Smbliq C 079 Business Exit During The Covid 19 Pandemic Non Traditional Measures In Historical 2021 04
Summary
Federal Reserve Board Finance and Economics Discussion Series paper 2020-089, Business Exit During the COVID-19 Pandemic: Non-Traditional Measures in Historical Context, by Leland D. Crane, Ryan A. Decker, Aaron Flaaen, Adrian Hamins-Puertolas and Christopher Kurz, April 2021. The abstract states that the authors review official data on business exit and explore alternative measures, including novel measures based on paycheck issuance and phone-tracking data. It states that business exit is countercyclical, with about 7.5 percent of firms exiting annually in recent years, and that the evidence is mixed, with exit elevated in certain sectors during the first year of the pandemic but many industries likely seeing lower-than-usual exit rates.
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
“Smbliq C 079 Business Exit During The Covid 19 Pandemic Non Traditional Measures In Historical 2021 04.”
Original at www.federalreserve.gov. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/smbliq-c-079-business-exit-during-the-covid-19-pandemic-non-traditiona-d369b89827/
, SHA-256 8f77b1e2534aa22df6fc6ecc0c9f3c1021c2a6b89edb58657ab9089a945b7509.Full text
Finance and Economics Discussion Series
Divisions of Research & Statistics and Monetary Affairs
Federal Reserve Board, Washington, D.C.
Business Exit During the COVID-19 Pandemic: Non-Traditional
Measures in Historical Context
Leland D. Crane, Ryan A. Decker, Aaron Flaaen, Adrian
Hamins-Puertolas, and Christopher Kurz
2020-089
Please cite this paper as:
Crane, Leland D., Ryan A. Decker, Aaron Flaaen, Adrian Hamins-Puertolas, and
Christopher Kurz (2021). “Business Exit During the COVID-19 Pandemic: Non-
Traditional Measures in Historical Context,” Finance and Economics Discussion Se-
ries 2020-089r1. Washington: Board of Governors of the Federal Reserve System,
https://doi.org/10.17016/FEDS.2020.089r1.
NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary
materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth
are those of the authors and do not indicate concurrence by other members of the research staff or the
Board of Governors. References in publications to the Finance and Economics Discussion Series (other than
acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.
Business Exit During the COVID-19 Pandemic:
Non-Traditional Measures in Historical Context
Leland D. Crane Ryan A. Decker Aaron Flaaen
Adrian Hamins-Puertolas Christopher Kurz
April 2021
Abstract
Lags in official data releases have forced economists and policymakers to leverage “alter-
native” or “non-traditional” data to measure business exit resulting from the COVID-
19 pandemic. We first review official data on business exit in recent decades to place
the alternative measures of exit within historical context. For the U.S., business exit
is countercyclical and fairly common, with about 7.5 percent of firms exiting annually
in recent years. Both the high level and the cyclicality of exit are driven by very small
firms and establishments. We then explore a range of alternative measures of business
exit, including novel measures based on paycheck issuance and phone-tracking data,
which indicate exit was elevated in certain sectors during the first year of the pandemic.
The evidence is mixed, however; many industries have likely seen lower-than-usual exit
rates, and exiting businesses do not appear to represent a large share of U.S. employ-
ment. Actual exit is likely to have been lower than widespread expectations from early
in the pandemic. Moreover, businesses have recently exhibited notable optimism about
their survival prospects.
JEL codes: D22, E32, C55, C81
All authors are from the Federal Reserve Board. First draft: October 2020. Tomaz Cajner provided
critical support and feedback for this project. We thank Tim Dore, Nathan Goldschlag, John Haltiwanger,
Traci Mach, Brendan Price, John Stevens, and seminar participants at the 2020 Banca d’Italia and FRB
Joint Conference, the CAED 2020 seminar series, the Federal Reserve Board, the Boston Fed, and the
International Monetary Fund for helpful discussions. We thank Anbar Aizenman and Betsy Vrankovich for
technical assistance and Rosemary Rhodes, Ashley Sexton, and Jacob Williams for research assistance. We
thank Ray Sandza, Adam Liem, and Homebase for providing Homebase data and technical support. We
thank SafeGraph for providing access to their data and Ryan Fox Squire in particular for answering questions
related to the data. We thank ADP and Matt Levin, Ahu Yildirmaz, and Sinem Buber for access to and
help with their payroll microdata. The analysis and conclusions set forth are those of the authors and do
not indicate concurrence by other members of the research staff or the Board of Governors of the Federal
Reserve System.
1 Introduction
Widespread business exit—death—resulting from the Pandemic Recession would have long-
lasting consequences for the U.S. economy. Unfortunately, actual business exit is difficult to
measure in real time since official statistics on business dynamics are released with substantial
lags: Bureau of Labor Statistics (BLS) data on establishment deaths during mid-2020 will
become available in late 2021, and Census Bureau data on firm deaths will likely not be
public until 2023. As a result, economic commentators and policymakers have been relying
on “alternative” or “non-traditional” data to measure business exit. For example, Figure
1 shows several popular indicators of business distress. Business electricity accounts show
little imprint of the recent economic stress while vacancy rates for office and retail are
reaching levels last seen during the Great Recession. Similarly, defaults have jumped, but
both “going out of business” search queries from Google and 30-day defaults have returned
to trend after brief elevation.1 While the many indicators made available by private firms
have improved understanding of recent economic developments, it is critical to be aware of
historical patterns of business shutdown and how popular alternative indicators compare.
In this paper we review official data on business closures and deaths before the pandemic,
providing a set of stylized facts that are necessary for evaluation of alternative indicators
of business shutdown. We then evaluate a range of alternative indicators—including several
new measures of exit we formulate—and discuss what they suggest about business exits
during the first year of the COVID-19 pandemic.
Official measures of business exit provide valuable historical context in which to evaluate
alternative indicators. These statistics reveal business death to be a common occurrence,
with about 7.5 percent of firms and 8.5 percent of establishments exiting annually in recent
years. Various measures of business closure—temporary and permanent—have been coun-
tercyclical in the past and rose notably during the Great Recession. Levels and cyclicality
of business death are driven primarily by extremely small firms and establishments—those
with fewer than 5 employees—though larger firms often permanently close individual estab-
lishments (locations) as part of geographic or industry restructuring. The facts we document
are interesting independent of COVID-19 considerations.
Alternative indicators of exit during the pandemic’s first year, on balance, suggest that
exit has been elevated at least among small firms and establishments and particularly in the
sectors most exposed to social distancing, though this elevated exit was partially offset by
reduced exit in pandemic-friendly industries. A rough estimate is that the most troubled
sector, other services (NAICS 81, which includes barber shops and nail salons), saw the
permanent exit of more than 100,000 establishments above and beyond historically normal
1
Monthly electricity account data available from the EIA Electricity Data Browser at
https://www.eia.gov/electricity/data/browser/; we thank Josh Blonz for this idea and Jacob Williams for
data assistance. Commercial real estate daily vacancy rates are a simple average across Chicago, Dallas,
New York, Los Angeles, and San Francisco from Costar, a provider of real estate analytics; see Costar
Group, Inc. (2020)). Google Trends monthly data taken from https://trends.google.com/trends/ with the
search term “going out of business.” Small business delinquency rates are from Paynet, a provider of lending
and banking analytics; see Paynet, Inc. (2020) and http://sbinsights.paynetonline.com/loan-performance/.
2
Business electricity accounts Commercial real estate vacancy rates
19 1000 16 11
Office (left)
950 15 Retail (right) 10
Industrial (right)
18.5 900 14 9
Retail, Industrial
850 13 8
Commercial Industrial
18 800 Office 12 7
750 11 6
17.5 700 10 5
Commercial (left) 650 9 4
Industrial (right)
17 600 8 3
08 10 12 14 16 18 20 22 06 08 10 12 14 16 18 20 22
20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20
Source: EIA through January 2021. Source: Costar Group, Inc. data through February 2021.
Google trends: Going out of business Small business delinquency rates
100 100 7 7
30-day
90 90 90-day
6 Defaults 6
80 80
70 70 5 5
60 60
Rate (%)
4 4
Index 50 50
3 3
40 40
30 30 2 2
20 20
1 1
10 10
0 0 0 0
06 08 10 12 14 16 18 20 22 06 08 10 12 14 16 18 20 22
20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20
Source: Google Trends through February 2021 Source: Paynet, Inc. (2020) data through January 2021.
Figure 1: High-frequency indicators of business distress
exit levels during the 12 months of March 2020 through February 2021. Results for other
sectors may have been more mixed; for example, within the leisure and hospitality sector,
some businesses—like full-service restaurants—saw significantly elevated exit, while other
businesses—such as those focused on outdoor recreation—saw exit rates similar to, or even
below, those of previous years such that exit was not dramatically elevated in the sector
overall. The retail trade sector appears similar in that some industries—such as clothing
stores—saw elevated exit, while others—such as grocery stores—saw below-normal exit, so
the sector overall likely saw only modestly elevated exit on net. Importantly, our best non-
traditional measures are more indicative of establishment than firm exit, though we do have
3
some firm-based indicators with useful insights.
Taken together along with some prudent guesswork, our sector-level results suggest econ-
omywide excess establishment exit—that is, exit above and beyond pre-pandemic rates—was
likely below 200,000 establishments during the first year of the pandemic, implying an exit
rate about one-quarter to one-third above normal. This is roughly consistent with what we
find from rough calculations on official BED data through the second quarter of 2020, which
suggest fewer than 150,000 excess establishment deaths during the first half of that year.
We have less insight into firm exit, though given historical patterns 200,000 excess establish-
ment exits would imply roughly 130,000 excess firm exits. Several sectors never saw elevated
exit expectations, at least among small businesses, and as of March 2021 forward-looking
firm exit expectations were actually below historical exit rates in most sectors. Relative
to popular discussion, then, our results may represent an optimistic update to views about
pandemic-related business failure. Throughout the paper, though, we emphasize the limita-
tions of our non-traditional data, which may lead us substantially to understate or overstate
exit. Moreover, even as the pandemic ebbs, business exit risk may remain material as busi-
ness owners face overdue credit bills, deferred rent, and other expenses that may have been
delayed but not eliminated by pandemic norms and policies (see, e.g., Simon and Haddon,
2021; Gournichas et al., 2021).
We draw these inferences from a number of timely, high-frequency business exit indicators,
some of which have been used in existing literature. Two key contributions in this respect,
however, are the construction of employment-weighted exit indicators from ADP payroll
data and a permanent business exit measure based on SafeGraph cell phone geolocation
data. We also review more commonly used data on small business operations (from Womply
and Homebase) as well as Census Bureau survey data on business exit expectations.2 As
we show, these alternative measures are most useful in the context of historical patterns of
business exit.
A key challenge is distinguishing between temporary shutdown and permanent shutdown
(death). U.S. statistical agencies provide data on business deaths, but identifying deaths is
more difficult in alternative data sources.3 Typically what can be measured is whether a
business is engaged in normal activities—e.g., receiving customer traffic, completing trans-
actions, or paying workers. We use the term “shutdown” to refer broadly to businesses not
engaged in normal activities, whether temporarily or permanently, and we attempt (loosely)
to make guesses about actual exits based on how long businesses have been inactive. We
use the terms “death” and “exit” interchangeably to refer to likely permanent shutdown.
We note, however, that the line between temporary and permanent shutdown may be blurry
at times, as even temporary shutdown may have significant economic implications if it is
prolonged or results in substantial business restructuring before reopening.
A handful of papers study business closure early in the pandemic. Bartik et al. (2020b)
2
We also provide brief discussion of more well-known private sector business registers like Yelp and Dun
& Bradstreet.
3
Permanent business death can be difficult to define and measure even in official data; for example,
Sadeghi (2008) shows that establishment death numbers depend materially on how long an establishment
must be inactive to count as a death (see Charts 10 and 11 therein).
4
surveyed firms in late March and early April and found temporary closure rates around
40 percent with permanent closure rates around 2 percent. Fairlie (Forthcoming) exploits
Current Population Survey (CPS) data on self-employment and finds that the number of
operating businesses in June was down 8 percent since March (note that we limit our focus
to employer businesses). Bartlett and Morse (2020) study business survival through April
in the City of Oakland using survey data and other sources. Wang et al. (2020) report that
2020 bankruptcy filings by small businesses through August were significantly lower than
in prior years, a topic also explored by Greenwood et al. (2020). Bartik et al. (2020a) and
Kurmann et al. (2020) measure early business closures in Homebase data, and Chetty et
al. (2020) measure early closures in Womply data. Hamilton (2020) uses Womply and Yelp
data to estimate that as of July 2020, roughly 400,000 businesses had permanently closed
during the pandemic; this number has been widely cited but is likely to be an overestimate
in light of official and non-traditional data we review here, which were not available at the
time Hamilton (2020) was written.4 A particular challenge with earlier estimates is that a
large number of early business closures may have been temporary, which is shown by Cajner
et al. (2020) and corroborated in the data we describe below.
Dalton et al. (2020a) and Dalton et al. (2020b) are the only papers studying business
closure in official business data; the authors use establishment microdata from the Cur-
rent Employment Statistics (CES)—the data behind the widely followed U.S. monthly jobs
report—along with firm information housed in internal BLS databases to study business
closure through November 2020 by firm size and establishment industry. The authors find
corroboration for the view that a large share of establishment closures during April and May
were temporary, though they note that closure appears to have stabilized by July. They
also find that establishment closure was far more elevated among small firms than among
large firms (and more elevated among large multi-establishment firms than large single-
establishment firms). As we argue below, the results in Dalton et al. (2020b) suggest that
the share of employment associated with cumulative establishment closures through Novem-
ber 2020 was not dramatically above that of previous years, consistent with our findings.
Interestingly, the authors find that about a quarter of job losses among restaurants are due
to establishment closure, while the closure share is closer to half among drinking places.
We first provide general background on the importance of business exit, drawing from
literature and the unique aspects of the COVID-19 pandemic (Section 2). We explore his-
torical patterns of business exit, summarizing them as a list of stylized facts, in Section 3.
We review a range of official and non-traditional measures of business shutdown during 2020
in Section 4, then we survey data on small business expectations in Section 5. We take stock
and conclude in Section 6.
4
The 400,000 figure has been cited not only by high-level government officials but also in academic
literature, e.g., Wang et al. (2020). Data from the BED (reviewed below) suggest that, at most, 350,000
establishments—and even fewer firms—closed permanently during the second quarter of 2020.
5
2 Background on business exit
Business exit is a natural phenomenon in healthy market economies, and while the closure
of weak businesses can be productivity enhancing, business exit imposes profound costs on,
workers, owners, and communities. Existing literature shows that death, as part of a broader
set of business dynamics patterns, enhances aggregate productivity as lower-productivity
exiting businesses are replaced by higher-productivity entrants.5 These patterns also release
resources, both employment and capital, to be used more efficiently elsewhere. Some business
exits occur as large firms restructure their activities across industry and geography, closing
some establishments while opening others to better meet demand or to adjust to changing
global supply chains (e.g., Davis et al., 2014; Fort et al., 2018).
The pandemic-induced recession raises concerns about whether exit can be as productiv-
ity enhancing as it has been in the past. In the current environment, patterns of business
exit may be driven by the geographic, industrial, and temporal onset of severe infection out-
breaks and lockdowns rather than business productivity. As a result, many high-productivity
businesses might fail during the COVID-19 episode, potentially implying negative effects on
aggregate productivity. In a broader sense, exit selection—–the relationship between business
productivity and exit–—has been weakening for decades (Decker et al., 2020), and patterns
of business exit were less productivity enhancing during the Great Recession than in previous
recessions (Foster et al., 2016). For these reasons, we might expect exit selection to be less
healthy during the pandemic than in the past.6 Moreover, to the extent that the industry
restructuring witnessed in 2020 (e.g., the decline in leisure & hospitality employment with
gains in pandemic-friendly industries) is temporary, business exits that occur in declining
industries may be especially wasteful.
Business exit implies permanent job destruction, potentially detaching workers from the
labor market and limiting the speed of the employment recovery.7 While the costs of exit-
induced layoffs may be manageable during periods of strong labor markets, releasing workers
onto labor markets at a time of high unemployment—such as the pandemic period—is more
concerning. Relatedly, business exit destroys the match-specific capital formed by a firm’s
relationship with its workers. As documented in Fujita and Moscarini (2017), a large share
of workers return to their former employers after being separated (“recall hires”). Business
exit eliminates this recall option and, potentially, implies longer unemployment spells for
5
For theoretical considerations around the role of exit in aggregate productivity see, e.g., Hopenhayn
(1992) and Hopenhayn and Rogerson (1993). For empirical explorations see, e.g., Bartelsman et al. (2013),
Decker et al. (2017), Decker et al. (2020), Foster et al. (2001), Foster et al. (2006), Foster et al. (2016),
and Syverson (2011). Exit, and firm turnover more generally, have been linked to industry cost structure
since industries with lower sunk costs exhibit higher firm turnover (both theoretically and empirically); see
Gschwandtner and Lambson (2012) and citations therein.
6
Additionally, policies designed to save businesses and jobs may have saved businesses that would have
exited had the pandemic never occurred, a source of potential misallocation that may result in additional
exit in the near future; for example, see Gournichas et al. (2021).
7
Exit is a significant contributor to overall job destruction. In Census Bureau Business Dynamics Statistics
data for 2015-2018, establishment exit accounted for 32 percent of annual gross job destruction, while firm
exit accounted for 17 percent.
6
workers that must form new employment relationships.
Business exit, particularly when it involves entire firms rather than single locations, also
means the destruction of firm-specific forms of intangible capital—brand value and tacit
knowledge about production or distribution—and costly reallocation of physical capital (see
Cooper and Haltiwanger, 2007). From the perspective of business owners, the exit of a firm
means not only the loss of a job and career disruption but also potentially the destruction of
household wealth. And from the perspective of local economies, widespread business deaths
may permanently alter the economic geography of neighborhoods and communities.
3 Historical patterns of business closure and death
Both the Bureau of Labor Statistics (BLS) and the Census Bureau publish official statis-
tics on business closure and death. The BLS publishes establishment closure and death
data through the quarterly Business Employment Dynamics (BED) product.8 These data
are based on the state and federal unemployment insurance data underlying the Quarterly
Census of Employment and Wages (QCEW) product, cover the near-universe of U.S. pri-
vate nonfarm business establishments with formal employees, and start in the early 1990s.9
The BLS provides two measures of business shutdown: establishment “closures” are estab-
lishments that had positive employment in the third month of the previous quarter but
zero employment (or no reported employment) in the third month of the current quarter,
and establishment “deaths” are establishments that have been closed for four consecutive
quarters.
Separately, the Census Bureau publishes both firm and establishment exit data through
the annual Business Dynamics Statistics (BDS) product.10 Firms and establishments with
positive employment in March of the previous year but no employment in March of the
current year are counted as deaths. While the BDS only provides annual data, rather than
quarterly as in the BED, the BDS has advantages of a longer time series (starting in the
late 1970s, though we focus on post-1983 data) and ability to distinguish between firm and
establishment deaths.
3.1 Levels and cyclicality
Figure 2 reports official data on business closures and deaths in recent decades. The top
panel reports annual firm and establishment death rates from the BDS through 2018, with un-
8
An “establishment” is a single business operating location (with few exceptions). A “firm” is a collection
of one or more establishments under common ownership or operational control.
9
The main omissions are railroads and some nonprofits that do not participate in state unemployment
insurance systems.
10
We use the current BDS redesign vintage introduced in September 2020. In the appendix we report
differences between the 2020 BDS vintage and the legacy vintage; our broad conclusions about historical
exit patterns are apparent in both vintages. BDS data are based on the Census Bureau’s Business Register,
which covers nearly the universe of private nonfarm business establishments with employees (the main
exceptions are railroads and certain financial entities).
7
weighted death rates (deaths as a share of establishments) on the left panel and employment-
weighted death rates (employment at deaths as a share of employment) on the right panel.
The bottom panel reports quarterly establishment closure and death rates from the BED
(seasonally adjusted) through 2019, again with unweighted rates on the left and weighted
rates on the right.11
BDS - Unweighted BDS - Employment weighted
14 14 7 7
Establishment deaths
Firm deaths
6 6
12 12
Annual rate (%) Annual rate (%)
5 5
10 10 4 4
3 3
8 8
2 2
6 6 1 1
82 86 90 94 98 02 06 10 14 18 82 86 90 94 98 02 06 10 14 18
19 19 19 19 19 20 20 20 20 20 19 19 19 19 19 20 20 20 20 20
BED - Unweighted BED - Employment weighted
8 8 2 2
Establishment deaths
Establishment closures
7 7
1.5 1.5
Quarterly rate (%) Quarterly rate (%)
6 6
5 5 1 1
4 4
.5 .5
3 3
2 2 0 0
92 96 00 04 08 12 16 20 92 96 00 04 08 12 16 20
19 19 20 20 20 20 20 20 19 19 20 20 20 20 20 20
Note: DHS denominators. BDS data are noisy around Economic Census years (2's and 7's).
Establishments are single operating business locations.
Firms are collections of one or more establishments under common ownership or operational control.
Source: BLS Business Employment Dynamics (BED), Census Bureau Business Dynamics Statistics (BDS).
Figure 2: Historical patterns of business shutdown
11
In all panels we use Davis et al. (1996) (DHS) denominators, where the current and previous quarters’ or
years’ values are averaged (using longitudinally precise lag values where possible). In the appendix, we repeat
the correlation and regression calculations that follow using the lag (i.e., initial) values only as denominators.
Our main stylized facts are unaffected by this specification choice.
8
Figure 2 shows that business shutdown and death are common occurrences. In recent
years (2015-2018), annual firm death rates have been around 7.5 percent of firms (about
400,000 per year), while establishment death rates have been around 8.5 percent at the annual
frequency (about 600,000 per year) or just over 2.5 percent at the quarterly frequency.12 A
comparison of the left and right panels reveals that business death comprises a much smaller
share of employment than of firms or establishments, implying that exit is concentrated
among smaller businesses (as we will explore further below).13 Figure 2 also suggests that
most measures of business shutdown are countercyclical, with particularly notable increases
during the Great Recession (when firm exit rates rose by roughly 1.5 percentage points
from the previous expansion low). Finally, a comparison of quarterly deaths and quarterly
closures in BED data indicates that temporary closure is common, affecting roughly 2 percent
of establishments or about 0.5 percent of employment each quarter (the difference between
the red and blue lines); this likely reflects some combination of typical seasonal business
suspensions and temporary periods of business distress.
We can more clearly observe the countercyclicality of business shutdown with some sim-
ple correlations, reported on Table 1. We compare our firm and establishment shutdown
measures—detrended with linear trends—with the change in the unemployment rate and
real GDP growth (at quarterly frequency for comparisons with BED measures and annual
frequency for comparisons with BDS measures).14
Table 1: Business cycle correlations
Unemployment GDP
Unweighted Weighted Unweighted Weighted
BED: Establishment closures (BLS) .34 .23 -.40 -.13
BED: Establishment deaths (BLS) .36 .31 -.44 -.21
BDS: Establishment deaths (Census) .64 .37 -.46 -.16
BDS: Firm deaths (Census) .41 .31 -.32 -.18
Note: Exit rates detrended linearly. BED correlated with quarterly change in unemploy-
ment rate or change in log real GDP. BDS correlated with annual change in unemployment
rate or change in log real GDP on BDS annual timing (April-March). BED data cover
1992q3-2019q1 (deaths) or 1992q3-2019q4 (closures). BDS data cover 1984-2018 (trends
estimated on full 1978-2018 sample).
12
Quarterly rates need not aggregate to annual rates due to short-lived establishments or potential dis-
crepancies between BDS and BED data.
13
The BLS BED program also publishes quarterly employment-weighted firm closure rates, not shown on
Figure 2, which have averaged about 0.7 percent per quarter in recent years. The BLS defines a firm based
on tax identifiers, while the Census Bureau definition is based on overall ownership or control.
14
We first take the average of the unemployment rate for the period; that is, we calculate the quarterly
average of monthly data for BED comparisons, and we calculate the annual average of monthly data for
BDS comparisons (where the BDS year t is defined as April t − 1 through March t). We then take the
difference in these quarterly or annual averages. Correlations with the level of unemployment rates (rather
than the change) would be difficult to interpret given persistence of unemployment rates in the aftermath
of recessions. For GDP correlations, we calculate the change in the log of real GDP at quarterly frequency
for BED comparisons or the change in the log of annual GDP (on BDS timing) for BDS comparisons, where
annual GDP is the average of the quarterly GDP level for the year.
9
Simple correlations are consistent with countercyclical shutdown and exit (i.e., positive
correlations with the change in unemployment and negative correlations with GDP growth).
Employment-weighted establishment shutdowns are less countercyclical than unweighted es-
tablishment shutdowns (again suggesting that smaller units drive exit cyclicality), and firm
death is less countercyclical than establishment death.15 The specifications in Table 1 show
consistent countercyclicality of both establishments and firms; in unreported results we find
that these are somewhat sensitive to inclusion of the 1980-1983 period, a volatile period for
exit measures in which firm exit is less countercyclical.16
In appendix Table A1 we detrend exit measures with the method of Hamilton (2018) and
find broadly similar results, though firm exit correlations with GDP are weaker under this
alternative detrending approach. In appendix Table A2 we report results specifying the GDP
component of the correlations using detrending methods of Hamilton (2018) and Hodrick
and Prescott (1997), in which countercyclical exit is still evident though the results for firms
versus establishments and weighted versus unweighted exit are less clear cut. In appendix
Table A12 we report correlations of exit with industrial production, another business cycle
concept, and find that exit is countercyclical with respect to this indicator as well.
The time series correlations above are suggestive but limited. We can gain more business
cycle variation using state-level data. Here we simplify by focusing on annual BDS data
during the post-2002 period, thereby avoiding the need to detrend the exit series (and also
avoiding the potentially problematic 2002 observation, discussed more in the appendix).
Table 2 reports results from regressions of exit rates on the annual change in unemployment
with state fixed effects (such that we study within-state business cycle fluctuations).
Business cycle correlations are substantial within states, with unweighted exit rates being
more cyclical than weighted rates for both establishments and firms (i.e., smaller units drive
the cyclicality) and firm exit being less cyclical than establishment exit.17 For example,
the interpretation of the first coefficient is that a one percentage point rise in the state
unemployment rate is associated with a 0.56 percentage point increase in the establishment
exit rate. Taken together with Table 1, the data consistently show that exit is countercyclical,
particularly in recent years.18
15
Correlations of the level of establishment deaths or firm deaths (rather than death rates) with the change
in unemployment also indicate countercyclical exit, as we show in appendix Table A3. Separately, as can be
seen from Figure 2, BDS data show notable spikes in exit in 2002 and other years ending in 2 or 7, suggesting
there may be data challenges created by semi-decadal Economic Censuses. The previous vintage of BDS
data displayed somewhat different patterns; see the appendix for analysis of Census year and vintage issues.
16
We intentionally begin the BDS-based sample in 1984 for Table 1 and all other results unless otherwise
noted. Firm countercyclicality is highly sensitive to specific years included in the 1980-1983 period. In
the appendix Table A8 we report correlations for the entire 1978-2018 period covered by the current BDS
vintage, in which firm exit countercyclicality is nevertheless apparent.
17
The 2003-2018 period included in the Table 2 estimates includes three Economic Census years (2007,
2012, and 2017), but the coefficients are little changed if Census years are omitted (see appendix Table A7).
Separately, coefficients are similar in regressions without state fixed effects or with both state and year fixed
effects (see appendix Table A9).
18
See Clementi and Palazzo (2016) for a model-based exploration of exit countercyclicality. Lee and
Mukoyama (2015) document the relative cyclicality of entry and exit of manufacturing plants.
10
Table 2: Business cycle comovement: States
Establishment death Firm death
Unweighted Weighted Unweighted Weighted
Change in unemp 0.56*** 0.22*** 0.37*** 0.11***
(0.03) (0.02) (0.02) (0.01)
Observations 816 816 816 816
State FE Yes Yes Yes Yes
Year FE No No No No
Note: Regression of annual exit rates on annual change in unemploy-
ment rates, 2003-2018. Unemployment rate changes timed to corre-
spond with BDS annual timing (April-March).
***denotes statistical significance with p < 0.01.
Source: Author calculations from Business Dynamics Statistics and
BLS unemployment data.
3.2 Exit by firm and establishment size
As noted above, the differences between unweighted and weighted exit rates suggest that
business exit is concentrated among smaller units, in terms of both overall exit rates and the
cyclicality of exit. This is made clear by Figure 3, which plots annual establishment death
rates (left panels) and firm death rates (right panel), reported by both firm (top panels) and
establishment (bottom panel) size categories.
The smallest firms—those with fewer than 5 employees—have the highest exit rates, with
typical rates markedly above any other firm size classes.19 Firm death rates decline mono-
tonically with firm size. Establishment death rates are likewise highest among the smallest
firms, but the largest firms—–those with 500 or more employees–—have the second-highest
establishment death rates; intuitively, large multi-unit firms may close many establishments
per year (and potentially open many others) as part of geographic or industrial restructur-
ing.20
The bottom panel shows that establishment death rates decline monotonically in estab-
lishment size, with the smallest establishments exiting at rates well above other size classes.
More broadly, Figure 3 makes apparent that the rise in business death during the Great
Recession was driven in large part by small firms and establishments.
19
Among firms with fewer than 5 employees, death rates are highest among firms that entered during the
previous year. In many years, death rates among these new, extremely small firms exceed 30 percent. Yet
even among firms at least 10 years old, death rates in this size category are around 10 percent. This may in
part reflect gradual downsizing of previously large firms that enter this small size category during their final
year.
20
Decker et al. (2016) find that, on net, multi-unit firms open establishments during expansions and close
establishments during recessions, which the authors rationalize using a model in which firms adjust product
or market in response to aggregate shocks; this can help explain the countercyclicality of firm volatility.
11
Establishment death by firm size Firm death by firm size
20 20 20 20
16 16 16 16
Percent of establishments
Percent of firms
12 12 12 12
8 8 8 8
4 4 4 4
0 19 82 0 0 19 82 0
19
19
19
19 86
90
94 19
19
19
19 86
90
94
20
20
20 98
02
06
10 20
20
20 98
02
06
10
20
20 14
18 20
20 14
18
Establishment death by establishment size
20 20
16 16
Percent of bottomleft
12 12
8 8
4 4
0 19 82 0
19
19
19
19 86
90
94
20
20
20 98
02
06
10
20
20 14
18
1-4 employees 5-9 employees 10-19 employees
20-99 employees 100-499 employees 500+ employees
Note: Unweighted exit rates with DHS denominators. BDS data are noisy in Economic Census years (2's and 7's).
Source: Census Bureau Business Dynamics Statistics (BDS).
Figure 3: Business death rates by firm and establishment size
12
3.3 Stylized facts about business shutdown
The above data suggest a set of stylized facts that must be kept in mind as alternative
measures of business shutdown are examined:
• Annual firm exit rates have averaged around 7.5 percent in recent years—roughly
400,000 firms—or 2 percent of employment.
• Annual establishment exit rates have averaged around 8.5 percent in recent years—
roughly 600,000 establishments—or 3.5 percent of employment.
• Quarterly establishment death rates have averaged about 2.5 percent, or about 0.5
percent of employment.
• Business exit is countercyclical; in particular, firm exit rates rose by about 1.5 percent-
age points in the Great Recession.
• The overall rate of business exit and the countercyclicality of exit are driven primarily
by very small firms and establishments.
• Temporary business closure is common, affecting about 2 percent of establishments
per quarter.
4 Has COVID-19 sparked a surge in business exit?
Official data on business exit are released with a lag: BED data on establishment closures
are currently available for the second quarter of 2020, but BED deaths for that quarter will
not be available until fall of 2021. BDS data on firm deaths during 2020 will (presumably)
not be available until 2023, as currently available data extend only through March of 2018.21
In the meantime, we must rely on some guesswork along with non-traditional measures of
business activity to assess the magnitude of business exit during 2020. We now describe
several such measures.
21
BED data are released with a lag of about two quarters, but deaths, by definition, are not observed
until three quarters later. The BDS reports business activity as of March 12 of each year and is typically
released with a lag of about two years (similar to the County Business Patterns product). Since much of the
economic fallout from COVID-19 occurred after March 12, 2020, most pandemic-related exit will be observed
in the BDS data for 2021. Statistical agencies face significant challenges when measuring business exit (and
entry), and this paper is in no way a criticism of the agencies for the lags with which these data are released.
Administrative source data are complicated and require careful processing, particularly when longitudinal
aspects are important (as is the case for business exit). Moreover, in the past business exit has not typically
received the intense research, policy, and media focus that has characterized the pandemic period; in recent
decades, exit has been less volatile than entry, and the latter tends to receive more focus given its relationship
to entrepreneurship. Therefore it is not surprising that exit is not measured and published in real time, even
in proxy form. The Census Bureau publishes proxy indicators of business entry—in the form of business tax
application data—in near real time as part of the Business Formation Statistics product.
13
4.1 Business Employment Dynamics data
While official data on firm and establishment deaths are not yet available, we can observe
BLS data on establishment closures—which include potentially temporary closures—during
the first and second quarters of 2020 in the BED. These data, which measure establishments
paying no employees in the third month of a given quarter (but with positive employment
in the third month of the previous quarter), provide a view into establishment shutdowns
during the first half of the year from high-quality administrative data.22 Figure 4 shows
establishment closure rates (left panel) and numbers of establishments and jobs associated
with closures (right panel).
Rates Levels
10 10 800 3000
Unweighted Establishments (left) 2750
Employment weighted 700 Employment (right)
2500
Establishments (thousands)
8 8
Employment (thousands)
600 2250
2000
Quarterly rate
6 6 500
1750
400 1500
4 4 1250
300
1000
200 750
2 2
500
100
250
0 0 0 0
q1 q2 q3 q4 q1 q2 q1 q2 q3 q4 q1 q2
20 20 20 20 20 20 20 20 20 20 20 20
19 19 19 19 20 20 19 19 19 19 20 20
Note: Includes potentially temporary closures.
Source: Census Bureau Business Dynamics Statistics (BDS).
Figure 4: Establishment closure rates, 2019-2020
Focusing on the left panel of Figure 4, the quarterly share of establishments closing rose
from just above 5 percent during 2019 to 6 percent during the first quarter of 2020 and nearly
9 percent during the second quarter. Second-quarter closures accounted for nearly 2.5 percent
of private sector employment.23 The right panel shows that 500,000 establishments closed
during the first quarter followed by more than 700,000 establishments in the second quarter;
the second-quarter pace exceeded the 2019 average pace by nearly 300,000 establishments.
Those establishments that closed in the second quarter were associated with nearly 2.9
million jobs.
Historically, roughly half of establishment closures result in permanent deaths.24 If this
historical share held during the first quarter of 2020 it would imply about 250,000 estab-
lishment deaths, roughly 25,000 above the 2019 quarterly average. For the second quarter,
22
The BLS has noted that data collection for the BED (and the related QCEW product) may have
been affected by the pandemic; see https://www.bls.gov/covid19/effects-of-covid-19-pandemic-on-business-
employment-dynamics.htm.
23
The BED quarterly employment-weighted firm closure rate, not shown on Figure 4, rose from an average
rate of 0.7 percent in 2019 to 0.8 percent in 2020q1 and 1.8 percent in 2020q2.
24
This share averaged 53 percent during the 2018q1-2019q3 period.
14
the historical share would imply about 350,000 total establishment deaths, about 125,000
above the 2019 quarterly average. But we expect that establishment closures during the
early months of the pandemic are more likely than normal to have been temporary (Cajner
et al., 2020) such that 600,000 total establishment deaths and 150,000 excess establishment
deaths are extreme upper bound estimates for actual deaths during the entire first half of
2020. Firm deaths (and excess deaths) would be lower still, historically equalling about
two-thirds of establishment deaths. Moreover, as we argue in section 4.1 above, Dalton et al.
(2020b) estimates from BLS microdata suggest that cumulative establishment closures—at
least on an employment-weighted basis—were roughly back on historical patterns by late
2020. For these reasons, the Womply-based estimate of Hamilton (2020)—constructed well
before BED data were available—that 400,000 firms closed permanently between March and
early July is probably somewhat too high.
The elevated establishment closure patterns of 2020 varied widely across sectors. Figure
5 shows closure rates (top panel) and employment-weighted closure rates (bottom panel)
during the second quarter (in blue) compared with the average of 2015-2019 (red) and
the maximum quarterly closure rate seen during the Great Recession of 2007-2009. Every
sector saw higher rates in the second quarter than the 2015-2019 average. Most sectors
also saw higher closure rates than the peaks of the Great Recession; the exceptions are
construction, financial activities, and (in employment-weighted terms) natural resources &
mining and manufacturing. While all sectors saw high closure rates, other services and
leisure & hospitality saw extremely high rates, exceeding 13 percent of establishments and
6.5 percent of private sector employment.
The closure rates seen in Figures 4 and 5 are instructive about the early pandemic period,
but two caveats are notable. First, to count as closed, an establishment must have no paid
employment for the third month of the quarter; establishments with paid workers in June
but none in April or May do not count as closed. Second, existing literature and data
we discuss below find that a large share of businesses that closed in the early months of
the pandemic reopened thereafter; our rough permanent death calculations above are only
suggestive upper bounds.
Here it is also useful to note that Dalton et al. (2020b), using BLS microdata but not
actual BED data, find that about 2.3 percent of February 2020 employment at small firms
and roughly 0.5 percent of February 2020 employment at large firms was associated with
establishments that were closed three quarters later in November 2020, for a rough overall
employment-weighted closure rate of about 1.4 percent for that period (calculated as the
weighted average of small and large firm rates using 2019 BED firm size weights; see Bureau
of Labor Statistics, 2020b, for BED firm size data). Roughly speaking (and ignoring sea-
sonality), this implies an annualized employment-weighted closure rate of 1.9 percent. This
is close to the typical annual employment-weighted establishment death rate seen in BDS
data for recent years (top-right panel of Figure 2) and actually well below typical annual
employment-weighted establishment closure rates reported in BED data, consistent with the
notion that business death has not accounted for a dramatically elevated share of job de-
15
Unweighted
Other Services
Leisure & Hospitality
Retail Trade
Transp. & Warehous.
Economywide
Construction
Manufacturing
Education & Health
Wholesale Trade
Prof & Business Svcs
Information 2020q2
Nat. Resources & Mining 2015-2019 average
Financial Activities 2007-2009 max
0 2 4 6 8 10 12 14
Percent of establishments
Employment weighted
Leisure & Hospitality
Other Services
Economywide
Information
Retail Trade
Transp. & Warehous.
Education & Health
Prof & Business Svcs
Construction
Wholesale Trade
Nat. Resources & Mining
Financial Activities
Manufacturing
0 1 2 3 4 5 6 7 8
Percent of employment
Note: Quarterly rates (seasonally adjusted). Bars ordered by difference between 2020q2 and
2015-2019 average. Includes temporary closures.
Source: BLS Business Employment Dynamics.
Figure 5: Establishment closure rates, 2020q2 versus history
16
struction during the pandemic.25 Based on these data, we might expect BED closure rates
to come down in the third quarter reading (to be released in late April 2021).
4.2 ADP
The best non-traditional measures of business closure come from business databases in which
data are generated automatically from actual business activity such as revenue transactions,
payroll events, or customer traffic. We first describe business closure measures based on
microdata from ADP, a provider of payroll processing services for businesses comprising
about one-fifth of total private sector employment.26 Key advantages of ADP data are
their comprehensive coverage across business size and industry cells as well as the ability to
observe both unweighted and employment-weighted business closure measures. A challenge
of these data is that some ADP clients may process payroll at the establishment level,
while others may process at the firm level or something in between; we follow Cajner et
al. (2018) in treating ADP units as establishments. Another challenge—in the context of
business shutdown—is that true shutdown cannot be distinguished from ADP client turnover,
a limitation present in many non-traditional business microdata sources.
In the ADP data, we observe paycheck issuance events at the business level. Pay fre-
quency varies by business—weekly, biweekly, or monthly—so we measure business shutdown
based on the length of time a business goes without issuing pay. Since we have a long history
of ADP data, we focus on comparing the 2020-2021 experience to the average experience
from recent years (2015-2019); this allows us to abstract from typical patterns of ADP client
turnover and determine the extent to which business shutdown has been elevated in the
pandemic relative to normal. We begin in mid-February 2020 and, for each week thereafter,
we tally up the share of businesses that were operating in February 2020 but are in the midst
of a shutdown period. We compare this share to the same-week average for the 2015-2019
period (i.e., the average for February cohorts of businesses starting in each of 2015-2019).27
In all cases, we apply sampling weights to ADP payroll units from the QCEW (with weights
in terms of NAICS sector and establishment size as of March of 2020).28
25
In addition to quarterly data, the BLS BED program publishes year-over-year establishment closure
data timed to March of each year; see Bureau of Labor Statistics (2020a). Annual employment-weighted
establishment closure rates in BED data average 2.8 percent during 2015-2019.
26
Cajner et al. (2018) describe ADP microdata in detail and document representativeness across business
size and industry. Cajner et al. (2020) use ADP data to explore various dimensions of the early pandemic
recession including business shutdowns and reopenings, temporary versus permanent job losses, and wage
dynamics.
27
For example, the 2019 cohort data extend from February 2019 through February 2020.
28
We also take the 2-week trailing moving average of all series to smooth through volatility associated with
varying pay frequencies across businesses. Importantly, in this and subsequent sections our exit measure
differs slightly from the DHS-based measure from BDS and BED data to which we make comparisons. In
non-traditional sources we will generally calculate exit rates as the share of firms operating at the beginning
of the pandemic that are not operating at some later date. This is necessary so that we do not have to make
assumptions about firm entry in the intervening period, which would be required for calculating the DHS
denominator. As we show in the appendix, in BDS data the choice of denominator is not quantitatively
critical for the types of exit measures we study, so we prefer to view this conceptual discrepancy as a
17
The top-left panel of Figure 6 shows the results for various shutdown durations. The
blue line shows the share of businesses that are in a shutdown spell of at least 25 days, in
2020 relative to the 2015-2019 average for a given week. By late April of 2020, the share of
businesses in a 25-day (or more) shutdown spell was nearly 12 percentage points higher than
it was at the same time in past years. After that time, however, closed businesses reopened
such that the share of businesses that were shut down returned to the historical pace by late
August.
Importantly, businesses that issue paychecks at a monthly frequency routinely count as
shutdowns according to the blue line, as they typically issue paychecks more than 25 days
apart. Staying in the top-left panel of Figure 6, the red line uses a more stringent criteria for
measuring shutdowns, reporting the share of businesses that were in shutdown spells of at
least 36 days (relative to the same measure in previous years). This ensures that businesses
that pay on any pay frequency—weekly, biweekly, or monthly—do not count as spurious
shutdowns. The red line indicates that by late April 2020, the share of businesses that were
in shutdown spells of at least 36 days exceeded historical patterns by more than 6 percentage
points. The black line focuses on shutdown spells of 70 days or more; if permanent shutdown
were elevated among ADP businesses, we would expect to see it in this line at least. While
the share of businesses shut down for 70 days or more exceeded historical averages by about
2 percentage points in early May, it has since returned to normal.
In short, while business shutdown, including shutdown spells of more than two months,
was elevated in the late Spring of 2020, by the end of August we observe no evidence of
excessive, ongoing business inactivity; in fact, shutdown was well below normal by late 2020.
We even do not see evidence of a shutdown surge during the latest wave of Covid cases of
November-February. The historically low pace of shutdowns in late 2020 likely reflects, in
large part, increased client retention during 2020, which has been noted in ADP earnings
calls. This makes the data difficult to interpret; still, though, if permanent death were
extremely elevated in 2020, it would likely be reflected in this large dataset.29 So the data,
while not dispositive, are at least suggestive.
The top-right panel of Figure 6 shows the same concepts in an employment-weighted form.
To calculate employment-weighted shutdowns in any given week, we identify businesses that
meet a given shutdown criteria (25 days, 36 days, or 70 days) then calculate their employment
share based on their February 15 employment as a share of total February 15 employment.
Hence, the top-right panel of Figure 6 shows the share of February’s employment that is
associated with businesses that shut down in some weeks thereafter, in 2020 compared with
source of measurement error with magnitude that is small relative to other sources of error inherent in our
non-traditional measures.
29
ADP’s fiscal Q4 2020 earnings report (ADP, 2020b), which covered the calendar quarter ending June
30, 2020, noted an overall decline in client retention in that quarter. But retention rebounded, apparently;
the fiscal Q2 2021 report (ADP, 2020a), which covered the calendar quarter ending December 31, 2020,
noted ”an all-time high retention level” and ”record client retention.” In the associated earnings call, ADP
executives noted that retention has been trending higher in recent years generally and that composition
shifts from older to newer ADP products naturally raise retention. Higher-than-historical retention helps
explain why relative shutdown measures shown on Figure 6 moved below zero after the summer surge.
18
Share of businesses closed by duration, Share of employment closed by duration,
difference from historical share difference from historical share
12 12 6 6
At least 25 days At least 25 days
10 At least 36 days 10 At least 36 days
Percentage points, Percentage points,
At least 70 days 4 At least 70 days 4
8 8
6 6
2 2
4 4
2020 vs. 2015-2019 average 2020 vs. 2015-2019 average
2 2 0 0
0 0
-2 -2 -2 -2
-4 -4
-4 -4
ry ay st be r ry ry ay st be r ry
br M gu br br M gu br
ua Au em ua ua Au em ua
Fe ov Fe Fe ov Fe
N N
Note: Share of Feb. 15, 2020 businesses issuing no Note: Share of Feb. 15, 2020 employment at businesses
paychecks, current minus same-week 2015-2019 issuing no paychecks, current minus same-week
average. 2015-2019 average.
Source: ADP. Weekly data through Feb. 27, 2021. Source: ADP. Weekly data through Feb. 27, 2021.
Share of businesses closed by initial size, Share of businesses closed by supersector,
difference from historical share difference from historical share
2 2 6 6
4 4
Percentage points, Percentage points,
1 1
2 2
0 0
0 0
2020 vs. 2015-2019 average 2020 vs. 2015-2019 average
-1 -1 -2 -2
-4 Manufacturing -4
-2 -2 Trade/Transportation/Utilities
1-49 employees Professional & Business Services
50-499 employees Education & Health Services
-3 500+ employees -3 Leisure & Hospitality
ry ay gu
t
be r ry ry ay gu
t
be r ry
ua M s ua ua M s ua
Fe Au ov Fe Fe Au ov Fe
br em br br em br
N N
Note: Share of Feb. 15, 2020 businesses issuing no Note: Share of Feb. 15, 2020 businesses issuing no
paychecks for 70+ days, current minus same-week paychecks for 70+ days, current minus same-week
2015-2019 average. 2015-2019 average.
Source: ADP. Weekly data through Feb. 27, 2021. Source: ADP. Weekly data through Feb. 27, 2021.
Figure 6: Measures of business closure from ADP payroll data (2020-2021 relative to 2015-
2019 average)
19
the 2015-2019 average for that week. Employment-weighted shutdown also peaked in late
April/early May 2020, when businesses inactive for at least 25 weeks accounted for a share of
February employment that exceeded past years by about 5 percentage points. In employment
terms, extremely long shutdown spells of 70 days or more were barely more common in mid-
2020 than in past years; and by late August the share of employment attached to closed
businesses was lower than average. As we argue in section 4.1, BLS data studied by Dalton
et al. (2020b) are also consistent with the notion that employment-weighted shutdown was
not markedly above historical norms by the third quarter of 2020.
The differences between the top-left and the top-right panels of Figure 6 suggest that the
elevated shutdown in spring 2020 relative to past years was driven largely by smaller units.
We can see this more clearly in the bottom-left panel of Figure 6, which shows (unweighted)
shutdown rates in 2020 relative to 2015-2019 averages, separated by business size (we focused
on shutdown spells of at least 70 days). The black line shows that shutdown rates among
the largest units—those with at least 500 employees—were similar to the pace of previous
years for much of 2020 then dipped even lower by the end of the year. Smaller units saw
significantly elevated shutdown rates in late April/early May 2020, but by August all business
sizes saw shutdown rates well below historical patterns.
Shutdown patterns do vary some across sectors. The bottom-right panel of Figure 6
shows unweighted 70-day shutdown rates for selected supersectors. The leisure & hospitality
supersector (NAICS 71-72) had historically high shutdown rates during spring and summer
2020, with shutdown rates exceeding past patterns by more than 5 percentage points in early
May. By late August, however, shutdown among leisure & hospitality businesses was only
modestly elevated relative to previous years, and by late-2020 the sector had joined other
sectors in seeing historically low shutdown activity. Indeed, the cross-sector convergence is
striking.
Taken at face value, the ADP data suggest that business shutdown was elevated during
the middle of the year, but on net these excess shutdowns were all temporary; since then,
shutdown has moved well below historical rates, even among small units. This is a striking
result, since the ADP data are reasonably comprehensive in terms of coverage across sectors
and establishment size classes and allow us to study how shutdown activity is affecting
employment; the measures we will review below are more limited in at least some of these
dimensions. As such, the ADP data provide some good news, suggesting that permanent
shutdown has not been materially elevated since the pandemic began and that the industries
hit hardest in the spring also saw relatively large recoveries.
If permanent business shutdown were accounting for a sizeable share of businesses and
employment, we would expect to see evidence of it in the large sample of ADP clients. We
again emphasize, however, that ADP data can be affected by patterns of client turnover
in addition to true business shutdown, and elevated customer retention is readily appar-
ent during the second half of the pandemic. Moreover, ADP data, like other client-based
databases, are subject to unobservable selection around which businesses choose to engage
ADP for payroll services. For these reasons, it is critical that we add other data sources to
20
our study.30
4.3 Small business trackers
We next turn to popular measures of small business activity, shown on Figure 7. The left
panel reports daily data from Womply, a credit card transaction processor, on the share of
firms that have ceased processing point-of-sale transactions since mid-February.31 The right
panel reports weekly data from Homebase, a provider of clock-in/clock-out tracking software,
showing the share of firms that have stopped reporting clock events since mid-February (and,
conveniently, we can observe 2019 data for Homebase as well).32 In both cases, the sample
of businesses is restricted to those that were operating in February 2020 (or February 2019
for the 2019 Homebase line), abstracting from entry into the sample (consistent with our
ADP-based exercises above). Importantly, presence in these datasets may be less costly
than for ADP data; for example, many Homebase clients use a free tier of the service, which
implies different selection dynamics than may be present in ADP data.
These measures, which are focused on small firms in customer-facing industries, suggest
that business shutdown rose sharply in March and April, but many closed businesses re-
opened in May and June. Still, the recent observations indicate that shutdown was indeed
elevated during the first year of the pandemic. Homebase data suggest that, as of the end
of February 2021, shutdown in well-covered industries was elevated by roughly 3 percentage
points relative to the same time a year earlier.
Importantly, a limitation that is common to ADP, Womply, and Homebase data is the
possibility that exit patterns are driven by client attrition rather than business shutdown.
Our comparisons to past-year patterns in ADP and Homebase data are designed to provide
perspective on this; roughly speaking, the question is not whether we observe exit in these
data but, rather, whether we see excess exit relative to historical patterns. This makes the
Womply data particularly difficult to interpret; and, indeed, exit appears far more elevated
30
Additionally, as noted above, some ADP payroll units are closer to firms than establishments. As such,
we would expect relatively low exit rates in ADP data relative to pure establishment-based measures.
31
Womply is a credit card analytics firm that aggregates data on card transactions. Data reported by
Womply reflect card transactions (or lack thereof) among small businesses as defined by the Small Busi-
ness Administration (see Chetty et al. (2020) or or https://www.womply.com/blog/data-dashboard-how-
coronavirus-covid-19-is-impacting-local-business-revenue-across-the-u-s/). We follow Chetty et al. (2020) in
treating Womply businesses as firms, though it is unclear from Womply documentation whether this is
accurate.
32
Homebase provides clock-in/clock-out software for small businesses and can therefore observe employ-
ment activity in close to real time. Homebase data were a good predictor of official CES monthly job
gains during the initial late-spring recovery; see Reinicke (2020). As of early 2020, Homebase data in-
cluded over 60,000 establishments with about 500,000 (hourly) employees. Coverage is concentrated among
very small establishments (mostly those with fewer than 20 employees) in retail and service industries that
happened to be particularly affected by social distancing (coverage is best in accommodation and food
services—NAICS 72—where Homebase has almost 2.5 percent of all U.S. establishments with fewer than
50 employees). We aggregate Homebase establishment data to the firm level using their (anonymized) com-
pany identifier. See Kurmann et al. (2020) for extensive detail on Homebase representativeness, and see
http://joinhomebase.com/data for more details on Homebase data.
21
Womply Homebase
60 60 50 50
40 40
Percent of businesses Percent of businesses
40 40
30 30
20 20
20 20
All
Education & Health 10 10
0 0
Leisure & Hospitality 2019
Retail & Transportation 2020
0 0
ry ay gu
t r
be ry ry ay gu
t r
be ry
ua M s m ua ua M s m ua
Fe Au ov Fe Fe Au ov Fe
br e br br e br
N N
Note: Share of Feb. 15, 2020 firms with no transactions. Note: Share of Feb. 15, 2020 firms with no clock-ins.
Daily data through Feb. 28, 2021. Weekly data through week ending Feb. 27, 2021.
Source: Womply data from Source: Homebase.
https://www.tracktherecovery.org and Chetty et al. (2020).
Figure 7: Small consumer business closures
in Womply than in our other sources, suggesting that normal attrition may be higher in
Womply data.33 The Homebase data provide better context and are suggestive of, though
not unambiguous confirmation of, elevated exits during the pandemic. We next turn to a
measure that is not subject to the particular limitation of client attrition.
4.4 SafeGraph
We construct a final direct measure of business shutdown based on information from Safe-
Graph, a data company that aggregates anonymized location data from numerous mobile
device applications to provide insights about physical places.34 The company has temporar-
ily made their micro-level data available to researchers and government agencies studying
the impact of COVID-19. The location data from roughly 45 million mobile devices is linked
to a registry of around 6 million points of interest nationwide to record, at a daily frequency,
individual visits to these points of interest.35
33
As noted on the Womply website, businesses that change their method of accepting payment will appear
as closures in Womply data; for example, restaurants that close on-site dining and do business via third-
party delivery services will appear as closed. This may be partly responsible for the particularly high and
persistent closure rate in leisure & hospitality.
34
To enhance privacy, SafeGraph excludes census block group information if fewer than five devices visited
an establishment in a month from a given census block group.
35
While the exact universe of businesses covered by SafeGraph data is difficult to define, with 6 million
points of interest it is likely that a large share of U.S. establishments are covered. The BLS QCEW data for
March 2019 show about 10 million establishments in the U.S., which may be an overstatement of employer
businesses since Census Bureau BDS data for 2018 show about 7 million establishments. While there are
certain industry scope differences between the BLS and Census business registers, a discrepancy exists even
when the two sources are restricted to common scope (see Barnatchez et al. (2017)).
22
In the appendix we describe a methodology that translates patterns of visits to business
locations into indicators of temporary closure and likely permanent establishment exit (these
indicators are best thought of as establishment, not firm, indicators since they are based
on business operating locations). This is a customer and worker traffic-based measure of
business operation, as opposed to the payroll- or revenue-based measures of ADP, Womply,
and Homebase described above. Importantly, traffic-based measures of business shutdown are
more useful in some industries than in others; for example, in construction, traffic patterns
may not be useful as workers may report to various construction sites each day. Similarly,
such measures applied to industries like landscaping services or food trucks would also be
problematic. Generally speaking this methodology is appropriate for industries that rely on
consumer (and worker) visits to businesses at a stationary location, a situation that applies
to many retail and service businesses. It is convenient, though, that industries likely to be
most sensitive to social distancing concerns are also those for which our traffic-based closure
measure may be most appropriate.
We first illustrate these measures for full-service (“sit-down”) restaurants (NAICS 722511),
an industry that is well-suited to this methodology, has good coverage in SafeGraph data,
and is sensitive to social distancing concerns and restrictions; note that this industry ex-
cludes fast food and takeout establishments (NAICS 722513).36 Figure 8 shows estimates
of restaurants that were closed—–at least temporarily—during each month of 2020 and into
early 2021, based on severe drops in consumer visits relative to normal patterns for that es-
tablishment. This fraction was low, around 3-4 percent in the months before the pandemic,
but then jumped to over 50 percent in the months of March and April as social distancing
policies were put in place. The overall contour of the remaining months fits the pattern of
impacts of COVID-19: some declines in temporary closure in the summer months followed
by increases in November, 2020 to January, 2021 before declining once again. The significant
share of closures identified after the initial March-April surge could be comprised of restau-
rants responding to renewed social distancing policies, restaurants that have permanently
closed, or restaurants that are in fact open but operating at extremely low levels of activity.
Estimates of the flow of restaurants that may have permanently closed each month are
shown by the red bars in Figure 9; these are restaurants that closed during the indicated
month and had still not reopened by February, 2021. The monthly estimates for the short
benchmark period of January-February of 2020 (0.1 to 0.2 percent) are below the average
monthly rate of exit for restaurants from BDS data (0.7 percent), so our SafeGraph measure
might understate exit somewhat in other periods.37 The monthly exit rate jumps in March
36
Importantly, our measure is reasonably robust to the notable shift of restaurants to carry-out service,
which is evident in the SafeGraph data based on changes in the duration of consumer visits (see Appendix
Figure B1). Carry-out or delivery service still requires a visit to the business address, which is the basis of
the concept of SafeGraph visits.
37
BDS data for NAICS 7225 indicate recent (2015-2018) annual establishment death rates of 8.1 percent,
implying monthly rates of a bit under 0.7 percent. This suggests that SafeGraph closure rates may be a bit
too low, though it is difficult to tell since seasonality may play a large role in monthly patterns. Perhaps
more importantly, exit rates for the narrow full-service restaurant industry are not available in public data.
The overall rate for NAICS 7225 may understate exit rates for full-service restaurants if, for example, fast
food establishments (in NAICS 722513) have lower exit rates.
23
Percent of Restaurants Temporarily Closed
62%
60 58% 60
40 35% 40
28% 28%
25%
20% 21% 22%
18% 19%
20 17% 20
3.3% 3.6%
0 0
y y ch Ap ay Ju Ju st r er r r y y
ar ar ar ril ne ly gu be ob be be ar ar
nu ru M em em em nu ru
Ja Fe M Au ct Ja Fe
b pt O ov ec b
Se N D
Source: Author calculations from SafeGraph microdata.
Note: Temporary closure defined in Appendix.
Figure 8: Temporary restaurant closures inferred from weekly visits
to around 1.5 percent of restaurants and then increases again in July and August. The
cumulative 12-month exit rate of restaurants, from March 2020 to February 2021, based
on this measure (the black line in Figure 9) is 13.5 percent, about 70 percent above rates
seen in official data in recent years (around 8 percent). The data therefore suggest that
establishment exit has been substantially elevated in the full-service restaurant industry.
4.4.1 Case study of New York City and Oklahoma
A concrete way to evaluate the validity of the signal from this measure is to study variation in
business closure across two geographies with significant differences in the contour of COVID-
19 cases and social distancing policies: New York City and the state of Oklahoma.38 Panel
A of Figure 10 shows the early spike in cases in New York City as well as the much later
jump beginning in July in the state of Oklahoma. Panel B of Figure 10 shows one measure
of social distancing—also derived from SafeGraph data—which identifies the share of mobile
devices remaining close to home in a given day. This figure shows much larger and more
pervasive changes in behavior regarding social distancing practices in New York City than
in Oklahoma.39
The left panel of Figure 11 shows, perhaps unsurprisingly, that the relative magnitudes
38
It is worth noting that identified visits in dense urban locations such as New York City may be subject
to greater measurement error. SafeGraph uses a variety of tools to separate out visits from dense POIs;
however, occasionally a POI will be excluded from data on visits due to being completely enclosed within
another POI.
39
These social distancing practices are, of course, also influenced by official regulations in each geography.
For example, inside dining was not allowed in New York City through the end of September, whereas
restrictions on dining for Oklahoma were minimal.
24
Percent of Restaurants Permanently Closed
13.5% 14
Monthly Rate, by month of identified closing (left)
Cumulative: Mar. 2020 to Feb. 2021 (right)
2.0% 12
2 1.8%
10
1.5 1.4%
8
1 6
4
.5
2
0
ar ry ch ril M ne ly st r er r r ar ry
y ua ar Ap ay Ju gu be ob be be y ua
nu Ju em em em nu
Ja br M Au ct Ja br
Fe pt O ov ec Fe
Se N D
Source: Author calculations from SafeGraph microdata.
Note: Permanent closure defined in Appendix.
Figure 9: Permanent restaurant closures inferred from weekly visits
Positive Cases per Capita Fraction Staying at Home
.3 .3 .6 .6
New York City
.5 .5
New York City
Share of Devices at Home
.2 .2
Oklahoma .4 .4
.1 .1 Oklahoma
.3 .3
.2 .2
0 0 Feb05 Mar26 May15 Jul04 Aug23 Oct12
Feb05 Mar26 May15 Jul04 Aug23 Oct12
Source: Author calcualtions from SafeGraph microdata.
Source: New York Times and Covid Tracking Project Note: Staying at Home is defined as a device remaining within a 150m radius from home.
Figure 10: Cases per capita and social distancing: New York City vs Oklahoma
25
of temporary restaurant closure for New York City and Oklahoma align with the experience
of cases and social distancing practices. Over 80 percent of restaurants appeared to be
temporarily closed in New York City during the months of March and April, whereas this
figure was significantly below the national average in Oklahoma. The right panel of Figure 11
plots the cumulative share of restaurants that are permanently closed. The blue line indicates
that over 30 percent of sit-down restaurants in New York City could have permanently closed
during the 12-month period following the start of the pandemic.40 This cumulative share is
considerably lower in Oklahoma, and indeed appears to be lower than the (national) rate of
closure in a typical year.
Percent Temporarily Closed Cumulative Percent Permanently Closed
100 100 35 35
New York City New York City
Oklahoma Oklahoma
30 30
80 80
25 25
60 60
20 20
40 40 15 15
10 10
20 20
5 5
0 0 0 0
Ja nu
Fe ary
br ua r M ua
ar ry
ry
ch
M y
ar
Ap ch
ril ua Ap
M ayril
M
Ju ay Ju ne
A ly
Se ug June Ja br A ly
Se ugu
pt st Ju
pt ust
em
O ber
N tob
D c
ov e
em r
ec be
em r n Fe
em
O er
N obe
D ov ct
em r
ec ber
em
b
b
Ja ber
nu
Fe ary
br ua ry Ja er
nu
Fe ary
br ua ry
Source: Author calculations from SafeGraph microdata. Source: Author calculations from SafeGraph microdata.
Note: Temporary closure defined in Appendix. Note: Permanent closure defined in Appendix.
Figure 11: Restaurant closures: New York City vs Oklahoma
4.4.2 Industry analysis
Finally, we calculate cumulative (March 2020 to February 2021) permanent closure rates
using this SafeGraph-based measure for a number of other industries and compare the ex-
perience of the first 12 months of the COVID-19 pandemic to historical exit rates. The
industry-level variation in excess business closure largely aligns with expectations, as shown
in Figure 12. The top panel shows these comparisons for select 2-digit NAICS sectors, while
the bottom panel shows select 4-digit industry-group detail. The highest rates of excess clo-
sure are in industries relying heavily on in-person contact (other services, such as personal
care services), experiencing COVID-related reductions in demand (automotive repair), or
close substitutes with e-commerce (clothing stores). Some industries with lower-than-typical
40
While New York City was uniquely impacted by the pandemic, with dramatic declines in the number of
people commuting into the city on a regular basis, there are some reasons to treat these numbers with even
greater caution. Allocating a given GPS signal to a particular establishment is inherently more difficult with
multi-storey buildings as is the case in urban areas such as New York City.
26
closure rates include grocery stores and electronics and appliance stores, two industries that
have seen increased demand during the pandemic.
A few other industry results are worth discussing. The excess closure rate for accommo-
dation and food services (and traveler accommodation in particular) is lower than expected
given the large drops in travel and restaurant dining induced by the pandemic (and the ele-
vated exit we observe among full-service restaurants on Figure 9). This sector saw dramatic
employment declines and is still (as of March 2021) well below its pre-pandemic employ-
ment level. That said, anecdotes suggest that certain areas of this sector have been resilient
to the pandemic; indeed, the aggregated industry group of restaurants and other eating
places (NAICS 7225) saw exit rates only modestly above historical averages, suggesting
that the high rate we document for full-service restaurants was partially offset by low exit
rates among limited-service restaurants like pizza delivery, fast food, and takeout that likely
benefited from shifts toward more at-home eating. More broadly, the Paycheck Protection
Program (PPP) targeted the accommodation and food services sector for special treatment,
allowing virtually the entire sector to qualify rather than just small businesses, potentially
facilitating continued operation despite large declines in revenue.41
The low excess closure rate for “arts, entertainment, and recreation” is also surprising,
though this industry is a mix between establishments that specialize in pandemic-friendly
outdoor recreation (i.e., golf courses, skiing facilities) and those that specialize in riskier
indoor recreation (i.e., museums, fitness centers, and bowling alleys). The retail trade sec-
tor likely saw considerable heterogeneity in exit rates, combining high-exit industries like
clothing stores with low-exit industries like grocery stores and (likely) nonstore retail. We
expect exit rates to have been lower in the broad sectors not reported here (based on BED
and ADP data reviewed above, Census Bureau survey data reviewed below, and anecdotal
discussions), with the possible exceptions of mining (due to oil market developments) and,
perhaps, education. Regardless, we again emphasize that our SafeGraph-based measures
depend on traffic, and industries vary in how accurately traffic tracks business operational
status; therefore, aggregated sector-level data as shown on the top of Figure 12 may not
have a straightforward interpretation.
We report the implications of our SafeGraph-based estimates on Table 3. For each of the
sectors and industries we show on Figure 12, we report the establishment count from the
first quarter of 2019 from BLS QCEW data, the average 2015-2018 establishment exit rate
from the BDS, the SafeGraph-based exit rate for March 2020-February 2021, the implied
excess exit rate, and the implied count of establishment deaths in excess of historical norms.
Our SafeGraph-based estimates suggest excess exit of about 123,000 establishments in other
services, with negligible figures in other sectors. Since we expect other sectors that are less
well measured by SafeGraph to have fared better than the social distance-sensitive sectors
we report here, excess exit for the economy as a whole was likely not far off from the figure
for the worst-off sector of other services. That is, these data suggest that excess exit was
41
While the PPP was generally only available to small firms, for the accommodation and food services
sector the size qualification was based on establishment size. Decker et al. (2021) find that more than 99
percent of establishments in that sector met the size qualification.
27
Other Services (except Public Administration)
Accommodation and Food Services
SafeGraph, Mar 2020-Feb 2021
2015-2018 average
Great Recession
Retail Trade
Arts, Entertainment, and Recreation
0 2 4 6 8 10 12 14 16 18 20 22 24
Percent of firms
Personal Care Services
Automotive Repair and Maintenance
Clothing Stores
Home Furnishings Stores
Furniture Stores
Museums, Historical Sites, and Similar Institutions
SafeGraph, Mar 2020-Feb 2021
Traveler Accommodation 2015-2018 average
Great Recession
Restaurants and Other Eating Places
Building Material and Supplies Dealers
Other Amusement and Recreation Industries
Grocery Stores
Electronics and Appliance Stores
0 2 4 6 8 10 12 14 16 18 20 22 24
Percent of firms
Note: Bars ordered by difference between SafeGraph 2020-2021 rate and BDS 2015-2018 average. Great Recession rate is average of years spanning
April 2008-March 2010.
Source: Author calculations from SafeGraph microdata and Business Dynamics Statistics.
Figure 12: Cumulative permanent closures for selected industries, SafeGraph
28
likely below (say) 200,000 establishments during the first year of the pandemic—broadly
consistent with our BED-based estimate described above, which suggested excess deaths
below 150,000 during the first half of 2020. Of course, this is a guess in which we assume
that sectors other than other services saw relatively low excess exit; if our guess is reasonable,
it would imply that overall exit has been elevated by about one-third, as the average number
of establishment exits during 2015-2018 was about 600,000 in BDS data.42
Table 3: Excess exit estimates for selected industries, SafeGraph
Initial
establishments Historical Pandemic Excess Excess exits
Industry (NAICS) (QCEW, 1000s) rate (BDS) rate rate (1000s)
Broad sectors
Retail trade 1,048 7.6 8.6 1.0 11
Arts, Entertainment, & recreation 165 10.1 7.6 -2.5 -4
Accommodation & food services 737 8.5 9.7 1.2 9
Other services 875 6.0 20.1 14.1 123
Four-digit industries
Furniture stores 22 7.2 10.2 3.0 1
Home furnishings stores 25 7.2 11.4 4.3 1
Electronics & appliance stores 44 14.1 12.1 -2.0 -1
Building material & supplies dealers 53 4.7 5.3 0.6 0
Grocery stores 90 8.4 7.2 -1.2 -1
Clothing stores 82 8.6 16.0 7.3 6
Museums, historical sites, etc. 9 3.8 6.1 2.4 0
Other amusement & recreation 85 9.3 8.3 -1.0 -1
Traveler accommodation 64 7.5 8.8 1.3 1
Restaurants & other eating places 578 8.1 9.3 1.2 7
Automotive repair & maintenance 165 7.8 16.7 8.9 15
Personal care services 135 10.4 22.7 12.3 17
Note: Industries listed in NAICS order. Initial establishment counts from QCEW, 2020q1. Historical rate
is average establishment exit rate for 2015-2018 from BDS. Pandemic rate is SafeGraph-based exit rate
estimate for the 12 months of March 2020-February 2021. Excess rate is difference between pandemic and
historical rates. Excess deaths are equal to excess rate multiplied by establishment count.
In summary, the measure inferred from mobile device location data—which is not subject
to the customer attrition concerns of our previously described data sources but does face
42
Our mixing of BLS and Census Bureau data on Table 3 is not without consequence and is its own source
of uncertainty around our estimates. As noted elsewhere, establishment counts in BLS and Census Bureau
sources have diverged in recent years, with higher levels in BLS sources. This is particularly driven by
divergence among the smallest establishments—which tend to have the highest exit rates. In 2018, BLS data
reported about 35 percent more private sector establishments than Census Bureau data. Our estimate of
fewer than 200,000 excess establishments is based on the BLS establishment count, while our suggestion that
this number is about one-third above normal is based on Census Bureau annual death data (whereas the BLS
does not, to our knowledge, publish annual establishment death figures). If we scale down the establishment
counts from Table 3 using the 1.35 ratio just mentioned, we might suggest an upper bound estimate of
150,000 excess establishment exits, which would imply overall exit was elevated by about one-fourth.
29
other measurement limitations—appears to show elevated exit rates for restaurants and other
industries that align with the effects from COVID-19. However, this unconventional measure
of business exit should be treated with caution as there remain several unique features of
our measurement that may not align with true business exit. Perhaps most importantly,
we do not know how well our SafeGraph measure would have performed in past years—the
January and February 2020 data might suggest some understatement—so the comparisons
on which we based our excess death estimates should be treated with caution.
4.5 Other private sector business registers
As noted above, an advantage of SafeGraph over the other non-traditional measures we
study is that it is designed to capture the near universe of business establishments, and
there are no customer attrition concerns. Certain other private sector business registers
share these qualities; however, we see them as significantly limited for tracking business exit.
In particular, business registers that do not have near-real-time sources of information on
actual business activity—such as revenue transactions, payroll events, or customer traffic—
for all businesses in the register are likely to poorly capture exit at high frequency, since
business shutdown can typically be identified only with analyst effort.
One such business register is maintained by Yelp, the online platform for rating businesses.
The company occasionally provides shutdown data including permanent shutdown estimates,
but we find these difficult to interpret in their current form (see Yelp: Local Economic
Impact Report (2020)). For example, Yelp reports the highest permanent closure rates among
restaurants selling burgers and sandwiches, where estimated permanent closure rates from
March through August were around 3.5 percent. On an annualized basis, this would imply
exit rates around 7 percent, still below recent restaurant exit rates of around 8 percent.
Other industries reported by Yelp show exit rates that are even lower relative to historical
patterns. These results are not necessarily implausible, but still we find them difficult to
interpret. In particular, we assume that Yelp underestimates exit because the company
cannot directly verify business activity in real time but instead can only observe exit if website
users contribute information or if analysts engage in significant effort. That said, Yelp data
may be useful for studying differences across industries. For example, Yelp indicates the
lowest exit rates among lawyers, architects, and accountants, consistent with ADP data on
the relative success of professional and business services industries early in the pandemic.
The most prominent private sector business register is that of Dun & Bradstreet (D&B)—
or its derivative, the National Establishment Time Series (NETS) of Walls & Associates.
D&B uses many sources to maintain their business list, including yellow pages and surveys
by analysts. Such business lists have many uses, but we expect them to be poorly suited
for measuring recent exits. D&B-like business registers do not automatically update when
business activity drops, in contrast to measures based on automated observation of business
status. Instead, analysts must somehow identify business shutdown, an enormous task par-
ticularly for the millions of small businesses with high exit rates (and one for which even
Yelp is likely better suited, given its access to effectively crowdsourced information). Indeed,
Crane and Decker (2020) find that D&B/NETS data capture business dynamics poorly.
30
5 Business death expectations
Many currently operating businesses are concerned about exit risk going forward, though
these expectations have fluctuated throughout the pandemic. The Census Bureau has con-
ducted surveys of small businesses (those with one establishment and fewer than 500 em-
ployees) at weekly frequency through several phases since the pandemic began.43 Among
other questions, businesses are asked about their expectations for the six months following
the survey; the left panel of Figure 13 reports the share of businesses expecting permanently
to close within six months of the survey week; the red dashed line shows the historical actual
exit rate for firms with fewer than 500 employees from BDS data for 2015-2018.44 The right
panel reports the share of businesses expecting to need to obtain financial assistance or credit
over the same period.
Close permanently Obtain finance
6 6 35 35
5 5 30 30
Percent of firms Percent of firms
4 4 25 25
Historical rate
3 3 20 20
2 2 15 15
st be r ber br Ap st be r ber br Ap
Au ct m ua ril Au ct m ua ril
gu o ec ry gu o ec ry
O e Fe O e Fe
D D
Note: Six month expectations. Data correspond to end of survey week. Gaps indicate between-survey periods. Historical rate is 2015-2018 actual small business death rate.
Source: Census Bureau Business Dynamics Statistics and Small Business Pulse Survey; data through March 29 - April 4 2021.
Figure 13: Small firms’ expected needs over next six months
Death expectations rose in November and December of 2020 corresponding with rising
Covid case counts in the U.S.; this pessimism declined some through December and January
then dropped substantially by mid-February, possibly related to the late-December passage
of an additional round of Paycheck Protection Program loans and other fiscal support.45
During most of the pandemic death expectations have been well above the level consistent
43
See https://www.census.gov/data/experimental-data-products/small-business-pulse-survey.html for sur-
vey data, including time series views; for methodological details see Buffington et al. (2020).
44
We estimate the historical six-month exit rate by taking half of the historical annual exit rate. This is
only approximate—actual six-month exit rates can be higher due to firms that enter and exit within a year.
45
The Pulse survey also asks questions about past closures; these are more difficult to interpret given
the possibility that non-response to the survey is correlated with business death. Even the forward-looking
measure we report may suffer from selection bias among respondents, and it is possible that industries
showing relatively low expectations of exit have already seen many exits. That said, the Pulse survey
is constructed using the Census Bureau’s Business Register as a sampling frame, and the Census Bureau
applies appropriate sampling weights to responses, so the survey is of high scientific quality. Importantly,
the sample is limited to small employer businesses with only one establishment.
31
with actual death rates from recent years (the dashed red line), though we do not know how
well these death expectations would predict actual death in normal times.
Importantly, the expected exit rates vary widely by sector, and expected exit rates relative
to historical averages vary widely as well. Figure 14 shows Pulse Survey expected exit as
of early April 2021, in blue, compared with average six-month actual death rates among
small firms (fewer than 500 employees) from BDS data from 2015-2018, in red, and from the
Great Recession, in black.46 In some sectors, recent expected death rates do indeed imply
historically elevated rates (i.e., the blue bars are longer than the red and black bars). This is
most notable in mining, other services, and education.47 Some of these sectors have already
seen considerable exit such that these expectations, if realized, would result in material
permanent impacts; for example, consider the possibility that the other services sector has
already seen exit rates at more than three times the normal rate over the past year (as
suggested by our SafeGraph analysis) and then experiences exit rates at nearly double the
usual pace over the next six months.
On the other hand, some sectors have recent exit expectations close to, or even below,
historical averages. The most notable of these are accommodation and food services and
arts, entertainment, and recreation—that is, the leisure and hospitality supersector. This is
a notable development.
Some of the sectors with low exit expectations in recent surveys had much higher expec-
tations earlier in the pandemic. Figure 15 reports excess exit expectations—that is, expected
exit rates minus historical actual exit rates among firms with fewer than 500 employees—by
sector (top panel) and by state (bottom panel). The green bars show expectations as of
late November, when overall exit expectations were highest; the orange bars show expecta-
tions in the latest data from early April 2021. Sectors and states are listed in order of their
improvement; the most improved sector is accommodation and food services, where excess
death expectations once exceeded historical death rates by about six percentage points but
are now below historical rates. Arts, entertainment, and recreation has also improved sub-
stantially. Notably, several sectors did not see dramatically elevated death expectations even
during the November peak. Data on states also show wide heterogeneity in improvement—
and in peak exit expectations.48
46
BDS exit rates are divided by two to approximate 6-month exit rates comparable to Pulse expectations.
Sector-level historical actual exit rates are approximate; in particular, the lagged firm count in the DHS
exit rate denominator is constructed with the previous-year total firm count (f irmst−1 ) rather than with
longitudinally precise lagged firm count (f irmst − entrantst + deathst ) because publicly available BDS data
are not tabulated at the sector-by-age-by-size level, and age is necessary for constructing the longitudinal
measure. The resulting error is likely negligible.
47
The education sector includes not only K-12 schools and colleges/universities but also trade training
programs such as cosmetology schools, flight training, and technical programs as well as sports and recreation
training, exam preparation programs, and driving schools.
48
The Census Bureau suppresses data for some states; all states with reported data in both time periods
are shown on Figure 14. State historical actual exit rates are approximate (see footnote 46).
32
Mining
Other Services
Education
Information
Wholesale Trade
Manufacturing
Accomm. & Food
Transp. & Warehous.
Retail Trade
Admin. & Waste
Finance & Insurance
Health Care
Arts, Ent., & Rec
Construction Expected rate
Prof., Sci., & Tech. 2015-2018 average
Real Estate Great Recession
0 2 4 6 8
Percent of firms
Note: Bars ordered by difference between expected rate and 2015-2018 average. Firms with fewer than 500 employees. Great Recession
rate is average of years spanning April 2008-March 2010.
Source: Census Bureau Small Business Pulse Survey, Mar. 29-Apr. 4 2021 and Business Dynamics Statistics.
Figure 14: Exit expectations versus historical rates
6 Taking stock
Our analysis suggests that business exit likely was elevated during the first year of the pan-
demic. We find direct evidence of increased establishment deaths among full-service restau-
rants, personal care services, automotive repair, and certain retailers (apparent in SafeGraph
data) and suggestive direct evidence for elevated deaths among small firms in related in-
dustries generally (apparent in Womply and Homebase data). But within sectors, variation
across industries appears to be partially or fully offsetting, such that most sectors likely have
not seen dramatically elevated exit; the primary exception is other services, where we esti-
mate that establishment exits exceed prior years’ pace by about 120,000. From this figure,
and considering our relatively modest estimates for other sectors, we judgmentally estimate
that excess establishment exits were below 200,000 during the first year of the pandemic.
Separately, our ADP data suggest that any rise in business exits has not reached larger
business units or even enough smaller units to account for a material share of employment,
which would be consistent with past patterns of employment-weighted exit cyclicality. This
view appears to be roughly corroborated in BLS data studied by Dalton et al. (2020b) (see
section 4.1 above).
Excess establishment exits below 200,000 during the first year of the pandemic—and ex-
cess firm exits below 150,000, if historical shares continued—with little associated excess job
33
Sectors
Accomm. & Food
Arts, Ent., & Rec
Education
Transp. & Warehous.
Real Estate
Other Services
Construction
Mining
Admin. & Waste
Information
Prof., Sci., & Tech.
Health Care
Wholesale Trade
Retail Trade
Manufacturing November 23-29, 2020
Finance & Insurance March 29 - April 4, 2021
-2 0 2 4 6
Expected excess death rate (percentage points)
Source: Census Bureau Small Business Pulse Survey and Business Dynamics Statistics (BDS).
Expected excess death rate is expected rate (over six months) minus half the sector average firm exit rate among firms
with <500 employees, 2015-2018. Sectors ordered by change in excess death rate.
Reported states
OR
NC
PA
LA
NJ
OK
IL
CA
NY
CO
MD
MI
WA
MN
TX
OH
VA
GA
FL
-2 -1 0 1 2 3 4
Expected excess death rate (percentage points)
Source: Census Bureau Small Business Pulse Survey and Business Dynamics Statistics (BDS).
Expected excess death rate is expected rate (over six months) minus half the state average firm exit rate among firms
with <500 employees, 2015-2018. States ordered by change in excess death rate.
Figure 15: Excess exit expectations, peak versus early March 2021
34
destruction would likely be a positive outcome relative to widespread expectations from early
in the pandemic.49 Some of the detrimental consequences of elevated exit—permanent job
dislocation, potential productivity impacts if selection works adversely, and the destruction
of intangible and physical capital—may be modest in aggregate if exit does not reach large
firms or a greater number of small firms. Other detrimental consequences—loss of job and
wealth for business owners and abrupt changes to local economic geography—have welfare
implications even if exit is not substantial on an activity-weighted basis.
Given various dataset-specific caveats mentioned in the main text, we view our evidence
as suggestive, not conclusive. Whether the patterns we document will continue is difficult
to tell; for example, a continued economic recovery could reduce exit rates, and if some
deaths apparent in existing data were “pulled forward” from later this year then annualized
exit rates could even be subdued in coming months. Exit expectations remain elevated in
a handful of sectors and states but have generally improved dramatically since November
2020 and are significantly below normal in several sectors. Early and ongoing policy actions
may have helped businesses in the relatively optimistic sectors to survive the worst of the
pandemic and sustainably reopen.50
Alternatively, it may be that our imperfect indicators understate the magnitude of the
surge in exit. Anecdotes of business failure certainly abound. Moreover, the situation could
deteriorate further from here. Considerable uncertainty about the virus and its various
strains remains. Business owners who have managed to stay open this long may face overdue
financial obligations that will prove unmanageable, and liquidity contraction or distortions
resulting from 2020 policies could leave many firms at high risk (Gournichas et al., 2021).
Cross-industry demand reallocation associated with a transition out of the pandemic could
be significant. Large firms may have been taking time to implement restructuring plans that
could result in waves of establishment exits or the failure of smaller suppliers. Ultimately, we
will not have certainty about business exit during 2020-2021 until the pandemic is behind
us and high-quality official data are published. In this respect, our work highlights the
importance of timely data production and reminds us of the value of official data producers.
49
Separately, early evidence suggests business entry may have surged during the first year of the pandemic
(Dinlersoz et al., 2021). Combined with our evidence of lower-than-expected exit, it may be that the effect
of the pandemic on the net change in establishment or firm numbers will turn out to be modest.
50
Autor et al. (2020) estimate that the Paycheck Protection Program (PPP) had a material effect, boosting
June payroll employment by about 2.3 million. Given the particular vulnerability of the smallest businesses,
even a small number of jobs saved could correspond with a large number of businesses; moreover, the Autor
et al. (2020) identification strategies may understate effects on the smallest, most exit-prone businesses. We
also note that the dramatic improvement in exit expectations shown in the Pulse survey coincided with the
late-December refunding of the PPP program alongside other fiscal measures.
35
References
ADP, “ADP Reports First Quarter Fiscal 2021 Results,” Earnings release, October 2020.
, “ADP Reports Fourth Quarter Fiscal 2020 Results; Provides Fiscal 2021 Outlook,”
Earnings release, July 2020.
Autor, David, David Cho, Leland D. Crane, Mita Goldar, Byron Lutz,
Joshua Montes, William B. Peterman, David Ratner, Daniel Villar, and Ahu
Yildirmaz, “An Evaluation of the Paycheck Protection Program Using Administrative
Payroll Microdata,” Working paper 2020.
Barnatchez, Keith, Leland D. Crane, and Ryan A. Decker, “An Assessment of the
National Establishment Time Series (NETS) Database,” Finance and Economics Discus-
sion Series 2017-110, Federal Reserve Board 2017.
Bartelsman, Eric, John Haltiwanger, and Stefano Scarpetta, “Cross-Country Differ-
ences in Productivity: The Role of Allocation and Selection,” American Economic Review,
2013, 103 (1), 305–334.
Bartik, Alexander W., Marianne Bertrand, Feng Lin, Jesse Rothstein, and Matt
Unrath, “Measuring the Labor Market at the Onset of the COVID-19 Crisis,” BPEA
Conference Draft, Summer, 2020.
Bartik, Alexander W, Marianne Bertrand, Zoë B Cullen, Edward L Glaeser,
Michael Luca, and Christopher T Stanton, “The Impact of COVID-19 on Small
Business Outcomes and Expectations,” Proceedings of the National Academy of Sciences,
2020, 117 (30), 17656–17666.
Bartlett, Robert P. and Adair Morse, “Small Business Survival Capabilities and Policy
Effectiveness: Evidence from Oakland,” NBER Working Papers 27629, National Bureau
of Economic Research 2020.
Buffington, Catherine, Carrie Dennis, Emin Dinlersoz, Lucia Foster, and Shawn
Klimek, “Measuring the Effect of COVID-19 on U.S. Small Businesses: The Small Busi-
ness Pulse Survey,” CES Working Paper CES-20-16, U.S. Census Bureau: Center for
Economic Studies 2020.
Bureau of Labor Statistics, “Annual Business Employment Dynamics Data,” website
2020.
, “National Business Employment Dynamics Data by Firm Size Class,” website 2020.
Cajner, Tomaz, Leland Crane, Ryan A. Decker, Adrian Hamins-Puertolas,
Christopher Kurz, and Tyler Radler, “Using Payroll Processor Microdata to Measure
Aggregate Labor Market Activity,” FEDS Working Paper 2018-005, Board of Governors
of the Federal Reserve System (U.S.) January 2018.
36
, Leland D. Crane, Ryan A. Decker, John Grigsby, Adrian Hamins-Puertolas,
Erik Hurst, Christopher Kurz, and Ahu Yildirmaz, “The U.S. Labor Market during
the Beginning of the Pandemic Recession,” BPEA Conference Draft, Summer, 2020.
Chetty, Raj, John N. Friedman, Nathaniel Hendren, Michael Stepner, and the
Opportunity Insights Team, “How Did COVID-19 and Stabilization Policies Affect
Spending and Employment? A New Real Time Economic Tracker Based on Private Sector
Data,” Working Paper June 2020.
Clementi, Gian Luca and Berardino Palazzo, “Entry, Exit, Firm Dynamics, and Ag-
gregate Fluctuations,” American Economic Journal: Macroeconomics, 2016, 8 (3), 1–41.
Cooper, Russell W. and John C. Haltiwanger, “On the Nature of Capital Adjustment
Costs,” The Review of Economic Studies, 2007, 73 (3), 611–633.
Costar Group, Inc., “Costar Suite (US),” http://www.costar.com September 2020.
Crane, Leland and Ryan Decker, “Research with Private Sector Business Microdata:
The Case of NETS/D&B,” Working paper 2020.
Dalton, Michael, Elizabeth Weber Handwerker, and Mark A. Loewenstein, “Em-
ployment Changes by Employer Size Using the Current Employment Statistics Survey
Microdata during the COVID-19 Panedmic,” Covid Economics, September 2020, 1 (46),
123–136.
, , and , “An Update on Employment Changes by Employer Size during the COVID-
19 Pandemic: A Look at the Current Employment Statistics Microdata,” BLS Working
Papers, December 2020, (532).
Davis, Steven J, John C Haltiwanger, and Scott Schuh, Job Creation and Destruc-
tion, The MIT Press, 1996.
Davis, Steven J., John Haltiwanger, Kyle Handley, Ron Jarmin, Josh Lerner,
and Javier Miranda, “Private Equity, Jobs, and Productivity,” American Economic
Association, 2014, 104 (12), 2956–3990.
Decker, Ryan A., John Haltiwanger, Ron S. Jarmin, and Jarmin Miranda, “De-
clining Dynamism, Allocative Efficiency, and the Productivity Slowdown,” American Eco-
nomic Review: Papers & Proceedings, 2017, 107 (5), 322–326.
, , , and Javier Miranda, “Changing Business Dynamism and Productivity: Shocks
versus Responsiveness,” American Economic Review, 2020, 110 (12), 835–850.
, Pablo N. D’erasmo, and Hernan Moscoso Boedo, “Market Exposure and Endoge-
nous Firm Volatility over the Business Cycle,” American Economic Journal: Macroeco-
nomics, 2016, 8 (1), 148–198.
37
Decker, Ryan, Robert Kurtzman, Byron Lutz, and Christopher Nekarda, “Across
the Universe: Policy Support for Employment and Revenue in the Pandemic Recession,”
American Economic Association: Papers and Proceedings, May 2021.
Dinlersoz, Emin, Timothy Dunne, John Haltiwanger, and Veronika Penciakova,
“Business Formation: A Tale of Two Recessions,” American Economic Association: Pa-
pers and Proceedings, May 2021.
Fairlie, Robert, “The Impact of COVID-19 on Small Business Owners: The First Three
Months after Social-Distancing Restrictions,” Journal of Economics & Management Strat-
egy, Forthcoming.
Fort, Teresa C., Justin R. Pierce, and Peter K. Schott, “New Perspectives on the
Decline in US Manufacturing Employment,” Journal of Economic Perspectives, 2018, 32
(2), 47–72.
Foster, Lucia, Cheryl Grim, and John Haltiwanger, “Reallocation in the Great Re-
cession: Cleansing or Not?,” Journal of Labor Economics, 2016, 34 (S1), S293–S331.
, John C. Haltiwanger, and C. J. Krizan, “Aggregate Productivity Growth: Lessons
from Microeconomic Evidence,” in “New Developments in Productivity Analysis,” Uni-
versity of Chicago Press, January 2001, pp. 303–372.
, John Haltiwanger, and C. J. Krizan, “Market Selection, Reallocation, and Re-
structuring in the U.S. Retail Trade Sector in the 1990s,” The Review of Economics and
Statistics, 2006, 88 (4), 748–758.
Fujita, Shigeru and Giuseppe Moscarini, “Recall and Unemployment,” American Eco-
nomic Review, December 2017, 107 (12), 3875–3916.
Gournichas, Pierre Olivier, Sebnem Kalemli-Ozcan, Veronika Penciakova, and
Nick Sander, “COVID-19 and SMEs: A 2021 “Time Bomb”?,” American Economic
Association: Papers and Proceedings, May 2021.
Greenwood, Robin, Ben Iverson, and David Thesmar, “Sizing Up Corporate Re-
structuring in the COVID Crisis,” BPEA Conference Draft, Fall, 2020.
Gschwandtner, Adelina and Val E. Lambson, “Sunk Costs, Depreciation, and Industry
Dynamics,” Review of Economics and Statistics, 2012, 94 (4), 1059–1065.
Hamilton, James, “Why You Shoud Never Use the Hodrick-Prescott Filter,” Review of
Economics and Statistics, December 2018, 100, 831–843.
Hamilton, Steven, “From Survival to Revival: How to Help Small Businesses through the
COVID-19 Crisis,” The Hamilton Project Policy Proposal, 2020, (2020-14).
38
Hodrick, Robert J. and Edward C. Prescott, “Postwar U.S. Business Cycles: An
Empirical Investigation,” Journal of Money, Credit, and Banking, 1997, 29 (1), 1–16.
Hopenhayn, Hugo A., “Exit, Selection, and the Value of Firms,” Journal of Economic
Dynamics and Control, 1992, 16 (3-4), 621–653.
and Richard Rogerson, “Job Turnover and Policy Evaluation: A General Equilibrium
Analysis,” Journal of Political Economy, 1993, 101 (5), 915–938.
Kurmann, Andre, Etienne Lale, and Lien Ta, “The Impact of COVID-19 on U.S.
Employment and Hours: Real-Time Estimates with Homebase Data,” mimeo 2020.
Lee, Yoonsoo and Toshihiko Mukoyama, “Entry and Exit of Manufacturing Plants
over the Business Cycle,” European Economic Review, 2015, 77, 20–27.
Paynet, Inc., “Small Business Credit Conditions Report,” http://www.paynetonline.
com/ September 2020. Accessed: October 7, 2020.
Reinicke, Carmen, “How a Small-Business Software Company Became One of the Most
Accurate Forecasters of the Pandemic Recession,” Business Insider, August 23 2020.
Sadeghi, Akbar, “The Births and Deaths of Business Establishments in the United States,”
Monthly Labor Review, December 2008, pp. 3–18.
Simon, Ruth and Heather Haddon, “Small-Business Owners Feel Weight of Personal
Debt Guarantees,” The Wall Street Journal, Apr 2021.
Syverson, Chad, “What Determines Productivity?,” Journal of Economic Literature, 2011,
49 (2), 326–365.
Wang, Jialan, Jeyul Yang, Benjamin Iverson, and Raymond Kluender,
“Bankruptcy and the COVID-19 Crisis,” Working Paper, University of Illinois at Urbana-
Champaign 2020.
Yelp: Local Economic Impact Report
Yelp: Local Economic Impact Report, https: // www. yelpeconomicaverage. com/
business-closures-update-sep-2020 September 2020. Accessed: October 5, 2020.
39
A Appendix for historical results
A.1 Alternative detrending specifications
In our main results for nationwide business cycle correlations, we detrend BED and BDS
shutdown measures using simple linear trends. Table A1 repeats the results of Table 1 using
the detrending methodology of Hamilton (2018) instead; since the Hamilton (2018) method
loses the first three years of data from the detrended series, we also report correlations based
on linear trends covering the same periods. The results are largely unaffected by detrending
methodology, though we observe some differences for firm-based measures.
Table A1: Business cycle correlations: Hamilton (2018) detrending
Unemployment GDP
Unweighted Weighted Unweighted Weighted
Hamilton (2018) method:
BED: Establishment closures (BLS) .39 .18 -.35 -.14
BED: Establishment deaths (BLS) .36 .22 -.33 -.24
BDS: Establishment deaths (Census) .41 .13 -.19 -.02
BDS: Firm deaths (Census) .16 .09 -.05 -.05
Linear method:
BED: Establishment closures (BLS) .37 .20 -.41 -.11
BED: Establishment deaths (BLS) .37 .27 -.46 -.19
BDS: Establishment deaths (Census) .64 .37 -.46 -.16
BDS: Firm deaths (Census) .41 .31 -.32 -.18
Note: BED correlated with quarterly change in unemployment rate or change in log real
GDP. BDS correlated with annual change in unemployment rate or change in log real
GDP on BDS annual timing (April-March). BED trends estimated on data for 1992q3-
2019q1 (deaths) or 1992q3-2019q4 (closures); BED correlations omit first three quarters
of data consistent with Hamilton (2018) methodology (and therefore do not match Table
1 exactly). BDS trends estimated on data for 1978-2018; BDS correlations based on data
for 1984-2018.
Separately, our main specifications use the annual or quarterly change in the log of real
GDP for GDP-based correlations. An alternative approach to specifying GDP is to extract
the cyclical component using a detrending method. Table A2 reports our main correlations
in which GDP is expressed as the cyclical component of real GDP from the Hamilton (2018)
method (left panel) and the HP filter of Hodrick and Prescott (1997) parameterized as is
common in the literature (right panel); exit rates are detrended linearly as in the main text.
The finding that establishment exit is more countercyclical than firm exit and unweighted
exit rates are more countercyclical than employment-weighted exit rates still hold under the
Hamilton (2018) method for GDP detrending. The HP filter results are more puzzling;
Hamilton (2018) recommends against using the HP filter, and we show it here only for
reader convenience.
40
Table A2: Business cycle correlations: Alternative GDP specifications
Hamilton (2018) Hodrick and Prescott (1997)
Unweighted Weighted Unweighted Weighted
BED: Establishment closures (BED) -.57 -.04 -.15 -.14
BED: Establishment deaths (BED) -.44 .04 -.07 -.08
BDS: Establishment deaths (BDS) -.29 .02 -.27 -.22
BDS: Firm deaths (BDS) -.17 .00 -.08 -.22
Note: Exit rates detrended linearly. BED correlated with quarterly change in unemployment
rate or growth of GDP. BDS correlated with annual change in unemployment rate or detrended
real GDP on BDS annual timing (April-March). HP filter parameter set at 6.25 for annual data
(BDS correlations) and 1600 for quarterly data (BED correlations. BED data cover 1992q3-
2019q1 (deaths) or 1992q3-2019q4 (closures). BDS data cover 1984-2018 (trends estimated on
1978-2018 data).
A.2 Exit levels
In the main text we focus on rates of establishment and firm exit. The behavior of estab-
lishment firm exit levels (in terms of number of firms or establishments and their associated
employment) behave in a broadly similar manner over the business cycle, though some in-
teresting different trends are apparent. Figure A1 replicates Figure 2 using levels instead of
rates.
The cyclicality of exit levels, particularly in recent years, is apparent in Figure A1.
Interestingly, the number of exiting firms and establishments rose during the 1990s and
2000s (as the number of firms and establishments in the U.S. grew), but the number of exits
declined after the Great Recession.
Table A3 reports business cycle correlations of exit level measures; that is, we correlate
the (detrended) number of establishment closures and deaths and firm deaths with the change
in the unemployment rate and the change in log real GDP. We report exit level measures
detrended linearly (top) and using the Hamilton (2018) method (bottom).
The countercyclicality of establishment exit is apparent in levels, as is the weaker coun-
tercyclicality of employment-weighted quantities relative to unweighted quantities (with one
exception of firm deaths under the Hamilton (2018) method). The countercyclicality of firm
death is, as in other results, weaker than establishment countercyclicality.
A.3 Exit rate denominators
In the main text, we calculate exit rates as the number of establishments (or firms) divided by
the average aggregate number of establishments (or firms) in the two periods (the so-called
“DHS denominator” after Davis et al. (1996)). Likewise, we calculate employment-weighted
exit rates as the jobs destroyed by exiting establishments (or firms) divided by average
aggregate employment in the two periods. This convention follows the business dynamics
literature generally and is consistent with how the BLS and the Census Bureau provide pre-
calculated exit rates. But the denominators in these calculations are endogenous to exit and
may be unintuitive to some readers, and our non-traditional measures use only initial values
41
BDS - Establishments and firms BDS - Employment
7500 7500
850 Establishment deaths 850
Firm deaths
6500 6500
750 750
Thousands (annual) Thousands (annual)
5500 5500
650 650
4500 4500
550 550
450 450 3500 3500
350 350 2500 2500
250 250 1500 1500
82 86 90 94 98 02 06 10 14 18 8
19 2
8
19 6
9
19 0
9
19 4
9
20 8
19 19 19 19 19 20 20 20 20 20 19 0
20 2
0
20 6
1
20 0
1
20 4
18
BED - Establishments and firms BED - Employment
2000 2000
Establishment deaths
500 Establishment closures 500
1750 1750
Thousands (quarterly) Thousands (quarterly)
400 400 1500 1500
1250 1250
300 300
1000 1000
200 200
750 750
100 100 500 500
92 96 00 04 08 12 16 20 92 96 00 04 08 12 16 20
19 19 20 20 20 20 20 20 19 19 20 20 20 20 20 20
Note: BDS data are noisy around Economic Census years (2's and 7's). Establishments are single operating business locations.
Firms are collections of one or more establishments under common ownership or operational control.
Source: BLS Business Employment Dynamics (BED), Census Bureau Business Dynamics Statistics (BDS).
Figure A1: Historical patterns of business shutdown levels
42
Table A3: Business cycle correlations: Exit levels
Unemployment GDP
Unweighted Weighted Unweighted Weighted
Linear method:
BED: Establishment closures (BLS) .35 .21 -.40 -.10
BED: Establishment deaths (BLS) .36 .23 -.45 -.13
BDS: Establishment deaths (Census) .57 .34 -.29 -.05
BDS: Firm deaths (Census) .39 .32 -.16 -.10
Hamilton (2018) method:
BED: Establishment closures (BLS) .29 .10 -.35 -.09
BED: Establishment deaths (BLS) .29 .10 -.49 -.09
BDS: Establishment deaths (Census) .32 .10 -.26 -.06
BDS: Firm deaths (Census) .08 .13 -.13 -.12
Note: BED correlated with quarterly change in unemployment rate or change in log real
GDP. BDS correlated with annual change in unemployment rate or change in log real GDP
on BDS annual timing (April-March). BED trends estimated on data for 1992q3-2019q1
(deaths) or 1992q3-2019q4 (closures). BDS trends estimated on data for 1978-2018 but
correlations estimated on data for 1984-2018.
in denominators, so we examine the robustness of our main results to using the lagged (i.e.,
initial) establishment (or firm) count as the denominator for unweighted exit rates and using
lagged aggregate employment as the denominator for weighted exit rates. Importantly, our
main results about recent average exit rates (described in our stylized facts bullets in the
main text) are not affected, to rounding, by this alternative denominator.
We study the sensitivity to denominator choice on Table A4 where we repeat our main
correlations from Table 1 instead using the lag denominator. For simplicity, we focus on
BDS data only, and we repeat the main results for ease of comparison (shown in the lines
with the “(DHS)” parenthetical).
Table A4: Business cycle correlations: Different denominator
Unemployment GDP
Unweighted Weighted Unweighted Weighted
Establishment deaths (DHS) .64 .37 -.46 -.16
Establishment deaths (lag denominator) .60 .31 -.40 -.07
Firm deaths (DHS) .41 .31 -.32 -.18
Firm deaths (lag denominator) .35 .25 -.25 -.12
Note: Exit rates detrended linearly. BDS exit rates correlated with annual change in unem-
ployment rate or change in log real GDP on BDS annual timing (April-March). BDS trends
estimated on data for 1978-2018 but correlations estimated on data for 1984-2018. “DHS”
refers to the Davis et al. (1996) two-period averaged denominator.
Business cycle correlations for unweighted exit rates are somewhat weaker—but still
apparent—under the lag denominator, though employment-weighted exit rates are clearly
less cyclical. We repeat our state-level regressions (from Table 1) with the lag denominator
on Table A5; these are similar to the results with DHS denominators.
43
Table A5: Business cycle comovement: States (different de-
nominator)
Establishment death Firm death
Unweighted Weighted Unweighted Weighted
Change in unemp 0.52*** 0.19*** 0.33*** 0.10***
(0.03) (0.02) (0.02) (0.01)
Observations 816 816 816 816
State FE Yes Yes Yes Yes
Year FE No No No No
Note: Regression of annual exit rates on annual change in unemploy-
ment rates, 2003-2018. Exit rates use lag denominator. Unemploy-
ment rate changes timed to correspond with BDS annual timing (April-
March). Compare to Table 2.
***denotes statistical significance with p < 0.01.
Source: Author calculations from Business Dynamics Statistics and
BLS unemployment data.
A.4 Using different sample periods
In the main text we explore official measures of business shutdown from two sources, BED
(from the BLS) and BDS (from the Census Bureau). While BDS data have key advantages
in terms of time coverage and firm identification, BDS data do face some limitations around
years in which the Economic Census is conducted (those ending in 2 or 7) arising from
difficulties with longitudinal linkages; these appear most significant in 2002, which was also a
year in which the underlying Business Register (formerly Standard Statistical Establishment
List) source data were reorganized. In this section we briefly explore these limitations of the
BDS.
We recreate the business cycle correlations in Table 1 omitting the year 2002 and, sep-
arately, omitting all Economic Census years; these can be found on Table A6. For ease of
comparison, we repeat the correlations from 1 (calculated on all years in the data) and also
show correlations in which specified years are omitted. Excluding 2002 or all Census years
has little effect on the business cycle correlations of unweighted exit rates, but weighted
rates are sensitive to these exclusions. Importantly, omitting these years is not necessarily
the best practice, as they are likely to reflect real data in addition to potential noise.
Table A7 repeats the state-level panel regressions shown on Table 2 with the omission of
Economic Census years.
Economic Census years do not substantially alter the result that exit rates are counter-
cyclical, correlating positively with the change in unemployment rates and negatively with
GDP growth, and that this cyclicality is driven by smaller units.
Our main results for BDS data use the years 1984-2018. BDS data are available for 1978-
2018, but we omit the early period because correlations are heavily sensitive to the omission
or inclusion of certain specific years during 1978-1983 (and particularly 1980-1983). This
may be because the early 1980s recession years did not see elevated business exit in BDS
44
Table A6: Business cycle correlations: Census year robustness
Unemployment GDP
Unweighted Weighted Unweighted Weighted
Establishment deaths (all years) .64 .37 -.46 -.16
Establishment deaths (ex. 2002) .61 .29 -.42 -.03
Establishment deaths (ex. EC years) .67 .34 -.52 -.11
Firm deaths (all years) .41 .31 -.32 -.18
Firm deaths (ex. 2002) .41 .22 -.32 -.09
Firm deaths (ex. EC years) .44 .21 -.43 -.16
Note: Exit rates detrended linearly. BDS correlated with annual change in unemployment
rate or change in log real GDP on BDS annual timing (April-March). “EC” is Economic
Census (years ending in 2 or 7). BDS trends estimated on data for 1978-2018 but corre-
lations estimated on data for 1984-2018.
Source: Author calculations from Business Dynamics Statistics, BLS (unemployment
rates), and BEA (GDP).
Table A7: Business cycle comovement: States (excluding Cen-
sus years)
Establishment death Firm death
Unweighted Weighted Unweighted Weighted
Change in unemp 0.57*** 0.23*** 0.36*** 0.11***
(0.03) (0.02) (0.03) (0.01)
Observations 663 663 663 663
State FE Yes Yes Yes Yes
Year FE No No No No
Note: Regression of state annual exit rates on annual change in unem-
ployment rates, 2003-2018 excluding Economic Census years. Unem-
ployment rate changes timed to correspond with BDS annual timing
(April-March). Compare to Table 2.
***denotes statistical significance with p < 0.01.
Source: Author calculations from Business Dynamics Statistics and
BLS unemployment data.
45
data. That said, the inclusion of the entire 1978-1983 period does not qualitatively matter
for our main results, as can be seen on Table A8.
Table A8: Business cycle correlations: Additional BDS years
Unemployment GDP
Unweighted Weighted Unweighted Weighted
1978-2018:
BDS: Establishment deaths (Census) .39 .16 -.28 -.01
BDS: Firm deaths (Census) .16 .06 -.15 -.02
1984-2018:
BDS: Establishment deaths (Census) .64 .37 -.46 -.16
BDS: Firm deaths (Census) .41 .31 -.32 -.18
Note: Exit rates detrended linearly and correlated with annual change in unemployment
rate or change in log real GDP on BDS annual timing (April-March). Trends estimated
on full 1978-2018 sample.
A.5 Additional state results
Table 2 reports panel regressions relating state-level exit rates and changes in unemployment
rates with state fixed effects. Table A9 reports the same regressions without state fixed
effects (top panel) or with both state and year fixed effects (bottom panel). The fixed effects
specification makes little difference. Moreover, in unreported results we find little effect of
omitting Economic Census years from these regressions.
A.6 2020 BDS redesign
Our main results feature the most recent vintage of the BDS, released in September 2020
after a substantial redesign. Relative to the previous BDS vintage, the redesigned product
features extensive industry detail under consistent NAICS classification along with narrower
tabulations across geography. The redesign also introduced expanded source data and im-
proved longitudinal linking processes. The documentation released with the BDS redesign
explores these issues and highlights ongoing challenges.51 Importantly, overall patterns of
business exit, and patterns around Economic Census years in particular, were slightly differ-
ent in the previous vintage of BDS data. Figure A2 shows establishment (top panels) and
firm (bottom panels) exit rates, unweighted (left panels) and employment weighted (right
panels).
These vintage differences also introduce differences in our business cycle correlations.
Table A10 reports business cycle correlations for both the legacy BDS vintage and the
current (2020) vintage, where we omit the years 2017 and 2018 from the latter (including
the detrending estimation) to have the same time coverage as the legacy vintage.
51
See https://www.census.gov/programs-surveys/bds/news-updates/updates/2018-bds-release.html.
46
Table A9: Business cycle comovement: Alternative controls
Establishment death Firm death
Unweighted Weighted Unweighted Weighted
Change in unemp 0.55*** 0.22*** 0.36*** 0.11***
(0.04) (0.03) (0.04) (0.01)
Observations 816 816 816 816
State FE No No No No
Year FE No No No No
Change in unemp 0.46*** 0.18*** 0.40*** 0.10***
(0.03) (0.03) (0.03) (0.02)
Observations 816 816 816 816
State FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Note: Regression of state annual exit rates on annual change in un-
employment rates, 2003-2018. Unemployment rate changes timed to
correspond with BDS annual timing (April-March). Compare to Table
2.
***denotes statistical significance with p < 0.01.
Source: Author calculations from Business Dynamics Statistics and
BLS data.
Table A10: Business cycle correlations: Vintage comparisons
Unemployment GDP
Unweighted Weighted Unweighted Weighted
Establishment deaths (2020 vintage) .63 .36 -.46 -.15
Establishment deaths (legacy vintage) .40 .19 -.34 -.03
Firm deaths (2020 vintage) .43 .30 -.34 -.19
Firm deaths (legacy vintage) .39 .29 -.36 -.22
Note: Exit rates detrended linearly. BDS correlated with annual change in unemployment
rate or growth of GDP on BDS annual timing (April-March). BDS trends estimated on
data for 1978-2018 but correlations estimated on data for 1984-2018.
47
Establishment death - Unweighted Establishment death - Employment weighted
14 14 7 7
2020 vintage
Legacy vintage
6 6
Annual rate (%) Annual rate (%)
12 12
5 5
10 10
4 4
8 8 8 3 3
19 2
8
19 6
9
19 0
9
19 4
9
20 8 82 86 90 94 98 02 06 10 14 18
19 0
20 2
0
20 6
1
20 0
1
20 4
18 19 19 19 19 19 20 20 20 20 20
Firm death - Unweighted Firm death - Employment weighted
11 11 4 4
2020 vintage
Legacy vintage
10 10
Annual rate (%) Annual rate (%)
3 3
9 9
2 2
8 8
7 7 1 1
82 8
19 6
9
19 0
9
19 4
9
20 8
0
20 2 82 86 90 94 98 02 06 10 14 18
19 19 0
20 6
1
20 0
1
20 4
18 19 19 19 19 19 20 20 20 20 20
Note: DHS denominators. Establishments are single operating business locations.
Firms are collections of one or more establishments under common ownership or operational control.
Source: Census Bureau Business Dynamics Statistics (BDS).
Figure A2: Vintage differences in BDS data
48
Differences between vintages are material for our quantitative estimates. On Table A11,
we repeat our state-level regressions for vintage comparisons; our legacy vintage state-level
data ended in 2014 so we first show regressions on the 2020 redesign data ending in 2014.
We then show regressions on the legacy vintage data for the same years.
Table A11: Business cycle comovement: States, BDS vintages
Establishment death Firm death
Unweighted Weighted Unweighted Weighted
2020 vintage, 2003-2014 data
Change in unemp 0.43*** 0.13*** 0.32*** 0.09***
(0.03) (0.02) (0.03) (0.01)
Observations 612 612 612 612
State FE Yes Yes Yes Yes
Year FE No No No No
Legacy vintage, 2003-2014 data
Change in unemp 0.30*** 0.04* 0.30*** 0.05***
(0.02) (0.02) (0.02) (0.01)
Observations 612 612 612
State FE Yes Yes Yes Yes
Year FE No No No No
Note: Regression of annual exit rates on annual change in unemployment rates, 2003-
2014. Unemployment rate changes timed to correspond with BDS annual timing
(April-March).
***denotes statistical significance with p < 0.01.
** denotes statistical significance with p < 0.05.
* denotes statistical significance with p < 0.10.
Source: Author calculations from Business Dynamics Statistics and BLS unemploy-
ment data.
In the 2020 vintage, business cycle comovements with data for 2003-2014 are somewhat
weaker than the comovements with data for 2003-2018 shown on Table 2. Within the 2003-
2014 period, the legacy BDS vintage has notably weaker cyclicality in all categories than does
the 2020 BDS vintage. That said, the countercyclicality of unweighted exit rates is confirmed
in the older vintage, and there still exists modest countercyclicality even of weighted exit.
A.7 Industrial production
Our main business cycle correlations related exit rates with unemployment and GDP. Table
A12 reports correlations using the growth rate of industrial production (IP) at quarterly
frequency for BED correlations and annual frequency for BDS correlations. The left panel
reports correlations with manufacturing IP, while the right panel reports correlations with
total IP (which consists of manufacturing, mining, and utilities). Overall, exit measures
are consistently countercyclical with respect to IP. Manufacturing IP correlations follow the
49
familiar pattern in which establishment exit is more cyclical than firm exit, and employment-
weighted exit is less cyclical than unweighted exit.
Table A12: Business cycle correlations: Industrial production
Manufacturing IP Total IP
Unweighted Weighted Unweighted Weighted
BED: Establishment closures (BLS) -.39 -.17 -.39 -.17
BED: Establishment deaths (BLS) -.52 -.28 -.53 -.28
BDS: Establishment deaths (Census) -.41 -.14 -.14 -.17
BDS: Firm deaths (Census) -.33 -.19 -.34 -.22
Note: Exit rates detrended linearly. BED correlated with quarterly growth of industrial
production. BDS correlated with annual growth of industrial production on BDS annual
timing (April-March). BED data cover 1992q3-2019q1 (deaths) or 1992q3-2019q4 (clo-
sures). BDS trends estimated on data for 1978-2018 but correlations estimated on data
for 1984-2018.
B Data appendix for SafeGraph closure measure
B.1 Details on SafeGraph data
The closure measure based on SafeGraph mobile device data comes from a combination of two
different sources of information. First, the “Core” dataset is a registry of points of interest
(POIs) in the U.S., with each POI listed with a SafeGraph identifier, industry information,
location, and other time-invariant data.52 SafeGraph has somewhere between 5-6 million
POIs in their Core files, though not all of these POIs would be considered establishments
with employment, as parks, etc., are also listed in the data. Coverage varies across industries
but is generally good for consumer-facing businesses that have an incentive in advertising
their business location to consumers. Overall, the POI database likely covers a significant
share of U.S. business establishments, as Census Bureau and BLS data include between 7
and 10 million establishments.
The second dataset is the set of weekly “patterns” datasets that contain information
about daily/weekly visits to individual establishments. The sample of devices delivering
these data varies over time, but is often in the range of 40-45 million. The weekly pattern
files are released each week and include visits from the previous week (a lag of 4 days or so).
B.2 Creating a longitudinally consistent sample
There are a number of challenges in creating a longitudinally consistent sample (across
establishments, visits data, and the sample of devices). The first challenge comes from the
impact of periodic revisions to the SafeGraph sample. Because establishments deemed closed
52
The nature of this dataset is roughly comparable to the Business Register maintained by the Census
Bureau but is not as comprehensive.
50
are removed from the Core files, one must integrate historical Core snapshots in creating the
universe of potential businesses. The earliest of these is from March 2020; hence, successfully
identifying businesses during the process of closure becomes more challenging the further
back one moves from March 2020.53 Practically speaking, this feature limits the ability to
compare measures of business exit between 2020 and equivalent months from earlier years.
A second challenge comes from revisions to the periodic patterns datasets that contain
information on visits. These revisions often remove the historical visits from businesses
deemed closed as of the time of the revision, a feature which could obviously bias estimates
of business closure. The latest of these revisions to the weekly data that extend back to early
2019 occurred in May, 2020. To correct for this potential bias in months prior to May, 2020,
we utilize an earlier dataset from the monthly version of the patterns information dated from
March 2020. These monthly snapshots that extend to early 2019 contain the daily detail to
construct equivalent weekly estimates of visits. We use this supplementary information to
fill out the sample with any POIs that could have been removed from the weekly patterns
files.54
Finally, we adjust the weekly visits measure based on natural fluctuations in the un-
derlying sample of mobile devices that are used to generate the data. This is particularly
important during the peak of social distancing in March and April, as the sample of devices
fell significantly during this time.55 Using the SafeGraph-provided summary data on mobile
devices recorded in given locations, we gross up the visits by the share of devices in a given
location population, and use this variable for all of our longitudinal comparisons.
B.3 Sample selection
Sit-down restaurants (NAICS 722511) is a good fit for the SafeGraph data for several reasons.
First, this industry has particularly good coverage in the SafeGraph data. When aggregating
up to the four-digit NAICS category (7225) which includes limited service restaurants, the
sample of SafeGraph establishments is reasonably close to official statistics (520 thousand vs
580 thousand according to the BLS QCEW). Second, unlike some industries where expendi-
ture switching to e-commerce can translate to sales without physical movement, restaurants
require movement to the geographic location of some form.56 While there has clearly been
a migration from in-person dining to carry-out service at these establishments in recent
53
This ability declines gradually, as it typically takes some time for SafeGraph to identify a business as
closed using their traditional methods.
54
Because the monthly patterns files use different criteria for identifying the sample of devices for measuring
visits, we construct an establishment-specific scalar to adjust the visits derived from the monthly files based
on overlap between the monthly-based and weekly-based estimates of visits. Those establishments lacking
an overlap are adjusted using the overall sample adjustment ratio.
55
This was due, at least in part, to the fact that mobile devices are not counted when there is no location
data being recorded, a feature that occurs more often when devices are stationary.
56
In reality, it is actually unclear whether COVID-induced switch to e-commerce would mitigate the impact
on business closure. To the extent a greater reliance on internet-based shopping translates to direct delivery
of merchandise, retail locations may be increasingly dispensable, with obvious implications for employment,
at least at a spatial level.
51
months, the carry-out transactions still require a visit that would be picked up by cellphone
tracking data (and this is true whether it be a delivery service or picked up by the ultimate
consumer). Figure B1 shows the shift in the distribution of visits toward carry-out, plotting
the median duration of a visit to restaurants in the sample. The median duration of a visit
before the onset of COVID-19 of just under 40 minutes indicates that casual dining and
some form of carry-out service was likely a feature for these establishments even before the
pandemic. Finally, sit-down restaurants have been particularly hard hit by the COVID-19
recession and have thus received substantial attention in the popular press in the context of
closure.
40
35
Median number of minutes per visit
30
25
20
15
Jan01 Apr01 Jul01 Oct01
Source: Author calculations from SafeGraph microdata.
Figure B1: Median duration of restaurant visits
We merge the Core datasets for all restaurant establishments to the weekly patterns files
that encompass January 2019 to August 2020, retaining establishment data whether or not
visits are recorded. From there we create a balanced panel of weekly observations for each
establishment, filling in visits as zero if no data were recorded in the patterns files. We impose
a number of sample restrictions such as removing restaurants that record implausibly high
numbers of visits in a given week. The resulting dataset identifies 335 thousand restaurants
and covers around 85 weeks (≈28.5 million observations).
B.4 Defining temporary and permanent closures
The approach taken here is to infer the likelihood of temporary and permanent business
closure based on the degree and duration of declines in customer visits. Although one might
expect business closure to be associated with zero visits from customers, there are several
reasons that a non-zero threshold is appropriate. First, SafeGraph visits are based on a
sample of devices that is not universal. Second, even a closed business would likely continue
to record some visits due to periodic trips to the establishment by owners, management,
52
or employees. Even after a prolonged or permanent closure, visits could be recorded if
contractors or landlords continue to maintain the property. A third reason for a non-zero
threshold to record closures is due to data error: while cellphone tracking from GPS signals is
reasonably precise, there is some noise in where SafeGraph identifies a given visit, particularly
if establishments are located close together.
For each case of temporary and permanent closures, we rely on additional data provided
by SafeGraph to discipline the definitions based on visits data.
For temporary closures, we utilize an additional ad-hoc SafeGraph dataset resulting from
an experimental analysis using machine-learning techniques to infer the operating status of
POIs for the week of April 5th. The SafeGraph team trained a logistic model based on
some POIs with known operating status, using a series of visit attributes during this week
relative to a baseline week in early March. The resulting dataset released to researchers has
a set of POIs with the predicted operational status from this model along with a confidence
value.57 Working backward, we retain only those POIs with high confidence values and use
these data to identify simple rules in our sample of restaurants that most accurately align
with temporary closure status. In addition to the year-over-year percent decline in visits, we
consider lower and upper bounds of the absolute value of visits associated with not operating
and operating status, respectively, as well as various forms of weekly smoothing (3-week vs
5-week moving averages) for visit declines.
Operating
.2
Not operating
Year-over-year percent change
0
-.2
-.4
-.6
-.8
Jan01 Apr01 Jul01 Sep30
Source: Author's calculations using SafeGraph microdata.
Figure B2: Year-over-year percent change in visits, by operational status
Week of April 5-11
Figure B2 shows the year-over-year percent change in visits based on operational status
for the week of April 5th (identified by the dashed line in the chart). As expected, restaurants
identified as closed had considerably lower average visits (78 percent declines) versus those
57
Further information on this dataset is available at https://docs.safegraph.com/docs/operatingnot-
operating-poi.
53
remaining open (31 percent decline). Narrowing in on the heterogeneity in visit declines
specifically for the week of April 5th, Figure plots the densities of the year-over-year
change in visits, separately for those identified as operational and not operational. While
the overlap of the densities in Figure [B3]indicates that any rule based on visit declines will
be imperfect, there is significant separation such that visit patterns are highly informative
for closure.
3f Operating
Not operating
Density
0 : 1 ; ;
-1 -5 0 5 1
3-week moving average of year-over-year percent change in visits (for week of April 5-11)
Source: Author's calcualtions using SafeGraph microdata.
Figure B3: Density of year-over-year percent changes in visits, by operational status
Week of April 5-11
More formally, we calculate false positive (identifying closed establishments that are iden-
tified as operational) and false negative (identifying open establishments that are identified
as not operational) rates across a wide variety of parameter values || Ultimately, the best
combination of parameters was a threshold of a 65 percent year-over-year decline in visits
(using a 3-week moving average), a lower bound of 4 visits and an upper bound of 50 vis-
its. This value of y/y declines is identified by the dashed line in Figure For a formal
definition of temporary closure, a business i is temporarily closed in week n if 0j,, = 1:
if Ly (cts) <—0.65 and vin <50
1 j=n-1 Vi,j-—52
1 if jn <4
Vin = ue ntl Vij —Vi,j —52 (BI)
0 if ann (“ucts ) > —0.65 and u;,>4
0 if vm > 50
To guide definitions of permanent closures, we rely on recent additions to the “Core” data
that identify the opening and closing dates of POIs. These data are not directly applicable for
°SThis terminology is somewhat misleading as the “true” data from SafeGraph are themselves estimates
subject to error.
54
purposes of closure due to poor coverage; however, the patterns evident for POIs identified
as permanently closed are nevertheless useful for constructing a universal definition based
on observable characteristics. To verify these closures identified by SafeGraph, we group
all restaurants according to closure date (including a category for those remaining open)
and plot their average weekly visits over the sample period. Figure B4 plots the year-
over-year percent change in weekly visits for some select closure dates. Overall, there is
some evidence that the pattern of visits align with closure: visits drop in February for
those restaurants identified as closing in that month, though the average visits of these
establishments do not fall immediately to zero. The closures occurring during or after the
COVID-19 pandemic show a similar pattern, with average visits remaining well below those
that are identified as remaining open. The numbers corresponding to each line represent
the number of establishment comprising the average visits—highlighting the lack of coverage
for these indicators. The large jump in closures in July, 2020, could be due at least in part
to increased surveillance of closures by SafeGraph. It is also somewhat puzzling that the
contour of visits for the average of restaurants identified as closing in this month is only
slightly below those identified as open.
.5
Open
Year-over-Year Percent Change in Weekly Visits )
Closed Feb.2020
Closed May.2020
Closed Jul.2020
0
328,830
-.5 4,546
109
104
-1
Jan01 Apr01 Jul01 Sep30
Source: Author calculations from SafeGraph microdata. Numbers correspond to national number of restaurants in each group.
Figure B4: Weekly visits by identified closure Status
Figure B4 illustrates how the timing of a permanent closure is made difficult by the
widespread temporary closures in March and April of 2020. Moreover, even for the closures
identified as before the lockdowns resulting from COVID-19 (for example, the February
2020 closures shown by the red line), the weekly visits do not immediately drop to zero in
subsequent weeks. Hence, the patterns in Figure B4 demonstrate that a threshold rule is
likely also necessary for identifying a permanent closure. For simplicity, we therefore define
a closure to be permanent if subsequent visits never rise above the visits threshold identified
for temporary closure. Formally, an establishment i is identified as permanently closed in
p
week n (e vi,n ) if:
p t t
vei,n = 1 if vei,n = 1 and ∀m > n vei,m =1 (B2)
55
t
We translate these measures to a monthly frequency t such that vei,t equals one if any weeks
n ∈ t are equal to one; for permanent closures all subsequent (smoothed) weeks must be
below the threshold to be recorded as closed in the month.
Using prior contour of revisions to refine estimate of closures
Unlike temporary closures, an important feature of the definition of permanent closures is
that it is subject to revision as additional data become available. The set of permanently
closed establishments in a month t can only decline as additional weeks of data outside of
that month become available. Hence the initial estimate of permanent closures for a given
month is an upper bound and will decline over time. One method of accounting for potential
future revisions in real time is to calculate the equivalent revision schedule from prior months’
data and apply that forecasted revision to the appropriate vintage of more recent data. As
an example, Figure B5 illustrates the monthly closure rates for several months in our sample
based on the number of weeks following the end of the respective month. As is clear from
Panel A of Figure B5, the revisions are initially significant and then quickly decline. Panel
B translates these estimates into a ratio relative to the initial estimate, and shows how one
can apply the revision schedule from prior months to arrive at a forecast estimate for a
subsequent vintage of a more recent estimate.59
B.5 Alternate threshold
Figures B6 and B7 show measures of temporary and potentially permanent closures when
using a threshold of 80 percent y/y visit declines (versus 65 percent as in our main specifi-
cations); these figures can be compared to Figures 8 and 9. This more stringent threshold
naturally leads to lower closure rates, with a cumulative death rate estimate of 9.3 percent
(versus 13.5 percent in our main specification).
59
Thanks to Brendan Price for helpful conversations on this methodology.
56
Figure B5: Revisions by month of identified permanent closure
(a) Panel A: Monthly estimate by week
Revisions to Monthly Permanent Closure Rate
8 May
June
July
Percent of establishments
6
4
2
0
0 5 10 15
Weeks since end of month
Source: Author calculations using SafeGraph microdata.
(b) Panel B: Revisions contour relative to initial reading
Revisions to Monthly Permanent Closure Rate
1
May
June
.8 July
Fraction of initial value
.6
.4
.2
0 5 10 15
Weeks since end of month
Source: Author calculations using SafeGraph microdata. Dashed line extrapolate from prior months.
57
Figure B6: Temporary closures: 80 percent threshold
Percent of Restaurants Temporarily Closed (80% Threshold)
43%
40 38% 40
30
22%
19% 19%
20 20
16% 16%
12% 12% 11% 12% 13%
10
2% 3%
0 0
y ry ch ril ay Ju ly st r er r r y ry
ar ua ar Ap ne Ju gu be ob be be ar ua
nu M em em em nu
Ja br M Au ct Ja br
Fe pt O ov ec Fe
Se N D
Source: Author calculations from SafeGraph microdata.
Note: Temporary closure defined in Appendix.
Figure B7: Potentially permanent closures: 80 percent threshold
Percent of Restaurants Permanently Closed (80% Threshold)
Monthly Rate, by month of identified closing (left) 9.3% 10
Cumulative for 2020 (right)
1.6% 8
1.5
6
1
.8%
4
.5
2
0 0
ar ry ar ril ay ne ly st r r r r ar ry
y br ch Ap Ju gu be be be be y br
nu ua M Ju em ct em em nu ua
Ja M Au o Ja
Fe pt O ov ec Fe
Se N D
Source: Author calculations using SafeGraph microdata.
Note: Permanent closure defined in Appendix. Last 3 months are preliminary and may revise.
58
File and source
- File
- SMBLIQ-C-079_business-exit-during-the-covid-19-pandemic-non-traditional-measures-in-historical_2021-04.pdf
- Size
- 1,076,597 bytes
- SHA-256
- 8f77b1e2534aa22df6fc6ecc0c9f3c1021c2a6b89edb58657ab9089a945b7509
- Original
- www.federalreserve.gov