.9: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment Growth Rate in NAICS 71, 82, and 81 Industries
All Counties Urban Smaller Urban Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Jan 2020 × Share Delayed 0.008 0.007 0.007 0.024∗ 0.015 0.023∗ 0.015 0.006 0.006
(0.010) (0.010) (0.010) (0.014) (0.013) (0.014) (0.013) (0.012) (0.012)
Feb 2020 × Share Delayed 0.003 0.003 0.003 0.009 0.003 0.008 0.003 0.003 0.004
(0.008) (0.008) (0.008) (0.012) (0.012) (0.012) (0.012) (0.010) (0.010)
Apr 2020 × Share Delayed 0.068∗∗∗ 0.068∗∗∗ 0.068∗∗∗ 0.098∗∗ 0.087∗ 0.098∗∗ 0.087∗ 0.055∗ 0.058∗
(0.026) (0.026) (0.026) (0.044) (0.044) (0.044) (0.045) (0.031) (0.031)
May 2020 × Share Delayed 0.039∗∗ 0.039∗∗ 0.041∗∗ 0.067∗ 0.064 0.066 0.063 0.029 0.035
(0.019) (0.019) (0.019) (0.041) (0.041) (0.041) (0.041) (0.022) (0.022)
Jun 2020 × Share Delayed 0.023 0.025 0.028∗ 0.047 0.053 0.048 0.054 0.015 0.021
(0.017) (0.017) (0.017) (0.035) (0.035) (0.035) (0.035) (0.019) (0.019)
Jul 2020 × Share Delayed –0.003 –0.002 0.001 –0.006 –0.001 –0.005 0.000 –0.005 –0.001
(0.018) (0.017) (0.017) (0.033) (0.033) (0.033) (0.033) (0.021) (0.021)
Aug 2020 × Share Delayed –0.021 –0.020 –0.017 –0.000 0.001 0.000 0.002 –0.028 –0.024
(0.017) (0.017) (0.017) (0.030) (0.029) (0.030) (0.029) (0.021) (0.021)
Sept 2020 × Share Delayed –0.011 –0.010 –0.008 0.001 –0.002 0.002 –0.001 –0.016 –0.012
(0.018) (0.018) (0.018) (0.032) (0.031) (0.032) (0.031) (0.022) (0.022)
Oct 2020 × Share Delayed –0.017 –0.017 –0.014 –0.020 –0.019 –0.020 –0.019 –0.017 –0.013
(0.018) (0.018) (0.018) (0.030) (0.031) (0.030) (0.031) (0.021) (0.021)
Nov 2020 × Share Delayed –0.006 –0.006 –0.004 –0.004 –0.005 –0.004 –0.005 –0.008 –0.005
(0.018) (0.018) (0.018) (0.034) (0.035) (0.034) (0.035) (0.021) (0.021)
Dec 2020 × Share Delayed –0.001 –0.001 0.001 0.009 0.008 0.008 0.007 –0.005 –0.002
(0.018) (0.018) (0.018) (0.033) (0.035) (0.033) (0.035) (0.021) (0.021)
Jan 2021 × Share Delayed –0.011 –0.010 –0.008 –0.025 –0.021 –0.024 –0.020 –0.009 –0.005
(0.023) (0.023) (0.023) (0.042) (0.042) (0.042) (0.042) (0.026) (0.026)
Feb 2021 × Share Delayed –0.011 –0.011 –0.009 –0.005 –0.006 –0.005 –0.006 –0.012 –0.010
(0.023) (0.023) (0.023) (0.038) (0.041) (0.039) (0.041) (0.027) (0.027)
Mar 2021 × Share Delayed –0.012 –0.011 –0.009 0.009 0.013 0.009 0.013 –0.017 –0.015
(0.023) (0.023) (0.023) (0.039) (0.040) (0.040) (0.041) (0.027) (0.028)
Apr 2021 × Share Delayed –0.033 –0.030 –0.028 –0.069 –0.053 –0.070 –0.055 –0.017 –0.015
(0.029) (0.029) (0.029) (0.050) (0.050) (0.050) (0.050) (0.035) (0.035)
May 2021 × Share Delayed –0.013 –0.009 –0.009 –0.027 –0.012 –0.026 –0.011 –0.003 –0.003
(0.026) (0.026) (0.026) (0.049) (0.048) (0.049) (0.048) (0.031) (0.031)
Jun 2021 × Share Delayed 0.003 0.005 0.005 0.006 0.012 0.006 0.011 0.009 0.009
(0.023) (0.023) (0.023) (0.040) (0.039) (0.040) (0.039) (0.028) (0.028)
Jul 2021 × Share Delayed 0.002 0.003 0.003 0.019 0.024 0.019 0.023 –0.001 –0.001
(0.024) (0.024) (0.024) (0.037) (0.037) (0.038) (0.037) (0.029) (0.029)
Aug 2021 × Share Delayed 0.017 0.019 0.018 0.020 0.028 0.021 0.027 0.017 0.017
(0.023) (0.023) (0.023) (0.041) (0.040) (0.041) (0.041) (0.028) (0.028)
Sept 2021 × Share Delayed 0.034 0.035 0.035 0.039 0.046 0.039 0.046 0.033 0.033
(0.025) (0.025) (0.025) (0.042) (0.042) (0.043) (0.042) (0.030) (0.030)
Average Private Employment –0.04 –0.06 –0.06 –0.03
St. Dev. of Private Employment 0.22 0.24 0.24 0.21
Within R-squared 0.13 0.16 0.16 0.21 0.28 0.21 0.27 0.07 0.10
Observations 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes No Yes
CARES Act Controls No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the
top 1 percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation
(2). Preexisting Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013
NCHS urban-rural designation, population, community-bank share of deposits, largest four banks’ share of deposits, bank branch
density, and 2019 small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and
deaths per population, share of days in April 2020 in early lockdown, share of employment in essential industries, and share of
employment in most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits replacement
rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
52
Table A.10: Effects of Share of PPP Loans Delayed on QCEW County Private Employ-
ment in NAICS 71, 72, and 81: Non-Industry-Specific Share of PPP Loans Delayed
All Counties Urban Smaller Rural
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Jan 2020 × Share Delayed 41∗∗ 71 67 67 215∗∗ 179 127 77 –0 2
(21) (46) (46) (46) (93) (117) (82) (103) (4) (4)
Feb 2020 × Share Delayed 50∗∗∗ 80∗ 79∗ 79∗ 220∗∗∗ 195∗ 139∗ 106 –4 –2
(18) (42) (42) (42) (81) (102) (72) (96) (3) (4)
Apr 2020 × Share Delayed –35 14 62 60 –383 –102 –344 –100 –14 –8
(91) (69) (70) (70) (409) (299) (347) (266) (18) (18)
May 2020 × Share Delayed –181 –122 –76 –75 –551 –152 –643 –277 –23 –18
(112) (92) (92) (92) (504) (400) (457) (385) (20) (20)
Jun 2020 × Share Delayed –361∗∗ –262∗∗ –218∗∗ –213∗ –1,147 –490 –1,142 –550 –20 –13
(156) (114) (110) (110) (719) (505) (695) (499) (17) (17)
Jul 2020 × Share Delayed –440∗∗∗ –356∗∗∗ –311∗∗∗ –303∗∗∗ –1,000 –298 –1,045 –422 –29∗ –25
(157) (116) (112) (112) (718) (495) (746) (508) (15) (15)
Aug 2020 × Share Delayed –438∗∗∗ –366∗∗∗ –339∗∗∗ –329∗∗∗ –817 –175 –932 –359 –31∗∗ –28∗
(152) (119) (116) (116) (687) (498) (717) (499) (15) (15)
Sept 2020 × Share Delayed –375∗∗∗ –313∗∗∗ –283∗∗∗ –275∗∗∗ –722 –142 –842 –314 –25∗ –22
(139) (107) (104) (104) (631) (448) (656) (456) (14) (14)
Oct 2020 × Share Delayed –284∗∗ –223∗∗ –189∗∗ –183∗∗ –480 26 –607 –139 –30∗∗∗ –25∗∗
(121) (91) (89) (89) (556) (396) (572) (399) (12) (11)
Nov 2020 × Share Delayed –237∗∗ –185∗∗ –150∗ –143 –503 9 –554 –73 –23∗ –18
(121) (91) (88) (88) (552) (390) (570) (386) (12) (12)
Dec 2020 × Share Delayed –228∗ –177∗ –144 –132 –254 264 –354 103 –24∗ –21∗
(124) (103) (102) (101) (553) (446) (564) (407) (13) (13)
Jan 2021 × Share Delayed –382∗∗∗ –320∗∗∗ –287∗∗ –271∗∗ –607 27 –694 –175 –38∗∗∗ –35∗∗
(146) (121) (119) (119) (641) (503) (666) (461) (14) (14)
Feb 2021 × Share Delayed –316∗∗ –255∗∗ –221∗∗ –234∗∗ –638 –156 –762 –344 –35∗∗∗ –33∗∗
(142) (106) (102) (102) (638) (434) (662) (442) (13) (13)
Mar 2021 × Share Delayed –303∗∗ –243∗∗ –209∗∗ –230∗∗ –623 –149 –762 –348 –35∗∗∗ –34∗∗∗
(141) (102) (99) (99) (640) (427) (661) (435) (13) (13)
Apr 2021 × Share Delayed –209 –142 –109 –135 –299 115 –470 –104 –16 –23
(135) (96) (94) (92) (637) (434) (629) (425) (20) (21)
May 2021 × Share Delayed –196 –126 –93 –121 –256 140 –388 –33 –4 –12
(128) (92) (91) (89) (612) (414) (610) (406) (19) (20)
Jun 2021 × Share Delayed –181 –122 –88 –116 –76 280 –185 130 –11 –21
(117) (90) (90) (88) (563) (414) (550) (390) (16) (18)
Jul 2021 × Share Delayed –161 –96 –65 –82 –140 199 –224 68 –13 –21
(103) (79) (80) (80) (495) (361) (490) (354) (14) (16)
Aug 2021 × Share Delayed –139 –76 –61 –74 –152 244 –228 114 –13 –21
(96) (73) (76) (76) (461) (343) (451) (332) (14) (15)
Sept 2021 × Share Delayed –63 –14 –2 –9 27 221 –93 52 –13 –19
(77) (60) (61) (61) (361) (267) (332) (255) (12) (13)
Average Private Employment 5,666 13,272 11,221 939
St. Dev. of Private Employment 20,173 31,065 20,276 1,451
Within R-squared 0.89 0.89 0.89 0.89 0.89 0.90 0.85 0.86 0.75 0.76
Observations 61,173 61,173 61,173 61,173 23,436 23,436 23,184 23,184 37,716 37,716
County and State by Mth FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes Yes No Yes No Yes No Yes
CARES Act Controls No No Yes Yes No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No No Yes No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those in the top
1 percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2).
Preexisting Conditions Controls: median family income, commuter-to-residential-population ratio, indicators for 2013 NCHS
urban-rural designation, population, community-bank share of deposits, largest four banks’ share of deposits, bank branch density,
and 2019 small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-19 cases and deaths per
population, share of days in April 2020 in early lockdown, share of employment in essential industries, and share of employment in
most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits replacement rate, and rebates
(“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
53
Table A.11: Effects of Share of PPP Loans Delayed on the Number of Establishments by
Size and Industry
All Estabs with up to NAICS Imp
Estabs 5 50 500 71, 72, 81 Ind
All Counties
2020Q2 × Share Delayed –10.71∗∗ –6.18∗ –10.12∗∗ –10.66∗ –0.95 –1.54
(5.46) (3.24) (5.16) (5.44) (1.37) (1.54)
2020Q3 × Share Delayed –12.37 –6.67 –11.60 –12.30 –2.78 –4.38∗∗
(8.38) (4.82) (7.89) (8.35) (2.18) (2.21)
2020Q4 × Share Delayed –8.46 –5.12 –7.96 –8.39 –1.90 –4.13
(17.18) (9.69) (16.27) (17.14) (3.01) (2.71)
2021Q1 × Share Delayed –18.96 –11.31 –17.91 –18.89 –3.63 –7.42∗
(14.50) (8.39) (13.67) (14.45) (3.59) (3.85)
2021Q2 × Share Delayed –22.43 –13.45 –21.21 –22.35 –4.39 –8.38∗∗
(16.11) (9.27) (15.21) (16.06) (4.14) (4.02)
2021Q3 × Share Delayed –22.19 –13.84 –21.06 –22.12 –1.94 –6.52∗
(16.12) (9.19) (15.21) (16.07) (4.05) (3.58)
Average Nu Private Estabs 3,202 1,754 3,026 3,193 563 563
St. Dev. of Private Estabs 13,252 7,864 12,547 13,218 2,107 2,107
Within R-squared 0.64 0.66 0.64 0.64 0.36 0.50
Observations 20,391 20,391 20,391 20,391 20,391 20,391
Notes: “Estabs with up to 5 (50 or 500)”: number of establishments with fewer than 5, 50,
or 500 employees. “Imp Ind”: number of establishments in the impacted industries 44, 61,
62, 71, 72, and 81. Regressions include the same controls as column (4) of Table 3.
54
Table A.12: Determinants of Lending Delay, April 2020. CBSA Sample
Share of Volume Delay Share of Number Delay
(All) (Smaller) (All) (Smaller)
Cum. COVID-19 Cases per billion up to 4/15/2020 –0.045 –0.045 –0.022 –0.026
(0.043) (0.044) (0.024) (0.023)
Cum. COVID-19 Deaths per billion up to 4/15/2020 0.884∗ 0.818 0.660∗∗ 1.123∗∗
(0.522) (1.101) (0.278) (0.440)
Share of days in lockdown (pre-4/17/2020) –0.049 –0.051 –0.000 –0.011
(0.074) (0.079) (0.046) (0.047)
Share of days in lockdown (4/17–4/30/2020) 0.024 0.024 0.007 0.010
(0.112) (0.112) (0.074) (0.075)
Share of Emp. in Essential Industries –0.140 –0.144 –0.159 –0.150
(0.272) (0.267) (0.205) (0.202)
Share of Wages in Impacted Industries –0.114 –0.116 –0.200∗∗ –0.196∗∗
(0.092) (0.092) (0.087) (0.086)
UI Benefits Replacement Rate (Industry-Wtd.) 0.053 0.054 0.050 0.050
(0.056) (0.057) (0.042) (0.041)
Most Populouse CBSA (Top 1%) 0.081∗∗∗ 0.038∗∗
(0.019) (0.017)
Ln Residential Population –0.013∗∗ –0.013∗∗ –0.015∗∗∗ –0.015∗∗∗
(0.005) (0.005) (0.004) (0.004)
Commuter to Residential Population Ratio 0.064 0.062 0.147∗∗ 0.147∗∗
(0.080) (0.081) (0.060) (0.061)
Ln Median Family Income 0.028 0.028 0.028 0.028
(0.055) (0.056) (0.033) (0.034)
Community Bank Share of Deposits –0.016 –0.016 –0.005 –0.005
(0.023) (0.023) (0.023) (0.023)
Big4 Bank Share of Deposits –0.153 –0.153 –0.112 –0.114
(0.131) (0.131) (0.081) (0.081)
Ln Bank Branch Density –0.013 –0.013 –0.013 –0.012
(0.020) (0.020) (0.012) (0.012)
SBL Volume per Small Estab. (< 500Emp.) (CBP 2019Q1) –0.003∗∗∗ –0.003∗∗∗ –0.001∗∗∗ –0.001∗∗∗
(0.001) (0.001) (0.000) (0.000)
Ratio of Small Employment in 2020Q1 to 2019Q1, QWI 0.189∗ 0.189∗ 0.074 0.073
(0.109) (0.109) (0.072) (0.072)
Constant 0.305 0.309 0.459 0.445
(0.788) (0.798) (0.420) (0.424)
Adjusted R-squared 0.14 0.14 0.32 0.33
Observations 874 865 874 865
State FE Yes Yes Yes Yes
Notes: “Smaller” refers to CBSAs excluding those in the top 1 percent by population.
55
Table A.13: Effects of Share of PPP Loans Delayed on QCEW Private Employment:
CBSA Regressions
Total NAICS 71, 72, and 81 Industries
All Smaller All Smaller
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Jan 2020 × Share Delayed 336 775 721 109 24 32 250 415 395 –85 –99 –98
(231) (1,170) (1,121) (161) (235) (229) (242) (367) (352) (125) (137) (135)
Feb 2020 × Share Delayed 312 780 705 281∗∗ 193 192 300 424 412 –4 –28 –26
(300) (1,147) (1,095) (129) (207) (199) (200) (333) (320) (90) (102) (101)
Apr 2020 × Share Delayed –2,415 –872 –77 338 –77 253 389 1,562∗∗∗ 1,626∗∗∗ 956∗∗ 1,006∗∗ 1,044∗∗
(3,312) (2,566) (2,532) (1,870) (2,018) (1,993) (876) (588) (585) (475) (504) (494)
May 2020 × Share Delayed –3,571 –3,801 –2,445 –692 –1,401 –902 696 1,446∗ 1,646∗∗ 601 521 655
(3,348) (2,842) (2,726) (1,782) (2,014) (2,002) (982) (864) (835) (707) (752) (721)
Jun 2020 × Share Delayed –4,828 –5,386∗ –3,593 –821 –1,112 –658 –947 224 591 –67 –88 69
(3,530) (3,063) (2,811) (1,232) (1,267) (1,309) (1,452) (1,127) (1,071) (793) (824) (787)
Jul 2020 × Share Delayed –4,432 –4,546∗ –2,923 –1,293 –1,419 –978 –945 –400 –129 –398 –365 –260
(2,912) (2,448) (2,292) (1,069) (1,096) (1,138) (1,008) (901) (851) (676) (694) (669)
Aug 2020 × Share Delayed –3,918 –3,721∗ –2,254 –1,194 –1,371 –967 –537 –327 –55 –384 –385 –263
(2,633) (2,155) (2,095) (1,038) (1,064) (1,111) (906) (906) (860) (678) (701) (676)
Sept 2020 × Share Delayed –3,264 –2,905 –1,687 –1,068 –1,308 –967 –234 –78 154 –272 –272 –160
(2,186) (1,812) (1,817) (960) (972) (1,015) (779) (796) (755) (617) (643) (618)
Oct 2020 × Share Delayed –2,340 –1,952 –1,022 –302 –573 –309 –98 117 286 –241 –235 –147
(1,964) (1,583) (1,570) (933) (931) (960) (655) (670) (641) (527) (548) (529)
Nov 2020 × Share Delayed –2,214 –1,544 –672 224 68 332 19 342 505 –60 –21 63
(1,950) (1,530) (1,487) (986) (981) (989) (617) (617) (593) (459) (471) (458)
Dec 2020 × Share Delayed –1,418 –851 191 417 216 626 950 1,305 1,509∗ 115 242 361
(1,873) (1,571) (1,581) (1,105) (1,094) (1,109) (760) (810) (793) (588) (592) (582)
Jan 2021 × Share Delayed –2,250 –1,145 272 266 167 553 615 1,086 1,417 39 147 267
(2,314) (1,848) (1,826) (1,114) (1,107) (1,127) (951) (961) (951) (618) (618) (609)
Feb 2021 × Share Delayed –2,375 –1,142 2 805 631 853 87 623 840 6 86 151
(2,464) (1,767) (1,676) (1,066) (1,049) (1,051) (863) (797) (775) (527) (532) (519)
Mar 2021 × Share Delayed –2,323 –1,114 –87 532 401 634 –115 427 611 –125 –50 –2
(2,322) (1,671) (1,604) (1,019) (1,000) (1,008) (868) (771) (744) (510) (513) (500)
Apr 2021 × Share Delayed –2,025 –1,282 –621 460 260 406 1 456 550 –249 –177 –155
(2,068) (1,435) (1,379) (896) (879) (878) (842) (717) (688) (500) (502) (492)
May 2021 × Share Delayed –2,018 –1,330 –837 474 276 384 –21 342 338 –278 –201 –237
(2,007) (1,391) (1,328) (889) (876) (874) (753) (670) (640) (465) (461) (456)
Jun 2021 × Share Delayed –1,475 –1,111 –758 350 136 214 53 350 287 –288 –194 –258
(1,757) (1,381) (1,347) (867) (871) (875) (681) (670) (637) (478) (476) (470)
Jul 2021 × Share Delayed –2,060 –1,315 –1,128 475 295 382 95 329 240 –80 –5 –73
(2,075) (1,468) (1,416) (931) (934) (930) (584) (589) (567) (458) (458) (450)
Aug 2021 × Share Delayed –2,274 –1,226 –938 696 616 696 –40 173 159 –53 3 –26
(2,238) (1,511) (1,440) (918) (912) (910) (536) (536) (528) (419) (420) (419)
Sept 2021 × Share Delayed –1,273 –295 –41 1,123 1,099 1,122 165 367 306 82 119 79
(1,921) (1,372) (1,324) (907) (899) (903) (448) (445) (432) (337) (334) (331)
Average Private Employment 123,224 91,460 18,021 13,622
St. Dev. of Private Employment 430,927 223,528 60,322 32,258
Within R-squared 0.92 0.93 0.93 0.86 0.86 0.86 0.93 0.94 0.94 0.89 0.89 0.89
Observations 18,333 18,333 18,333 18,165 18,165 18,165 18,333 18,333 18,333 18,165 18,165 18,165
CBSA and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes No Yes No Yes Yes
CARES Act Controls No No Yes No Yes No Yes No Yes No No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to CBSAs excluding those in the top 1 percent by population.
Share Delayed is the share (by volume) of county-level PPP loans delayed as defined in Equation (2). Preexisting Conditions Controls: me-
dian family income, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation, population, community-bank share
of deposits, largest four banks’ share of deposits, bank branch density, and 2019 small-business-loan volume per small establishment; COVID-19
Controls: cumulative COVID-19 cases and deaths per population, share of days in April 2020 in early lockdown, share of employment in essential
industries, and share of employment in most impacted industries; CARES Act Controls: industry-employment-share-weighted UI benefits replace-
ment rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
56
Table A.14: Determinants of Lending Delay, April 2020. CPS Sample
Share of Volume Delay Share of Number Delay
(All) (Urban) (Smaller) (All) (Urban) (Smaller)
Cum. COVID-19 Cases per billion up to 4/15/2020 0.046∗ 0.048∗ 0.047∗ 0.051∗ 0.053∗∗ 0.053∗∗
(0.026) (0.025) (0.025) (0.025) (0.024) (0.024)
Cum. COVID-19 Deaths per billion up to 4/15/2020 0.143∗∗∗ 0.139∗∗∗ 0.139∗∗∗ 0.098∗∗∗ 0.094∗∗∗ 0.093∗∗∗
(0.027) (0.029) (0.029) (0.026) (0.027) (0.027)
Share of days in lockdown (pre-4/17/2020) 0.017 0.019 0.021 -0.054 -0.050 -0.052
(0.154) (0.155) (0.156) (0.065) (0.065) (0.067)
Share of days in lockdown (4/17–4/30/2020) -0.725∗∗∗ -0.746∗∗∗ -0.743∗∗∗ -0.245 -0.284 -0.289
(0.198) (0.209) (0.210) (0.279) (0.298) (0.293)
Share of Emp. in Essential Industries -0.382 -0.330 -0.329 0.015 0.063 0.060
(0.574) (0.559) (0.557) (0.555) (0.544) (0.545)
Share of Wages in Impacted Industries 0.024 -0.005 -0.010 -0.014 -0.052 -0.049
(0.162) (0.167) (0.168) (0.137) (0.137) (0.140)
UI Benefits Replacement Rate (Industry-Wtd.) 0.047 0.050 0.050 0.062 0.071 0.071
(0.048) (0.049) (0.049) (0.044) (0.046) (0.047)
Rural County Dummy -0.015 -0.030
(0.032) (0.031)
Ln Residential Population -0.001 -0.001 -0.001 -0.002 -0.002 -0.001
(0.013) (0.013) (0.013) (0.009) (0.009) (0.011)
Commuter to Residential Population Ratio -0.003 -0.004 -0.004 0.014 0.014 0.014
(0.053) (0.053) (0.053) (0.042) (0.041) (0.041)
Ln Median Family Income 0.043 0.041 0.042 0.014 0.012 0.011
(0.035) (0.034) (0.034) (0.025) (0.025) (0.025)
Community Bank Share of Deposits 0.053 0.045 0.045 0.048 0.039 0.039
(0.044) (0.045) (0.045) (0.037) (0.038) (0.038)
Big4 Bank Share of Deposits 0.533∗ 0.542∗ 0.533∗ 0.425∗∗ 0.413∗ 0.416∗
(0.280) (0.290) (0.292) (0.194) (0.207) (0.209)
Ln Bank Branch Density 0.060∗∗ 0.058∗∗ 0.059∗∗ 0.017 0.015 0.015
(0.027) (0.028) (0.028) (0.018) (0.019) (0.020)
SBL Volume per Small Estab. (< 500Emp.) (CBP 2019Q1) -0.001 -0.001 -0.001 -0.000 -0.000 -0.000
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Proportion of Small Employment in 2020Q1 to 2019Q1, QWI 0.002 0.002 0.002 0.001 0.000 0.000
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Proportion of Small Estabs in 2020Q1 to 2019Q1, CBP -0.001 -0.001 -0.001 0.004 0.005 0.005
(0.008) (0.008) (0.008) (0.004) (0.004) (0.004)
Constant 0.431 0.392 0.377 -0.059 -0.082 -0.081
(1.365) (1.333) (1.338) (1.096) (1.086) (1.088)
Adjusted R-squared 0.45 0.44 0.44 0.51 0.50 0.50
Observations 279 273 270 279 273 270
State FE Yes Yes Yes Yes Yes Yes
Notes: “Smaller” refers to counties excluding those in the top 1 percent by population.
57
Table A.15: Effects of Share of PPP Loans Delayed on QCEW Private Employment:
CPS Sample
All Counties Urban Smaller Urban
(1) (2) (3) (4) (5) (6) (7)
Jan 2020 × Share Delayed 496 –98 –547 518 –607 461 –585
(994) (1,575) (1,517) (1,016) (1,535) (933) (1,525)
Feb 2020 × Share Delayed 1,823 856 485 1,679 338 2,134∗∗∗ 1,513
(1,499) (1,759) (1,666) (1,543) (1,708) (769) (1,235)
Apr 2020 × Share Delayed –15,460∗ –14,243 –13,329 –14,715 –13,120 –11,185 –9,456
(8,855) (9,176) (9,156) (9,052) (9,314) (7,883) (8,258)
May 2020 × Share Delayed –17,560∗ –12,118 –9,384 –17,263 –10,023 –11,838 –5,159
(10,385) (10,114) (9,863) (10,633) (9,928) (8,338) (8,217)
Jun 2020 × Share Delayed –22,714∗ –14,821 –13,336 –23,083∗ –14,285 –18,002∗ –9,322
(12,461) (11,034) (10,890) (12,800) (11,053) (10,356) (9,195)
Jul 2020 × Share Delayed –23,085∗∗ –15,619 –14,692 –23,392∗∗ –15,500 –19,831∗∗ –11,637
(11,217) (9,473) (9,466) (11,551) (9,672) (9,816) (8,257)
Aug 2020 × Share Delayed –22,218∗∗ –14,122 –13,400 –22,597∗∗ –14,383 –18,971∗∗ –10,478
(10,961) (9,068) (9,076) (11,292) (9,309) (9,630) (7,893)
Sept 2020 × Share Delayed –20,097∗ –11,979 –11,282 –20,441∗ –12,260 –16,743∗ –8,369
(10,688) (8,793) (8,757) (11,000) (8,972) (9,314) (7,545)
Oct 2020 × Share Delayed –16,337 –8,969 –7,791 –16,656 –8,599 –12,284 –4,156
(10,602) (8,688) (8,536) (10,917) (8,766) (8,686) (7,053)
Nov 2020 × Share Delayed –16,954 –9,972 –8,416 –17,349 –9,127 –12,241 –4,066
(11,413) (9,165) (8,933) (11,768) (9,226) (8,886) (7,132)
Dec 2020 × Share Delayed –14,750 –7,818 –6,362 –14,790 –6,827 –9,935 –1,838
(11,003) (8,991) (8,859) (11,347) (9,164) (8,741) (7,178)
Jan 2021 × Share Delayed –15,484 –8,082 –7,202 –15,644 –7,952 –11,197 –3,235
(11,414) (9,570) (9,464) (11,765) (9,787) (9,323) (7,876)
Feb 2021 × Share Delayed –16,976 –10,464 –8,843 –17,447 –9,672 –11,502 –3,716
(12,710) (10,357) (10,093) (13,101) (10,449) (9,517) (7,883)
Mar 2021 × Share Delayed –15,776 –9,355 –7,794 –16,164 –8,456 –10,505 –2,741
(12,188) (9,878) (9,647) (12,569) (9,985) (9,172) (7,559)
Apr 2021 × Share Delayed –12,037 –5,918 –4,918 –12,402 –5,514 –7,377 –374
(11,000) (8,792) (8,587) (11,382) (8,918) (8,576) (7,106)
May 2021 × Share Delayed –12,066 –6,587 –5,158 –12,396 –5,549 –7,647 –642
(10,904) (8,625) (8,382) (11,280) (8,701) (8,385) (6,831)
Jun 2021 × Share Delayed –11,692 –6,430 –5,140 –11,693 –5,193 –7,820 –1,151
(8,967) (7,235) (7,081) (9,258) (7,315) (7,293) (5,911)
Jul 2021 × Share Delayed –14,310 –9,988 –8,373 –14,321 –8,378 –10,202 –4,210
(9,470) (7,483) (7,370) (9,798) (7,640) (7,363) (5,942)
Aug 2021 × Share Delayed –15,319 –11,486 –9,430 –15,417 –9,515 –10,703 –4,695
(9,878) (7,819) (7,647) (10,215) (7,924) (7,267) (5,936)
Sept 2021 × Share Delayed –11,219 –7,986 –6,369 –11,293 –6,531 –6,807 –2,204
(8,316) (6,715) (6,540) (8,587) (6,754) (6,106) (5,089)
Average Private Employment 194,470 197,888 175,234
St. Dev. of Private Employment 321,895 324,671 220,779
Within R-squared 0.90 0.91 0.91 0.90 0.91 0.84 0.86
Observations 5,733 5,733 5,733 5,607 5,607 5,544 5,544
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes
CARES Act Controls No No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties excluding those
in the top 1 percent by population. Share Delayed is the share (by volume) of county-level PPP loans delayed as de-
fined in Equation (2). Preexisting Conditions Controls: median family income, commuter-to-residential-population
ratio, indicators for 2013 NCHS urban-rural designation, population, community-bank share of deposits, largest four
banks’ share of deposits, bank branch density, and 2019 small-business-loan volume per small establishment; COVID-
19 Controls: cumulative COVID-19 cases and deaths per population, share of days in April 2020 in early lockdown,
share of employment in essential industries, and share of employment in most impacted industries; CARES Act Con-
trols: industry-employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per capita.
Source: Multiple data sources described in Section 3.2.
58
Table A.16: Effects of the Share of PPP Loans Delayed on QCEW Private Employment:
Non-CPS Sample.
All Counties Urban Smaller Urban
(1) (2) (3) (4) (5) (6) (7)
Jan 2020 × Share Delayed –9 –53 –51 –9 –98 –39 –37
(21) (41) (43) (85) (190) (68) (68)
Feb 2020 × Share Delayed 15 22 22 31 52 –21 –8
(17) (19) (19) (67) (72) (57) (57)
Apr 2020 × Share Delayed 207 192 204 495 342 495 422
(131) (135) (138) (496) (529) (431) (453)
May 2020 × Share Delayed 72 70 82 8 –35 300 281
(150) (157) (159) (623) (639) (455) (473)
Jun 2020 × Share Delayed –40 –63 –49 –175 –238 214 157
(134) (138) (140) (588) (587) (431) (431)
Jul 2020 × Share Delayed –113 –140 –125 –193 –216 248 207
(112) (113) (115) (476) (469) (397) (391)
Aug 2020 × Share Delayed –113 –137 –128 –104 –143 339 281
(107) (109) (111) (451) (449) (376) (370)
Sept 2020 × Share Delayed –119 –145 –134 –143 –194 231 162
(108) (109) (112) (456) (453) (378) (372)
Oct 2020 × Share Delayed –73 –96 –84 –96 –97 228 205
(114) (116) (119) (479) (474) (350) (344)
Nov 2020 × Share Delayed –35 –63 –50 –159 –91 244 258
(133) (135) (139) (567) (557) (340) (337)
Dec 2020 × Share Delayed 1 –37 –25 –20 60 391 413
(143) (145) (148) (611) (617) (389) (392)
Jan 2021 × Share Delayed –25 –63 –52 –79 –22 431 415
(149) (153) (156) (647) (660) (393) (399)
Feb 2021 × Share Delayed 36 –3 –6 82 41 519 487
(145) (147) (151) (610) (617) (378) (383)
Mar 2021 × Share Delayed 16 –23 –32 46 –3 411 407
(135) (137) (142) (582) (581) (362) (362)
Apr 2021 × Share Delayed –88 –114 –133 –66 –142 258 283
(119) (119) (123) (516) (501) (335) (328)
May 2021 × Share Delayed –37 –59 –85 133 25 434 451
(114) (115) (118) (494) (476) (326) (322)
Jun 2021 × Share Delayed 44 30 5 523 438 763∗∗ 797∗∗
(103) (104) (107) (444) (431) (358) (358)
Jul 2021 × Share Delayed 76 67 38 506 457 808∗∗ 880∗∗
(117) (119) (122) (508) (495) (396) (397)
Aug 2021 × Share Delayed 97 91 70 498 442 760∗ 804∗∗
(117) (118) (122) (506) (495) (390) (393)
Sept 2021 × Share Delayed 156 159 137 669 688 786∗∗ 898∗∗
(108) (110) (113) (465) (455) (353) (349)
Average Private Employment 22,307 56,128 45,077
St. Dev. of Private Employment 87,868 149,646 86,914
Within R-squared 0.89 0.89 0.89 0.89 0.90 0.85 0.85
Observations 55,314 55,314 55,314 17,661 17,661 17,472 17,472
County and State-by-Month FE Yes Yes Yes Yes Yes Yes Yes
Lag of Dependent Var Yes Yes Yes Yes Yes Yes Yes
Preexisting Conditions Controls Yes Yes Yes Yes Yes Yes Yes
COVID-19 Controls No Yes Yes No Yes No Yes
CARES Act Controls No No Yes No Yes No Yes
Cum PPP per Emp in Small Estab (t-1) No No Yes No Yes No Yes
Notes: Standard errors clustered at the county level in parentheses. “Smaller” refers to urban counties
excluding those in the top 1 percent by population. Share Delayed is the share (by volume) of county-level
PPP loans delayed as defined in Equation (2). Preexisting Conditions Controls: median family in-
come, commuter-to-residential-population ratio, indicators for 2013 NCHS urban-rural designation, popu-
lation, community-bank share of deposits, largest four banks’ share of deposits, bank branch density, and
2019 small-business-loan volume per small establishment; COVID-19 Controls: cumulative COVID-
19 cases and deaths per population, share of days in April 2020 in early lockdown, share of employment
in essential industries, and share of employment in most impacted industries; CARES Act Controls:
industry-employment-share-weighted UI benefits replacement rate, and rebates (“stimulus checks”) per
capita.
Source: Multiple data sources described in Section 3.2.
59
File and source
- File
- gorbachev-2023-the-tail-that-wagged-the-dog-what-explains-the-persistent-employment-effect.pdf
- Size
- 6,359,391 bytes
- SHA-256
- 8381377752c60c9a4ec01fdd932028e5d83669eb88f6117b2067c70b2c06e116
- Our copy
- gorbachev-2023-the-tail-that-wagged-the-dog-what-explains-the-persistent-employment-effect.pdf
- Original
- www.bostonfed.org