Olivia Kim et al., Revenue Collapses and the Consumption of Small Business Owners in the Early Stages of the COVID-19 Pandemic (2020)
Summary
NBER Working Paper No. 28151 by Olivia S. Kim, Jonathan A. Parker and Antoinette Schoar, Revenue Collapses and the Consumption of Small Business Owners in the Early Stages of the COVID-19 Pandemic, dated November 2020 and revised March 2025. Using JPMorgan Chase Institute account data linking small businesses to their owners' households, it examines how owners' consumption responded to revenue changes during the COVID-19 crisis. The authors report that business revenues declined by 40 percent in the first two months after the National Emergency, but that each dollar of revenue loss reduced owner consumption by only 1.6 cents. They find that this muted pass-through persisted through 2021 and attribute it to federal subsidies and pandemic-induced reductions in spending opportunities.
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
“Olivia Kim et al., Revenue Collapses and the Consumption of Small Business Owners in the Early Stages of the COVID-19 Pandemic (2020).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/olivia-kim-et-al-revenue-collapses-and-the-consumption-of-small-busine-5e8327cc8c/
, SHA-256 f4726dc3b3d765d5dbeed09ae057b6f651d350ffe3f4442cd37874c49adabc98.Full text
NBER WORKING PAPER SERIES
REVENUE COLLAPSES AND THE CONSUMPTION OF SMALL BUSINESS OWNERS IN
THE EARLY STAGES OF THE COVID-19 PANDEMIC
Olivia S. Kim
Jonathan A. Parker
Antoinette Schoar
Working Paper 28151
http://www.nber.org/papers/w28151
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
November 2020, Revised March 2025
For useful discussions, we thank Diana Farrell, Peter Ganong, Fiona Greig, Michaela Pagel, Chris
Wheat, as well as participants at presentations at HBS, MIT, and the 2022Winter Meetings of the
American Economic Association. Sarwari Das provided outstanding research assistance. This
research was made possible by a data-use agreement between the authors and the JPMorgan Chase
Institute (JPMCI), which has created deidentified data assets that are selectively available to be
used for academic research. All statistics from JPMCI data, including medians, reflect cells with
multiple observations. The opinions expressed are those of the authors alone and do not represent
the views of JPMorgan Chase & Co. The Kauffman Foundation provided support for this research
through the Knowledge Challenge Research Grant. The views expressed herein are those of the
authors and do not necessarily reflect the views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2020 by Olivia S. Kim, Jonathan A. Parker, and Antoinette Schoar. All rights reserved. Short
sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided
that full credit, including © notice, is given to the source.
Revenue Collapses and the Consumption of Small Business Owners in the Early Stages of
the COVID-19 Pandemic
Olivia S. Kim, Jonathan A. Parker, and Antoinette Schoar
NBER Working Paper No. 28151
November 2020, Revised March 2025
JEL No. D12, D15, D25, E2, E63, E65, G32, G5
ABSTRACT
Using financial account data linking small businesses to their owner households, we examine how
business owners’ consumption responded to changes in business revenues during the COVID-19
crisis. In the first two months following the National Emergency, business revenues declined by 40
percent, largely driven by national factors rather than local infection rates or policies. However, the
pass-through of revenue losses to owner consumption was limited: each dollar of revenue loss
resulted in only a 1.6-cent decline in consumption. This muted pass-through persisted through
2021, even after the introduction of COVID-19 vaccines. Our findings suggest that federal
subsidies and pandemic-induced reductions in spending opportunities explain the limited impact.
Olivia S. Kim Antoinette Schoar
60 North Harvard st MIT Sloan School of Management
Harvard Business School 100 Main Street, E62-638
Rock Center 319 Cambridge, MA 02142
Boston, MA 02163 and NBER
USA aschoar@mit.edu
okim@hbs.edu
Jonathan A. Parker
MIT Sloan School of Management
100 Main Street, E62-642
Cambridge, MA 02142-1347
and NBER
JAParker@MIT.edu
A data appendix is available at http://www.nber.org/data-appendix/w28151
The small business sector is a critical driver of the U.S. economy, contributing two-thirds
of net job growth and 44 percent of U.S. economic activity (SBA, 2019). But unlike most
large firms, small businesses are typically not owned by diversified shareholders but by a
few individuals whose standards of living are often closely tied to their businesses (FRBNY,
2019). Small businesses are also riskier, typically contracting earlier and more severely than
large firms during economic downturns, partly because they operate with lower overhead
and fixed costs (Davis et al., 1996). Thus, the Covid-19 pandemic presented unprecedented
challenges for small businesses and their owners. This economic strain, stemming from
demand collapses, supply chain disruptions, and production slowdowns due to unsafe
work environments was a major concern during the pandemic and a driving force behind
the enactment of the Coronavirus Aid, Relief, and Economic Security (CARES) Act of 2020.
In this paper, we measure the declines in small business revenues, costs, and profits during
the pandemic, and measure how exposed the consumption spending of small business
owners was to these declines in their businesses.
Using de-identified data on the checking, credit-, and debit-card accounts of small
businesses and households from JPMorgan Chase (JPMC) bank account records, we docu-
ment that local infection rates and Non-Pharmaceutical Interventions (NPI) caused only
modest declines in revenues for most small businesses and that these revenue losses had
limited impact on the consumption of the owners of these small business during the first
eighteen months of the pandemic. We use transaction-level information to construct a
monthly panel dataset on the revenues, expenses, and profits of small businesses linked
with the consumption of their owners from January 2019 to the end of September 2021.
We categorize spending using the counterparty of financial account transactions, and use
these components of spending to shed light on how and why businesses and their owner
households adjusted spending patterns. Unlike other datasets covering small businesses,
our sample of businesses has significant coverage of non-employer businesses and closely
mirrors the national distribution of the small business sector.
We first document the enormous impact that COVID-19 pandemic had on small busi-
nesses and their owners. Average revenues fell by over forty percent in the first three
1
months following the declaration of a national emergency on March 13, 2020, consistent
with other studies reporting steep declines in small business employment (Bartik et al.,
2020a; Chetty et al., 2024) and finances (Farrell et al., 2020b). After this initial drop, aver-
age revenues rebounded steadily until March 2021, with a sharp increase following the
widespread availability of COVID-19 vaccines in April 2021. Average expenses tracked
revenues closely, indicating that businesses quickly downsized in response to the sudden
revenue drop and scaled back up as revenues recovered. Similarly, consumption spending
of small business owners declined by more than forty percent on average in the first quar-
ter of the pandemic, rebounded to pre-pandemic levels within five months, and surged
significantly after the vaccine rollout.
The large declines in business performance and owner consumption during the first
two months following the national emergency were similar across businesses with different
liquidity buffers and employer versus non-employer status. However, we find a greater
divergence in revenue and consumption patterns based on business size and industry.
Larger businesses and those in non-essential industries experienced more pronounced
revenue declines than smaller businesses or those in essential industries.1 These differences
persisted through the recovery period until September 2021. Small or essential businesses
began recovering as early as three months into the pandemic, while larger or non-essential
businesses continued to see revenue declines until the end of 2020.2 After the vaccine
rollout, businesses in non-essential industries saw a stronger rebound in average revenues
compared to essential industries.
Second, we show that the collapse of small business revenues and the consumption of
their owners was predominantly explained by nationwide factors, and that only a small
part of the collapse was related to county-level variation in new infection rates or state-level
variation in policies such as Shelter-in-Place (SIP) orders. Following the declaration of the
national emergency, both new infection rates and state-level policies differed significantly
across counties and states and had low correlation with each other. Regressing outcomes
1 Essential businesses are classified according to the CISA guideline and verified via news searches.
2 This finding is consistent with Bartlett and Morse (2021), who find that microbusinesses show greater
revenue resiliency relative to bigger small businesses.
2
onto infection rates and SIP orders and including weekly time fixed effects to control for
national factors, we find that local infections and SIP orders have only modest direct effects
on small businesses and their owners. Specifically, a one standard deviation increase in
new infections leads to a 0.5 percent decline in business revenues. The introduction of a
SIP order leads to an 11 percent incremental decline in business revenues and a 5.4 percent
decline in owner consumption.
One possibility for the relatively moderate direct effects of SIP orders is that states
adopted a variety NPI policies, so focusing solely on SIP effects may underestimate the
broader impact of NPIs. However, our results hold when using an alternative summary
measure of ”NPI strictness” that captures the intensity of state-specific packaged NPI
policies, with a standard deviation increase in NPI strictness reducing business revenues,
expenses, and owners’ consumption by less than 5 percentage points.3 Our finding that
NPIs had modest effects where they were imposed is consistent with Correia et al. (2022),
which finds that cities with stricter NPI policies did not perform worse than those with less
stringent policies during the 1918 flu pandemic. Compared to Correia et al. (2022), which
attributes the main source of economic disruption to the pandemic itself, we find more
modest effects of local disease incidence.
Our final and most novel result is that the decline in a small business’s revenues only
had a small direct effect on the consumption spending of its owner. The challenge in
identifying the magnitude of the consumption response to revenue declines – particularly
in the first couple of months of the pandemic – is that the pandemic affected household
consumption decisions not only through the drop in business revenues but also directly
through the local prevalence of the disease and the restrictions on mobility and consumer-
facing businesses. For example, business owners in hard-hit areas might experience a
revenue drop but also reduce spending due to the infection risk when leaving the house.
Indeed, for wage-earning households, Cox et al. (2020) finds a reduction in spending
on average in the initial months of the pandemic, indicating that the drop in household
spending extended even to those whose income was not directly affected by the pandemic.
3 Specifically, we construct a composite measure of NPI strictness that reflects a state’s NPI strictness
relative to other states using principal component analysis.
3
We estimate the propensity of owners to reduce consumption in response to revenue
declines using a two-stage least squares (2SLS) strategy that compares businesses in the
same county and month, which experience varying degrees of revenue losses due to local
infections and SIP orders, because they are in different industries. Our identification
strategy relies on the idea that the direct effect of local factors on the consumption of
business owners is driven by local conditions that affect households independently of
their own business revenues. To control for this endogenous variation in consumption, we
include county x time effects in our regression. Furthermore, because consumption and
revenues are co-determined for any given business, we use the interactions between 4-digit
NAICS industry indicators and local infections and SIP orders to generate revenue changes
that are plausibly orthogonal to any remaining endogenous variation in consumption. The
changes in revenues for the five least affected industries since the onset of the national
emergency range from zero to a 20 percent increase, while the changes in revenues for the
five most affected industries are as large as a 90 percent decline.
Our estimates show that for every dollar decline in business revenues or expenses,
the owner of that business decreases consumption spending by only 1.6 and 2.6 cents,
respectively. To put these estimates in perspective, the average monthly drop in revenues
during the lockdown period, relative to its pre-pandemic average, was -$2,400. Therefore,
the implied consumption drop due to revenue losses is roughly $36. This decline accounts
for less than 10 percent of the average monthly change in consumption during the lockdown
period, which was $396. We find little heterogeneity in the pass-through response from
business revenue to owner consumption based on employer status and ex-ante liquidity.
However, owners of smaller businesses or pass-through entities exhibit greater sensitivity
to revenue losses than owners of larger or incorporated businesses.
Our findings are consistent with federal fiscal support providing liquidity to stabilize
owners’ consumption and the pandemic limiting spending opportunities, especially in the
early phase. Consistent with the important role of the CARES Act, cash balances in both
business and personal accounts rose sharply in April and May of 2020 when the Economic
Impact Payment (EIP) and Paycheck Protection Program (PPP) grants were disbursed and
4
remained high until the end of our sample period. The pandemic imposed significant
restrictions on consumers’ ability to spend, particularly in its early months, which could
make consumption relatively insensitive to revenue declines.4 Consistent with this, the
magnitude of revenue-to-consumption pass-through more than doubled after the vaccine
rollout relative to the onset of the pandemic.
In sum, the large average drop in the consumption of small business owners appears to
be largely divorced from the specific performance of their individual businesses and mainly
driven by the national crisis. The low direct effects of business revenues on consumption
seem to be due both to pandemic-related restrictions on consumption (at least in the early
stages of the pandemic) and to the government’s large fiscal responses. Thus although
the common view is that the standard of living of a small business owner is closely tied
to the success of their business, this was not the case during the pandemic. And while
businesses and their owners’ cash balances remained elevated until the end of our sample
period, the pass-through of business losses into owners’ consumption may rise over time
as government funds are exhausted.
Related literature Our study complements a number of studies of the impact of the
pandemic on small businesses by providing well-identified estimates of the effect of local
infections and NPIs on small business revenues and expenses and by quantifying the
extent to which disruptions to business revenues impacted the living standards of their
owners. Several papers conducted surveys of small businesses and their owners in the
early phases of the pandemic and find mass layoffs, temporary closures, and downsizing
of businesses.5 Bartik et al. (2020a) finds that 43 percent of businesses were temporarily
closed by the end of March, 2020, and estimates from the U.S. Census Small Business Pulse
Survey show that only 25 percent of firms had enough cash on hand to cover 3 months
of operations at the end of May (U.S. Census, 2020). Bartlett and Morse (2021) finds that
larger small businesses experienced 14 percent higher revenue declines than smaller small
4 We find disproportionate declines in spending on goods and services that are luxuries or that require close
personal contact, such as travel, eating out, or personal services in the first three months of the pandemic.
5 Despite the negative impact on businesses, Wang et al. (2022) finds that small business bankruptcy filings
remained low, partly due to the availability of generous unemployment insurance programs.
5
businesses, and Humphries et al. (2020) reports that 60 percent of small businesses had laid
off at least one worker by the end of April (see, also, Bartik et al. (2020b), Balla-Elliott et al.
(2022), and Fairlie et al. (2023)). Alekseev et al. (2022) documents that increased household
responsibilities, such as taking care of children and self-isolating household members,
affected business owners’ ability to focus on work during the crisis. Studies that make
use of administrative data document a sudden 12.7 percent drop in median business cash
balances at the onset of the pandemic (Farrell et al., 2020b) and substantial heterogeneity in
the speed of recovery by owner’s race or by income profiles of the neighborhood in which
businesses operate. African-American businesses recovered at a slower rate than White-
owned businesses (Fairlie, 2020) and business located in less affluent areas experienced
smaller revenue losses than those located in more affluent areas (Chetty et al., 2024).6
Although our focus on small business income and owners is novel and highlights the
risks associated with small business ownership, our work also contributes to the extensive
literature estimating the marginal propensity to consume (MPC) out of transitory income
shocks. This literature typically examines the extent to which households smooth transitory
income fluctuations generated by factors such as the randomized timing of economic stim-
ulus disbursements (Parker et al., 2013; Broda and Parker, 2014), the timing and amount of
tax refunds (Baugh et al., 2021), variations in income losses due to unemployment insurance
(Ganong and Noel, 2019), or changes in income net of installment debt payments (Stephens,
2008; Di Maggio et al., 2017).7 In the context of the pandemic, we examined an income
shock that was both more unusual and less predictable than those previously studied.
For example, while the existing literature shows that households (mainly wage earners)
saved a significant fraction of Federal economic impact payments (Parker et al., 2022),
those experiencing greater income declines exhibited stronger MPC responses (Baker et al.,
6 Fairlie (2020) documents that African-American business owners experienced a drop of 26 percent in
business activity from pre-COVID-19 levels compared to only an 11 percent drop for White business owners
by May, 2020. Farrell et al. (2020a) similarly finds that cash balances of White-owned restaurants doubled in
May compared to only 38 percent increase for Black-owned restaurants.
7 In addition to income shocks, there is a substantial literature measuring spending responses to changes
in non-labor wealth (e.g., lottery winnings (Golosov et al., 2024; Kotsogiannis and Sakellaris, 2024) or stock
market wealth (Di Maggio et al., 2020)) and in debt capacity (e.g., credit card limits (Gross and Souleles, 2002;
Aydin, 2022) or mortgage values (Ganong and Noel, 2023)).
6
2023) and faster spending rebounds during recovery (Cox et al., 2020) (see Yannelis and
Amato (2023) for a comprehensive overview). Given that business owners’ incomes were
more severely affected than those of wage earners, coupled with substantial uncertainty
surrounding the timing of recovery in the small business sector, one might expect higher
MPC responses compared to prior studies. However, this variation occurred within a
context in which income fluctuations were associated with reduced spending capacity and
significant public assistance. As a result, the predicted magnitude of pass-through remains
ambiguous and presents an open question.8
The remainder of the paper proceeds as follows. We describe the data used in this
study in Section 1, including our sample construction procedures and the definition of
primary outcomes considered in the study. Section 2 presents descriptive evidence of how
small businesses performed and owners’ consumption evolved in the early months of the
pandemic. We present our main findings on the effect of infections and NPIs in Section
3. In Section 4, we describe our estimation strategy for quantifying the causal effect of
revenue losses on owners’ consumption and present the pass-through estimates. Section 5
discusses potential explanations behind the modest average pass-through effect of business
revenue losses. Section 6 concludes.
1 Data
Our analysis makes use of de-identified financial account data provided by JPMorgan
Chase Institute (JPMCI). We use transaction-level data from both small business accounts
and personal accounts to construct a panel dataset on the revenues, expenses, and net
income of small businesses linked with the consumption of their owners. Our final dataset
provides monthly business outcomes and household consumption for 380,532 businesses
and 333,128 business owners between September 2019 and September 2021.
8 While our proxy for owners’ consumption captures comprehensive spending patterns, it reflects spending
responses rather than pure consumption responses, as durable expenses are excluded.
7
1.1 Samples
We start by constructing a dataset of the universe of small business checking accounts.
We define a small business as a collection of small business checking accounts linked to the
same signer of the account.
The all businesses sample. We apply several screening criteria to identify active busi-
nesses that primarily use financial accounts provided by JPMC to manage their business
finances. First, we exclude businesses with more than two business checking accounts and
those with multiple locations or industry assignments.9 We next apply several account
activity filters to ensure that the set of firms we consider are actively operating businesses
prior to the pandemic. We limit the sample to firms with ”open” business checking account
status for at least twelve consecutive calendar months. We also require that a checking
account has at least three transactions per month for at least ten months in 2019 (i.e., our
baseline period). Finally, we require a business to have an ”open” account status for at least
one month in 2020. This reduces our sample of 3.44 million small businesses to 2.4 million
businesses with active accounts as of the beginning of 2020. This sample is henceforth
referred to as the all businesses sample.
The business owners sample. From this all businesses sample, we next create a sub-
sample of accounts for which we can match at least one of the owners of the small business
to a personal account at the same large financial institution. We construct a business owners
sample of paired small-business accounts and personal accounts where each observation
represents a small business matched with the personal accounts of one of its owners.
Specifically, we start from the all businesses sample and apply several additional screening
criteria to ensure that the set of owner households we consider also use financial services
provided by JPMC to manage their personal finances. We also require that a business owner
has an active personal checking account (or accounts) that is ”open” for at least twelve
consecutive months and has at least three transactions during all months in 2019. Relative
9 Firms with more than two business checking accounts are rare. We drop these firms as they are more
likely to have multiple industry and location assignments.
8
to the business account activity filter that we impose – i.e., at least three transactions for
ten months in 2019– we require personal accounts to have activity in all months. This
is because business account activity tends to be more volatile due to variations in cash
flows and seasonality, whereas personal accounts are not subject to the same concern. This
procedure leads to a sample of 363,682 small business-owner pairs.
Further subsamples We categorize observations in the business owners sample by type
of business along the following dimensions: employer versus non-employer, essential
versus non-essential, small versus big, and low versus high liquidity.
A business is categorized as employer if it has payroll expenses for at least 6 months in
2019. In our sample, 11 percent are employer firms. We next identify essential and non-
essential businesses based on their 4-digit North American Industry Classification System
(NAICS) sub-sector. Specifically, we categorize businesses that operate in sectors classified
as “critical workforce” by the Department of Homeland Security (HLS)’s advisory list as
essential businesses. We make a few exceptions to this list.10 Namely, we categorize several
sub-sectors in the food and agriculture industry, such as bakeries, caterers, or full-service
restaurants, as non-essential because they are heavily affected by stay-at-home restrictions
even if food and agriculture sectors were considered to be essential and technically not
closed. According to this measure, roughly 60% are essential businesses.
We use two measures of firm size. In our descriptive statistics, we define firm size based
on its average monthly revenues during 2019. Businesses with average monthly revenue
below the first tercile are classified as smaller, and those with weekly revenues greater
than the third tercile are classified as larger. For later estimation (in section 4), we define
firm size based on the within-industry distribution of annual revenue in 2019. Since our
estimation exploits differential industry exposure to local NPIs and infection rates, using
the within-industry distribution provides a better measure of relative size. Businesses with
2019 revenue below the median for their industry are classified as small, and those above
10 This list is intended to help local officials make informed decisions, so individual jurisdictions may differ
in their own requirements of essential versus non-essential distinctions. It is nonetheless a good proxy for
whether a business is considered to be essential at the local level.
9
the median for their industry are classified as large.
Business liquidity is computed as the ratio of average cash balances at the end of each
month in 2019 to average monthly expenses in 2019. We then multiply this figure by thirty
to express liquidity as the number of days of operating expenses that a business could pay
out of its cash balances were its revenues to stop. Businesses are classified based on the
within-industry distribution. Businesses in the bottom quarter of the distribution of cash
buffer days within its NAICS4 sub-sector are classified as low liquidity, those in the top
quarter are classified as high liquidity.
Supplemental Data. We supplement this administrative financial accounts data with
county-level infections data from the New York Times (New York Times, 2020) and state-
level non-pharmaceutical interventions from Keystone Strategy (Keystone, 2020) and the
U.S. Department of Health and Human Services.
1.2 Measurement of Business and Owner Outcomes
Our dataset captures businesses and their owners’ financial activity on their personal
and business credit cards, debit cards, and checking accounts. We construct monthly
business revenues, expenses, profits, and household consumption from this dataset.
Business outcomes of firms To construct operating revenues for each small business, we
first compute total credit transactions (i.e., inflows) into business checking accounts for
each firm and month. We next identify financial transactions or non-business income that
are unlikely to represent operating revenues received from providing goods and services,
and subtract these amounts from total inflows.
Operating Revenue = Total Inflows − Financial Inflows − Non-business income (1)
Financial inflows include any inter-personal transfers, fee reversals, or miscellaneous
account activities such as SWEEP inflows or loans from financial institutions. Non-business
10
income includes government transfers, such as unemployment insurance, tax refunds,
veterans benefits, income from gig platforms, or other interest income.
To construct operating expenses, we categorize all debit transactions (i.e., outflows) on
business checking, debit card, and credit card accounts for each firm and week.11
Operating Expense = Fuel + Equipment + Groceries + Materials + Retail+
Retail Durable + Wholesale + Entertainment + Food + Insurance + MiscBizExpense+
Services + Travel + Payroll + Tax + Debt Payment + Utilities+
Cash Withdrawal + Check + Uncategorized (2)
Spending categories are identified using a combination of transaction tags provided by
JPMCI, such as the Merchant Category Code (for spending on cards), the identity of
a transaction counterparty, or the channel of payment. We are able to classify detailed
categories of business expenses but not of operating revenues because counterparty identity
for credit transactions is often redacted to preserve the anonymity of the business.
We code both operating revenues and expenses as continuing zeros following account
closures. This approach eliminates a possible survivorship bias that could spuriously make
business outcomes appear better by dropping exiting firms that have performed the worst
through the pandemic. We define Profit as the difference between revenues and expenses
and Profit Margin as profit divided by average operating revenue in 2019.
Finally, we infer business exit from a closure of business checking accounts or from its
inactivity. Specifically, if a business is inactive (near zero revenues, expenses, or balances)
from t until the end of the sample period, we consider this firm as having exited even if the
business account is not officially closed out. Exit is defined as a binary variable that equals
one if the small business closed and zero otherwise. If a business has two deposit accounts,
both accounts must be closed to be coded as having exited.
11 Deposit account transactions refer to non-debit checking account transactions, while debit transactions
refer to those using debit cards.
11
Household consumption of owners We construct business owner households’ consump-
tion by categorizing all debit transactions on the owners’ personal deposit accounts, debit
card, and credit card accounts:
Consumption = Fuel + Groceries + Pharmacy + Retail + Retail Durable+
Auto Repair + Insurance + Medical + Entertainment + FoodAway + Personal Svcs+
Professional Svcs + Other Svcs + Travel + Rent + Gov’t + Utilities+
Cash Withdrawal + Check + Uncategorized (3)
For credit card spending, we follow Ganong and Noel (2019) and measure spending
as of the time when the goods and services are purchased rather than when the card bill
is paid. In addition to household consumption, we categorize and track household debt
payments as they are major household expenses. We supplement business and household
outcomes with demographic information about businesses and their owners, such as the
owner’s gender and age, business industry, incorporation type, and location.
Relative to other account-level analyses, one advantage of our use of linked small-
business and personal accounts is that we observe some business expenses in personal
accounts and some personal expenses in business accounts and can re-classify them. That is,
there are instances where it appears that the business owner uses their business accounts for
transactions that are clearly for personal use (e.g., child care, medical expenses, hair salon,
etc). We exclude these transactions from operating expenses and re-categorize them as
household spending. Similarly, when we observe business spending (e.g., payroll, business
insurance, etc) from personal accounts, we re-categorize them as operating expenses.
One disadvantage of account-level data, in general, is that we cannot capture business
activities or household spending patterns if a business or household has financial accounts
with another financial services company. However, given that the sample of households
we study have both their business and personal checking accounts provided by JPMC
to manage their finances, combined with our activity filters, we believe that the account
activity that we can track in our data captures the majority, if not all, of their business and
12
personal finances.
Our dataset does not provide information beyond financial activity on credit cards,
debit cards, and checking accounts. For example, we do not observe firms’ balance sheets
or households’ brokerage or investment accounts. Thus, our dataset primarily captures
cash flows and short-term liquidity of businesses and their owners rather than their total
wealth, including illiquid assets (e.g., housing, land, physical capital, retirement accounts).
Scaled outcome variables To compare businesses of different sizes, we normalize our
outcome variables (except profits) using two alternative scaling factors. The first scaling
factor is the monthly average of the outcome in 2019. We denote variables scaled with this
factor by the superscript avg (for 2019 average). Thus, for business-owner pair i in month t:
avg Yi,t
Yi,t = (4)
Y i,2019
1
where Y i,2019 = 12 ∑s∈2019 Yi,s and Yi,s represents operating revenue, expenses, or con-
sumption. This normalization allows us to capture changes in outcome relative to a
firm-household pair-specific constant baseline.
The second scaling factor is the centered 3-month average of the same outcome a
year ago. We denote variables scaled with this factor by the superscript sa (for seasonal
adjustment):
sa Yi,t
Yi,t = (5)
Y i,(t−13,t−11)
t−11
where Y i,(t−13,t−11) = 31 ∑ss=
=t−13 Yi,s . The second normalization factor has the advantage of
adjusting for seasonal fluctuations so that one is comparing the firm-household outcome
to a similar period in the previous year. We take the 3-month average so that the scaling
factor does not add month-to-month volatility.
Unlike revenues, expenses, or consumption, each scaled by its own monthly average
in 2019, we transform profits (i.e., Revenuei,t − Expensei,t ) into a profit margin measure,
or percentage of sales turned into profits, by normalizing profits by monthly average or
13
centered-3 month average of operating revenue a year ago. All of the scaled and unscaled
outcome variables are winsorized at the 2nd and 98th percentile.
1.3 Descriptive Statistics
Businesses in the business owners sample are smaller – both on average and across
the distribution – relative to those in the all businesses sample, as shown in Panels A (all
businesses sample) and F (business owners sample) of Table 1. Panels B through E show
the distributions of revenues, expenses, and profits for subsamples of all businesses sample.
Average revenues and expenses track each other closely for all types of businesses, with
the exception of smaller small businesses (Panel D), which have higher average monthly
expenses than revenues. Employer businesses (Panel C) tend to be larger and are similar in
size to large small businesses (those in the top tercile of monthly revenues in 2019, Panel E).
Across all business types, a median firm does not break even (median profits are negative).
Consumption represents about 23% of business revenues and expenses (Table 1 Panel F).
At the bottom of Table 1, we show that businesses in the all businesses sample are
concentrated in 5 industries, in which more than half of all businesses in the sample
operate. A large share of businesses (more than 76%) are pass-through entities, and roughly
30% of all businesses are female-owned.
How do the characteristics of the small businesses that use our financial institution
compare to the national distribution? Table 2 compares our data to external benchmarks.
Roughly 85% of businesses sampled in our data are nonemployer businesses, similar to 81%
in the U.S. overall. Thus, a key advantage of our data is better coverage of nonemployer
businesses relative to traditional data sources. However, relative to the benchmark, our
sample under-represents businesses that have existed for more than ten years and over-
represents businesses that operate in professional services, real estate, and transportation
sectors. Among nonemployer businesses, our sample of firms tends to be bigger in terms
of annual receipts relative to the nationwide distribution.
14
1.4 Measurement of Infections and NPIs
To estimate the effect of local disease prevalence on business performance and owner’s
consumption, we obtain measures of new infections from New York Times (2020). We
aggregate new cases in every county from the daily to the monthly level and divide by
ex-ante population to obtain the monthly rate of new infections per 1, 000 residents at the
county level:
New Casesc,t
LIRc(i),t = × 1000 (6)
Populationc
where LIR denotes local infection rate, c indexes counties and c(i ) denotes the county in
which business i is located. For studying exit rates, we cumulate this variable across weeks:
LIRCum
c(i ),t
= ∑ts=0 LIRc(i),t . It is important to note that the infection rates that we use may
underestimate the true infection rates due to limited testing capability or efforts. However,
these rates reflect the available public information about the prevalence of the disease.
We obtain measures of NPI policies at the state level from Keystone (2020) and the U.S.
Department of Health and Human Services.
2 Average business and owner outcomes
This section shows that the performance of small businesses and the consumption
of their owners declined dramatically in the early months of the pandemic. Following
this analysis, Section 3 studies the relative roles of national vs. local infection rates and
state-level policies in these severe declines. Section 4 then measures the extent to which
owners’ living standards are affected by their own business’ revenue losses and how this
differs across businesses.
Our data shows sharp declines in all measures of economic activity following the
declaration of the COVID-19 national emergency on March 13, 2020. Figure 1 presents
the average monthly dollar amounts (unscaled) of business revenue, expenses, profit,
and owner’s consumption for our sample of business owners. Figure 2 illustrates the
15
average percent change in seasonally adjusted outcomes for this same sample, relative
to the six-month average preceding the national emergency. In the first month, revenues,
expenses, and consumption decline by over 40 percent, before rebounding to pre-pandemic
levels about three months later as shelter-in-place (SIP) orders are lifted. These outcomes
experience a sharp increase in May 2021, following the nationwide expansion of COVID-19
vaccine eligibility to all residents aged 16 and older on April 19, 2021.12
Figure 3 shows the average changes in business revenues across different types of
small businesses. Panel A shows that essential businesses experienced a 40% decline in
revenues one month into a national emergency, whereas non-essential businesses saw a
significantly larger drop. This discrepancy is likely due to non-essential businesses being
more affected by consumer and worker responses to the pandemic and facing stricter
operational restrictions imposed by local governments. The performance gap between
essential and non-essential businesses remains relatively stable until March 2021, after
which non-essential businesses experience a substantial revenue increase of 200% following
the vaccine rollout. Panel B indicates that larger small businesses faced greater revenue
declines compared to their smaller counterparts. However, the differences in revenue
changes between employer and nonemployer businesses (Panel C), as well as between
businesses with high and low liquidity levels (Panel D), are relatively minor.
The initial collapse in business revenues, expenses, and profits is unlikely to be driven
by business exits, as exit rates remained low during the first year of the pandemic. Figure
4 shows the monthly survival rates of businesses. We define a firm as having exited if it
closed its business checking account or remained inactive from a specific time point until
the end of our sample period. At the start of 2020, nearly 99 percent of firms had not exited.
However, exit rates began to increase in March 2021, and by the end of our sample period,
approximately 7 percent of firms had fully exited. A potential reason for the delayed exits
could be federal stimulus programs, such as the Paycheck Protection Program (PPP), which
provided ample liquidity to help businesses weather the economic downturn in 2020.
12 While the set of businesses in the business owners sample tend to be smaller than those in our all
businesses sample, we find that the experiences in the two samples are very similar. Appendix Figures A.1
and A.2 repeat Figures 1 and 2 for a random subset of all businesses and show a very similar pattern to that
of our business owners sample in the early stages of the pandemic.
16
3 The Effects of Infections and State-level Policies
This section demonstrates that both business outcomes and owner consumption de-
clined sharply as local infection rates rose and states implemented policies to curb the
spread of the virus. However, the direct effects of infections and state policy responses on
business and consumption outcomes appear modest when comparing businesses across
counties with differing levels of exposure to infections and policy interventions.
3.1 Effects of Covid-19 Infections and Shelter in Place Orders
We estimate the impact of infection rates and SIP orders by comparing businesses across
geographic areas with varying infection rates and NPI policies. There was considerable
heterogeneity in the incidence of infection rates and NPI policies across regions. Figure 5
shows the share of states with NPI orders by month, illustrating the variation in the timing
and imposition of these measures across states. Infection rates also varied significantly
across counties. Figure 6 shows the evolution of infection rates per 1,000 residents in
three counties representing low, medium, and high-growth regions. Panel A displays new
cases, and panel B shows cumulative cases per 1, 000 residents. The counties with low
and medium infection rates (Glenn and Nowata) experience an order of magnitude lower
infection rates than those of New York, the county with the highest caseload.
We quantify the impact of infections and shelter-in-place (SIP) orders on businesses and
their owners’ consumption by estimating variants of the following regression model:
Yi,t = αi + β 1 LIRc(i),t + β 2 1[SIPs(i),t ] + Xi,t + ei,t (7)
where Yi,t denotes our outcomes of interest (as defined in Section 1.2), and our main
explanatory variables, LIRc(i),t and 1[SIPs(i),t ], represent local infection rates per 1,000
residents and an indicator for whether a state has a SIP order in effect, respectively. We
include firm fixed effects, αi , in all specifications to account for time-invariant, unobserved
differences across firms. We also include different sets of fixed effects, denoted by Xi,t ,
to compare firms within the same month, industry, and/or size category. While states
17
implemented a variety of NPIs in response to rising infection rates, we focus on the effects of
shelter-in-place (SIP) orders, which were among the most common and stringent measures.
SIP orders imposed direct restrictions on both the ability of many businesses to operate
and the ability of individuals to consume various goods and services.
Local infection rates and SIP orders together explain the majority of the decline in
business outcomes and owners’ consumption in a naive regression that does not account
for nationwide factors. These results are reported in Table 3. The first four columns report
avg
the estimates with dependent variables normalized by the prior-year average, Yi,t , while
the last four columns report estimates using seasonally adjusted dependent variables,
normalized by the centered three-month average of the respective outcomes from the
sa . The reported coefficients capture percentage point change relative to
previous year, Yi,t
the mean of the previous year. Columns 1 and 5 of Table 3 shows that each new case per
1, 000 residents is associated with a 0.3 or 0.4 percentage point decline in monthly business
revenues while the imposition of a SIP order leads to a 27 percent decline in revenues.
Local infection rates and SIP orders account for only a small portion of the significant
decline in business revenues once nationwide (i.e., time) effects, as well as sectoral and size
differences, are controlled for. Columns 2 and 6 of Table 3 include time (month-year) fixed
effects in Xi,t . Each new case per 1, 000 residents in a week leads to a 9 percent decline in
revenues in that month, and an SIP order leads up to an 11 percent decline in revenues.
The infection effect is economically small in magnitude: a standard deviation increase in
infection rates leads to a 0.5 percent decline in revenues (≈ 6.16 cases per thousand ×0.08).
The remaining columns show that the estimated effects remain stable and robust across
different specifications that include time × industry or time × size bin fixed effects.
We similarly observe modest effects on business expenses, profits, exit rates, and the
consumption of small business owners. Panel B shows that the impact of infections and
SIP orders on business expenses was comparable to their effect on revenues, suggesting
that businesses immediately scaled down, potentially by depleting inventory, postponing
bill payments, or ”eating” into capital. Panel C shows that while SIP policies had a small
negative impact on profits (up to a 2 percent decline), infections had a small positive effect
18
on profits. Panel D shows that consumption declined by 0.13 percent per new case per
thousand residents. Similarly, state-level SIPs reduced consumption by 1.7 or 5.6 percent,
depending on the specification. Panel E shows that local infections and SIP had little effect
on exit rates.13
Overall, comparing businesses in different geographic regions exposed to varying levels
of infection rates and SIP orders shows that infection rates and SIP orders had minimal
direct effects on business performance once nationwide factors are controlled for. However,
SIP orders represent only one of many policies states implemented to limit the spread of
the virus. Given that many states adopted multiple NPI policies, focusing solely on SIP
effects may underestimate the broader impact of NPIs. We next explore the broader impact
of NPIs.
3.2 Effects of Infections and NPI tightness
We construct a summary measure of NPIs and show that the conclusions of the previous
subsection – that both local infections and NPIs had modest effects – still hold for this
alternative measure of state policies. Figure 5 shows that many states adopted multiple
NPI policies at the same time. Since packaged policies can reinforce and complement one
another, correlated policies and heterogeneity in policy duration across states complicate the
measurement of NPI effects. We address this challenge by conducting Principal-Component
analysis (PCA) and constructing a simple ”NPI strictness” measure that captures the
intensity of state-specific packaged NPI policies relative to other states.
We perform PCA on the NPIs listed in Figure 5 during the period after the national
emergency is declared. Appendix Table A.1 reports the detailed PCA results. We focus on
the first principal component (denoted Strictness), which explains 58 percent of variance
and weighs positively on all restrictions.14 We estimate the impact of infections and NPI
13 Figure A.3 helps to visualize this result by showing that the majority of early declines in business
performance are explained by nationwide factors and that these effects are not persistent.
14 The first principal component explains up to 65 percent of variance in the first 6 months of the pandemic.
19
tightness on business outcomes and owner consumption using the following specification
Yi,t = α + αi + β 1 LIRc(i),t + β 2 Strictnesss(i),t + Xi,t + ei,t (8)
which is analogous to equation (7).
We find that NPI strictness has effects on business and owner outcomes that are very
similar to the modest effects that we found for SIP orders. And the use of NPI strictness
in place of SIP does not alter any of our conclusions about the modest effect of infections.
Table 4 reports the effect of infection rates on outcomes (simply the estimated β 1 ) and the
effect of NPI strictness per standard deviation increase in NPI strictness (the estimated β 2
times the standard deviation of the first NPI factor). A one standard deviation increase in
NPI strictness reduces business revenues, expenses, and owner’s consumption by less than
two percentage points when we account for time effects.
In sum, we find only modest effects of local infection rates and SIP policies on business
revenues and owner consumption, suggesting that the primary drivers of business dis-
ruptions were national factors rather than local infections. Moreover, the low correlation
between county-level infections and state-level policies allows us to reasonably estimate
the contribution of each separately. Our finding of a weak effect of SIP policies aligns with
Correia et al. (2022), who report that cities with stricter NPI measures did not perform
worse than those with less stringent measures, indicating that the main source of economic
disruption was the pandemic itself rather than NPIs during the 1918 Flu pandemic.
4 The Effect of Revenue Losses on Owner Consumption
This section investigates the causal impact of declines in individual business revenues
on the consumption of business owners. While local infection rates and state-level policies
had modest direct effects on business performance and owner consumption, we observe
substantial heterogeneity in business performance across industries. We analyze how these
variations in business outcomes pass through to the living standards of business owners.
20
4.1 The Effect of Business Revenues on Owner Consumption
Small businesses experienced substantial variation in performance across industries,
but the consumption of their owners remained strikingly similar across industries. Figure 7
plots changes in average revenues and owner consumption for the least and most affected
NAICS 4-digit industries. In the first two months of the pandemic, average revenues for
the least affected industries (solid blue line) remained stable, whereas those for the most
affected industries (dashed blue line) plummeted by nearly 90 percent. Despite this sharp
decline, the consumption patterns of business owners showed minimal variation across
sectors, and this trend persisted even after the vaccine rollout in April 2021.15
The stark disconnect between the severity of revenue shocks and adjustments in owner
consumption motivates the causal analysis of how well owners could insulate their spend-
ing from firm-specific revenue losses. While the earlier figure hints at owners’ self-insurance
capabilities, geographic differences in infection rates and NPIs could obscure the direct
effect of business performance on consumption. If, for instance, the least affected industries
are concentrated were high-infection areas and the most affected ones were in low-infection
areas.16 To address this, we conduct a regression analysis that accounts for local conditions.
To estimate the decline in owner consumption caused by their business’s revenue drop,
we compare owners of businesses within the same county but in industries differentially
impacted by local infection rates and SIP orders. The key assumption is that industry-
specific exposure to these local shocks does not directly influence owner consumption
except through its impact on business performance. For example, restaurants are more
affected by local conditions than chemical manufacturing, creating revenue differences
15 Appendix Figure A.4 also plots average revenues and consumption for the bottom and the middle three
industry performance deciles and show similar patterns in the early stages of the pandemic. Appendix
Figures A.5 and A.6 plot changes in revenues and consumption for each individual sector.
16 For example, if some of the least affected industries are located in areas with high infection rates (e.g.,
finance in Boston), where consumption dropped the most, while some of the most affected industries
are concentrated in regions with low infection rates and minimal NPIs (e.g., potato farmers), we might
observe a negative relationship between industry revenue declines and owner consumption. In this scenario,
the consumption of owners in the least affected industries would be more heavily influenced by high
local infection rates and stringent NPIs, reducing their consumption. This negative correlation would
counterbalance the direct impact of revenue declines on owner consumption, making it appear as though
changes in business performance have little effect on owner spending.
21
driven by industry exposure rather than the owner’s personal consumption behavior.
We instrument for business revenues, expenses, and profits using industry-specific
exposure to local infection rates and state-level NPIs. The exclusion restriction is that the
differential effects of infections and NPIs on owner consumption operate solely through
their impact on the owner’s business industry. This leads us to estimate the following
two-stage least squares (2SLS) regression model:
Yi,t = αi + ∑ β FS
j 1[ j= j(i )] NPIs(i ),t + ∑ δj 1[ j= j(i )] LIRc(i ),t + γc(i ),t + ei,t
FS
(9)
j j
Ci,t = αi + β IV Y
ci,t + γc(i ),t + ηi,t (10)
where the two key variables in equation (9) are interactions of NAICS 4 industry indicators
with state-level NPIs and with county-level infection rates, respectively. These terms
measure the industry-specific effect of local infections and state-level policies on business
outcomes. The term γc(i),t represents month × county fixed effects, which control for
differences in the average effect of infections and NPI on revenues through all channels, as
well as for location-specific industry effects. In the second-stage equation (10), we use the
same fixed effects to isolate the causal effect of revenue declines on owner consumption
based solely on industry-level differences in revenue responses to local infection rates and
state-level policies within a given county and time period.
Table 5 shows the results from estimating equation (9) and (10) with Yi,t and Ci,t mea-
sured in levels (dollars) to directly assess the marginal propensity to adjust consumption
in response to business losses. In this table, the odd columns use variation in industry
exposure to SIP or NPI strictness, while the even columns use both industry exposure
to infections and SIP/NPI strictness as instruments. Panel A includes county × time
fixed effects, and Panel B includes state × time fixed effects. All specifications control for
firm-household pair fixed effects.
Our main finding is that the marginal propensity to cut consumption in response to
business losses was modest. Table 5 shows that for each dollar reduction in business
revenues and expenses, consumption declined by between 1.6 and 2.6 cents, respectively.
22
The third row shows that a dollar reduction in profits leads to a 3.3 cent decrease in con-
sumption.17 To put these estimates in perspective, the average monthly drop in revenues
in the lockdown period relative to its pre-pandemic average was about −$2, 400. Therefore,
the implied consumption drop due to revenue losses is 2, 400× (roughly) 0.016 ≈ $38 per
month. This decline is modest relative to an average monthly decline in consumption of
roughly $396 during the lockdown period.
4.2 The Consumption Sensitivity to Revenues Across Business Types
In this section, we investigate whether some business owners are less able to insure
against revenue losses by conducting a subsample heterogeneity analysis based on four
business characteristics: employer status, ex-ante liquidity, business size, and ownership
structure. Ex-ante liquidity is measured by the ratio of 2019 average account balances to
typical spending, or ”cash buffer days.” Firm size is proxied by 2019 average monthly
revenues. We define subgroups based on the within-industry distribution of liquidity and
size due to significant variation across industries in business size and liquidity levels.
We find little heterogeneity in the pass-through response from business revenues to
owner consumption based on employer status and ex-ante liquidity. Table 6 reports β IV
from equations (9) and (10) by subgroup. Panels A and B show that for every dollar
reduction in revenue, owners of businesses with employees reduce their consumption by
1.3 cents, which is similar to the 1.7-cent reduction for owners of non-employer businesses.
Similarly, Panels C and D show that the pass-through of businesses with low vs. high
ex-ante liquidity are 1.2 cents and 1 cent, respectively, suggesting that both low and high
liquidity businesses experience similar consumption adjustments.
Owners of smaller or pass-through businesses exhibit greater sensitivity to revenue
losses than larger or incorporated businesses. Panels E and F of Table 6 show that smaller
businesses reduce consumption by 2.9 cents per dollar revenue decline, whereas the pass-
through sensitivity is half this size (1.4 cents per dollar decline) for owners of larger small
17 These results are robust to using scaled measures of business outcomes, which accounts for potential
differences in consumption effects by business size. These analyses are shown in Appendix Table A.2 for the
first few weeks of the pandemic.
23
businesses. Table 7 shows that the living standards of owners of pass-through entities are
more sensitive to business losses relative to C-corporations. A dollar reduction in revenue
leads consumption to drop by 2.3 cents for pass-through entities (i.e., sole proprietors and
S-corporations) but only by 0.8 cents for C-corporations.
5 Explanations for the Modest Impact of Revenue Losses
on Business Owners’ Consumption
The limited pass-through of revenue losses to the living standards of business owner
households is consistent with two explanations: (1) pandemic-induced restrictions on the
ability to spend and (2) generous federal fiscal support.
First, the pandemic imposed significant restrictions on everyone’s ability to spend,
which reduced the sensitivity of owner consumption to individual business revenue de-
clines, especially during the early phase of the pandemic before the vaccine rollout. Figures
8 and 9 show the average monthly dollar amounts spent on various categories for both busi-
nesses and their owners and confirm that business and household spending was depressed
throughout the pandemic until the vaccine rollout. While spending on some categories
(e.g., fuel and groceries) increased, mirroring the household stocking-up behavior observed
in Baker et al. (2020), expenditures on travel, food services (e.g., restaurants, bars, bakeries),
entertainment, and personal services were sharply reduced during the lockdown period
(March to May 2020) and remained depressed until the vaccine rollout in April 2021.
To further validate that pandemic-induced spending restrictions contributed to the
limited pass-through, we conduct two event studies and re-estimate our marginal propen-
sity to consume (MPC) by focusing on two distinct periods: the six months surrounding
March 2020 (”early-phase”) and the six months around March 2021 (”vaccine-phase”).
MPC more than doubles in the vaccine period (2.4 cents) relative to the early phase (1.1
cents), consistent with the pandemic limiting everyone’s ability to spend in the early phase
of the pandemic. Tables A.3 and A.4 further confirm that the pass-through is larger for
”discretionary” spending than for ”committed” spending categories that were difficult to
24
adjust during the pandemic, such as groceries, utilities, insurance payments, etc. Further
decomposing consumption types into early vs. vaccine periods shows that MPC was
muted for both ”committed” and ”discretionary” consumption and that the pass-through
rises for both consumption types after the vaccine rollout (see Table A.5). Overall, these
results suggest that pandemic-induced restrictions on the ability to spend contributed to
the limited pass-through in the early phase of the pandemic.
Second, businesses and households’ liquid cash balances increased significantly at the
onset of COVID-19 and remained elevated throughout our sample period, suggesting that
federal fiscal support may have helped stabilize consumption for hard-hit business owners.
The Coronavirus Aid, Relief, and Economic Security Act (CARES Act), which included
Economic Impact Payments (EIP) and Federal Pandemic Unemployment Compensation
(FPUC) for households,18 as well as the Paycheck Protection Program (PPP) and expanded
Economic Injury Disaster Loans for small businesses, played a key role. Figure 10 illustrates
the sharp rise in financial buffers: business and personal checking account balances began
to grow significantly when the first EIP payments were distributed (April 15) and during
the rollout of the first (April 3) and second (April 27) rounds of PPP. By the end of May,
median business and owner account balances were 60 percent and 50 percent higher,
respectively, compared to the six months preceding the COVID-19 crisis.
Business and household liquid balances showed similar trends for both the most and
least affected industries, which helps explain why owner consumption remained stable
despite significant volatility in business revenues. Figure 11 illustrates median changes in
business and personal account balances of business owners, compared to levels from six
months before the crisis, for firms in both the most and least affected industries. Business
account balances surged by over 60 percent within three months of the pandemic onset
across both groups and remained elevated until the vaccine rollout. Similarly, the personal
account balances of the owners trended in parallel across sectors, indicating a consistent
buildup of financial buffers irrespective of industry impact. The generous fiscal support
during the sample period helps to explain the minimal variation in pass-through across
18 In addition to increasing UI generosity, the FPUC expanded UI eligibility criteria to include business
owners and self-employed individuals who would traditionally not be eligible to receive UI benefits.
25
households with low versus high ex-ante liquidity.
6 Conclusion
This paper documents that small businesses and their owners experienced unprece-
dented disruptions, with monthly revenues, expenses, and consumption dropping by up to
40 percent in the early phases of the pandemic. However, most of this decline was driven
by nationwide factors, while local infections and state-level policies, such as shelter-in-
place orders and NPI strictness, had only moderate additional direct effects on business
outcomes and owner consumption. A one standard deviation increase in the new infection
rate resulted in a 0.5 percentage point decline in business revenues, and the imposition of a
shelter-in-place order led to an 11 percentage point decline in business revenues.
Using differential industry exposure to NPIs and infection rates, we find that small
business losses had only a modest impact on their owners’ consumption. Despite significant
variation in small business performance across industries, owner consumption remained
strikingly similar. For every dollar reduction in business revenues, consumption declined
by 1.6 cents, which translates to an economically small effect of a -$38 monthly decline in
consumption. This represents less than 10 percent of the average monthly consumption
decline of $396 during the lockdown period.
We find corroborating evidence that the limited pass-through of business losses to
owner consumption can be attributed to widespread access to federal support programs
and reduced spending opportunities during the pandemic. Although the pass-through
increased over time, it remained small even after the vaccine rollout, when spending
opportunities expanded significantly. Additionally, the elevated liquid cash balances
throughout the first 18 months of the pandemic indicate that federal stimulus provided a
sufficient financial buffer for business owners to weather disruptions. Whether business
owners’ living standards will continue to be insulated from revenue losses once these cash
buffers are depleted remains an open question.
26
References
A LEKSEEV, G., S. A MER , M. G OPAL , K UCHLER , J. S TROEBEL , AND N. W ERNERFELT (2022):
“The Effects of COVID-19 on U.S. Small Businesses: Evidence from Owners, Managers,
and Employees,” Management Science, 69, 7–24.
AYDIN , D. (2022): “Consumption Response to Credit Expansions: Evidence from Experi-
mental Assignment of 45,307 Credit Lines,” American Economic Review, 112, 1–40.
B AKER , S., R. FARROKHNIA , S. M EYER , M. PAGEL , AND C. YANNELIS (2020): “How
Does Household Spending Respond to an Epidemic? Consumption During the 2020
COVID-19 Pandemic,” Review of Asset Pricing Studies, 10, 834–862.
B AKER , S. R., R. A. FARROKHNIA , S. M EYER , M. PAGEL , AND C. YANNELIS (2023):
“Income, Liquidity, and the Consumption Response to the 2020 Economic Stimulus
Payments,” Review of Finance, 27, 2271–2304.
B ALLA -E LLIOTT, D., Z. B. C ULLEN , E. L. G LAESER , M. L UCA , AND C. S TANTON (2022):
“Determinants of Small Business Reopening Decisions After Covid Restrictions Were
Lifted,” Journal of Policy Analysis and Management, 41, 278–317.
B ARTIK , A. W., M. B ERTRAND , Z. C ULLEN , E. G LAESER , M. L UCA , AND C. T. S TANTON
(2020a): “How Are Small Businesses Adjusting to COVID-19? Early Evidence from a
Survey,” Proceedings of the Natural Academy of Sciences, 117, 17656–17666.
B ARTIK , A. W., M. B ERTRAND , Z. B. C ULLEN , E. L. G LAESER , M. L UCA , AND C. T. S TAN -
TON (2020b): “The Impact of COVID-19 on Small Business Outcomes and Expectations,”
Proceedings of the National Academy of Sciences, 117, 17656–17666.
B ARTLETT, R. P. AND A. M ORSE (2021): “Small-Business Survival Capabilities and Fiscal
Programs: Evidence from Oakland,” Journal of Financial and Quantitative Analysis, 56,
2500–2544.
B AUGH , B., I. B EN -D AVID , H. PARK , AND J. A. PARKER (2021): “Asymmetric Consumption
Smoothing,” American Economic Review, 111, 192–230.
BDS (2016): “Business Dynamics Statistics,” U.S. Census Bureau.
B RODA , C. AND J. A. PARKER (2014): “The Economic Stimulus Payments of 2008 and the
Aggregate Demand for Consumption,” Journal of Monetary Economics, 68, S20–S36.
C HETTY, R., J. N. F RIEDMAN , M. S TEPNER , AND O. I. T EAM (2024): “The Economic
Impacts of COVID-19: Evidence from a New Public Database Built Using Private Sector
Data,” The Quarterly Journal of Economics, 139, 829–889.
C ORREIA , S., S. L UCK , AND E. V ERNER (2022): “Pandemics Depress the Economy, Public
Health Interventions Do Not: Evidence from the 1918 Flu,” The Journal of Economic
History, 82, 917–957.
27
C OX , N., P. G ANONG , P. N OEL , J. VAVRA , A. W ONG , D. FARRELL , AND F. G REIG (2020):
“Initial Impacts of the Pandemic on Consumer Behavior: Evidence from Linked Income,
Spending, and Savings Data,” Brookings Papers on Economic Activity.
D AVIS , S. J., J. H ALTIWANGER , AND S. S CHUH (1996): “Small business and job creation:
Dissecting the myth and reassessing the facts,” Small Business Economics.
D I M AGGIO , M., A. K ERMANI , B. J. K EYS , T. P ISKORSKI , R. R AMCHARAN , A. S ERU , AND
V. YAO (2017): “Interest Rate Pass-Through: Mortgage Rates, Household Consumption,
and Voluntary Deleveraging,” American Economic Review, 107, 3550–88.
D I M AGGIO , M., A. K ERMANI , AND K. M AJLESI (2020): “Stock Market Returns and
Consumption,” Journal of Finance, 75, 3175–3219.
FAIRLIE , R. W. (2020): “The Impact of COVID-19 on Small Business Owners: Evidence of
Early-Stage Losses from the April 2020 Current Population Survey,” National Bureau of
Economic Research Working Paper.
FAIRLIE , R. W., F. M. F OSSEN , R. J OHNSEN , AND G. D ROBONIKU (2023): “Were Small
Businesses More Likely to Permanently Close in the Pandemic?” Small Business Economics,
60, 1613–1629.
FARRELL , D., C. W HEAT, AND C. M AC (2020a): “Small Business Financial Outcomes
during the COVID-19 Pandemic,” JP Morgan Chase Institute.
——— (2020b): “Small Business Financial Outcomes during the Onset of COVID-19,” JP
Morgan Chase Institute.
FRBNY (2019): “Report on Nonemployer Firms: Small Business Credit Survey,” Tech. rep.,
Federal Reserve Bank of New York.
G ANONG , P. AND P. N OEL (2019): “Consumer Spending During Unemployment: Positive
and Normative Implications,” American Economic Review, 109, 2383–2424.
——— (2023): “Why do Borrowers Default on Mortgages?” The Quarterly Journal of
Economics, 138, 1001–1065.
G OLOSOV, M., M. G RABER , M. M OGSTAD , AND D. N OVGORODSKY (2024): “How Amer-
icans Respond to Idiosyncratic and Exogenous Changes in Household Wealth and
Unearned Income,” The Quarterly Journal of Economics, 139, 1321–1395.
G ROSS , D. B. AND N. S. S OULELES (2002): “Do Liquidity Constraints and Interest Rates
Matter for Consumer Behavior? Evidence from Credit Card Data,” Quarterly Journal of
Economics, 117, 149–185.
H UMPHRIES , J. E., C. N EILSON , AND G. U LYSSEA (2020): “The Evolving Impacts of
COVID-19 on Small Businesses Since the CARES Act,” Cowles Foundation Discussion
Paper.
28
K EYSTONE (2020): “Coronavirus City and County Non-pharmaceutical Intervention Roll-
out Date Dataset,” Keystone Strategy.
K OTSOGIANNIS , C. AND P. S AKELLARIS (2024): “MPCs Estimated from Tax Lottery, Survey
and Administrative Data,” Manuscript.
NES (2018): “Nonemployer Statistics,” U.S. Census Bureau.
N EW Y ORK T IMES (2020): “Coronavirus (Covid-19) Data in the United States,” The New
York Times.
PARKER , J., N. S OULELES , D. S. J OHNSON , AND R. M C C LELLAND (2013): “Consumer
Spending and the Economic Stimulus Payments of 2008,” American Economic Review, 103,
2530–2553.
PARKER , J. A., J. S CHILD , L. E RHARD , AND D. S. J OHNSON (2022): “Economic Impact
Payments and Household Spending during the Pandemic,” Brookings Papers on Economic
Activity, 53, 81–156.
SBA (2019): “Small Businesses Generate 44 Percent Of U.S. Economic Activity,” Tech. rep.,
Small Business Administration, Office of Advocacy.
S TEPHENS , M. (2008): “The Consumption Response to Predictable Changes in Discretionary
Income: Evidence from the Repayment of Vehicle Loans,” The Review of Economics and
Statistics, 90, 241–252.
SUSB (2017): “Statistics of U.S. Businesses,” U.S. Census Bureau.
U.S. C ENSUS (2020): “Small Business Pulse Survey: Tracking Changes During The Coron-
avirus Pandemic,” U.S. Census Bureau.
WANG , J., J. YANG , B. I VERSON , AND R. J IANG (2022): “Bankruptcy and the COVID-19
Crisis,” Working Paper.
YANNELIS , C. AND L. A MATO (2023): “Household Behavior (Consumption, Credit, and
Investments) During the COVID-19 Pandemic,” Annual Review of Financial Economics, 15,
91–113.
29
Table 1: Descriptive statistics of monthly outcomes in 2019 ($)
Notes: This table reports monthly business and household outcomes in 2019 in dollars. Outcomes
are winsorized at the 2nd and 98th percentile. Columns 1 and 2 report the sample mean and standard
deviation. Columns 3 to 5 report the pseudo-distribution presented as means of 10 observations in the
pth percentiles. Columns 6 and 7 report the number of firms and households in the sample. Panel A
reports statistics using the all businesses sample. Panels B and C report those using only non-employer
and employer samples. A firm is considered to be an employer firm if a business had payroll expenses
for at least 6 months in 2019. Panels D and E sample small and large firms. Firm size is determined by
2019 average monthly revenues– firms with less than the first tercile of average revenue ($1,434) are
”small,” and those with greater than the third tercile ($10,034) are ”large.” Panel F uses the business
owners’ sample, which serves as the main analysis. Business and owner characteristics are reported
below Panel F and use the all businesses sample.
Mean SD p20 p50 p80 N firms N HHs
Winsorized
(1) (2) (3) (4) (5) (6) (7)
A. All Sample
Revenue 21,914 52,199 0 3,056 22,631 2,420,484 —
Expense 22,018 46,399 600 5,121 25,253 2,420,484 —
Profit -283 20,440 -5,217 -304 2,987 2,420,484 —
B. Nonemployer
Revenue 14,798 39,611 0 2,092 14,692 2,150,414 —
Expense 15,137 34,885 440 3,894 16,983 2,150,414 —
Profit -535 17,273 -4,420 -300 2,156 2,150,414 —
C. Employer
Revenue 71,723 89,326 6,067 33,026 122,794 270,070 —
Expense 70,186 77,797 11,516 37,419 119,485 270,070 —
Profit 1,481 35,358 -16,979 -501 17,053 270,070 —
D. Small
Revenue 426 1,079 0 0 592 800,194 —
Expense 3,353 12,740 32 715 3,260 800,194 —
Profit -2,542 7,722 -2,751 -377 0 800,194 —
E. Large
Revenue 56,289 75,575 9,344 25,639 81,648 810,145 —
Expense 51,813 65,674 9,380 25,480 76,455 810,145 —
Profit 3,359 31,639 -11,289 495 15,892 810,145 —
F. Owner Subsample
Revenue 17,780 39,526 0 2,796 20,363 363,682 333,434
Expense 18,217 36,437 220 4,408 22,550 363,682 333,434
Profit -416 15,171 -4,287 -26 2,799 363,682 333,434
Consumption 4,185 5,565 601 2,359 6,009 363,682 333,434
G. Business and Owner Characteristics
Business Age 7.5 7.0 2.3 5.7 11.1 2,420,484 –
Owner Age 47.8 13.3 36.0 47.0 60.0 – 333,434
Industry N Firms Sh (%) Business Location N Firms Sh (%)
Professional Services 400,483 16.5 California 486,338 20.1
Real Estate and Leasing 325,614 13.5 New York 438,619 18.1
Other Services 283,167 11.7 Texas 316,259 13.1
Construction 220,798 9.1 Florida 243,979 10.1
Health Care and Social Asst. 172,052 7.1 Illinois 205,574 8.5
Business Ownership N Firms Sh (%) Owner Gender N HHs Sh (%)
S-Corp 589,712 24.4 M 158,112 47.4
Sole Prop 498,030 20.6 F 99,309 29.8
LLC - Member Managed 485,926 20.1 Missing 76,013 22.8
C-Corp 298,892 12.3
LLC - Manager Managed 273,236 11.3
30
Table 2: Sample Representativeness
Notes: This table compares the representativeness of the sample used in this study to various U.S. Census
external benchmarks. Column 1 reports nationwide shares. Columns 2 and 3 report the same statistics using
the 2019 all businesses and the business owners samples. See section 1.1 for details on the construction of the
all businesses and the business owners samples. Panel A compares the share of employer and nonemployer
firms. The population statistic is from 2017 Statistics of U.S. Business (SUSB, 2017). We classify establishments
with less than 5 employees in the SUSB data or those with no payroll expenses in our data to be nonemployer
firm. Panel B compares the share of firms by firm age. We exclude new firms (age =0) to make the population
statistic more aligned with our sample criteria because we require firms to have existed for at least a year to
be included in our sample. The population statistic for firm age is from 2016 Business Dynamics Statistics
(BDS, 2016). Panel C compares annual receipts in dollars for nonemployer firms using 2018 Nonemployer
Statistics (NES, 2018). To make our sample comparable to NES, we also restrict our sample to nonemployer
firms. Panel D compares industry shares using 2017 SUSB.
Table 1. Sample Benchmarking
Population Sample
Nationwide All Business Owner Sample
Share (%) Share (%) Share (%)
(1) (2) (3)
A. Employer vs. Nonemployer
Nonemployer 81.00 85.11 85.82
Employer 19.00 14.89 14.18
B. Firm age (excluding new firms)
1 7.36 15.09 16.09
2 6.34 11.75 12.61
3 5.63 9.37 9.89
4 5.16 7.95 8.24
5 4.63 7.30 7.45
6 ~ 10 20.17 29.50 31.05
11 ~ 15 50.70 19.04 14.66
C. Annual Receipts in dollars (nonemployer only)
< $5,000 24.48 11.32 14.28
$5,000-$9,999 15.54 5.12 5.80
$10,000 - $24,999 23.70 11.90 12.42
$25,000 - $49,999 14.30 14.04 13.97
$50,000 - $99,999 10.36 17.09 16.88
$100,000 - $249,999 7.81 20.13 19.19
$250,000 - $499,999 2.52 9.81 8.78
> $500,000 1.29 10.60 8.68
D. Industry
Agriculture, Forestry, Fishing/Hunting 0.37 0.62 0.52
Mining, Quarrying, and Oil/Gas Extraction 0.31 0.28 0.23
Utilities 0.10 0.11 0.10
Construction 11.57 10.76 10.79
Manufacturing 4.09 3.10 2.87
Wholesale Trade 4.92 3.69 3.37
Retail Trade 10.68 7.78 8.10
Transportation and Warehousing 3.05 5.62 6.20
Information 1.31 2.26 2.65
Finance and Insurance 3.93 2.06 1.90
Real Estate and Rental and Leasing 5.10 13.26 10.99
Professional, Scientific, and Technical Svcs 13.38 15.81 16.39
Management of Companies and Enterprises 0.44 0.51 0.34
Administrative and Waste Manag. 5.74 6.02 6.47
Educational Services 1.54 1.81 1.95
Health Care and Social Assistance 10.80 7.58 7.47
Arts, Entertainment, and Recreation 2.15 2.84 3.43
Accommodation and Food Services 8.90 4.01 4.22
Other Services (excl. Public Administration) 11.49 11.78 11.96
Industries not classified 0.13 0.13 0.11
31
Table 3: Effects of Shelter in Place (SIP) controlling for Infections (%)
Notes: This table reports estimates of local infections and shelter in place (SIP) on business outcomes and
consumption of the owners. For panels A through D, the first row of each panel reports the effect of each new
case per 1,000 residents, and the second row reports the effect of SIP. The first row of panel E reports the effect of
cumulative infections per 1,000 and that of the cumulative number of weeks that SIP has been in effect. Columns
1 through 4 report estimates using outcomes normalized 2019 monthly average, and the estimated coefficients
can be interpreted as change as a percent of the 2019 monthly average. Columns 5 through 8 report estimates
using seasonally-adjusted outcomes, and the coefficients can be interpreted as change as a percent of the centered
3-month average from 2019. All regressions include firm and household pair fixed effects. Columns 2 and 6
include time effects, columns 3 and 7 include time × NAICS 2-digit industry effects, and columns 4 and 8 include
time × size bin effects to flexibly control for time-varying factors related to industry and firm size. Size bins are
as defined in Table 1. Coefficients are multiplied by 100 and represented in a percent unit. Standard errors are
reported in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.
Increase as percent of 2019 monthly average Increase as percent of 3-month centered average
(1) (2) (3) (4) (5) (6) (7) (8)
A. Revenues
New cases -0.401 *** -0.055 *** -0.063 *** -0.068 *** -0.253 *** -0.087 *** -0.069 *** -0.092 ***
(0.007) (0.01) (0.01) (0.01) (0.014) (0.019) (0.019) (0.019)
Shelter in place 6.372 *** -4.851 *** -4.586 *** -4.854 *** -27.03 *** -11.43 *** -10.52 *** -11.40 ***
(0.176) (0.292) (0.293) (0.289) (0.338) (0.563) (0.565) (0.563)
B. Expenses
New cases -0.351 *** -0.101 *** -0.103 *** -0.109 *** 0.009 -0.131 *** -0.115 *** -0.131 ***
(0.005) (0.007) (0.007) (0.007) (0.009) (0.012) (0.012) (0.012)
Shelter in place 11.163 *** -3.885 *** -3.626 *** -3.908 *** -20.09 *** -6.298 *** -5.899 *** -6.285 ***
(0.134) (0.222) (0.222) (0.219) (0.211) (0.351) (0.352) (0.351)
C. Profit
New cases -0.021 *** 0.055 *** 0.053 *** 0.054 *** -0.088 *** 0.064 *** 0.062 *** 0.065 ***
(0.006) (0.009) (0.009) (0.009) (0.008) (0.011) (0.012) (0.011)
Shelter in place -9.049 *** -1.337 *** -1.271 *** -1.320 *** -5.618 *** -1.933 *** -1.789 *** -1.956 ***
(0.16) (0.266) (0.267) (0.266) (0.208) (0.348) (0.349) (0.348)
D. Consumption
New cases -0.054 *** -0.130 *** -0.126 *** -0.130 *** 0.073 *** -0.137 *** -0.130 *** -0.137 ***
(0.005) (0.007) (0.007) (0.007) (0.011) (0.014) (0.014) (0.014)
Shelter in place -1.895 *** -1.678 *** -1.591 *** -1.669 *** -25.63 *** -5.635 *** -5.485 *** -5.597 ***
(0.133) (0.221) (0.221) (0.22) (0.257) (0.43) (0.432) (0.43)
E. Exit
Cumulative cases 0.012 *** -0.001 *** -0.001 *** 0.000 0.012 *** -0.001 *** -0.001 *** 0.000
(0.000) (0.0001) (0.0001) (0.0001) (0.000) (0.0001) (0.0001) (0.0001)
Shelter in place 0.048 *** 0.003 0.0001 0.001 0.048 *** 0.003 0.0001 0.001
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Number of Obs 5,854,504 5,854,504 5,854,504 5,854,504 5,461,485 5,461,485 5,461,485 5,461,485
Firm-Household FE X X X X X X X X
Time FE X X
Time x Industry FE X X
Time x Size Bin FE X X
32
Table 4: Effects of Non-Pharmaceutical Intervention (NPI) Strictness and Infections (%)
Notes: This table reports estimates of infections and NPI strictness on business outcomes and consumption
of the owners. The first rows of panels A through D report the effect of each new case per 1,000 residents,
and the first row of panel E reports the effect of cumulative infections per 1,000 residents. The second row
of each panel reports the effect of NPI strictness per standard deviation increase in NPI strictness. NPI strict-
ness is the first principal component in a principal component analysis of state-level NPIs and captures the
intensity of state-specific packaged NPI policies relative to other states. The first component explains 76% of
the variance and weighs positively on all restrictions. Columns 1 through 4 report estimates using outcomes
normalized 2019 monthly average, and the estimated coefficients can be interpreted as change as a percent of
the 2019 monthly average. Columns 5 through 8 report estimates using seasonally-adjusted outcomes, and
the coefficients can be interpreted as change as a percent of the centered 3-month average from 2019. All re-
gressions include firm and household pair fixed effects. Columns 2 and 6 include time effects, columns 3 and
7 include time × NAICS 2-digit industry effects, and columns 4 and 8 include time × firm size bin effects to
flexibly control for time-varying factors related to industry and firm size. Size bins are as defined in Table 1.
Coefficients are multiplied by 100 and represented in a percent unit. Standard errors are reported in parentheses.
*** p < 0.01, ** p < 0.05, * p < 0.10.
Increase as percent of 2019 monthly average Increase as percent of 3-month centered average
(1) (2) (3) (4) (5) (6) (7) (8)
A. Revenues
New cases -0.397 *** -0.065 *** -0.073 *** -0.079 *** -0.267 *** -0.109 *** -0.090 *** -0.114 ***
(0.007) (0.01) (0.01) (0.01) (0.014) (0.019) (0.019) (0.019)
NPI Strictness 3.013 *** -0.575 *** -0.478 *** -0.547 *** -10.38 *** -1.906 *** -1.662 *** -1.889 ***
(0.067) (0.067) (0.067) (0.066) (0.13) (0.13) (0.129) (0.129)
B. Expenses
New cases -0.343 *** -0.109 *** -0.111 *** -0.117 *** -0.001 -0.141 *** -0.125 *** -0.142 ***
(0.005) (0.007) (0.007) (0.007) (0.009) (0.012) (0.012) (0.012)
NPI Strictness 5.266 *** -0.544 *** -0.464 *** -0.523 *** -7.722 *** -1.493 *** -1.367 *** -1.486 ***
(0.051) (0.051) (0.051) (0.05) (0.081) (0.081) (0.081) (0.081)
C. Profit
New cases -0.027 *** 0.052 *** 0.049 *** 0.051 *** -0.092 *** 0.060 *** 0.058 *** 0.061 ***
(0.006) (0.009) (0.009) (0.009) (0.008) (0.011) (0.012) (0.011)
NPI Strictness -4.059 *** -0.109 . -0.081 -0.113 . -2.605 *** -0.281 *** -0.229 ** -0.284 ***
(0.061) (0.061) (0.061) (0.061) (0.08) (0.08) (0.08) (0.08)
D. Consumption
New cases -0.053 *** -0.133 *** -0.128 *** -0.132 *** 0.064 *** -0.150 *** -0.142 *** -0.149 ***
(0.005) (0.007) (0.007) (0.007) (0.011) (0.014) (0.014) (0.014)
NPI Strictness -0.346 *** -0.407 *** -0.393 *** -0.403 *** -9.216 *** -0.450 *** -0.424 *** -0.452 ***
(0.051) (0.051) (0.051) (0.051) (0.098) (0.099) (0.099) (0.099)
E. Exit
Cumulative cases 0.010 *** -0.001 *** -0.001 *** 0.000 0.010 *** -0.001 *** -0.001 *** 0.000
(0.000) (0.0001) (0.0001) (0.0001) (0.000) (0.0001) (0.0001) (0.0001)
NPI Strictness -0.280 *** 0.021 *** 0.021 *** 0.021 *** -0.280 *** 0.021 *** 0.021 *** 0.021 ***
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Number of Obs 5,854,504 5,854,504 5,854,504 5,854,504 5,461,485 5,461,485 5,461,485 5,461,485
Firm-Household FE X X X X X X X X
Time FE X X
Time x Industry FE X X
Time x Size Bin FE X X
33
Table 5: Marginal Propensity to Consume out of Business Outcomes
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar change in
business revenue, expense, and profit margin using equation (10). Columns 1 and 2 use variation by industry due
to SIP or SIP and infections and Columns 3 and 4 use that due to NPI strictness or NPI strictness and infections
as the excluded instruments. NPI strictness is as defined in Table 4. Outcomes are in dollars (level), and all
regressions include firm and household pair fixed effects and time × county fixed effects. Therefore, the estimated
coefficients can be interpreted as consumption declines (in dollar unit) per each dollar reduction in business
outcomes. Firms that operate in sub-industries with less than 30 firms are dropped from the estimation. ***
p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to
SIP NPI strictness
Endogenous SIP and infections NPI strictness and infections
Variable (1) (2) (3) (4)
Revenue 0.013 *** 0.016 *** 0.014 *** 0.016 ***
(.002) (.002) (.002) (.002)
Expense 0.018 *** 0.025 *** 0.022 *** 0.026 ***
(.003) (.002) (.003) (.003)
Net Income 0.033 *** 0.024 *** 0.022 *** 0.017 ***
(.007) (.005) (.006) (.005)
Number of Obs 5,119,324 5,119,324 5,119,324 5,119,324
Firm-Household FE x x x x
Time x County FE x x x x
34
Table 6: Marginal Propensity to Consume out of Business by Subgroup
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar
change in business revenue, expense, and profit margin using equation (10) by business type. Odd numbered
columns in each panel use variation by industry due to SIP or NPI strictness and even numbered columns in
each panel use that due to SIP and infections or NPI strictness and infections as the excluded instruments.
NPI strictness is as defined in Table 4. Outcomes are in dollars (level), and all regressions include firm
and household pair fixed effects and time × county fixed effects. Therefore, the estimated coefficients can
be interpreted as consumption declines (in dollar unit) per each dollar reduction in business outcomes.
Panels A and B reports estimates using subsamples of nonemployer and employer firms. Panels C and
D reports estimates using subsamples of low and high liquidity firms. Panels E and F report estimates
using subsamples of small and large firms. Liquidity is computed as the ratio of 2019 average monthly cash
balances to expenses multiplied by 30 and can be interpreted as a firm’s average cash buffer days, or the
number of days of operating expenses that a business could pay out of its cash balances were its revenues
to stop. ”Low (high) liquidity” sample includes firms with lower (higher) than the first (third) quartile of
cash buffer days within its sub-industry (NAICS 4-digit). ”Small” (”Large”) firms includes those with lower
(higher) than median annual sales in 2019 within its sub-industry. Firms that operate in sub-industries with
less than 30 firms are dropped from the estimation. *** p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to Variation by industry due to
SIP NPI strictness SIP NPI strictness
SIP and infections NPI strictness and infections SIP and infections NPI strictness and infections
Endogenous variable (1) (2) (3) (4) (5) (6) (7) (8)
A. Non-employer B. Employer
Revenue 0.009 ** 0.017 *** 0.008 ** 0.015 *** 0.013 *** 0.013 *** 0.013 *** 0.013 ***
(.003) (.003) (.003) (.003) (.003) (.002) (.002) (.002)
Expense 0.011 * 0.025 *** 0.014 *** 0.026 *** 0.021 *** 0.020 *** 0.021 *** 0.021 ***
(.004) (.004) (.004) (.003) (.003) (.003) (.003) (.003)
Net Income 0.026 ** 0.025 *** 0.008 0.013 . 0.008 0.008 0.001 0.003
(.009) (.007) (.009) (.007) (.007) (.006) (.007) (.005)
Number of Obs 4,296,538 4,296,538 4,296,538 762,095 762,095 762,095 762,095
C. Low-liquidity D. High Liquidity
Revenue 0.006 . 0.009 *** 0.010 *** 0.012 *** 0.010 * 0.010 * 0.007 0.007 .
(.003) (.003) (.003) (.003) (.005) (.004) (.005) (.004)
Expense 0.012 ** 0.015 *** 0.015 *** 0.017 *** 0.018 * 0.023 ** 0.017 * 0.021 **
(.005) (.003) (.004) (.003) (.008) (.007) (.008) (.007)
Net Income 0.003 0.003 0.000 0.001 0.018 . 0.010 0.010 0.006
(.011) (.008) (.011) (.008) (.010) (.008) (.009) (.008)
Number of Obs 1,300,907 1,300,907 1,300,907 1,300,907 1,222,284 1,222,284 1,222,284 1,222,284
E. Small F. Big
Revenue 0.014 0.029 *** 0.016 . 0.028 *** 0.013 *** 0.014 *** 0.014 *** 0.014 ***
(.009) (.008) (.009) (.008) (.002) (.002) (.002) (.001)
Expense 0.008 0.032 *** 0.021 * 0.041 *** 0.020 *** 0.022 *** 0.021 *** 0.023 ***
(.010) (.009) (.009) (.008) (.002) (.002) (.002) (.002)
Net Income 0.060 * 0.025 0.020 0.002 0.026 *** 0.017 *** 0.020 *** 0.014 **
(.025) (.019) (.023) (.019) (.006) (.005) (.005) (.004)
Number of Obs 2,483,924 2,483,924 2,483,924 2,483,924 2,633,915 2,633,915 2,718,773 2,718,773
Firm-Household FE x x x x x x x x
Time x County FE x x x x x x x x
35
Table 7: Marginal Propensity to Consume out of Business by Incorporation Status
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar
change in business revenue, expense, and profit margin using equation (10) by business incorporation status.
Panel A reports estimates using a sample of pass-through (sole proprietors and S-corporations) entities and
panel B reports estimates using a sample of C-corporations. Columns 1 and 3 use variation by industry
due to SIP and Columns 3 and 4 use that due to SIP and infections. Outcomes are in dollars (level), and all
regressions include firm and household pair fixed effects and time × county fixed effects. Therefore, the
estimated coefficients can be interpreted as consumption declines (in dollar unit) per each dollar reduction
in business outcomes. Firms that operate in sub-industries with less than 30 firms are dropped from the
estimation. Standard errors are reported in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to
SIP SIP
Endogenous SIP and infections SIP and infections
Variable (1) (2) (3) (4)
A. C-corp B. Pass-through
Revenue 0.006 0.008 * 0.018 *** 0.023 ***
(.004) (.004) (.003) (.003)
Expense 0.004 0.011 * 0.026 *** 0.035 ***
(.006) (.005) (.004) (.004)
Profit Margin 0.033 ** 0.012 0.042 *** 0.035 ***
(.011) (.008) (.009) (.007)
Number of Obs 607,890 607,890 2,322,113 2,322,113
Firm-Household FE x x x x
Time x County FE x x x x
36
Figure 1: Average business and household outcomes
Notes: This figure shows the average monthly dollar levels of business revenues, expenses, profits, and
household consumption from September 2019 to September 2021. The dotted vertical lines mark two
key points: the month of the national emergency declaration (March 9th, 2020) and one year after the
national emergency. The blue horizontal lines represent the average levels of these outcomes during the
six months prior to the national emergency.
A. Revenues B. Expenses
26,000 Mar 20 Mar 21 26,000 Mar 20 Mar 21
Level ($) Level ($)
22,000 22,000
18,000 18,000
14,000 14,000
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
C. Profits D. Consumption
3,000
Mar 20 Mar 21
Mar 20 Mar 21
6,000
2,000
Level ($) Level ($)
1,000 5,000
0
4,000
−1,000
−2,000 3,000
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
37
Figure 2: Average percent change in business and owner outcomes relative to 2019
Notes: This figure shows the average monthly percent change in business revenues, expenses, profits,
and household consumption from September 2019 to September 2021. The dotted vertical lines indicate
two key points: the month of the national emergency declaration (March 9th, 2020) and one year after the
declaration. Outcomes are normalized based on the centered 3-month average from the previous year,
with changes defined as percent deviations from their averages between September 2019 and March
2020 (i.e., the six months prior to the national emergency).
A. Revenues B. Expenses
120 120
Mar 20 Mar 21 Mar 20 Mar 21
Percent change Percent change
80 80
40 40
0 0
−40 −40
−80 −80
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
C. Profit Margins D. Consumption
40 120
Mar 20 Mar 21
Mar 20 Mar 21
Percent change Percent change
80
20
40
0
0
−40
−20 −80
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
38
Figure 3: Average changes in business revenues in 2020 by subgroup
Notes: This figure shows the average monthly percent change in business revenues, expenses, profits,
and household consumption from September 2019 to September 2021. The dotted vertical lines indicate
two key points: the month of the national emergency declaration (March 9th, 2020) and one year after the
declaration. Outcomes are normalized based on the centered 3-month average from the previous year,
with changes defined as percent deviations from their averages between September 2019 and March
2020 (i.e., the six months prior to the national emergency). Panel A plots monthly changes in revenues
for essential and non-essential businesses; Panel B for small and large businesses; Panel C by employer
and non-employer firms; and panel D by low vs. high liquidity firms. Essential industry categorization
based on the advisory list provided by the Department of Homeland Security (HLS). ”Small” (”Large”)
firms includes those with lower (higher) than median annual sales in 2019 within its NAICS 4-digit
sub-industry. A firm is considered to be an employer firm if a business had payroll expenses for at least
6 months in 2019. Low (high) liquidity sample includes firms with lower (higher) than the first (third)
quartile of cash buffer days within its sub-industry.
A. Essential vs. Non−essential B. Small vs. Big
120
Mar 20 Mar 21 Mar 20 Mar 21
220
Percent change Percent change
180 80
140
100 40
60
20 0
−20 −40
−60
−100 −80
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
Non−essential Essential Big Small
C. Employer vs. Non−employer D. Low vs. High Liquidity
120 120
90 Mar 20 Mar 21 Mar 20 Mar 21
Percent change Percent change
80
60
30 40
0 0
−30
−40
−60
−90 −80
−6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency
Non−employer Employer High liquidity Low liquidity
39
Figure 4: Business survival
Notes: This figure shows the number of business survival by month. This sample includes 1.8mil
businesses that were active in 2019 and have an open account for at least one month in 2020. Exit is
defined as the closure of a business checking account or inactivity until the end of the sample period. If
a business has two business checking accounts, both accounts must be closed to be coded as exit.
100
99.97%
98.75%
Percent of firms 95
Mar 20 Mar 21 92.58%
90
−6 0 6 12 18
Months Since National Emergency
Firms with account closures or exits
40
Figure 5: Share of states with NPI policies in effect
Notes: This figure shows the share of states that have respective NPIs enacted over time. The dotted
vertical lines indicate the month of the national emergency declaration (March 2020). For example,
panel A shows that more than 80% of the states in our sample imposed shelter-in-place restrictions five
weeks into the national emergency. ”Nonessential,” ”Public venue,” ”Religious gathering,” and ”School”
refer to closures or restrictions on the said activities. The numbers in parenthesis for ”Gathering limit”
restrictions refer to gathering limits (e.g., a limit of 10 people). Source: State-level NPI data are obtained
from Keystone Strategy and U.S. Department of Health and Human Services.
A. Shelter in place E. Religious gathering I. Closed biz overnight
1.0 1.0 1.0
0.8 0.8 0.8
0.6 0.6 0.6
Share Share Share
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency Months Since National Emergency
B. State of Emergency F. Public venue J. School
1.0 1.0 1.0
0.8 0.8 0.8
0.6 0.6 0.6
Share Share Share
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency Months Since National Emergency
C. Gathering limit (0−10) G. Nonessential K. Mask mandate
1.0 1.0 1.0
0.8 0.8 0.8
0.6 0.6 0.6
Share Share Share
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency Months Since National Emergency
D. Gathering limit (11−100) H. Theatres L. Restaurants
1.0 1.0 1.0
0.8 0.8 0.8
0.6 0.6 0.6
Share Share Share
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Months Since National Emergency Months Since National Emergency Months Since National Emergency
41
Figure 6: County-level infections per capita
Notes: This figure displays county-level infections per 1,000 residents for three counties representing low,
medium, and high-risk areas for disease growth, determined by terciles of cumulative infection rates.
Counties with cumulative infection rates below the first tercile (1.01) are classified as ”low-risk”; those
between the first and third terciles (up to 2.94) are ”medium-risk”; and counties above the third tercile
are classified as ”high-risk.” The counties illustrated have the highest cumulative infections per 1,000
residents within each risk category. Panel A plots new cases, using the left axis, for example, counties
with low (Glenn, CA) and medium (Nowata, OK) caseloads, while the right axis represents the high-risk
county (New York, NY). Panel B plots cumulative cases per 1,000 residents. County-level population
estimates are based on total population data as of July 1, 2019. The dotted vertical lines indicate two
key points: the month of the national emergency declaration (March 9th, 2020) and one year after the
declaration. Source: Population data from the U.S. Census Bureau and coronavirus data from The New
York Times, based on reports from state and local health agencies.
A. Growth in cases per 1,000 residents
80
150
New York
Mar 20 Mar 21
Cases per 1,000 Cases per 1,000
60 County, NY
Nowata County,
(right) 100
Glenn County, OK (left)
40
CA (left)
50
20
0 0
−6 0 6 12 18
Months Since National Emergency
B. Cumulative Number of Cases per 1,000 People
500 New York 1000
Mar 20 Mar 21
County, NY
Cases per 1,000 Cases per 1,000
400
(right) 750
300 Nowata County,
OK (left) 500
200 Glenn
County, CA 250
100
(left)
0 0
−6 0 6 12 18
Months Since National Emergency
42
Figure 7: Average changes in business revenues and owners’ consumption by industry performance
Notes: This figure shows average monthly changes in business revenues (blue) and owners’ consumption
(red) for businesses in the most and the least affected industries. Outcomes are normalized based on the
centered 3-month average from the previous year, with changes defined as percent deviations from their
averages between September 2019 and March 2020 (i.e., the six months prior to the national emergency).
The dotted vertical lines indicate two key points: the month of the national emergency declaration
(March 9th, 2020) and one year after the declaration. Solid lines show the average change in outcomes
for the top 5% best performing (i.e., least affected) NAICS 4-digit industries in terms of their average
drop in revenues since the onset of the national emergency, and dashed lines show the average change
in outcomes for the bottom 5% worst performing industries. For this analysis, we restrict the sample to
industries with at least 100 businesses. The least affected industries include beer, wine, and liquor stores;
nursing care facilities; funeral homes and cemeteries; toilet preparation and detergent manufacturing;
and chemical product manufacturing firms. The most affected industries include vending machine
operators, taxi or limo services, consumer goods rentals, travel/tour agencies, and drinking places.
150
Mar 20 Mar 21
Percent difference from pre−period
100
50
0
−50
−100
−150
−6 −4 −2 0 2 4 6 8 10 12 14 16 18
Months Since National Emergency
Revenue: Least Revenue: Most Consumption: Least Consumption: Most
affected industries affected industries affected industries affected industries
43
Figure 8: Average business expenses across spending categories
Notes: This figure plots the average monthly dollar levels of detailed business expense categories from
September 2019 to September 2021. See section 1.2 for details on the categorization of business expenses.
”Goods” expenses are plotted in black. ”Services” are plotted in blue. ”Other major expenses” are plotted
in red. Uncategorizable cash, check, or wire transfer expenses are plotted in green. The dotted vertical
lines indicate two key points: the month of the national emergency declaration (March 9th, 2020) and
one year after the declaration.
Fuel Durable retail Misc. biz svcs Utilities
160 100 210
90 30
140 200
80 25 190
120 70 180
100 60 20 170
50 160
80
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Groceries Wholesale Prof. biz svcs Debt pymt
130 75 1200
120 70 1100
65 1100
110 1000 1000
100 60
55 900
90 50 800 900
80 45 700
70 600 800
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Equipment Entertainment Travel Cash
50 30 60 300
45 50
20 250
40 40
35 10 30
30 200
0 20
25
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Materials Food Payroll Check
2000
180 120 8000
160 100 1800
7000
140 80 1600
120 6000
60
100 1400 5000
80 40
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Retail Insurance Tax Uncat
130 200 240
120 190 7000
220
110 180 6000
100 170 200
90 160 180 5000
80 150 160
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
44
Figure 9: Average household expenses across spending categories
Notes: This figure plots the average monthly dollar levels of detailed categories of business owners’
household expenses from September 2019 to September 2021. See section 1.2 for details on the catego-
rization of household expenses. ”Goods” expenses are plotted in black. ”Services” are plotted in blue.
”Other major expenses” are plotted in red. Uncategorizable cash, check, or wire transfer expenses are
plotted in green. The dotted vertical lines indicate two key points: the month of the national emergency
declaration (March 9th, 2020) and one year after the declaration.
Fuel Auto repair Travel Debt pymt
100 100 1700
90 30
1600
80 80
25 1500
70 60 1400
60 20 40 1300
50 20 1200
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Groceries Entertainment Insurance Cash
50 550
200 210 500
45
180 40 200 450
160 35 190 400
30 180 350
140 25 300
20 170 250
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Pharmacy Food away Medical Check
20 200 120 1000
18 150 100 900
16 80 800
14 100 700
60
12
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Retail Prof. svcs Utilities Uncat
400 210 900
350 18 200 800
300 16 190 700
250 180
14 170 600
200
160 500
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
Durable retail Persnl. svcs Gov't
140 80
70 100
120 60 80
100 50
40 60
80 30 40
−6 0 6 12 18 −6 0 6 12 18 −6 0 6 12 18
45
Figure 10: Median business and owner’s checking account balances from January
Notes: This figure shows median end-of-month business (blue) and owner’s personal (red) checking
account balances. Panel A shows the median dollar levels of account balances, and Panel B shows the
median percent change in account balances since September 2019. For panel A, the owner’s personal
checking account balances use the left axis, and business checking account balances use the right axis.
Green lines indicate the months when Economic Impact Payments and Paycheck Protection Program
funds were disbursed.
A. Median Business and Household Checking Account Balances ($)
9,000
Mar 20 PPP 2
10,000
EIP 2
Business Checking
8,000 EIP PPP
Account (right)
7,000
8,000
6,000 Mar 21
Household Checking
5,000 6,000
Account (left)
4,000
4,000
3,000
2,000
2,000
1,000
−6 −3 0 3 6 9 12 15 18
Months Since National Emergency
B. Median Change in Business and Household Checking Account Balances
100
EIP PPP EIP 2 PPP 2
80 Mar 20
Percent difference
60
40
from pre−period 20 Mar 21 Household Checking
Account
0
Business Checking
Account
−20
−6 −3 0 3 6 9 12 15 18
Months Since National Emergency
46
Figure 11: Median changes in checking account balance by industry performance
Notes: This figure shows median changes in the end-of-month business (blue) and owner’s personal
(red) checking account balances from September 2019 by industry performance. Solid lines show the
top 10% of best-performing NAICS 4-digit industries in terms of their average drop in revenues since
the onset of the national emergency, and dashed lines show the average change in outcomes for the
bottom 10% of worst-performing industries. For this analysis, we restrict the sample to industries with
at least 100 businesses. Green lines indicate the months when Economic Impact Payments and Paycheck
Protection Program funds were disbursed.
120
EIP PPP EIP 2 PPP 2
100
80
Mar 20
Percent difference 60
from pre−period
40
20
Mar 21
0
−20
−6 −3 0 3 6 9 12 15 18
Months Since National Emergency
Biz Chk: Least affected Biz Chk: Most affected HH Chk: Least affected HH Chk: Most affected
industries industries industries industries
120
EIP PPP EIP 2 PPP 2
100
80
Mar 20
Percent difference 60
from pre−period
40
20
Mar 21
0
−20
−6 −3 0 3 6 9 12 15 18
47 National Emergency
Months Since
Biz Chk: Least affected Biz Chk: Most affected HH Chk: Least affected HH Chk: Most affected
industries industries industries industries
Table A.1: Principal Component Analysis of State-level NPIs
Notes: This table reports results from the principal component analysis using state-level NPI data obtained from Keystone
Strategy. Panel A reports the standard deviation, proportion of variance, and cumulative proportion of each component.
Panel B reports factor loadings. Gathering restriction indicators are transformed such that less restrictive measures are
automatically considered to be in effect if more restrictive measures are in place. For example, if a state has a gathering limit
of 10 people, then less restrictive gathering restrictions (limit of 25, 100, or 500 people) are assumed to be in effect. We use the
first component (PC 1) as the primary measure of NPI strictness.
PC 1 PC 2 PC 3 PC 4 PC 5 PC 6 PC 7 PC 8 PC 9 PC 10 PC 11 PC 12 PC 13
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (10) (10) (10)
A. Importance of Components
Standard Deviation 2.762 1.102 0.905 0.868 0.831 0.639 0.586 0.537 0.526 0.431 0.410 0.347 0.316
Proportion of Variance 0.587 0.093 0.063 0.058 0.053 0.031 0.026 0.022 0.021 0.014 0.013 0.009 0.008
Cumulative Proportion 0.587 0.680 0.743 0.801 0.855 0.886 0.912 0.935 0.956 0.970 0.983 0.992 1.000
B. Factor Loadings (Rotated and Scaled)
Shelter State of Closed Closed Closed Biz. closed Public Closed Closed Religious Non essential Gathering Gathering
in place emergency gyms theater bars overnight venues schools restaurants gathering biz. 100 10
48 Shelter in place 0.301 0.141 -0.096 -0.020 0.093 0.333 0.600 -0.183 0.522 -0.270 -0.085 0.120 -0.005
State of emergency 0.137 -0.550 0.360 0.376 0.627 0.096 0.008 -0.041 -0.024 0.045 -0.031 -0.019 0.005
Closed gyms 0.319 0.019 0.043 -0.357 0.142 0.032 0.093 0.291 -0.285 -0.154 0.044 -0.116 0.730
Closed theater 0.309 -0.086 0.088 -0.416 0.057 0.079 0.143 0.297 -0.376 -0.128 -0.112 0.045 -0.652
Closed bars 0.280 -0.240 0.088 -0.394 -0.048 -0.180 -0.552 -0.261 0.425 -0.238 -0.229 -0.059 -0.002
Biz. closed overnight 0.111 -0.540 -0.808 0.092 -0.093 -0.002 0.042 0.108 -0.035 0.057 -0.089 -0.023 0.028
Public venues 0.296 0.234 -0.163 0.279 0.098 -0.265 -0.098 -0.417 -0.419 -0.327 -0.081 0.448 0.026
Closed schools 0.292 0.153 0.013 0.206 0.079 -0.642 0.071 0.517 0.342 0.142 0.029 0.133 -0.051
Closed restaurants 0.315 0.061 -0.098 -0.285 0.163 0.054 -0.038 -0.308 0.020 0.682 0.419 0.196 -0.022
Religious gathering 0.272 0.270 -0.101 0.295 0.007 0.583 -0.517 0.358 0.087 -0.006 0.025 0.105 -0.033
Non essential biz. 0.318 0.253 -0.114 0.202 0.069 -0.096 0.010 -0.207 -0.095 0.023 0.020 -0.834 -0.135
Gathering 100 0.264 -0.308 0.242 0.192 -0.537 0.005 0.056 -0.019 -0.016 -0.239 0.625 -0.006 -0.018
Gathering 10 0.291 -0.088 0.272 0.167 -0.477 0.075 0.129 -0.071 -0.112 0.419 -0.586 0.048 0.129
Table A.2: Transmission of Business Shocks to Owner’s Consumption (Elasticity)
Notes: This table reports elasticities of owner households’ consumption with respect to changes in business
revenues, expenses, and profit margins using equations (10). The first four columns use outcomes normalized by
2019 weekly average and the last four columns use seasonally-adjusted outcomes that are normalized by 2019
9-week centered average. Columns 1, 2, 5 and 6 use variation by industry due to SIP or SIP and infections as the
excluded instruments. Columns 3, 4, 7, and 8 use variation by industry due to NPI strictness or NPI strictness and
infections as the excluded instruments. NPI strictness is as defined in Table 4. All regressions include firm and
household pair fixed effects. Panel A reports estimates including county × time fixed effects and panel B reports
estimates controlling for state × time effects. Firms that operate in sub-industries with less than 30 firms are
dropped from the estimation. Standard errors
Table 9. are of
reported
Transmission in Shocks
Business to Owner's*** p <
parentheses. 0.01, ** p < 0.05, * p < 0.10.
Consumption
(Elasticity of Consumption to Business Outcomes)
Increase as percent of 2019 weekly average Increase as percent of 2019 9-week centered average
Variation by industry due to Variation by industry due to
SIP NPI NPI Strictness SIP NPI NPI Strictness
Endogenous SIP & Infections Strictness & Infections SIP & Infections Strictness & Infections
Variable (1) (2) (3) (4) (5) (6) (7) (8)
A. County x time fixed effects
Revenue .149 *** .119 *** .156 *** .127 *** .163 *** .099 *** .173 *** .108 ***
(.007) (.007) (.006) (.006) (.009) (.009) (.009) (.008)
Expense .195 *** .148 *** .197 *** .153 *** .205 *** .127 *** .208 *** .130 ***
49 (.008) (.008) (.008) (.008) (.011) (.011) (.01) (.01)
Profit .237 *** .190 *** .256 *** .205 *** .170 *** .124 *** .188 *** .140 ***
(.012) (.011) (.012) (.01) (.016) (.014) (.014) (.013)
B. State x time fixed effects
Revenue .120 *** .111 *** .132 *** .121 *** .080 *** .062 *** .089 *** .070 ***
(.007) (.007) (.006) (.006) (.01) (.009) (.009) (.009)
Expense .154 *** .141 *** .164 *** .149 *** .093 *** .086 *** .096 *** .086 ***
(.009) (.008) (.008) (.008) (.012) (.012) (.011) (.011)
Profit .210 *** .175 *** .230 *** .190 *** .170 *** .116 *** .189 *** .134 ***
(.012) (.011) (.012) (.01) (.016) (.014) (.014) (.013)
Number of Obs 7,248,362 6,918,891 7,248,362 6,918,891 6,482,239 6,190,218 6,482,239 6,190,218
Firm-Household FE X X X X X X X X
Table A.3: MPC Event Study: Early-Phase vs. Vaccine-Phase
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar
change in business revenue, expense, and profit margin using equation (10). Columns 1 and 2 use variation
by industry due to SIP or SIP and infections and Columns 3 and 4 use that due to NPI strictness or NPI
strictness and infections as the excluded instruments. NPI strictness is as defined in Table 4. Outcomes are in
dollars (level), and all regressions include firm and household pair fixed effects and time × county fixed
effects. Therefore, the estimated coefficients can be interpreted as consumption declines (in dollar unit) per
each dollar reduction in business outcomes. Firms that operate in sub-industries with less than 30 firms are
dropped from the estimation. *** p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to
SIP NPI strictness
Endogenous SIP and infections NPI strictness and infections
Variable (1) (2) (3) (4)
A. March 2020 Event Study
Revenue 0.001 0.006 ** 0.008 *** 0.011 ***
(.002) (.002) (.002) (.002)
Expense 0.008 * 0.012 *** 0.021 *** 0.021 ***
(.003) (.003) (.003) (.003)
Net Income -0.019 ** 0.001 -0.017 ** 0.001
(.007) (.005) (.006) (.005)
Number of Obs 2,353,988 2,353,988 2,353,988 2,353,988
B. March 2021 Event Study
Revenue 0.025 *** 0.022 *** 0.025 *** 0.024 ***
(.005) (.003) (.003) (.002)
Expense 0.036 *** 0.035 *** 0.033 *** 0.034 ***
(.007) (.005) (.003) (.003)
Net Income 0.035 *** 0.019 ** 0.048 *** 0.024 ***
(.010) (.007) (.009) (.007)
Number of Obs 3,011,272 3,011,272 3,011,272 3,011,272
Firm-Household FE x x x x
Time x County FE x x x x
50
Table A.4: Committed vs. Discretionary Consumption
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar change in
business revenue, expense, and profit margin using equation (10). Columns 1 and 2 use variation by industry due
to SIP or SIP and infections and Columns 3 and 4 use that due to NPI strictness or NPI strictness and infections
as the excluded instruments. NPI strictness is as defined in Table 4. Outcomes are in dollars (level), and all
regressions include firm and household pair fixed effects and time × county fixed effects. Therefore, the estimated
coefficients can be interpreted as consumption declines (in dollar unit) per each dollar reduction in business
outcomes. Committed consumption includes the following categories: groceries, insurance, rent, government
payments, utilities, uncategorized check payments, uncategorized payments. Discretionary consumption includes
the following categories: fuel, retail, durable retail, auto repair, medical, entertainment, food away, personal
services, professional services, other services, travel, cash. *** p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to
SIP NPI strictness
Endogenous SIP and infections NPI strictness and infections
Variable (1) (2) (3) (4)
A. Committed Consumption
Revenue 0.006 *** 0.007 *** 0.005 *** 0.007 ***
(.002) (.001) (.001) (.001)
Expense 0.009 *** 0.012 *** 0.009 *** 0.011 ***
(.002) (.002) (.002) (.002)
Net Income 0.012 * 0.009 * 0.003 0.003
(.005) (.004) (.004) (.004)
Number of Obs 5,119,324 5,119,324 5,119,324 5,119,324
B. Discretionary Consumption
Revenue 0.007 *** 0.009 *** 0.009 *** 0.010 ***
(.001) (.001) (.001) (.001)
Expense 0.010 *** 0.013 *** 0.013 *** 0.015 ***
(.002) (.001) (.001) (.001)
Net Income 0.022 *** 0.015 *** 0.020 *** 0.014 ***
(.004) (.003) (.003) (.003)
Number of Obs 5,119,324 5,119,324 5,119,324 5,119,324
Firm-Household FE x x x x
Time x County FE x x x x
51
Table A.5: Marginal Propensity to Consume out of Business Outcomes
Notes: This table reports 2SLS-IV estimates of the owner households’ consumption response per dollar change in business revenue, expense,
and profit margin using equation (10). Columns 1 and 2 use variation by industry due to SIP or SIP and infections and Columns 3 and 4 use
that due to NPI strictness or NPI strictness and infections as the excluded instruments. NPI strictness is as defined in Table 4. Outcomes are in
dollars (level), and all regressions include firm and household pair fixed effects and time × county fixed effects. Therefore, the estimated
coefficients can be interpreted as consumption declines (in dollar unit) per each dollar reduction in business outcomes. Firms that operate in
sub-industries with less than 30 firms are dropped from the estimation. *** p < 0.01, ** p < 0.05, * p < 0.10.
Variation by industry due to Variation by industry due to
SIP NPI strictness SIP NPI strictness
Endogenous SIP and infections NPI strictness and infections SIP and infections NPI strictness and infections
Variable (1) (2) (3) (4) (5) (6) (7) (8)
Committed Consumption Discretionary Consumption
A. March 2020 Event Study
Revenue -0.003 . 0.000 0.000 0.002 . 0.005 *** 0.005 *** 0.008 *** 0.008 ***
(.002) (.001) (.002) (.001) (.001) (.001) (.001) (.001)
Expense 0.000 0.004 * 0.007 ** 0.008 *** 0.008 *** 0.009 *** 0.014 *** 0.013 ***
(.002) (.002) (.002) (.002) (.002) (.001) (.002) (.001)
52
Net Income -0.025 *** -0.011 ** -0.023 *** -0.011 ** 0.006 . 0.012 *** 0.006 . 0.012 ***
(.005) (.004) (.005) (.004) (.004) (.003) (.003) (.003)
Number of Obs 2,353,988 2,353,988 2,353,988 2,353,988 2,353,988 2,353,988 2,353,988 2,353,988
B. March 2021 Event Study
Revenue 0.018 *** 0.013 *** 0.012 *** 0.012 *** 0.007 ** 0.010 *** 0.013 *** 0.012 ***
(.003) (.002) (.002) (.002) (.003) (.002) (.001) (.001)
Expense 0.027 *** 0.020 *** 0.016 *** 0.017 *** 0.009 * 0.015 *** 0.016 *** 0.017 ***
(.005) (.003) (.002) (.002) (.004) (.002) (.002) (.002)
Net Income 0.020 ** 0.010 * 0.020 ** 0.009 * 0.015 ** 0.009 * 0.028 *** 0.015 ***
(.007) (.005) (.006) (.005) (.006) (.004) (.005) (.004)
Number of Obs 3,011,272 3,011,272 3,011,272 3,011,272 3,011,272 3,011,272 3,011,272 3,011,272
Firm-Household FE x x x x x x x x
Time x County FE x x x x x x x x
Figure A.1: Average business outcomes of a random subset of all businesses (levels)
Notes: This figure shows average weekly dollar levels of business revenues, expenses, profits, and
household consumption from the week starting December 30th, 2019 to the week starting May 25th,
2020 for a random subset (20%) of businesses in the all businesses sample. Dotted vertical lines denote
the week of national emergency, which was declared the week starting March 9th, 2020. Blue horizontal
lines denote the average of respective outcomes between January 13, 2020 to February 9, 2020 (i.e., two
months before the week of national emergency). This figure illustrates that the business owners sample
that we use for our main analysis have similar declining patterns as the broader sample of businesses in
the all businesses sample.
A. Revenues B. Expenses
6500 ●
6500 ●
6000 ● 6000 ●
●
● ●
●
● ● ●
●
●
5500 5500 ●
Level ($) Level ($)
● ● ● ●
●
●
● ●
5000 5000 ●
● ● ● ●
● ●
● ●
4500 ● ● 4500 ● ●
●
●
●
● ● ● ●
● ●
4000 4000
−10 −8 −6 −4 −2 0 2 4 6 8 10 12 −10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency Weeks Since National Emergency
C. Profits
200
●
100 ●
●
●
0 ●
●
● ●
Level ($)
●
−100
●
● ● ●
−200 ●
● ●
● ● ●
−300
●
●
−400 ●
−500
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
53
Figure A.2: Average changes in business outcomes of a random subset of all businesses
Notes: This figure shows average weekly changes in business revenues, expenses, profits, and household
consumption from the week starting December 30th, 2019 to the week starting May 25th, 2020 for
a random subset (20%) of businesses in the all businesses sample. Outcomes are normalized by the
centered 9-week average from a year ago, and the change is defined as a percent change from its own
average between January 13, 2020 and February 9, 2020 (i.e., two months before the week of national
emergency). Dotted vertical lines denote the week of national emergency, which was declared the week
starting March 9th, 2020.
A. Revenue B. Expense
30 30
20 ● 20 ●
●
●
10 ● 10 ●
● ●
Percent change Percent change
● ●
0 ● ● 0
● ●
● ●
●
−10 −10 ●
●
●
● ● ●
● ●
−20 −20 ●
●
● ● ● ●
● ●
−30 ● −30
● ●
●
−40 ● ● −40 ● ●
● ● ●
−50 −50
−10 −8 −6 −4 −2 0 2 4 6 8 10 12 −10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency Weeks Since National Emergency
C. Profit
10
8
●
6 ●
● ●
4
Percent change
●
2 ●
● ●
0
●
−2 ● ●
● ●
−4 ● ●
● ●
−6 ●
●
● ●
−8
●
−10
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
54
Figure A.3: Decomposition of average changes in business outcomes and household consumption
Notes: This figure shows a decomposition of the observed decline in business outcomes and owner’s
consumption. Outcomes are normalized based on the centered 3-month average from the previous year,
with changes defined as percent deviations from their averages between September 2019 and March
2020 (i.e., the six months prior to the national emergency). Black lines plot average monthly changes in
respective outcomes. Red and blue lines plot average changes net of changes predicted by the effects
of local infections and SIP on these outcomes. Specifically, we subtract predicted changes in outcomes
using the estimated effects of local infections and SIP reported in columns 5 and 6 of Table 3 that include
time effects (blue) and do not include time effects (red): Yi,t − β̂ 1 LIRc(i),t − β̂ 2 1[SIPs(i),t ] − Xi,t . Since
the combined effects of local infections and SIP on revenues, for example, are negative, the red line in
panel A can be interpreted as average changes in revenues that would have prevailed in the absence of
changes in infections, SIP, or other factors that correlate with infections and SIP. The gap between the
black and red lines captures the effect of revenue changes explained by infections and SIP. Since the blue
lines are constructed using estimates including time-effects, the gap between the blue and black lines
captures the effect of local infections and SIP that is solely driven by cross-sectional differences. The
dotted vertical lines indicate two key points: the month of the national emergency declaration (March
9th, 2020) and one year after the declaration.
A. Revenues B. Expenses
100 100
80 Mar 20 Mar 21 80 Mar 20 Mar 21
60 60
40 40
20 20
0 0
−20 −20
−40 −40
−60 −60
−6 0 6 12 18 −6 −4 −2 0 2 4 6 8 10 12 14 16 18
Months Since National Emergency Months Since National Emergency
C. Profit Margins D. Consumption
80
Mar 20 Mar 21 60 Mar 20 Mar 21
40
6
20
0
−20
−40
−14 −60
−6 −4 −2 0 2 4 6 8 10 12 14 16 18 −6 −4 −2 0 2 4 6 8 10 12 14 16 18
Months Since National Emergency Months Since National Emergency
55
Figure A.4: Average changes in business revenues and owners’ consumption by industry perfor-
mance for the lowest and the middle three deciles
Notes: This figure shows average weekly changes in business revenues (blue) and owners’ consumption
(red) for businesses in the middle and bottom three deciles of industry performance. Outcomes are
normalized by the centered 9-week average from a year ago, and the change is defined as a percent
change from its own average between January 13, 2020 and February 9, 2020 (i.e., two months before the
week of national emergency). Dotted vertical lines denote the week of national emergency, which was
declared the week starting March 9th, 2020. Line types capture different industry rankings in terms of
their average drop in revenues since the onset of the national emergency. Each decile contains roughly
20 industries. For this analysis, we restrict the sample to industries with at least 100 businesses.
A. Middle Three Deciles
40
20 ●
●
● ●
●
● ●
● ● ● ●
Percent change
●
0 ●
●
● ●
●
●
●
● ●
●
● ● ● ●
● ●
● ●
● ●
−20 ●
●
● ●
●
●
●
● ●
● ● ● ●
● ● ●
● ● ●
−40 ●
●
●
●
●
●
●
● ●
● ●
−60
−80
−100
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
4th Lowest 5th Lowest 6th Lowest ● Consumption ● Revenue
B. Bottom Three Deciles
40
20
● ●
●
●
● ● ● ●
Percent change
0 ●
●
●
●
● ● ●
●
●
●
● ●
● ● ●
● ●
● ● ●
●
−20 ●
●
● ● ●
●
−40 ● ●
● ●
●
●
●
●
● ●
● ● ●
● ●
●
−60 ● ●
●
●
●
● ●
●
● ●
−80 ●
●
−100
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
Lowest 2nd Lowest 3rd Lowest ● Consumption ● Revenue
56
Figure A.5: Average changes in business revenue by top and bottom performing sectors
Notes: This figure shows average weekly changes in business revenues for businesses in the top five and
the bottom five performing industries. Industry performance is defined as the average drop in revenues
by NAICS 4-digit industries since the onset of the national emergency. Panel A plots revenue series
for businesses in the five least affected (i.e., best performing) industries, and Panel B plots those for
businesses in the five most affected (i.e., worst performing) industries. Industries shown in this figure
correspond to the underlying sectors in the least and the most affected industries shown in Figure 7. For
this analysis, we restrict the sample to industries with at least 100 businesses. Outcomes are normalized
by the centered 9-week average from a year ago, and the change is defined as a percent change from
its own average between January 13, 2020 and February 9, 2020 (i.e., two months before the week of
national emergency). Dotted vertical lines denote the week of national emergency, which was declared
the week starting March 9th, 2020.
A. Least affected industries
80 ●
●
60 ●
● ●
●
●
40 ●
● ● ●
Percent change
● ● ●
●
● ● ●
● ● ● ● ● ●
● ●
20 ●
●
● ●
●
●
●
●
● ●
●
●
●
●
● ● ● ● ●
● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ●
● ● ● ●
0 ●
●
● ●
●
● ●
●
● ●
●
●
●
● ● ● ●
●
●
●
● ● ●
● ●
●
−20 ● ●
●
● ●
●
● ● ●
● ●
−40 ●
−60
−80
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
● Beer, Wine, and Liquor Stores ● Nursing Care Facilities ● Funeral Homes/Cemeteries
● Toilet Preparation/Detergent Manfg ● Chemical Product Manfg
B. Most affected industries
40
● ●
20 ●
●
●
● ●
●
● ● ● ● ● ●
● ● ● ● ●
0 ● ● ● ● ●
● ●
Percent change
●
● ● ● ●
● ● ● ● ● ● ●
● ● ● ●
−20 ● ●
● ●
● ●
● ●
●
−40 ●
●
●
−60 ●
● ●
●
● ●
● ●
● ● ● ● ●
−80 ● ●
●
●
●
●
●
● ● ●
●
●
●
● ● ● ●
● ●
● ● ● ● ● ● ●
● ● ● ●
● ●
−100 ●
● ●
−120
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
● Vending Machine Operators ● Taxi/Limo Service ● Consumer Goods Rental
● Travel/Tour Agencies ● Drinking Places
57
Figure A.6: Average changes in household consumption by top and bottom performing sectors
Notes: This figure shows average weekly changes in business owner’s consumption for businesses in
the top five and the bottom five performing industries. Industry performance is defined as the average
drop in revenues by NAICS 4-digit industries since the onset of the national emergency. Panel A plots
revenue series for businesses in the five least affected (i.e., best performing) industries, and Panel B plots
those for businesses in the five most affected (i.e., worst performing) industries. Industries shown in
this figure correspond to the underlying sectors in the least and the most affected industries shown in
Figure 7. For this analysis, we restrict the sample to industries with at least 100 businesses. Outcomes
are normalized by the centered 9-week average from a year ago, and the change is defined as a percent
change from its own average between January 13, 2020 and February 9, 2020 (i.e., two months before the
week of national emergency). Dotted vertical lines denote the week of national emergency, which was
declared the week starting March 9th, 2020.
A. Least affected industries
40
●
20 ●
●
● ●
●
●
●
●
●
● ●
● ● ● ●
● ●
● ● ●
Percent change
●
0 ●
●
●
●
●
● ●
● ●
●
● ●
● ●
● ● ●
● ●
● ● ● ●
● ●
● ● ● ●
● ●
−20 ●
●
●
●
●
●
● ●
● ●
● ●
● ● ●
●
● ●
● ● ●
−40 ●
● ●
●
●
●
●
● ● ●
●
● ● ●
● ● ●
−60 ●
●
● ●
● ●
●
●
●
−80 ● ●
●
−100
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
● Beer, Wine, and Liquor Stores ● Nursing Care Facilities ● Funeral Homes/Cemeteries
● Toilet Preparation/Detergent Manfg ● Chemical Product Manfg
B. Most affected industries
40 ●
●
20 ●
●
●
●
●
●
●
● ●
●
●
● ● ● ●
● ● ● ● ● ●
●
Percent change
0 ● ●
●
● ● ●
●
●
●
●
● ●
●
●
●
● ● ●
●
● ● ● ● ●
● ● ● ● ●
● ●
●
−20 ●
●
●
●
● ●
●
●
● ●
●
●
● ● ● ● ● ●
● ● ●
● ● ● ● ●
● ●
−40 ● ●
●
●
● ●
● ●
●
●
●
● ● ●
● ●
−60 ●
−80
−100
−10 −8 −6 −4 −2 0 2 4 6 8 10 12
Weeks Since National Emergency
● Vending Machine Operators ● Taxi/Limo Service ● Consumer Goods Rental
● Travel/Tour Agencies ● Drinking Places
58
File and source
- File
- kim-2020-revenue-collapses-and-the-consumption-of-small-business-owners-in-the-early-stage.pdf
- Size
- 572,335 bytes
- SHA-256
- f4726dc3b3d765d5dbeed09ae057b6f651d350ffe3f4442cd37874c49adabc98
- Our copy
- kim-2020-revenue-collapses-and-the-consumption-of-small-business-owners-in-the-early-stage.pdf
- Original
- www.nber.org