Full text
Journal of Economic Perspectives—Volume 36, Number 2—Spring 2022—Pages 55–80
I
n the early weeks of the COVID-19 pandemic, many small businesses in the
United States were in precarious financial condition: revenues had plunged,
access to credit was in many cases inadequate or absent, and large-scale layoffs
and closures had already occurred (Bartik et al. 2020a, b). The potential conse-
quences of widespread business failure were not confined to business owners. Since
approximately 47 percent of US workers were employed by small businesses prior to
the pandemic (SBA 2019), these closures held the potential for vast job loss. Over
the longer term, widespread firm closures could slow the subsequent economic
recovery by destroying intangible firm capital, liquidating high quality worker-firm
matches, and forcing the costly reallocation of physical capital.
To aid these distressed businesses, Congress enacted the Paycheck Protec-
tion Program (PPP), which provided uncollateralized, low-interest loans of up to
$10 million to firms with fewer than 500 employees—loans that were forgivable
on the condition that recipient firms maintained employment and wages at close
The $800 Billion Paycheck Protection
Program: Where Did the Money Go and
Why Did It Go There?
■
■ David Autor is Professor of Economics, Massachusetts Institute of Technology, and also
David Autor is Professor of Economics, Massachusetts Institute of Technology, and also
a Research Associate, National Bureau of Economic Research, both in Cambridge, Massa-
a Research Associate, National Bureau of Economic Research, both in Cambridge, Massa-
chusetts. David Cho is an Economist, Leland Crane is a Senior Economist, Byron Lutz is
chusetts. David Cho is an Economist, Leland Crane is a Senior Economist, Byron Lutz is
an Assistant Director, Joshua Montes is a Senior Economist, William Peterman is Chief of
an Assistant Director, Joshua Montes is a Senior Economist, William Peterman is Chief of
the Fiscal Analysis Section, David Ratner is a Principal Economist, and Daniel Villar is an
the Fiscal Analysis Section, David Ratner is a Principal Economist, and Daniel Villar is an
Economist, all at the Board of Governors of the Federal Reserve System, Washington, DC.
Economist, all at the Board of Governors of the Federal Reserve System, Washington, DC.
Mita Goldar is an independent researcher. Ahu Yildermaz is CEO of the Coleridge Initiative,
Mita Goldar is an independent researcher. Ahu Yildermaz is CEO of the Coleridge Initiative,
Chevy Chase, Maryland. David Autor is the corresponding author at dautor@mit.edu.
Chevy Chase, Maryland. David Autor is the corresponding author at dautor@mit.edu.
For supplementary materials such as appendices, datasets, and author disclosure statements, see the
article page at https://doi.org/10.1257/jep.36.2.55.
David Autor, David Cho, Leland D. Crane, Mita
Goldar, Byron Lutz, Joshua Montes, William B.
Peterman, David Ratner, Daniel Villar, and Ahu
Yildirmaz
56 Journal of Economic Perspectives
to pre-crisis levels in the two to six months following loan receipt. The scale of the
aid provided was extraordinary. By the time the program concluded in mid-2021,
around $800 billion in loans had been extended. Despite facing initial capacity
constraints, the Paycheck Protection Program was notably successful in distributing
a vast number of loans in short order: the take-up rate among eligible firms was
94 percent. Crucial to this rapid rollout was the decision to enlist the private sector
to oversee the origination of all PPP loans, with the Small Business Administration
(SBA) serving as the guarantor.
The Paycheck Protection Program was ultimately comparable in size to the
two other major federal transfer programs enacted in response to the pandemic:
expenditures on household payments—that is, stimulus checks—were around
$800 billion; and expenditures on expanded unemployment benefits totaled roughly
$680 billion under the Federal Pandemic Unemployment Compensation program
(FPUC), Pandemic Unemployment Assistance program (PUA), and Pandemic
Emergency Unemployment Compensation (PEUC) (CRFB 2021). As another stan-
dard of comparison, each of these three programs was roughly comparable in size
to the entire American Recovery and Reinvestment Act of 2009 (ARRA), the prin-
cipal fiscal stimulus enacted in response to the Great Recession of 2007–2009.
This paper explores who ultimately benefited from those $800 billion in
Paycheck Protection Program loans: concretely, where did the money go and why
did it go there? We provide an answer in three steps. First, we consider how PPP
funds flowed to three proximate sets of actors: workers who otherwise would have
been laid off; creditors and suppliers of PPP-receiving businesses (for example,
landlords, and utilities) who would otherwise not have received payments; and
windfall transfers to PPP-recipient businesses (owners and shareholders) that
would have maintained employment and met other financial obligations absent
the PPP. Second, we calculate how these recipients were distributed across the
household income distribution. Finally, we compare this allocation of funds to the
household incidence of the two other major federal pandemic transfer programs:
unemployment assistance and direct household payments. Our analysis combines
lessons from existing research, including some of our own, and also presents new
analysis using anonymized and aggregated payroll data from the private firm ADP,
which processes payrolls for over 26 million individual workers in the United
States per month.
The Paycheck Protection Program had measurable impacts. It meaningfully
blunted pandemic job losses, preserving somewhere between 1.98 and 3.0 million
job-years of employment during and after the pandemic at a substantial cost of
$169,000 to $258,000 per job-year saved. PPP also reduced the rate of temporary
closures among small firms, though it is less clear whether it reduced permanent
closures. The majority of PPP loan dollars issued in 2020—66 to 77 percent—
did not go to paychecks, however, but instead accrued to business owners and
shareholders. And because business ownership and share-holding are concen-
trated among high-income households, the incidence of the program across the
household income distribution was highly regressive. We estimate that about
Autor et al. 57
three-quarters of PPP benefits accrued to the top quintile of household income.
By comparison, the incidence of federal pandemic unemployment insurance and
household stimulus payments was far more equally distributed.
Ironically, the program feature that arguably made the Paycheck Protection
Program’s meteoric scale-up possible is also the feature that made it potentially the
most problematic: the program was essentially untargeted, aside from excluding
firms with more than 500 workers (a rule further relaxed for some sectors). Small
firms merely needed to attest that they were “substantially affected by COVID-19”
to qualify, and almost all did so. Evidence strongly suggests that the program did
not ultimately differentiate among firms or geographic areas according to need.
This near absence of targeting virtually guaranteed that a large fraction of the
first two tranches of $525 billion in PPP loan dollars went to businesses that would
have remained viable and retained their employees even absent PPP. Perhaps
recognizing this program limitation, Congress explicitly targeted the final tranche
($285 billion) of PPP loans in 2021 toward firms that had experienced revenue
losses.
The Paycheck Protection Program’s meteoric scale-up, its lack of targeting,
and its highly regressive incidence reflect a key tradeoff that policymakers faced
in March 2020 when crafting an emergency pandemic business loan program
under severe time constraints: a lack of existing administrative infrastructure for
overseeing large-scale, targeted federal support to US small businesses. Congress
accordingly authorized the Small Business Administration (SBA) to harness the
private sector to originate forgivable PPP loans and stipulated only a few coarse limi-
tations on which firms could receive loans. These decisions rapidly opened the PPP
floodgates to essentially all firms with fewer than 500 employees. Had policymakers
instead insisted on better targeting, this would have likely substantially slowed aid
delivery and reduced program efficacy. A key takeaway from the PPP experience is
that building administrative capacity now would enable greatly improved targeting
of either employment or business liquidity when the next pandemic or other large-
scale economic emergency occurs, as it surely will.
The Basics
The Basics
The Paycheck Protection Program sought to issue forgivable loans to small firms
facing financial distress.1 Businesses were permitted to draw PPP loans worth up to
ten weeks of payroll costs—including wage and salary compensation not to exceed
$100,000 per worker, as well as paid leave, health insurance costs, other benefit
costs, and state and local taxes—with a maximum loan size of $10 million dollars.
1 The Paycheck Protection Program was one of four large government direct-lending programs
introduced during the pandemic; the other three programs were the Main Street Lending Program,
Corporate Credit Facilities, and Municipal Liquidity Facility. These programs jointly covered a large
swath of the US economy (Decker et al. 2021).
58 Journal of Economic Perspectives
Although the Small Business Administration issued the loan guarantees and would
ultimately determine whether loans would be forgiven, PPP loans were processed
and delivered through the nation’s banking system.
The program received three tranches of funding. The Coronavirus Aid, Relief,
and Economic Security Act of 2020 (CARES) established the Paycheck Protection
Program and provided $350 billion in appropriations on March 27, 2020. Subse-
quently, the Paycheck Protection Program and Health Care Enhancement Act,which
passed on April 24, 2020, provided an additional $320 billion in appropriations. A
third tranche of $285 billion was signed in to law on December 27, 2020, as part
of the Consolidated Appropriations Act of 2021. Finally, early on in the pandemic,
the Federal Reserve introduced the Paycheck Protection Program Liquidity Facility
(PPLF) to bolster the ability of the banking system to provide PPP loans (Anbil et al.
2021).
Loans from the first two tranches were issued in 2020 and available to firms
meeting the definition of a small business in the Paycheck Protection Program. In
most industries, but not all, this required having fewer than 500 employees. The third
tranche provided loans to firms in 2021 that had not previously taken out a PPP loan.
It also provided “second draw” loans for firms that had already taken out a PPP loan,
had fewer than 300 employees, and had experienced a significant revenue loss in
2020. About 75 percent of the third tranche of funding went to second-draw loans.
While the moniker Paycheck Protection Program suggests that the program
was focused solely on employment, the criteria for loan forgiveness reveal another
complementary goal: providing firms with liquidity to meet non-compensation
obligations to creditors (like suppliers, banks, and landlords). Businesses had to
do four things to qualify for PPP loan forgiveness: 1) spend at least 60 percent of
the loan amount on payroll expenses; 2) spend (at least) the full loan amount
on total qualifying expenses, including payroll, utilities, rent, and mortgage
payments; 3) maintain average full-time equivalent employment at its pre-crisis
level; and 4) maintain employee wages at no lower than 75 percent of their pre-
crisis level. These criteria applied to a “covered period” that started on the date of
loan disbursal and ran for 8 to 24 weeks, with the interval at the firm’s discretion.
If these criteria were not met, the Small Business Administration offered
alternative routes to forgiveness. Businesses could exercise a “safe harbor” option
to meet the employment and wage criteria by restoring their full-time equivalent
employment and wage rates to their pre-COVID level by the end of 2020 (or by
the end of the covered period for loans issued in 2021). A second safe harbor
absolved firms of the need to restore full pre-pandemic employment levels if they
could document in “good faith” that other pandemic provisions (for example,
lockdowns) made it infeasible to return to full business activity. These safe harbor
provisions made the employment criteria far less onerous. Moreover, firms that
did not meet all criteria could also receive partial loan forgiveness. As of late 2021,
94 percent of PPP loans issued in 2020 had applied for forgiveness and virtually all
such applications had been approved by the Small Business Administration (SBA
2021).
The $800 Billion Paycheck Protection Program 59
One reason that almost all firms were able to meet these criteria is that they
were retroactively loosened in June 2020, well after most PPP loans were issued
(the discussion above pertains to the revised rules). Adding to the windfall,
Congress amended the tax treatment of PPP loans in January 2021 to enable busi-
nesses to claim deductions for expenses paid with PPP loans (for example, wages,
rent, utilities, etc.) without treating PPP loans as taxable business revenue. This
retroactive change, which cost the Treasury an estimated $100 billion in fore-
gone tax revenue, effectively allowed some firms to pay a negative tax rate on
PPP income (Harney and Mott 2021). For simplicity, our primary distributional
accounting exercise below does not adjust for the additional tax subsidy provided
to PPP recipients through this provision, though we briefly estimate its distribu-
tional implications—which are highly regressive.
A Timeliness versus Targeting Tradeoff
A Timeliness versus Targeting Tradeoff
Fiscal interventions during economic downturns are often judged based on
whether they are targeted, timely, and temporary (Elmendorf and Furman 2008).
The Paycheck Protection Program was clearly temporary. How did it do on the other
two T’s?
Timeliness
Timeliness
The program deserves high marks for timeliness. When the pandemic began,
no existing federal program had the scale to quickly distribute hundreds of billions
of dollars to small businesses. The only other possible mechanism seemed to be
state unemployment insurance systems (Bernstein and Rothstein 2020), but these
systems struggled to handle the flood of initial unemployment insurance claims,
and struggled further when tasked with distributing the enhanced unemployment
benefits provided by the Coronavirus Aid, Relief, and Economic Security Act of
2020. It seems unlikely that state unemployment insurance systems could have
handled an additional novel burden (Hubbard and Strain 2020).
Despite these obstacles, the Paycheck Protection Program succeeded in deliv-
ering a staggering sum of money over a two-month period in spring 2020. This can
be seen in Table 1. As shown in column 3 of panel A, $505 billion in first draw loans
were issued to firms with fewer than 500 employees (column 3) and all but 7 percent
of these were issued in 2020 (column 6). A very large share of these loans were issued
in April and May (not shown). Finally, the memo lines show that non-employer busi-
nesses—for example, the self-employed—received $43 billion in first draw loans and
employers with more than 500 employees received a relatively small $18 billion.
One emblem of the Paycheck Protection Program’s success is its market pene-
tration, which we define as the employment-weighted share of firms that received
PPP loans and will refer to as the takeup rate. We make use of loan-level data from
the PPP on the size of each firm that received a PPP loan, along with publicly
available employment data from the Census Bureau’s Statistics of US Businesses
60 Journal of Economic Perspectives
(SUSB). SUSB data provide total employment for a number of categories of firm
size which we use to form the rows of Table 1. For each size category, the takeup
rate is the ratio of the total number of employees at PPP-receiving firms from the
PPP loan data divided by total employment from SUSB. For example, in the PPP
loan data, in the size bin 10–49, there were 1.3 million first-draw loans to firms
with a total of 21.4 million employees over 2020 and 2021. In the aggregate, the
SUSB data from 2018 (the latest available) report that there were 21.4 million
employees in firms with between 10 and 49 workers; accounting for the growth
Table 1
PPP Loans by Employer Size
Employer
size
Employment
share
Loan
$ (billions)
Share of $
Take-up rate
% of $ received
in 2020
(1)
(2)
(3)
(4)
(5)
(6)
A. First draw loans
1–4
10%
44
9%
81%
84%
5–9
11%
54
11%
98%
93%
10–49
35%
182
36%
99%
96%
50–149
23%
122
24%
97%
98%
150–299
13%
64
13%
91%
98%
300–499
9%
40
8%
87%
97%
1–499
100%
505
100%
94%
93%
Memo:
Non-employers
--
43
--
--
25%
Employers 500+
--
18
--
--
93%
B. Second draw loans
1–4
10%
17
9%
30%
--
5–9
11%
25
13%
43%
--
10–49
35%
87
46%
45%
--
50–149
23%
45
24%
34%
--
150–299
13%
14
8%
29%
--
300–499
9%
1
0%
3%
--
1–499
100%
189
100%
34%
--
Memo:
Non-employers
--
11
--
--
--
Employers 500+
--
0.2
--
--
--
Source: Authors’ analysis of Census Bureau Statistics of US Businesses (SUSB) 2018, BLS BED, and SBA
PPP data.
Note: Panels A and B reflect data on employer businesses. The main panels exclude loans to the self-
employed, sole proprietors, independent contractors, and single-member LLCs with only one reported
job because non-employers are excluded from the SUSB data used to calculate the denominator of the
takeup rates displayed in column 5. The roughly 4.6 million non-employer loans (constituting about 8
percent of total loan dollars) are reported in the first memo lines of each panel. As PPP loan-level data
censor firm size at 500, in the main panels of the table we restrict attention to loans to businesses smaller
than 500; loans to businesses reported as having 500 employees in the PPP loan-level data are reported in
the second memo line of each panel. Loans to businesses in Guam, Puerto Rico, and the Virgin Islands
are excluded. Loans to businesses in the following NAICS industries are excluded as they are out of
scope for the SUSB data used in columns 2 and 5: 111, 112, 482, 491, 525110, 525120, 525190, 525920,
541120, 814, and 92.
Autor et al. 61
of employment between 2018 and before the pandemic, aggregate employment
between 10 and 49 was 21.7 million. Thus, the takeup rate in this group is 99 percent
(21.4 million/21.7 million), as given in column 5. We note that these estimated
takeup rates are constrained by significant data limitations in determining the set of
firms eligible for a PPP loan, inaccuracies in the reporting of firm size in PPP loan-
receipt data, the possibility of fraudulent loans, and other measurement issues. (See
the online Appendix, available with this article at the JEP website, for further details
on the methodology underlying Table 1, as well as additional information on the
subsequent analysis in this paper.)
Overall, we estimate that 94 percent of employers with fewer than 500 employees
took up a Paycheck Protection Program loan; consistent with this high takeup
rate, the distribution of loan dollars is tightly in line with employment shares—
compare columns 2 and 4. Indeed, the fact that the second tranche of PPP funding
concluded without exhausting all available funds suggests that the program had
achieved something close to saturation in its first five months of operation. While
near-universal participation in a government program is not altogether surprising
since the program in most cases constituted a pure cash transfer, it is nevertheless a
substantial administrative accomplishment: merely handing out $500 billion dollars
in two months takes many hands. As noted above, this accomplishment would likely
have been infeasible had Congress not authorized the Small Business Administra-
tion to enlist the private banking sector to issue PPP loans.
The early rollout of the Paycheck Protection Program in April and May
2020 did, however, stumble on two hurdles. First, initial demand for loans signifi-
cantly exceeded the ability of banks to deliver them. In the face of these capacity
constraints, banks appear to have prioritized firms with which they had a preexisting
relationship (Amiram and Rabetti 2020; Cororaton and Rosen 2021; Joaquim and
Netto 2021; Granja et al. 2020; Li and Strahan, 2020). Larger firms, which tend to
have ongoing banking relationships, accessed PPP funds sooner than smaller firms
on average. Moreover, as most small business lending is sourced from local banks
(Brevoort et al. 2010), the aptitude and willingness of local banks to process loan
applications generated significant geographic heterogeneity in the initial distribu-
tion of loans (Bartik et al. 2021; Li and Strahan 2020).
The second hurdle was the significant uncertainty and confusion among busi-
nesses and banks over the specifics of the program, particularly over whether the
loans would be forgiven. For example, in April 2020, the Small Business Admin-
istration announced that publicly traded companies were unlikely to satisfy the
required good faith certification of need for a loan from the Paycheck Protection
Program and stipulated a time window in which firms could return loans. Simul-
taneously, the Treasury Department announced that loans in excess of $2 million
would be subject to review and warned of possible criminal charges for those who
failed the review. These issues were resolved over the course of several months. By
the second round of funding, confusion about eligibility and forgiveness terms had
abated. Meanwhile, initially underperforming banks upped their loan tempo, and
non-banks stepped into fill gaps in local loan provision (Granja et al. 2020; Erel
62 Journal of Economic Perspectives
and Liebersohn 2020). By July 2020, virtually all firms that would access a PPP loan
in 2020 had done so.
The delay in delivering funds in April and May 2020 had real consequences.
Doniger and Kay (2021) and Kurmann et al.(2021) find that loans received even a
little earlier had a more pronounced effect on employment than those issued a bit
later. Meanwhile, as we show below, the third tranche of loans, which did not go out
until 2021, had no discernible effect on employment, perhaps because this tranche
was issued when the labor market was already rapidly recovering.
Targeting
Targeting
The rapid, near-universal takeup of Paycheck Protection Program loans in 2020
is inseparable from the reality that the program was essentially untargeted. That
takeup was around 94 percent of all small businesses means that loans reached the
most and least distressed firms—and all those in between—in nearly equal propor-
tions. This observation helps to explain why there is little geographic correlation
between the size of the initial COVID local economic shock, prior to PPP’s passage,
and subsequent PPP participation (Granja et al. 2020).
Around $200 billion in so-called second draw loans were issued in 2021—see
column 3 of Table 1, panel B. Unlike the first two tranches of Paycheck Protection
Program funds, these loans were explicitly targeted at firms that had experienced
significant revenue losses over the course of the pandemic (and had already
received a first PPP loan). We find a much higher correlation between PPP loan
volumes and state-level employment declines for loans issued in 2021 than those
issued in 2020 (see online Appendix Figure B.1), suggesting that this targeting
was more than nominal. Nevertheless these loans do not appear to have boosted
employment, as we show below.
What Did the Paycheck Protection Program Accomplish?
What Did the Paycheck Protection Program Accomplish?
Supporting Employment
Supporting Employment
A first step in calculating where the Paycheck Protection Program money went
is to determine what fraction of funding went to paychecks that would otherwise not
have been paid. Because PPP was ultimately taken up by almost all small businesses,
we lack an ideal control group for making experimental comparisons. Nevertheless,
a burgeoning literature, our own analysis included, indicates that PPP substantially
boosted payroll employment.
The simplest and arguably most credible—though not necessarily most
complete—method to assess the employment effects of the Paycheck Protection
Plan is to compare the trajectory of employment at firms below the 500-employee
initial-eligibility threshold to employment at ineligible firms above this threshold
during the course of the pandemic. Figure 1—which is similar to our analysis in
Autor et al.(2020)—presents this comparison using ADP payroll data. Employ-
ment is indexed to each firm’s average level of employment in February 2020
The $800 Billion Paycheck Protection Program 63
(immediately before the pandemic) for two employment size classes: 401–500
employees (in blue) and 501–600 employees(in red). Employment declines in
parallel for these groups of firms at the start of the crisis. Following the launch of
PPP, these trends diverge, with employment at firms that are likely eligible for PPP
loans (401–500 employees) falling by substantially less than employment at firms
that are likely ineligible (501–600). Approximately a month after the start of the
PPP, employment had fallen by approximately 4 percent less at likely-eligible firms
than at likely-ineligible firms. In the months thereafter, employment levels rela-
tive to baseline at likely-eligible and likely-ineligible firms gradually converged,
with the difference falling to less than 2 percent by the start of July 2020. It disap-
peared altogether by September 2020.
Our formal econometric analysis of the employment effects of the Paycheck
Protection Program in Autor et al.(2020) exploits this comparison of firms above
versus below the size eligibility threshold, while additionally controlling for the
differential impact of the pandemic across industries and states. After accounting
for the fact that not all eligible firms received a loan, particularly in the initial
months of the program, we estimate that taking out a PPP loan boosted firm
employment by between 4 and 10 percent in mid-May and by 0 to 6 percent by the
1.02
401–500
501–600
CARES
PPP1
PPP2
Index (February 2020 = 1)
Week ended on Saturday
0.98
0.96
0.94
0.92
0.9
0.88
0.86
0.84
0.82
1
Feb 8
Mar 7
Apr 4
May 2
May 30
Jun 27
Jul 25
Aug 22
Sep 19
Oct 17
Nov 14
Dec 12
Figure 1
Employment by Firm Size for Industries with PPP Eligibility at 500 Workers
Source: Author’s calculations using ADP data.
Note: Each series represents average employment for firms with that particular range of workers during
2019 (on average) and in February 2020. Data are weighted by each firm’s employment as of February
2020. Sample reflects firms that were present in the ADP data for all 12 months of 2019.
64 Journal of Economic Perspectives
end of the year.2 Our best evidence is that about 2.97 million jobs per week were
preserved by the Paycheck Protection Program in the second quarter of 2020,
and 1.75 million jobs per week were preserved in the fourth quarter. Chetty et al.
(2020) and Hubbard and Strain (2020) conduct similar analysis exploiting the
eligibility size threshold, using non-ADP data sources, and reach broadly similar
conclusions. Assuming that the employment effect declines linearly from its peak
in May 2020 to zero by June 2021 implies that PPP saved 1.98 million worker years
of employment at the very substantial cost of $258,000 per worker-year retained.
These estimates based on eligibility thresholds are subject to an important
caveat: because they focus on firms just above and below the 500 employee size-eligi-
bility threshold for Paycheck Protection Program loans, they may not capture the
effect of such loans on smaller firms. If smaller firms are more liquidity constrained
and hence more likely to shrink or shut down during the pandemic (Chodorow-
Reich et al. 2021), the threshold-based estimates will likely underestimate the effects
of PPP at these firms and, by implication, understate the full effect of PPP.
To develop causal effect estimates that cover a broader set of treated firms, a
number of papers exploit an event-study approach that compares employment at
firms receiving a loan early in the program period to employment at firms receiving
a loan later. This approach potentially captures the effect of PPP loans on small firms
that are well below the eligibility threshold, though it comes at a cost of focusing
only on the early months of the program, before most firms had taken loans.3
We complement existing event-study estimates using the vast ADP database,
which offers substantial precision and a sample frame identical to that used for
the size-threshold analysis above. To implement the event-study using the timing of
loan takeup, we merge PPP loan-level data from the Small Business Administration
to our sample of employers from ADP. This provides the precise date of PPP loan
approval for each matched firm within a sizeable sample of firms with fewer than 50
employees.
Figure 2 presents our timing-based estimates which trace out the effect of
receipt of a Paycheck Protection Program loan on employment at firms with fewer
than 50 employees.4 The employment trend prior to loan approval is roughly
2 Adjusting for incomplete takeup means rescaling our Intent-to-Treat (ITT) estimates by the takeup rate
to obtain Treatment-on-the-Treated estimates (TOT).
3 Papers using the event-study approach obtain a range of employment effect estimates. The first to
employ this approach, Granja et al. (2020), finds aggregate employment effects that are comparable to
those found by the eligibility threshold papers. Estimates in Li and Strahan (2020) imply a much smaller
boost to employment, however, while those in Bartik et al. (2021), Doniger and Kay (2021), Faulkender
et al. (2020), and Kurmann et al. (2021) point toward a larger employment effect.
4 A rapidly growing literature—for example, Goodman-Bacon (2021), Callaway and Sant’Anna (2020),
Sun and Abraham (2020)—highlights the problems that arise in event-study estimates when the magni-
tude of the treatment effect is correlated with the timing of treatment. We resolve this issue using the
approach proposed by Sun and Abraham (2020): we estimate and then average separate treatment
effects for each of the first eleven cohorts of borrowers, where cohort refers to week of loan issuance,
while using the final seven cohorts as a comparison group. We choose the final seven cohorts to ensure a
sufficient sample size. Using only those firms receiving a PPP loan in the final week of the program yields
qualitatively similar results.
Autor et al. 65
flat and about equal to zero, but begins rising on loan approval. Five weeks later,
employment is roughly 12 percent higher, where it remains through the close of
We estimate the following specification:
yit = α + ∑
c∈T
∑
g=−8
11
(βc,g ∗ PPPg,it) ∗ Dc + θjt + θst + ϵit,
where yit is total employment for firm i at week t indexed to equal 1 in February of 2020, θjt is a vector of
NAICS 3-digit industry j–by–week t fixed effects, θst is a set of state s–by–week t fixed effects, and PPPg,it is
a dummy variable equaling one if firm i at time t was approved for a PPP loan g weeks ago; g = 0 denotes
the week of approval and the week prior to approval (g = −1) is the omitted category. Dc is a dummy vari-
able denoting the week of PPP receipt for each cohort in the treatment set T (the first week through
the eleventh week of the program). For additional details of how we implement the equation, see the
online Appendix.
−0.04
−8
−7
−6
−5
5
6
7
9
10
11
8
−4
−3
−2
−1
0
1
2
3
4
−0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Weeks since initial PPP receipt
Index (February 2020 = 1)
Figure 2
Event-Study Employment Effects at Firms Sized 1–49
Source: Authors’ analysis of SBA and ADP data using Sun and Abraham (2020) “eventstudyinteract”
STATA implementation.
Notes: Estimates from Sun and Abraham (2020) event-study interaction estimator on the sample of
loan-matched ADP firms with between 1-49 employees where firm size is defined using the average size
in February 2020. The outcome variable—firm-level employment—is indexed to equal 1 in February
2020. The estimates are weighted by each firm’s employment as of February 2020 and include controls
for 3-digit industry-by-week and state-by-week fixed effects. Standard errors are clustered at the 3-digit
industry. All points to the right of the solid line represent post-treatment periods. Alternatively, accounting
for the biweekly pay schedule of most ADP employers, and the back-filling used to establish start dates,
all periods to the right of the dashed line can be viewed as post-treatment. See online Appendix Section
D.4 for more details.
66 Journal of Economic Perspectives
the outcome window. The relatively flat pre-trend centered around zero, and the
sharp upward break after approval, are consistent with the interpretation that we
are detecting a causal effect of PPP loans on small firm employment. We emphasize
that these results indicate that small firms shrank relatively less after receiving a PPP
loan as compared to firms not yet receiving a loan—not that their employment rose
during the pandemic. The fact that the estimated effect on small firm employment
is roughly twice as large as what we estimate for larger firms supports the view that
smaller firms received a bigger employment boost from PPP.
Combining the results from Autor et al. (2020) for larger firms with the smaller
firm results in Figure 2, we estimate that PPP loans originating in 2020 preserved
about 3.0 million job-years at an average cost of $169.300 per job-year saved. We use
this result below when calculating the share of PPP funds that accrued to paychecks.
While our findings in Figures 1 and 2 capture the employment effects of loans
issued in 2020 from the first two tranches of PPP funding, we know of no similar
evidence on the consequences of the third major tranche of $278 billion in PPP
loans issued in 2021. To complete this picture, we estimate difference-in-difference
threshold eligibility results analogous to those in Autor et al.(2020) for the second
draw PPP loans which constituted the majority of third tranche loans issued in
2021 (comparing employment at firms above and below the 300 worker eligibility
threshold for second draw loans).
Despite seemingly better targeting than the 2020 loans, we find no evidence in
Figure 3 that the 2021 second-draw loans boosted employment, perhaps because
they were issued too late to be relevant, after the economic recovery was well
underway. If this interpretation is correct, it affirms that Congress was wise to priori-
tize speed over precision in dispatching the initial two tranches of PPP loans.
Preventing Firm Exits
Preventing Firm Exits
The spike in business closings during the COVID pandemic was historic. The
Business Employment Dynamics database collected by the Bureau of Labor Statistics
finds that employment at closing firms, which hovered at about 1 million worker
per quarter for the last three decades, spiked to 2.1 million workers in the second
quarter of 2020. Evidence in Crane et al. (2020) and Kurmann et al. (2021) corrobo-
rates these trends using a number of alternative indicators, including data from ADP.
A key justification for the Paycheck Protection Program was to prevent a conta-
gion of business closures that would cause longer-term economic damage (Hubbard
and Strain 2020). Business deaths—as distinct from business contractions and
temporary closures—may potentially produce lasting economic harm not only by
forcing the costly reallocation of physical capital, but also by permanently destroying
worker-firm relationships and the associated match-specific capital (Farooq et al.
2020). Indeed, the prevalence of recall hires—as opposed to new hires—when firms
rebound from contractions underscores the importance of match-specific capital to
both employers and employees (for example, Fujita and Moscarini 2017).
We can observe the importance of firm closures for employment losses during
the pandemic in the ADP data. Figure 4 groups firms into size classes based on their
The $800 Billion Paycheck Protection Program 67
February 2020 (pre-pandemic) employment and reports the share of their orig-
inal employment that is lost due to firm shutdowns in each week between February
and December 2020. Shutdowns are heavily concentrated among small firms: fully
10 percent of employment at firms that had 1–50 employees in February 2020 was
lost due to shutdowns by early April 2020. For firms with more than 50 workers,
these losses were only one-tenth to one-third as large. (We note that the general
upward slope of the series in this figure is expected since some fraction of firms
inevitably closes each year.5)
5 We define a firm as shutdown if it has no paid employment in a given week. Although we cannot
definitively determine whether firms that appear to be shutdown in the ADP data have shuttered busi-
ness or rather stopped utilizing ADP’s payroll services, we expect that the spike in apparent shutdowns
during the pandemic primarily reflects firms dropping to zero employment rather than discontinuing
ADP’s services.
Index (December 2020 = 1)
Week ended on Saturday
0.12
0.1
0.08
0.06
0.04
0.02
−0.02
−0.04
−0.06
−0.08
0
Consolidated
Appropriations
Act
Dec 12
Dec 26
Jan 9
Jan 23
Feb 6
Feb 20
Mar 6
Mar 20
Apr 3
Apr 17
Figure 3
Difference-in-Differences Eligibility Threshold Employment Effect, Second Draw
Loans in 2021
Source: Authors’ analysis of ADP data.
Note: Each firm’s size is determined using average employment in 2019 and employment in February
2020, the same size measure as used in Autor et al. (2020). Regressions are weighted by firm size as
of February 2020 and include controls for state-by-week and industry-by-week effects. Standard errors
are clustered at the 3-digit NAICS industry level. Sample comprises firms that were present in the ADP
data for all 12 months of 2019. The specification is analogous to the difference-in-difference eligibility
threshold specification in Autor et al. (2020); treatment and control groups are based on being within a
range of 100 below and above, respectively, the 300 eligibility threshold for second draw loans. See the
online Appendix for additional discussion.
68 Journal of Economic Perspectives
Figure 4 also offers tantalizing evidence that the Paycheck Protection Program
may have inhibited firm closures or spurred reopenings. Among firms sized 1–50 and
51–100, firm shutdowns peaked shortly after PPP loans began flowing and rapidly
reversed course thereafter. By June 2020, the fraction of employment at small firms lost
due to closure was only half as large as in April—meaning that many had reopened.
Following recent work by Dalton (2021), we test whether the receipt of a
Paycheck Protection Program loan affects the probability that firms with fewer than
50 employees remain open (or reopen after closure). Using event study estimates
akin to those above for small-firm employment, we find in Figure 5 that PPP loans
reduced employment losses due to small-firm closures by about 8 percentage points
five weeks after loan receipt. Since our earlier results in Figure 2 found a peak PPP
effect on small-firm employment of 12 percentage points at week five, we infer that
about two-thirds of the employment-preserving effect of PPP loans on very small
firm employment was due to PPP keeping the lights on at establishments that would
have otherwise shuttered—at least temporarily.6
6 One anomaly in our results in Figure 5 is that the estimated employment effects of PPP receipt at
small firms appear to start a week too early relative to loan receipt. A possible explanation is that a large
Percent of February 2020 empolyment
Week ended on Saturday
Feb 8
Mar 7
12
11
10
9
8
7
6
5
4
3
2
1
0
Apr 4
May 2 May 30 Jun 27 Jul 25 Aug 22 Sep 19 Oct 17 Nov 14 Dec 12
51–100
1–50
CARES
PPP1 PPP2
101–250
501–750
251–500
Figure 4
Share of February 2020 Employment at a Firm with Zero Employment
Source: Authors’ analysis of ADP data.
Note: Each series represents the size-weighted share of firms with zero employment in a given week, using
February 2020 employment as the weight. Sample reflects firms that were present in the ADP data for all
12 months of 2019, and excludes firms in NAICS 72.
Autor et al. 69
Ultimately, permanent business closure proved less pervasive than many had
anticipated at the pandemic’s onset. The Paycheck Protection Program may be part
of the reason. Because our methodology permits examining firm closures only over
the short run, we cannot assess whether PPP averted permanent firm exits or mainly
temporary closures. Using a related methodology, Dalton (2021) finds that the
fraction of ADP paycheck recipients are paid biweekly, and this payment scheme blurs the observable
timing of any discrete event over the prior 13 days. Concretely, imagine that a firm’s two-week pay period
begins July 17 and ends July 31. After receiving a PPP loan, that firm reopens its doors on July 30th. In
our estimation, this firm will show an employment jump on July 17, even though all of its hires occurred
13 days later, when the loan was issued.
Weeks since initial PPP receipt
Index (February 2020 = 1)
−8
−7
−0.16
−0.14
−0.12
−0.1
−0.08
−0.06
−0.04
−0.02
0.02
0.04
0
−6
−5
0
1
2
3
4
5
6
7
8
9
10
11
−4
−3
−2
−1
Figure 5
Employment Change Due to Firm Closure, Firms Sized 1–49
Source: Authors’ analysis of SBA and ADP data using Sun and Abraham (2020) “eventstudyinteract”
STATA implementation.
Notes: Estimates from Sun and Abraham (2020) event-study estimator on the sample of loan-matched
ADP firms with between 1–49 employees where firm size is defined using the average size in February
2020. The analysis is implemented by setting yit in the equation in footnote 4 equal to one if firm i has
zero employment in week t and zero if it has positive employment. The specification includes controls
for 3-digit industry-by-week and state-by-week fixed effects. Standard errors are clustered at the 3-digit
industry. The coefficient vector displayed in the figure can be interpreted as the percentage point effect
of the PPP on employment due to firm closure as the specification is weighted by pre-pandemic firm
employment. All points to the right of the solid line represent post-treatment periods. Alternatively,
accounting for the biweekly pay schedule of most ADP employers, and the back-filling used to establish
start dates, all periods to the right of the dashed line can be viewed as post-treatment. See online
appendix Section D.4 for more details.
70 Journal of Economic Perspectives
PPP effect on small-firm closures waned somewhat over the ensuing seven months,
indicating that some of the PPP effect on closure was temporary, not permanent,
in nature. For larger firms around the 500 employee eligibility threshold, we find no
consistent evidence that the PPP influenced shutdowns, either over the short- or
longer-term (for details, see online Appendix figure E.1). Despite bolstering jobs
during the pandemic, PPP may not have had a pronounced effect on preserving
intangible business capital. More work is needed to definitively assess the effect of
the PPP on permanent business closure.7
Reducing Commercial Delinquency
Reducing Commercial Delinquency
Alongside preserving jobs and keeping firms open during the pandemic, the
Paycheck Protection Program may have indirectly benefited creditors of small busi-
nesses—landlords, banks, holders of mortgage-backed securities, and suppliers—by
keeping payments flowing. There is limited evidence on the effect of PPP on loan
recipients’ ability to pay creditors, but the evidence that exists suggests the impact
was positive. Exploiting differences in the tendency of small versus large commer-
cial properties to have PPP-eligible tenants, Agarwal et al. (2021) find that PPP
significantly blunted the rise in commercial mortgage delinquency rates during
the pandemic, particularly in the retail sector. Using survey data and a variant of
the event-study strategy discussed above, Granja et al. (2020) also find that PPP
decreased delinquency on mortgages and other payments.
Where Did the Money Go?
Where Did the Money Go?
The estimates above provide a key input for answering our motivating ques-
tion—that is, where did the money go? Using the employment effects estimated
above, along with many other data sources, we estimate the incidence of the
$510 billion in Paycheck Protection Program loans issued in 2020 across the house-
hold income distribution. We further compare this incidence to other pandemic
economic assistance programs. Additional information on these calculations can be
found in section F of the online Appendix.
Proximate Recipients: Workers versus Non-workers
Proximate Recipients: Workers versus Non-workers
Paycheck Protection Program funds were paid to businesses. In turn, businesses
used these funds to pay three proximate groups of beneficiaries: workers who other-
wise would have been laid off; creditors and suppliers who otherwise would not have
been paid; and owners and shareholders of PPP-receiving firms as residual claim-
ants in cases where businesses would have met some or all of their payroll and other
financial obligations absent PPP (also known as “windfall profits”). The distribution
7 Other recent work provides mixed evidence on these outcomes. Granja et al. (2020) find little evidence
of a PPP effect on firm shutdown. Bartik et al. (2021) and Kurmann et al. (2021), though, find that PPP
mitigated business shutdowns.
The $800 Billion Paycheck Protection Program 71
of PPP funds among these groups—workers versus non-workers, in particular—
matters for our accounting exercise because different groups represent different
parts of the household income distribution.
We focus first on payments to workers. As documented above, Paycheck Protec-
tion Program loans issued in 2020 modestly raised employment at recipient firms.
To convert these employment effects into payroll expenditures, we use the main
estimates reported above from Autor et al. (2020), who find that PPP boosted
employment by about 6 percent in mid-May 2020, with effects tapering off gradu-
ally thereafter. These numbers imply that PPP preserved about 2.97 million jobs
per week in the second quarter of 2020 and about 1.75 million jobs per week by the
fourth quarter of 2020. Assuming a linear trend decline in this program effect, PPP
would have had zero employment effects by June 2021. Converting these weekly job
numbers into job-years (that is, one worker for one year), implies that PPP preserved
about 1.98 million job-years of employment at a cost of $258,000 per job-year saved
(that is, $510 billion/1.98 million). We assume that actual employee compensa-
tion for each saved job averaged $58,200 since the average weekly wage from the
Current Population Survey in February 2020 is $786 (truncating at an annual wage
of $100,000 above which the PPP did not provide additional support per worker)
and, on average, total compensation is 42 percent larger than wages according to
BLS Employer Costs for Employee Compensation data ($786 × 52 × 1.42). The 1.98
million job-years saved then imply that $115 billion in PPP loans ($58,200 × 1.98
million) accrued to employee paychecks.
We also produce an alternative estimate of the amount of Paycheck Protec-
tion Program loans accruing to compensation based on our 3.0 million job-years
saved estimate which combines the results from Autor et al. (2020) and the larger
effects for smaller firms in Figure 2. Continuing to assume that compensation at
retained jobs averaged $58,200 implies that $175 billion in PPP compensation went
to paychecks. It is likely that this $175 billion estimate is an upper bound on the
share of PPP funds flowing to worker compensation. Some event-study estimates
for the entire size distribution of PPP-eligible firms, including small firms, find an
overall peak PPP employment effect of approximately 6 to 8 percent (for example,
see online Appendix Figure D.1 and Dalton 2021). These estimates are more in line
with our smaller $115 billion estimate. Additionally, our assumption of a smooth
trend decline in PPP’s impacts through June 2021 is generous.8 Moreover, we are
not accounting for loans issued in 2021 where our evidence suggests the PPP failed
to boost employment; doing so would further lower the estimated share of PPP
loans that flowed to workers relative to non-workers and hence substantially raise
the estimated cost per job-year preserved.
These bounds of $115 billion to $175 billion in Paycheck Protection Program
funds accruing directly to paychecks imply that between 23 percent and 34 percent
8 In Autor et al. (2020), we detect no statistically significant impact of PPP on employment after July 2020.
Because the point estimates remain non-zero through December 2020, we extrapolate the entire series
out until it is numerically zero in June 2021.
72 Journal of Economic Perspectives
of the first two tranches of PPP dollars totaling $510 billion supported jobs that
would otherwise have been lost. By implication, the remaining $335 to $395 billion
(66 to 77 percent) accrued to owners of business and corporate stakeholders,
including creditors and suppliers, and others.
The Household Distributional Incidence
The Household Distributional Incidence
To trace the flow of Paycheck Protection Program payments from their proxi-
mate recipients to their household incidence requires information on the income
distributions of both worker and non-worker (that is, owner) beneficiaries. Starting
with the worker beneficiaries, we estimated above that, at the high end, $175 billion
in PPP money flowed to workers whose jobs were saved. We assume that the distri-
butional incidence of those funds followed the distribution of job loss in 2020 by
household income quintile. To estimate this distribution, we first measure employment
declines across the weekly wage distribution using the Current Population Survey
Outgoing Rotation Group (CPSORG) files. Pandemic job losses were largest for
low-paid workers: total employment fell by 17.8 percent from March 2020 through
the end of 2020 among the lowest-paid (first) quintile of workers; by 10.6 percent
in the second quintile; by 6.0 and 2.2 percent in the third and fourth quintiles,
respectively; and by a substantial 8.7 percent among the highest quintile of earners.
We convert these job loss percentages into average weekly wage losses by multiplying
each by the February 2020 pre-COVID average weekly wage within quintile. From
there, it is straightforward to calculate the share of compensation lost by weekly
wage quintile, which we impute to the household income distribution using March
CPS data on the joint distribution of weekly wages and household income.9
We make an analogous (but simpler) imputation for the household incidence
of the $335 billion in Paycheck Protection Program fund payments that flowed to
non-worker beneficiaries, such as creditors and suppliers who otherwise would not
have been paid and owners and shareholders of PPP-receiving firms. Specifically,
we use the Congressional Budget Office’s (2020a) most recent estimates on the
distribution of capital incomes by type to distribute the funds across households.
We do not attempt to account for the flow of PPP funds from proximate and subse-
quent recipients—for example, a PPP-receiving firm’s supplier pays its workers or a
worker at a PPP-receiving firm pays her landlord. Thus, our exercise is in the spirit
of the static distributional incidence analyses performed for tax policies by the Joint
Committee on Taxation.
Unlike Paycheck Protection Program payments, which went to businesses,
transfer payments made by the two other major federal pandemic emergency assis-
tance programs—pandemic unemployment insurance payments and household
9 Specifically, we calculate Sq ≡ T ×
Uq × Wq
_________
∑ q=1
5 Uq × Wq
, where T is total PPP dollars that support employment, Uq
is quintile q′s share of job losses during the pandemic, and Wq is quintile q′s wage in February 2020 prior
to the pandemic. We then map from the weekly wage distribution to the household income distribution
using March CPS supplement data on the probability that a worker in a given weekly wage quintile is in
each household income quintile.
Autor et al. 73
stimulus payments—went directly to households and workers. The size of these
payments rivaled those of PPP, as noted above. To facilitate comparison with PPP, we
calculate the distributional incidence of these as well.
For household payments, we use incidence data from Bhutta et al. (2020), who
analyzed the effect of stimulus payments on household finances using the Survey of
Consumer Finances. For unemployment insurance, we calculate approximate shares
of benefits paid during the pandemic—including regular state programs and the
pandemic enhancements to unemployment insurance—using, as a starting point,
our estimates of average wages lost during the pandemic, and applying the method-
ology above for apportioning the PPP funds to paychecks. We combine these data on
wages lost by quintile with simple estimates of the unemployment insurance replace-
ment rate by quintile, which we estimate using same CPS ORG data. This calculation
accounts for both pandemic supplements to weekly unemployment insurance bene-
fits and the portion of Pandemic Unemployment Assistance payments that went to
the self-employed, as estimated by Boesch et al. (2021).
Panel A of Figure 6 reports the distribution of Paycheck Protection Program,
unemployment insurance, and household payments in billions of dollars across
household quintiles. As the figure makes clear, the distribution of PPP loans over-
whelmingly accrued to high-income households. Of the $510 billion in PPP loans
provided in 2020, we estimate that only $13.2 billion ultimately flowed to house-
holds in the bottom fifth of the income distribution, and that $130.8 billion flowed
to the second through fourth quintiles. The remaining $365.9 billion (72 percent)
400
Billions
Percent of annual income
20
15
10
5
0
300
200
100
Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
0
PPP compensation
Unemployment insurance
PPP capital income
Stimulus checks
Figure 6
Distribution of Paycheck Protection Program, Unemployment Insurance, and
Stimulus Check Payments
Source: Authors’ analysis of CBO, Census Bureau, BEA, BLS ECEC, and Current Population Survey
microdata, and estimates from Autor et al. (2020), Bhutta et al. (2020), and Boesch et al. (2021).
Note: See online Appendix for details of calculations. Calculations are based on our upper-end (most
generous) estimate of the wage compensation supported by PPP. Online Appendix Figure F.1 reports the
same calculations using our lower-end estimate of the wage compensation supported by PPP.
74 Journal of Economic Perspectives
flowed to the top fifth of household income. This skew reflects two features of PPP.
First, high-wage earners are found in high-income households. Though the PPP
offered loans to support only up to $100,000 in annual earnings, even with this
truncation, the top fifth of households accounts for about 35 percent of wage and
salary earnings. Second, the distribution of capital ownership is even more right-
skewed than the distribution of wage earnings—with the top fifth of households
commanding 86.2 percent of capital income—meaning that subsidies to businesses
are ultimately subsidies to high-income households.
If we additionally account for the $100 billion in tax credits that Congress
granted to PPP-receiving businesses in January 2021 by making PPP payments
nontaxable, we conclude that an additional $85 billion flowed to the top quintile
of households, raising that quintile’s total PPP distribution further from $365.9 to
$450.9 billion.10
In comparison, both household stimulus payments and pandemic unemploy-
ment insurance payments were far less regressive than in the Paycheck Protection
Program. The incidence of household stimulus checks in dollar terms was close to
uniform across the lower four income quintiles. Moreover, due to the income caps
that Congress set on household payments, the incidence of these payments was
much smaller for the highest quintile of households.
Meanwhile, the incidence of unemployment insurance during the pandemic
was weighted towards both the upper and lower tails of the household income
distribution. We estimate that 31.5 percent ($175.6 billion) and 20.7 percent
($115.4 billion) in pandemic unemployment insurance payments went to the
bottom fifth and second-to-bottom fifth of households, respectively (red diamonds
in panel A of Figure 6). Surprisingly, the top fifth of households received a bit more
than one-quarter of unemployment insurance benefits. This occurred both because
the highest income quintile of wage and salary workers sustained substantial
employment losses during the pandemic (as documented above), and because the
Pandemic Unemployment Assistance (PUA) program allowed self-employed busi-
ness owners—who tend to have high incomes—to collect unemployment insurance
benefits. Estimates from Boesch et al. (2021) suggest that self-employed business
owners received about 40 percent of Pandemic Unemployment Assistance insur-
ance benefits.
Panel B of Figure 6 recasts these distributional incidence figures into house-
hold annual income replacement rates rather than dollar transfers. Both stimulus
checks and unemployment insurance payments replaced about 17 percent of the
incomes of the lowest quintile of households, with much lower shares at higher quin-
tiles. Thus, although the combination of these three programs is highly regressive
10 We make this calculation by attributing the $100 billion in forgone tax revenues to the share of busi-
ness income that goes to the top fifth of households, which is approximately 85 percent. An alternative
calculation yields a comparable conclusion: the Congressional Budget Office estimates that the average
marginal tax rate on capital income is 18.6 percent; noting that PPP distributed $510 billion in 2020 and
that the top fifth of households commands about 85 percent of capital income, we obtain a tax benefit
for the top fifth of $81 billion.
The $800 Billion Paycheck Protection Program 75
in dollar terms, it is roughly progressive in replacement rate terms due to the highly
skewed distribution of US household incomes.
Macroeconomic Benefits
Macroeconomic Benefits
An additional benefit of these transfers programs is that they provided stim-
ulus during a time of rapid economic contraction. The short-term macroeconomic
boost of a program during a recessionary period is conventionally linked to the
marginal propensity to consume (MPC) of those who receive benefits from the
program. Cashin et al. (2018) provide estimates of the MPC for different types of
fiscal shocks based on characteristics such as: the type of policy change (say, tax
versus transfer payment); who is receiving the benefit (say, low-income households
versus corporations); and whether the flow of benefits is temporary or permanent.
These MPC estimates are informed by a publicly available macroeconomic model,
FRB/US, used by the Federal Reserve Board staff (described in Brayton et al. 2014),
and by the relevant empirical literature.
Using these marginal propensity to consume estimates, we offer a back-of-
the-envelope comparison of the degree of stimulus provided by the three main
programs mentioned above: the Payment Protection Program, stimulus payments,
and pandemic-enhanced unemployment insurance. This calculation relies on the
following assumptions:
1. Since unemployment insurance payments are generally made to house-
holds that are highly liquidity-constrained, the marginal propensity to
consume out of unemployment insurance payments is one (Cashin et al.
2018).
2. Because stimulus payments are made to a broad mixture of households
across the income distribution, we use the estimate from Cashin et al.
(2018) for the MPC of general, temporary transfers to households of 0.5.
3. The part of the Paycheck Protection Program that flows through to wages
is similar to unemployment insurance, and thus has a plausible marginal
propensity to consume of one.
4. For the part of PPP that flows to non-workers, we use the estimates in Cashin
et al. (2018) for the MPC of temporary corporate tax cuts of 0.2; this rela-
tively low marginal propensity to consume is consistent with these funds
flowing disproportionately to the upper quintile of the income distribution.
Weighting these last two components together, we obtain an overall marginal
propensity to consume out of PPP loans of about 0.5, which is comparable to
stimulus checks (where we have an imputed MPC of 0.5) and much lower than
unemployment insurance payments (where we have imputed an MPC of one).11
11 We noted in the previous section that a substantial share of Pandemic Unemployment Assistance
recipients were likely high income self-employed business owners who might be expected to have a lower
marginal propensity to consume than the one we assume here for unemployment insurance recipients.
76 Journal of Economic Perspectives
This illustrative calculation thus suggests the PPP loans and stimulus checks were
roughly equally effective at boosting spending, and both were much less effective on
this margin than pandemic unemployment insurance.
These estimates have the virtue of transparency. They also have shortcom-
ings. The pandemic environment surely generated non-normal household and
business behavior. Extraordinarily high replacement rates delivered by enhanced
unemployment benefits may have diminished the marginal propensity to consume
of recipients. The substantial share of payroll income received by the top quintile
from the Paycheck Protection Program also suggests that treating this income as
similar to unemployment insurance probably overstates the MPC. Finally, these
estimates quantify only the transfer’s initial boost to aggregate demand; they
do not capture aggregate supply effects, such as that arising from preventing
firm bankruptcies, or subsequent general equilibrium effects. Fortunately, the
Congressional Budget Office (2020b) has also estimated the boost to GDP per
dollar for these same pandemic programs, carefully accounting for the pandemic
environment and the specifics of each program. CBO also strives to capture the
full, general equilibrium effect of each program, including its potential impact on
business closure. CBO concludes that the enhanced unemployment and stimulus
checks were far more effective at boosting GDP than was PPP. Specifically, the
CBO estimates a per dollar boost to GDP of 0.36 for the PPP and 0.60 and 0.67 for
stimulus checks and enhanced unemployment insurance benefits, respectively.
Taking account of the highly distributionally-skewed incidence of PPP payments,
we concur that PPP was likely the least effective of the three programs in boosting
the macroeconomy.
Lessons Learned from the Paycheck Protection Program Experience
Lessons Learned from the Paycheck Protection Program Experience
The US small business sector appeared at risk of collapse at the outset of
the pandemic. To avert this collapse, Congress enacted the Paycheck Protection
Program, which successfully distributed vast amounts of aid to the near-universe of
eligible small businesses in the space of a few months. Our best evidence to date
indicates that the PPP’s economic impacts were less than hoped: it preserved only
a moderate number of jobs at a high cost per job-year retained and transferred
resources overwhelmingly to the highest quintile of households.
These outcomes should not, however, be viewed first and foremost as program-
matic failures. The PPP’s regressive distributional incidence and its limited efficacy
as economic stimulus stem from the program’s absence of targeting. This absence,
in turn, reflected necessity. Given the time constraints and, more profoundly, the
Nonetheless, even if we assumed an MPC of zero for self-employed PUA recipients, the overall MPC out
of unemployment insurance would be about 0.7 (since non-PUA benefits were about 70 percent of total
unemployment insurance), which is still higher than our estimates of the MPC out of PPP and stimulus
checks.
Autor et al. 77
lack of existing administrative infrastructure for overseeing targeted federal support
to the entire population of US small businesses at the onset of the pandemic, we
strongly suspect that Congress could not have better targeted the Paycheck Protec-
tion Program without substantially slowing its delivery. We thus concur with Bartik
et al. (2021) that policymakers made a defensible trade-off between speed and
targeting in the PPP’s design.
However, if the Paycheck Protection Program was a logical answer to a highly
constrained question, a forward-looking lesson from the PPP experience is that the
United States should invest now to relax those constraints. We have emphasized that
the PPP had dual goals: preserving jobs and providing liquidity. These goals could
be better served with the administrative capacity to address these issues directly and
separately, thus enabling better targeting and a more progressive incidence. The
primary job retention goal of the Paycheck Protection Program could in the future
be better achieved through an expanded program to encourage “work-sharing,”
which refers to a policy in which employers, when faced with an economic downturn,
are encouraged to reduce hours worked more broadly across the workforce rather
than laying off a narrower group outright. In effect, the government program ends
up paying partial unemployment to many, rather than full unemployment to some.
Currently, 26 US states have a work-sharing program through their unemploy-
ment insurance systems, but these were not well-subscribed where available during
the COVID recession. A number of proposals over the last decade that advocate
for expanded work-sharing suggest that to reach broader coverage, such programs
should be simplified and automated (Abraham and Houseman 2014; Strain 2020;
Dube 2021). A work-sharing program can target firms of all sizes that are cutting
hours or employment, not just small firms. Additionally, with sufficient administra-
tive capacity developed in normal times, the progressivity of the program could be
altered as policymakers deem appropriate.
A separate liquidity provision program could then be targeted primarily at
small firms, which are more likely to be liquidity-constrained. Moreover, with better
information systems operational, liquidity could be provided in proportion to firms’
decline in revenues as well as firms’ actual fixed obligations.
Distinct from the United States, many other high-income countries responded
to the pandemic with a mixture of job retention incentives, including 1) work-sharing
programs that allowed either partial or complete furloughs; 2) newly introduced
wage subsidy programs, similar in many ways to the Paycheck Protection Program,
that provided businesses with direct support for at least some fraction of their wage
bill (OECD 2021). Both work-sharing and wage subsidy programs were targeted.
Wage subsidy programs were explicitly targeted to firms that had experienced
declines in revenue: for example, Canada’s Employer Wage Subsidy was available
to firms that experienced a year-over-year revenue drop of 30 percent (reduced to
10 percent later). In some countries, firms were entitled to wage subsidies on a
sliding scale in proportion to their declines in revenues. By contrast, work-sharing
programs were not explicitly targeted to distressed firms. But the requirement that
firms reduce workers’ hours to obtain assistance generally makes firm participation
78 Journal of Economic Perspectives
unattractive absent a negative shock (for additional discussion in this symposium,
see the paper by Giupponi, Landais, and Lapeyre). Of course, the details of these
programs, such as the length of benefits and the extent to which non-payroll
expenses are covered, as well as their efficacy, varies across countries.
A key lesson from these cross-national comparisons is that targeted business
support systems were feasible and rapidly scalable in other high-income coun-
tries because administrative systems for monitoring worker hours and topping up
paychecks were already in place prior to the pandemic. Lacking such systems, the
United States chose to administer emergency aid using a fire hose rather than a fire
extinguisher, with the predictable consequence that virtually the entire small busi-
ness sector was doused with money. This approach may have been necessary, but it
was desirable only because the United States lacked viable alternatives. By building
administrative capacity in the years ahead, the United States could more deftly
target, calibrate, and deploy its emergency business response systems when most
needed. This investment will pay off in the near term in improvements of existing
programs, such as work sharing—and will pay off again when the next crisis arises.
■
■ At the Federal Reserve Board, we thank Christopher Kurz and Norman Morin for support,
Kendra Robbins and Eleanor Warren for excellent research assistance, and Kevin Moore for
updating the analysis in Bhutta et al. (2020). We thank Tolga Tuncoglu of ADP for superb
assistance with matching the SBA PPP loan data into the ADP data. We thank Michael
Dalton, Andrew Goodman-Bacon, Erik Hurst, Katie Lim, Adam Looney, Joseph Nichols,
Ryan Nunn, Matthew Shapiro, Liyang Sun, and Eric Zwick for helpful discussion. Autor
acknowledges financial support from the Smith Richardson Foundation (#20202252),
Accenture LLP (#027843-0001), the Andrew Carnegie Fellowship (G-F-19-56882), and the
Washington Center for Equitable Growth (APP-01666). The analysis and conclusions set
forth here are those of the authors and do not indicate concurrence by other members of the
Federal Reserve Board research staff, by the Board of Governors, or by ADP (ADP’s data
privacy policy can be found at https://www.adp.com/about-adp/data-privacy.aspx).
References
Abraham, Katherine G., and Susan N. Houseman. 2014. “Proposal 12: Work Sharing to Reduce Unem-
ployment.” Policies to Address Poverty in America, edited by Melissa S. Kearney and Benjamin H.
Harris, 1–10. Washington, DC: The Brookings Institution.
Agarwal, Sumit, Brent W. Ambrose, Luis A. Lopez, and Xue Xiao. 2021. “Did the Paycheck Protection
Program Help Small Businesses? Evidence from Commercial Mortgage-backed Securities.” https://
papers.ssrn.com/sol3/papers.cfm?abstract_id=3674960.
Amiram, Dan, and Daniel Rabetti. 2020. “The Relevance of Relationship Lending in Times of Crisis.”
The $800 Billion Paycheck Protection Program 79
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3701587.
Anbil, Sriya, Mark Carlson, and Mary-Frances Styczynski. 2021. “The Effect of the PPPLF on PPP Lending
by Commercial Banks.” FEDS Working Paper 2021-030.
Autor, David, David Cho, Leland Crane, Mita Goldar, Byron Lutz, Joshua Montes, William B. Peterman,
David Ratner, Daniel Villar, and Ahu Yildirmaz. 2020. “An Evaluation of the Paycheck Protection
Program Using Administrative Payroll Microdata.” https://economics.mit.edu/files/20094.
Bartik, Alexander W., Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matthew Unrath. 2020a.
“Measuring the Labor Market at the Onset of the COVID-19 Crisis.” Brookings Papers on Economic
Activity.
Bartik, Alexander W., Marianne Bertrand, Zoe Cullen, Edward L. Glaeser, Michael Luca, and Christo-
pher Stanton. 2020b. “The Impact of COVID-19 on Small Business Outcomes and Expectations.”
Proceedings of the National Academy of Sciences 117 (30): 17656–17666.
Bartik, Alexander W., Zoe B. Cullen, Edward L. Glaeser, Michael Luca, Christopher T. Stanton, and
Adi Sunderam. 2021. “The Targeting and Impact of Paycheck Protection Program Loans to Small
Businesses.” NBER Working Paper 27623.
Bernstein, Jared, and Jesse Rothstein. 2020. “A Fast, Simple Way to Get Support to Workers without
Paid Leave.” Washington Post, March 10. https://www.washingtonpost.com/outlook/2020/03/10/
fast-simple-way-get-support-workers-without-paid-leave/.
Bhutta, Neil, Jacqueline Blair, Lisa Dettling, and Kevin Moore. 2020. “COVID-19, The CARES act, and
Families’ Financial Security.” National Tax Journal 73 (3): 645–72.
Boesch, Tyler, Katherine Lim, and Ryan Nunn. 2021. “What Did and Didn’t Work in Unemployment
Insurance during the Pandemic.” Federal Reserve Bank of Minneapolis.
Brayton, Flint, Thomas Laubach, and David Reifschneider. 2014. “The FRB/US Model: A Tool for
Macroeconomic
Policy
Analysis.”
https://www.federalreserve.gov/econresdata/notes/feds-
notes/2014/a-tool-for-macroeconomic-policy-analysis.html.
Brevoort, Kenneth P., John A. Holmes, and John D. Wolken. 2010. “Distance Still Matters: The Informa-
tion Revolution in Small Business Lending and the Persistent Role of Location, 1993–2003.” FEDS
Working Paper 2010-08.
Callaway, Brantly, and Pedro H.C. Sant’Anna. 2020. “Difference-in-Differences with Multiple Time
Periods.” Journal of Econometrics 225 (2): 200–30.
Cashin, David, Jamie Lenney, Byron Lutz, and William Peterman. 2018. “Fiscal Policy and Aggregate
Demand in the USA before, during, and following the Great Recession.” International Tax and
Public Finance 25: 1519–58.
Congressional Budget Office. 2020a. The Distribution of Household Income, 2017. Congressional Budget
Office.
Congressional Budget Office. 2020b. “The Effects of Pandemic-Related Legislation on Output.” Congres-
sional Budget Office.
Chetty, Raj, John N. Friedman, Nathaniel Hendren, Michael Stepner, and The Opportunity Insights
Team. 2020. “How Did COVID-19 and Stabilization Policies Affect Spending and Employment?
A New Real-Time Economic Tracker Based on Private Sector Data.” NBER Working Paper 2743.
Chodorow-Reich, Gabriel, Olivier Darmouni, Stephan Luck, and Matthew Plosser. 2021. “Bank Liquidity
Provision across the Firm Size Distribution.” Journal of Financial Economics.
Cororaton, Anna, and Samuel Rosen. 2021. “Public Firm Borrowers of the US Paycheck Protection
Program.” Review of Corporate Finance Studies 10 (4): 641–93.
Crane, Leland D., Ryan A. Decker, Aaron B. Flaaen, Adrian Hamins-Puertolas, and Christopher J. Kurz.
2020. “Business Exit During the COVID-19 Pandemic: Non-Traditional Measures in Historical
Context.” Finance and Economics Discussion Series 2020-089.
Center for Economic Policy Research. 2021. “CPS ORG Uniform Extracts, Version 2.5.” Washington, DC.
CRFB. 2021. “Covid Money Tracker.” https://www.covidmoneytracker.org/.
Dalton, Michael. 2021. “Putting the Paycheck Protection Program into Perspective: An Analysis Using
Administrative and Survey Data.” Bureau of Labor Statistics Working Paper 542.
Decker, Ryan A., Robert J. Kurtzman, Byron F. Lutz, and Christopher J. Nekarda. 2021. “Across the
Universe: Policy Support for Employment and Revenue in the Pandemic Recession.” AEA Papers
and Proceedings 111 (May): 267–71.
Doniger, Cynthia L., and Benjamin Kay. 2021. “Ten Days Late and Billions of Dollars Short: The Employ-
ment Effects of Delays in Paycheck Protection Program Financing.” https://papers.ssrn.com/
sol3/papers.cfm?abstract_id=3747223.
80 Journal of Economic Perspectives
Dube, Arindrajit. 2021. “A Plan to Reform the Unemployment Insurance System in the United States.”
Hamilton Project Policy Proposal 2021-03.
Elmendorf, Douglas W., and Jason Furman. 2008. “If, When, How: A Primer on Fiscal Stimulus.”
Hamilton Project Strategy Paper January 2008.
Erel, Isil, and Jack Liebersohn. 2020. “Does FinTech Substitute for Banks? Evidence from the Paycheck
Protection Program.” NBER Working Paper 27659.
Farooq, Ammar, Adriana D. Kugler, and Umberto Muratori. 2020. “Do Unemployment Insurance
Benefits Improve Match Quality? Evidence from Recent U.S. Recessions.” NBER Working Paper
27574.
Faulkender, Michael, Robert Jackman, and Stephen I. Miran. 2020. “The Job-Preservation Effects of
Paycheck Protection Program Loans.” US Treasury, Office of Economic Policy Working Paper
2020-01.
lood, Sarah, Miriam King, Renae Rodgers, Steven J. Ruggles, Robert Warren, and Michael Westberry.
2021. “Integrated Public Use Microdata Series, Current Population Survey: Version 9.0, Annual
Social and Economic Supplement 2020.” Minneapolis, MN: IPUMS.
Fujita, Shigeru, and Giuseppe Moscarini. 2017. “Recall and Unemployment.” American Economic Review
107 (12): 3875–3916.
Goodman-Bacon, Andrew. 2021.“Difference-in-Differences with Variation in Treatment Timing.” Journal
of Econometrics 225 (2): 254–77.
Granja, Jõao, Christos Makridis, Constantine Yannelis, and Eric Zwick. 2020. “Did the Paycheck Protec-
tion Program Hit the Target?” NBER Working Paper 27095.
Harney, John, and Gregory Mott. 2021. “IRS Blesses Tax Breaks on Forgiven PPP Loans after Law
Change.” Bloomberg Tax, January 6. https://news.bloombergtax.com/daily-tax-report/irs-blesses-
tax-breaks-on-forgiven-ppp-loans-after-law-change.
Hubbard, Glenn, and Michael R. Strain. 2020. “Has the Paycheck Protection Program Succeeded?”
Brookings Papers on Economic Activity.
Joaquim, Gustavo, and Felipe Netto. 2021. “Bank Incentives and the Impact of the Paycheck Protection
Program.” Unpublished.
Kurmann, Andre, Etienne Lale, and Lien Ta. 2021. “The Impact of COVID-19 on Small Business Dynamics
and Employment: Real-Time Estimates with Homebase Data.” Unpublished.
Li, Lei, and Philip Strahan. 2020. “Who Supplies PPP Loans (And Does it Matter)? Banks, Relationships
and the COVID Crisis.” NBER Working Paper 28286.
OECD. 2021. OECD Employment Outlook 2021. https://www.oecd.org/employment-outlook/#Read-the-
report.
Small Business Administration (SBA). 2019. “2019 Small Business Profile.” Technical Report, Office of
Advocacy 2019. https://cdn.advocacy.sba.gov/wp-content/uploads/2019/04/23142719/2019-
Small-Business-Profiles-US.pdf.
Small Business Administration (SBA). 2021. “Forgiveness Platform Lender Submission Metrics, November
28,
2021.
https://www.sba.gov/document/report-ppp-forgiveness-platform-lender-submission-
metrics-reports?utm_medium=email&utm_source=govdelivery (accessed November 30, 2021).
Strain, Michael. 2020. “Let’s Provide Unemployment Benefits Without the Layoffs.” Bloomberg Opinion,
May 7.
Sun, Liyang, and Sarah Abraham. 2020. “Estimating Dynamic Treatment Effects in Event Studies with
Heterogeneous Treatment Effects.” Journal of Econometrics 225 (22): 175–99.